{ "schema_version": 1, "snapshot_at": "2026-09-15T01:19:08.614705+00:00", "title": "Qwen3-14B DEX trainer comparison, September 14, 2026 snapshot", "protocol": { "model": "Qwen/Qwen3-14B", "revision": "40c069824f4251a91eefaf281ebe4c544efd3e18", "hardware": "One NVIDIA RTX PRO 6000 Blackwell 96 GB per run", "precision": "NF4 weights, BF16 compute", "lora_rank": 32, "lora_alpha": 64, "seed": 42, "optimizer_steps": 50, "group_size": 4, "fresh_batch_completions": 8, "updates_per_rollout": 2, "learning_rate": 1e-05, "reward_weights": { "native": 1, "efficiency": 0.05, "reliability": 0.2 }, "holdout": "5 trained families x 2 held-out task seeds x 4 samples = 40 episodes, 10 distinct tasks", "checkpoint_selection": "Final checkpoint, fixed training budget; no holdout-based checkpoint selection" }, "caveats": [ "One training seed per configuration; no statistical winner established.", "Thinking uses a 2048 thinking-token cap and 3072 total tokens per turn; no-think uses 1024 total tokens per turn. These are budget configurations, not equal-token comparisons.", "Fresh reward is logged only on odd optimizer steps. Even steps reuse the preceding rollout.", "Reward is the logged composite rollout score, not native holdout reward or financial profit.", "Refill steps average reward logs across generation rounds. Saved completion rows cover the last generated round, not the final merged optimizer batch.", "Step hours sum optimizer-step timers including rollout generation; exclude provisioning, installation, model loading, holdout and uploads. They are not billed VM hours.", "Memory values are recorded training CUDA allocated/reserved peaks. Historical no-think runs lack these fields.", "The hybrid refill run is partial at the snapshot cutoff; no holdout exists in this snapshot.", "Held-out instances are synthetic and within trained families; these measurements do not establish live trading performance." ], "runs": [ { "run_id": "1789188652", "mode": "no-think", "method": "initial", "steps": 0, "reward_points": 0, "train_mean": null, "train_last10": null, "holdout_native": 0.38555043464607947, "holdout_episodes": 40, "holdout_successes": 16, "mean_step_seconds": null, "label": "Starting model", "state_at_cutoff": "complete", "step_hours": null, "peak_allocated_gib": null, "peak_reserved_gib": null, "extra_refill_rounds": 0, "groups_refilled": 0, "holdout_component_means": { "native": 0.38555043464607947, "efficiency": 0.1625, "reliability": -0.15 }, "configuration": { "model": { "id": "Qwen/Qwen3-14B", "revision": "40c069824f4251a91eefaf281ebe4c544efd3e18", "loss": "dapo", "algorithm": "grpo", "no_think": true, "lora": true, "quantize": true, "lora_r": 32, "lora_alpha": 64, "use_dora": false }, "loss": "dapo", "normalization": "joint", "entropy_control": { "method": "monitor", "epsilon_low": 0.2, "epsilon_high": 0.28, "zeta_initial": 0.001, "zeta_min": 0.0001, "zeta_max": 0.05, "holdout_split": "benchmark", "holdout_seeds_per_family": 2, "num_iterations": 2 }, "dynamic_sampling": false, "sampling": { "group_size": 4, "temperature": 1.0, "max_completion_tokens": 8192, "top_entropy_quantile": 1.0, "temperature_decay_to": null }, "learning_rate": 1e-05, "turn_max_tokens": 1024 }, "source_sha256": { "entropy_holdout_episodes.jsonl": "61b34dafee4b443d1f7a1966e06e93ec722ac1f8f030714c7ae4f0573804c00f", "resolved_training_config.json": "7447b43fa5ec879c6d319e7548b69d5ec8f4680d73d125e4cbd0a260e4551ca9" }, "training_steps": [], "holdout": [ { "seed": 900000, "sample": 0, "native_reward": 0.615643296169515, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 290, "reward_channels": { "native": 0.615643296169515, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "two_pool_cycle" }, { "seed": 900000, "sample": 1, "native_reward": 0.6156419091142415, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 317, "reward_channels": { "native": 0.6156419091142415, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "two_pool_cycle" }, { "seed": 900000, "sample": 2, "native_reward": 0.6156430169577571, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 251, "reward_channels": { "native": 0.6156430169577571, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "two_pool_cycle" }, { "seed": 900000, "sample": 3, "native_reward": 0.175, "promotion_success": false, "done": false, "errors": 0, "tool_steps": 5, "model_tokens": 550, "reward_channels": { "native": 0.175, "efficiency": 0.0, "reliability": -1.0 }, "verification_level": null, "failed_gates": [ "positive_net_profit", "stress_positive", "finalized", "task_complete" ], "family": "two_pool_cycle" }, { "seed": 900001, "sample": 0, "native_reward": 0.6154821465692866, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 519, "reward_channels": { "native": 0.6154821465692866, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "two_pool_cycle" }, { "seed": 900001, "sample": 1, "native_reward": 0.25, "promotion_success": false, "done": false, "errors": 0, "tool_steps": 5, "model_tokens": 546, "reward_channels": { "native": 0.25, "efficiency": 0.0, "reliability": -1.0 }, "verification_level": null, "failed_gates": [ "finalized" ], "family": "two_pool_cycle" }, { "seed": 900001, "sample": 2, "native_reward": 0.175, "promotion_success": false, "done": false, "errors": 0, "tool_steps": 5, "model_tokens": 389, "reward_channels": { "native": 0.175, "efficiency": 0.0, "reliability": -1.0 }, "verification_level": null, "failed_gates": [ "positive_net_profit", "stress_positive", "finalized", "task_complete" ], "family": "two_pool_cycle" }, { "seed": 900001, "sample": 3, "native_reward": 0.175, "promotion_success": false, "done": false, "errors": 0, "tool_steps": 5, "model_tokens": 362, "reward_channels": { "native": 0.175, "efficiency": 0.0, "reliability": -1.0 }, "verification_level": null, "failed_gates": [ "positive_net_profit", "stress_positive", "finalized", "task_complete" ], "family": "two_pool_cycle" }, { "seed": 900000, "sample": 0, "native_reward": 0.175, "promotion_success": false, "done": false, "errors": 0, "tool_steps": 4, "model_tokens": 0, "reward_channels": { "native": 0.175, "efficiency": 0.0, "reliability": -1.0 }, "verification_level": null, "failed_gates": [ "positive_net_profit", "stress_positive", "finalized", "task_complete" ], "family": "triangular_cycle" }, { "seed": 900000, "sample": 1, "native_reward": 0.5848008209671872, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 606, "reward_channels": { "native": 0.5848008209671872, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "triangular_cycle" }, { "seed": 900000, "sample": 2, "native_reward": 0.175, "promotion_success": false, "done": false, "errors": 0, "tool_steps": 4, "model_tokens": 0, "reward_channels": { "native": 0.175, "efficiency": 0.0, "reliability": -1.0 }, "verification_level": null, "failed_gates": [ "positive_net_profit", "stress_positive", "finalized", "task_complete" ], "family": "triangular_cycle" }, { "seed": 900000, "sample": 3, "native_reward": 0.175, "promotion_success": false, "done": false, "errors": 0, "tool_steps": 4, "model_tokens": 736, "reward_channels": { "native": 0.175, "efficiency": 0.0, "reliability": -1.0 }, "verification_level": null, "failed_gates": [ "positive_net_profit", "stress_positive", "finalized", "task_complete" ], "family": "triangular_cycle" }, { "seed": 900001, "sample": 0, "native_reward": 0.175, "promotion_success": false, "done": true, "errors": 0, "tool_steps": 5, "model_tokens": 483, "reward_channels": { "native": 0.175, "efficiency": 0.16666666666666663, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [ "positive_net_profit", "stress_positive", "task_complete" ], "family": "triangular_cycle" }, { "seed": 900001, "sample": 1, "native_reward": 0.175, "promotion_success": false, "done": false, "errors": 0, "tool_steps": 4, "model_tokens": 0, "reward_channels": { "native": 0.175, "efficiency": 0.0, "reliability": -1.0 }, "verification_level": null, "failed_gates": [ "positive_net_profit", "stress_positive", "finalized", "task_complete" ], "family": "triangular_cycle" }, { "seed": 900001, "sample": 2, "native_reward": 0.175, "promotion_success": false, "done": false, "errors": 0, "tool_steps": 4, "model_tokens": 0, "reward_channels": { "native": 0.175, "efficiency": 0.0, "reliability": -1.0 }, "verification_level": null, "failed_gates": [ "positive_net_profit", "stress_positive", "finalized", "task_complete" ], "family": "triangular_cycle" }, { "seed": 900001, "sample": 3, "native_reward": 0.175, "promotion_success": false, "done": false, "errors": 0, "tool_steps": 4, "model_tokens": 0, "reward_channels": { "native": 0.175, "efficiency": 0.0, "reliability": -1.0 }, "verification_level": null, "failed_gates": [ "positive_net_profit", "stress_positive", "finalized", "task_complete" ], "family": "triangular_cycle" }, { "seed": 900000, "sample": 0, "native_reward": 0.25, "promotion_success": false, "done": false, "errors": 0, "tool_steps": 4, "model_tokens": 684, "reward_channels": { "native": 0.25, "efficiency": 0.0, "reliability": -1.0 }, "verification_level": null, "failed_gates": [ "finalized" ], "family": "competing_cycles" }, { "seed": 900000, "sample": 1, "native_reward": 0.25, "promotion_success": false, "done": false, "errors": 0, "tool_steps": 3, "model_tokens": 304, "reward_channels": { "native": 0.25, "efficiency": 0.0, "reliability": -1.0 }, "verification_level": null, "failed_gates": [ "finalized" ], "family": "competing_cycles" }, { "seed": 900000, "sample": 2, "native_reward": 0.25, "promotion_success": false, "done": false, "errors": 0, "tool_steps": 3, "model_tokens": 514, "reward_channels": { "native": 0.25, "efficiency": 0.0, "reliability": -1.0 }, "verification_level": null, "failed_gates": [ "finalized" ], "family": "competing_cycles" }, { "seed": 900000, "sample": 3, "native_reward": 0.25, "promotion_success": false, "done": false, "errors": 0, "tool_steps": 4, "model_tokens": 760, "reward_channels": { "native": 0.25, "efficiency": 0.0, "reliability": -1.0 }, "verification_level": null, "failed_gates": [ "finalized" ], "family": "competing_cycles" }, { "seed": 900001, "sample": 0, "native_reward": 0.6192995951616761, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 696, "reward_channels": { "native": 0.6192995951616761, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "competing_cycles" }, { "seed": 900001, "sample": 1, "native_reward": 0.25, "promotion_success": false, "done": false, "errors": 0, "tool_steps": 4, "model_tokens": 656, "reward_channels": { "native": 0.25, "efficiency": 0.0, "reliability": -1.0 }, "verification_level": null, "failed_gates": [ "finalized" ], "family": "competing_cycles" }, { "seed": 900001, "sample": 2, "native_reward": 0.25, "promotion_success": false, "done": false, "errors": 0, "tool_steps": 4, "model_tokens": 694, "reward_channels": { "native": 0.25, "efficiency": 0.0, "reliability": -1.0 }, "verification_level": null, "failed_gates": [ "finalized" ], "family": "competing_cycles" }, { "seed": 900001, "sample": 3, "native_reward": 0.6212167624035315, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 3, "model_tokens": 406, "reward_channels": { "native": 0.6212167624035315, "efficiency": 0.5, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "competing_cycles" }, { "seed": 900000, "sample": 0, "native_reward": 0.6161298837365798, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 3, "model_tokens": 387, "reward_channels": { "native": 0.6161298837365798, "efficiency": 0.5, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "four_hop_cycle" }, { "seed": 900000, "sample": 1, "native_reward": 0.6161282892735525, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 3, "model_tokens": 366, "reward_channels": { "native": 0.6161282892735525, "efficiency": 0.5, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "four_hop_cycle" }, { "seed": 900000, "sample": 2, "native_reward": 0.6179612389087088, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 332, "reward_channels": { "native": 0.6179612389087088, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "four_hop_cycle" }, { "seed": 900000, "sample": 3, "native_reward": 0.6153297728643924, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 569, "reward_channels": { "native": 0.6153297728643924, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "four_hop_cycle" }, { "seed": 900001, "sample": 0, "native_reward": 0.175, "promotion_success": false, "done": false, "errors": 0, "tool_steps": 4, "model_tokens": 575, "reward_channels": { "native": 0.175, "efficiency": 0.0, "reliability": -1.0 }, "verification_level": null, "failed_gates": [ "positive_net_profit", "stress_positive", "finalized", "task_complete" ], "family": "four_hop_cycle" }, { "seed": 900001, "sample": 1, "native_reward": 0.6163578745967674, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 3, "model_tokens": 391, "reward_channels": { "native": 0.6163578745967674, "efficiency": 0.5, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "four_hop_cycle" }, { "seed": 900001, "sample": 2, "native_reward": 0.25, "promotion_success": false, "done": false, "errors": 0, "tool_steps": 4, "model_tokens": 514, "reward_channels": { "native": 0.25, "efficiency": 0.0, "reliability": -1.0 }, "verification_level": null, "failed_gates": [ "finalized" ], "family": "four_hop_cycle" }, { "seed": 900001, "sample": 3, "native_reward": 0.6140284978433885, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 301, "reward_channels": { "native": 0.6140284978433885, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "four_hop_cycle" }, { "seed": 900000, "sample": 0, "native_reward": 0.175, "promotion_success": false, "done": false, "errors": 0, "tool_steps": 4, "model_tokens": 756, "reward_channels": { "native": 0.175, "efficiency": 0.0, "reliability": -1.0 }, "verification_level": null, "failed_gates": [ "positive_net_profit", "stress_positive", "finalized", "task_complete" ], "family": "size_sensitive_cycle" }, { "seed": 900000, "sample": 1, "native_reward": 0.8308833579638738, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 320, "reward_channels": { "native": 0.8308833579638738, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "size_sensitive_cycle" }, { "seed": 900000, "sample": 2, "native_reward": 0.175, "promotion_success": false, "done": false, "errors": 0, "tool_steps": 4, "model_tokens": 724, "reward_channels": { "native": 0.175, "efficiency": 0.0, "reliability": -1.0 }, "verification_level": null, "failed_gates": [ "positive_net_profit", "stress_positive", "finalized", "task_complete" ], "family": "size_sensitive_cycle" }, { "seed": 900000, "sample": 3, "native_reward": 0.175, "promotion_success": false, "done": false, "errors": 0, "tool_steps": 4, "model_tokens": 737, "reward_channels": { "native": 0.175, "efficiency": 0.0, "reliability": -1.0 }, "verification_level": null, "failed_gates": [ "positive_net_profit", "stress_positive", "finalized", "task_complete" ], "family": "size_sensitive_cycle" }, { "seed": 900001, "sample": 0, "native_reward": 0.8287354292640952, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 369, "reward_channels": { "native": 0.8287354292640952, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "size_sensitive_cycle" }, { "seed": 900001, "sample": 1, "native_reward": 0.25, "promotion_success": false, "done": false, "errors": 0, "tool_steps": 4, "model_tokens": 821, "reward_channels": { "native": 0.25, "efficiency": 0.0, "reliability": -1.0 }, "verification_level": null, "failed_gates": [ "finalized" ], "family": "size_sensitive_cycle" }, { "seed": 900001, "sample": 2, "native_reward": 0.8287354940486263, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 263, "reward_channels": { "native": 0.8287354940486263, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "size_sensitive_cycle" }, { "seed": 900001, "sample": 3, "native_reward": 0.25, "promotion_success": false, "done": false, "errors": 0, "tool_steps": 4, "model_tokens": 464, "reward_channels": { "native": 0.25, "efficiency": 0.0, "reliability": -1.0 }, "verification_level": null, "failed_gates": [ "finalized" ], "family": "size_sensitive_cycle" } ] }, { "run_id": "1789151457", "mode": "no-think", "method": "grpo", "steps": 50, "reward_points": 25, "train_mean": 0.703773136138916, "train_last10": 0.7729044497013092, "holdout_native": 0.6423985158115345, "holdout_episodes": 40, "holdout_successes": 38, "mean_step_seconds": 177.08361022171974, "label": "GRPO", "state_at_cutoff": "complete", "step_hours": 2.459494586412774, "peak_allocated_gib": null, "peak_reserved_gib": null, "extra_refill_rounds": 0.0, "groups_refilled": 0.0, "holdout_component_means": { "native": 0.6423985158115345, "efficiency": 0.4708333333333334, "reliability": 0.9 }, "configuration": { "model": { "id": "Qwen/Qwen3-14B", "revision": "40c069824f4251a91eefaf281ebe4c544efd3e18", "loss": "grpo", "algorithm": "grpo", "no_think": true, "lora": true, "quantize": true, "lora_r": 32, "lora_alpha": 64, "use_dora": false }, "loss": "grpo", "normalization": "joint", "entropy_control": { "method": "monitor", "epsilon_low": 0.2, "epsilon_high": 0.2, "zeta_initial": 0.001, "zeta_min": 0.0001, "zeta_max": 0.05, "holdout_split": "benchmark", "holdout_seeds_per_family": 2, "num_iterations": 2 }, "dynamic_sampling": false, "sampling": { "group_size": 4, "temperature": 1.0, "max_completion_tokens": 8192, "top_entropy_quantile": 1.0, "temperature_decay_to": null }, "learning_rate": 1e-05, "turn_max_tokens": 1024 }, "source_sha256": { "metrics.jsonl": "bd297cd2a864b080aec383fe534465a878ce0d7795e76bb5e1ad6846445367ab", "entropy_holdout_episodes.jsonl": "4546dfd4faba49ca69699ded44a29683c3e2ca5161182a3fc6567061ba47594b", "resolved_training_config.json": "a6a46f534eb34e2b299247249bfc4f357b28e6f9d2eb502e18be9539e6ab1c8a" }, "training_steps": [ { "step": 1, "loss": -1.1175870895385742e-08, "grad_norm": 0.2431640625, "learning_rate": 1e-05, "num_tokens": 27996.0, "completions/mean_length": 603.0, "completions/min_length": 278.0, "completions/max_length": 1190.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 603.0, "completions/min_terminated_length": 278.0, "completions/max_terminated_length": 1190.0, "tools/call_frequency": 4.375, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.5008535981178284, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 0.21723572909832, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.25, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.1666666716337204, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/mean": 0.3369031250476837, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/std": 0.3238062560558319, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/mean": 0.1666666716337204, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/std": 0.19245010614395142, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/mean": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/std": 1.154700517654419, "reward": 0.4792949855327606, "reward_std": 0.45436012744903564, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.04867766337702051, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 352.69348045599986, "entropy_control/entropy": 0.03031625726079178, "entropy_control/target": 0.03031625726079178, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "epoch": 0.004, "adapter_sync_time": 0.0, "generation_time": 249.134806, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 269.704387, "forward_backward_time": 82.877189, "rollout_overhead_time": 20.569581 }, { "step": 2, "loss": -0.0008399337530136108, "grad_norm": 0.2060546875, "learning_rate": 9.800000000000001e-06, "entropy": 0.048708565649576485, "clip_ratio/low_mean": 0.0005130928620928898, "clip_ratio/low_min": 0.0005130928620928898, "clip_ratio/high_mean": 0.002514466643333435, "clip_ratio/high_max": 0.002514466643333435, "clip_ratio/region_mean": 0.003027559505426325, "step_time": 82.91500510800006, "entropy_control/entropy": 0.03025453078373544, "entropy_control/target": 0.03031625726079178, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "epoch": 0.008, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 82.799137, "rollout_overhead_time": 0.0, "reward": null }, { "step": 3, "loss": -7.399643072858453e-05, "grad_norm": 0.15625, "learning_rate": 9.600000000000001e-06, "num_tokens": 52569.0, "completions/mean_length": 355.625, "completions/min_length": 299.0, "completions/max_length": 526.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 355.625, "completions/min_terminated_length": 299.0, "completions/max_terminated_length": 526.0, "tools/call_frequency": 4.25, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.5030288696289062, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 0.2024640142917633, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.2708333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.12400397658348083, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 0.75, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 0.7071067690849304, "reward": 0.6665704846382141, "reward_std": 0.3200153410434723, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.5, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.07991488021798432, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 258.22600335200013, "entropy_control/entropy": 0.07515834817317896, "entropy_control/target": 0.03031625726079178, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "epoch": 0.012, "adapter_sync_time": 0.0, "generation_time": 166.98614, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 184.568514, "forward_backward_time": 73.54299, "rollout_overhead_time": 17.582374 }, { "step": 4, "loss": -0.0005477418890222907, "grad_norm": 0.0888671875, "learning_rate": 9.4e-06, "entropy": 0.0792204940225929, "clip_ratio/low_mean": 0.001218765857629478, "clip_ratio/low_min": 0.001218765857629478, "clip_ratio/high_mean": 0.002375894458964467, "clip_ratio/high_max": 0.002375894458964467, "clip_ratio/region_mean": 0.003594660316593945, "step_time": 73.627856651, "entropy_control/entropy": 0.07454071336359021, "entropy_control/target": 0.03031625726079178, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "epoch": 0.016, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 73.520765, "rollout_overhead_time": 0.0, "reward": null }, { "step": 5, "loss": 4.470348358154297e-07, "grad_norm": 0.224609375, "learning_rate": 9.200000000000002e-06, "num_tokens": 77055.0, "completions/mean_length": 351.75, "completions/min_length": 288.0, "completions/max_length": 521.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 351.75, "completions/min_terminated_length": 288.0, "completions/max_terminated_length": 521.0, "tools/call_frequency": 4.75, "tools/failure_frequency": 0.1315789520740509, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.5532469749450684, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 0.15319469571113586, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.2916666865348816, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.1178511381149292, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 0.75, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 0.7071067690849304, "reward": 0.717830240726471, "reward_std": 0.3003321588039398, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.06839368864893913, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 292.5127315850002, "entropy_control/entropy": 0.06708223354780454, "entropy_control/target": 0.03031625726079178, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "epoch": 0.02, "adapter_sync_time": 0.0, "generation_time": 200.99866, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 218.492077, "forward_backward_time": 73.912911, "rollout_overhead_time": 17.493417 }, { "step": 6, "loss": -0.0005029439926147461, "grad_norm": 0.236328125, "learning_rate": 9e-06, "entropy": 0.06814951589331031, "clip_ratio/low_mean": 0.0007854699215386063, "clip_ratio/low_min": 0.0007854699215386063, "clip_ratio/high_mean": 0.00350716634420678, "clip_ratio/high_max": 0.00350716634420678, "clip_ratio/region_mean": 0.0042926362657453865, "step_time": 74.01661177599931, "entropy_control/entropy": 0.06685699916611106, "entropy_control/target": 0.03031625726079178, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "epoch": 0.024, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 73.911966, "rollout_overhead_time": 0.0, "reward": null }, { "step": 7, "loss": -0.06248396262526512, "grad_norm": 0.162109375, "learning_rate": 8.8e-06, "num_tokens": 109119.0, "completions/mean_length": 536.5, "completions/min_length": 246.0, "completions/max_length": 1023.0, "completions/clipped_ratio": 0.125, "completions/mean_terminated_length": 472.2857360839844, "completions/min_terminated_length": 246.0, "completions/max_terminated_length": 1023.0, "tools/call_frequency": 4.0, "tools/failure_frequency": 0.0625, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/mean": 0.3410455584526062, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/std": 0.1820911467075348, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/mean": 0.0833333358168602, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/std": 0.1666666716337204, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/mean": -0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/std": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/mean": 0.6818172931671143, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/std": 0.28787821531295776, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/mean": 0.25, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/std": 0.1666666716337204, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/std": 1.0, "reward": 0.51976478099823, "reward_std": 0.5069523453712463, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.02533304301323369, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 337.1979487909998, "entropy_control/entropy": 0.02212925314251013, "entropy_control/target": 0.03031625726079178, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "epoch": 0.028, "adapter_sync_time": 0.0, "generation_time": 218.429661, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 240.530696, "forward_backward_time": 96.561975, "rollout_overhead_time": 22.101035 }, { "step": 8, "loss": -0.06261745095252991, "grad_norm": 0.09326171875, "learning_rate": 8.6e-06, "entropy": 0.025826407712884247, "clip_ratio/low_mean": 0.0008170681248884648, "clip_ratio/low_min": 0.0008170681248884648, "clip_ratio/high_mean": 0.0014168911147862673, "clip_ratio/high_max": 0.0014168911147862673, "clip_ratio/region_mean": 0.002233959239674732, "step_time": 96.67439458799981, "entropy_control/entropy": 0.022419002595210895, "entropy_control/target": 0.03031625726079178, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "epoch": 0.032, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 96.568858, "rollout_overhead_time": 0.0, "reward": null }, { "step": 9, "loss": -1.4901161193847656e-08, "grad_norm": 0.1572265625, "learning_rate": 8.400000000000001e-06, "num_tokens": 137769.0, "completions/mean_length": 293.25, "completions/min_length": 254.0, "completions/max_length": 398.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 293.25, "completions/min_terminated_length": 254.0, "completions/max_terminated_length": 398.0, "tools/call_frequency": 3.75, "tools/failure_frequency": 0.13333334028720856, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.6143291592597961, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 0.0010098718339577317, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.4166666865348816, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.1666666716337204, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/mean": 0.25, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/mean": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/mean": -1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/std": 0.0, "reward": 0.4425812363624573, "reward_std": 0.4197315573692322, "frac_reward_zero_std": 0.5, "frac_groups_zero_std": 0.5, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.06191796762868762, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 246.38061882300053, "entropy_control/entropy": 0.0668467061896424, "entropy_control/target": 0.03031625726079178, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "epoch": 0.036, "adapter_sync_time": 0.0, "generation_time": 133.529362, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 154.472101, "forward_backward_time": 91.798267, "rollout_overhead_time": 20.94274 }, { "step": 10, "loss": -0.00017190352082252502, "grad_norm": 0.1318359375, "learning_rate": 8.2e-06, "entropy": 0.061034107115119696, "clip_ratio/low_mean": 0.0008351178839802742, "clip_ratio/low_min": 0.0008351178839802742, "clip_ratio/high_mean": 0.0009422110742889345, "clip_ratio/high_max": 0.0009422110742889345, "clip_ratio/region_mean": 0.0017773289582692087, "step_time": 91.90123079600062, "entropy_control/entropy": 0.06596986373564874, "entropy_control/target": 0.03031625726079178, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "epoch": 0.04, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 91.791034, "rollout_overhead_time": 0.0, "reward": null }, { "step": 11, "loss": 2.4065375328063965e-06, "grad_norm": 0.33984375, "learning_rate": 8.000000000000001e-06, "num_tokens": 166825.0, "completions/mean_length": 319.0, "completions/min_length": 264.0, "completions/max_length": 409.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 319.0, "completions/min_terminated_length": 264.0, "completions/max_terminated_length": 409.0, "tools/call_frequency": 3.5, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__native/mean": 0.6147550344467163, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__native/std": 0.0012653579469770193, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__efficiency/mean": 0.4166666865348816, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__efficiency/std": 0.09622503817081451, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/mean": 0.6125867366790771, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/std": 0.001267216051928699, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/mean": 0.4166666865348816, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/std": 0.09622503817081451, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/std": 0.0, "reward": 0.8345041275024414, "reward_std": 0.005744827911257744, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.06843931006733328, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 235.28879666799958, "entropy_control/entropy": 0.07227286487301786, "entropy_control/target": 0.03031625726079178, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "epoch": 0.044, "adapter_sync_time": 0.0, "generation_time": 137.09465, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 155.86287, "forward_backward_time": 79.322734, "rollout_overhead_time": 18.768221 }, { "step": 12, "loss": 0.0006765499711036682, "grad_norm": 0.482421875, "learning_rate": 7.800000000000002e-06, "entropy": 0.0681472544092685, "clip_ratio/low_mean": 0.0004734848625957966, "clip_ratio/low_min": 0.0004734848625957966, "clip_ratio/high_mean": 0.0015520643792115152, "clip_ratio/high_max": 0.0015520643792115152, "clip_ratio/region_mean": 0.0020255492418073118, "step_time": 79.40832818699982, "entropy_control/entropy": 0.07203012919287269, "entropy_control/target": 0.03031625726079178, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "epoch": 0.048, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 79.305767, "rollout_overhead_time": 0.0, "reward": null }, { "step": 13, "loss": 7.3577044531703e-05, "grad_norm": 0.15234375, "learning_rate": 7.600000000000001e-06, "num_tokens": 195955.0, "completions/mean_length": 352.25, "completions/min_length": 252.0, "completions/max_length": 722.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 352.25, "completions/min_terminated_length": 252.0, "completions/max_terminated_length": 722.0, "tools/call_frequency": 4.0, "tools/failure_frequency": 0.09375, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.6102476119995117, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 1.2516975402832031e-06, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.3333333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/mean": 0.3416963219642639, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/std": 0.18339261412620544, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/mean": 0.125, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/std": 0.25, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/mean": -0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/std": 1.0, "reward": 0.5374302864074707, "reward_std": 0.40366190671920776, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.04598454339429736, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 308.6096049810003, "entropy_control/entropy": 0.04070698001055162, "entropy_control/target": 0.03031625726079178, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "epoch": 0.052, "adapter_sync_time": 0.0, "generation_time": 194.807216, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 215.960182, "forward_backward_time": 92.546204, "rollout_overhead_time": 21.152966 }, { "step": 14, "loss": 0.00019895355217158794, "grad_norm": 0.14453125, "learning_rate": 7.4e-06, "entropy": 0.04630886914674193, "clip_ratio/low_mean": 0.00017313018906861544, "clip_ratio/low_min": 0.00017313018906861544, "clip_ratio/high_mean": 0.0005924170836806297, "clip_ratio/high_max": 0.0005924170836806297, "clip_ratio/region_mean": 0.0007655472727492452, "step_time": 92.64190135599938, "entropy_control/entropy": 0.041020644994067545, "entropy_control/target": 0.03031625726079178, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "epoch": 0.056, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 92.532393, "rollout_overhead_time": 0.0, "reward": null }, { "step": 15, "loss": -2.2351741790771484e-08, "grad_norm": 0.26953125, "learning_rate": 7.2000000000000005e-06, "num_tokens": 226261.0, "completions/mean_length": 300.75, "completions/min_length": 252.0, "completions/max_length": 430.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 300.75, "completions/min_terminated_length": 252.0, "completions/max_terminated_length": 430.0, "tools/call_frequency": 3.5, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/mean": 0.25, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/mean": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/mean": -1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/mean": 0.5386062860488892, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/std": 0.3332538306713104, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/mean": 0.1666666716337204, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/std": 0.19245010614395142, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/mean": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/std": 1.154700517654419, "reward": 0.29846981167793274, "reward_std": 0.4600767195224762, "frac_reward_zero_std": 0.5, "frac_groups_zero_std": 0.5, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.02504690084606409, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 210.34915412499868, "entropy_control/entropy": 0.02372330784831911, "entropy_control/target": 0.03031625726079178, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "epoch": 0.06, "adapter_sync_time": 0.0, "generation_time": 111.033873, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 129.705962, "forward_backward_time": 80.52408, "rollout_overhead_time": 18.672089 }, { "step": 16, "loss": 0.00030812621116638184, "grad_norm": 0.19921875, "learning_rate": 7e-06, "entropy": 0.025833570398390293, "clip_ratio/low_mean": 0.0002906976733356714, "clip_ratio/low_min": 0.0002906976733356714, "clip_ratio/high_mean": 0.0004716981202363968, "clip_ratio/high_max": 0.0004716981202363968, "clip_ratio/region_mean": 0.0007623957935720682, "step_time": 80.62655930799974, "entropy_control/entropy": 0.02445330641371228, "entropy_control/target": 0.03031625726079178, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "epoch": 0.064, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 80.516377, "rollout_overhead_time": 0.0, "reward": null }, { "step": 17, "loss": 7.450580596923828e-08, "grad_norm": 0.03173828125, "learning_rate": 6.800000000000001e-06, "num_tokens": 259877.0, "completions/mean_length": 507.0, "completions/min_length": 317.0, "completions/max_length": 974.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 507.0, "completions/min_terminated_length": 317.0, "completions/max_terminated_length": 974.0, "tools/call_frequency": 4.125, "tools/failure_frequency": 0.03030303120613098, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/mean": 0.5229074358940125, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/std": 0.1819382905960083, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/mean": 0.25, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/std": 0.1666666716337204, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/std": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/mean": 0.6089195609092712, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/mean": 0.3333333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/std": 0.0, "reward": 0.7304967641830444, "reward_std": 0.27497315406799316, "frac_reward_zero_std": 0.5, "frac_groups_zero_std": 0.5, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.01884047588100657, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 301.0209074270001, "entropy_control/entropy": 0.016145197240740677, "entropy_control/target": 0.03031625726079178, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "epoch": 0.068, "adapter_sync_time": 0.0, "generation_time": 196.888417, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 216.662419, "forward_backward_time": 84.253635, "rollout_overhead_time": 19.774001 }, { "step": 18, "loss": -0.00015524029731750488, "grad_norm": 0.0260009765625, "learning_rate": 6.600000000000001e-06, "entropy": 0.01891381130553782, "clip_ratio/low_mean": 0.00012833676009904593, "clip_ratio/low_min": 0.00012833676009904593, "clip_ratio/high_mean": 0.00021739130897913128, "clip_ratio/high_max": 0.00021739130897913128, "clip_ratio/region_mean": 0.0003457280690781772, "step_time": 84.37778744900015, "entropy_control/entropy": 0.01608851904818625, "entropy_control/target": 0.03031625726079178, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "epoch": 0.072, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 84.264873, "rollout_overhead_time": 0.0, "reward": null }, { "step": 19, "loss": -0.0001468014670535922, "grad_norm": 0.0035552978515625, "learning_rate": 6.4000000000000006e-06, "num_tokens": 285422.0, "completions/mean_length": 298.625, "completions/min_length": 259.0, "completions/max_length": 399.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 298.625, "completions/min_terminated_length": 259.0, "completions/max_terminated_length": 399.0, "tools/call_frequency": 4.0, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.6152946352958679, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 1.5055335325087071e-06, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.3333333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/mean": 0.8315019607543945, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/std": 1.088771000468114e-06, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/mean": 0.3333333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/std": 0.0, "reward": 0.9400649666786194, "reward_std": 0.11556772887706757, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.5, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.061668465728871524, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 221.8670776459976, "entropy_control/entropy": 0.06537337239155366, "entropy_control/target": 0.03031625726079178, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "epoch": 0.076, "adapter_sync_time": 0.0, "generation_time": 118.79643, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 138.001979, "forward_backward_time": 83.748845, "rollout_overhead_time": 19.205548 }, { "step": 20, "loss": -0.00013157277135178447, "grad_norm": 0.00445556640625, "learning_rate": 6.200000000000001e-06, "entropy": 0.06107214139774442, "clip_ratio/low_mean": 0.0013392106629908085, "clip_ratio/low_min": 0.0013392106629908085, "clip_ratio/high_mean": 0.0004006410308647901, "clip_ratio/high_max": 0.0004006410308647901, "clip_ratio/region_mean": 0.0017398516938555986, "step_time": 83.84809216400117, "entropy_control/entropy": 0.06471683871963352, "entropy_control/target": 0.03031625726079178, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "epoch": 0.08, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 83.74383, "rollout_overhead_time": 0.0, "reward": null }, { "step": 21, "loss": -1.3262033462524414e-06, "grad_norm": 0.3828125, "learning_rate": 6e-06, "num_tokens": 318967.0, "completions/mean_length": 490.125, "completions/min_length": 265.0, "completions/max_length": 925.0, "completions/clipped_ratio": 0.125, "completions/mean_terminated_length": 428.0000305175781, "completions/min_terminated_length": 265.0, "completions/max_terminated_length": 781.0, "tools/call_frequency": 3.625, "tools/failure_frequency": 0.03448275849223137, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/mean": 0.25, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/mean": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/mean": -1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/mean": 0.6108242273330688, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/std": 0.002814680337905884, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/mean": 0.4166666865348816, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/std": 0.1666666716337204, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/std": 0.0, "reward": 0.440828800201416, "reward_std": 0.41787728667259216, "frac_reward_zero_std": 0.5, "frac_groups_zero_std": 0.5, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.01661260286346078, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 333.5141935849979, "entropy_control/entropy": 0.01655680273067031, "entropy_control/target": 0.03031625726079178, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "epoch": 0.084, "adapter_sync_time": 0.0, "generation_time": 215.193628, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 237.129194, "forward_backward_time": 96.257753, "rollout_overhead_time": 21.935566 }, { "step": 22, "loss": -0.000545695424079895, "grad_norm": 0.330078125, "learning_rate": 5.8e-06, "entropy": 0.017196238273754716, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0011980830458924174, "clip_ratio/high_max": 0.0011980830458924174, "clip_ratio/region_mean": 0.0011980830458924174, "step_time": 96.38356308099901, "entropy_control/entropy": 0.01702333176424779, "entropy_control/target": 0.03031625726079178, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "epoch": 0.088, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 96.262762, "rollout_overhead_time": 0.0, "reward": null }, { "step": 23, "loss": 0.00044046342372894287, "grad_norm": 0.16015625, "learning_rate": 5.600000000000001e-06, "num_tokens": 344242.0, "completions/mean_length": 269.875, "completions/min_length": 238.0, "completions/max_length": 391.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 269.875, "completions/min_terminated_length": 238.0, "completions/max_terminated_length": 391.0, "tools/call_frequency": 4.375, "tools/failure_frequency": 0.17142857611179352, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.6148279905319214, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 0.0008064718567766249, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.4583333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.0833333283662796, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/mean": 0.8294905424118042, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/std": 1.501989686403249e-06, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/mean": 0.3333333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/std": 0.0, "reward": 0.9419509172439575, "reward_std": 0.11144879460334778, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.042385022912640125, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 245.29134741899816, "entropy_control/entropy": 0.04561349127379264, "entropy_control/target": 0.03031625726079178, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "epoch": 0.092, "adapter_sync_time": 0.0, "generation_time": 143.773406, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 162.690954, "forward_backward_time": 82.476143, "rollout_overhead_time": 18.917548 }, { "step": 24, "loss": 0.00047101080417633057, "grad_norm": 0.1357421875, "learning_rate": 5.400000000000001e-06, "entropy": 0.04200521484017372, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0020697777217719704, "clip_ratio/high_max": 0.0020697777217719704, "clip_ratio/region_mean": 0.0020697777217719704, "step_time": 82.59226387000035, "entropy_control/entropy": 0.045218182585161575, "entropy_control/target": 0.03031625726079178, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "epoch": 0.096, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 82.471461, "rollout_overhead_time": 0.0, "reward": null }, { "step": 25, "loss": -9.89530235528946e-09, "grad_norm": 0.173828125, "learning_rate": 5.2e-06, "num_tokens": 371488.0, "completions/mean_length": 311.75, "completions/min_length": 250.0, "completions/max_length": 375.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 311.75, "completions/min_terminated_length": 250.0, "completions/max_terminated_length": 375.0, "tools/call_frequency": 3.875, "tools/failure_frequency": 0.12903225421905518, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.6115735769271851, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 7.555148613391793e-07, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.3333333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/mean": 0.6130543947219849, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/std": 0.0014879548689350486, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/mean": 0.5416666865348816, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/std": 0.15957120060920715, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/std": 0.0, "reward": 0.8341889977455139, "reward_std": 0.008876294828951359, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.5, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.04961282107979059, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 277.141689441999, "entropy_control/entropy": 0.05045514122645926, "entropy_control/target": 0.03031625726079178, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "epoch": 0.1, "adapter_sync_time": 0.0, "generation_time": 168.794268, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 188.926659, "forward_backward_time": 88.100639, "rollout_overhead_time": 20.132391 }, { "step": 26, "loss": -0.00029454356990754604, "grad_norm": 0.10595703125, "learning_rate": 5e-06, "entropy": 0.049299571895971894, "clip_ratio/low_mean": 0.0019183356780558825, "clip_ratio/low_min": 0.0019183356780558825, "clip_ratio/high_mean": 0.0013235294609330595, "clip_ratio/high_max": 0.0013235294609330595, "clip_ratio/region_mean": 0.003241865138988942, "step_time": 88.20175407399893, "entropy_control/entropy": 0.05016574730078988, "entropy_control/target": 0.03031625726079178, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "epoch": 0.104, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 88.088165, "rollout_overhead_time": 0.0, "reward": null }, { "step": 27, "loss": 7.271557115018368e-05, "grad_norm": 0.2138671875, "learning_rate": 4.800000000000001e-06, "num_tokens": 400961.0, "completions/mean_length": 377.625, "completions/min_length": 252.0, "completions/max_length": 603.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 377.625, "completions/min_terminated_length": 252.0, "completions/max_terminated_length": 603.0, "tools/call_frequency": 3.0, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/mean": 0.6162011623382568, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/std": 1.1866166005347623e-06, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/mean": 0.6666666865348816, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/mean": 0.8260114789009094, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/std": 0.00435878848657012, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/mean": 0.3333333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/std": 0.2357022762298584, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/std": 0.0, "reward": 0.9461063146591187, "reward_std": 0.10377224534749985, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.05742222792468965, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 251.29658381099944, "entropy_control/entropy": 0.05643979459697434, "entropy_control/target": 0.03031625726079178, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "epoch": 0.108, "adapter_sync_time": 0.0, "generation_time": 143.217945, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 163.201814, "forward_backward_time": 87.985604, "rollout_overhead_time": 19.983869 }, { "step": 28, "loss": -0.00024202116765081882, "grad_norm": 0.09814453125, "learning_rate": 4.600000000000001e-06, "entropy": 0.05681797885335982, "clip_ratio/low_mean": 0.0006218905473360792, "clip_ratio/low_min": 0.0006218905473360792, "clip_ratio/high_mean": 0.002575607068138197, "clip_ratio/high_max": 0.002575607068138197, "clip_ratio/region_mean": 0.0031974976154742762, "step_time": 87.88071318700077, "entropy_control/entropy": 0.05581093909786784, "entropy_control/target": 0.03031625726079178, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "epoch": 0.112, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 87.766821, "rollout_overhead_time": 0.0, "reward": null }, { "step": 29, "loss": 2.9802322387695312e-08, "grad_norm": 0.08056640625, "learning_rate": 4.4e-06, "num_tokens": 430422.0, "completions/mean_length": 396.125, "completions/min_length": 234.0, "completions/max_length": 857.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 396.125, "completions/min_terminated_length": 234.0, "completions/max_terminated_length": 857.0, "tools/call_frequency": 3.75, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.6086075305938721, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.3333333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/mean": 0.34096160531044006, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/std": 0.18192321062088013, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/mean": 0.0833333358168602, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/std": 0.1666666716337204, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/mean": -0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/std": 1.0, "reward": 0.5352011919021606, "reward_std": 0.4017886817455292, "frac_reward_zero_std": 0.5, "frac_groups_zero_std": 0.5, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.04723928542807698, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 316.6319273319996, "entropy_control/entropy": 0.03989102720396147, "entropy_control/target": 0.03031625726079178, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "epoch": 0.116, "adapter_sync_time": 0.0, "generation_time": 203.490923, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 224.434911, "forward_backward_time": 92.086401, "rollout_overhead_time": 20.943988 }, { "step": 30, "loss": -8.894503116607666e-05, "grad_norm": 0.064453125, "learning_rate": 4.2000000000000004e-06, "entropy": 0.047549558454193175, "clip_ratio/low_mean": 0.0011081278935307637, "clip_ratio/low_min": 0.0011081278935307637, "clip_ratio/high_mean": 0.00021222411305643618, "clip_ratio/high_max": 0.00021222411305643618, "clip_ratio/region_mean": 0.0013203520065871999, "step_time": 92.2031235839977, "entropy_control/entropy": 0.040190997737661806, "entropy_control/target": 0.03031625726079178, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "epoch": 0.12, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 92.090239, "rollout_overhead_time": 0.0, "reward": null }, { "step": 31, "loss": 0.00014733633724972606, "grad_norm": 0.15234375, "learning_rate": 4.000000000000001e-06, "num_tokens": 458054.0, "completions/mean_length": 366.5, "completions/min_length": 245.0, "completions/max_length": 875.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 366.5, "completions/min_terminated_length": 245.0, "completions/max_terminated_length": 875.0, "tools/call_frequency": 4.0, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/mean": 0.744084894657135, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/std": 0.22995175421237946, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/mean": 0.3125, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/std": 0.13908715546131134, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/mean": 0.75, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/std": 0.7071067690849304, "reward": 0.9097099304199219, "reward_std": 0.37770724296569824, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.5, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.020744612906128168, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 290.5032301099991, "entropy_control/entropy": 0.02546891146378123, "entropy_control/target": 0.03031625726079178, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "epoch": 0.124, "adapter_sync_time": 0.0, "generation_time": 197.472944, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 215.364881, "forward_backward_time": 75.02767, "rollout_overhead_time": 17.891936 }, { "step": 32, "loss": 0.00017066244618035853, "grad_norm": 0.1630859375, "learning_rate": 3.8000000000000005e-06, "entropy": 0.020651699567679316, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0012177512398920953, "clip_ratio/high_max": 0.0012177512398920953, "clip_ratio/region_mean": 0.0012177512398920953, "step_time": 75.1522636180016, "entropy_control/entropy": 0.025430285065361977, "entropy_control/target": 0.03031625726079178, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "epoch": 0.128, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 75.043663, "rollout_overhead_time": 0.0, "reward": null }, { "step": 33, "loss": 0.00014857196947559714, "grad_norm": 0.5546875, "learning_rate": 3.6000000000000003e-06, "num_tokens": 483745.0, "completions/mean_length": 311.875, "completions/min_length": 260.0, "completions/max_length": 408.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 311.875, "completions/min_terminated_length": 260.0, "completions/max_terminated_length": 408.0, "tools/call_frequency": 3.5, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.6119922995567322, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 1.5102457382454304e-06, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.3333333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/mean": 0.8293945789337158, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/std": 0.001506144879385829, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/std": 0.13608276844024658, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/std": 0.0, "reward": 0.9415267705917358, "reward_std": 0.12078285962343216, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.07779017789289355, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 263.0930055120007, "entropy_control/entropy": 0.07918635040678239, "entropy_control/target": 0.03031625726079178, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "epoch": 0.132, "adapter_sync_time": 0.0, "generation_time": 160.597135, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 179.652276, "forward_backward_time": 83.324273, "rollout_overhead_time": 19.055141 }, { "step": 34, "loss": -0.0004045320674777031, "grad_norm": 0.248046875, "learning_rate": 3.4000000000000005e-06, "entropy": 0.07796472008340061, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0020050798484589905, "clip_ratio/high_max": 0.0020050798484589905, "clip_ratio/region_mean": 0.0020050798484589905, "step_time": 83.42563158399844, "entropy_control/entropy": 0.07932321980524848, "entropy_control/target": 0.03031625726079178, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "epoch": 0.136, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 83.313219, "rollout_overhead_time": 0.0, "reward": null }, { "step": 35, "loss": -0.062484174966812134, "grad_norm": 0.2158203125, "learning_rate": 3.2000000000000003e-06, "num_tokens": 518902.0, "completions/mean_length": 527.625, "completions/min_length": 217.0, "completions/max_length": 972.0, "completions/clipped_ratio": 0.125, "completions/mean_terminated_length": 464.14288330078125, "completions/min_terminated_length": 217.0, "completions/max_terminated_length": 758.0, "tools/call_frequency": 3.5, "tools/failure_frequency": 0.0714285746216774, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/mean": 0.3409881293773651, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/std": 0.16847732663154602, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/mean": 0.1041666716337204, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/std": 0.19795581698417664, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/mean": -0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/std": 0.9258201122283936, "reward": 0.246196448802948, "reward_std": 0.3632914125919342, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.019984100246801972, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 343.58754661199964, "entropy_control/entropy": 0.024188085687493994, "entropy_control/target": 0.03031625726079178, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "epoch": 0.14, "adapter_sync_time": 0.0, "generation_time": 224.77326, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 246.9039, "forward_backward_time": 96.566652, "rollout_overhead_time": 22.13064 }, { "step": 36, "loss": -0.06240898370742798, "grad_norm": 0.1767578125, "learning_rate": 3e-06, "entropy": 0.019961558049544692, "clip_ratio/low_mean": 0.002036069316091016, "clip_ratio/low_min": 0.002036069316091016, "clip_ratio/high_mean": 0.00020161290012765676, "clip_ratio/high_max": 0.00020161290012765676, "clip_ratio/region_mean": 0.0022376822162186727, "step_time": 96.65855460000057, "entropy_control/entropy": 0.024303713555688736, "entropy_control/target": 0.03031625726079178, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "epoch": 0.144, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 96.552414, "rollout_overhead_time": 0.0, "reward": null }, { "step": 37, "loss": -0.00022276118397712708, "grad_norm": 0.1337890625, "learning_rate": 2.8000000000000003e-06, "num_tokens": 544877.0, "completions/mean_length": 362.875, "completions/min_length": 257.0, "completions/max_length": 611.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 362.875, "completions/min_terminated_length": 257.0, "completions/max_terminated_length": 611.0, "tools/call_frequency": 4.25, "tools/failure_frequency": 0.3529411852359772, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.615119218826294, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 0.0009891889058053493, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.4166666865348816, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.1666666716337204, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__native/mean": 0.6181600093841553, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__native/std": 1.0589964176688227e-06, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__efficiency/mean": 0.6666666865348816, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__efficiency/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__reliability/std": 0.0, "reward": 0.843722939491272, "reward_std": 0.010307898744940758, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.5, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.06896448787301779, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 270.79194448, "entropy_control/entropy": 0.06577244641272037, "entropy_control/target": 0.03031625726079178, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "epoch": 0.148, "adapter_sync_time": 0.0, "generation_time": 167.833593, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 186.991626, "forward_backward_time": 83.689616, "rollout_overhead_time": 19.158032 }, { "step": 38, "loss": -0.0005158446729183197, "grad_norm": 0.123046875, "learning_rate": 2.6e-06, "entropy": 0.0692716515623033, "clip_ratio/low_mean": 0.0011267974332440645, "clip_ratio/low_min": 0.0011267974332440645, "clip_ratio/high_mean": 0.0018399592663627118, "clip_ratio/high_max": 0.0018399592663627118, "clip_ratio/region_mean": 0.0029667566996067762, "step_time": 83.78387444800137, "entropy_control/entropy": 0.06597887336379939, "entropy_control/target": 0.03031625726079178, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "epoch": 0.152, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 83.676588, "rollout_overhead_time": 0.0, "reward": null }, { "step": 39, "loss": 0.00022244127467274666, "grad_norm": 0.1376953125, "learning_rate": 2.4000000000000003e-06, "num_tokens": 572401.0, "completions/mean_length": 345.5, "completions/min_length": 261.0, "completions/max_length": 436.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 345.5, "completions/min_terminated_length": 261.0, "completions/max_terminated_length": 436.0, "tools/call_frequency": 3.75, "tools/failure_frequency": 0.4000000059604645, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__native/mean": 0.6180810928344727, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__native/std": 0.0004928380949422717, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__efficiency/mean": 0.625, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__efficiency/std": 0.07715168595314026, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__reliability/std": 0.0, "reward": 0.8493311405181885, "reward_std": 0.0041687930934131145, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.07258317992091179, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 195.94552210799975, "entropy_control/entropy": 0.07133666538272915, "entropy_control/target": 0.03031625726079178, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "epoch": 0.156, "adapter_sync_time": 0.0, "generation_time": 115.964442, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 131.192693, "forward_backward_time": 64.650655, "rollout_overhead_time": 15.22825 }, { "step": 40, "loss": 0.0003291212487965822, "grad_norm": 0.162109375, "learning_rate": 2.2e-06, "entropy": 0.07260472676716745, "clip_ratio/low_mean": 0.0007578451477456838, "clip_ratio/low_min": 0.0007578451477456838, "clip_ratio/high_mean": 0.0038300505839288235, "clip_ratio/high_max": 0.0038300505839288235, "clip_ratio/region_mean": 0.004587895731674507, "step_time": 64.75038468500134, "entropy_control/entropy": 0.07133906537595829, "entropy_control/target": 0.03031625726079178, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "epoch": 0.16, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 64.643563, "rollout_overhead_time": 0.0, "reward": null }, { "step": 41, "loss": 0.0001470815623179078, "grad_norm": 0.1572265625, "learning_rate": 2.0000000000000003e-06, "num_tokens": 603510.0, "completions/mean_length": 407.625, "completions/min_length": 273.0, "completions/max_length": 694.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 407.625, "completions/min_terminated_length": 273.0, "completions/max_terminated_length": 694.0, "tools/call_frequency": 2.625, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__native/mean": 0.6172691583633423, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__native/std": 1.3644126966028125e-06, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__efficiency/mean": 0.6666666865348816, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__efficiency/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/mean": 0.5252507925033569, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/std": 0.1835046112537384, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/mean": 0.3333333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/std": 0.2357022762298584, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/std": 1.0, "reward": 0.7462599277496338, "reward_std": 0.2814081013202667, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.0738231724826619, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 260.5962830739991, "entropy_control/entropy": 0.0669646793525761, "entropy_control/target": 0.03031625726079178, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "epoch": 0.164, "adapter_sync_time": 0.0, "generation_time": 157.18409, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 176.584739, "forward_backward_time": 83.907308, "rollout_overhead_time": 19.400649 }, { "step": 42, "loss": 0.0007077767513692379, "grad_norm": 0.49609375, "learning_rate": 1.8000000000000001e-06, "entropy": 0.07389897236134857, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0009240323852282017, "clip_ratio/high_max": 0.0009240323852282017, "clip_ratio/region_mean": 0.0009240323852282017, "step_time": 84.02459510299923, "entropy_control/entropy": 0.06693753250717527, "entropy_control/target": 0.03031625726079178, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "epoch": 0.168, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 83.915515, "rollout_overhead_time": 0.0, "reward": null }, { "step": 43, "loss": 1.516193151473999e-06, "grad_norm": 0.291015625, "learning_rate": 1.6000000000000001e-06, "num_tokens": 632765.0, "completions/mean_length": 348.875, "completions/min_length": 265.0, "completions/max_length": 457.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 348.875, "completions/min_terminated_length": 265.0, "completions/max_terminated_length": 457.0, "tools/call_frequency": 3.375, "tools/failure_frequency": 0.29629629850387573, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__native/mean": 0.6178746223449707, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__native/std": 0.0008552604122087359, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__efficiency/mean": 0.5833333730697632, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__efficiency/std": 0.09622505307197571, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/mean": 0.6169817447662354, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/std": 0.0007432500715367496, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/mean": 0.625, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/std": 0.0833333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/std": 0.0, "reward": 0.8476365208625793, "reward_std": 0.00494955200701952, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.07686966517940164, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 208.75057098300022, "entropy_control/entropy": 0.07832597108465249, "entropy_control/target": 0.03031625726079178, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "epoch": 0.172, "adapter_sync_time": 0.0, "generation_time": 121.30173, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 138.014003, "forward_backward_time": 70.635289, "rollout_overhead_time": 16.712273 }, { "step": 44, "loss": -0.0003779083490371704, "grad_norm": 0.1865234375, "learning_rate": 1.4000000000000001e-06, "entropy": 0.07664299523457885, "clip_ratio/low_mean": 0.0007155550993047655, "clip_ratio/low_min": 0.0007155550993047655, "clip_ratio/high_mean": 0.003679176967125386, "clip_ratio/high_max": 0.003679176967125386, "clip_ratio/region_mean": 0.0043947320664301515, "step_time": 70.72571368899844, "entropy_control/entropy": 0.07810436976089705, "entropy_control/target": 0.03031625726079178, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "epoch": 0.176, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 70.627355, "rollout_overhead_time": 0.0, "reward": null }, { "step": 45, "loss": -3.8817524909973145e-06, "grad_norm": 0.193359375, "learning_rate": 1.2000000000000002e-06, "num_tokens": 665829.0, "completions/mean_length": 459.0, "completions/min_length": 287.0, "completions/max_length": 639.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 459.0, "completions/min_terminated_length": 287.0, "completions/max_terminated_length": 639.0, "tools/call_frequency": 3.5, "tools/failure_frequency": 0.1428571492433548, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/mean": 0.3412134647369385, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/std": 0.18242695927619934, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/mean": 0.0833333358168602, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/std": 0.1666666716337204, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/mean": -0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/std": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/mean": 0.6146758198738098, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/std": 0.0014483736595138907, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/mean": 0.625, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/std": 0.0833333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/std": 0.0, "reward": 0.5456529855728149, "reward_std": 0.41048502922058105, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.05643290979787707, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 289.95140555000035, "entropy_control/entropy": 0.05734015628450203, "entropy_control/target": 0.03031625726079178, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "epoch": 0.18, "adapter_sync_time": 0.0, "generation_time": 193.896552, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 212.203438, "forward_backward_time": 77.64727, "rollout_overhead_time": 18.306887 }, { "step": 46, "loss": 0.0005619600415229797, "grad_norm": 0.251953125, "learning_rate": 1.0000000000000002e-06, "entropy": 0.056736917700618505, "clip_ratio/low_mean": 0.0010034948209067807, "clip_ratio/low_min": 0.0010034948209067807, "clip_ratio/high_mean": 0.002066292188828811, "clip_ratio/high_max": 0.002066292188828811, "clip_ratio/region_mean": 0.0030697870097355917, "step_time": 77.75101602400173, "entropy_control/entropy": 0.05755260381565072, "entropy_control/target": 0.03031625726079178, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "epoch": 0.184, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 77.654512, "rollout_overhead_time": 0.0, "reward": null }, { "step": 47, "loss": -0.00021885125897824764, "grad_norm": 0.3203125, "learning_rate": 8.000000000000001e-07, "num_tokens": 694625.0, "completions/mean_length": 309.0, "completions/min_length": 238.0, "completions/max_length": 364.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 309.0, "completions/min_terminated_length": 238.0, "completions/max_terminated_length": 364.0, "tools/call_frequency": 2.5, "tools/failure_frequency": 0.05000000074505806, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/mean": 0.6166108846664429, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/std": 1.3838036920788e-06, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/mean": 0.6666666865348816, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/mean": 0.8314213752746582, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/std": 0.0010649171890690923, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/mean": 0.5416666865348816, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/std": 0.15957120060920715, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/std": 0.0, "reward": 0.9542244672775269, "reward_std": 0.11163643002510071, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.08227347710635513, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 191.2153643629972, "entropy_control/entropy": 0.08541661103700529, "entropy_control/target": 0.03031625726079178, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "epoch": 0.188, "adapter_sync_time": 0.0, "generation_time": 100.440939, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 117.617273, "forward_backward_time": 73.502514, "rollout_overhead_time": 17.176334 }, { "step": 48, "loss": -0.00024396576918661594, "grad_norm": 0.138671875, "learning_rate": 6.000000000000001e-07, "entropy": 0.08149674907326698, "clip_ratio/low_mean": 0.0005252101109363139, "clip_ratio/low_min": 0.0005252101109363139, "clip_ratio/high_mean": 0.0033328040735796094, "clip_ratio/high_max": 0.0033328040735796094, "clip_ratio/region_mean": 0.0038580141845159233, "step_time": 73.5989647940005, "entropy_control/entropy": 0.08463391225243246, "entropy_control/target": 0.03031625726079178, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "epoch": 0.192, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 73.500912, "rollout_overhead_time": 0.0, "reward": null }, { "step": 49, "loss": 6.984919309616089e-09, "grad_norm": 0.1298828125, "learning_rate": 4.0000000000000003e-07, "num_tokens": 723945.0, "completions/mean_length": 370.0, "completions/min_length": 244.0, "completions/max_length": 651.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 370.0, "completions/min_terminated_length": 244.0, "completions/max_terminated_length": 651.0, "tools/call_frequency": 3.5, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/mean": 0.5234451293945312, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/std": 0.18229824304580688, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/mean": 0.4583333730697632, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/std": 0.3154948949813843, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/std": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/mean": 0.8265383839607239, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/std": 1.6115315020215348e-06, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/mean": 0.3333333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/std": 0.0, "reward": 0.844783365726471, "reward_std": 0.3357786536216736, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.036323420819826424, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 274.59873056799916, "entropy_control/entropy": 0.0429283501309586, "entropy_control/target": 0.03031625726079178, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "epoch": 0.196, "adapter_sync_time": 0.0, "generation_time": 176.057319, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 194.636102, "forward_backward_time": 79.864304, "rollout_overhead_time": 18.578783 }, { "step": 50, "loss": 7.894332520663738e-05, "grad_norm": 0.087890625, "learning_rate": 2.0000000000000002e-07, "entropy": 0.036797207430936396, "clip_ratio/low_mean": 0.0014213859685696661, "clip_ratio/low_min": 0.0014213859685696661, "clip_ratio/high_mean": 0.0018518302822485566, "clip_ratio/high_max": 0.0018518302822485566, "clip_ratio/region_mean": 0.0032732162508182228, "step_time": 79.95465855900147, "entropy_control/entropy": 0.04327416916451558, "entropy_control/target": 0.03031625726079178, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "epoch": 0.2, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 79.856897, "rollout_overhead_time": 0.0, "reward": null } ], "holdout": [ { "seed": 900000, "sample": 0, "native_reward": 0.6155619431579329, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 273, "reward_channels": { "native": 0.6155619431579329, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "two_pool_cycle" }, { "seed": 900000, "sample": 1, "native_reward": 0.615560891281058, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 343, "reward_channels": { "native": 0.615560891281058, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "two_pool_cycle" }, { "seed": 900000, "sample": 2, "native_reward": 0.6155606664493392, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 252, "reward_channels": { "native": 0.6155606664493392, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "two_pool_cycle" }, { "seed": 900000, "sample": 3, "native_reward": 0.6164592739162532, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 3, "model_tokens": 274, "reward_channels": { "native": 0.6164592739162532, "efficiency": 0.5, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "two_pool_cycle" }, { "seed": 900001, "sample": 0, "native_reward": 0.6175072328484663, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 273, "reward_channels": { "native": 0.6175072328484663, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "two_pool_cycle" }, { "seed": 900001, "sample": 1, "native_reward": 0.6175057798087202, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 298, "reward_channels": { "native": 0.6175057798087202, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "two_pool_cycle" }, { "seed": 900001, "sample": 2, "native_reward": 0.6175061121677046, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 298, "reward_channels": { "native": 0.6175061121677046, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "two_pool_cycle" }, { "seed": 900001, "sample": 3, "native_reward": 0.6175069963866499, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 286, "reward_channels": { "native": 0.6175069963866499, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "two_pool_cycle" }, { "seed": 900000, "sample": 0, "native_reward": 0.6169302065811881, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 3, "model_tokens": 320, "reward_channels": { "native": 0.6169302065811881, "efficiency": 0.5, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "triangular_cycle" }, { "seed": 900000, "sample": 1, "native_reward": 0.6184861076618522, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 268, "reward_channels": { "native": 0.6184861076618522, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "triangular_cycle" }, { "seed": 900000, "sample": 2, "native_reward": 0.6184856035464226, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 258, "reward_channels": { "native": 0.6184856035464226, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "triangular_cycle" }, { "seed": 900000, "sample": 3, "native_reward": 0.618486948443024, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 872, "reward_channels": { "native": 0.618486948443024, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "triangular_cycle" }, { "seed": 900001, "sample": 0, "native_reward": 0.6176954349414234, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 3, "model_tokens": 243, "reward_channels": { "native": 0.6176954349414234, "efficiency": 0.5, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "triangular_cycle" }, { "seed": 900001, "sample": 1, "native_reward": 0.6176957130694314, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 3, "model_tokens": 316, "reward_channels": { "native": 0.6176957130694314, "efficiency": 0.5, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "triangular_cycle" }, { "seed": 900001, "sample": 2, "native_reward": 0.6184896583035329, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 409, "reward_channels": { "native": 0.6184896583035329, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "triangular_cycle" }, { "seed": 900001, "sample": 3, "native_reward": 0.6184886158270289, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 341, "reward_channels": { "native": 0.6184886158270289, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "triangular_cycle" }, { "seed": 900000, "sample": 0, "native_reward": 0.25, "promotion_success": false, "done": false, "errors": 0, "tool_steps": 4, "model_tokens": 723, "reward_channels": { "native": 0.25, "efficiency": 0.0, "reliability": -1.0 }, "verification_level": null, "failed_gates": [ "finalized" ], "family": "competing_cycles" }, { "seed": 900000, "sample": 1, "native_reward": 0.6201801608919191, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 3, "model_tokens": 421, "reward_channels": { "native": 0.6201801608919191, "efficiency": 0.5, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "competing_cycles" }, { "seed": 900000, "sample": 2, "native_reward": 0.25, "promotion_success": false, "done": false, "errors": 0, "tool_steps": 4, "model_tokens": 594, "reward_channels": { "native": 0.25, "efficiency": 0.0, "reliability": -1.0 }, "verification_level": null, "failed_gates": [ "finalized" ], "family": "competing_cycles" }, { "seed": 900000, "sample": 3, "native_reward": 0.6182885244004738, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 562, "reward_channels": { "native": 0.6182885244004738, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "competing_cycles" }, { "seed": 900001, "sample": 0, "native_reward": 0.6185548230263636, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 583, "reward_channels": { "native": 0.6185548230263636, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "competing_cycles" }, { "seed": 900001, "sample": 1, "native_reward": 0.6207064180086292, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 3, "model_tokens": 392, "reward_channels": { "native": 0.6207064180086292, "efficiency": 0.5, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "competing_cycles" }, { "seed": 900001, "sample": 2, "native_reward": 0.6207075640003088, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 3, "model_tokens": 406, "reward_channels": { "native": 0.6207075640003088, "efficiency": 0.5, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "competing_cycles" }, { "seed": 900001, "sample": 3, "native_reward": 0.6207058199470863, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 3, "model_tokens": 685, "reward_channels": { "native": 0.6207058199470863, "efficiency": 0.5, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "competing_cycles" }, { "seed": 900000, "sample": 0, "native_reward": 0.6175192419545681, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 291, "reward_channels": { "native": 0.6175192419545681, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "four_hop_cycle" }, { "seed": 900000, "sample": 1, "native_reward": 0.6167561987402518, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 3, "model_tokens": 316, "reward_channels": { "native": 0.6167561987402518, "efficiency": 0.5, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "four_hop_cycle" }, { "seed": 900000, "sample": 2, "native_reward": 0.6175174288668923, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 335, "reward_channels": { "native": 0.6175174288668923, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "four_hop_cycle" }, { "seed": 900000, "sample": 3, "native_reward": 0.6167551775519118, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 3, "model_tokens": 367, "reward_channels": { "native": 0.6167551775519118, "efficiency": 0.5, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "four_hop_cycle" }, { "seed": 900001, "sample": 0, "native_reward": 0.6168364043457519, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 273, "reward_channels": { "native": 0.6168364043457519, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "four_hop_cycle" }, { "seed": 900001, "sample": 1, "native_reward": 0.616834342416455, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 503, "reward_channels": { "native": 0.616834342416455, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "four_hop_cycle" }, { "seed": 900001, "sample": 2, "native_reward": 0.6168348950551854, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 324, "reward_channels": { "native": 0.6168348950551854, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "four_hop_cycle" }, { "seed": 900001, "sample": 3, "native_reward": 0.6154729421455175, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 286, "reward_channels": { "native": 0.6154729421455175, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "four_hop_cycle" }, { "seed": 900000, "sample": 0, "native_reward": 0.8346269127250262, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 3, "model_tokens": 307, "reward_channels": { "native": 0.8346269127250262, "efficiency": 0.5, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "size_sensitive_cycle" }, { "seed": 900000, "sample": 1, "native_reward": 0.8346271191178777, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 3, "model_tokens": 319, "reward_channels": { "native": 0.8346271191178777, "efficiency": 0.5, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "size_sensitive_cycle" }, { "seed": 900000, "sample": 2, "native_reward": 0.8346283987501825, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 3, "model_tokens": 333, "reward_channels": { "native": 0.8346283987501825, "efficiency": 0.5, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "size_sensitive_cycle" }, { "seed": 900000, "sample": 3, "native_reward": 0.8346278914886784, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 3, "model_tokens": 302, "reward_channels": { "native": 0.8346278914886784, "efficiency": 0.5, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "size_sensitive_cycle" }, { "seed": 900001, "sample": 0, "native_reward": 0.8316970841532554, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 3, "model_tokens": 351, "reward_channels": { "native": 0.8316970841532554, "efficiency": 0.5, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "size_sensitive_cycle" }, { "seed": 900001, "sample": 1, "native_reward": 0.8316978576102672, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 3, "model_tokens": 317, "reward_channels": { "native": 0.8316978576102672, "efficiency": 0.5, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "size_sensitive_cycle" }, { "seed": 900001, "sample": 2, "native_reward": 0.8316978253037827, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 3, "model_tokens": 351, "reward_channels": { "native": 0.8316978253037827, "efficiency": 0.5, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "size_sensitive_cycle" }, { "seed": 900001, "sample": 3, "native_reward": 0.8307404175709702, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 242, "reward_channels": { "native": 0.8307404175709702, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "size_sensitive_cycle" } ] }, { "run_id": "1789152686", "mode": "no-think", "method": "dapo", "steps": 50, "reward_points": 25, "train_mean": 0.6519488525390625, "train_last10": 0.7141775667667389, "holdout_native": 0.646690062246073, "holdout_episodes": 40, "holdout_successes": 39, "mean_step_seconds": 176.37912706865995, "label": "DAPO", "state_at_cutoff": "complete", "step_hours": 2.4497100981758324, "peak_allocated_gib": null, "peak_reserved_gib": null, "extra_refill_rounds": 0.0, "groups_refilled": 0.0, "holdout_component_means": { "native": 0.646690062246073, "efficiency": 0.5791666666666667, "reliability": 0.95 }, "configuration": { "model": { "id": "Qwen/Qwen3-14B", "revision": "40c069824f4251a91eefaf281ebe4c544efd3e18", "loss": "dapo", "algorithm": "grpo", "no_think": true, "lora": true, "quantize": true, "lora_r": 32, "lora_alpha": 64, "use_dora": false }, "loss": "dapo", "normalization": "joint", "entropy_control": { "method": "monitor", "epsilon_low": 0.2, "epsilon_high": 0.28, "zeta_initial": 0.001, "zeta_min": 0.0001, "zeta_max": 0.05, "holdout_split": "benchmark", "holdout_seeds_per_family": 2, "num_iterations": 2 }, "dynamic_sampling": false, "sampling": { "group_size": 4, "temperature": 1.0, "max_completion_tokens": 8192, "top_entropy_quantile": 1.0, "temperature_decay_to": null }, "learning_rate": 1e-05, "turn_max_tokens": 1024 }, "source_sha256": { "metrics.jsonl": "585b3af741dbe64511cb6be1c1f1b58f5386ec0fa897b7ce4da57439318d1327", "entropy_holdout_episodes.jsonl": "413cfe0af0e836728d41fa258792fb938cfdb5a93d80e3eb5293f24b9b3ed42b", "resolved_training_config.json": "1629d5372941013330db5b2061b5713115746e4b1feb94b8552b37f64d0240e3" }, "training_steps": [ { "step": 1, "loss": 0.3692533075809479, "grad_norm": 0.177734375, "learning_rate": 1e-05, "num_tokens": 27996.0, "completions/mean_length": 603.0, "completions/min_length": 278.0, "completions/max_length": 1190.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 603.0, "completions/min_terminated_length": 278.0, "completions/max_terminated_length": 1190.0, "tools/call_frequency": 4.375, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.5008535981178284, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 0.21723572909832, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.25, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.1666666716337204, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/mean": 0.3369031250476837, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/std": 0.3238062560558319, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/mean": 0.1666666716337204, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/std": 0.19245010614395142, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/mean": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/std": 1.154700517654419, "reward": 0.4792949855327606, "reward_std": 0.45436012744903564, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.04867766337702051, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 340.52523792299996, "entropy_control/entropy": 0.03031625726079178, "entropy_control/target": 0.03031625726079178, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.004, "adapter_sync_time": 0.0, "generation_time": 237.549989, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 257.894146, "forward_backward_time": 82.543066, "rollout_overhead_time": 20.344157 }, { "step": 2, "loss": 0.3688158392906189, "grad_norm": 0.1240234375, "learning_rate": 9.800000000000001e-06, "entropy": 0.048358485801145434, "clip_ratio/low_mean": 0.0006929489754838869, "clip_ratio/low_min": 0.0006929489754838869, "clip_ratio/high_mean": 0.0011890835303347558, "clip_ratio/high_max": 0.0011890835303347558, "clip_ratio/region_mean": 0.0018820325058186427, "step_time": 82.57729190400005, "entropy_control/entropy": 0.030192915657265968, "entropy_control/target": 0.03031625726079178, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.008, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 82.483635, "rollout_overhead_time": 0.0, "reward": null }, { "step": 3, "loss": -6.873492384329438e-05, "grad_norm": 0.00157928466796875, "learning_rate": 9.600000000000001e-06, "num_tokens": 52250.0, "completions/mean_length": 315.75, "completions/min_length": 268.0, "completions/max_length": 441.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 315.75, "completions/min_terminated_length": 268.0, "completions/max_terminated_length": 441.0, "tools/call_frequency": 4.125, "tools/failure_frequency": 0.03030303120613098, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.6156482696533203, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 0.0003735041245818138, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.3333333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 0.0, "reward": 0.8323149681091309, "reward_std": 0.0003735041245818138, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 1.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.073490837123245, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 225.7903535270002, "entropy_control/entropy": 0.07463982666444095, "entropy_control/target": 0.03031625726079178, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.012, "adapter_sync_time": 0.0, "generation_time": 127.254493, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 145.6265, "forward_backward_time": 80.06869, "rollout_overhead_time": 18.372007 }, { "step": 4, "loss": -6.031058728694916e-05, "grad_norm": 0.002166748046875, "learning_rate": 9.4e-06, "entropy": 0.07436912599951029, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 80.15263358700031, "entropy_control/entropy": 0.07551235631313276, "entropy_control/target": 0.03031625726079178, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.016, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 80.060743, "rollout_overhead_time": 0.0, "reward": null }, { "step": 5, "loss": -0.00023701813188381493, "grad_norm": 0.0037689208984375, "learning_rate": 9.200000000000002e-06, "num_tokens": 76602.0, "completions/mean_length": 335.0, "completions/min_length": 298.0, "completions/max_length": 412.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 335.0, "completions/min_terminated_length": 298.0, "completions/max_terminated_length": 412.0, "tools/call_frequency": 4.5, "tools/failure_frequency": 0.1111111119389534, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.6156157851219177, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 4.692983566201292e-05, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.3333333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 0.0, "reward": 0.8322824239730835, "reward_std": 4.692983566201292e-05, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.5, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.09296132577583194, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 233.43878169599998, "entropy_control/entropy": 0.0943122746853232, "entropy_control/target": 0.03031625726079178, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.02, "adapter_sync_time": 0.0, "generation_time": 134.944179, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 153.30447, "forward_backward_time": 80.030852, "rollout_overhead_time": 18.360291 }, { "step": 6, "loss": -0.00024486094480380416, "grad_norm": 0.003082275390625, "learning_rate": 9e-06, "entropy": 0.09225809015333652, "clip_ratio/low_mean": 0.0010964519169647247, "clip_ratio/low_min": 0.0010964519169647247, "clip_ratio/high_mean": 0.0012057877611368895, "clip_ratio/high_max": 0.0012057877611368895, "clip_ratio/region_mean": 0.002302239678101614, "step_time": 80.14518817899989, "entropy_control/entropy": 0.09359416556827227, "entropy_control/target": 0.03031625726079178, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.024, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 80.040992, "rollout_overhead_time": 0.0, "reward": null }, { "step": 7, "loss": -0.07327378541231155, "grad_norm": 0.2333984375, "learning_rate": 8.8e-06, "num_tokens": 107048.0, "completions/mean_length": 334.25, "completions/min_length": 247.0, "completions/max_length": 582.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 334.25, "completions/min_terminated_length": 247.0, "completions/max_terminated_length": 582.0, "tools/call_frequency": 3.625, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/mean": 0.3412603735923767, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/std": 0.1825207769870758, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/mean": 0.0833333358168602, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/std": 0.1666666716337204, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/mean": -0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/std": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/mean": 0.6824637651443481, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/std": 0.2883092164993286, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/mean": 0.25, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/std": 0.1666666716337204, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/std": 1.0, "reward": 0.5201954245567322, "reward_std": 0.5074084997177124, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.026138133136555552, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 259.96634932300003, "entropy_control/entropy": 0.02442254099857785, "entropy_control/target": 0.03031625726079178, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.028, "adapter_sync_time": 0.0, "generation_time": 164.944698, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 183.033936, "forward_backward_time": 76.835146, "rollout_overhead_time": 18.089238 }, { "step": 8, "loss": -0.07358121126890182, "grad_norm": 0.162109375, "learning_rate": 8.6e-06, "entropy": 0.026456466526724398, "clip_ratio/low_mean": 0.0005252101109363139, "clip_ratio/low_min": 0.0005252101109363139, "clip_ratio/high_mean": 0.0004180601972620934, "clip_ratio/high_max": 0.0004180601972620934, "clip_ratio/region_mean": 0.0009432703081984073, "step_time": 76.91976198899988, "entropy_control/entropy": 0.024563744591242016, "entropy_control/target": 0.03031625726079178, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.032, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 76.824628, "rollout_overhead_time": 0.0, "reward": null }, { "step": 9, "loss": 0.18830087780952454, "grad_norm": 0.0986328125, "learning_rate": 8.400000000000001e-06, "num_tokens": 138420.0, "completions/mean_length": 633.5, "completions/min_length": 286.0, "completions/max_length": 924.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 633.5, "completions/min_terminated_length": 286.0, "completions/max_terminated_length": 924.0, "tools/call_frequency": 4.25, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.3922732174396515, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 0.2508854866027832, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.1666666716337204, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.19245010614395142, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 1.154700517654419, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/mean": 0.3410886824131012, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/std": 0.1821773648262024, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/mean": 0.0833333358168602, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/std": 0.1666666716337204, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/mean": -0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/std": 1.0, "reward": 0.3229309320449829, "reward_std": 0.4192386567592621, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.026950508821755648, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 326.76431794399923, "entropy_control/entropy": 0.018794556699214083, "entropy_control/target": 0.03031625726079178, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.036, "adapter_sync_time": 0.0, "generation_time": 213.805217, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 234.773073, "forward_backward_time": 91.891128, "rollout_overhead_time": 20.967856 }, { "step": 10, "loss": 0.18811774253845215, "grad_norm": 0.08203125, "learning_rate": 8.2e-06, "entropy": 0.02686283423099667, "clip_ratio/low_mean": 0.0011273209820501506, "clip_ratio/low_min": 0.0011273209820501506, "clip_ratio/high_mean": 0.00021186440426390618, "clip_ratio/high_max": 0.00021186440426390618, "clip_ratio/region_mean": 0.0013391853863140568, "step_time": 91.97408711800017, "entropy_control/entropy": 0.018791692803420808, "entropy_control/target": 0.03031625726079178, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.04, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 91.873202, "rollout_overhead_time": 0.0, "reward": null }, { "step": 11, "loss": 0.17540164291858673, "grad_norm": 0.173828125, "learning_rate": 8.000000000000001e-06, "num_tokens": 168150.0, "completions/mean_length": 403.25, "completions/min_length": 269.0, "completions/max_length": 771.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 403.25, "completions/min_terminated_length": 269.0, "completions/max_terminated_length": 771.0, "tools/call_frequency": 3.625, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__native/mean": 0.4943929612636566, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__native/std": 0.21344728767871857, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__efficiency/mean": 0.25, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__efficiency/std": 0.1666666716337204, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__reliability/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__reliability/std": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/mean": 0.6113900542259216, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/std": 0.0027538298163563013, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/mean": 0.4583333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/std": 0.15957120060920715, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/std": 0.0, "reward": 0.7205997705459595, "reward_std": 0.30162128806114197, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.0341033375589177, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 273.963373522, "entropy_control/entropy": 0.03113980793012608, "entropy_control/target": 0.03031625726079178, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.044, "adapter_sync_time": 0.0, "generation_time": 176.041426, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 194.856423, "forward_backward_time": 78.9992, "rollout_overhead_time": 18.814997 }, { "step": 12, "loss": 0.175516739487648, "grad_norm": 0.171875, "learning_rate": 7.800000000000002e-06, "entropy": 0.03414776746649295, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.00020833333837799728, "clip_ratio/high_max": 0.00020833333837799728, "clip_ratio/region_mean": 0.00020833333837799728, "step_time": 79.09812102199976, "entropy_control/entropy": 0.031178688400708438, "entropy_control/target": 0.03031625726079178, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.048, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 78.99074, "rollout_overhead_time": 0.0, "reward": null }, { "step": 13, "loss": 0.2720063328742981, "grad_norm": 0.173828125, "learning_rate": 7.600000000000001e-06, "num_tokens": 197090.0, "completions/mean_length": 328.5, "completions/min_length": 236.0, "completions/max_length": 743.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 328.5, "completions/min_terminated_length": 236.0, "completions/max_terminated_length": 743.0, "tools/call_frequency": 3.75, "tools/failure_frequency": 0.03333333507180214, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.5028147101402283, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 0.2185431271791458, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.25, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.1666666716337204, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/mean": 0.25, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/mean": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/mean": -1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/std": 0.0, "reward": 0.3326573371887207, "reward_std": 0.41158783435821533, "frac_reward_zero_std": 0.5, "frac_groups_zero_std": 0.5, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.04868619958870113, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 287.9703269009997, "entropy_control/entropy": 0.048345302553217145, "entropy_control/target": 0.03031625726079178, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.052, "adapter_sync_time": 0.0, "generation_time": 172.52584, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 193.699104, "forward_backward_time": 94.159163, "rollout_overhead_time": 21.173264 }, { "step": 14, "loss": 0.271788626909256, "grad_norm": 0.11669921875, "learning_rate": 7.4e-06, "entropy": 0.04911805596202612, "clip_ratio/low_mean": 0.00016823687474243343, "clip_ratio/low_min": 0.00016823687474243343, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.00016823687474243343, "step_time": 94.26841234099993, "entropy_control/entropy": 0.0487854574763968, "entropy_control/target": 0.03031625726079178, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.056, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 94.156891, "rollout_overhead_time": 0.0, "reward": null }, { "step": 15, "loss": -0.33696749806404114, "grad_norm": 0.33203125, "learning_rate": 7.2000000000000005e-06, "num_tokens": 227675.0, "completions/mean_length": 335.625, "completions/min_length": 239.0, "completions/max_length": 613.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 335.625, "completions/min_terminated_length": 239.0, "completions/max_terminated_length": 613.0, "tools/call_frequency": 3.5, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/mean": 0.34176012873649597, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/std": 0.18352025747299194, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/mean": 0.125, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/std": 0.25, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/mean": -0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/std": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/mean": 0.8657005429267883, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/std": 0.08177393674850464, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/mean": 0.3333333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/std": 0.0, "reward": 0.6651886701583862, "reward_std": 0.5186328887939453, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.030811693228315562, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 247.40627943599975, "entropy_control/entropy": 0.03400098974383643, "entropy_control/target": 0.03031625726079178, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.06, "adapter_sync_time": 0.0, "generation_time": 151.867485, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 170.110212, "forward_backward_time": 77.186803, "rollout_overhead_time": 18.242728 }, { "step": 16, "loss": -0.3372400104999542, "grad_norm": 0.32421875, "learning_rate": 7e-06, "entropy": 0.030748213626793586, "clip_ratio/low_mean": 0.00048449612222611904, "clip_ratio/low_min": 0.00048449612222611904, "clip_ratio/high_mean": 0.00044853356666862965, "clip_ratio/high_max": 0.00044853356666862965, "clip_ratio/region_mean": 0.0009330296888947487, "step_time": 77.28687519600044, "entropy_control/entropy": 0.03413712302972143, "entropy_control/target": 0.03031625726079178, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.064, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 77.185206, "rollout_overhead_time": 0.0, "reward": null }, { "step": 17, "loss": -0.15046238899230957, "grad_norm": 0.3046875, "learning_rate": 6.800000000000001e-06, "num_tokens": 260344.0, "completions/mean_length": 388.625, "completions/min_length": 271.0, "completions/max_length": 587.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 388.625, "completions/min_terminated_length": 271.0, "completions/max_terminated_length": 587.0, "tools/call_frequency": 3.125, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/mean": 0.5247765779495239, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/std": 0.1831853687763214, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/mean": 0.3333333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/std": 0.2357022762298584, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/std": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/mean": 0.6125996112823486, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/std": 0.0023507894948124886, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/std": 0.19245010614395142, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/std": 0.0, "reward": 0.7395214438438416, "reward_std": 0.2787351608276367, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.05782427568919957, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 254.2655020920006, "entropy_control/entropy": 0.0555637324306606, "entropy_control/target": 0.03031625726079178, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.068, "adapter_sync_time": 0.0, "generation_time": 151.547989, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 170.920427, "forward_backward_time": 83.241832, "rollout_overhead_time": 19.372439 }, { "step": 18, "loss": -0.1504790335893631, "grad_norm": 0.173828125, "learning_rate": 6.600000000000001e-06, "entropy": 0.05739053594879806, "clip_ratio/low_mean": 0.0008724388317205012, "clip_ratio/low_min": 0.0008724388317205012, "clip_ratio/high_mean": 0.00042517005931586027, "clip_ratio/high_max": 0.00042517005931586027, "clip_ratio/region_mean": 0.0012976088910363615, "step_time": 83.31394233899982, "entropy_control/entropy": 0.055122998891002195, "entropy_control/target": 0.03031625726079178, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.072, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 83.211267, "rollout_overhead_time": 0.0, "reward": null }, { "step": 19, "loss": 0.14910073578357697, "grad_norm": 0.291015625, "learning_rate": 6.4000000000000006e-06, "num_tokens": 286477.0, "completions/mean_length": 372.125, "completions/min_length": 255.0, "completions/max_length": 526.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 372.125, "completions/min_terminated_length": 255.0, "completions/max_terminated_length": 526.0, "tools/call_frequency": 3.875, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.6136557459831238, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 0.0015253729652613401, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.375, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.0833333283662796, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/mean": 0.394774854183197, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/std": 0.28954967856407166, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/mean": 0.0833333358168602, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/std": 0.1666666716337204, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/mean": -0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/std": 1.0, "reward": 0.5656735897064209, "reward_std": 0.43308326601982117, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.052682810463011265, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 272.061771614, "entropy_control/entropy": 0.048758424207517456, "entropy_control/target": 0.03031625726079178, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.076, "adapter_sync_time": 0.0, "generation_time": 167.421362, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 187.098197, "forward_backward_time": 84.851514, "rollout_overhead_time": 19.676835 }, { "step": 20, "loss": 0.14920605719089508, "grad_norm": 0.2314453125, "learning_rate": 6.200000000000001e-06, "entropy": 0.05313685396686196, "clip_ratio/low_mean": 0.0005488480965141207, "clip_ratio/low_min": 0.0005488480965141207, "clip_ratio/high_mean": 0.000469924823846668, "clip_ratio/high_max": 0.000469924823846668, "clip_ratio/region_mean": 0.0010187729203607887, "step_time": 84.94119402100023, "entropy_control/entropy": 0.04929954233653071, "entropy_control/target": 0.03031625726079178, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.08, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 84.833103, "rollout_overhead_time": 0.0, "reward": null }, { "step": 21, "loss": 0.01824653148651123, "grad_norm": 0.19921875, "learning_rate": 6e-06, "num_tokens": 319256.0, "completions/mean_length": 394.375, "completions/min_length": 245.0, "completions/max_length": 666.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 394.375, "completions/min_terminated_length": 245.0, "completions/max_terminated_length": 666.0, "tools/call_frequency": 3.5, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/mean": 0.4329206645488739, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/std": 0.21122044324874878, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/mean": 0.2083333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/std": 0.25, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/mean": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/std": 1.154700517654419, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/mean": 0.5206031203269958, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/std": 0.18040494620800018, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/mean": 0.2916666865348816, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/std": 0.20971763134002686, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/std": 1.0, "reward": 0.5392618775367737, "reward_std": 0.4051787257194519, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.031799523741938174, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 277.84228195500054, "entropy_control/entropy": 0.03242920339025762, "entropy_control/target": 0.03031625726079178, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.084, "adapter_sync_time": 0.0, "generation_time": 168.251986, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 188.42333, "forward_backward_time": 89.319367, "rollout_overhead_time": 20.171344 }, { "step": 22, "loss": 0.01798143982887268, "grad_norm": 0.171875, "learning_rate": 5.8e-06, "entropy": 0.03149362048134208, "clip_ratio/low_mean": 0.00037537538446485996, "clip_ratio/low_min": 0.00037537538446485996, "clip_ratio/high_mean": 0.0009347896557301283, "clip_ratio/high_max": 0.0009347896557301283, "clip_ratio/region_mean": 0.0013101650401949883, "step_time": 89.41478857799848, "entropy_control/entropy": 0.03225091970596338, "entropy_control/target": 0.03031625726079178, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.088, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 89.308033, "rollout_overhead_time": 0.0, "reward": null }, { "step": 23, "loss": 0.17221468687057495, "grad_norm": 0.1875, "learning_rate": 5.600000000000001e-06, "num_tokens": 345422.0, "completions/mean_length": 381.25, "completions/min_length": 238.0, "completions/max_length": 748.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 381.25, "completions/min_terminated_length": 238.0, "completions/max_terminated_length": 748.0, "tools/call_frequency": 3.875, "tools/failure_frequency": 0.09677419066429138, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.613832950592041, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 0.0028983973897993565, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.19245010614395142, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/mean": 0.6828960180282593, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/std": 0.28859731554985046, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/mean": 0.25, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/std": 0.1666666716337204, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/std": 1.0, "reward": 0.8171144723892212, "reward_std": 0.3262479603290558, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.05550521705299616, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 269.0507772079991, "entropy_control/entropy": 0.04823267800406146, "entropy_control/target": 0.03031625726079178, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.092, "adapter_sync_time": 0.0, "generation_time": 167.54063, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 186.482361, "forward_backward_time": 82.463555, "rollout_overhead_time": 18.941731 }, { "step": 24, "loss": 0.17187626659870148, "grad_norm": 0.146484375, "learning_rate": 5.400000000000001e-06, "entropy": 0.055170910665765405, "clip_ratio/low_mean": 0.0012920362350996584, "clip_ratio/low_min": 0.0012920362350996584, "clip_ratio/high_mean": 0.001443713903427124, "clip_ratio/high_max": 0.001443713903427124, "clip_ratio/region_mean": 0.0027357501385267824, "step_time": 82.55980333700063, "entropy_control/entropy": 0.048020255119890526, "entropy_control/target": 0.03031625726079178, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.096, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 82.453754, "rollout_overhead_time": 0.0, "reward": null }, { "step": 25, "loss": 0.036988019943237305, "grad_norm": 0.353515625, "learning_rate": 5.2e-06, "num_tokens": 372580.0, "completions/mean_length": 300.75, "completions/min_length": 265.0, "completions/max_length": 356.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 300.75, "completions/min_terminated_length": 265.0, "completions/max_terminated_length": 356.0, "tools/call_frequency": 3.375, "tools/failure_frequency": 0.14814814925193787, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.6139856576919556, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 0.0010810233652591705, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.4583333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.15957120060920715, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/mean": 0.6145153045654297, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/std": 0.001134276739321649, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/mean": 0.5833333730697632, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/std": 0.1666666716337204, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/std": 0.0, "reward": 0.8402920961380005, "reward_std": 0.009312445297837257, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.0713640390895307, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 246.0319621259996, "entropy_control/entropy": 0.0702834473347908, "entropy_control/target": 0.03031625726079178, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.1, "adapter_sync_time": 0.0, "generation_time": 138.623042, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 158.496741, "forward_backward_time": 87.430078, "rollout_overhead_time": 19.873699 }, { "step": 26, "loss": 0.036650970578193665, "grad_norm": 0.330078125, "learning_rate": 5e-06, "entropy": 0.07126661529764533, "clip_ratio/low_mean": 0.0003511235991027206, "clip_ratio/low_min": 0.0003511235991027206, "clip_ratio/high_mean": 0.0012477763812057674, "clip_ratio/high_max": 0.0012477763812057674, "clip_ratio/region_mean": 0.001598899980308488, "step_time": 87.54783761100043, "entropy_control/entropy": 0.07012220591020303, "entropy_control/target": 0.03031625726079178, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.104, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 87.441913, "rollout_overhead_time": 0.0, "reward": null }, { "step": 27, "loss": 0.2653420865535736, "grad_norm": 0.1025390625, "learning_rate": 4.800000000000001e-06, "num_tokens": 405260.0, "completions/mean_length": 778.5, "completions/min_length": 316.0, "completions/max_length": 1940.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 778.5, "completions/min_terminated_length": 316.0, "completions/max_terminated_length": 1940.0, "tools/call_frequency": 3.5, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/mean": 0.502800464630127, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/std": 0.21853949129581451, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/mean": 0.4166666865348816, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/std": 0.3191424012184143, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/std": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/mean": 0.2994753420352936, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/std": 0.20206774771213531, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/mean": 0.0833333358168602, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/std": 0.1666666716337204, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/mean": -0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/std": 1.0, "reward": 0.4136379063129425, "reward_std": 0.44966861605644226, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.03606779675465077, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 434.5469909829999, "entropy_control/entropy": 0.023290043880667306, "entropy_control/target": 0.03031625726079178, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.108, "adapter_sync_time": 0.0, "generation_time": 323.716093, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 344.14945, "forward_backward_time": 90.298629, "rollout_overhead_time": 20.433357 }, { "step": 28, "loss": 0.26548078656196594, "grad_norm": 0.1162109375, "learning_rate": 4.600000000000001e-06, "entropy": 0.03613937651971355, "clip_ratio/low_mean": 0.0008235201530624181, "clip_ratio/low_min": 0.0008235201530624181, "clip_ratio/high_mean": 0.0007868059328757226, "clip_ratio/high_max": 0.0007868059328757226, "clip_ratio/region_mean": 0.0016103260859381407, "step_time": 90.39110960100152, "entropy_control/entropy": 0.02332148559095522, "entropy_control/target": 0.03031625726079178, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.112, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 90.289309, "rollout_overhead_time": 0.0, "reward": null }, { "step": 29, "loss": -0.0823657289147377, "grad_norm": 0.10009765625, "learning_rate": 4.4e-06, "num_tokens": 434741.0, "completions/mean_length": 398.625, "completions/min_length": 250.0, "completions/max_length": 742.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 398.625, "completions/min_terminated_length": 250.0, "completions/max_terminated_length": 742.0, "tools/call_frequency": 4.0, "tools/failure_frequency": 0.0625, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.6098519563674927, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 1.183618792310881e-06, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.3333333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/mean": 0.3412742018699646, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/std": 0.1825484037399292, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/mean": 0.0833333358168602, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/std": 0.1666666716337204, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/mean": -0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/std": 1.0, "reward": 0.5359797477722168, "reward_std": 0.4024333357810974, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.04406218009535223, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 309.39598267899873, "entropy_control/entropy": 0.03689289156406395, "entropy_control/target": 0.03031625726079178, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.116, "adapter_sync_time": 0.0, "generation_time": 194.145289, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 215.223729, "forward_backward_time": 94.070809, "rollout_overhead_time": 21.07844 }, { "step": 30, "loss": -0.08245766907930374, "grad_norm": 0.0908203125, "learning_rate": 4.2000000000000004e-06, "entropy": 0.043836229597218335, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 94.16109022100045, "entropy_control/entropy": 0.036719057926245253, "entropy_control/target": 0.03031625726079178, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.12, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 94.057874, "rollout_overhead_time": 0.0, "reward": null }, { "step": 31, "loss": 0.22658193111419678, "grad_norm": 0.1318359375, "learning_rate": 4.000000000000001e-06, "num_tokens": 463118.0, "completions/mean_length": 459.625, "completions/min_length": 273.0, "completions/max_length": 1194.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 459.625, "completions/min_terminated_length": 273.0, "completions/max_terminated_length": 1194.0, "tools/call_frequency": 3.125, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/mean": 0.6656099557876587, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/std": 0.3028259575366974, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/mean": 0.4166666865348816, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/std": 0.29546844959259033, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/std": 0.9258201122283936, "reward": 0.7864433526992798, "reward_std": 0.500939130783081, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.5, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.053054060554131866, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 290.2619876309991, "entropy_control/entropy": 0.042977884104551334, "entropy_control/target": 0.03031625726079178, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.124, "adapter_sync_time": 0.0, "generation_time": 189.00532, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 207.787733, "forward_backward_time": 82.369938, "rollout_overhead_time": 18.782413 }, { "step": 32, "loss": 0.22693459689617157, "grad_norm": 0.1474609375, "learning_rate": 3.8000000000000005e-06, "entropy": 0.05277115071658045, "clip_ratio/low_mean": 0.0007332969398703426, "clip_ratio/low_min": 0.0007332969398703426, "clip_ratio/high_mean": 0.001598130736965686, "clip_ratio/high_max": 0.001598130736965686, "clip_ratio/region_mean": 0.0023314276768360287, "step_time": 82.46991769599936, "entropy_control/entropy": 0.04284019018248579, "entropy_control/target": 0.03031625726079178, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.128, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 82.372028, "rollout_overhead_time": 0.0, "reward": null }, { "step": 33, "loss": 0.26303285360336304, "grad_norm": 0.2890625, "learning_rate": 3.6000000000000003e-06, "num_tokens": 489378.0, "completions/mean_length": 383.0, "completions/min_length": 267.0, "completions/max_length": 849.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 383.0, "completions/min_terminated_length": 267.0, "completions/max_terminated_length": 849.0, "tools/call_frequency": 3.75, "tools/failure_frequency": 0.13333334028720856, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.6145089864730835, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 0.002659555757418275, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.19245010614395142, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/mean": 0.6661649942398071, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/std": 0.3274487555027008, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/mean": 0.375, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/std": 0.28463754057884216, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/std": 1.0, "reward": 0.8122120499610901, "reward_std": 0.35481512546539307, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.099782669916749, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 271.0022347350023, "entropy_control/entropy": 0.08817868029873602, "entropy_control/target": 0.03031625726079178, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.132, "adapter_sync_time": 0.0, "generation_time": 173.740596, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 192.280987, "forward_backward_time": 78.621158, "rollout_overhead_time": 18.540391 }, { "step": 34, "loss": 0.2627889811992645, "grad_norm": 0.26953125, "learning_rate": 3.4000000000000005e-06, "entropy": 0.09929254441522062, "clip_ratio/low_mean": 0.0004646840097848326, "clip_ratio/low_min": 0.0004646840097848326, "clip_ratio/high_mean": 0.00037202381645329297, "clip_ratio/high_max": 0.00037202381645329297, "clip_ratio/region_mean": 0.0008367078262381256, "step_time": 78.72165960399889, "entropy_control/entropy": 0.0879312352183957, "entropy_control/target": 0.03031625726079178, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.136, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 78.616983, "rollout_overhead_time": 0.0, "reward": null }, { "step": 35, "loss": 0.10388702899217606, "grad_norm": 0.1015625, "learning_rate": 3.2000000000000003e-06, "num_tokens": 525987.0, "completions/mean_length": 709.125, "completions/min_length": 281.0, "completions/max_length": 1039.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 709.125, "completions/min_terminated_length": 281.0, "completions/max_terminated_length": 1039.0, "tools/call_frequency": 3.875, "tools/failure_frequency": 0.032258063554763794, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/mean": 0.2958386540412903, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/std": 0.1296512633562088, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/mean": 0.0625, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/std": 0.1767766922712326, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/mean": -0.75, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/std": 0.7071067690849304, "reward": 0.14896363019943237, "reward_std": 0.27991142868995667, "frac_reward_zero_std": 0.5, "frac_groups_zero_std": 0.5, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.0200532489689067, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 312.12456131, "entropy_control/entropy": 0.014173528316416882, "entropy_control/target": 0.03031625726079178, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.14, "adapter_sync_time": 0.0, "generation_time": 209.030083, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 228.521845, "forward_backward_time": 83.503519, "rollout_overhead_time": 19.491762 }, { "step": 36, "loss": 0.10383187979459763, "grad_norm": 0.06103515625, "learning_rate": 3e-06, "entropy": 0.019959633063990623, "clip_ratio/low_mean": 0.0001203079882543534, "clip_ratio/low_min": 0.0001203079882543534, "clip_ratio/high_mean": 0.0002659574383869767, "clip_ratio/high_max": 0.0002659574383869767, "clip_ratio/region_mean": 0.0003862654266413301, "step_time": 83.60138981300133, "entropy_control/entropy": 0.014069342739958808, "entropy_control/target": 0.03031625726079178, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.144, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 83.504691, "rollout_overhead_time": 0.0, "reward": null }, { "step": 37, "loss": 0.003299206495285034, "grad_norm": 0.1767578125, "learning_rate": 2.8000000000000003e-06, "num_tokens": 551569.0, "completions/mean_length": 313.75, "completions/min_length": 200.0, "completions/max_length": 463.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 313.75, "completions/min_terminated_length": 200.0, "completions/max_terminated_length": 463.0, "tools/call_frequency": 3.75, "tools/failure_frequency": 0.4000000059604645, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.6169520020484924, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 0.000885089801158756, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.5833333730697632, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.1666666716337204, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__native/mean": 0.618954598903656, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__native/std": 9.291444484915701e-07, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__efficiency/mean": 0.6666666865348816, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__efficiency/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__reliability/std": 0.0, "reward": 0.84920334815979, "reward_std": 0.0068770782090723515, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.5, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.07350466074422002, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 212.67323126400152, "entropy_control/entropy": 0.0739125223995948, "entropy_control/target": 0.03031625726079178, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.148, "adapter_sync_time": 0.0, "generation_time": 111.956567, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 130.605717, "forward_backward_time": 81.966546, "rollout_overhead_time": 18.64915 }, { "step": 38, "loss": 0.00309772789478302, "grad_norm": 0.16796875, "learning_rate": 2.6e-06, "entropy": 0.07349386159330606, "clip_ratio/low_mean": 0.0007055184687487781, "clip_ratio/low_min": 0.0007055184687487781, "clip_ratio/high_mean": 0.0007942890515550971, "clip_ratio/high_max": 0.0007942890515550971, "clip_ratio/region_mean": 0.0014998075203038752, "step_time": 82.06627910099905, "entropy_control/entropy": 0.07389654169173025, "entropy_control/target": 0.03031625726079178, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.152, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 81.967482, "rollout_overhead_time": 0.0, "reward": null }, { "step": 39, "loss": 0.021948106586933136, "grad_norm": 0.228515625, "learning_rate": 2.4000000000000003e-06, "num_tokens": 578839.0, "completions/mean_length": 313.75, "completions/min_length": 247.0, "completions/max_length": 436.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 313.75, "completions/min_terminated_length": 247.0, "completions/max_terminated_length": 436.0, "tools/call_frequency": 2.875, "tools/failure_frequency": 0.17391304671764374, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__native/mean": 0.618468165397644, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__native/std": 0.000908317684661597, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__efficiency/mean": 0.6041666865348816, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__efficiency/std": 0.08625821024179459, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__reliability/std": 0.0, "reward": 0.8486765623092651, "reward_std": 0.005111454986035824, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.08312045782804489, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 158.74121129800187, "entropy_control/entropy": 0.08195730024252483, "entropy_control/target": 0.03031625726079178, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.156, "adapter_sync_time": 0.0, "generation_time": 82.87084, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 97.525424, "forward_backward_time": 61.113497, "rollout_overhead_time": 14.654584 }, { "step": 40, "loss": 0.02142183482646942, "grad_norm": 0.2041015625, "learning_rate": 2.2e-06, "entropy": 0.08254547649994493, "clip_ratio/low_mean": 0.0007418397581204772, "clip_ratio/low_min": 0.0007418397581204772, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0007418397581204772, "step_time": 61.216063511001266, "entropy_control/entropy": 0.08144057412950294, "entropy_control/target": 0.03031625726079178, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.16, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 61.115791, "rollout_overhead_time": 0.0, "reward": null }, { "step": 41, "loss": 0.02038617990911007, "grad_norm": 0.09619140625, "learning_rate": 2.0000000000000003e-06, "num_tokens": 610333.0, "completions/mean_length": 455.75, "completions/min_length": 264.0, "completions/max_length": 659.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 455.75, "completions/min_terminated_length": 264.0, "completions/max_terminated_length": 659.0, "tools/call_frequency": 2.75, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__native/mean": 0.6167720556259155, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__native/std": 3.006737870236975e-06, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__efficiency/mean": 0.6666666865348816, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__efficiency/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/mean": 0.5251070857048035, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/std": 0.1834070235490799, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/mean": 0.3333333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/std": 0.2357022762298584, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/std": 1.0, "reward": 0.7459396123886108, "reward_std": 0.28127336502075195, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.04792513488791883, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 257.0219260349986, "entropy_control/entropy": 0.043204792956577386, "entropy_control/target": 0.03031625726079178, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.164, "adapter_sync_time": 0.0, "generation_time": 153.105383, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 172.651199, "forward_backward_time": 84.269766, "rollout_overhead_time": 19.545815 }, { "step": 42, "loss": 0.020487722009420395, "grad_norm": 0.103515625, "learning_rate": 1.8000000000000001e-06, "entropy": 0.04826283315196633, "clip_ratio/low_mean": 0.00044802867341786623, "clip_ratio/low_min": 0.00044802867341786623, "clip_ratio/high_mean": 0.0006631661963183433, "clip_ratio/high_max": 0.0006631661963183433, "clip_ratio/region_mean": 0.0011111948697362095, "step_time": 84.34784766300072, "entropy_control/entropy": 0.04347036650723186, "entropy_control/target": 0.03031625726079178, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.168, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 84.253063, "rollout_overhead_time": 0.0, "reward": null }, { "step": 43, "loss": 0.09651297330856323, "grad_norm": 0.154296875, "learning_rate": 1.6000000000000001e-06, "num_tokens": 639941.0, "completions/mean_length": 393.0, "completions/min_length": 258.0, "completions/max_length": 660.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 393.0, "completions/min_terminated_length": 258.0, "completions/max_terminated_length": 660.0, "tools/call_frequency": 3.5, "tools/failure_frequency": 0.25, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__native/mean": 0.6175951361656189, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__native/std": 0.0009088596561923623, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__efficiency/mean": 0.5833333730697632, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__efficiency/std": 0.09622505307197571, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/mean": 0.5253363847732544, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/std": 0.18355761468410492, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/std": 0.3333333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/std": 1.0, "reward": 0.7485491037368774, "reward_std": 0.2822871804237366, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.06372657674364746, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 248.04164702799972, "entropy_control/entropy": 0.05962481169471561, "entropy_control/target": 0.03031625726079178, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.172, "adapter_sync_time": 0.0, "generation_time": 151.149313, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 169.465944, "forward_backward_time": 78.475008, "rollout_overhead_time": 18.316631 }, { "step": 44, "loss": 0.09647795557975769, "grad_norm": 0.1494140625, "learning_rate": 1.4000000000000001e-06, "entropy": 0.06359893176704645, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 78.56714497000394, "entropy_control/entropy": 0.059532171890339464, "entropy_control/target": 0.03031625726079178, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.176, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 78.470305, "rollout_overhead_time": 0.0, "reward": null }, { "step": 45, "loss": 0.14501114189624786, "grad_norm": 0.1962890625, "learning_rate": 1.2000000000000002e-06, "num_tokens": 673778.0, "completions/mean_length": 555.625, "completions/min_length": 309.0, "completions/max_length": 840.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 555.625, "completions/min_terminated_length": 309.0, "completions/max_terminated_length": 840.0, "tools/call_frequency": 3.375, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/mean": 0.4321771264076233, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/std": 0.21036000549793243, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/mean": 0.1666666716337204, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/std": 0.19245010614395142, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/mean": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/std": 1.154700517654419, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/mean": 0.524140477180481, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/std": 0.1827603131532669, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/std": 0.3333333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/std": 1.0, "reward": 0.5448254346847534, "reward_std": 0.40982091426849365, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.05191244976595044, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 291.9126464979945, "entropy_control/entropy": 0.04314513979878438, "entropy_control/target": 0.03031625726079178, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.18, "adapter_sync_time": 0.0, "generation_time": 182.660712, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 202.765055, "forward_backward_time": 89.049637, "rollout_overhead_time": 20.104343 }, { "step": 46, "loss": 0.14456185698509216, "grad_norm": 0.10009765625, "learning_rate": 1.0000000000000002e-06, "entropy": 0.05136946530546993, "clip_ratio/low_mean": 0.000510081197717227, "clip_ratio/low_min": 0.000510081197717227, "clip_ratio/high_mean": 0.0017875390622066334, "clip_ratio/high_max": 0.0017875390622066334, "clip_ratio/region_mean": 0.0022976202599238604, "step_time": 89.14526429899888, "entropy_control/entropy": 0.0426339945196989, "entropy_control/target": 0.03031625726079178, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.184, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 89.046254, "rollout_overhead_time": 0.0, "reward": null }, { "step": 47, "loss": -0.018097413703799248, "grad_norm": 0.08544921875, "learning_rate": 8.000000000000001e-07, "num_tokens": 703353.0, "completions/mean_length": 406.375, "completions/min_length": 308.0, "completions/max_length": 603.0, "completions/clipped_ratio": 0.125, "completions/mean_terminated_length": 379.71429443359375, "completions/min_terminated_length": 308.0, "completions/max_terminated_length": 603.0, "tools/call_frequency": 3.375, "tools/failure_frequency": 0.14814814925193787, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/mean": 0.6172248721122742, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/std": 1.1022831358786789e-06, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/mean": 0.6666666865348816, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/mean": 0.5410884618759155, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/std": 0.3361206352710724, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/mean": 0.2916666865348816, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/std": 0.34359216690063477, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/mean": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/std": 1.154700517654419, "reward": 0.7031150460243225, "reward_std": 0.41348356008529663, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.5, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.049183032708242536, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 243.1568956810006, "entropy_control/entropy": 0.05395608048497949, "entropy_control/target": 0.03031625726079178, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.188, "adapter_sync_time": 0.0, "generation_time": 132.256004, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 152.536721, "forward_backward_time": 90.521028, "rollout_overhead_time": 20.280717 }, { "step": 48, "loss": -0.01824425347149372, "grad_norm": 0.083984375, "learning_rate": 6.000000000000001e-07, "entropy": 0.048936380771920085, "clip_ratio/low_mean": 0.0004145936982240528, "clip_ratio/low_min": 0.0004145936982240528, "clip_ratio/high_mean": 0.00032552084303461015, "clip_ratio/high_max": 0.00032552084303461015, "clip_ratio/region_mean": 0.0007401145412586629, "step_time": 90.61540986599903, "entropy_control/entropy": 0.0536625850921848, "entropy_control/target": 0.03031625726079178, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.192, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 90.519903, "rollout_overhead_time": 0.0, "reward": null }, { "step": 49, "loss": -0.01925632357597351, "grad_norm": 0.1787109375, "learning_rate": 4.0000000000000003e-07, "num_tokens": 732287.0, "completions/mean_length": 321.75, "completions/min_length": 277.0, "completions/max_length": 382.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 321.75, "completions/min_terminated_length": 277.0, "completions/max_terminated_length": 382.0, "tools/call_frequency": 2.375, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/mean": 0.6156848669052124, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/std": 1.5359042890850105e-06, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/mean": 0.6666666865348816, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/mean": 0.8315935730934143, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/std": 0.0012110897805541754, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/mean": 0.5416666865348816, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/std": 0.15957120060920715, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/std": 0.0, "reward": 0.9538475275039673, "reward_std": 0.11222782731056213, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.08955142507329583, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 194.973901905998, "entropy_control/entropy": 0.09202889173549136, "entropy_control/target": 0.03031625726079178, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.196, "adapter_sync_time": 0.0, "generation_time": 103.317614, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 120.437101, "forward_backward_time": 74.436733, "rollout_overhead_time": 17.119486 }, { "step": 50, "loss": -0.01872306317090988, "grad_norm": 0.1650390625, "learning_rate": 2.0000000000000002e-07, "entropy": 0.08925818745046854, "clip_ratio/low_mean": 0.0004448398540262133, "clip_ratio/low_min": 0.0004448398540262133, "clip_ratio/high_mean": 0.00037537538446485996, "clip_ratio/high_max": 0.00037537538446485996, "clip_ratio/region_mean": 0.0008202152384910733, "step_time": 74.52270755099926, "entropy_control/entropy": 0.09178309330982862, "entropy_control/target": 0.03031625726079178, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.2, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 74.429617, "rollout_overhead_time": 0.0, "reward": null } ], "holdout": [ { "seed": 900000, "sample": 0, "native_reward": 0.617827415078968, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 273, "reward_channels": { "native": 0.617827415078968, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "two_pool_cycle" }, { "seed": 900000, "sample": 1, "native_reward": 0.6178268262001205, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 277, "reward_channels": { "native": 0.6178268262001205, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "two_pool_cycle" }, { "seed": 900000, "sample": 2, "native_reward": 0.6178278680985384, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 295, "reward_channels": { "native": 0.6178278680985384, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "two_pool_cycle" }, { "seed": 900000, "sample": 3, "native_reward": 0.6178281013971713, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 307, "reward_channels": { "native": 0.6178281013971713, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "two_pool_cycle" }, { "seed": 900001, "sample": 0, "native_reward": 0.6177086905727631, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 224, "reward_channels": { "native": 0.6177086905727631, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "two_pool_cycle" }, { "seed": 900001, "sample": 1, "native_reward": 0.6177077954531537, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 320, "reward_channels": { "native": 0.6177077954531537, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "two_pool_cycle" }, { "seed": 900001, "sample": 2, "native_reward": 0.6177076480229584, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 294, "reward_channels": { "native": 0.6177076480229584, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "two_pool_cycle" }, { "seed": 900001, "sample": 3, "native_reward": 0.6168370557757319, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 294, "reward_channels": { "native": 0.6168370557757319, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "two_pool_cycle" }, { "seed": 900000, "sample": 0, "native_reward": 0.6196029689418718, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 316, "reward_channels": { "native": 0.6196029689418718, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "triangular_cycle" }, { "seed": 900000, "sample": 1, "native_reward": 0.6196020525787467, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 276, "reward_channels": { "native": 0.6196020525787467, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "triangular_cycle" }, { "seed": 900000, "sample": 2, "native_reward": 0.6196037362934148, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 350, "reward_channels": { "native": 0.6196037362934148, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "triangular_cycle" }, { "seed": 900000, "sample": 3, "native_reward": 0.6196031742367349, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 268, "reward_channels": { "native": 0.6196031742367349, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "triangular_cycle" }, { "seed": 900001, "sample": 0, "native_reward": 0.6194053855289625, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 3, "model_tokens": 333, "reward_channels": { "native": 0.6194053855289625, "efficiency": 0.5, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "triangular_cycle" }, { "seed": 900001, "sample": 1, "native_reward": 0.6200496131721656, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 258, "reward_channels": { "native": 0.6200496131721656, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "triangular_cycle" }, { "seed": 900001, "sample": 2, "native_reward": 0.6200484743316188, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 252, "reward_channels": { "native": 0.6200484743316188, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "triangular_cycle" }, { "seed": 900001, "sample": 3, "native_reward": 0.6200497891838649, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 297, "reward_channels": { "native": 0.6200497891838649, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "triangular_cycle" }, { "seed": 900000, "sample": 0, "native_reward": 0.618495692979087, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 726, "reward_channels": { "native": 0.618495692979087, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "competing_cycles" }, { "seed": 900000, "sample": 1, "native_reward": 0.6207112328656104, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 3, "model_tokens": 446, "reward_channels": { "native": 0.6207112328656104, "efficiency": 0.5, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "competing_cycles" }, { "seed": 900000, "sample": 2, "native_reward": 0.6207140123110011, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 3, "model_tokens": 489, "reward_channels": { "native": 0.6207140123110011, "efficiency": 0.5, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "competing_cycles" }, { "seed": 900000, "sample": 3, "native_reward": 0.62071255280854, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 3, "model_tokens": 550, "reward_channels": { "native": 0.62071255280854, "efficiency": 0.5, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "competing_cycles" }, { "seed": 900001, "sample": 0, "native_reward": 0.6222763451244105, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 3, "model_tokens": 446, "reward_channels": { "native": 0.6222763451244105, "efficiency": 0.5, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "competing_cycles" }, { "seed": 900001, "sample": 1, "native_reward": 0.6222766070858753, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 3, "model_tokens": 587, "reward_channels": { "native": 0.6222766070858753, "efficiency": 0.5, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "competing_cycles" }, { "seed": 900001, "sample": 2, "native_reward": 0.6222768966053871, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 3, "model_tokens": 394, "reward_channels": { "native": 0.6222768966053871, "efficiency": 0.5, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "competing_cycles" }, { "seed": 900001, "sample": 3, "native_reward": 0.6222726313160121, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 3, "model_tokens": 411, "reward_channels": { "native": 0.6222726313160121, "efficiency": 0.5, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "competing_cycles" }, { "seed": 900000, "sample": 0, "native_reward": 0.6182188046812087, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 302, "reward_channels": { "native": 0.6182188046812087, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "four_hop_cycle" }, { "seed": 900000, "sample": 1, "native_reward": 0.0, "promotion_success": false, "done": false, "errors": 0, "tool_steps": 0, "model_tokens": 251, "reward_channels": { "native": 0.0, "efficiency": 0.0, "reliability": -1.0 }, "verification_level": null, "failed_gates": null, "family": "four_hop_cycle" }, { "seed": 900000, "sample": 2, "native_reward": 0.6182191455836307, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 366, "reward_channels": { "native": 0.6182191455836307, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "four_hop_cycle" }, { "seed": 900000, "sample": 3, "native_reward": 0.6182196745883963, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 353, "reward_channels": { "native": 0.6182196745883963, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "four_hop_cycle" }, { "seed": 900001, "sample": 0, "native_reward": 0.6183858537745408, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 286, "reward_channels": { "native": 0.6183858537745408, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "four_hop_cycle" }, { "seed": 900001, "sample": 1, "native_reward": 0.6183862626624119, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 314, "reward_channels": { "native": 0.6183862626624119, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "four_hop_cycle" }, { "seed": 900001, "sample": 2, "native_reward": 0.6183844270508299, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 0, "reward_channels": { "native": 0.6183844270508299, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "four_hop_cycle" }, { "seed": 900001, "sample": 3, "native_reward": 0.6183847253722361, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 331, "reward_channels": { "native": 0.6183847253722361, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "four_hop_cycle" }, { "seed": 900000, "sample": 0, "native_reward": 0.8345685745689324, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 3, "model_tokens": 335, "reward_channels": { "native": 0.8345685745689324, "efficiency": 0.5, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "size_sensitive_cycle" }, { "seed": 900000, "sample": 1, "native_reward": 0.8345688851902214, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 3, "model_tokens": 363, "reward_channels": { "native": 0.8345688851902214, "efficiency": 0.5, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "size_sensitive_cycle" }, { "seed": 900000, "sample": 2, "native_reward": 0.834567592653991, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 3, "model_tokens": 318, "reward_channels": { "native": 0.834567592653991, "efficiency": 0.5, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "size_sensitive_cycle" }, { "seed": 900000, "sample": 3, "native_reward": 0.8352183059394792, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 313, "reward_channels": { "native": 0.8352183059394792, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "size_sensitive_cycle" }, { "seed": 900001, "sample": 0, "native_reward": 0.832439310785638, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 3, "model_tokens": 345, "reward_channels": { "native": 0.832439310785638, "efficiency": 0.5, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "size_sensitive_cycle" }, { "seed": 900001, "sample": 1, "native_reward": 0.8333153584278061, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 345, "reward_channels": { "native": 0.8333153584278061, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "size_sensitive_cycle" }, { "seed": 900001, "sample": 2, "native_reward": 0.8333146420131576, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 302, "reward_channels": { "native": 0.8333146420131576, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "size_sensitive_cycle" }, { "seed": 900001, "sample": 3, "native_reward": 0.8324383605877279, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 3, "model_tokens": 444, "reward_channels": { "native": 0.8324383605877279, "efficiency": 0.5, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "size_sensitive_cycle" } ] }, { "run_id": "1789165288", "mode": "no-think", "method": "dapo-refill", "steps": 50, "reward_points": 25, "train_mean": 0.6564315838118394, "train_last10": 0.7584913248817126, "holdout_native": 0.5839457972260582, "holdout_episodes": 40, "holdout_successes": 34, "mean_step_seconds": 223.9525490545201, "label": "DAPO-refill", "state_at_cutoff": "complete", "step_hours": 3.110452070201668, "peak_allocated_gib": null, "peak_reserved_gib": null, "extra_refill_rounds": 19.0, "groups_refilled": 16.0, "holdout_component_means": { "native": 0.5839457972260582, "efficiency": 0.37083333333333335, "reliability": 0.7 }, "configuration": { "model": { "id": "Qwen/Qwen3-14B", "revision": "40c069824f4251a91eefaf281ebe4c544efd3e18", "loss": "dapo", "algorithm": "grpo", "no_think": true, "lora": true, "quantize": true, "lora_r": 32, "lora_alpha": 64, "use_dora": false }, "loss": "dapo", "normalization": "joint", "entropy_control": { "method": "monitor", "epsilon_low": 0.2, "epsilon_high": 0.28, "zeta_initial": 0.001, "zeta_min": 0.0001, "zeta_max": 0.05, "holdout_split": "benchmark", "holdout_seeds_per_family": 2, "num_iterations": 2 }, "dynamic_sampling": true, "sampling": { "group_size": 4, "temperature": 1.0, "max_completion_tokens": 8192, "top_entropy_quantile": 1.0, "temperature_decay_to": null }, "learning_rate": 1e-05, "turn_max_tokens": 1024 }, "source_sha256": { "metrics.jsonl": "c38d539b34cbe05efd14d603ae07cb9e0c0e80ccd1d228f4042856dd6466aed5", "entropy_holdout_episodes.jsonl": "8f26e23909f81fa52ff519d31682cfc906fd7fcb9f104a97ccb5374f1892c87d", "resolved_training_config.json": "837029a3a7e45c5235cee2f925cc5ce8856cda8aff9f496a9d920a3f62b4f382" }, "training_steps": [ { "step": 1, "loss": 0.35472699999809265, "grad_norm": 0.1748046875, "learning_rate": 1e-05, "num_tokens": 28078.0, "completions/mean_length": 613.25, "completions/min_length": 278.0, "completions/max_length": 1190.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 613.25, "completions/min_terminated_length": 278.0, "completions/max_terminated_length": 1190.0, "tools/call_frequency": 4.375, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.5011641383171082, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 0.21744278073310852, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.25, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.1666666716337204, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/mean": 0.33700618147850037, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/std": 0.3240123689174652, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/mean": 0.1666666716337204, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/std": 0.19245010614395142, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/mean": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/std": 1.154700517654419, "reward": 0.4795018434524536, "reward_std": 0.45456814765930176, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.06061301758745685, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 333.8126870860002, "entropy_control/entropy": 0.038171986696735806, "entropy_control/target": 0.038171986696735806, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.004, "adapter_sync_time": 0.0, "generation_time": 230.269955, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 250.540646, "forward_backward_time": 83.174382, "rollout_overhead_time": 20.270691 }, { "step": 2, "loss": 0.35448992252349854, "grad_norm": 0.1064453125, "learning_rate": 9.800000000000001e-06, "entropy": 0.060262446582783014, "clip_ratio/low_mean": 0.0006929489754838869, "clip_ratio/low_min": 0.0006929489754838869, "clip_ratio/high_mean": 0.000953999362536706, "clip_ratio/high_max": 0.000953999362536706, "clip_ratio/region_mean": 0.001646948338020593, "step_time": 83.20406460700008, "entropy_control/entropy": 0.03790323962936314, "entropy_control/target": 0.038171986696735806, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.008, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 83.104259, "rollout_overhead_time": 0.0, "reward": null }, { "step": 3, "loss": 0.18107134103775024, "grad_norm": 0.265625, "learning_rate": 9.600000000000001e-06, "num_tokens": 77061.0, "completions/mean_length": 348.9375, "completions/min_length": 270.0, "completions/max_length": 477.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 348.9375, "completions/min_terminated_length": 270.0, "completions/max_terminated_length": 477.0, "tools/call_frequency": 4.375, "tools/failure_frequency": 0.08435708656907082, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.5571814775466919, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 0.15460200607776642, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.3229166716337204, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.10661446303129196, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 0.875, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 0.3535533845424652, "reward": 0.748327225446701, "reward_std": 0.23009877651929855, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.5, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 1.0, "n_refills": 1.0, "entropy": 0.07343449490144849, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 422.89352600200027, "entropy_control/entropy": 0.07352014494481669, "entropy_control/target": 0.038171986696735806, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.012, "adapter_sync_time": 0.0, "generation_time": 307.411631, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 342.455792, "forward_backward_time": 80.341637, "rollout_overhead_time": 35.044162 }, { "step": 4, "loss": 0.1805187314748764, "grad_norm": 0.197265625, "learning_rate": 9.4e-06, "entropy": 0.07333209784701467, "clip_ratio/low_mean": 0.0008283458300866187, "clip_ratio/low_min": 0.0008283458300866187, "clip_ratio/high_mean": 0.0016232220805250108, "clip_ratio/high_max": 0.0016232220805250108, "clip_ratio/region_mean": 0.0024515679106116295, "step_time": 80.6114583420001, "entropy_control/entropy": 0.07342028031964923, "entropy_control/target": 0.038171986696735806, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.016, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 80.513781, "rollout_overhead_time": 0.0, "reward": null }, { "step": 5, "loss": 0.22802186012268066, "grad_norm": 0.208984375, "learning_rate": 9.200000000000002e-06, "num_tokens": 101505.0, "completions/mean_length": 346.5, "completions/min_length": 266.0, "completions/max_length": 728.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 346.5, "completions/min_terminated_length": 266.0, "completions/max_terminated_length": 728.0, "tools/call_frequency": 4.25, "tools/failure_frequency": 0.05882352963089943, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.5583271384239197, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 0.1548900455236435, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.3125, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.13908715546131134, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 0.75, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 0.7071067690849304, "reward": 0.7239521145820618, "reward_std": 0.3026461601257324, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.07284741546027362, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 258.5455324069999, "entropy_control/entropy": 0.06509570859049835, "entropy_control/target": 0.038171986696735806, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.02, "adapter_sync_time": 0.0, "generation_time": 167.327621, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 184.709988, "forward_backward_time": 73.736424, "rollout_overhead_time": 17.382367 }, { "step": 6, "loss": 0.22797930240631104, "grad_norm": 0.185546875, "learning_rate": 9e-06, "entropy": 0.07223051064647734, "clip_ratio/low_mean": 0.0008824660908430815, "clip_ratio/low_min": 0.0008824660908430815, "clip_ratio/high_mean": 0.0025362651504110545, "clip_ratio/high_max": 0.0025362651504110545, "clip_ratio/region_mean": 0.003418731241254136, "step_time": 73.84239936600011, "entropy_control/entropy": 0.06451345940291593, "entropy_control/target": 0.038171986696735806, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.024, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 73.745749, "rollout_overhead_time": 0.0, "reward": null }, { "step": 7, "loss": 0.18092216551303864, "grad_norm": 0.33984375, "learning_rate": 8.8e-06, "num_tokens": 164867.0, "completions/mean_length": 486.625, "completions/min_length": 240.0, "completions/max_length": 1010.5, "completions/clipped_ratio": 0.0625, "completions/mean_terminated_length": 453.96429443359375, "completions/min_terminated_length": 240.0, "completions/max_terminated_length": 1010.5, "tools/call_frequency": 3.6875, "tools/failure_frequency": 0.01666666753590107, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/mean": 0.25, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/mean": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/mean": -1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/mean": 0.384564109146595, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/std": 0.18495657108724117, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/mean": 0.0833333358168602, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/std": 0.09622505307197571, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/mean": -0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/std": 0.5773502588272095, "reward": 0.16936538740992546, "reward_std": 0.2429595785215497, "frac_reward_zero_std": 0.5, "frac_groups_zero_std": 0.5, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 1.0, "n_refills": 1.0, "entropy": 0.021431849687360227, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.00027593818958848715, "clip_ratio/high_max": 0.00027593818958848715, "clip_ratio/region_mean": 0.00027593818958848715, "step_time": 574.8180244990001, "entropy_control/entropy": 0.020956970628637567, "entropy_control/target": 0.038171986696735806, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.028, "adapter_sync_time": 0.0, "generation_time": 436.32049, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 478.264278, "forward_backward_time": 96.434945, "rollout_overhead_time": 41.943788 }, { "step": 8, "loss": 0.18057852983474731, "grad_norm": 0.22265625, "learning_rate": 8.6e-06, "entropy": 0.02112600638065487, "clip_ratio/low_mean": 0.0007281596481334418, "clip_ratio/low_min": 0.0007281596481334418, "clip_ratio/high_mean": 0.0015063803584780544, "clip_ratio/high_max": 0.0015063803584780544, "clip_ratio/region_mean": 0.002234540006611496, "step_time": 96.54974007300007, "entropy_control/entropy": 0.020737553577686675, "entropy_control/target": 0.038171986696735806, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.032, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 96.431935, "rollout_overhead_time": 0.0, "reward": null }, { "step": 9, "loss": -0.00012810673797503114, "grad_norm": 0.00579833984375, "learning_rate": 8.400000000000001e-06, "num_tokens": 251261.0, "completions/mean_length": 314.4166666666667, "completions/min_length": 255.66666666666666, "completions/max_length": 510.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 314.4166666666667, "completions/min_terminated_length": 255.66666666666666, "completions/max_terminated_length": 510.0, "tools/call_frequency": 3.5833333333333335, "tools/failure_frequency": 0.01149425283074379, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.6132153272628784, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 1.979118527136355e-06, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.3333333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/mean": 0.25, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/mean": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/mean": -1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/std": 0.0, "reward": 0.4399409890174866, "reward_std": 0.4168644646803538, "frac_reward_zero_std": 0.6666666666666666, "frac_groups_zero_std": 1.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 2.0, "n_refills": 2.0, "entropy": 0.08356596855446696, "clip_ratio/low_mean": 0.0003881987649947405, "clip_ratio/low_min": 0.0003881987649947405, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0003881987649947405, "step_time": 565.6080998629991, "entropy_control/entropy": 0.08513529738600786, "entropy_control/target": 0.038171986696735806, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.036, "adapter_sync_time": 0.0, "generation_time": 410.810525, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 473.260004, "forward_backward_time": 92.248829, "rollout_overhead_time": 62.449479 }, { "step": 10, "loss": -0.000141437747515738, "grad_norm": 0.01019287109375, "learning_rate": 8.2e-06, "entropy": 0.08336706086993217, "clip_ratio/low_mean": 0.001187002082588151, "clip_ratio/low_min": 0.001187002082588151, "clip_ratio/high_mean": 0.0003846153849735856, "clip_ratio/high_max": 0.0003846153849735856, "clip_ratio/region_mean": 0.0015716174675617367, "step_time": 92.34179596000013, "entropy_control/entropy": 0.08494570650452599, "entropy_control/target": 0.038171986696735806, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.04, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 92.239984, "rollout_overhead_time": 0.0, "reward": null }, { "step": 11, "loss": -0.010657824575901031, "grad_norm": 0.326171875, "learning_rate": 8.000000000000001e-06, "num_tokens": 280258.0, "completions/mean_length": 311.625, "completions/min_length": 267.0, "completions/max_length": 361.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 311.625, "completions/min_terminated_length": 267.0, "completions/max_terminated_length": 361.0, "tools/call_frequency": 3.0, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__native/mean": 0.6160637140274048, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__native/std": 0.0016426424263045192, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__efficiency/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__efficiency/std": 0.19245010614395142, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/mean": 0.6137489080429077, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/std": 0.0011741066118702292, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/std": 0.13608276844024658, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/std": 0.0, "reward": 0.8399062752723694, "reward_std": 0.00911726988852024, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.057969769928604364, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 218.35304513600022, "entropy_control/entropy": 0.05833197585289479, "entropy_control/target": 0.038171986696735806, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.044, "adapter_sync_time": 0.0, "generation_time": 120.34968, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 139.021158, "forward_backward_time": 79.23106, "rollout_overhead_time": 18.671477 }, { "step": 12, "loss": -0.01145172119140625, "grad_norm": 0.287109375, "learning_rate": 7.800000000000002e-06, "entropy": 0.05749161960557103, "clip_ratio/low_mean": 0.001531512476503849, "clip_ratio/low_min": 0.001531512476503849, "clip_ratio/high_mean": 0.0008183718600776047, "clip_ratio/high_max": 0.0008183718600776047, "clip_ratio/region_mean": 0.0023498843365814537, "step_time": 79.30637174300045, "entropy_control/entropy": 0.057858433400582165, "entropy_control/target": 0.038171986696735806, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.048, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 79.207744, "rollout_overhead_time": 0.0, "reward": null }, { "step": 13, "loss": 0.010552559047937393, "grad_norm": 0.154296875, "learning_rate": 7.600000000000001e-06, "num_tokens": 365944.0, "completions/mean_length": 279.5833333333333, "completions/min_length": 242.0, "completions/max_length": 338.3333333333333, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 279.5833333333333, "completions/min_terminated_length": 242.0, "completions/max_terminated_length": 338.3333333333333, "tools/call_frequency": 3.6666666666666665, "tools/failure_frequency": 0.06388889004786809, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.6148615876833597, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 0.00035375153769715933, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.3611111243565877, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.055555557211240135, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/mean": 0.25, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/mean": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/mean": -1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/std": 0.0, "reward": 0.4414585630098979, "reward_std": 0.4185018142064412, "frac_reward_zero_std": 0.5, "frac_groups_zero_std": 0.5, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 1.0, "n_refills": 2.0, "entropy": 0.08887322247028351, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0009293680195696652, "clip_ratio/high_max": 0.0009293680195696652, "clip_ratio/region_mean": 0.0009293680195696652, "step_time": 501.12077764199876, "entropy_control/entropy": 0.08723790235639847, "entropy_control/target": 0.038171986696735806, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.052, "adapter_sync_time": 0.0, "generation_time": 344.869233, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 407.671669, "forward_backward_time": 93.346994, "rollout_overhead_time": 62.802436 }, { "step": 14, "loss": 0.010183563455939293, "grad_norm": 0.150390625, "learning_rate": 7.4e-06, "entropy": 0.08818746590986848, "clip_ratio/low_mean": 0.0008457895892206579, "clip_ratio/low_min": 0.0008457895892206579, "clip_ratio/high_mean": 0.0016981883090920746, "clip_ratio/high_max": 0.0016981883090920746, "clip_ratio/region_mean": 0.0025439778983127326, "step_time": 93.43431715799761, "entropy_control/entropy": 0.08655305600153379, "entropy_control/target": 0.038171986696735806, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.056, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 93.332763, "rollout_overhead_time": 0.0, "reward": null }, { "step": 15, "loss": 0.20478859543800354, "grad_norm": 0.310546875, "learning_rate": 7.2000000000000005e-06, "num_tokens": 426178.0, "completions/mean_length": 280.625, "completions/min_length": 240.5, "completions/max_length": 435.5, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 280.625, "completions/min_terminated_length": 240.5, "completions/max_terminated_length": 435.5, "tools/call_frequency": 3.5, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/mean": 0.25, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/mean": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/mean": -1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/mean": 0.684032142162323, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/std": 0.28935477137565613, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/mean": 0.25, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/std": 0.1666666716337204, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/std": 1.0, "reward": 0.42326606810092926, "reward_std": 0.5151561200618744, "frac_reward_zero_std": 0.5, "frac_groups_zero_std": 0.5, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 1.0, "n_refills": 1.0, "entropy": 0.023842402908485383, "clip_ratio/low_mean": 0.00029137529782019556, "clip_ratio/low_min": 0.00029137529782019556, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.00029137529782019556, "step_time": 318.04057367399855, "entropy_control/entropy": 0.024695198106487473, "entropy_control/target": 0.038171986696735806, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.06, "adapter_sync_time": 0.0, "generation_time": 200.226438, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 237.342114, "forward_backward_time": 80.595195, "rollout_overhead_time": 37.115676 }, { "step": 16, "loss": 0.20481272041797638, "grad_norm": 0.193359375, "learning_rate": 7e-06, "entropy": 0.023706270498223603, "clip_ratio/low_mean": 0.0011569313064683229, "clip_ratio/low_min": 0.0011569313064683229, "clip_ratio/high_mean": 0.0028163990064058453, "clip_ratio/high_max": 0.0028163990064058453, "clip_ratio/region_mean": 0.003973330312874168, "step_time": 80.69576919999872, "entropy_control/entropy": 0.024632493184943767, "entropy_control/target": 0.038171986696735806, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.064, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 80.598971, "rollout_overhead_time": 0.0, "reward": null }, { "step": 17, "loss": -0.3950302004814148, "grad_norm": 0.2216796875, "learning_rate": 6.800000000000001e-06, "num_tokens": 459122.0, "completions/mean_length": 423.0, "completions/min_length": 279.0, "completions/max_length": 1048.0, "completions/clipped_ratio": 0.125, "completions/mean_terminated_length": 333.71429443359375, "completions/min_terminated_length": 279.0, "completions/max_terminated_length": 601.0, "tools/call_frequency": 3.25, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/mean": 0.34096914529800415, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/std": 0.1819382905960083, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/mean": 0.0833333358168602, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/std": 0.1666666716337204, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/mean": -0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/std": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/mean": 0.45668965578079224, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/std": 0.3044597804546356, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/mean": 0.25, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/std": 0.1666666716337204, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/std": 1.0, "reward": 0.40716272592544556, "reward_std": 0.45604321360588074, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 0.5, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.020758527258294635, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 360.7394638119995, "entropy_control/entropy": 0.023500461558263214, "entropy_control/target": 0.038171986696735806, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.068, "adapter_sync_time": 0.0, "generation_time": 250.957018, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 271.182472, "forward_backward_time": 89.454093, "rollout_overhead_time": 20.225454 }, { "step": 18, "loss": -0.3949372470378876, "grad_norm": 0.314453125, "learning_rate": 6.600000000000001e-06, "entropy": 0.02104511100333184, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0004385965003166348, "clip_ratio/high_max": 0.0004385965003166348, "clip_ratio/region_mean": 0.0004385965003166348, "step_time": 89.55805717299972, "entropy_control/entropy": 0.023838089914430677, "entropy_control/target": 0.038171986696735806, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.072, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 89.455236, "rollout_overhead_time": 0.0, "reward": null }, { "step": 19, "loss": 0.1308889389038086, "grad_norm": 0.1220703125, "learning_rate": 6.4000000000000006e-06, "num_tokens": 484831.0, "completions/mean_length": 319.125, "completions/min_length": 240.0, "completions/max_length": 450.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 319.125, "completions/min_terminated_length": 240.0, "completions/max_terminated_length": 450.0, "tools/call_frequency": 4.25, "tools/failure_frequency": 0.05882352963089943, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.6140448451042175, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 1.3739270343648968e-06, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.3333333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/mean": 0.5401251316070557, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/std": 0.3350076377391815, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/mean": 0.1666666716337204, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/std": 0.19245010614395142, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/mean": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/std": 1.154700517654419, "reward": 0.6895849704742432, "reward_std": 0.40588119626045227, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.04395945614669472, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 231.9451002750011, "entropy_control/entropy": 0.04250542961572484, "entropy_control/target": 0.038171986696735806, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.076, "adapter_sync_time": 0.0, "generation_time": 129.726252, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 148.674262, "forward_backward_time": 83.167571, "rollout_overhead_time": 18.94801 }, { "step": 20, "loss": 0.13052202761173248, "grad_norm": 0.0634765625, "learning_rate": 6.200000000000001e-06, "entropy": 0.043755472055636346, "clip_ratio/low_mean": 0.001440864900359884, "clip_ratio/low_min": 0.001440864900359884, "clip_ratio/high_mean": 0.0009417087712790817, "clip_ratio/high_max": 0.0009417087712790817, "clip_ratio/region_mean": 0.0023825736716389656, "step_time": 83.27529721300107, "entropy_control/entropy": 0.04237263877625146, "entropy_control/target": 0.038171986696735806, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.08, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 83.171485, "rollout_overhead_time": 0.0, "reward": null }, { "step": 21, "loss": -0.00013608962763100863, "grad_norm": 0.004791259765625, "learning_rate": 6e-06, "num_tokens": 549685.0, "completions/mean_length": 356.375, "completions/min_length": 252.5, "completions/max_length": 846.5, "completions/clipped_ratio": 0.0625, "completions/mean_terminated_length": 315.61607360839844, "completions/min_terminated_length": 252.5, "completions/max_terminated_length": 538.0, "tools/call_frequency": 3.625, "tools/failure_frequency": 0.017241379246115685, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/mean": 0.25, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/mean": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/mean": -1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/mean": 0.6098255217075348, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/std": 3.031099936379178e-06, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/mean": 0.3333333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/std": 0.0, "reward": 0.43824608623981476, "reward_std": 0.41505251824855804, "frac_reward_zero_std": 0.5, "frac_groups_zero_std": 0.5, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 1.0, "n_refills": 1.0, "entropy": 0.02123649511486292, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.00039308174746111035, "clip_ratio/high_max": 0.00039308174746111035, "clip_ratio/region_mean": 0.00039308174746111035, "step_time": 495.37745529899803, "entropy_control/entropy": 0.021338803376996295, "entropy_control/target": 0.038171986696735806, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.084, "adapter_sync_time": 0.0, "generation_time": 356.510989, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 399.145373, "forward_backward_time": 96.112842, "rollout_overhead_time": 42.634384 }, { "step": 22, "loss": -0.00013892760034650564, "grad_norm": 0.004180908203125, "learning_rate": 5.8e-06, "entropy": 0.022265722160227597, "clip_ratio/low_mean": 0.0008104770968202502, "clip_ratio/low_min": 0.0008104770968202502, "clip_ratio/high_mean": 0.0012040542787872255, "clip_ratio/high_max": 0.0012040542787872255, "clip_ratio/region_mean": 0.0020145313756074756, "step_time": 96.24851294900054, "entropy_control/entropy": 0.022364956513355468, "entropy_control/target": 0.038171986696735806, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.088, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 96.127461, "rollout_overhead_time": 0.0, "reward": null }, { "step": 23, "loss": 0.0001172204501926899, "grad_norm": 0.002716064453125, "learning_rate": 5.600000000000001e-06, "num_tokens": 574957.0, "completions/mean_length": 269.5, "completions/min_length": 238.0, "completions/max_length": 302.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 269.5, "completions/min_terminated_length": 238.0, "completions/max_terminated_length": 302.0, "tools/call_frequency": 4.25, "tools/failure_frequency": 0.05882352963089943, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.6153919696807861, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 1.2611230886250269e-06, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.3333333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/mean": 0.8312658071517944, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/std": 1.2029707932015299e-06, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/mean": 0.3333333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/std": 0.0, "reward": 0.9399955868721008, "reward_std": 0.11538945138454437, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.0361948530189693, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 204.48686326700226, "entropy_control/entropy": 0.03717981087405208, "entropy_control/target": 0.038171986696735806, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.092, "adapter_sync_time": 0.0, "generation_time": 102.312062, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 121.247969, "forward_backward_time": 83.132366, "rollout_overhead_time": 18.935906 }, { "step": 24, "loss": 0.0001247867476195097, "grad_norm": 0.0019073486328125, "learning_rate": 5.400000000000001e-06, "entropy": 0.036643278086557984, "clip_ratio/low_mean": 0.0005252101109363139, "clip_ratio/low_min": 0.0005252101109363139, "clip_ratio/high_mean": 0.0022539790661539882, "clip_ratio/high_max": 0.0022539790661539882, "clip_ratio/region_mean": 0.002779189177090302, "step_time": 83.22265478900044, "entropy_control/entropy": 0.03764647798762703, "entropy_control/target": 0.038171986696735806, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.096, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 83.120204, "rollout_overhead_time": 0.0, "reward": null }, { "step": 25, "loss": -0.0010135704651474953, "grad_norm": 0.0031280517578125, "learning_rate": 5.2e-06, "num_tokens": 656996.0, "completions/mean_length": 318.625, "completions/min_length": 283.0, "completions/max_length": 395.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 318.625, "completions/min_terminated_length": 283.0, "completions/max_terminated_length": 395.0, "tools/call_frequency": 4.083333333333333, "tools/failure_frequency": 0.01960784321029981, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.6138690710067749, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 1.2543287274032384e-06, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.3333333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/mean": 0.6138824025789896, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/std": 8.413857509973847e-07, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/mean": 0.3333333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/std": 0.0, "reward": 0.8305423855781555, "reward_std": 0.000395485646246622, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.5, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 1.0, "n_refills": 2.0, "entropy": 0.08206695504486561, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 519.9165100769987, "entropy_control/entropy": 0.08336239241412931, "entropy_control/target": 0.038171986696735806, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.1, "adapter_sync_time": 0.0, "generation_time": 370.307891, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 430.216418, "forward_backward_time": 89.601474, "rollout_overhead_time": 59.908527 }, { "step": 26, "loss": -0.0010216929949820042, "grad_norm": 0.00225830078125, "learning_rate": 5e-06, "entropy": 0.08219606801867485, "clip_ratio/low_mean": 0.0008117653196677566, "clip_ratio/low_min": 0.0008117653196677566, "clip_ratio/high_mean": 0.0006885900220368057, "clip_ratio/high_max": 0.0006885900220368057, "clip_ratio/region_mean": 0.0015003553417045623, "step_time": 89.6818983230014, "entropy_control/entropy": 0.08349240696196723, "entropy_control/target": 0.038171986696735806, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.104, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 89.580678, "rollout_overhead_time": 0.0, "reward": null }, { "step": 27, "loss": 0.06667087972164154, "grad_norm": 0.2333984375, "learning_rate": 4.800000000000001e-06, "num_tokens": 685976.0, "completions/mean_length": 316.0, "completions/min_length": 239.0, "completions/max_length": 443.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 316.0, "completions/min_terminated_length": 239.0, "completions/max_terminated_length": 443.0, "tools/call_frequency": 3.875, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/mean": 0.6123619079589844, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/std": 0.0016762717859819531, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/mean": 0.375, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/std": 0.0833333283662796, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/mean": 0.6825577616691589, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/std": 0.28837186098098755, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/mean": 0.25, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/std": 0.1666666716337204, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/std": 1.0, "reward": 0.8130848407745361, "reward_std": 0.32576289772987366, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.027348426810931414, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 248.77518401499856, "entropy_control/entropy": 0.029340359165355667, "entropy_control/target": 0.038171986696735806, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.108, "adapter_sync_time": 0.0, "generation_time": 150.421425, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 168.96586, "forward_backward_time": 79.706083, "rollout_overhead_time": 18.544434 }, { "step": 28, "loss": 0.06668534874916077, "grad_norm": 0.228515625, "learning_rate": 4.600000000000001e-06, "entropy": 0.02710607065819204, "clip_ratio/low_mean": 0.0003918495203834027, "clip_ratio/low_min": 0.0003918495203834027, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0003918495203834027, "step_time": 79.8105911370003, "entropy_control/entropy": 0.029073715001409702, "entropy_control/target": 0.038171986696735806, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.112, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 79.709371, "rollout_overhead_time": 0.0, "reward": null }, { "step": 29, "loss": -0.008208900690078735, "grad_norm": 0.82421875, "learning_rate": 4.4e-06, "num_tokens": 772488.0, "completions/mean_length": 305.5, "completions/min_length": 238.33333333333334, "completions/max_length": 490.6666666666667, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 305.5, "completions/min_terminated_length": 238.33333333333334, "completions/max_terminated_length": 490.6666666666667, "tools/call_frequency": 3.4583333333333335, "tools/failure_frequency": 0.025641026596228283, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.6136381228764852, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 0.00048718280838026357, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.3888888955116272, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.06415003538131714, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/mean": 0.25, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/mean": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/mean": -1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/std": 0.0, "reward": 0.44154128432273865, "reward_std": 0.41859601934750873, "frac_reward_zero_std": 0.5, "frac_groups_zero_std": 1.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 1.0, "n_groups_refilled": 1.0, "n_refills": 2.0, "entropy": 0.053886870853602886, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 555.8580370990003, "entropy_control/entropy": 0.10694760189390767, "entropy_control/target": 0.038171986696735806, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.116, "adapter_sync_time": 0.0, "generation_time": 398.705007, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 461.088238, "forward_backward_time": 94.668508, "rollout_overhead_time": 62.383231 }, { "step": 30, "loss": -0.00959441065788269, "grad_norm": 0.48828125, "learning_rate": 4.2000000000000004e-06, "entropy": 0.05362746957689524, "clip_ratio/low_mean": 0.00041666667675599456, "clip_ratio/low_min": 0.00041666667675599456, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.00041666667675599456, "step_time": 94.76554830300029, "entropy_control/entropy": 0.10645655898498135, "entropy_control/target": 0.038171986696735806, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.12, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 94.666383, "rollout_overhead_time": 0.0, "reward": null }, { "step": 31, "loss": -0.001027640886604786, "grad_norm": 0.002716064453125, "learning_rate": 4.000000000000001e-06, "num_tokens": 826457.0, "completions/mean_length": 273.3125, "completions/min_length": 249.0, "completions/max_length": 312.5, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 273.3125, "completions/min_terminated_length": 249.0, "completions/max_terminated_length": 312.5, "tools/call_frequency": 4.0, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/mean": 0.8302797377109528, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/std": 0.0009997881716117263, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/mean": 0.3333333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/std": 0.0, "reward": 1.046946406364441, "reward_std": 0.0009997921297326684, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.5, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 1.0, "n_refills": 1.0, "entropy": 0.026015853974968195, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0013187458680476993, "clip_ratio/high_max": 0.0013187458680476993, "clip_ratio/region_mean": 0.0013187458680476993, "step_time": 263.7495171710025, "entropy_control/entropy": 0.025940921572166056, "entropy_control/target": 0.038171986696735806, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.124, "adapter_sync_time": 0.0, "generation_time": 159.29829, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 191.688536, "forward_backward_time": 71.956727, "rollout_overhead_time": 32.390245 }, { "step": 32, "loss": -0.0010281403083354235, "grad_norm": 0.00183868408203125, "learning_rate": 3.8000000000000005e-06, "entropy": 0.02501611760817468, "clip_ratio/low_mean": 0.00045454545761458576, "clip_ratio/low_min": 0.00045454545761458576, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.00045454545761458576, "step_time": 72.05432988400025, "entropy_control/entropy": 0.024931831726151377, "entropy_control/target": 0.038171986696735806, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.128, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 71.952371, "rollout_overhead_time": 0.0, "reward": null }, { "step": 33, "loss": 0.15645533800125122, "grad_norm": 0.16015625, "learning_rate": 3.6000000000000003e-06, "num_tokens": 903268.0, "completions/mean_length": 299.2916666666667, "completions/min_length": 244.0, "completions/max_length": 432.3333333333333, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 299.2916666666667, "completions/min_terminated_length": 244.0, "completions/max_terminated_length": 432.3333333333333, "tools/call_frequency": 4.0, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.6140478849411011, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 1.0916741833473982e-06, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.3333333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/mean": 0.7814156214396158, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/std": 0.09639175283120949, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/mean": 0.3055555621782939, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/std": 0.055555557211240135, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/mean": 0.8333333333333334, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/std": 0.3333333333333333, "reward": 0.8970373074213663, "reward_std": 0.18561929712692896, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 1.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 2.0, "n_refills": 2.0, "entropy": 0.049092638306319714, "clip_ratio/low_mean": 0.0022937626927159727, "clip_ratio/low_min": 0.0022937626927159727, "clip_ratio/high_mean": 0.0005387931014411151, "clip_ratio/high_max": 0.0005387931014411151, "clip_ratio/region_mean": 0.002832555794157088, "step_time": 503.61257829799797, "entropy_control/entropy": 0.04636318241141942, "entropy_control/target": 0.038171986696735806, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.132, "adapter_sync_time": 0.0, "generation_time": 362.20372, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 419.299724, "forward_backward_time": 84.208469, "rollout_overhead_time": 57.096004 }, { "step": 34, "loss": 0.156344473361969, "grad_norm": 0.126953125, "learning_rate": 3.4000000000000005e-06, "entropy": 0.04920581495389342, "clip_ratio/low_mean": 0.001404082984663546, "clip_ratio/low_min": 0.001404082984663546, "clip_ratio/high_mean": 0.0009505703346803784, "clip_ratio/high_max": 0.0009505703346803784, "clip_ratio/region_mean": 0.0023546533193439245, "step_time": 84.46763328200177, "entropy_control/entropy": 0.046510019564284355, "entropy_control/target": 0.038171986696735806, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.136, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 84.364044, "rollout_overhead_time": 0.0, "reward": null }, { "step": 35, "loss": -0.09821812808513641, "grad_norm": 0.283203125, "learning_rate": 3.2000000000000003e-06, "num_tokens": 1004209.0, "completions/mean_length": 345.2083333333333, "completions/min_length": 239.33333333333334, "completions/max_length": 727.6666666666666, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 345.2083333333333, "completions/min_terminated_length": 239.33333333333334, "completions/max_terminated_length": 727.6666666666666, "tools/call_frequency": 3.1666666666666665, "tools/failure_frequency": 0.012820513298114141, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/mean": 0.31105149785677594, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/std": 0.10636565089225769, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/mean": 0.07638889054457347, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/std": 0.13749298950036368, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/mean": -0.6666666666666666, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/std": 0.5807350675264994, "reward": 0.18153761327266693, "reward_std": 0.22929534316062927, "frac_reward_zero_std": 0.5, "frac_groups_zero_std": 1.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 2.0, "n_refills": 2.0, "entropy": 0.04938523378223181, "clip_ratio/low_mean": 0.0002025931898970157, "clip_ratio/low_min": 0.0002025931898970157, "clip_ratio/high_mean": 0.00036443150020204484, "clip_ratio/high_max": 0.00036443150020204484, "clip_ratio/region_mean": 0.0005670246900990605, "step_time": 653.0792657429993, "entropy_control/entropy": 0.04729186669328827, "entropy_control/target": 0.038171986696735806, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.14, "adapter_sync_time": 0.0, "generation_time": 501.373948, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 558.763755, "forward_backward_time": 94.211908, "rollout_overhead_time": 57.389807 }, { "step": 36, "loss": -0.09740924835205078, "grad_norm": 0.326171875, "learning_rate": 3e-06, "entropy": 0.04937561647966504, "clip_ratio/low_mean": 0.0010391335235908628, "clip_ratio/low_min": 0.0010391335235908628, "clip_ratio/high_mean": 0.00036443150020204484, "clip_ratio/high_max": 0.00036443150020204484, "clip_ratio/region_mean": 0.0014035650237929076, "step_time": 94.32943631900162, "entropy_control/entropy": 0.04731695139599768, "entropy_control/target": 0.038171986696735806, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.144, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 94.226868, "rollout_overhead_time": 0.0, "reward": null }, { "step": 37, "loss": 0.05287720263004303, "grad_norm": 0.310546875, "learning_rate": 2.8000000000000003e-06, "num_tokens": 1029472.0, "completions/mean_length": 273.875, "completions/min_length": 218.0, "completions/max_length": 321.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 273.875, "completions/min_terminated_length": 218.0, "completions/max_terminated_length": 321.0, "tools/call_frequency": 2.875, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.6158080101013184, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 0.0009923677425831556, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.4166666865348816, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.1666666716337204, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__native/mean": 0.6185405850410461, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__native/std": 0.0006287001306191087, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__efficiency/mean": 0.625, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__efficiency/std": 0.0833333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__reliability/std": 0.0, "reward": 0.8432159423828125, "reward_std": 0.00982489064335823, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.07710135239176452, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 201.25111297200237, "entropy_control/entropy": 0.07798839822258334, "entropy_control/target": 0.038171986696735806, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.148, "adapter_sync_time": 0.0, "generation_time": 100.288398, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 119.024786, "forward_backward_time": 82.122121, "rollout_overhead_time": 18.736387 }, { "step": 38, "loss": 0.05305519700050354, "grad_norm": 0.271484375, "learning_rate": 2.6e-06, "entropy": 0.07741276384331286, "clip_ratio/low_mean": 0.0009779252286534756, "clip_ratio/low_min": 0.0009779252286534756, "clip_ratio/high_mean": 0.00048449612222611904, "clip_ratio/high_max": 0.00048449612222611904, "clip_ratio/region_mean": 0.0014624213508795947, "step_time": 82.232110277002, "entropy_control/entropy": 0.07839585139765556, "entropy_control/target": 0.038171986696735806, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.152, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 82.124298, "rollout_overhead_time": 0.0, "reward": null }, { "step": 39, "loss": 0.009634837508201599, "grad_norm": 0.32421875, "learning_rate": 2.4000000000000003e-06, "num_tokens": 1056904.0, "completions/mean_length": 334.0, "completions/min_length": 236.0, "completions/max_length": 573.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 334.0, "completions/min_terminated_length": 236.0, "completions/max_terminated_length": 573.0, "tools/call_frequency": 3.375, "tools/failure_frequency": 0.29629629850387573, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__native/mean": 0.6180243492126465, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__native/std": 0.0007692081853747368, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__efficiency/mean": 0.6041666865348816, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__efficiency/std": 0.08625821024179459, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__reliability/std": 0.0, "reward": 0.8482327461242676, "reward_std": 0.004943262320011854, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.09590593725442886, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 189.21507756300343, "entropy_control/entropy": 0.0965752983336617, "entropy_control/target": 0.038171986696735806, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.156, "adapter_sync_time": 0.0, "generation_time": 113.086696, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 127.815637, "forward_backward_time": 61.291094, "rollout_overhead_time": 14.728941 }, { "step": 40, "loss": 0.010024309158325195, "grad_norm": 0.287109375, "learning_rate": 2.2e-06, "entropy": 0.09636079519987106, "clip_ratio/low_mean": 0.0026782010099850595, "clip_ratio/low_min": 0.0026782010099850595, "clip_ratio/high_mean": 0.001626634126296267, "clip_ratio/high_max": 0.001626634126296267, "clip_ratio/region_mean": 0.004304835136281326, "step_time": 61.378676403999634, "entropy_control/entropy": 0.09697112428913258, "entropy_control/target": 0.038171986696735806, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.16, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 61.281396, "rollout_overhead_time": 0.0, "reward": null }, { "step": 41, "loss": -0.10985058546066284, "grad_norm": 0.2578125, "learning_rate": 2.0000000000000003e-06, "num_tokens": 1117511.0, "completions/mean_length": 308.4375, "completions/min_length": 221.5, "completions/max_length": 521.5, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 308.4375, "completions/min_terminated_length": 221.5, "completions/max_terminated_length": 521.5, "tools/call_frequency": 3.3125, "tools/failure_frequency": 0.07407407462596893, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__native/mean": 0.6147098541259766, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__native/std": 0.0012609988334588706, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__efficiency/mean": 0.5000000149011612, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__efficiency/std": 0.12145226448774338, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/mean": 0.2958032041788101, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/std": 0.09160640835762024, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/mean": 0.0416666679084301, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/std": 0.0833333358168602, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/mean": -0.75, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/std": 0.5, "reward": 0.4937981963157654, "reward_std": 0.4152926653623581, "frac_reward_zero_std": 0.25, "frac_groups_zero_std": 0.5, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 1.0, "n_refills": 1.0, "entropy": 0.0840687305899337, "clip_ratio/low_mean": 0.0004340277810115367, "clip_ratio/low_min": 0.0004340277810115367, "clip_ratio/high_mean": 0.0013882395578548312, "clip_ratio/high_max": 0.0013882395578548312, "clip_ratio/region_mean": 0.001822267338866368, "step_time": 400.15355072100465, "entropy_control/entropy": 0.09080790058207101, "entropy_control/target": 0.038171986696735806, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.164, "adapter_sync_time": 0.0, "generation_time": 280.551752, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 316.728637, "forward_backward_time": 83.323281, "rollout_overhead_time": 36.176885 }, { "step": 42, "loss": -0.10962048172950745, "grad_norm": 0.298828125, "learning_rate": 1.8000000000000001e-06, "entropy": 0.08486917038680986, "clip_ratio/low_mean": 0.001293134322622791, "clip_ratio/low_min": 0.001293134322622791, "clip_ratio/high_mean": 0.0027587802323978394, "clip_ratio/high_max": 0.0027587802323978394, "clip_ratio/region_mean": 0.00405191455502063, "step_time": 83.41378324700054, "entropy_control/entropy": 0.09167548157514109, "entropy_control/target": 0.038171986696735806, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.168, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 83.31063, "rollout_overhead_time": 0.0, "reward": null }, { "step": 43, "loss": -0.10128452628850937, "grad_norm": 0.318359375, "learning_rate": 1.6000000000000001e-06, "num_tokens": 1175905.0, "completions/mean_length": 342.625, "completions/min_length": 244.5, "completions/max_length": 602.5, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 342.625, "completions/min_terminated_length": 244.5, "completions/max_terminated_length": 602.5, "tools/call_frequency": 3.125, "tools/failure_frequency": 0.1610305979847908, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__native/mean": 0.6161404550075531, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__native/std": 0.0015597647870890796, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__efficiency/mean": 0.5416666865348816, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__efficiency/std": 0.15957120060920715, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/mean": 0.6150930523872375, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/std": 0.0011504920076959024, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/mean": 0.5833333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/std": 0.09622505307197571, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/std": 0.0, "reward": 0.8437418043613434, "reward_std": 0.008564013754948974, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.5, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 1.0, "n_refills": 1.0, "entropy": 0.05819309735670686, "clip_ratio/low_mean": 0.00033602150506339967, "clip_ratio/low_min": 0.00033602150506339967, "clip_ratio/high_mean": 0.0004310344811528921, "clip_ratio/high_max": 0.0004310344811528921, "clip_ratio/region_mean": 0.0007670559862162918, "step_time": 382.6277843600001, "entropy_control/entropy": 0.06136870889889593, "entropy_control/target": 0.038171986696735806, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.172, "adapter_sync_time": 0.0, "generation_time": 268.219008, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 303.843175, "forward_backward_time": 78.680384, "rollout_overhead_time": 35.624167 }, { "step": 44, "loss": -0.10096702724695206, "grad_norm": 0.388671875, "learning_rate": 1.4000000000000001e-06, "entropy": 0.05854875745717436, "clip_ratio/low_mean": 0.0007353825203608721, "clip_ratio/low_min": 0.0007353825203608721, "clip_ratio/high_mean": 0.0015142027841648087, "clip_ratio/high_max": 0.0015142027841648087, "clip_ratio/region_mean": 0.002249585304525681, "step_time": 78.79109064600016, "entropy_control/entropy": 0.06175642943044847, "entropy_control/target": 0.038171986696735806, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.176, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 78.682704, "rollout_overhead_time": 0.0, "reward": null }, { "step": 45, "loss": -0.12669306993484497, "grad_norm": 0.318359375, "learning_rate": 1.2000000000000002e-06, "num_tokens": 1208489.0, "completions/mean_length": 399.0, "completions/min_length": 227.0, "completions/max_length": 809.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 399.0, "completions/min_terminated_length": 227.0, "completions/max_terminated_length": 809.0, "tools/call_frequency": 3.25, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/mean": 0.3412669599056244, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/std": 0.18253393471240997, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/mean": 0.0833333358168602, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/std": 0.1666666716337204, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/mean": -0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/std": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/mean": 0.6126610040664673, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/std": 0.0026778648607432842, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/std": 0.19245010614395142, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/std": 0.0, "reward": 0.5415472984313965, "reward_std": 0.407124400138855, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.029463766026310623, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 269.876155730999, "entropy_control/entropy": 0.02931172858487994, "entropy_control/target": 0.038171986696735806, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.18, "adapter_sync_time": 0.0, "generation_time": 162.155096, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 182.159972, "forward_backward_time": 87.617081, "rollout_overhead_time": 20.004876 }, { "step": 46, "loss": -0.12641510367393494, "grad_norm": 0.26953125, "learning_rate": 1.0000000000000002e-06, "entropy": 0.02980194427073002, "clip_ratio/low_mean": 0.0011407125857658684, "clip_ratio/low_min": 0.0011407125857658684, "clip_ratio/high_mean": 0.0006992460694164038, "clip_ratio/high_max": 0.0006992460694164038, "clip_ratio/region_mean": 0.0018399586551822722, "step_time": 87.70896454199828, "entropy_control/entropy": 0.029673521435050657, "entropy_control/target": 0.038171986696735806, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.184, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 87.609553, "rollout_overhead_time": 0.0, "reward": null }, { "step": 47, "loss": -0.07695724815130234, "grad_norm": 0.2890625, "learning_rate": 8.000000000000001e-07, "num_tokens": 1237272.0, "completions/mean_length": 307.375, "completions/min_length": 238.0, "completions/max_length": 373.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 307.375, "completions/min_terminated_length": 238.0, "completions/max_terminated_length": 373.0, "tools/call_frequency": 3.125, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/mean": 0.6148351430892944, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/std": 0.0013752857921645045, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/mean": 0.5416666865348816, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/std": 0.15957120060920715, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/mean": 0.8294572234153748, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/std": 0.0007778152939863503, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/mean": 0.4166666865348816, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/std": 0.09622503817081451, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/std": 0.0, "reward": 0.9461045265197754, "reward_std": 0.11160773038864136, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.05825923592783511, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 216.35137124500216, "entropy_control/entropy": 0.062156407920524556, "entropy_control/target": 0.038171986696735806, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.188, "adapter_sync_time": 0.0, "generation_time": 119.592444, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 137.843519, "forward_backward_time": 78.410537, "rollout_overhead_time": 18.251074 }, { "step": 48, "loss": -0.07648591697216034, "grad_norm": 0.21875, "learning_rate": 6.000000000000001e-07, "entropy": 0.05864103720523417, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0012007733748760074, "clip_ratio/high_max": 0.0012007733748760074, "clip_ratio/region_mean": 0.0012007733748760074, "step_time": 78.48169384499852, "entropy_control/entropy": 0.06260346602252487, "entropy_control/target": 0.038171986696735806, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.192, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 78.390733, "rollout_overhead_time": 0.0, "reward": null }, { "step": 49, "loss": -0.03466889634728432, "grad_norm": 0.2470703125, "learning_rate": 4.0000000000000003e-07, "num_tokens": 1265889.0, "completions/mean_length": 282.125, "completions/min_length": 244.0, "completions/max_length": 380.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 282.125, "completions/min_terminated_length": 244.0, "completions/max_terminated_length": 380.0, "tools/call_frequency": 3.5, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/mean": 0.6141222715377808, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/std": 0.0015448754420503974, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/std": 0.19245010614395142, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/mean": 0.8297139406204224, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/std": 1.3350198742045905e-06, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/mean": 0.3333333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/std": 0.0, "reward": 0.942751407623291, "reward_std": 0.11102522164583206, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.0250300100306049, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 209.28487726499588, "entropy_control/entropy": 0.027301900069967827, "entropy_control/target": 0.038171986696735806, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.196, "adapter_sync_time": 0.0, "generation_time": 112.177665, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 130.548996, "forward_backward_time": 78.631275, "rollout_overhead_time": 18.371331 }, { "step": 50, "loss": -0.03522661328315735, "grad_norm": 0.1025390625, "learning_rate": 2.0000000000000002e-07, "entropy": 0.02534905809443444, "clip_ratio/low_mean": 0.0004448398540262133, "clip_ratio/low_min": 0.0004448398540262133, "clip_ratio/high_mean": 0.0018459486309438944, "clip_ratio/high_max": 0.0018459486309438944, "clip_ratio/region_mean": 0.0022907884849701077, "step_time": 78.7290867219981, "entropy_control/entropy": 0.02766516995308736, "entropy_control/target": 0.038171986696735806, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.2, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 78.626182, "rollout_overhead_time": 0.0, "reward": null } ], "holdout": [ { "seed": 900000, "sample": 0, "native_reward": 0.6169376468238315, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 386, "reward_channels": { "native": 0.6169376468238315, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "two_pool_cycle" }, { "seed": 900000, "sample": 1, "native_reward": 0.6169388080321323, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 304, "reward_channels": { "native": 0.6169388080321323, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "two_pool_cycle" }, { "seed": 900000, "sample": 2, "native_reward": 0.6169379690916243, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 423, "reward_channels": { "native": 0.6169379690916243, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "two_pool_cycle" }, { "seed": 900000, "sample": 3, "native_reward": 0.6169384763792611, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 317, "reward_channels": { "native": 0.6169384763792611, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "two_pool_cycle" }, { "seed": 900001, "sample": 0, "native_reward": 0.6172326900050679, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 319, "reward_channels": { "native": 0.6172326900050679, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "two_pool_cycle" }, { "seed": 900001, "sample": 1, "native_reward": 0.6172314366859272, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 304, "reward_channels": { "native": 0.6172314366859272, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "two_pool_cycle" }, { "seed": 900001, "sample": 2, "native_reward": 0.6172326385674898, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 252, "reward_channels": { "native": 0.6172326385674898, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "two_pool_cycle" }, { "seed": 900001, "sample": 3, "native_reward": 0.6172317395034662, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 300, "reward_channels": { "native": 0.6172317395034662, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "two_pool_cycle" }, { "seed": 900000, "sample": 0, "native_reward": 0.6168411288846061, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 3, "model_tokens": 337, "reward_channels": { "native": 0.6168411288846061, "efficiency": 0.5, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "triangular_cycle" }, { "seed": 900000, "sample": 1, "native_reward": 0.618344568706403, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 260, "reward_channels": { "native": 0.618344568706403, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "triangular_cycle" }, { "seed": 900000, "sample": 2, "native_reward": 0.61834576808361, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 264, "reward_channels": { "native": 0.61834576808361, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "triangular_cycle" }, { "seed": 900000, "sample": 3, "native_reward": 0.6183447103195866, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 923, "reward_channels": { "native": 0.6183447103195866, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "triangular_cycle" }, { "seed": 900001, "sample": 0, "native_reward": 0.6174224465462672, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 3, "model_tokens": 381, "reward_channels": { "native": 0.6174224465462672, "efficiency": 0.5, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "triangular_cycle" }, { "seed": 900001, "sample": 1, "native_reward": 0.6174236248934742, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 3, "model_tokens": 378, "reward_channels": { "native": 0.6174236248934742, "efficiency": 0.5, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "triangular_cycle" }, { "seed": 900001, "sample": 2, "native_reward": 0.6183223973018531, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 420, "reward_channels": { "native": 0.6183223973018531, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "triangular_cycle" }, { "seed": 900001, "sample": 3, "native_reward": 0.6183236836817555, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 341, "reward_channels": { "native": 0.6183236836817555, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "triangular_cycle" }, { "seed": 900000, "sample": 0, "native_reward": 0.25, "promotion_success": false, "done": false, "errors": 0, "tool_steps": 3, "model_tokens": 266, "reward_channels": { "native": 0.25, "efficiency": 0.0, "reliability": -1.0 }, "verification_level": null, "failed_gates": [ "finalized" ], "family": "competing_cycles" }, { "seed": 900000, "sample": 1, "native_reward": 0.25, "promotion_success": false, "done": false, "errors": 0, "tool_steps": 3, "model_tokens": 232, "reward_channels": { "native": 0.25, "efficiency": 0.0, "reliability": -1.0 }, "verification_level": null, "failed_gates": [ "finalized" ], "family": "competing_cycles" }, { "seed": 900000, "sample": 2, "native_reward": 0.25, "promotion_success": false, "done": false, "errors": 0, "tool_steps": 3, "model_tokens": 235, "reward_channels": { "native": 0.25, "efficiency": 0.0, "reliability": -1.0 }, "verification_level": null, "failed_gates": [ "finalized" ], "family": "competing_cycles" }, { "seed": 900000, "sample": 3, "native_reward": 0.6212279479623682, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 3, "model_tokens": 419, "reward_channels": { "native": 0.6212279479623682, "efficiency": 0.5, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "competing_cycles" }, { "seed": 900001, "sample": 0, "native_reward": 0.6189484372109143, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 555, "reward_channels": { "native": 0.6189484372109143, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "competing_cycles" }, { "seed": 900001, "sample": 1, "native_reward": 0.6211425116085901, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 3, "model_tokens": 547, "reward_channels": { "native": 0.6211425116085901, "efficiency": 0.5, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "competing_cycles" }, { "seed": 900001, "sample": 2, "native_reward": 0.6189508764907581, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 689, "reward_channels": { "native": 0.6189508764907581, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "competing_cycles" }, { "seed": 900001, "sample": 3, "native_reward": 0.6189480592874964, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 571, "reward_channels": { "native": 0.6189480592874964, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "competing_cycles" }, { "seed": 900000, "sample": 0, "native_reward": 0.6181305094102322, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 307, "reward_channels": { "native": 0.6181305094102322, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "four_hop_cycle" }, { "seed": 900000, "sample": 1, "native_reward": 0.6173034664748611, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 3, "model_tokens": 370, "reward_channels": { "native": 0.6173034664748611, "efficiency": 0.5, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "four_hop_cycle" }, { "seed": 900000, "sample": 2, "native_reward": 0.6173049009084548, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 3, "model_tokens": 363, "reward_channels": { "native": 0.6173049009084548, "efficiency": 0.5, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "four_hop_cycle" }, { "seed": 900000, "sample": 3, "native_reward": 0.6181315855299198, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 357, "reward_channels": { "native": 0.6181315855299198, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "four_hop_cycle" }, { "seed": 900001, "sample": 0, "native_reward": 0.6177471050538963, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 286, "reward_channels": { "native": 0.6177471050538963, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "four_hop_cycle" }, { "seed": 900001, "sample": 1, "native_reward": 0.617748567686201, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 286, "reward_channels": { "native": 0.617748567686201, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "four_hop_cycle" }, { "seed": 900001, "sample": 2, "native_reward": 0.617746137146201, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 286, "reward_channels": { "native": 0.617746137146201, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "four_hop_cycle" }, { "seed": 900001, "sample": 3, "native_reward": 0.6177484115209666, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 286, "reward_channels": { "native": 0.6177484115209666, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "four_hop_cycle" }, { "seed": 900000, "sample": 0, "native_reward": 0.175, "promotion_success": false, "done": false, "errors": 0, "tool_steps": 4, "model_tokens": 396, "reward_channels": { "native": 0.175, "efficiency": 0.0, "reliability": -1.0 }, "verification_level": null, "failed_gates": [ "positive_net_profit", "stress_positive", "finalized", "task_complete" ], "family": "size_sensitive_cycle" }, { "seed": 900000, "sample": 1, "native_reward": 0.8332660773209832, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 269, "reward_channels": { "native": 0.8332660773209832, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "size_sensitive_cycle" }, { "seed": 900000, "sample": 2, "native_reward": 0.175, "promotion_success": false, "done": false, "errors": 0, "tool_steps": 4, "model_tokens": 396, "reward_channels": { "native": 0.175, "efficiency": 0.0, "reliability": -1.0 }, "verification_level": null, "failed_gates": [ "positive_net_profit", "stress_positive", "finalized", "task_complete" ], "family": "size_sensitive_cycle" }, { "seed": 900000, "sample": 3, "native_reward": 0.8332646993695965, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 244, "reward_channels": { "native": 0.8332646993695965, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "size_sensitive_cycle" }, { "seed": 900001, "sample": 0, "native_reward": 0.175, "promotion_success": false, "done": false, "errors": 0, "tool_steps": 4, "model_tokens": 392, "reward_channels": { "native": 0.175, "efficiency": 0.0, "reliability": -1.0 }, "verification_level": null, "failed_gates": [ "positive_net_profit", "stress_positive", "finalized", "task_complete" ], "family": "size_sensitive_cycle" }, { "seed": 900001, "sample": 1, "native_reward": 0.8317237538429428, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 242, "reward_channels": { "native": 0.8317237538429428, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "size_sensitive_cycle" }, { "seed": 900001, "sample": 2, "native_reward": 0.831724450047142, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 242, "reward_channels": { "native": 0.831724450047142, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "size_sensitive_cycle" }, { "seed": 900001, "sample": 3, "native_reward": 0.8317246586644467, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 242, "reward_channels": { "native": 0.8317246586644467, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "size_sensitive_cycle" } ] }, { "run_id": "1789166523", "mode": "no-think", "method": "gdpo", "steps": 50, "reward_points": 25, "train_mean": 0.6487702184915543, "train_last10": 0.7727015286684036, "holdout_native": 0.6244599752567455, "holdout_episodes": 40, "holdout_successes": 36, "mean_step_seconds": 167.15891383194008, "label": "GDPO", "state_at_cutoff": "complete", "step_hours": 2.3216515809991676, "peak_allocated_gib": null, "peak_reserved_gib": null, "extra_refill_rounds": 0.0, "groups_refilled": 0.0, "holdout_component_means": { "native": 0.6244599752567455, "efficiency": 0.4083333333333334, "reliability": 0.8 }, "configuration": { "model": { "id": "Qwen/Qwen3-14B", "revision": "40c069824f4251a91eefaf281ebe4c544efd3e18", "loss": "dapo", "algorithm": "grpo", "no_think": true, "lora": true, "quantize": true, "lora_r": 32, "lora_alpha": 64, "use_dora": false }, "loss": "dapo", "normalization": "gdpo", "entropy_control": { "method": "monitor", "epsilon_low": 0.2, "epsilon_high": 0.28, "zeta_initial": 0.001, "zeta_min": 0.0001, "zeta_max": 0.05, "holdout_split": "benchmark", "holdout_seeds_per_family": 2, "num_iterations": 2 }, "dynamic_sampling": false, "sampling": { "group_size": 4, "temperature": 1.0, "max_completion_tokens": 8192, "top_entropy_quantile": 1.0, "temperature_decay_to": null }, "learning_rate": 1e-05, "turn_max_tokens": 1024 }, "source_sha256": { "metrics.jsonl": "bdf092940782709c651cd88ff100d0d5792e48c113165e62feb9bca73f12d1b1", "entropy_holdout_episodes.jsonl": "073f623323f7d481e5983b67cd8834440980d42d23c61d6434c18e9eb0f04ca6", "resolved_training_config.json": "33c8f43a44a7fbefbed916b1e7508136f3678340caa956196a7f542916d1496b" }, "training_steps": [ { "step": 1, "loss": 0.38727250695228577, "grad_norm": 0.1884765625, "learning_rate": 1e-05, "num_tokens": 28004.0, "completions/mean_length": 604.0, "completions/min_length": 278.0, "completions/max_length": 1190.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 604.0, "completions/min_terminated_length": 278.0, "completions/max_terminated_length": 1190.0, "tools/call_frequency": 4.375, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.5011546015739441, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 0.21743638813495636, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.25, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.1666666716337204, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/mean": 0.33700284361839294, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/std": 0.324005663394928, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/mean": 0.1666666716337204, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/std": 0.19245010614395142, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/mean": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/std": 1.154700517654419, "reward": 0.47949540615081787, "reward_std": 0.45456162095069885, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.0511504570604302, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 333.16587591500013, "entropy_control/entropy": 0.031669787994955255, "entropy_control/target": 0.031669787994955255, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.004, "adapter_sync_time": 0.0, "generation_time": 230.422178, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 250.553306, "forward_backward_time": 82.5198, "rollout_overhead_time": 20.131129 }, { "step": 2, "loss": 0.38700491189956665, "grad_norm": 0.1640625, "learning_rate": 9.800000000000001e-06, "entropy": 0.051021368126384914, "clip_ratio/low_mean": 0.0006221849689609371, "clip_ratio/low_min": 0.0006221849689609371, "clip_ratio/high_mean": 0.00044169611646793783, "clip_ratio/high_max": 0.00044169611646793783, "clip_ratio/region_mean": 0.001063881085428875, "step_time": 82.55144402600024, "entropy_control/entropy": 0.03164224714123032, "entropy_control/target": 0.031669787994955255, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.008, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 82.454305, "rollout_overhead_time": 0.0, "reward": null }, { "step": 3, "loss": 0.09951995313167572, "grad_norm": 0.375, "learning_rate": 9.600000000000001e-06, "num_tokens": 52480.0, "completions/mean_length": 343.5, "completions/min_length": 276.0, "completions/max_length": 476.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 343.5, "completions/min_terminated_length": 276.0, "completions/max_terminated_length": 476.0, "tools/call_frequency": 4.25, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.5125374794006348, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 0.1862691342830658, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.2708333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.12400397658348083, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 0.75, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 0.7071067690849304, "reward": 0.6760792136192322, "reward_std": 0.2972210943698883, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.07134074857458472, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 273.98975242299935, "entropy_control/entropy": 0.07246907367917993, "entropy_control/target": 0.031669787994955255, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.012, "adapter_sync_time": 0.0, "generation_time": 174.958456, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 193.567245, "forward_backward_time": 80.309605, "rollout_overhead_time": 18.608788 }, { "step": 4, "loss": 0.09931477159261703, "grad_norm": 0.302734375, "learning_rate": 9.4e-06, "entropy": 0.07043077307753265, "clip_ratio/low_mean": 0.0006172069115564227, "clip_ratio/low_min": 0.0006172069115564227, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0006172069115564227, "step_time": 80.4203468559997, "entropy_control/entropy": 0.07162203048895709, "entropy_control/target": 0.031669787994955255, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.016, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 80.304211, "rollout_overhead_time": 0.0, "reward": null }, { "step": 5, "loss": 0.08481939136981964, "grad_norm": 0.2177734375, "learning_rate": 9.200000000000002e-06, "num_tokens": 77286.0, "completions/mean_length": 391.75, "completions/min_length": 241.0, "completions/max_length": 541.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 391.75, "completions/min_terminated_length": 241.0, "completions/max_terminated_length": 541.0, "tools/call_frequency": 4.625, "tools/failure_frequency": 0.1621621549129486, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.5576282739639282, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 0.1546093374490738, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.3333333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.1781741827726364, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 0.75, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 0.7071067690849304, "reward": 0.724294900894165, "reward_std": 0.3028410077095032, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.5, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.07192487828433514, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 305.7092941430003, "entropy_control/entropy": 0.06914766435804792, "entropy_control/target": 0.031669787994955255, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.02, "adapter_sync_time": 0.0, "generation_time": 213.427861, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 231.105861, "forward_backward_time": 74.503197, "rollout_overhead_time": 17.678 }, { "step": 6, "loss": 0.08426021039485931, "grad_norm": 0.1337890625, "learning_rate": 9e-06, "entropy": 0.07182961772195995, "clip_ratio/low_mean": 0.002259135013446212, "clip_ratio/low_min": 0.002259135013446212, "clip_ratio/high_mean": 0.001180436956929043, "clip_ratio/high_max": 0.001180436956929043, "clip_ratio/region_mean": 0.0034395719703752548, "step_time": 74.60743086900015, "entropy_control/entropy": 0.0689600109913863, "entropy_control/target": 0.031669787994955255, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.024, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 74.508268, "rollout_overhead_time": 0.0, "reward": null }, { "step": 7, "loss": 0.0, "grad_norm": 0.0, "learning_rate": 8.8e-06, "num_tokens": 108425.0, "completions/mean_length": 420.875, "completions/min_length": 273.0, "completions/max_length": 730.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 420.875, "completions/min_terminated_length": 273.0, "completions/max_terminated_length": 730.0, "tools/call_frequency": 3.625, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/mean": 0.25, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/mean": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/mean": -1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/mean": 0.25, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/mean": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/mean": -1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/std": 0.0, "reward": 0.04999999701976776, "reward_std": 0.0, "frac_reward_zero_std": 1.0, "frac_groups_zero_std": 1.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.02595269726589322, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 258.437109813, "entropy_control/entropy": 0.023225980352899613, "entropy_control/target": 0.031669787994955255, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.028, "adapter_sync_time": 0.0, "generation_time": 161.861848, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 180.322231, "forward_backward_time": 78.015972, "rollout_overhead_time": 18.460383 }, { "step": 8, "loss": 0.0, "grad_norm": 0.0, "learning_rate": 8.6e-06, "entropy": 0.02600592316593975, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 78.10814379600015, "entropy_control/entropy": 0.023177197137843157, "entropy_control/target": 0.031669787994955255, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.032, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 78.011566, "rollout_overhead_time": 0.0, "reward": null }, { "step": 9, "loss": -0.5584689378738403, "grad_norm": 0.18359375, "learning_rate": 8.400000000000001e-06, "num_tokens": 137937.0, "completions/mean_length": 401.0, "completions/min_length": 283.0, "completions/max_length": 1078.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 401.0, "completions/min_terminated_length": 283.0, "completions/max_terminated_length": 1078.0, "tools/call_frequency": 3.625, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.609546422958374, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.3333333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/mean": 0.3410886824131012, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/std": 0.1821773648262024, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/mean": 0.0833333358168602, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/std": 0.1666666716337204, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/mean": -0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/std": 1.0, "reward": 0.5357341766357422, "reward_std": 0.4022294282913208, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.5, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.06978383508976549, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 331.3484992100002, "entropy_control/entropy": 0.054703001032588155, "entropy_control/target": 0.031669787994955255, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.036, "adapter_sync_time": 0.0, "generation_time": 218.681912, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 239.45927, "forward_backward_time": 91.792435, "rollout_overhead_time": 20.777357 }, { "step": 10, "loss": -0.5580927729606628, "grad_norm": 0.2041015625, "learning_rate": 8.2e-06, "entropy": 0.06931559549411759, "clip_ratio/low_mean": 0.0016626695287413895, "clip_ratio/low_min": 0.0016626695287413895, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0016626695287413895, "step_time": 91.88299751500108, "entropy_control/entropy": 0.05430791565556925, "entropy_control/target": 0.031669787994955255, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.04, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 91.786498, "rollout_overhead_time": 0.0, "reward": null }, { "step": 11, "loss": 0.1361590027809143, "grad_norm": 0.2060546875, "learning_rate": 8.000000000000001e-06, "num_tokens": 167796.0, "completions/mean_length": 419.375, "completions/min_length": 243.0, "completions/max_length": 728.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 419.375, "completions/min_terminated_length": 243.0, "completions/max_terminated_length": 728.0, "tools/call_frequency": 3.875, "tools/failure_frequency": 0.12903225421905518, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__native/mean": 0.4959903657436371, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__native/std": 0.21461595594882965, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__efficiency/mean": 0.3333333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__efficiency/std": 0.27216553688049316, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__reliability/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__reliability/std": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/mean": 0.6115941405296326, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/std": 0.002659526187926531, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/mean": 0.4583333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/std": 0.15957120060920715, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/std": 0.0, "reward": 0.7235839366912842, "reward_std": 0.3029329180717468, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.03277585830073804, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 282.3033622840003, "entropy_control/entropy": 0.029470079562793985, "entropy_control/target": 0.031669787994955255, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.044, "adapter_sync_time": 0.0, "generation_time": 183.821638, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 202.476008, "forward_backward_time": 79.726722, "rollout_overhead_time": 18.65437 }, { "step": 12, "loss": 0.13542595505714417, "grad_norm": 0.1689453125, "learning_rate": 7.800000000000002e-06, "entropy": 0.033271659864112735, "clip_ratio/low_mean": 0.00034340660204179585, "clip_ratio/low_min": 0.00034340660204179585, "clip_ratio/high_mean": 0.0016978764615487307, "clip_ratio/high_max": 0.0016978764615487307, "clip_ratio/region_mean": 0.0020412830635905266, "step_time": 79.83501741999999, "entropy_control/entropy": 0.029904279471660317, "entropy_control/target": 0.031669787994955255, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.048, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 79.737631, "rollout_overhead_time": 0.0, "reward": null }, { "step": 13, "loss": -0.019682511687278748, "grad_norm": 0.1943359375, "learning_rate": 7.600000000000001e-06, "num_tokens": 196405.0, "completions/mean_length": 287.125, "completions/min_length": 242.0, "completions/max_length": 355.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 287.125, "completions/min_terminated_length": 242.0, "completions/max_terminated_length": 355.0, "tools/call_frequency": 3.5, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.6150088310241699, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 1.1664858448057203e-06, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.3333333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/mean": 0.25, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/mean": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/mean": -1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/std": 0.0, "reward": 0.4408377408981323, "reward_std": 0.41782310605049133, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.5, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.05726331286132336, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 224.77536741099993, "entropy_control/entropy": 0.06022444738495296, "entropy_control/target": 0.031669787994955255, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.052, "adapter_sync_time": 0.0, "generation_time": 111.617821, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 132.484956, "forward_backward_time": 92.193062, "rollout_overhead_time": 20.867135 }, { "step": 14, "loss": -0.019479095935821533, "grad_norm": 0.197265625, "learning_rate": 7.4e-06, "entropy": 0.0561726544983685, "clip_ratio/low_mean": 0.0007928410486783832, "clip_ratio/low_min": 0.0007928410486783832, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0007928410486783832, "step_time": 92.27920960200072, "entropy_control/entropy": 0.05913788510035027, "entropy_control/target": 0.031669787994955255, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.056, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 92.184823, "rollout_overhead_time": 0.0, "reward": null }, { "step": 15, "loss": 0.12864768505096436, "grad_norm": 0.361328125, "learning_rate": 7.2000000000000005e-06, "num_tokens": 227279.0, "completions/mean_length": 371.75, "completions/min_length": 248.0, "completions/max_length": 651.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 371.75, "completions/min_terminated_length": 248.0, "completions/max_terminated_length": 651.0, "tools/call_frequency": 3.625, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/mean": 0.25, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/mean": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/mean": -1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/mean": 0.3937557339668274, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/std": 0.2875114977359772, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/mean": 0.0833333358168602, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/std": 0.1666666716337204, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/mean": -0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/std": 1.0, "reward": 0.17396119236946106, "reward_std": 0.3506152331829071, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.5, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.02486211364157498, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 247.53113513699964, "entropy_control/entropy": 0.024271421530545952, "entropy_control/target": 0.031669787994955255, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.06, "adapter_sync_time": 0.0, "generation_time": 148.448901, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 167.003873, "forward_backward_time": 80.431971, "rollout_overhead_time": 18.554972 }, { "step": 16, "loss": 0.1289738416671753, "grad_norm": 0.1923828125, "learning_rate": 7e-06, "entropy": 0.02516411303076893, "clip_ratio/low_mean": 0.001231933245435357, "clip_ratio/low_min": 0.001231933245435357, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.001231933245435357, "step_time": 80.5280956029992, "entropy_control/entropy": 0.02442323211793303, "entropy_control/target": 0.031669787994955255, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.064, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 80.434229, "rollout_overhead_time": 0.0, "reward": null }, { "step": 17, "loss": -0.3020651340484619, "grad_norm": 0.2890625, "learning_rate": 6.800000000000001e-06, "num_tokens": 259498.0, "completions/mean_length": 332.375, "completions/min_length": 213.0, "completions/max_length": 577.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 332.375, "completions/min_terminated_length": 213.0, "completions/max_terminated_length": 577.0, "tools/call_frequency": 3.625, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/mean": 0.3412719666957855, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/std": 0.18254393339157104, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/mean": 0.0833333358168602, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/std": 0.1666666716337204, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/mean": -0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/std": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/mean": 0.6101261377334595, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/std": 2.9887635264458368e-06, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/mean": 0.3333333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/std": 0.0, "reward": 0.5361157059669495, "reward_std": 0.4025455117225647, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.020030950428918004, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 271.0126385839999, "entropy_control/entropy": 0.020303291219384022, "entropy_control/target": 0.031669787994955255, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.068, "adapter_sync_time": 0.0, "generation_time": 159.057043, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 179.745435, "forward_backward_time": 91.169739, "rollout_overhead_time": 20.688392 }, { "step": 18, "loss": -0.3024434447288513, "grad_norm": 0.1875, "learning_rate": 6.600000000000001e-06, "entropy": 0.02001496049342677, "clip_ratio/low_mean": 0.002471604704624042, "clip_ratio/low_min": 0.002471604704624042, "clip_ratio/high_mean": 0.000216637781704776, "clip_ratio/high_max": 0.000216637781704776, "clip_ratio/region_mean": 0.002688242486328818, "step_time": 91.25010780400044, "entropy_control/entropy": 0.020251737270828584, "entropy_control/target": 0.031669787994955255, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.072, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 91.154146, "rollout_overhead_time": 0.0, "reward": null }, { "step": 19, "loss": 0.22143757343292236, "grad_norm": 0.31640625, "learning_rate": 6.4000000000000006e-06, "num_tokens": 285248.0, "completions/mean_length": 324.25, "completions/min_length": 270.0, "completions/max_length": 532.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 324.25, "completions/min_terminated_length": 270.0, "completions/max_terminated_length": 532.0, "tools/call_frequency": 4.0, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.6130162477493286, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 1.3635444702231325e-06, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.3333333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/mean": 0.6844160556793213, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/std": 0.28961068391799927, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/mean": 0.25, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/std": 0.1666666716337204, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/std": 1.0, "reward": 0.8132994771003723, "reward_std": 0.32645106315612793, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.05664578056894243, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 242.5849810360005, "entropy_control/entropy": 0.05519685239657242, "entropy_control/target": 0.031669787994955255, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.076, "adapter_sync_time": 0.0, "generation_time": 140.450356, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 159.358096, "forward_backward_time": 83.130439, "rollout_overhead_time": 18.90774 }, { "step": 20, "loss": 0.2211138904094696, "grad_norm": 0.2080078125, "learning_rate": 6.200000000000001e-06, "entropy": 0.0572046481538564, "clip_ratio/low_mean": 0.000234962411923334, "clip_ratio/low_min": 0.000234962411923334, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.000234962411923334, "step_time": 83.22634736600048, "entropy_control/entropy": 0.05571636632192484, "entropy_control/target": 0.031669787994955255, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.08, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 83.128502, "rollout_overhead_time": 0.0, "reward": null }, { "step": 21, "loss": 0.0011397600173950195, "grad_norm": 0.060546875, "learning_rate": 6e-06, "num_tokens": 317237.0, "completions/mean_length": 295.625, "completions/min_length": 255.0, "completions/max_length": 310.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 295.625, "completions/min_terminated_length": 255.0, "completions/max_terminated_length": 310.0, "tools/call_frequency": 3.5, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/mean": 0.25, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/mean": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/mean": -1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/mean": 0.6147057414054871, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/std": 1.8970764585901634e-06, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/mean": 0.3333333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/std": 0.0, "reward": 0.4406861960887909, "reward_std": 0.4176611304283142, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.5, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.023078047670423985, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 199.58957194699906, "entropy_control/entropy": 0.022820973648635442, "entropy_control/target": 0.031669787994955255, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.084, "adapter_sync_time": 0.0, "generation_time": 89.412534, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 109.751578, "forward_backward_time": 89.716512, "rollout_overhead_time": 20.339044 }, { "step": 22, "loss": 0.0010419189929962158, "grad_norm": 0.060546875, "learning_rate": 5.8e-06, "entropy": 0.02251318353228271, "clip_ratio/low_mean": 0.0004208754107821733, "clip_ratio/low_min": 0.0004208754107821733, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0004208754107821733, "step_time": 89.83986751499924, "entropy_control/entropy": 0.022247935782540286, "entropy_control/target": 0.031669787994955255, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.088, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 89.723634, "rollout_overhead_time": 0.0, "reward": null }, { "step": 23, "loss": 0.32887518405914307, "grad_norm": 0.19921875, "learning_rate": 5.600000000000001e-06, "num_tokens": 343185.0, "completions/mean_length": 354.0, "completions/min_length": 238.0, "completions/max_length": 784.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 354.0, "completions/min_terminated_length": 238.0, "completions/max_terminated_length": 784.0, "tools/call_frequency": 4.0, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.6096732020378113, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 1.4813483630859992e-06, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.3333333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/mean": 0.5377726554870605, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/std": 0.33229130506515503, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/mean": 0.1666666716337204, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/std": 0.19245010614395142, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/mean": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/std": 1.154700517654419, "reward": 0.6862229704856873, "reward_std": 0.40382930636405945, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.05229326005792245, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 298.13588110399996, "entropy_control/entropy": 0.04591277142366583, "entropy_control/target": 0.031669787994955255, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.092, "adapter_sync_time": 0.0, "generation_time": 196.645474, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 215.577114, "forward_backward_time": 82.45823, "rollout_overhead_time": 18.93164 }, { "step": 24, "loss": 0.32888084650039673, "grad_norm": 0.1845703125, "learning_rate": 5.400000000000001e-06, "entropy": 0.052134542376734316, "clip_ratio/low_mean": 0.0001594387722434476, "clip_ratio/low_min": 0.0001594387722434476, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0001594387722434476, "step_time": 82.56295539700022, "entropy_control/entropy": 0.04576141435445114, "entropy_control/target": 0.031669787994955255, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.096, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 82.460355, "rollout_overhead_time": 0.0, "reward": null }, { "step": 25, "loss": -0.037824004888534546, "grad_norm": 0.1376953125, "learning_rate": 5.2e-06, "num_tokens": 370439.0, "completions/mean_length": 312.75, "completions/min_length": 280.0, "completions/max_length": 438.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 312.75, "completions/min_terminated_length": 280.0, "completions/max_terminated_length": 438.0, "tools/call_frequency": 4.125, "tools/failure_frequency": 0.03030303120613098, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.6147221326828003, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 1.3013301440878422e-06, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.3333333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/mean": 0.6143907308578491, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/std": 1.6236112969636451e-06, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/mean": 0.3333333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/std": 0.0, "reward": 0.8312231302261353, "reward_std": 0.00017713104898575693, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.06422434281557798, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 235.52427597500082, "entropy_control/entropy": 0.07347538832630605, "entropy_control/target": 0.031669787994955255, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.1, "adapter_sync_time": 0.0, "generation_time": 126.756355, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 147.002874, "forward_backward_time": 88.422992, "rollout_overhead_time": 20.24652 }, { "step": 26, "loss": -0.03768989443778992, "grad_norm": 0.138671875, "learning_rate": 5e-06, "entropy": 0.06442170706577599, "clip_ratio/low_mean": 0.00042372880852781236, "clip_ratio/low_min": 0.00042372880852781236, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.00042372880852781236, "step_time": 88.50265450799907, "entropy_control/entropy": 0.07369115508136113, "entropy_control/target": 0.031669787994955255, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.104, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 88.40374, "rollout_overhead_time": 0.0, "reward": null }, { "step": 27, "loss": -0.027953416109085083, "grad_norm": 0.31640625, "learning_rate": 4.800000000000001e-06, "num_tokens": 399193.0, "completions/mean_length": 287.75, "completions/min_length": 239.0, "completions/max_length": 336.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 287.75, "completions/min_terminated_length": 239.0, "completions/max_terminated_length": 336.0, "tools/call_frequency": 3.75, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/mean": 0.614199697971344, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/std": 0.0012803486315533519, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/mean": 0.4166666865348816, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/std": 0.1666666716337204, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/mean": 0.828781008720398, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/std": 1.2249201972736046e-06, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/mean": 0.3333333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/std": 0.0, "reward": 0.9402403235435486, "reward_std": 0.11264732480049133, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.027143670595251024, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 213.40258446899952, "entropy_control/entropy": 0.02844951199716676, "entropy_control/target": 0.031669787994955255, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.108, "adapter_sync_time": 0.0, "generation_time": 115.68746, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 134.058492, "forward_backward_time": 79.244631, "rollout_overhead_time": 18.371032 }, { "step": 28, "loss": -0.028048425912857056, "grad_norm": 0.322265625, "learning_rate": 4.600000000000001e-06, "entropy": 0.02642580756219104, "clip_ratio/low_mean": 0.000899292150279507, "clip_ratio/low_min": 0.000899292150279507, "clip_ratio/high_mean": 0.0027432642818894237, "clip_ratio/high_max": 0.0027432642818894237, "clip_ratio/region_mean": 0.0036425564321689308, "step_time": 79.33389681799963, "entropy_control/entropy": 0.027779768298991425, "entropy_control/target": 0.031669787994955255, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.112, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 79.235427, "rollout_overhead_time": 0.0, "reward": null }, { "step": 29, "loss": -0.03274300694465637, "grad_norm": 0.1572265625, "learning_rate": 4.4e-06, "num_tokens": 427738.0, "completions/mean_length": 281.625, "completions/min_length": 234.0, "completions/max_length": 311.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 281.625, "completions/min_terminated_length": 234.0, "completions/max_terminated_length": 311.0, "tools/call_frequency": 3.5, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.614264965057373, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 1.7530252307551564e-06, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.3333333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/mean": 0.25, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/mean": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/mean": -1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/std": 0.0, "reward": 0.4404658079147339, "reward_std": 0.4174254834651947, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.5, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.05104556283913553, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 214.97263195699907, "entropy_control/entropy": 0.05276931485631387, "entropy_control/target": 0.031669787994955255, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.116, "adapter_sync_time": 0.0, "generation_time": 102.219213, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 123.103668, "forward_backward_time": 91.761182, "rollout_overhead_time": 20.884456 }, { "step": 30, "loss": -0.032407574355602264, "grad_norm": 0.1513671875, "learning_rate": 4.2000000000000004e-06, "entropy": 0.05202084640040994, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.000803858507424593, "clip_ratio/high_max": 0.000803858507424593, "clip_ratio/region_mean": 0.000803858507424593, "step_time": 91.84039551899878, "entropy_control/entropy": 0.05378079177366842, "entropy_control/target": 0.031669787994955255, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.12, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 91.745394, "rollout_overhead_time": 0.0, "reward": null }, { "step": 31, "loss": -0.06863882392644882, "grad_norm": 0.443359375, "learning_rate": 4.000000000000001e-06, "num_tokens": 454653.0, "completions/mean_length": 276.875, "completions/min_length": 245.0, "completions/max_length": 337.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 276.875, "completions/min_terminated_length": 245.0, "completions/max_terminated_length": 337.0, "tools/call_frequency": 3.875, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/mean": 0.8293575048446655, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/std": 0.001310738967731595, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/mean": 0.3541666865348816, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/std": 0.058925557881593704, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/std": 0.0, "reward": 1.0470658540725708, "reward_std": 0.003988455515354872, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.5, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.028321846970357, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 189.37133719200028, "entropy_control/entropy": 0.030888804714896197, "entropy_control/target": 0.031669787994955255, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.124, "adapter_sync_time": 0.0, "generation_time": 107.358861, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 123.11143, "forward_backward_time": 66.16048, "rollout_overhead_time": 15.752569 }, { "step": 32, "loss": -0.06865701079368591, "grad_norm": 0.34375, "learning_rate": 3.8000000000000005e-06, "entropy": 0.02845800167415291, "clip_ratio/low_mean": 0.0005081300623714924, "clip_ratio/low_min": 0.0005081300623714924, "clip_ratio/high_mean": 0.0008554160012863576, "clip_ratio/high_max": 0.0008554160012863576, "clip_ratio/region_mean": 0.00136354606365785, "step_time": 66.25698401200225, "entropy_control/entropy": 0.031066890709079847, "entropy_control/target": 0.031669787994955255, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.128, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 66.158577, "rollout_overhead_time": 0.0, "reward": null }, { "step": 33, "loss": 0.18087846040725708, "grad_norm": 0.361328125, "learning_rate": 3.6000000000000003e-06, "num_tokens": 480114.0, "completions/mean_length": 283.125, "completions/min_length": 237.0, "completions/max_length": 420.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 283.125, "completions/min_terminated_length": 237.0, "completions/max_terminated_length": 420.0, "tools/call_frequency": 4.0, "tools/failure_frequency": 0.0625, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.6134346127510071, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 1.212775146086642e-06, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.3333333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/mean": 0.6849815845489502, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/std": 0.28999248147010803, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/mean": 0.3333333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/std": 0.27216553688049316, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/std": 1.0, "reward": 0.8158747553825378, "reward_std": 0.32845762372016907, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.0443328854162246, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 238.86322008099887, "entropy_control/entropy": 0.0440509048773095, "entropy_control/target": 0.031669787994955255, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.132, "adapter_sync_time": 0.0, "generation_time": 136.054203, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 155.232774, "forward_backward_time": 83.528896, "rollout_overhead_time": 19.178571 }, { "step": 34, "loss": 0.1809271275997162, "grad_norm": 0.5078125, "learning_rate": 3.4000000000000005e-06, "entropy": 0.04486644431017339, "clip_ratio/low_mean": 0.0007286535401362926, "clip_ratio/low_min": 0.0007286535401362926, "clip_ratio/high_mean": 0.001054852269589901, "clip_ratio/high_max": 0.001054852269589901, "clip_ratio/region_mean": 0.0017835058097261935, "step_time": 83.6242890880003, "entropy_control/entropy": 0.04458745232396342, "entropy_control/target": 0.031669787994955255, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.136, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 83.524807, "rollout_overhead_time": 0.0, "reward": null }, { "step": 35, "loss": 0.02203717827796936, "grad_norm": 0.1298828125, "learning_rate": 3.2000000000000003e-06, "num_tokens": 516202.0, "completions/mean_length": 644.0, "completions/min_length": 252.0, "completions/max_length": 972.0, "completions/clipped_ratio": 0.25, "completions/mean_terminated_length": 534.6666870117188, "completions/min_terminated_length": 252.0, "completions/max_terminated_length": 805.0, "tools/call_frequency": 3.625, "tools/failure_frequency": 0.06896551698446274, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/mean": 0.3413922190666199, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/std": 0.16922716796398163, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/mean": 0.1041666716337204, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/std": 0.19795581698417664, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/mean": -0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/std": 0.9258201122283936, "reward": 0.24660053849220276, "reward_std": 0.364045649766922, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.5, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.021460627438500524, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 325.1009769660004, "entropy_control/entropy": 0.0255883422361711, "entropy_control/target": 0.031669787994955255, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.14, "adapter_sync_time": 0.0, "generation_time": 206.78301, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 228.799691, "forward_backward_time": 96.20013, "rollout_overhead_time": 22.016681 }, { "step": 36, "loss": 0.022266268730163574, "grad_norm": 0.14453125, "learning_rate": 3e-06, "entropy": 0.021530213358346373, "clip_ratio/low_mean": 0.00031407034839503467, "clip_ratio/low_min": 0.00031407034839503467, "clip_ratio/high_mean": 0.0003696877392940223, "clip_ratio/high_max": 0.0003696877392940223, "clip_ratio/region_mean": 0.000683758087689057, "step_time": 96.2894305789996, "entropy_control/entropy": 0.025569877103817792, "entropy_control/target": 0.031669787994955255, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.144, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 96.18998, "rollout_overhead_time": 0.0, "reward": null }, { "step": 37, "loss": -0.025003258138895035, "grad_norm": 0.392578125, "learning_rate": 2.8000000000000003e-06, "num_tokens": 541783.0, "completions/mean_length": 313.625, "completions/min_length": 249.0, "completions/max_length": 583.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 313.625, "completions/min_terminated_length": 249.0, "completions/max_terminated_length": 583.0, "tools/call_frequency": 3.25, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.6136628985404968, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 1.1413420679673436e-06, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.3333333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__native/mean": 0.6176912784576416, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__native/std": 0.0016456745797768235, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__efficiency/mean": 0.5833333730697632, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__efficiency/std": 0.1666666716337204, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__reliability/std": 0.0, "reward": 0.8385937213897705, "reward_std": 0.010987779125571251, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.5, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.06341225933283567, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 234.417130584, "entropy_control/entropy": 0.06292146563919619, "entropy_control/target": 0.031669787994955255, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.148, "adapter_sync_time": 0.0, "generation_time": 130.063177, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 149.50278, "forward_backward_time": 84.813356, "rollout_overhead_time": 19.439603 }, { "step": 38, "loss": -0.025411099195480347, "grad_norm": 0.1982421875, "learning_rate": 2.6e-06, "entropy": 0.06308346474543214, "clip_ratio/low_mean": 0.0023818135377950966, "clip_ratio/low_min": 0.0023818135377950966, "clip_ratio/high_mean": 0.00021440823911689222, "clip_ratio/high_max": 0.00021440823911689222, "clip_ratio/region_mean": 0.002596221776911989, "step_time": 84.90608804000021, "entropy_control/entropy": 0.062550013713793, "entropy_control/target": 0.031669787994955255, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.152, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 84.798238, "rollout_overhead_time": 0.0, "reward": null }, { "step": 39, "loss": -0.0020994246006011963, "grad_norm": 0.291015625, "learning_rate": 2.4000000000000003e-06, "num_tokens": 569192.0, "completions/mean_length": 331.125, "completions/min_length": 237.0, "completions/max_length": 581.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 331.125, "completions/min_terminated_length": 237.0, "completions/max_terminated_length": 581.0, "tools/call_frequency": 2.75, "tools/failure_frequency": 0.1818181872367859, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__native/mean": 0.6184960007667542, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__native/std": 0.0007463989895768464, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__efficiency/mean": 0.625, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__efficiency/std": 0.07715168595314026, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__reliability/std": 0.0, "reward": 0.8497459888458252, "reward_std": 0.004449996631592512, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.08218858856707811, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 171.43775163800092, "entropy_control/entropy": 0.07829559810819023, "entropy_control/target": 0.031669787994955255, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.156, "adapter_sync_time": 0.0, "generation_time": 95.560153, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 110.213943, "forward_backward_time": 61.121981, "rollout_overhead_time": 14.65379 }, { "step": 40, "loss": -0.002934083342552185, "grad_norm": 0.1826171875, "learning_rate": 2.2e-06, "entropy": 0.08189644711092114, "clip_ratio/low_mean": 0.0011502679553814232, "clip_ratio/low_min": 0.0011502679553814232, "clip_ratio/high_mean": 0.001474380391300656, "clip_ratio/high_max": 0.001474380391300656, "clip_ratio/region_mean": 0.002624648346682079, "step_time": 61.22994593600197, "entropy_control/entropy": 0.07795156602255307, "entropy_control/target": 0.031669787994955255, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.16, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 61.128949, "rollout_overhead_time": 0.0, "reward": null }, { "step": 41, "loss": -0.3022345304489136, "grad_norm": 0.298828125, "learning_rate": 2.0000000000000003e-06, "num_tokens": 599963.0, "completions/mean_length": 365.375, "completions/min_length": 221.0, "completions/max_length": 707.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 365.375, "completions/min_terminated_length": 221.0, "completions/max_terminated_length": 707.0, "tools/call_frequency": 2.875, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__native/mean": 0.6151466369628906, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__native/std": 0.0030156085267663, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__efficiency/mean": 0.5833333730697632, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__efficiency/std": 0.1666666716337204, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/mean": 0.4327476918697357, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/std": 0.21102263033390045, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/mean": 0.2083333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/std": 0.25, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/mean": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/std": 1.154700517654419, "reward": 0.643738865852356, "reward_std": 0.3665681481361389, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.053320204198826104, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 299.5808600929995, "entropy_control/entropy": 0.05409244801659026, "entropy_control/target": 0.031669787994955255, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.164, "adapter_sync_time": 0.0, "generation_time": 195.03342, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 214.75164, "forward_backward_time": 84.702605, "rollout_overhead_time": 19.71822 }, { "step": 42, "loss": -0.3024761378765106, "grad_norm": 0.34375, "learning_rate": 1.8000000000000001e-06, "entropy": 0.05294223362579942, "clip_ratio/low_mean": 0.0009981368493754417, "clip_ratio/low_min": 0.0009981368493754417, "clip_ratio/high_mean": 0.00101072974212002, "clip_ratio/high_max": 0.00101072974212002, "clip_ratio/region_mean": 0.0020088665914954618, "step_time": 84.81286498499958, "entropy_control/entropy": 0.05380071883417856, "entropy_control/target": 0.031669787994955255, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.168, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 84.690381, "rollout_overhead_time": 0.0, "reward": null }, { "step": 43, "loss": -0.2147461324930191, "grad_norm": 0.1884765625, "learning_rate": 1.6000000000000001e-06, "num_tokens": 629722.0, "completions/mean_length": 411.875, "completions/min_length": 310.0, "completions/max_length": 651.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 411.875, "completions/min_terminated_length": 310.0, "completions/max_terminated_length": 651.0, "tools/call_frequency": 3.375, "tools/failure_frequency": 0.29629629850387573, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__native/mean": 0.6168264746665955, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__native/std": 0.00071199971716851, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__efficiency/mean": 0.5416666865348816, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__efficiency/std": 0.0833333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/mean": 0.6166301965713501, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/std": 1.2825390740545117e-06, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/mean": 0.6666666865348816, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/std": 0.0, "reward": 0.8469367027282715, "reward_std": 0.004546592943370342, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.05981270456686616, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 250.78653634400234, "entropy_control/entropy": 0.05563764885202113, "entropy_control/target": 0.031669787994955255, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.172, "adapter_sync_time": 0.0, "generation_time": 166.051046, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 182.080745, "forward_backward_time": 68.607153, "rollout_overhead_time": 16.029699 }, { "step": 44, "loss": -0.2156495749950409, "grad_norm": 0.1943359375, "learning_rate": 1.4000000000000001e-06, "entropy": 0.05967620899900794, "clip_ratio/low_mean": 0.0004032258002553135, "clip_ratio/low_min": 0.0004032258002553135, "clip_ratio/high_mean": 0.00034916200092993677, "clip_ratio/high_max": 0.00034916200092993677, "clip_ratio/region_mean": 0.0007523878011852503, "step_time": 68.71801683599915, "entropy_control/entropy": 0.055452486125669935, "entropy_control/target": 0.031669787994955255, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.176, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 68.617309, "rollout_overhead_time": 0.0, "reward": null }, { "step": 45, "loss": -0.2568885087966919, "grad_norm": 0.28515625, "learning_rate": 1.2000000000000002e-06, "num_tokens": 661718.0, "completions/mean_length": 325.5, "completions/min_length": 239.0, "completions/max_length": 564.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 325.5, "completions/min_terminated_length": 239.0, "completions/max_terminated_length": 564.0, "tools/call_frequency": 2.875, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/mean": 0.3418276309967041, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/std": 0.183655247092247, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/mean": 0.0833333358168602, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/std": 0.1666666716337204, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/mean": -0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/std": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/mean": 0.614968478679657, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/std": 0.0015859014820307493, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/mean": 0.5833333730697632, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/std": 0.1666666716337204, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/std": 0.0, "reward": 0.5450646877288818, "reward_std": 0.41001835465431213, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.03933716518804431, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 231.37294147800094, "entropy_control/entropy": 0.03803405947033658, "entropy_control/target": 0.031669787994955255, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.18, "adapter_sync_time": 0.0, "generation_time": 123.82628, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 143.807056, "forward_backward_time": 87.467611, "rollout_overhead_time": 19.980777 }, { "step": 46, "loss": -0.2571696639060974, "grad_norm": 0.216796875, "learning_rate": 1.0000000000000002e-06, "entropy": 0.03926300717284903, "clip_ratio/low_mean": 0.00048638132284395397, "clip_ratio/low_min": 0.00048638132284395397, "clip_ratio/high_mean": 0.0015322959225159138, "clip_ratio/high_max": 0.0015322959225159138, "clip_ratio/region_mean": 0.002018677245359868, "step_time": 87.56447466299505, "entropy_control/entropy": 0.03820173846617607, "entropy_control/target": 0.031669787994955255, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.184, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 87.465584, "rollout_overhead_time": 0.0, "reward": null }, { "step": 47, "loss": -0.10692477971315384, "grad_norm": 0.44140625, "learning_rate": 8.000000000000001e-07, "num_tokens": 690657.0, "completions/mean_length": 326.875, "completions/min_length": 269.0, "completions/max_length": 462.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 326.875, "completions/min_terminated_length": 269.0, "completions/max_terminated_length": 462.0, "tools/call_frequency": 3.0, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/mean": 0.6150550842285156, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/std": 0.0019571385346353054, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/mean": 0.5833333730697632, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/std": 0.1666666716337204, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/mean": 0.8290933966636658, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/std": 0.0014574839733541012, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/mean": 0.4166666865348816, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/std": 0.09622503817081451, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/std": 0.0, "reward": 0.947074294090271, "reward_std": 0.11023653298616409, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.06172364507801831, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 232.82608543000242, "entropy_control/entropy": 0.0650129977897534, "entropy_control/target": 0.031669787994955255, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.188, "adapter_sync_time": 0.0, "generation_time": 135.819103, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 154.21941, "forward_backward_time": 78.509995, "rollout_overhead_time": 18.400307 }, { "step": 48, "loss": -0.1069703996181488, "grad_norm": 0.6171875, "learning_rate": 6.000000000000001e-07, "entropy": 0.06248296645935625, "clip_ratio/low_mean": 0.0004646840097848326, "clip_ratio/low_min": 0.0004646840097848326, "clip_ratio/high_mean": 0.0010518127819523215, "clip_ratio/high_max": 0.0010518127819523215, "clip_ratio/region_mean": 0.0015164967917371541, "step_time": 78.5874558070027, "entropy_control/entropy": 0.06575856846425679, "entropy_control/target": 0.031669787994955255, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.192, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 78.485602, "rollout_overhead_time": 0.0, "reward": null }, { "step": 49, "loss": -0.005773954559117556, "grad_norm": 0.2275390625, "learning_rate": 4.0000000000000003e-07, "num_tokens": 719568.0, "completions/mean_length": 318.875, "completions/min_length": 263.0, "completions/max_length": 426.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 318.875, "completions/min_terminated_length": 263.0, "completions/max_terminated_length": 426.0, "tools/call_frequency": 3.125, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/mean": 0.6151204109191895, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/std": 0.0008677542791701853, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/mean": 0.625, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/std": 0.0833333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/mean": 0.8296027183532715, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/std": 8.28054169232928e-07, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/mean": 0.3333333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/std": 0.0, "reward": 0.9463198781013489, "reward_std": 0.10690131783485413, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.5, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.06097144330851734, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 201.91460961000303, "entropy_control/entropy": 0.06719659389612954, "entropy_control/target": 0.031669787994955255, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.196, "adapter_sync_time": 0.0, "generation_time": 114.081295, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 130.872651, "forward_backward_time": 70.938517, "rollout_overhead_time": 16.791356 }, { "step": 50, "loss": -0.0063142236322164536, "grad_norm": 0.232421875, "learning_rate": 2.0000000000000002e-07, "entropy": 0.06048206565901637, "clip_ratio/low_mean": 0.0012763740378431976, "clip_ratio/low_min": 0.0012763740378431976, "clip_ratio/high_mean": 0.0008598968852311373, "clip_ratio/high_max": 0.0008598968852311373, "clip_ratio/region_mean": 0.002136270923074335, "step_time": 71.03282021299674, "entropy_control/entropy": 0.06666982661029804, "entropy_control/target": 0.031669787994955255, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.2, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 70.932853, "rollout_overhead_time": 0.0, "reward": null } ], "holdout": [ { "seed": 900000, "sample": 0, "native_reward": 0.6174455440146712, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 313, "reward_channels": { "native": 0.6174455440146712, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "two_pool_cycle" }, { "seed": 900000, "sample": 1, "native_reward": 0.6174466939589095, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 224, "reward_channels": { "native": 0.6174466939589095, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "two_pool_cycle" }, { "seed": 900000, "sample": 2, "native_reward": 0.6174463633467221, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 302, "reward_channels": { "native": 0.6174463633467221, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "two_pool_cycle" }, { "seed": 900000, "sample": 3, "native_reward": 0.617446325597054, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 242, "reward_channels": { "native": 0.617446325597054, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "two_pool_cycle" }, { "seed": 900001, "sample": 0, "native_reward": 0.6178422403872945, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 324, "reward_channels": { "native": 0.6178422403872945, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "two_pool_cycle" }, { "seed": 900001, "sample": 1, "native_reward": 0.6178429537705367, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 267, "reward_channels": { "native": 0.6178429537705367, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "two_pool_cycle" }, { "seed": 900001, "sample": 2, "native_reward": 0.6178425815094428, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 298, "reward_channels": { "native": 0.6178425815094428, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "two_pool_cycle" }, { "seed": 900001, "sample": 3, "native_reward": 0.6178426923321967, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 298, "reward_channels": { "native": 0.6178426923321967, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "two_pool_cycle" }, { "seed": 900000, "sample": 0, "native_reward": 0.6195805269881217, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 322, "reward_channels": { "native": 0.6195805269881217, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "triangular_cycle" }, { "seed": 900000, "sample": 1, "native_reward": 0.6195813048573601, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 260, "reward_channels": { "native": 0.6195813048573601, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "triangular_cycle" }, { "seed": 900000, "sample": 2, "native_reward": 0.6195814813085709, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 275, "reward_channels": { "native": 0.6195814813085709, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "triangular_cycle" }, { "seed": 900000, "sample": 3, "native_reward": 0.6195818791400944, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 294, "reward_channels": { "native": 0.6195818791400944, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "triangular_cycle" }, { "seed": 900001, "sample": 0, "native_reward": 0.6193224283775953, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 3, "model_tokens": 318, "reward_channels": { "native": 0.6193224283775953, "efficiency": 0.5, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "triangular_cycle" }, { "seed": 900001, "sample": 1, "native_reward": 0.6193222054013648, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 3, "model_tokens": 306, "reward_channels": { "native": 0.6193222054013648, "efficiency": 0.5, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "triangular_cycle" }, { "seed": 900001, "sample": 2, "native_reward": 0.6199607515405053, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 268, "reward_channels": { "native": 0.6199607515405053, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "triangular_cycle" }, { "seed": 900001, "sample": 3, "native_reward": 0.6199609601615601, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 259, "reward_channels": { "native": 0.6199609601615601, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "triangular_cycle" }, { "seed": 900000, "sample": 0, "native_reward": 0.6183800407568019, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 3, "model_tokens": 423, "reward_channels": { "native": 0.6183800407568019, "efficiency": 0.5, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "competing_cycles" }, { "seed": 900000, "sample": 1, "native_reward": 0.25, "promotion_success": false, "done": false, "errors": 0, "tool_steps": 3, "model_tokens": 231, "reward_channels": { "native": 0.25, "efficiency": 0.0, "reliability": -1.0 }, "verification_level": null, "failed_gates": [ "finalized" ], "family": "competing_cycles" }, { "seed": 900000, "sample": 2, "native_reward": 0.25, "promotion_success": false, "done": false, "errors": 0, "tool_steps": 3, "model_tokens": 213, "reward_channels": { "native": 0.25, "efficiency": 0.0, "reliability": -1.0 }, "verification_level": null, "failed_gates": [ "finalized" ], "family": "competing_cycles" }, { "seed": 900000, "sample": 3, "native_reward": 0.25, "promotion_success": false, "done": false, "errors": 0, "tool_steps": 4, "model_tokens": 1176, "reward_channels": { "native": 0.25, "efficiency": 0.0, "reliability": -1.0 }, "verification_level": null, "failed_gates": [ "finalized" ], "family": "competing_cycles" }, { "seed": 900001, "sample": 0, "native_reward": 0.6189366314748597, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 696, "reward_channels": { "native": 0.6189366314748597, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "competing_cycles" }, { "seed": 900001, "sample": 1, "native_reward": 0.6212638547151527, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 3, "model_tokens": 564, "reward_channels": { "native": 0.6212638547151527, "efficiency": 0.5, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "competing_cycles" }, { "seed": 900001, "sample": 2, "native_reward": 0.25, "promotion_success": false, "done": false, "errors": 0, "tool_steps": 4, "model_tokens": 694, "reward_channels": { "native": 0.25, "efficiency": 0.0, "reliability": -1.0 }, "verification_level": null, "failed_gates": [ "finalized" ], "family": "competing_cycles" }, { "seed": 900001, "sample": 3, "native_reward": 0.6212623637910315, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 3, "model_tokens": 472, "reward_channels": { "native": 0.6212623637910315, "efficiency": 0.5, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "competing_cycles" }, { "seed": 900000, "sample": 0, "native_reward": 0.618051336943904, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 489, "reward_channels": { "native": 0.618051336943904, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "four_hop_cycle" }, { "seed": 900000, "sample": 1, "native_reward": 0.6172147132400564, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 3, "model_tokens": 337, "reward_channels": { "native": 0.6172147132400564, "efficiency": 0.5, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "four_hop_cycle" }, { "seed": 900000, "sample": 2, "native_reward": 0.6180514963968338, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 318, "reward_channels": { "native": 0.6180514963968338, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "four_hop_cycle" }, { "seed": 900000, "sample": 3, "native_reward": 0.6180516145176738, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 358, "reward_channels": { "native": 0.6180516145176738, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "four_hop_cycle" }, { "seed": 900001, "sample": 0, "native_reward": 0.6172641546083886, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 304, "reward_channels": { "native": 0.6172641546083886, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "four_hop_cycle" }, { "seed": 900001, "sample": 1, "native_reward": 0.6172660292773925, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 286, "reward_channels": { "native": 0.6172660292773925, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "four_hop_cycle" }, { "seed": 900001, "sample": 2, "native_reward": 0.6172664365064744, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 304, "reward_channels": { "native": 0.6172664365064744, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "four_hop_cycle" }, { "seed": 900001, "sample": 3, "native_reward": 0.6172659797088377, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 286, "reward_channels": { "native": 0.6172659797088377, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "four_hop_cycle" }, { "seed": 900000, "sample": 0, "native_reward": 0.8331320541859051, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 340, "reward_channels": { "native": 0.8331320541859051, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "size_sensitive_cycle" }, { "seed": 900000, "sample": 1, "native_reward": 0.8331336674653191, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 255, "reward_channels": { "native": 0.8331336674653191, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "size_sensitive_cycle" }, { "seed": 900000, "sample": 2, "native_reward": 0.8331322965015887, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 244, "reward_channels": { "native": 0.8331322965015887, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "size_sensitive_cycle" }, { "seed": 900000, "sample": 3, "native_reward": 0.8339143527755926, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 3, "model_tokens": 316, "reward_channels": { "native": 0.8339143527755926, "efficiency": 0.5, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "size_sensitive_cycle" }, { "seed": 900001, "sample": 0, "native_reward": 0.8314103908148568, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 242, "reward_channels": { "native": 0.8314103908148568, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "size_sensitive_cycle" }, { "seed": 900001, "sample": 1, "native_reward": 0.8325324301229429, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 3, "model_tokens": 317, "reward_channels": { "native": 0.8325324301229429, "efficiency": 0.5, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "size_sensitive_cycle" }, { "seed": 900001, "sample": 2, "native_reward": 0.8314105724246614, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 242, "reward_channels": { "native": 0.8314105724246614, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "size_sensitive_cycle" }, { "seed": 900001, "sample": 3, "native_reward": 0.8314116613495444, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 242, "reward_channels": { "native": 0.8314116613495444, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "size_sensitive_cycle" } ] }, { "run_id": "1789178251", "mode": "no-think", "method": "cispo", "steps": 50, "reward_points": 25, "train_mean": 0.6486137694120407, "train_last10": 0.7485604628920555, "holdout_native": 0.6190075072883925, "holdout_episodes": 40, "holdout_successes": 36, "mean_step_seconds": 163.6310592721801, "label": "CISPO", "state_at_cutoff": "complete", "step_hours": 2.272653601002501, "peak_allocated_gib": null, "peak_reserved_gib": null, "extra_refill_rounds": 0.0, "groups_refilled": 0.0, "holdout_component_means": { "native": 0.6190075072883925, "efficiency": 0.3875, "reliability": 0.8 }, "configuration": { "model": { "id": "Qwen/Qwen3-14B", "revision": "40c069824f4251a91eefaf281ebe4c544efd3e18", "loss": "cispo", "algorithm": "grpo", "no_think": true, "lora": true, "quantize": true, "lora_r": 32, "lora_alpha": 64, "use_dora": false }, "loss": "cispo", "normalization": "joint", "entropy_control": { "method": "monitor", "epsilon_low": 0.2, "epsilon_high": 0.2, "zeta_initial": 0.001, "zeta_min": 0.0001, "zeta_max": 0.05, "holdout_split": "benchmark", "holdout_seeds_per_family": 2, "num_iterations": 2 }, "dynamic_sampling": false, "sampling": { "group_size": 4, "temperature": 1.0, "max_completion_tokens": 8192, "top_entropy_quantile": 1.0, "temperature_decay_to": null }, "learning_rate": 1e-05, "turn_max_tokens": 1024 }, "source_sha256": { "metrics.jsonl": "b5f6d56e778e7b7185368956126b5c73f5ea511c07051569c6e23190d76d7d3f", "entropy_holdout_episodes.jsonl": "0ce052904992b6a3ba17a48b7ce2966199ab0f1a18530420655f693d4d4cb8a0", "resolved_training_config.json": "7fab45861fcdba479397552688d9c2b9c172cf703270245f93426187a863932b" }, "training_steps": [ { "step": 1, "loss": 0.011696956120431423, "grad_norm": 0.17578125, "learning_rate": 1e-05, "num_tokens": 28015.0, "completions/mean_length": 605.375, "completions/min_length": 278.0, "completions/max_length": 1190.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 605.375, "completions/min_terminated_length": 278.0, "completions/max_terminated_length": 1190.0, "tools/call_frequency": 4.375, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.5011405944824219, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 0.217427060008049, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.25, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.1666666716337204, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/mean": 0.33699893951416016, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/std": 0.3239978551864624, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/mean": 0.1666666716337204, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/std": 0.19245010614395142, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/mean": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/std": 1.154700517654419, "reward": 0.4794864058494568, "reward_std": 0.4545527994632721, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.0576646868721582, "cispo_clip_ratio": 0.0, "step_time": 333.7465740409999, "entropy_control/entropy": 0.03520253447176984, "entropy_control/target": 0.03520253447176984, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.004, "adapter_sync_time": 0.0, "generation_time": 230.747036, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 251.066438, "forward_backward_time": 82.57693, "rollout_overhead_time": 20.319402 }, { "step": 2, "loss": 0.011686078272759914, "grad_norm": 0.40625, "learning_rate": 9.800000000000001e-06, "entropy": 0.05742358561838046, "cispo_clip_ratio": 0.002370120317209512, "step_time": 82.6701556299995, "entropy_control/entropy": 0.03503373747956739, "entropy_control/target": 0.03520253447176984, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.008, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 82.570555, "rollout_overhead_time": 0.0, "reward": null }, { "step": 3, "loss": -0.0008278244058601558, "grad_norm": 0.2001953125, "learning_rate": 9.600000000000001e-06, "num_tokens": 52690.0, "completions/mean_length": 368.375, "completions/min_length": 266.0, "completions/max_length": 541.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 368.375, "completions/min_terminated_length": 266.0, "completions/max_terminated_length": 541.0, "tools/call_frequency": 4.25, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.5124270915985107, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 0.1862013339996338, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.25, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.15430335700511932, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 0.9258201122283936, "reward": 0.6249271035194397, "reward_std": 0.37852942943573, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.07531320885755122, "cispo_clip_ratio": 0.0, "step_time": 288.856015051, "entropy_control/entropy": 0.07677366946745573, "entropy_control/target": 0.03520253447176984, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.012, "adapter_sync_time": 0.0, "generation_time": 194.042281, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 212.102961, "forward_backward_time": 76.655759, "rollout_overhead_time": 18.06068 }, { "step": 4, "loss": -0.00040398360579274595, "grad_norm": 0.1640625, "learning_rate": 9.4e-06, "entropy": 0.07414890546351671, "cispo_clip_ratio": 0.001132071134634316, "step_time": 76.74044861300013, "entropy_control/entropy": 0.07566043482977687, "entropy_control/target": 0.03520253447176984, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.016, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 76.638718, "rollout_overhead_time": 0.0, "reward": null }, { "step": 5, "loss": 0.009572505950927734, "grad_norm": 0.1552734375, "learning_rate": 9.200000000000002e-06, "num_tokens": 77191.0, "completions/mean_length": 353.625, "completions/min_length": 267.0, "completions/max_length": 508.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 353.625, "completions/min_terminated_length": 267.0, "completions/max_terminated_length": 508.0, "tools/call_frequency": 4.25, "tools/failure_frequency": 0.029411764815449715, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.5593600273132324, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 0.1553049236536026, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.2916666865348816, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.1178511381149292, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 0.75, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 0.7071067690849304, "reward": 0.7239433526992798, "reward_std": 0.3026188313961029, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.5, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.08100612135604024, "cispo_clip_ratio": 0.0, "step_time": 244.47076902000003, "entropy_control/entropy": 0.07980892798655008, "entropy_control/target": 0.03520253447176984, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.02, "adapter_sync_time": 0.0, "generation_time": 151.916381, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 169.603811, "forward_backward_time": 74.764521, "rollout_overhead_time": 17.68743 }, { "step": 6, "loss": 0.009266003966331482, "grad_norm": 0.150390625, "learning_rate": 9e-06, "entropy": 0.08065496874041855, "cispo_clip_ratio": 0.0017237341089639813, "step_time": 74.86873130699996, "entropy_control/entropy": 0.07953020139680542, "entropy_control/target": 0.03520253447176984, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.024, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 74.766905, "rollout_overhead_time": 0.0, "reward": null }, { "step": 7, "loss": 0.0002855815691873431, "grad_norm": 0.1044921875, "learning_rate": 8.8e-06, "num_tokens": 107672.0, "completions/mean_length": 338.625, "completions/min_length": 246.0, "completions/max_length": 702.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 338.625, "completions/min_terminated_length": 246.0, "completions/max_terminated_length": 702.0, "tools/call_frequency": 3.625, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/mean": 0.25, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/mean": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/mean": -1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/mean": 0.6818172931671143, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/std": 0.28787821531295776, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/mean": 0.25, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/std": 0.1666666716337204, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/std": 1.0, "reward": 0.4221586585044861, "reward_std": 0.5136277675628662, "frac_reward_zero_std": 0.5, "frac_groups_zero_std": 0.5, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.02798425412038341, "cispo_clip_ratio": 0.0, "step_time": 289.23557070100014, "entropy_control/entropy": 0.026933905109590678, "entropy_control/target": 0.03520253447176984, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.028, "adapter_sync_time": 0.0, "generation_time": 193.720329, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 211.982536, "forward_backward_time": 77.150814, "rollout_overhead_time": 18.262208 }, { "step": 8, "loss": 0.00026433286257088184, "grad_norm": 0.1171875, "learning_rate": 8.6e-06, "entropy": 0.0280180653790012, "cispo_clip_ratio": 0.000961028621532023, "step_time": 77.24971800500043, "entropy_control/entropy": 0.02699372181632257, "entropy_control/target": 0.03520253447176984, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.032, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 77.150474, "rollout_overhead_time": 0.0, "reward": null }, { "step": 9, "loss": 0.00012150872498750687, "grad_norm": 0.00250244140625, "learning_rate": 8.400000000000001e-06, "num_tokens": 136369.0, "completions/mean_length": 299.125, "completions/min_length": 248.0, "completions/max_length": 367.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 299.125, "completions/min_terminated_length": 248.0, "completions/max_terminated_length": 367.0, "tools/call_frequency": 3.75, "tools/failure_frequency": 0.06666667014360428, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.6150286793708801, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 1.9500225789670367e-06, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.3333333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/mean": 0.25, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/mean": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/mean": -1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/std": 0.0, "reward": 0.4408476650714874, "reward_std": 0.41783374547958374, "frac_reward_zero_std": 0.5, "frac_groups_zero_std": 0.5, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.056480550207197666, "cispo_clip_ratio": 0.0, "step_time": 224.80640497600007, "entropy_control/entropy": 0.057690104543256626, "entropy_control/target": 0.03520253447176984, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.036, "adapter_sync_time": 0.0, "generation_time": 111.54092, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 132.451977, "forward_backward_time": 92.253094, "rollout_overhead_time": 20.911057 }, { "step": 10, "loss": 0.00011808892304543406, "grad_norm": 0.0024566650390625, "learning_rate": 8.2e-06, "entropy": 0.05634190049022436, "cispo_clip_ratio": 0.000728798215277493, "step_time": 92.2049963720001, "entropy_control/entropy": 0.05751369690179424, "entropy_control/target": 0.03520253447176984, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.04, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 92.10017, "rollout_overhead_time": 0.0, "reward": null }, { "step": 11, "loss": 0.02621203288435936, "grad_norm": 0.251953125, "learning_rate": 8.000000000000001e-06, "num_tokens": 165484.0, "completions/mean_length": 326.375, "completions/min_length": 257.0, "completions/max_length": 474.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 326.375, "completions/min_terminated_length": 257.0, "completions/max_terminated_length": 474.0, "tools/call_frequency": 3.75, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__native/mean": 0.6135157346725464, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__native/std": 0.000880372419487685, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__efficiency/mean": 0.375, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__efficiency/std": 0.0833333283662796, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/mean": 0.6113469004631042, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/std": 0.0008795859757810831, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/mean": 0.375, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/std": 0.0833333283662796, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/std": 0.0, "reward": 0.8311812877655029, "reward_std": 0.0048139505088329315, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.04162628436461091, "cispo_clip_ratio": 0.0, "step_time": 246.77832825500036, "entropy_control/entropy": 0.04206859201563675, "entropy_control/target": 0.03520253447176984, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.044, "adapter_sync_time": 0.0, "generation_time": 147.041174, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 165.856464, "forward_backward_time": 80.81578, "rollout_overhead_time": 18.81529 }, { "step": 12, "loss": 0.02587161585688591, "grad_norm": 0.2578125, "learning_rate": 7.800000000000002e-06, "entropy": 0.04161024373024702, "cispo_clip_ratio": 0.0007668711477890611, "step_time": 80.91245346300002, "entropy_control/entropy": 0.04211295551149884, "entropy_control/target": 0.03520253447176984, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.048, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 80.815475, "rollout_overhead_time": 0.0, "reward": null }, { "step": 13, "loss": -8.809672726783901e-05, "grad_norm": 0.001556396484375, "learning_rate": 7.600000000000001e-06, "num_tokens": 194048.0, "completions/mean_length": 281.5, "completions/min_length": 248.0, "completions/max_length": 317.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 281.5, "completions/min_terminated_length": 248.0, "completions/max_terminated_length": 317.0, "tools/call_frequency": 3.75, "tools/failure_frequency": 0.06666667014360428, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.6150178909301758, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 1.087138230104756e-06, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.3333333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/mean": 0.25, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/mean": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/mean": -1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/std": 0.0, "reward": 0.44084227085113525, "reward_std": 0.41782793402671814, "frac_reward_zero_std": 0.5, "frac_groups_zero_std": 1.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.05512358667328954, "cispo_clip_ratio": 0.0, "step_time": 224.1840341080001, "entropy_control/entropy": 0.056365220677807186, "entropy_control/target": 0.03520253447176984, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.052, "adapter_sync_time": 0.0, "generation_time": 112.100676, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 132.685049, "forward_backward_time": 91.39898, "rollout_overhead_time": 20.584373 }, { "step": 14, "loss": -8.895885548554361e-05, "grad_norm": 0.001617431640625, "learning_rate": 7.4e-06, "entropy": 0.055093816947191954, "cispo_clip_ratio": 0.0, "step_time": 91.49797025500084, "entropy_control/entropy": 0.05631458216630492, "entropy_control/target": 0.03520253447176984, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.056, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 91.398171, "rollout_overhead_time": 0.0, "reward": null }, { "step": 15, "loss": 0.0021335906349122524, "grad_norm": 0.1474609375, "learning_rate": 7.2000000000000005e-06, "num_tokens": 224565.0, "completions/mean_length": 327.125, "completions/min_length": 248.0, "completions/max_length": 466.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 327.125, "completions/min_terminated_length": 248.0, "completions/max_terminated_length": 466.0, "tools/call_frequency": 3.5, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/mean": 0.25, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/mean": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/mean": -1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/mean": 0.39442187547683716, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/std": 0.2888437807559967, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/mean": 0.0833333358168602, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/std": 0.1666666716337204, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/mean": -0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/std": 1.0, "reward": 0.17429426312446594, "reward_std": 0.35155731439590454, "frac_reward_zero_std": 0.5, "frac_groups_zero_std": 0.5, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.025739588076248765, "cispo_clip_ratio": 0.0, "step_time": 202.12111372400022, "entropy_control/entropy": 0.025104306259425127, "entropy_control/target": 0.03520253447176984, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.06, "adapter_sync_time": 0.0, "generation_time": 102.96408, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 121.557889, "forward_backward_time": 80.46287, "rollout_overhead_time": 18.593809 }, { "step": 16, "loss": 0.0022731395438313484, "grad_norm": 0.1328125, "learning_rate": 7e-06, "entropy": 0.025588660966604948, "cispo_clip_ratio": 0.0004770992381963879, "step_time": 80.55344558900015, "entropy_control/entropy": 0.02477525346074803, "entropy_control/target": 0.03520253447176984, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.064, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 80.454056, "rollout_overhead_time": 0.0, "reward": null }, { "step": 17, "loss": -0.0017667042557150126, "grad_norm": 0.201171875, "learning_rate": 6.800000000000001e-06, "num_tokens": 256711.0, "completions/mean_length": 323.25, "completions/min_length": 250.0, "completions/max_length": 412.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 323.25, "completions/min_terminated_length": 250.0, "completions/max_terminated_length": 412.0, "tools/call_frequency": 3.5, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/mean": 0.342079758644104, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/std": 0.1841595470905304, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/mean": 0.125, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/std": 0.25, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/mean": -0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/std": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/mean": 0.6121729612350464, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/std": 1.1282977538940031e-06, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/mean": 0.3333333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/std": 0.0, "reward": 0.5385847091674805, "reward_std": 0.40461617708206177, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.5, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.037226142128929496, "cispo_clip_ratio": 0.0, "step_time": 234.3388837819998, "entropy_control/entropy": 0.037284315651470536, "entropy_control/target": 0.03520253447176984, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.068, "adapter_sync_time": 0.0, "generation_time": 124.121481, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 144.41138, "forward_backward_time": 89.824994, "rollout_overhead_time": 20.289899 }, { "step": 18, "loss": -0.0016487125540152192, "grad_norm": 0.1591796875, "learning_rate": 6.600000000000001e-06, "entropy": 0.036753456108272076, "cispo_clip_ratio": 0.0012653078301809728, "step_time": 89.91730763899932, "entropy_control/entropy": 0.03685362702606528, "entropy_control/target": 0.03520253447176984, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.072, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 89.81534, "rollout_overhead_time": 0.0, "reward": null }, { "step": 19, "loss": -0.002122009638696909, "grad_norm": 0.150390625, "learning_rate": 6.4000000000000006e-06, "num_tokens": 282669.0, "completions/mean_length": 350.25, "completions/min_length": 249.0, "completions/max_length": 533.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 350.25, "completions/min_terminated_length": 249.0, "completions/max_terminated_length": 533.0, "tools/call_frequency": 4.0, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.6127638220787048, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 1.384231609335984e-06, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.3333333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/mean": 0.5394852757453918, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/std": 0.33426880836486816, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/mean": 0.1666666716337204, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/std": 0.19245010614395142, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/mean": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/std": 1.154700517654419, "reward": 0.6886245608329773, "reward_std": 0.40530484914779663, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.05153785599395633, "cispo_clip_ratio": 0.0, "step_time": 249.91120985899988, "entropy_control/entropy": 0.05100663568874365, "entropy_control/target": 0.03520253447176984, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.076, "adapter_sync_time": 0.0, "generation_time": 147.614831, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 166.628373, "forward_backward_time": 83.180292, "rollout_overhead_time": 19.013543 }, { "step": 20, "loss": -0.002266033086925745, "grad_norm": 0.162109375, "learning_rate": 6.200000000000001e-06, "entropy": 0.05132424784824252, "cispo_clip_ratio": 0.0017184022581204772, "step_time": 83.28241945400032, "entropy_control/entropy": 0.05073171256210918, "entropy_control/target": 0.03520253447176984, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.08, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 83.181347, "rollout_overhead_time": 0.0, "reward": null }, { "step": 21, "loss": -0.004757974296808243, "grad_norm": 0.1484375, "learning_rate": 6e-06, "num_tokens": 315376.0, "completions/mean_length": 385.375, "completions/min_length": 253.0, "completions/max_length": 755.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 385.375, "completions/min_terminated_length": 253.0, "completions/max_terminated_length": 755.0, "tools/call_frequency": 3.25, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/mean": 0.25, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/mean": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/mean": -1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/mean": 0.45706266164779663, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/std": 0.30470848083496094, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/mean": 0.25, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/std": 0.1666666716337204, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/std": 1.0, "reward": 0.3097813129425049, "reward_std": 0.43581223487854004, "frac_reward_zero_std": 0.5, "frac_groups_zero_std": 0.5, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 0.5, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.021211489394772798, "cispo_clip_ratio": 0.0, "step_time": 296.8628074249991, "entropy_control/entropy": 0.018510403000621463, "entropy_control/target": 0.03520253447176984, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.084, "adapter_sync_time": 0.0, "generation_time": 184.644747, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 205.48839, "forward_backward_time": 91.273117, "rollout_overhead_time": 20.843643 }, { "step": 22, "loss": -0.00441549438983202, "grad_norm": 0.142578125, "learning_rate": 5.8e-06, "entropy": 0.020908002101350576, "cispo_clip_ratio": 0.000708465842762962, "step_time": 91.50680776800073, "entropy_control/entropy": 0.01813845544415134, "entropy_control/target": 0.03520253447176984, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.088, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 91.407153, "rollout_overhead_time": 0.0, "reward": null }, { "step": 23, "loss": -0.00011591057409532368, "grad_norm": 0.002410888671875, "learning_rate": 5.600000000000001e-06, "num_tokens": 340604.0, "completions/mean_length": 264.0, "completions/min_length": 238.0, "completions/max_length": 328.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 264.0, "completions/min_terminated_length": 238.0, "completions/max_terminated_length": 328.0, "tools/call_frequency": 4.25, "tools/failure_frequency": 0.05882352963089943, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.6159777641296387, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 1.4424653045352898e-06, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.3333333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/mean": 0.831850528717041, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/std": 9.824225344345905e-07, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/mean": 0.3333333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/std": 0.0, "reward": 0.9405807852745056, "reward_std": 0.11538887768983841, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.5, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.04471729957731441, "cispo_clip_ratio": 0.0, "step_time": 200.02424894900014, "entropy_control/entropy": 0.050502290532894366, "entropy_control/target": 0.03520253447176984, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.092, "adapter_sync_time": 0.0, "generation_time": 97.802155, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 116.759051, "forward_backward_time": 83.170146, "rollout_overhead_time": 18.956896 }, { "step": 24, "loss": -0.00011218983854632825, "grad_norm": 0.0024566650390625, "learning_rate": 5.400000000000001e-06, "entropy": 0.04477547685382888, "cispo_clip_ratio": 0.0007621950935572386, "step_time": 83.2709051560023, "entropy_control/entropy": 0.05050063915829728, "entropy_control/target": 0.03520253447176984, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.096, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 83.169735, "rollout_overhead_time": 0.0, "reward": null }, { "step": 25, "loss": -0.005243090912699699, "grad_norm": 0.14453125, "learning_rate": 5.2e-06, "num_tokens": 368122.0, "completions/mean_length": 345.75, "completions/min_length": 270.0, "completions/max_length": 537.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 345.75, "completions/min_terminated_length": 270.0, "completions/max_terminated_length": 537.0, "tools/call_frequency": 4.25, "tools/failure_frequency": 0.11764705926179886, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.610882043838501, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 1.5732314295746619e-06, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.3333333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/mean": 0.5216195583343506, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/std": 0.18109402060508728, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/mean": 0.3333333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/std": 0.27216553688049316, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/std": 1.0, "reward": 0.7329174280166626, "reward_std": 0.2760407626628876, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.051616732496768236, "cispo_clip_ratio": 0.0, "step_time": 291.8929681670015, "entropy_control/entropy": 0.04924595824718981, "entropy_control/target": 0.03520253447176984, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.1, "adapter_sync_time": 0.0, "generation_time": 184.755983, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 204.647948, "forward_backward_time": 87.142992, "rollout_overhead_time": 19.891965 }, { "step": 26, "loss": -0.00530831515789032, "grad_norm": 0.14453125, "learning_rate": 5e-06, "entropy": 0.05175798456184566, "cispo_clip_ratio": 0.001137894083512947, "step_time": 87.24682994000068, "entropy_control/entropy": 0.04931038512443623, "entropy_control/target": 0.03520253447176984, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.104, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 87.145873, "rollout_overhead_time": 0.0, "reward": null }, { "step": 27, "loss": 0.014035399071872234, "grad_norm": 0.16015625, "learning_rate": 4.800000000000001e-06, "num_tokens": 397130.0, "completions/mean_length": 319.5, "completions/min_length": 264.0, "completions/max_length": 420.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 319.5, "completions/min_terminated_length": 264.0, "completions/max_terminated_length": 420.0, "tools/call_frequency": 3.625, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/mean": 0.612784743309021, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/std": 1.0756396022770787e-06, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/mean": 0.3333333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/mean": 0.8294360637664795, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/std": 0.0009538255399093032, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/mean": 0.4583333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/std": 0.0833333283662796, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/std": 0.0, "reward": 0.940902054309845, "reward_std": 0.1191929280757904, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.5, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.04855222738115117, "cispo_clip_ratio": 0.0, "step_time": 228.7263617339986, "entropy_control/entropy": 0.05083852091353351, "entropy_control/target": 0.03520253447176984, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.108, "adapter_sync_time": 0.0, "generation_time": 128.891346, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 147.820181, "forward_backward_time": 80.808708, "rollout_overhead_time": 18.928835 }, { "step": 28, "loss": 0.01435482781380415, "grad_norm": 0.1806640625, "learning_rate": 4.600000000000001e-06, "entropy": 0.048650684067979455, "cispo_clip_ratio": 0.0013880650221835822, "step_time": 80.99899548400208, "entropy_control/entropy": 0.05092927913916793, "entropy_control/target": 0.03520253447176984, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.112, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 80.901329, "rollout_overhead_time": 0.0, "reward": null }, { "step": 29, "loss": 3.172431024722755e-05, "grad_norm": 0.004302978515625, "learning_rate": 4.4e-06, "num_tokens": 425612.0, "completions/mean_length": 273.75, "completions/min_length": 251.0, "completions/max_length": 308.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 273.75, "completions/min_terminated_length": 251.0, "completions/max_terminated_length": 308.0, "tools/call_frequency": 3.625, "tools/failure_frequency": 0.03448275849223137, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.6146688461303711, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 2.791042334138183e-06, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.3333333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/mean": 0.25, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/mean": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/mean": -1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/std": 0.0, "reward": 0.4406677484512329, "reward_std": 0.41764140129089355, "frac_reward_zero_std": 0.5, "frac_groups_zero_std": 0.5, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.05797431152313948, "cispo_clip_ratio": 0.0, "step_time": 213.71142901199983, "entropy_control/entropy": 0.06040345585483202, "entropy_control/target": 0.03520253447176984, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.116, "adapter_sync_time": 0.0, "generation_time": 100.988978, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 121.743333, "forward_backward_time": 91.862674, "rollout_overhead_time": 20.754355 }, { "step": 30, "loss": 4.620966501533985e-05, "grad_norm": 0.003753662109375, "learning_rate": 4.2000000000000004e-06, "entropy": 0.05839818390086293, "cispo_clip_ratio": 0.000811688310932368, "step_time": 91.9580735519994, "entropy_control/entropy": 0.060853806023601555, "entropy_control/target": 0.03520253447176984, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.12, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 91.859478, "rollout_overhead_time": 0.0, "reward": null }, { "step": 31, "loss": 9.76059673121199e-05, "grad_norm": 0.003814697265625, "learning_rate": 4.000000000000001e-06, "num_tokens": 452381.0, "completions/mean_length": 258.625, "completions/min_length": 245.0, "completions/max_length": 283.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 258.625, "completions/min_terminated_length": 245.0, "completions/max_terminated_length": 283.0, "tools/call_frequency": 4.0, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/mean": 0.8310146331787109, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/std": 0.0009932927787303925, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/mean": 0.3333333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/std": 0.0, "reward": 1.0476813316345215, "reward_std": 0.0009933008113875985, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.5, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.01822751300642267, "cispo_clip_ratio": 0.0, "step_time": 156.64428471800056, "entropy_control/entropy": 0.01881681340022675, "entropy_control/target": 0.03520253447176984, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.124, "adapter_sync_time": 0.0, "generation_time": 75.723421, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 91.369715, "forward_backward_time": 65.168527, "rollout_overhead_time": 15.646294 }, { "step": 32, "loss": 9.737124491948634e-05, "grad_norm": 0.0034027099609375, "learning_rate": 3.8000000000000005e-06, "entropy": 0.017943263635970652, "cispo_clip_ratio": 0.0009506253118161112, "step_time": 65.24445897799887, "entropy_control/entropy": 0.018511078376458594, "entropy_control/target": 0.03520253447176984, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.128, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 65.149254, "rollout_overhead_time": 0.0, "reward": null }, { "step": 33, "loss": 0.03442241623997688, "grad_norm": 0.2080078125, "learning_rate": 3.6000000000000003e-06, "num_tokens": 478063.0, "completions/mean_length": 310.75, "completions/min_length": 237.0, "completions/max_length": 457.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 310.75, "completions/min_terminated_length": 237.0, "completions/max_terminated_length": 457.0, "tools/call_frequency": 3.875, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.6115765571594238, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 6.63729224470444e-07, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.3333333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/mean": 0.8277045488357544, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/std": 0.001518408884294331, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/mean": 0.375, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/std": 0.0833333283662796, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/std": 0.0, "reward": 0.9373488426208496, "reward_std": 0.1166982427239418, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.5, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.05297732213512063, "cispo_clip_ratio": 0.0, "step_time": 263.94113257299887, "entropy_control/entropy": 0.058404171561802644, "entropy_control/target": 0.03520253447176984, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.132, "adapter_sync_time": 0.0, "generation_time": 161.802222, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 180.909272, "forward_backward_time": 82.932263, "rollout_overhead_time": 19.10705 }, { "step": 34, "loss": 0.03507761284708977, "grad_norm": 0.3515625, "learning_rate": 3.4000000000000005e-06, "entropy": 0.05305020906962454, "cispo_clip_ratio": 0.0014242389588616788, "step_time": 83.02269980999881, "entropy_control/entropy": 0.05845580962765313, "entropy_control/target": 0.03520253447176984, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.136, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 82.923604, "rollout_overhead_time": 0.0, "reward": null }, { "step": 35, "loss": 0.0, "grad_norm": 0.0, "learning_rate": 3.2000000000000003e-06, "num_tokens": 511441.0, "completions/mean_length": 305.25, "completions/min_length": 217.0, "completions/max_length": 765.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 305.25, "completions/min_terminated_length": 217.0, "completions/max_terminated_length": 765.0, "tools/call_frequency": 3.125, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/mean": 0.25, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/mean": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/mean": -1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/std": 0.0, "reward": 0.04999999701976776, "reward_std": 0.0, "frac_reward_zero_std": 1.0, "frac_groups_zero_std": 1.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.023512376297730953, "cispo_clip_ratio": 0.0, "step_time": 228.89222622999932, "entropy_control/entropy": 0.02142604650120126, "entropy_control/target": 0.03520253447176984, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.14, "adapter_sync_time": 0.0, "generation_time": 131.530378, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 150.163666, "forward_backward_time": 78.62982, "rollout_overhead_time": 18.633287 }, { "step": 36, "loss": 0.0, "grad_norm": 0.0, "learning_rate": 3e-06, "entropy": 0.02357284730533138, "cispo_clip_ratio": 0.0, "step_time": 78.71243763899838, "entropy_control/entropy": 0.021456983127719577, "entropy_control/target": 0.03520253447176984, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.144, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 78.621809, "rollout_overhead_time": 0.0, "reward": null }, { "step": 37, "loss": -0.0072151473723351955, "grad_norm": 0.16796875, "learning_rate": 2.8000000000000003e-06, "num_tokens": 536898.0, "completions/mean_length": 298.125, "completions/min_length": 267.0, "completions/max_length": 340.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 298.125, "completions/min_terminated_length": 267.0, "completions/max_terminated_length": 340.0, "tools/call_frequency": 3.125, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.6144766807556152, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 1.1261965937592322e-06, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.3333333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__native/mean": 0.6181135177612305, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__native/std": 0.0007229947950690985, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__efficiency/mean": 0.625, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__efficiency/std": 0.0833333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__reliability/std": 0.0, "reward": 0.8402534127235413, "reward_std": 0.010251680389046669, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.5, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.07811009837314487, "cispo_clip_ratio": 0.0, "step_time": 216.00361499799965, "entropy_control/entropy": 0.0783364595707996, "entropy_control/target": 0.03520253447176984, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.148, "adapter_sync_time": 0.0, "generation_time": 114.402923, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 133.348161, "forward_backward_time": 82.557679, "rollout_overhead_time": 18.945238 }, { "step": 38, "loss": -0.007680712733417749, "grad_norm": 0.2001953125, "learning_rate": 2.6e-06, "entropy": 0.07916032942011952, "cispo_clip_ratio": 0.002167983795516193, "step_time": 82.63343949400041, "entropy_control/entropy": 0.07941339372757932, "entropy_control/target": 0.03520253447176984, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.152, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 82.539924, "rollout_overhead_time": 0.0, "reward": null }, { "step": 39, "loss": 0.012925932183861732, "grad_norm": 0.1611328125, "learning_rate": 2.4000000000000003e-06, "num_tokens": 564177.0, "completions/mean_length": 314.875, "completions/min_length": 239.0, "completions/max_length": 396.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 314.875, "completions/min_terminated_length": 239.0, "completions/max_terminated_length": 396.0, "tools/call_frequency": 3.0, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__native/mean": 0.6163939833641052, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__native/std": 0.0010152828181162477, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__efficiency/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__efficiency/std": 0.0890870913863182, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__reliability/std": 0.0, "reward": 0.8413939476013184, "reward_std": 0.005392301827669144, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.09527003375114873, "cispo_clip_ratio": 0.0, "step_time": 196.96659392399852, "entropy_control/entropy": 0.10476005763898033, "entropy_control/target": 0.03520253447176984, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.156, "adapter_sync_time": 0.0, "generation_time": 112.033927, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 128.062265, "forward_backward_time": 68.808447, "rollout_overhead_time": 16.028338 }, { "step": 40, "loss": 0.012451570481061935, "grad_norm": 0.1787109375, "learning_rate": 2.2e-06, "entropy": 0.09531991847325116, "cispo_clip_ratio": 0.0027822678384836763, "step_time": 68.90001421400211, "entropy_control/entropy": 0.10475741206601762, "entropy_control/target": 0.03520253447176984, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.16, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 68.804511, "rollout_overhead_time": 0.0, "reward": null }, { "step": 41, "loss": 0.017922621220350266, "grad_norm": 0.390625, "learning_rate": 2.0000000000000003e-06, "num_tokens": 594895.0, "completions/mean_length": 358.75, "completions/min_length": 221.0, "completions/max_length": 673.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 358.75, "completions/min_terminated_length": 221.0, "completions/max_terminated_length": 673.0, "tools/call_frequency": 3.625, "tools/failure_frequency": 0.13793103396892548, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__native/mean": 0.6144246459007263, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__native/std": 0.003429231932386756, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__efficiency/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__efficiency/std": 0.19245010614395142, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/mean": 0.3419440984725952, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/std": 0.18388816714286804, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/mean": 0.125, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/std": 0.25, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/mean": -0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/std": 1.0, "reward": 0.5438093543052673, "reward_std": 0.4090038239955902, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.025755474634934217, "cispo_clip_ratio": 0.0, "step_time": 286.34767421000106, "entropy_control/entropy": 0.02877487767957807, "entropy_control/target": 0.03520253447176984, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.164, "adapter_sync_time": 0.0, "generation_time": 185.103894, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 204.370775, "forward_backward_time": 81.876508, "rollout_overhead_time": 19.266881 }, { "step": 42, "loss": 0.018034562468528748, "grad_norm": 0.322265625, "learning_rate": 1.8000000000000001e-06, "entropy": 0.025734930008184165, "cispo_clip_ratio": 0.0010989642469212413, "step_time": 81.9797362069994, "entropy_control/entropy": 0.02869584076593105, "entropy_control/target": 0.03520253447176984, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.168, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 81.875234, "rollout_overhead_time": 0.0, "reward": null }, { "step": 43, "loss": -0.00139893451705575, "grad_norm": 0.15234375, "learning_rate": 1.6000000000000001e-06, "num_tokens": 624258.0, "completions/mean_length": 362.375, "completions/min_length": 294.0, "completions/max_length": 537.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 362.375, "completions/min_terminated_length": 294.0, "completions/max_terminated_length": 537.0, "tools/call_frequency": 2.625, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__native/mean": 0.6174289584159851, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__native/std": 1.7607767404115293e-06, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__efficiency/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__efficiency/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/mean": 0.6172435879707336, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/std": 0.0007184449350461364, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/mean": 0.625, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/std": 0.0833333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/std": 0.0, "reward": 0.8454612493515015, "reward_std": 0.004553718492388725, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.08050087280571461, "cispo_clip_ratio": 0.0, "step_time": 206.41304068699992, "entropy_control/entropy": 0.08267451758847522, "entropy_control/target": 0.03520253447176984, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.172, "adapter_sync_time": 0.0, "generation_time": 119.114559, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 135.783519, "forward_backward_time": 70.529068, "rollout_overhead_time": 16.66896 }, { "step": 44, "loss": -0.001586355036124587, "grad_norm": 0.1533203125, "learning_rate": 1.4000000000000001e-06, "entropy": 0.08013763697817922, "cispo_clip_ratio": 0.0003324467979837209, "step_time": 70.62139491300059, "entropy_control/entropy": 0.08240471169399924, "entropy_control/target": 0.03520253447176984, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.176, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 70.522778, "rollout_overhead_time": 0.0, "reward": null }, { "step": 45, "loss": -0.002852009143680334, "grad_norm": 0.2041015625, "learning_rate": 1.2000000000000002e-06, "num_tokens": 657398.0, "completions/mean_length": 468.5, "completions/min_length": 245.0, "completions/max_length": 885.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 468.5, "completions/min_terminated_length": 245.0, "completions/max_terminated_length": 885.0, "tools/call_frequency": 4.125, "tools/failure_frequency": 0.12121212482452393, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/mean": 0.43202048540115356, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/std": 0.2101791650056839, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/mean": 0.1666666716337204, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/std": 0.19245010614395142, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/mean": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/std": 1.154700517654419, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/mean": 0.6105724573135376, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/std": 0.002514779567718506, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/mean": 0.4166666865348816, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/std": 0.1666666716337204, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/std": 0.0, "reward": 0.6358797550201416, "reward_std": 0.3616826832294464, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.022907740203663707, "cispo_clip_ratio": 0.0, "step_time": 327.8458718129996, "entropy_control/entropy": 0.020962253492300336, "entropy_control/target": 0.03520253447176984, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.18, "adapter_sync_time": 0.0, "generation_time": 217.615397, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 237.898564, "forward_backward_time": 89.847216, "rollout_overhead_time": 20.283167 }, { "step": 46, "loss": -0.0024753485340625048, "grad_norm": 0.1962890625, "learning_rate": 1.0000000000000002e-06, "entropy": 0.022694767743814737, "cispo_clip_ratio": 0.000433275563409552, "step_time": 89.92750606700247, "entropy_control/entropy": 0.020681492880492976, "entropy_control/target": 0.03520253447176984, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.184, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 89.830958, "rollout_overhead_time": 0.0, "reward": null }, { "step": 47, "loss": 0.0020990422926843166, "grad_norm": 0.25, "learning_rate": 8.000000000000001e-07, "num_tokens": 686504.0, "completions/mean_length": 347.75, "completions/min_length": 238.0, "completions/max_length": 752.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 347.75, "completions/min_terminated_length": 238.0, "completions/max_terminated_length": 752.0, "tools/call_frequency": 3.875, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/mean": 0.610345184803009, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/std": 1.3778002312392346e-06, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/mean": 0.3333333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/mean": 0.6640961170196533, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/std": 0.32606759667396545, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/mean": 0.2916666865348816, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/std": 0.20971763134002686, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/std": 1.0, "reward": 0.8028456568717957, "reward_std": 0.3517226576805115, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.029519525123760104, "cispo_clip_ratio": 0.0, "step_time": 263.3829375919995, "entropy_control/entropy": 0.02818433686974614, "entropy_control/target": 0.03520253447176984, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.188, "adapter_sync_time": 0.0, "generation_time": 165.998242, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 184.559773, "forward_backward_time": 78.721451, "rollout_overhead_time": 18.561531 }, { "step": 48, "loss": 0.0020241173915565014, "grad_norm": 0.3671875, "learning_rate": 6.000000000000001e-07, "entropy": 0.029669266426935792, "cispo_clip_ratio": 0.0015464519965462387, "step_time": 78.81570005100366, "entropy_control/entropy": 0.02836964253633302, "entropy_control/target": 0.03520253447176984, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.192, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 78.712458, "rollout_overhead_time": 0.0, "reward": null }, { "step": 49, "loss": 0.048718616366386414, "grad_norm": 0.302734375, "learning_rate": 4.0000000000000003e-07, "num_tokens": 715171.0, "completions/mean_length": 288.375, "completions/min_length": 244.0, "completions/max_length": 368.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 288.375, "completions/min_terminated_length": 244.0, "completions/max_terminated_length": 368.0, "tools/call_frequency": 3.625, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/mean": 0.6128493547439575, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/std": 0.0016547844279557467, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/mean": 0.4166666865348816, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/std": 0.1666666716337204, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/mean": 0.8294295072555542, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/std": 0.0009544495260342956, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/mean": 0.375, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/std": 0.0833333283662796, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/std": 0.0, "reward": 0.9409310817718506, "reward_std": 0.11488863080739975, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.0329138221568428, "cispo_clip_ratio": 0.0, "step_time": 222.13894051999705, "entropy_control/entropy": 0.038517393573914205, "entropy_control/target": 0.03520253447176984, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.196, "adapter_sync_time": 0.0, "generation_time": 125.186869, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 143.556854, "forward_backward_time": 78.480469, "rollout_overhead_time": 18.369985 }, { "step": 50, "loss": 0.047011349350214005, "grad_norm": 0.29296875, "learning_rate": 2.0000000000000002e-07, "entropy": 0.032883982348721474, "cispo_clip_ratio": 0.0010513716551940888, "step_time": 78.57328194000002, "entropy_control/entropy": 0.03844938819068443, "entropy_control/target": 0.03520253447176984, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.2, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 78.474572, "rollout_overhead_time": 0.0, "reward": null } ], "holdout": [ { "seed": 900000, "sample": 0, "native_reward": 0.6173862425683626, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 347, "reward_channels": { "native": 0.6173862425683626, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "two_pool_cycle" }, { "seed": 900000, "sample": 1, "native_reward": 0.6173873191268587, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 224, "reward_channels": { "native": 0.6173873191268587, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "two_pool_cycle" }, { "seed": 900000, "sample": 2, "native_reward": 0.6173867850859016, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 304, "reward_channels": { "native": 0.6173867850859016, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "two_pool_cycle" }, { "seed": 900000, "sample": 3, "native_reward": 0.6173872431842025, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 383, "reward_channels": { "native": 0.6173872431842025, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "two_pool_cycle" }, { "seed": 900001, "sample": 0, "native_reward": 0.6177540296404975, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 284, "reward_channels": { "native": 0.6177540296404975, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "two_pool_cycle" }, { "seed": 900001, "sample": 1, "native_reward": 0.6177549542583101, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 277, "reward_channels": { "native": 0.6177549542583101, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "two_pool_cycle" }, { "seed": 900001, "sample": 2, "native_reward": 0.61775547305956, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 228, "reward_channels": { "native": 0.61775547305956, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "two_pool_cycle" }, { "seed": 900001, "sample": 3, "native_reward": 0.6177551720649312, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 276, "reward_channels": { "native": 0.6177551720649312, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "two_pool_cycle" }, { "seed": 900000, "sample": 0, "native_reward": 0.6188794449135906, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 3, "model_tokens": 404, "reward_channels": { "native": 0.6188794449135906, "efficiency": 0.5, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "triangular_cycle" }, { "seed": 900000, "sample": 1, "native_reward": 0.6188811748917741, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 3, "model_tokens": 364, "reward_channels": { "native": 0.6188811748917741, "efficiency": 0.5, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "triangular_cycle" }, { "seed": 900000, "sample": 2, "native_reward": 0.619441818157028, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 451, "reward_channels": { "native": 0.619441818157028, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "triangular_cycle" }, { "seed": 900000, "sample": 3, "native_reward": 0.6188807265349773, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 3, "model_tokens": 348, "reward_channels": { "native": 0.6188807265349773, "efficiency": 0.5, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "triangular_cycle" }, { "seed": 900001, "sample": 0, "native_reward": 0.6193344927032203, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 3, "model_tokens": 314, "reward_channels": { "native": 0.6193344927032203, "efficiency": 0.5, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "triangular_cycle" }, { "seed": 900001, "sample": 1, "native_reward": 0.6193346308406813, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 3, "model_tokens": 274, "reward_channels": { "native": 0.6193346308406813, "efficiency": 0.5, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "triangular_cycle" }, { "seed": 900001, "sample": 2, "native_reward": 0.6193351808036109, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 3, "model_tokens": 326, "reward_channels": { "native": 0.6193351808036109, "efficiency": 0.5, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "triangular_cycle" }, { "seed": 900001, "sample": 3, "native_reward": 0.6193356116082984, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 3, "model_tokens": 341, "reward_channels": { "native": 0.6193356116082984, "efficiency": 0.5, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "triangular_cycle" }, { "seed": 900000, "sample": 0, "native_reward": 0.25, "promotion_success": false, "done": false, "errors": 0, "tool_steps": 3, "model_tokens": 331, "reward_channels": { "native": 0.25, "efficiency": 0.0, "reliability": -1.0 }, "verification_level": null, "failed_gates": [ "finalized" ], "family": "competing_cycles" }, { "seed": 900000, "sample": 1, "native_reward": 0.25, "promotion_success": false, "done": false, "errors": 0, "tool_steps": 3, "model_tokens": 317, "reward_channels": { "native": 0.25, "efficiency": 0.0, "reliability": -1.0 }, "verification_level": null, "failed_gates": [ "finalized" ], "family": "competing_cycles" }, { "seed": 900000, "sample": 2, "native_reward": 0.25, "promotion_success": false, "done": false, "errors": 0, "tool_steps": 3, "model_tokens": 228, "reward_channels": { "native": 0.25, "efficiency": 0.0, "reliability": -1.0 }, "verification_level": null, "failed_gates": [ "finalized" ], "family": "competing_cycles" }, { "seed": 900000, "sample": 3, "native_reward": 0.6208290054711377, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 3, "model_tokens": 425, "reward_channels": { "native": 0.6208290054711377, "efficiency": 0.5, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "competing_cycles" }, { "seed": 900001, "sample": 0, "native_reward": 0.6188329288659339, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 696, "reward_channels": { "native": 0.6188329288659339, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "competing_cycles" }, { "seed": 900001, "sample": 1, "native_reward": 0.621148871236344, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 3, "model_tokens": 563, "reward_channels": { "native": 0.621148871236344, "efficiency": 0.5, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "competing_cycles" }, { "seed": 900001, "sample": 2, "native_reward": 0.6211485251098207, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 3, "model_tokens": 414, "reward_channels": { "native": 0.6211485251098207, "efficiency": 0.5, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "competing_cycles" }, { "seed": 900001, "sample": 3, "native_reward": 0.621151492362887, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 3, "model_tokens": 429, "reward_channels": { "native": 0.621151492362887, "efficiency": 0.5, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "competing_cycles" }, { "seed": 900000, "sample": 0, "native_reward": 0.6179127510059353, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 485, "reward_channels": { "native": 0.6179127510059353, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "four_hop_cycle" }, { "seed": 900000, "sample": 1, "native_reward": 0.6172749174515798, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 3, "model_tokens": 325, "reward_channels": { "native": 0.6172749174515798, "efficiency": 0.5, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "four_hop_cycle" }, { "seed": 900000, "sample": 2, "native_reward": 0.6172738687728689, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 3, "model_tokens": 360, "reward_channels": { "native": 0.6172738687728689, "efficiency": 0.5, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "four_hop_cycle" }, { "seed": 900000, "sample": 3, "native_reward": 0.617274716620154, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 3, "model_tokens": 380, "reward_channels": { "native": 0.617274716620154, "efficiency": 0.5, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "four_hop_cycle" }, { "seed": 900001, "sample": 0, "native_reward": 0.6176202006734276, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 286, "reward_channels": { "native": 0.6176202006734276, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "four_hop_cycle" }, { "seed": 900001, "sample": 1, "native_reward": 0.6176201558103611, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 286, "reward_channels": { "native": 0.6176201558103611, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "four_hop_cycle" }, { "seed": 900001, "sample": 2, "native_reward": 0.6176187713000487, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 286, "reward_channels": { "native": 0.6176187713000487, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "four_hop_cycle" }, { "seed": 900001, "sample": 3, "native_reward": 0.6176187478306736, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 286, "reward_channels": { "native": 0.6176187478306736, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "four_hop_cycle" }, { "seed": 900000, "sample": 0, "native_reward": 0.832945101641491, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 340, "reward_channels": { "native": 0.832945101641491, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "size_sensitive_cycle" }, { "seed": 900000, "sample": 1, "native_reward": 0.8329464305758074, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 255, "reward_channels": { "native": 0.8329464305758074, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "size_sensitive_cycle" }, { "seed": 900000, "sample": 2, "native_reward": 0.8329452279628777, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 244, "reward_channels": { "native": 0.8329452279628777, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "size_sensitive_cycle" }, { "seed": 900000, "sample": 3, "native_reward": 0.833727046042741, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 3, "model_tokens": 316, "reward_channels": { "native": 0.833727046042741, "efficiency": 0.5, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "size_sensitive_cycle" }, { "seed": 900001, "sample": 0, "native_reward": 0.25, "promotion_success": false, "done": false, "errors": 0, "tool_steps": 4, "model_tokens": 471, "reward_channels": { "native": 0.25, "efficiency": 0.0, "reliability": -1.0 }, "verification_level": null, "failed_gates": [ "finalized" ], "family": "size_sensitive_cycle" }, { "seed": 900001, "sample": 1, "native_reward": 0.8311179670606771, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 3, "model_tokens": 301, "reward_channels": { "native": 0.8311179670606771, "efficiency": 0.5, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "size_sensitive_cycle" }, { "seed": 900001, "sample": 2, "native_reward": 0.8311177434017124, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 3, "model_tokens": 639, "reward_channels": { "native": 0.8311177434017124, "efficiency": 0.5, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "size_sensitive_cycle" }, { "seed": 900001, "sample": 3, "native_reward": 0.8281855228974155, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 242, "reward_channels": { "native": 0.8281855228974155, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "size_sensitive_cycle" } ] }, { "run_id": "1789190592", "mode": "no-think", "method": "adapo", "steps": 50, "reward_points": 25, "train_mean": 0.6694054710865021, "train_last10": 0.7525363206863404, "holdout_native": 0.6332133409563498, "holdout_episodes": 40, "holdout_successes": 37, "mean_step_seconds": 167.80051783482006, "label": "DAPO + ADAPO", "state_at_cutoff": "complete", "step_hours": 2.330562747705834, "peak_allocated_gib": null, "peak_reserved_gib": null, "extra_refill_rounds": 0.0, "groups_refilled": 0.0, "holdout_component_means": { "native": 0.6332133409563498, "efficiency": 0.4291666666666667, "reliability": 0.85 }, "configuration": { "model": { "id": "Qwen/Qwen3-14B", "revision": "40c069824f4251a91eefaf281ebe4c544efd3e18", "loss": "dapo", "algorithm": "grpo", "no_think": true, "lora": true, "quantize": true, "lora_r": 32, "lora_alpha": 64, "use_dora": false }, "loss": "dapo", "normalization": "joint", "entropy_control": { "method": "adapo", "epsilon_low": 0.2, "epsilon_high": 0.28, "zeta_initial": 0.001, "zeta_min": 0.0001, "zeta_max": 0.05, "holdout_split": "benchmark", "holdout_seeds_per_family": 2, "num_iterations": 2 }, "dynamic_sampling": false, "sampling": { "group_size": 4, "temperature": 1.0, "max_completion_tokens": 8192, "top_entropy_quantile": 1.0, "temperature_decay_to": null }, "learning_rate": 1e-05, "turn_max_tokens": 1024 }, "source_sha256": { "metrics.jsonl": "d38b42e6891e4777a1319a4e10316751b1c7d31c9eed645f898c44db632d688a", "entropy_holdout_episodes.jsonl": "f0235f44cc7b00197bb33a5d2a5f4923bf3998d1625d4b771fd97b86ee131a20", "resolved_training_config.json": "77b2ac4e48d9a0262c1ebe23fa0f7632bd4f260e67ae5563606e24c65b6bd9d2" }, "training_steps": [ { "step": 1, "loss": 0.36477282643318176, "grad_norm": 0.1767578125, "learning_rate": 1e-05, "num_tokens": 28021.0, "completions/mean_length": 606.125, "completions/min_length": 234.0, "completions/max_length": 1190.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 606.125, "completions/min_terminated_length": 234.0, "completions/max_terminated_length": 1190.0, "tools/call_frequency": 4.375, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.5010793209075928, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 0.21738621592521667, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.25, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.1666666716337204, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/mean": 0.33697786927223206, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/std": 0.3239557445049286, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/mean": 0.1666666716337204, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/std": 0.19245010614395142, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/mean": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/std": 1.154700517654419, "reward": 0.47944527864456177, "reward_std": 0.45451125502586365, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.05266462272265926, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 335.65679702800014, "entropy_control/entropy": 0.03387283189740995, "entropy_control/target": 0.03387283189740995, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.004, "adapter_sync_time": 0.0, "generation_time": 232.156644, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 252.486975, "forward_backward_time": 83.074623, "rollout_overhead_time": 20.330332 }, { "step": 2, "loss": 0.3643573820590973, "grad_norm": 0.10595703125, "learning_rate": 9.800000000000001e-06, "entropy": 0.05260619398904964, "clip_ratio/low_mean": 0.0006929489754838869, "clip_ratio/low_min": 0.0006929489754838869, "clip_ratio/high_mean": 0.001230676076374948, "clip_ratio/high_max": 0.001230676076374948, "clip_ratio/region_mean": 0.001923625051858835, "step_time": 83.12946474800015, "entropy_control/entropy": 0.033707522106509624, "entropy_control/target": 0.03387283189740995, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.29400000000000004, "epoch": 0.008, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 83.034529, "rollout_overhead_time": 0.0, "reward": null }, { "step": 3, "loss": 0.13224834203720093, "grad_norm": 0.162109375, "learning_rate": 9.600000000000001e-06, "num_tokens": 52556.0, "completions/mean_length": 350.875, "completions/min_length": 287.0, "completions/max_length": 530.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 350.875, "completions/min_terminated_length": 287.0, "completions/max_terminated_length": 530.0, "tools/call_frequency": 4.5, "tools/failure_frequency": 0.0555555559694767, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.5033039450645447, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 0.2026338279247284, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.25, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.15430335700511932, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 0.9258201122283936, "reward": 0.6158038973808289, "reward_std": 0.3955129086971283, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.5, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.05990427825599909, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 258.1928802289997, "entropy_control/entropy": 0.05504259035661093, "entropy_control/target": 0.03387283189740995, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.29400000000000004, "entropy_control/epsilon_high_next": 0.27930000000000005, "epoch": 0.012, "adapter_sync_time": 0.0, "generation_time": 166.934, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 184.405407, "forward_backward_time": 73.694281, "rollout_overhead_time": 17.471407 }, { "step": 4, "loss": 0.13253746926784515, "grad_norm": 0.21484375, "learning_rate": 9.4e-06, "entropy": 0.059316234197467566, "clip_ratio/low_mean": 0.00048386494745500386, "clip_ratio/low_min": 0.00048386494745500386, "clip_ratio/high_mean": 0.0016727226902730763, "clip_ratio/high_max": 0.0016727226902730763, "clip_ratio/region_mean": 0.00215658763772808, "step_time": 73.78107018699984, "entropy_control/entropy": 0.054435256890841485, "entropy_control/target": 0.03387283189740995, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.27930000000000005, "entropy_control/epsilon_high_next": 0.26533500000000004, "epoch": 0.016, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 73.689618, "rollout_overhead_time": 0.0, "reward": null }, { "step": 5, "loss": 0.20966660976409912, "grad_norm": 0.11767578125, "learning_rate": 9.200000000000002e-06, "num_tokens": 77296.0, "completions/mean_length": 383.5, "completions/min_length": 288.0, "completions/max_length": 745.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 383.5, "completions/min_terminated_length": 288.0, "completions/max_terminated_length": 745.0, "tools/call_frequency": 4.25, "tools/failure_frequency": 0.0882352963089943, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.5584877729415894, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 0.15495866537094116, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.3333333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.1781741827726364, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 0.75, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 0.7071067690849304, "reward": 0.7251543998718262, "reward_std": 0.3031940460205078, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.07745937688741833, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 267.63092621400006, "entropy_control/entropy": 0.0678428053406856, "entropy_control/target": 0.03387283189740995, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.26533500000000004, "entropy_control/epsilon_high_next": 0.25206825000000005, "epoch": 0.02, "adapter_sync_time": 0.0, "generation_time": 170.326377, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 188.575523, "forward_backward_time": 78.960512, "rollout_overhead_time": 18.249146 }, { "step": 6, "loss": 0.20974095165729523, "grad_norm": 0.10400390625, "learning_rate": 9e-06, "entropy": 0.0769259239314124, "clip_ratio/low_mean": 0.0014703015913255513, "clip_ratio/low_min": 0.0014703015913255513, "clip_ratio/high_mean": 0.0011569367488846183, "clip_ratio/high_max": 0.0011569367488846183, "clip_ratio/region_mean": 0.0026272383402101696, "step_time": 79.05084699599979, "entropy_control/entropy": 0.06738231559592471, "entropy_control/target": 0.03387283189740995, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.25206825000000005, "entropy_control/epsilon_high_next": 0.23946483750000003, "epoch": 0.024, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 78.955437, "rollout_overhead_time": 0.0, "reward": null }, { "step": 7, "loss": 0.07254389673471451, "grad_norm": 0.0673828125, "learning_rate": 8.8e-06, "num_tokens": 109162.0, "completions/mean_length": 511.75, "completions/min_length": 274.0, "completions/max_length": 989.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 511.75, "completions/min_terminated_length": 274.0, "completions/max_terminated_length": 989.0, "tools/call_frequency": 4.0, "tools/failure_frequency": 0.03125, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/mean": 0.25, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/mean": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/mean": -1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/mean": 0.5379344820976257, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/std": 0.3324781060218811, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/mean": 0.1666666716337204, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/std": 0.19245010614395142, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/mean": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/std": 1.154700517654419, "reward": 0.29813387989997864, "reward_std": 0.459454745054245, "frac_reward_zero_std": 0.5, "frac_groups_zero_std": 0.5, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.022110012534540147, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 296.3191180170004, "entropy_control/entropy": 0.0188676192898828, "entropy_control/target": 0.03387283189740995, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.23946483750000003, "entropy_control/epsilon_high_next": 0.251438079375, "epoch": 0.028, "adapter_sync_time": 0.0, "generation_time": 193.208976, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 212.684095, "forward_backward_time": 83.538242, "rollout_overhead_time": 19.475119 }, { "step": 8, "loss": 0.07248738408088684, "grad_norm": 0.06982421875, "learning_rate": 8.6e-06, "entropy": 0.022285712941084057, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 83.63829929600024, "entropy_control/entropy": 0.019028148946128042, "entropy_control/target": 0.03387283189740995, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.251438079375, "entropy_control/epsilon_high_next": 0.26400998334375003, "epoch": 0.032, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 83.543357, "rollout_overhead_time": 0.0, "reward": null }, { "step": 9, "loss": 0.0005241659237071872, "grad_norm": 0.001129150390625, "learning_rate": 8.400000000000001e-06, "num_tokens": 138440.0, "completions/mean_length": 371.75, "completions/min_length": 247.0, "completions/max_length": 891.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 371.75, "completions/min_terminated_length": 247.0, "completions/max_terminated_length": 891.0, "tools/call_frequency": 3.75, "tools/failure_frequency": 0.03333333507180214, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.6098536252975464, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 1.3639786402563914e-06, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.3333333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/mean": 0.25, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/mean": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/mean": -1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/std": 0.0, "reward": 0.43826013803482056, "reward_std": 0.415067583322525, "frac_reward_zero_std": 0.5, "frac_groups_zero_std": 0.5, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.061771039851009846, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 313.30607508199955, "entropy_control/entropy": 0.053312825129257745, "entropy_control/target": 0.03387283189740995, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.26400998334375003, "entropy_control/epsilon_high_next": 0.2508094841765625, "epoch": 0.036, "adapter_sync_time": 0.0, "generation_time": 200.57349, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 221.394916, "forward_backward_time": 91.811735, "rollout_overhead_time": 20.821426 }, { "step": 10, "loss": 0.0005249655805528164, "grad_norm": 0.00107574462890625, "learning_rate": 8.2e-06, "entropy": 0.06154174020048231, "clip_ratio/low_mean": 0.0003571428533177823, "clip_ratio/low_min": 0.0003571428533177823, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0003571428533177823, "step_time": 91.90846876399974, "entropy_control/entropy": 0.05307460054667331, "entropy_control/target": 0.03387283189740995, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2508094841765625, "entropy_control/epsilon_high_next": 0.23826900996773437, "epoch": 0.04, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 91.810975, "rollout_overhead_time": 0.0, "reward": null }, { "step": 11, "loss": -0.023890800774097443, "grad_norm": 0.16796875, "learning_rate": 8.000000000000001e-06, "num_tokens": 167571.0, "completions/mean_length": 328.375, "completions/min_length": 271.0, "completions/max_length": 390.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 328.375, "completions/min_terminated_length": 271.0, "completions/max_terminated_length": 390.0, "tools/call_frequency": 3.75, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__native/mean": 0.6143152117729187, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__native/std": 2.0457509890547954e-06, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__efficiency/mean": 0.3333333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__efficiency/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/mean": 0.6130292415618896, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/std": 0.0010174305643886328, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/mean": 0.4166666865348816, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/std": 0.09622503817081451, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/std": 0.0, "reward": 0.8324222564697266, "reward_std": 0.004114731214940548, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.05718625441659242, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 225.00398189199996, "entropy_control/entropy": 0.060523702909378495, "entropy_control/target": 0.03387283189740995, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.23826900996773437, "entropy_control/epsilon_high_next": 0.22635555946934763, "epoch": 0.044, "adapter_sync_time": 0.0, "generation_time": 125.877116, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 144.59141, "forward_backward_time": 80.315309, "rollout_overhead_time": 18.714294 }, { "step": 12, "loss": -0.02428354322910309, "grad_norm": 0.1767578125, "learning_rate": 7.800000000000002e-06, "entropy": 0.057206354453228414, "clip_ratio/low_mean": 0.001610824721865356, "clip_ratio/low_min": 0.001610824721865356, "clip_ratio/high_mean": 0.0006631299620494246, "clip_ratio/high_max": 0.0006631299620494246, "clip_ratio/region_mean": 0.0022739546839147806, "step_time": 80.40808471099945, "entropy_control/entropy": 0.060516099333305325, "entropy_control/target": 0.03387283189740995, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.22635555946934763, "entropy_control/epsilon_high_next": 0.21503778149588024, "epoch": 0.048, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 80.312477, "rollout_overhead_time": 0.0, "reward": null }, { "step": 13, "loss": 0.19698549807071686, "grad_norm": 0.255859375, "learning_rate": 7.600000000000001e-06, "num_tokens": 196590.0, "completions/mean_length": 338.375, "completions/min_length": 232.0, "completions/max_length": 614.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 338.375, "completions/min_terminated_length": 232.0, "completions/max_terminated_length": 614.0, "tools/call_frequency": 3.875, "tools/failure_frequency": 0.06451612710952759, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.6121081113815308, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 2.433836016280111e-06, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.3333333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/mean": 0.1875, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/std": 0.125, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/mean": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/mean": -1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/std": 0.0, "reward": 0.40813738107681274, "reward_std": 0.4570654034614563, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 0.5, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.04371588374488056, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 286.83379497499936, "entropy_control/entropy": 0.041613701535096394, "entropy_control/target": 0.03387283189740995, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.21503778149588024, "entropy_control/epsilon_high_next": 0.20428589242108622, "epoch": 0.052, "adapter_sync_time": 0.0, "generation_time": 171.245746, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 192.416197, "forward_backward_time": 94.317683, "rollout_overhead_time": 21.170451 }, { "step": 14, "loss": 0.1972992867231369, "grad_norm": 0.20703125, "learning_rate": 7.4e-06, "entropy": 0.04388912138529122, "clip_ratio/low_mean": 0.0002035830548265949, "clip_ratio/low_min": 0.0002035830548265949, "clip_ratio/high_mean": 0.001406942872563377, "clip_ratio/high_max": 0.001406942872563377, "clip_ratio/region_mean": 0.001610525927389972, "step_time": 94.41846370099938, "entropy_control/entropy": 0.04183795891587543, "entropy_control/target": 0.03387283189740995, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.20428589242108622, "entropy_control/epsilon_high_next": 0.2, "epoch": 0.056, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 94.321893, "rollout_overhead_time": 0.0, "reward": null }, { "step": 15, "loss": 0.07450078427791595, "grad_norm": 0.228515625, "learning_rate": 7.2000000000000005e-06, "num_tokens": 227814.0, "completions/mean_length": 415.5, "completions/min_length": 250.0, "completions/max_length": 697.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 415.5, "completions/min_terminated_length": 250.0, "completions/max_terminated_length": 697.0, "tools/call_frequency": 3.625, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/mean": 0.43368300795555115, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/std": 0.2120988667011261, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/mean": 0.25, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/std": 0.28867512941360474, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/mean": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/std": 1.154700517654419, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/mean": 0.5366948246955872, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/std": 0.3310466706752777, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/mean": 0.1666666716337204, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/std": 0.19245010614395142, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/mean": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/std": 1.154700517654419, "reward": 0.49560555815696716, "reward_std": 0.48219677805900574, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.031248711398802698, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 276.89079451900034, "entropy_control/entropy": 0.03159708998326311, "entropy_control/target": 0.03387283189740995, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.21000000000000002, "epoch": 0.06, "adapter_sync_time": 0.0, "generation_time": 176.239755, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 195.411499, "forward_backward_time": 81.380941, "rollout_overhead_time": 19.171743 }, { "step": 16, "loss": 0.07463432848453522, "grad_norm": 0.1875, "learning_rate": 7e-06, "entropy": 0.031175473937764764, "clip_ratio/low_mean": 0.0006493777327705175, "clip_ratio/low_min": 0.0006493777327705175, "clip_ratio/high_mean": 0.00022007041843608022, "clip_ratio/high_max": 0.00022007041843608022, "clip_ratio/region_mean": 0.0008694481512065977, "step_time": 81.47088516999929, "entropy_control/entropy": 0.03154816893489355, "entropy_control/target": 0.03387283189740995, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.21000000000000002, "entropy_control/epsilon_high_next": 0.22050000000000003, "epoch": 0.064, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 81.37529, "rollout_overhead_time": 0.0, "reward": null }, { "step": 17, "loss": -0.06576872617006302, "grad_norm": 0.09912109375, "learning_rate": 6.800000000000001e-06, "num_tokens": 261483.0, "completions/mean_length": 513.625, "completions/min_length": 303.0, "completions/max_length": 855.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 513.625, "completions/min_terminated_length": 303.0, "completions/max_terminated_length": 855.0, "tools/call_frequency": 4.125, "tools/failure_frequency": 0.03030303120613098, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/mean": 0.3694382905960083, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/std": 0.30013903975486755, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/mean": 0.1666666716337204, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/std": 0.19245010614395142, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/mean": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/std": 1.154700517654419, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/mean": 0.6089195609092712, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/mean": 0.3333333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/std": 0.0, "reward": 0.6016789078712463, "reward_std": 0.42297154664993286, "frac_reward_zero_std": 0.5, "frac_groups_zero_std": 0.5, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 0.5, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.02156463696155697, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 321.19880535000084, "entropy_control/entropy": 0.021361945963121964, "entropy_control/target": 0.03387283189740995, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.22050000000000003, "entropy_control/epsilon_high_next": 0.23152500000000004, "epoch": 0.068, "adapter_sync_time": 0.0, "generation_time": 215.755634, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 235.728368, "forward_backward_time": 85.372124, "rollout_overhead_time": 19.972734 }, { "step": 18, "loss": -0.06580959260463715, "grad_norm": 0.08447265625, "learning_rate": 6.600000000000001e-06, "entropy": 0.021777975372970104, "clip_ratio/low_mean": 0.00039802763785701245, "clip_ratio/low_min": 0.00039802763785701245, "clip_ratio/high_mean": 0.0006521098839584738, "clip_ratio/high_max": 0.0006521098839584738, "clip_ratio/region_mean": 0.0010501375218154863, "step_time": 85.4535766920003, "entropy_control/entropy": 0.02146843573961085, "entropy_control/target": 0.03387283189740995, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.23152500000000004, "entropy_control/epsilon_high_next": 0.24310125000000005, "epoch": 0.072, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 85.357488, "rollout_overhead_time": 0.0, "reward": null }, { "step": 19, "loss": 0.11337518692016602, "grad_norm": 0.169921875, "learning_rate": 6.4000000000000006e-06, "num_tokens": 287111.0, "completions/mean_length": 309.0, "completions/min_length": 240.0, "completions/max_length": 446.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 309.0, "completions/min_terminated_length": 240.0, "completions/max_terminated_length": 446.0, "tools/call_frequency": 4.0, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.6139358282089233, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 1.2258865353942383e-06, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.3333333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/mean": 0.6851068735122681, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/std": 0.2900712490081787, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/mean": 0.25, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/std": 0.1666666716337204, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/std": 1.0, "reward": 0.8141046762466431, "reward_std": 0.3267587125301361, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.04121474042767659, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 238.71769372400013, "entropy_control/entropy": 0.04206128059848401, "entropy_control/target": 0.03387283189740995, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.24310125000000005, "entropy_control/epsilon_high_next": 0.23094618750000004, "epoch": 0.076, "adapter_sync_time": 0.0, "generation_time": 135.987196, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 155.115913, "forward_backward_time": 83.503476, "rollout_overhead_time": 19.128717 }, { "step": 20, "loss": 0.11320944130420685, "grad_norm": 0.091796875, "learning_rate": 6.200000000000001e-06, "entropy": 0.041412135324208066, "clip_ratio/low_mean": 0.0007083512609824538, "clip_ratio/low_min": 0.0007083512609824538, "clip_ratio/high_mean": 0.0005208333604969084, "clip_ratio/high_max": 0.0005208333604969084, "clip_ratio/region_mean": 0.0012291846214793622, "step_time": 83.59727125000063, "entropy_control/entropy": 0.04231377992056128, "entropy_control/target": 0.03387283189740995, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.23094618750000004, "entropy_control/epsilon_high_next": 0.21939887812500003, "epoch": 0.08, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 83.498482, "rollout_overhead_time": 0.0, "reward": null }, { "step": 21, "loss": -9.638885967433453e-05, "grad_norm": 0.005279541015625, "learning_rate": 6e-06, "num_tokens": 319547.0, "completions/mean_length": 351.5, "completions/min_length": 250.0, "completions/max_length": 767.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 351.5, "completions/min_terminated_length": 250.0, "completions/max_terminated_length": 767.0, "tools/call_frequency": 3.625, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/mean": 0.25, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/mean": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/mean": -1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/mean": 0.610081672668457, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/std": 3.355729177201283e-06, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/mean": 0.3333333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/std": 0.0, "reward": 0.4383741617202759, "reward_std": 0.4151894748210907, "frac_reward_zero_std": 0.5, "frac_groups_zero_std": 0.5, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.023630270501598716, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 277.9648794860004, "entropy_control/entropy": 0.021144500537867875, "entropy_control/target": 0.03387283189740995, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.21939887812500003, "entropy_control/epsilon_high_next": 0.23036882203125003, "epoch": 0.084, "adapter_sync_time": 0.0, "generation_time": 168.310259, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 188.452262, "forward_backward_time": 89.411288, "rollout_overhead_time": 20.142004 }, { "step": 22, "loss": -8.972384966909885e-05, "grad_norm": 0.004852294921875, "learning_rate": 5.8e-06, "entropy": 0.023387800087220967, "clip_ratio/low_mean": 0.00038580247201025486, "clip_ratio/low_min": 0.00038580247201025486, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.00038580247201025486, "step_time": 89.51071626000066, "entropy_control/entropy": 0.02090090973537038, "entropy_control/target": 0.03387283189740995, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.23036882203125003, "entropy_control/epsilon_high_next": 0.24188726313281253, "epoch": 0.088, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 89.412838, "rollout_overhead_time": 0.0, "reward": null }, { "step": 23, "loss": 0.00031751743517816067, "grad_norm": 0.00173187255859375, "learning_rate": 5.600000000000001e-06, "num_tokens": 345012.0, "completions/mean_length": 293.625, "completions/min_length": 238.0, "completions/max_length": 398.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 293.625, "completions/min_terminated_length": 238.0, "completions/max_terminated_length": 398.0, "tools/call_frequency": 4.0, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.6155115365982056, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 1.0739868230302818e-06, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.3333333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/mean": 0.8313854932785034, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/std": 7.356607625297329e-07, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/mean": 0.3333333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/std": 0.0, "reward": 0.9401151537895203, "reward_std": 0.11538950353860855, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 1.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.053719160379841924, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 210.8280203840004, "entropy_control/entropy": 0.05752208560259863, "entropy_control/target": 0.03387283189740995, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.24188726313281253, "entropy_control/epsilon_high_next": 0.2297928999761719, "epoch": 0.092, "adapter_sync_time": 0.0, "generation_time": 108.128857, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 127.256432, "forward_backward_time": 83.474927, "rollout_overhead_time": 19.127574 }, { "step": 24, "loss": 0.00032108870800584555, "grad_norm": 0.00174713134765625, "learning_rate": 5.400000000000001e-06, "entropy": 0.05370692379074171, "clip_ratio/low_mean": 0.0004208754107821733, "clip_ratio/low_min": 0.0004208754107821733, "clip_ratio/high_mean": 0.0008392804593313485, "clip_ratio/high_max": 0.0008392804593313485, "clip_ratio/region_mean": 0.0012601558701135218, "step_time": 83.5664421800002, "entropy_control/entropy": 0.057481813048935224, "entropy_control/target": 0.03387283189740995, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2297928999761719, "entropy_control/epsilon_high_next": 0.21830325497736328, "epoch": 0.096, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 83.46936, "rollout_overhead_time": 0.0, "reward": null }, { "step": 25, "loss": -0.04335213452577591, "grad_norm": 0.349609375, "learning_rate": 5.2e-06, "num_tokens": 372514.0, "completions/mean_length": 343.75, "completions/min_length": 286.0, "completions/max_length": 441.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 343.75, "completions/min_terminated_length": 286.0, "completions/max_terminated_length": 441.0, "tools/call_frequency": 3.5, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.6124046444892883, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 0.0008573840605095029, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.375, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.0833333283662796, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/mean": 0.6130148768424988, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/std": 0.0017914436757564545, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/mean": 0.4583333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/std": 0.15957120060920715, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/std": 0.0, "reward": 0.833543062210083, "reward_std": 0.007631031796336174, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.08302592276595533, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 273.73329322100017, "entropy_control/entropy": 0.0872263392758156, "entropy_control/target": 0.03387283189740995, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.21830325497736328, "entropy_control/epsilon_high_next": 0.2073880922284951, "epoch": 0.1, "adapter_sync_time": 0.0, "generation_time": 165.887995, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 185.942353, "forward_backward_time": 87.690225, "rollout_overhead_time": 20.054358 }, { "step": 26, "loss": -0.043718062341213226, "grad_norm": 0.26171875, "learning_rate": 5e-06, "entropy": 0.08268096530809999, "clip_ratio/low_mean": 0.0004139072843827307, "clip_ratio/low_min": 0.0004139072843827307, "clip_ratio/high_mean": 0.0008907363517209888, "clip_ratio/high_max": 0.0008907363517209888, "clip_ratio/region_mean": 0.0013046436361037195, "step_time": 87.7848287859988, "entropy_control/entropy": 0.08683127245284247, "entropy_control/target": 0.03387283189740995, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2073880922284951, "entropy_control/epsilon_high_next": 0.2, "epoch": 0.104, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 87.686905, "rollout_overhead_time": 0.0, "reward": null }, { "step": 27, "loss": 0.1411801129579544, "grad_norm": 0.2451171875, "learning_rate": 4.800000000000001e-06, "num_tokens": 401848.0, "completions/mean_length": 360.25, "completions/min_length": 266.0, "completions/max_length": 632.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 360.25, "completions/min_terminated_length": 266.0, "completions/max_terminated_length": 632.0, "tools/call_frequency": 3.25, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/mean": 0.5240532159805298, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/std": 0.1827021688222885, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/std": 0.3333333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/std": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/mean": 0.8281166553497314, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/std": 0.0006886234623380005, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/mean": 0.375, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/std": 0.0833333283662796, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/std": 0.0, "reward": 0.8479599356651306, "reward_std": 0.3370180130004883, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.05922881024889648, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 239.18355853600042, "entropy_control/entropy": 0.05647923815915003, "entropy_control/target": 0.03387283189740995, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "epoch": 0.108, "adapter_sync_time": 0.0, "generation_time": 140.931675, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 159.455293, "forward_backward_time": 79.627677, "rollout_overhead_time": 18.523617 }, { "step": 28, "loss": 0.1409509927034378, "grad_norm": 0.2353515625, "learning_rate": 4.600000000000001e-06, "entropy": 0.05933056212961674, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0022114544408395886, "clip_ratio/high_max": 0.0022114544408395886, "clip_ratio/region_mean": 0.0022114544408395886, "step_time": 79.72936577699966, "entropy_control/entropy": 0.056529543312585945, "entropy_control/target": 0.03387283189740995, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "epoch": 0.112, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 79.630739, "rollout_overhead_time": 0.0, "reward": null }, { "step": 29, "loss": 0.023490160703659058, "grad_norm": 0.2265625, "learning_rate": 4.4e-06, "num_tokens": 430454.0, "completions/mean_length": 289.25, "completions/min_length": 257.0, "completions/max_length": 351.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 289.25, "completions/min_terminated_length": 257.0, "completions/max_terminated_length": 351.0, "tools/call_frequency": 3.375, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.6133197546005249, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 0.00048731043352745473, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.375, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.0833333283662796, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/mean": 0.25, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/mean": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/mean": -1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/std": 0.0, "reward": 0.44103488326072693, "reward_std": 0.4180449843406677, "frac_reward_zero_std": 0.5, "frac_groups_zero_std": 0.5, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.051102850353345275, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 235.61832257200058, "entropy_control/entropy": 0.05256730198180151, "entropy_control/target": 0.03387283189740995, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "epoch": 0.116, "adapter_sync_time": 0.0, "generation_time": 122.613903, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 143.461784, "forward_backward_time": 92.058724, "rollout_overhead_time": 20.847881 }, { "step": 30, "loss": 0.024162694811820984, "grad_norm": 0.169921875, "learning_rate": 4.2000000000000004e-06, "entropy": 0.051004725391976535, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.00044642857392318547, "clip_ratio/high_max": 0.00044642857392318547, "clip_ratio/region_mean": 0.00044642857392318547, "step_time": 92.13939696100078, "entropy_control/entropy": 0.052498454667273806, "entropy_control/target": 0.03387283189740995, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "epoch": 0.12, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 92.044089, "rollout_overhead_time": 0.0, "reward": null }, { "step": 31, "loss": -0.05649399757385254, "grad_norm": 0.173828125, "learning_rate": 4.000000000000001e-06, "num_tokens": 457858.0, "completions/mean_length": 338.0, "completions/min_length": 258.0, "completions/max_length": 508.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 338.0, "completions/min_terminated_length": 258.0, "completions/max_terminated_length": 508.0, "tools/call_frequency": 3.5, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/mean": 0.8280452489852905, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/std": 0.002457990776747465, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/mean": 0.4166666865348816, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/std": 0.12598817050457, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/std": 0.0, "reward": 1.0488786697387695, "reward_std": 0.008642906323075294, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.04904402093961835, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 233.9038784129989, "entropy_control/entropy": 0.05327178426842861, "entropy_control/target": 0.03387283189740995, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "epoch": 0.124, "adapter_sync_time": 0.0, "generation_time": 145.819175, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 162.543491, "forward_backward_time": 71.259922, "rollout_overhead_time": 16.724316 }, { "step": 32, "loss": -0.056584350764751434, "grad_norm": 0.10498046875, "learning_rate": 3.8000000000000005e-06, "entropy": 0.048565386212430894, "clip_ratio/low_mean": 0.00048449612222611904, "clip_ratio/low_min": 0.00048449612222611904, "clip_ratio/high_mean": 0.003227768116630614, "clip_ratio/high_max": 0.003227768116630614, "clip_ratio/region_mean": 0.003712264238856733, "step_time": 71.2344675839995, "entropy_control/entropy": 0.052824577840394786, "entropy_control/target": 0.03387283189740995, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "epoch": 0.128, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 71.134322, "rollout_overhead_time": 0.0, "reward": null }, { "step": 33, "loss": -0.008098140358924866, "grad_norm": 0.2294921875, "learning_rate": 3.6000000000000003e-06, "num_tokens": 483202.0, "completions/mean_length": 268.5, "completions/min_length": 237.0, "completions/max_length": 296.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 268.5, "completions/min_terminated_length": 237.0, "completions/max_terminated_length": 296.0, "tools/call_frequency": 3.875, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.614773690700531, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 0.000650485570076853, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.375, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.0833333283662796, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/mean": 0.8298177123069763, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/std": 1.8589261117085698e-06, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/mean": 0.3333333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/std": 0.0, "reward": 0.9400039911270142, "reward_std": 0.11387597024440765, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.046624069567769766, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 214.05489298000066, "entropy_control/entropy": 0.04925495917569158, "entropy_control/target": 0.03387283189740995, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "epoch": 0.132, "adapter_sync_time": 0.0, "generation_time": 111.961046, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 130.917313, "forward_backward_time": 83.037984, "rollout_overhead_time": 18.956267 }, { "step": 34, "loss": -0.008900672197341919, "grad_norm": 0.126953125, "learning_rate": 3.4000000000000005e-06, "entropy": 0.04707741632591933, "clip_ratio/low_mean": 0.0025366564514115453, "clip_ratio/low_min": 0.0025366564514115453, "clip_ratio/high_mean": 0.001825004001148045, "clip_ratio/high_max": 0.001825004001148045, "clip_ratio/region_mean": 0.00436166045255959, "step_time": 83.12838843599911, "entropy_control/entropy": 0.04968981134349904, "entropy_control/target": 0.03387283189740995, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "epoch": 0.136, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 83.027857, "rollout_overhead_time": 0.0, "reward": null }, { "step": 35, "loss": 0.09751167893409729, "grad_norm": 0.173828125, "learning_rate": 3.2000000000000003e-06, "num_tokens": 517228.0, "completions/mean_length": 386.25, "completions/min_length": 217.0, "completions/max_length": 788.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 386.25, "completions/min_terminated_length": 217.0, "completions/max_terminated_length": 788.0, "tools/call_frequency": 3.25, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/mean": 0.34177881479263306, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/std": 0.16994133591651917, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/mean": 0.125, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/std": 0.2314550280570984, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/mean": -0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/std": 0.9258201122283936, "reward": 0.24802881479263306, "reward_std": 0.3666780889034271, "frac_reward_zero_std": 0.5, "frac_groups_zero_std": 0.5, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.039438956999219954, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 268.3658710079999, "entropy_control/entropy": 0.03591177573991314, "entropy_control/target": 0.03387283189740995, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "epoch": 0.14, "adapter_sync_time": 0.0, "generation_time": 168.873184, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 187.784091, "forward_backward_time": 80.48275, "rollout_overhead_time": 18.910908 }, { "step": 36, "loss": 0.09772133827209473, "grad_norm": 0.1298828125, "learning_rate": 3e-06, "entropy": 0.0395521956961602, "clip_ratio/low_mean": 0.0005760368658229709, "clip_ratio/low_min": 0.0005760368658229709, "clip_ratio/high_mean": 0.0018618452886585146, "clip_ratio/high_max": 0.0018618452886585146, "clip_ratio/region_mean": 0.0024378821544814855, "step_time": 80.57577404099902, "entropy_control/entropy": 0.03620550101490342, "entropy_control/target": 0.03387283189740995, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "epoch": 0.144, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 80.48127, "rollout_overhead_time": 0.0, "reward": null }, { "step": 37, "loss": 0.026308799162507057, "grad_norm": 0.171875, "learning_rate": 2.8000000000000003e-06, "num_tokens": 542745.0, "completions/mean_length": 305.625, "completions/min_length": 280.0, "completions/max_length": 351.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 305.625, "completions/min_terminated_length": 280.0, "completions/max_terminated_length": 351.0, "tools/call_frequency": 3.125, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.6148286461830139, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 6.068756306376599e-07, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.3333333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__native/mean": 0.6183834075927734, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__native/std": 0.0005939810653217137, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__efficiency/mean": 0.625, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__efficiency/std": 0.0833333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__reliability/std": 0.0, "reward": 0.8405643701553345, "reward_std": 0.010183854959905148, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.5, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.09458077792078257, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 212.81755582700043, "entropy_control/entropy": 0.09513629756365732, "entropy_control/target": 0.03387283189740995, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "epoch": 0.148, "adapter_sync_time": 0.0, "generation_time": 111.815453, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 130.571178, "forward_backward_time": 82.146467, "rollout_overhead_time": 18.755725 }, { "step": 38, "loss": 0.02639860473573208, "grad_norm": 0.1689453125, "learning_rate": 2.6e-06, "entropy": 0.09390611946582794, "clip_ratio/low_mean": 0.0007961783558130264, "clip_ratio/low_min": 0.0007961783558130264, "clip_ratio/high_mean": 0.0004208754107821733, "clip_ratio/high_max": 0.0004208754107821733, "clip_ratio/region_mean": 0.0012170537665951997, "step_time": 82.24647135800114, "entropy_control/entropy": 0.09445156404649978, "entropy_control/target": 0.03387283189740995, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "epoch": 0.152, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 82.150283, "rollout_overhead_time": 0.0, "reward": null }, { "step": 39, "loss": -0.047864072024822235, "grad_norm": 0.162109375, "learning_rate": 2.4000000000000003e-06, "num_tokens": 570339.0, "completions/mean_length": 354.25, "completions/min_length": 291.0, "completions/max_length": 474.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 354.25, "completions/min_terminated_length": 291.0, "completions/max_terminated_length": 474.0, "tools/call_frequency": 3.25, "tools/failure_frequency": 0.1538461595773697, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__native/mean": 0.6177661418914795, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__native/std": 0.0005008924636058509, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__efficiency/mean": 0.5416666865348816, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__efficiency/std": 0.07715168595314026, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__reliability/std": 0.0, "reward": 0.8448494672775269, "reward_std": 0.004182360600680113, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.1054941964102909, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 189.7340894450008, "entropy_control/entropy": 0.10353898318414546, "entropy_control/target": 0.03387283189740995, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "epoch": 0.156, "adapter_sync_time": 0.0, "generation_time": 113.950666, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 128.642906, "forward_backward_time": 60.994263, "rollout_overhead_time": 14.69224 }, { "step": 40, "loss": -0.047934480011463165, "grad_norm": 0.1220703125, "learning_rate": 2.2e-06, "entropy": 0.10487919801380485, "clip_ratio/low_mean": 0.0012035470572300255, "clip_ratio/low_min": 0.0012035470572300255, "clip_ratio/high_mean": 0.0006148366665001959, "clip_ratio/high_max": 0.0006148366665001959, "clip_ratio/region_mean": 0.0018183837237302214, "step_time": 61.10121806999996, "entropy_control/entropy": 0.10289937086235329, "entropy_control/target": 0.03387283189740995, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "epoch": 0.16, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 61.005538, "rollout_overhead_time": 0.0, "reward": null }, { "step": 41, "loss": -0.18966902792453766, "grad_norm": 0.2578125, "learning_rate": 2.0000000000000003e-06, "num_tokens": 600809.0, "completions/mean_length": 327.75, "completions/min_length": 240.0, "completions/max_length": 540.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 327.75, "completions/min_terminated_length": 240.0, "completions/max_terminated_length": 540.0, "tools/call_frequency": 2.75, "tools/failure_frequency": 0.04545454680919647, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__native/mean": 0.6165226697921753, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__native/std": 0.0014098688261583447, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__efficiency/mean": 0.625, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__efficiency/std": 0.0833333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/mean": 0.43387043476104736, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/std": 0.21231532096862793, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/mean": 0.25, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/std": 0.28867512941360474, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/mean": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/std": 1.154700517654419, "reward": 0.6470715999603271, "reward_std": 0.36854520440101624, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.06314902822487056, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 231.81965113800015, "entropy_control/entropy": 0.06058914605878699, "entropy_control/target": 0.03387283189740995, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "epoch": 0.164, "adapter_sync_time": 0.0, "generation_time": 129.366009, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 148.594988, "forward_backward_time": 83.126482, "rollout_overhead_time": 19.228979 }, { "step": 42, "loss": -0.19033004343509674, "grad_norm": 0.294921875, "learning_rate": 1.8000000000000001e-06, "entropy": 0.06259556184522808, "clip_ratio/low_mean": 0.0010129593429155648, "clip_ratio/low_min": 0.0010129593429155648, "clip_ratio/high_mean": 0.0037230374291539192, "clip_ratio/high_max": 0.0037230374291539192, "clip_ratio/region_mean": 0.004735996772069484, "step_time": 83.21017399399989, "entropy_control/entropy": 0.05999337480007122, "entropy_control/target": 0.03387283189740995, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "epoch": 0.168, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 83.113883, "rollout_overhead_time": 0.0, "reward": null }, { "step": 43, "loss": 0.11023291200399399, "grad_norm": 0.203125, "learning_rate": 1.6000000000000001e-06, "num_tokens": 630059.0, "completions/mean_length": 348.25, "completions/min_length": 269.0, "completions/max_length": 563.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 348.25, "completions/min_terminated_length": 269.0, "completions/max_terminated_length": 563.0, "tools/call_frequency": 2.375, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__native/mean": 0.6185731887817383, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__native/std": 0.0006251955055631697, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__efficiency/mean": 0.625, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__efficiency/std": 0.0833333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/mean": 0.6169957518577576, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/std": 0.0007217058446258307, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/mean": 0.5833333730697632, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/std": 0.09622505307197571, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/std": 0.0, "reward": 0.8479927778244019, "reward_std": 0.00517592579126358, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.07912383368238807, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 193.35635953499786, "entropy_control/entropy": 0.08572732927896128, "entropy_control/target": 0.03387283189740995, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "epoch": 0.172, "adapter_sync_time": 0.0, "generation_time": 103.351117, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 120.433264, "forward_backward_time": 72.825862, "rollout_overhead_time": 17.082148 }, { "step": 44, "loss": 0.1103602945804596, "grad_norm": 0.181640625, "learning_rate": 1.4000000000000001e-06, "entropy": 0.07986166514456272, "clip_ratio/low_mean": 0.0005741375352954492, "clip_ratio/low_min": 0.0005741375352954492, "clip_ratio/high_mean": 0.0021349611051846296, "clip_ratio/high_max": 0.0021349611051846296, "clip_ratio/region_mean": 0.0027090986404800788, "step_time": 72.92475570000352, "entropy_control/entropy": 0.08636384081413018, "entropy_control/target": 0.03387283189740995, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "epoch": 0.176, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 72.825618, "rollout_overhead_time": 0.0, "reward": null }, { "step": 45, "loss": -0.04502106085419655, "grad_norm": 0.08056640625, "learning_rate": 1.2000000000000002e-06, "num_tokens": 663276.0, "completions/mean_length": 478.125, "completions/min_length": 229.0, "completions/max_length": 786.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 478.125, "completions/min_terminated_length": 229.0, "completions/max_terminated_length": 786.0, "tools/call_frequency": 2.875, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/mean": 0.25, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/mean": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/mean": -1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/mean": 0.46013420820236206, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/std": 0.3067595362663269, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/mean": 0.4166666865348816, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/std": 0.2886751592159271, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/std": 1.0, "reward": 0.3154837489128113, "reward_std": 0.4435495436191559, "frac_reward_zero_std": 0.5, "frac_groups_zero_std": 0.5, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 0.5, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.03741473110858351, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 270.28836933099774, "entropy_control/entropy": 0.02958827659738704, "entropy_control/target": 0.03387283189740995, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.21000000000000002, "epoch": 0.18, "adapter_sync_time": 0.0, "generation_time": 172.152937, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 190.757411, "forward_backward_time": 79.431845, "rollout_overhead_time": 18.604474 }, { "step": 46, "loss": -0.04480624198913574, "grad_norm": 0.0732421875, "learning_rate": 1.0000000000000002e-06, "entropy": 0.037528570857830346, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0007730482902843505, "clip_ratio/high_max": 0.0007730482902843505, "clip_ratio/region_mean": 0.0007730482902843505, "step_time": 79.52148914399913, "entropy_control/entropy": 0.02982904979034818, "entropy_control/target": 0.03387283189740995, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.21000000000000002, "entropy_control/epsilon_high_next": 0.22050000000000003, "epoch": 0.184, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 79.425489, "rollout_overhead_time": 0.0, "reward": null }, { "step": 47, "loss": -0.1024111956357956, "grad_norm": 0.40234375, "learning_rate": 8.000000000000001e-07, "num_tokens": 692331.0, "completions/mean_length": 341.375, "completions/min_length": 247.0, "completions/max_length": 509.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 341.375, "completions/min_terminated_length": 247.0, "completions/max_terminated_length": 509.0, "tools/call_frequency": 3.25, "tools/failure_frequency": 0.1538461595773697, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/mean": 0.6136386394500732, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/std": 0.0012079378357157111, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/mean": 0.4166666865348816, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/std": 0.1666666716337204, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/mean": 0.8311635255813599, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/std": 2.2774902390665375e-06, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/mean": 0.6666666865348816, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/std": 0.0, "reward": 0.9494844675064087, "reward_std": 0.12311206012964249, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.057108914596028626, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 231.68583597400357, "entropy_control/entropy": 0.056636118374944464, "entropy_control/target": 0.03387283189740995, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.22050000000000003, "entropy_control/epsilon_high_next": 0.20947500000000002, "epoch": 0.188, "adapter_sync_time": 0.0, "generation_time": 133.876683, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 152.548086, "forward_backward_time": 79.03583, "rollout_overhead_time": 18.671403 }, { "step": 48, "loss": -0.10202079266309738, "grad_norm": 0.1318359375, "learning_rate": 6.000000000000001e-07, "entropy": 0.05646660481579602, "clip_ratio/low_mean": 0.0011839762446470559, "clip_ratio/low_min": 0.0011839762446470559, "clip_ratio/high_mean": 0.0007812500116415322, "clip_ratio/high_max": 0.0007812500116415322, "clip_ratio/region_mean": 0.001965226256288588, "step_time": 79.13202099600312, "entropy_control/entropy": 0.05593077138747248, "entropy_control/target": 0.03387283189740995, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.20947500000000002, "entropy_control/epsilon_high_next": 0.2, "epoch": 0.192, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 79.032193, "rollout_overhead_time": 0.0, "reward": null }, { "step": 49, "loss": 0.15868572890758514, "grad_norm": 0.126953125, "learning_rate": 4.0000000000000003e-07, "num_tokens": 721455.0, "completions/mean_length": 345.5, "completions/min_length": 244.0, "completions/max_length": 638.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 345.5, "completions/min_terminated_length": 244.0, "completions/max_terminated_length": 638.0, "tools/call_frequency": 3.75, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/mean": 0.5223459601402283, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/std": 0.18156979978084564, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/mean": 0.3750000298023224, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/std": 0.28463754057884216, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/std": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/mean": 0.8282478451728821, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/std": 1.0963582326439791e-06, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/mean": 0.3333333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/std": 0.0, "reward": 0.8430052995681763, "reward_std": 0.33641016483306885, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.5, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.040065973764285445, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 244.29561989499962, "entropy_control/entropy": 0.043171080780193195, "entropy_control/target": 0.03387283189740995, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "epoch": 0.196, "adapter_sync_time": 0.0, "generation_time": 145.95122, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 164.472738, "forward_backward_time": 79.725501, "rollout_overhead_time": 18.521519 }, { "step": 50, "loss": 0.15876910090446472, "grad_norm": 0.10205078125, "learning_rate": 2.0000000000000002e-07, "entropy": 0.039461669395677745, "clip_ratio/low_mean": 0.0011050452158087865, "clip_ratio/low_min": 0.0011050452158087865, "clip_ratio/high_mean": 0.002213738247519359, "clip_ratio/high_max": 0.002213738247519359, "clip_ratio/region_mean": 0.0033187834633281454, "step_time": 79.96288616399761, "entropy_control/entropy": 0.042510970287708394, "entropy_control/target": 0.03387283189740995, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "epoch": 0.2, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 79.865639, "rollout_overhead_time": 0.0, "reward": null } ], "holdout": [ { "seed": 900000, "sample": 0, "native_reward": 0.6156027462878353, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 335, "reward_channels": { "native": 0.6156027462878353, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "two_pool_cycle" }, { "seed": 900000, "sample": 1, "native_reward": 0.6156035667268978, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 404, "reward_channels": { "native": 0.6156035667268978, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "two_pool_cycle" }, { "seed": 900000, "sample": 2, "native_reward": 0.6156035416276204, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 296, "reward_channels": { "native": 0.6156035416276204, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "two_pool_cycle" }, { "seed": 900000, "sample": 3, "native_reward": 0.6164455951305502, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 3, "model_tokens": 304, "reward_channels": { "native": 0.6164455951305502, "efficiency": 0.5, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "two_pool_cycle" }, { "seed": 900001, "sample": 0, "native_reward": 0.6176681372465913, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 285, "reward_channels": { "native": 0.6176681372465913, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "two_pool_cycle" }, { "seed": 900001, "sample": 1, "native_reward": 0.6176674544183491, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 285, "reward_channels": { "native": 0.6176674544183491, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "two_pool_cycle" }, { "seed": 900001, "sample": 2, "native_reward": 0.6176677963965522, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 316, "reward_channels": { "native": 0.6176677963965522, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "two_pool_cycle" }, { "seed": 900001, "sample": 3, "native_reward": 0.6176667988512006, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 314, "reward_channels": { "native": 0.6176667988512006, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "two_pool_cycle" }, { "seed": 900000, "sample": 0, "native_reward": 0.618972723875739, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 3, "model_tokens": 308, "reward_channels": { "native": 0.618972723875739, "efficiency": 0.5, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "triangular_cycle" }, { "seed": 900000, "sample": 1, "native_reward": 0.6189743849813638, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 3, "model_tokens": 271, "reward_channels": { "native": 0.6189743849813638, "efficiency": 0.5, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "triangular_cycle" }, { "seed": 900000, "sample": 2, "native_reward": 0.6195573225663248, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 281, "reward_channels": { "native": 0.6195573225663248, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "triangular_cycle" }, { "seed": 900000, "sample": 3, "native_reward": 0.6189739411487076, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 3, "model_tokens": 309, "reward_channels": { "native": 0.6189739411487076, "efficiency": 0.5, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "triangular_cycle" }, { "seed": 900001, "sample": 0, "native_reward": 0.6191941402051148, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 3, "model_tokens": 309, "reward_channels": { "native": 0.6191941402051148, "efficiency": 0.5, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "triangular_cycle" }, { "seed": 900001, "sample": 1, "native_reward": 0.6191957859660329, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 3, "model_tokens": 316, "reward_channels": { "native": 0.6191957859660329, "efficiency": 0.5, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "triangular_cycle" }, { "seed": 900001, "sample": 2, "native_reward": 0.6197781923240601, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 355, "reward_channels": { "native": 0.6197781923240601, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "triangular_cycle" }, { "seed": 900001, "sample": 3, "native_reward": 0.6191951726623806, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 3, "model_tokens": 335, "reward_channels": { "native": 0.6191951726623806, "efficiency": 0.5, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "triangular_cycle" }, { "seed": 900000, "sample": 0, "native_reward": 0.25, "promotion_success": false, "done": false, "errors": 0, "tool_steps": 4, "model_tokens": 721, "reward_channels": { "native": 0.25, "efficiency": 0.0, "reliability": -1.0 }, "verification_level": null, "failed_gates": [ "finalized" ], "family": "competing_cycles" }, { "seed": 900000, "sample": 1, "native_reward": 0.25, "promotion_success": false, "done": false, "errors": 0, "tool_steps": 3, "model_tokens": 245, "reward_channels": { "native": 0.25, "efficiency": 0.0, "reliability": -1.0 }, "verification_level": null, "failed_gates": [ "finalized" ], "family": "competing_cycles" }, { "seed": 900000, "sample": 2, "native_reward": 0.25, "promotion_success": false, "done": false, "errors": 0, "tool_steps": 3, "model_tokens": 319, "reward_channels": { "native": 0.25, "efficiency": 0.0, "reliability": -1.0 }, "verification_level": null, "failed_gates": [ "finalized" ], "family": "competing_cycles" }, { "seed": 900000, "sample": 3, "native_reward": 0.6176130096439503, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 799, "reward_channels": { "native": 0.6176130096439503, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "competing_cycles" }, { "seed": 900001, "sample": 0, "native_reward": 0.6190718400809729, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 696, "reward_channels": { "native": 0.6190718400809729, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "competing_cycles" }, { "seed": 900001, "sample": 1, "native_reward": 0.6210487380906995, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 3, "model_tokens": 430, "reward_channels": { "native": 0.6210487380906995, "efficiency": 0.5, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "competing_cycles" }, { "seed": 900001, "sample": 2, "native_reward": 0.621050450989801, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 3, "model_tokens": 541, "reward_channels": { "native": 0.621050450989801, "efficiency": 0.5, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "competing_cycles" }, { "seed": 900001, "sample": 3, "native_reward": 0.6210504229613636, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 3, "model_tokens": 406, "reward_channels": { "native": 0.6210504229613636, "efficiency": 0.5, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "competing_cycles" }, { "seed": 900000, "sample": 0, "native_reward": 0.6178382865214627, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 307, "reward_channels": { "native": 0.6178382865214627, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "four_hop_cycle" }, { "seed": 900000, "sample": 1, "native_reward": 0.6171504838289823, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 3, "model_tokens": 368, "reward_channels": { "native": 0.6171504838289823, "efficiency": 0.5, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "four_hop_cycle" }, { "seed": 900000, "sample": 2, "native_reward": 0.6178387261247049, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 333, "reward_channels": { "native": 0.6178387261247049, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "four_hop_cycle" }, { "seed": 900000, "sample": 3, "native_reward": 0.6171508614294314, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 3, "model_tokens": 386, "reward_channels": { "native": 0.6171508614294314, "efficiency": 0.5, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "four_hop_cycle" }, { "seed": 900001, "sample": 0, "native_reward": 0.6169854201482908, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 344, "reward_channels": { "native": 0.6169854201482908, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "four_hop_cycle" }, { "seed": 900001, "sample": 1, "native_reward": 0.6161734524620605, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 3, "model_tokens": 387, "reward_channels": { "native": 0.6161734524620605, "efficiency": 0.5, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "four_hop_cycle" }, { "seed": 900001, "sample": 2, "native_reward": 0.6161746136479783, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 3, "model_tokens": 394, "reward_channels": { "native": 0.6161746136479783, "efficiency": 0.5, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "four_hop_cycle" }, { "seed": 900001, "sample": 3, "native_reward": 0.6152915263031933, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 286, "reward_channels": { "native": 0.6152915263031933, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "four_hop_cycle" }, { "seed": 900000, "sample": 0, "native_reward": 0.8334000152269989, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 3, "model_tokens": 370, "reward_channels": { "native": 0.8334000152269989, "efficiency": 0.5, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "size_sensitive_cycle" }, { "seed": 900000, "sample": 1, "native_reward": 0.8323421942212565, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 244, "reward_channels": { "native": 0.8323421942212565, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "size_sensitive_cycle" }, { "seed": 900000, "sample": 2, "native_reward": 0.8323423715483269, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 278, "reward_channels": { "native": 0.8323423715483269, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "size_sensitive_cycle" }, { "seed": 900000, "sample": 3, "native_reward": 0.8333995836483465, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 3, "model_tokens": 359, "reward_channels": { "native": 0.8333995836483465, "efficiency": 0.5, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "size_sensitive_cycle" }, { "seed": 900001, "sample": 0, "native_reward": 0.8314558652540756, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 3, "model_tokens": 357, "reward_channels": { "native": 0.8314558652540756, "efficiency": 0.5, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "size_sensitive_cycle" }, { "seed": 900001, "sample": 1, "native_reward": 0.8314559817276693, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 3, "model_tokens": 353, "reward_channels": { "native": 0.8314559817276693, "efficiency": 0.5, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "size_sensitive_cycle" }, { "seed": 900001, "sample": 2, "native_reward": 0.8314556558584115, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 3, "model_tokens": 343, "reward_channels": { "native": 0.8314556558584115, "efficiency": 0.5, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "size_sensitive_cycle" }, { "seed": 900001, "sample": 3, "native_reward": 0.8304768381206966, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 315, "reward_channels": { "native": 0.8304768381206966, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "size_sensitive_cycle" } ] }, { "run_id": "1789191744", "mode": "no-think", "method": "repo-r", "steps": 50, "reward_points": 25, "train_mean": 0.6372784632444382, "train_last10": 0.7056789830327034, "holdout_native": 0.5724401019920183, "holdout_episodes": 40, "holdout_successes": 31, "mean_step_seconds": 155.50770398359992, "label": "DAPO + REPO-R", "state_at_cutoff": "complete", "step_hours": 2.1598292219944435, "peak_allocated_gib": null, "peak_reserved_gib": null, "extra_refill_rounds": 0.0, "groups_refilled": 0.0, "holdout_component_means": { "native": 0.5724401019920183, "efficiency": 0.25833333333333336, "reliability": 0.55 }, "configuration": { "model": { "id": "Qwen/Qwen3-14B", "revision": "40c069824f4251a91eefaf281ebe4c544efd3e18", "loss": "dapo", "algorithm": "grpo", "no_think": true, "lora": true, "quantize": true, "lora_r": 32, "lora_alpha": 64, "use_dora": false }, "loss": "dapo", "normalization": "joint", "entropy_control": { "method": "repo_r", "epsilon_low": 0.2, "epsilon_high": 0.28, "zeta_initial": 0.001, "zeta_min": 0.0001, "zeta_max": 0.05, "holdout_split": "benchmark", "holdout_seeds_per_family": 2, "num_iterations": 2 }, "dynamic_sampling": false, "sampling": { "group_size": 4, "temperature": 1.0, "max_completion_tokens": 8192, "top_entropy_quantile": 1.0, "temperature_decay_to": null }, "learning_rate": 1e-05, "turn_max_tokens": 1024 }, "source_sha256": { "metrics.jsonl": "bb5df7341e96f3d62594a192c7c782298b3ad56122a13e22516ddcf8807b4a6b", "entropy_holdout_episodes.jsonl": "a4b2eeb7b0a0f4a1f29e4a5da95672097552c97a1ce08d0bb3c60fa82798fb8d", "resolved_training_config.json": "c2e5ccd468c79aa40e12cff5deafe6e9c42c1671e5a1a9db7d24b1b50cd96acf" }, "training_steps": [ { "step": 1, "loss": 0.3677975833415985, "grad_norm": 0.1767578125, "learning_rate": 1e-05, "num_tokens": 28004.0, "completions/mean_length": 604.0, "completions/min_length": 278.0, "completions/max_length": 1190.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 604.0, "completions/min_terminated_length": 278.0, "completions/max_terminated_length": 1190.0, "tools/call_frequency": 4.375, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.5008783936500549, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 0.2172522395849228, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.25, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.1666666716337204, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/mean": 0.3369109332561493, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/std": 0.32382190227508545, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/mean": 0.1666666716337204, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/std": 0.19245010614395142, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/mean": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/std": 1.154700517654419, "reward": 0.4793113172054291, "reward_std": 0.4543764293193817, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.05132016638526693, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 339.4584226249999, "entropy_control/entropy": 0.03175173258437593, "entropy_control/target": 0.03175173258437593, "entropy_control/zeta_used": 0.001, "entropy_control/zeta_next": 0.001, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.004, "adapter_sync_time": 0.0, "generation_time": 236.640148, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 256.70967, "forward_backward_time": 82.656585, "rollout_overhead_time": 20.069522 }, { "step": 2, "loss": 0.36750611662864685, "grad_norm": 0.111328125, "learning_rate": 9.800000000000001e-06, "entropy": 0.05097695626318455, "clip_ratio/low_mean": 0.0005130928620928898, "clip_ratio/low_min": 0.0005130928620928898, "clip_ratio/high_mean": 0.0013298208068590611, "clip_ratio/high_max": 0.0013298208068590611, "clip_ratio/region_mean": 0.001842913668951951, "step_time": 82.7074748819997, "entropy_control/entropy": 0.03138903894117518, "entropy_control/target": 0.03175173258437593, "entropy_control/zeta_used": 0.001, "entropy_control/zeta_next": 0.002, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.008, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 82.601202, "rollout_overhead_time": 0.0, "reward": null }, { "step": 3, "loss": 0.07507390528917313, "grad_norm": 0.20703125, "learning_rate": 9.600000000000001e-06, "num_tokens": 52222.0, "completions/mean_length": 311.25, "completions/min_length": 264.0, "completions/max_length": 421.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 311.25, "completions/min_terminated_length": 264.0, "completions/max_terminated_length": 421.0, "tools/call_frequency": 4.375, "tools/failure_frequency": 0.05714285746216774, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.5584373474121094, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 0.15493246912956238, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.3125, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.0589255690574646, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 0.0, "reward": 0.7740623354911804, "reward_std": 0.15787874162197113, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.5, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.0679577412083745, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 242.9536198360006, "entropy_control/entropy": 0.06898518825867254, "entropy_control/target": 0.03175173258437593, "entropy_control/zeta_used": 0.002, "entropy_control/zeta_next": 0.001, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.012, "adapter_sync_time": 0.0, "generation_time": 151.631961, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 169.045374, "forward_backward_time": 73.80356, "rollout_overhead_time": 17.413413 }, { "step": 4, "loss": 0.0747268944978714, "grad_norm": 0.2021484375, "learning_rate": 9.4e-06, "entropy": 0.06841745739802718, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0013359230069909245, "clip_ratio/high_max": 0.0013359230069909245, "clip_ratio/region_mean": 0.0013359230069909245, "step_time": 73.88337641899943, "entropy_control/entropy": 0.06941178419734081, "entropy_control/target": 0.03175173258437593, "entropy_control/zeta_used": 0.001, "entropy_control/zeta_next": 0.0005, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.016, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 73.785005, "rollout_overhead_time": 0.0, "reward": null }, { "step": 5, "loss": 0.14878179132938385, "grad_norm": 0.18359375, "learning_rate": 9.200000000000002e-06, "num_tokens": 76528.0, "completions/mean_length": 329.25, "completions/min_length": 270.0, "completions/max_length": 545.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 329.25, "completions/min_terminated_length": 270.0, "completions/max_terminated_length": 545.0, "tools/call_frequency": 4.125, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.5586584806442261, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 0.15502142906188965, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.2916666865348816, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.1178511381149292, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 0.75, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 0.7071067690849304, "reward": 0.7232417464256287, "reward_std": 0.30233535170555115, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.5, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.08456735871732235, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 246.90380199200013, "entropy_control/entropy": 0.08560581798981172, "entropy_control/target": 0.03175173258437593, "entropy_control/zeta_used": 0.0005, "entropy_control/zeta_next": 0.00025, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.02, "adapter_sync_time": 0.0, "generation_time": 155.411498, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 172.942414, "forward_backward_time": 73.8652, "rollout_overhead_time": 17.530916 }, { "step": 6, "loss": 0.1488993614912033, "grad_norm": 0.1640625, "learning_rate": 9e-06, "entropy": 0.08421589806675911, "clip_ratio/low_mean": 0.0006574399885721505, "clip_ratio/low_min": 0.0006574399885721505, "clip_ratio/high_mean": 0.0008253562555182725, "clip_ratio/high_max": 0.0008253562555182725, "clip_ratio/region_mean": 0.001482796244090423, "step_time": 73.95877413000017, "entropy_control/entropy": 0.0853382331134385, "entropy_control/target": 0.03175173258437593, "entropy_control/zeta_used": 0.00025, "entropy_control/zeta_next": 0.000125, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.024, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 73.868693, "rollout_overhead_time": 0.0, "reward": null }, { "step": 7, "loss": 0.10186619311571121, "grad_norm": 0.16796875, "learning_rate": 8.8e-06, "num_tokens": 107241.0, "completions/mean_length": 367.625, "completions/min_length": 250.0, "completions/max_length": 733.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 367.625, "completions/min_terminated_length": 250.0, "completions/max_terminated_length": 733.0, "tools/call_frequency": 3.625, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/mean": 0.25, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/mean": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/mean": -1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/mean": 0.5384255051612854, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/std": 0.3330450654029846, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/mean": 0.1666666716337204, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/std": 0.19245010614395142, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/mean": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/std": 1.154700517654419, "reward": 0.29837942123413086, "reward_std": 0.4599093198776245, "frac_reward_zero_std": 0.5, "frac_groups_zero_std": 0.5, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.026633032015524805, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 257.80528411099954, "entropy_control/entropy": 0.023369627524382815, "entropy_control/target": 0.03175173258437593, "entropy_control/zeta_used": 0.000125, "entropy_control/zeta_next": 0.00025, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.028, "adapter_sync_time": 0.0, "generation_time": 160.900667, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 179.543973, "forward_backward_time": 78.175245, "rollout_overhead_time": 18.643306 }, { "step": 8, "loss": 0.10140529274940491, "grad_norm": 0.0966796875, "learning_rate": 8.6e-06, "entropy": 0.025937854894436896, "clip_ratio/low_mean": 0.001117386796977371, "clip_ratio/low_min": 0.001117386796977371, "clip_ratio/high_mean": 0.0013739545829594135, "clip_ratio/high_max": 0.0013739545829594135, "clip_ratio/region_mean": 0.0024913413799367845, "step_time": 78.26017540400039, "entropy_control/entropy": 0.022751461792769316, "entropy_control/target": 0.03175173258437593, "entropy_control/zeta_used": 0.00025, "entropy_control/zeta_next": 0.0005, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.032, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 78.17512, "rollout_overhead_time": 0.0, "reward": null }, { "step": 9, "loss": 0.0, "grad_norm": 0.0, "learning_rate": 8.400000000000001e-06, "num_tokens": 136500.0, "completions/mean_length": 369.375, "completions/min_length": 247.0, "completions/max_length": 890.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 369.375, "completions/min_terminated_length": 247.0, "completions/max_terminated_length": 890.0, "tools/call_frequency": 3.875, "tools/failure_frequency": 0.06451612710952759, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.609546422958374, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.3333333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/mean": 0.25, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/mean": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/mean": -1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/std": 0.0, "reward": 0.4381065368652344, "reward_std": 0.41490334272384644, "frac_reward_zero_std": 1.0, "frac_groups_zero_std": 1.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.05836797400843352, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 330.50201018900043, "entropy_control/entropy": 0.04967694137067181, "entropy_control/target": 0.03175173258437593, "entropy_control/zeta_used": 0.0005, "entropy_control/zeta_next": 0.00025, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.036, "adapter_sync_time": 0.0, "generation_time": 217.606914, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 238.441863, "forward_backward_time": 91.947054, "rollout_overhead_time": 20.834949 }, { "step": 10, "loss": 0.0, "grad_norm": 0.0, "learning_rate": 8.2e-06, "entropy": 0.0587232259567827, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 92.07023043800154, "entropy_control/entropy": 0.05004108105767101, "entropy_control/target": 0.03175173258437593, "entropy_control/zeta_used": 0.00025, "entropy_control/zeta_next": 0.000125, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.04, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 91.951372, "rollout_overhead_time": 0.0, "reward": null }, { "step": 11, "loss": -0.00042744772508740425, "grad_norm": 0.002349853515625, "learning_rate": 8.000000000000001e-06, "num_tokens": 165380.0, "completions/mean_length": 297.0, "completions/min_length": 256.0, "completions/max_length": 382.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 297.0, "completions/min_terminated_length": 256.0, "completions/max_terminated_length": 382.0, "tools/call_frequency": 4.0, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__native/mean": 0.6152178645133972, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__native/std": 2.2717633783031488e-06, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__efficiency/mean": 0.3333333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__efficiency/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/mean": 0.6130492687225342, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/std": 1.3786594763587345e-06, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/mean": 0.3333333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/std": 0.0, "reward": 0.8308002352714539, "reward_std": 0.001159140607342124, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.02870261576026678, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 209.886425788, "entropy_control/entropy": 0.02939591253793728, "entropy_control/target": 0.03175173258437593, "entropy_control/zeta_used": 0.000125, "entropy_control/zeta_next": 0.00025, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.044, "adapter_sync_time": 0.0, "generation_time": 110.477852, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 129.265464, "forward_backward_time": 80.514475, "rollout_overhead_time": 18.787612 }, { "step": 12, "loss": -0.00042418157681822777, "grad_norm": 0.002410888671875, "learning_rate": 7.800000000000002e-06, "entropy": 0.02758015098515898, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 80.60424775000092, "entropy_control/entropy": 0.028318129083174517, "entropy_control/target": 0.03175173258437593, "entropy_control/zeta_used": 0.00025, "entropy_control/zeta_next": 0.0005, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.048, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 80.502626, "rollout_overhead_time": 0.0, "reward": null }, { "step": 13, "loss": -0.0006278872024267912, "grad_norm": 0.0030364990234375, "learning_rate": 7.600000000000001e-06, "num_tokens": 194020.0, "completions/mean_length": 291.0, "completions/min_length": 254.0, "completions/max_length": 358.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 291.0, "completions/min_terminated_length": 254.0, "completions/max_terminated_length": 358.0, "tools/call_frequency": 3.5, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.6147884130477905, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 2.126363142451737e-06, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.3333333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/mean": 0.25, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/mean": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/mean": -1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/std": 0.0, "reward": 0.4407275319099426, "reward_std": 0.4177052974700928, "frac_reward_zero_std": 0.5, "frac_groups_zero_std": 0.5, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.05903345951810479, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 221.18660950699996, "entropy_control/entropy": 0.06082484628909123, "entropy_control/target": 0.03175173258437593, "entropy_control/zeta_used": 0.0005, "entropy_control/zeta_next": 0.00025, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.052, "adapter_sync_time": 0.0, "generation_time": 108.3372, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 129.130762, "forward_backward_time": 91.944136, "rollout_overhead_time": 20.793562 }, { "step": 14, "loss": -0.0006222964730113745, "grad_norm": 0.0029144287109375, "learning_rate": 7.4e-06, "entropy": 0.05921871028840542, "clip_ratio/low_mean": 0.001252432237379253, "clip_ratio/low_min": 0.001252432237379253, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.001252432237379253, "step_time": 92.04747875399926, "entropy_control/entropy": 0.060986647525984344, "entropy_control/target": 0.03175173258437593, "entropy_control/zeta_used": 0.00025, "entropy_control/zeta_next": 0.000125, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.056, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 91.938908, "rollout_overhead_time": 0.0, "reward": null }, { "step": 15, "loss": 0.09406601637601852, "grad_norm": 0.14453125, "learning_rate": 7.2000000000000005e-06, "num_tokens": 224524.0, "completions/mean_length": 325.5, "completions/min_length": 246.0, "completions/max_length": 443.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 325.5, "completions/min_terminated_length": 246.0, "completions/max_terminated_length": 443.0, "tools/call_frequency": 3.5, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/mean": 0.25, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/mean": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/mean": -1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/mean": 0.39422667026519775, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/std": 0.2884533107280731, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/mean": 0.0833333358168602, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/std": 0.1666666716337204, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/mean": -0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/std": 1.0, "reward": 0.17419666051864624, "reward_std": 0.35128122568130493, "frac_reward_zero_std": 0.5, "frac_groups_zero_std": 0.5, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.02758158289361745, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 215.4094084999997, "entropy_control/entropy": 0.026020437417879422, "entropy_control/target": 0.03175173258437593, "entropy_control/zeta_used": 0.000125, "entropy_control/zeta_next": 0.00025, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.06, "adapter_sync_time": 0.0, "generation_time": 116.292477, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 134.882183, "forward_backward_time": 80.414608, "rollout_overhead_time": 18.589707 }, { "step": 16, "loss": 0.09405205398797989, "grad_norm": 0.1181640625, "learning_rate": 7e-06, "entropy": 0.02729236683808267, "clip_ratio/low_mean": 0.0005840994126629084, "clip_ratio/low_min": 0.0005840994126629084, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0005840994126629084, "step_time": 80.51190537299908, "entropy_control/entropy": 0.025735738148851568, "entropy_control/target": 0.03175173258437593, "entropy_control/zeta_used": 0.00025, "entropy_control/zeta_next": 0.0005, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.064, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 80.402759, "rollout_overhead_time": 0.0, "reward": null }, { "step": 17, "loss": -6.423094600904733e-05, "grad_norm": 0.00116729736328125, "learning_rate": 6.800000000000001e-06, "num_tokens": 256259.0, "completions/mean_length": 271.875, "completions/min_length": 213.0, "completions/max_length": 306.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 271.875, "completions/min_terminated_length": 213.0, "completions/max_terminated_length": 306.0, "tools/call_frequency": 3.5, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/mean": 0.25, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/mean": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/mean": -1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/mean": 0.6142328977584839, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/std": 2.5271713184338296e-06, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/mean": 0.3333333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/std": 0.0, "reward": 0.4404497742652893, "reward_std": 0.4174083471298218, "frac_reward_zero_std": 0.5, "frac_groups_zero_std": 0.5, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.01683327781211119, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 198.84701016199915, "entropy_control/entropy": 0.01737684080312262, "entropy_control/target": 0.03175173258437593, "entropy_control/zeta_used": 0.0005, "entropy_control/zeta_next": 0.001, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.068, "adapter_sync_time": 0.0, "generation_time": 88.976497, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 109.227697, "forward_backward_time": 89.507457, "rollout_overhead_time": 20.2512 }, { "step": 18, "loss": -5.6417673476971686e-05, "grad_norm": 0.00135040283203125, "learning_rate": 6.600000000000001e-06, "entropy": 0.016886168712517247, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 89.61900508899998, "entropy_control/entropy": 0.017444355780958987, "entropy_control/target": 0.03175173258437593, "entropy_control/zeta_used": 0.001, "entropy_control/zeta_next": 0.002, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.072, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 89.502407, "rollout_overhead_time": 0.0, "reward": null }, { "step": 19, "loss": 0.15298286080360413, "grad_norm": 0.193359375, "learning_rate": 6.4000000000000006e-06, "num_tokens": 281939.0, "completions/mean_length": 315.5, "completions/min_length": 253.0, "completions/max_length": 521.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 315.5, "completions/min_terminated_length": 253.0, "completions/max_terminated_length": 521.0, "tools/call_frequency": 4.0, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.6129077672958374, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 1.0220068134003668e-06, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.3333333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/mean": 0.684334397315979, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/std": 0.28955626487731934, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/mean": 0.25, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/std": 0.1666666716337204, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/std": 1.0, "reward": 0.8132044076919556, "reward_std": 0.32641470432281494, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.5, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.044751655659638345, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 249.52503382100076, "entropy_control/entropy": 0.045542472828241955, "entropy_control/target": 0.03175173258437593, "entropy_control/zeta_used": 0.002, "entropy_control/zeta_next": 0.001, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.076, "adapter_sync_time": 0.0, "generation_time": 147.188196, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 166.216887, "forward_backward_time": 83.19432, "rollout_overhead_time": 19.028691 }, { "step": 20, "loss": 0.15252260863780975, "grad_norm": 0.12060546875, "learning_rate": 6.200000000000001e-06, "entropy": 0.044047977426089346, "clip_ratio/low_mean": 0.000973917602095753, "clip_ratio/low_min": 0.000973917602095753, "clip_ratio/high_mean": 0.0004789271915797144, "clip_ratio/high_max": 0.0004789271915797144, "clip_ratio/region_mean": 0.0014528447936754674, "step_time": 83.31169498500003, "entropy_control/entropy": 0.044899696954788675, "entropy_control/target": 0.03175173258437593, "entropy_control/zeta_used": 0.001, "entropy_control/zeta_next": 0.0005, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.08, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 83.196748, "rollout_overhead_time": 0.0, "reward": null }, { "step": 21, "loss": 0.0008379751816391945, "grad_norm": 0.00372314453125, "learning_rate": 6e-06, "num_tokens": 314181.0, "completions/mean_length": 327.25, "completions/min_length": 256.0, "completions/max_length": 512.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 327.25, "completions/min_terminated_length": 256.0, "completions/max_terminated_length": 512.0, "tools/call_frequency": 3.5, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/mean": 0.25, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/mean": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/mean": -1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/mean": 0.6129704117774963, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/std": 2.9602974791487213e-06, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/mean": 0.3333333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/std": 0.0, "reward": 0.43981853127479553, "reward_std": 0.4167335629463196, "frac_reward_zero_std": 0.5, "frac_groups_zero_std": 0.5, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.02073569670028519, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 232.0127544120005, "entropy_control/entropy": 0.02058484643387447, "entropy_control/target": 0.03175173258437593, "entropy_control/zeta_used": 0.0005, "entropy_control/zeta_next": 0.001, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.084, "adapter_sync_time": 0.0, "generation_time": 119.038225, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 140.031491, "forward_backward_time": 91.865806, "rollout_overhead_time": 20.993266 }, { "step": 22, "loss": 0.0008416565251536667, "grad_norm": 0.002777099609375, "learning_rate": 5.8e-06, "entropy": 0.020891853877401445, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 91.99613114599924, "entropy_control/entropy": 0.020715168776132165, "entropy_control/target": 0.03175173258437593, "entropy_control/zeta_used": 0.001, "entropy_control/zeta_next": 0.002, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.088, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 91.877733, "rollout_overhead_time": 0.0, "reward": null }, { "step": 23, "loss": 5.250214599072933e-06, "grad_norm": 0.00164031982421875, "learning_rate": 5.600000000000001e-06, "num_tokens": 339658.0, "completions/mean_length": 295.125, "completions/min_length": 258.0, "completions/max_length": 328.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 295.125, "completions/min_terminated_length": 258.0, "completions/max_terminated_length": 328.0, "tools/call_frequency": 4.25, "tools/failure_frequency": 0.05882352963089943, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.61496502161026, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 1.3495769053406548e-06, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.3333333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/mean": 0.8308384418487549, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/std": 6.769786295990343e-07, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/mean": 0.3333333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/std": 0.0, "reward": 0.9395684003829956, "reward_std": 0.11538917571306229, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.5, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.04453222232405096, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 216.0728001890011, "entropy_control/entropy": 0.04541582179304047, "entropy_control/target": 0.03175173258437593, "entropy_control/zeta_used": 0.002, "entropy_control/zeta_next": 0.001, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.092, "adapter_sync_time": 0.0, "generation_time": 113.720911, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 132.717597, "forward_backward_time": 83.234269, "rollout_overhead_time": 18.996686 }, { "step": 24, "loss": 6.836606189608574e-06, "grad_norm": 0.00170135498046875, "learning_rate": 5.400000000000001e-06, "entropy": 0.0446750185219571, "clip_ratio/low_mean": 0.00046641789958812296, "clip_ratio/low_min": 0.00046641789958812296, "clip_ratio/high_mean": 0.0008405257540289313, "clip_ratio/high_max": 0.0008405257540289313, "clip_ratio/region_mean": 0.0013069436536170542, "step_time": 83.3361953620024, "entropy_control/entropy": 0.04554446372616983, "entropy_control/target": 0.03175173258437593, "entropy_control/zeta_used": 0.001, "entropy_control/zeta_next": 0.0005, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.096, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 83.22157, "rollout_overhead_time": 0.0, "reward": null }, { "step": 25, "loss": -0.00014857194037176669, "grad_norm": 0.00225830078125, "learning_rate": 5.2e-06, "num_tokens": 366783.0, "completions/mean_length": 296.625, "completions/min_length": 240.0, "completions/max_length": 312.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 296.625, "completions/min_terminated_length": 240.0, "completions/max_terminated_length": 312.0, "tools/call_frequency": 4.125, "tools/failure_frequency": 0.03030303120613098, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.6142096519470215, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 1.2474328059397521e-06, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.3333333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/mean": 0.6138775944709778, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/std": 1.644627445784863e-06, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/mean": 0.3333333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/std": 0.0, "reward": 0.830710232257843, "reward_std": 0.00017749733524397016, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.04687308776192367, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 230.28595424500054, "entropy_control/entropy": 0.047156096221906994, "entropy_control/target": 0.03175173258437593, "entropy_control/zeta_used": 0.0005, "entropy_control/zeta_next": 0.00025, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.1, "adapter_sync_time": 0.0, "generation_time": 123.061737, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 142.856217, "forward_backward_time": 87.308331, "rollout_overhead_time": 19.794481 }, { "step": 26, "loss": -0.0001543594989925623, "grad_norm": 0.00157928466796875, "learning_rate": 5e-06, "entropy": 0.047321059624664485, "clip_ratio/low_mean": 0.0004071661096531898, "clip_ratio/low_min": 0.0004071661096531898, "clip_ratio/high_mean": 0.0008025702845770866, "clip_ratio/high_max": 0.0008025702845770866, "clip_ratio/region_mean": 0.0012097363942302763, "step_time": 87.43614393299958, "entropy_control/entropy": 0.04758420987263972, "entropy_control/target": 0.03175173258437593, "entropy_control/zeta_used": 0.00025, "entropy_control/zeta_next": 0.000125, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.104, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 87.320586, "rollout_overhead_time": 0.0, "reward": null }, { "step": 27, "loss": 0.07329875975847244, "grad_norm": 0.15234375, "learning_rate": 4.800000000000001e-06, "num_tokens": 395905.0, "completions/mean_length": 333.75, "completions/min_length": 303.0, "completions/max_length": 438.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 333.75, "completions/min_terminated_length": 303.0, "completions/max_terminated_length": 438.0, "tools/call_frequency": 4.0, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/mean": 0.612331748008728, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/std": 4.526291661477444e-07, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/mean": 0.3333333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/mean": 0.6831650733947754, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/std": 0.2887767553329468, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/mean": 0.25, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/std": 0.1666666716337204, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/std": 1.0, "reward": 0.8123317360877991, "reward_std": 0.32592231035232544, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.5, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.023141608107835054, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 235.536094047, "entropy_control/entropy": 0.022956123942045287, "entropy_control/target": 0.03175173258437593, "entropy_control/zeta_used": 0.000125, "entropy_control/zeta_next": 0.00025, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.108, "adapter_sync_time": 0.0, "generation_time": 136.594134, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 155.359358, "forward_backward_time": 80.057082, "rollout_overhead_time": 18.765225 }, { "step": 28, "loss": 0.07343291491270065, "grad_norm": 0.1943359375, "learning_rate": 4.600000000000001e-06, "entropy": 0.02303338935598731, "clip_ratio/low_mean": 0.0002853881160262972, "clip_ratio/low_min": 0.0002853881160262972, "clip_ratio/high_mean": 0.00040983606595546007, "clip_ratio/high_max": 0.00040983606595546007, "clip_ratio/region_mean": 0.0006952241819817573, "step_time": 80.15440302699972, "entropy_control/entropy": 0.022861084086465176, "entropy_control/target": 0.03175173258437593, "entropy_control/zeta_used": 0.00025, "entropy_control/zeta_next": 0.0005, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.112, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 80.04201, "rollout_overhead_time": 0.0, "reward": null }, { "step": 29, "loss": -0.0004251918289810419, "grad_norm": 0.0012054443359375, "learning_rate": 4.4e-06, "num_tokens": 424415.0, "completions/mean_length": 277.25, "completions/min_length": 236.0, "completions/max_length": 311.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 277.25, "completions/min_terminated_length": 236.0, "completions/max_terminated_length": 311.0, "tools/call_frequency": 3.5, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.6138591170310974, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 9.524317761133716e-07, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.3333333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/mean": 0.25, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/mean": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/mean": -1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/std": 0.0, "reward": 0.44026288390159607, "reward_std": 0.41720855236053467, "frac_reward_zero_std": 0.5, "frac_groups_zero_std": 1.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.05170884821563959, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 225.034533508001, "entropy_control/entropy": 0.05304620545559113, "entropy_control/target": 0.03175173258437593, "entropy_control/zeta_used": 0.0005, "entropy_control/zeta_next": 0.00025, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.116, "adapter_sync_time": 0.0, "generation_time": 112.046297, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 133.043454, "forward_backward_time": 91.872883, "rollout_overhead_time": 20.997157 }, { "step": 30, "loss": -0.0004242782015353441, "grad_norm": 0.00106048583984375, "learning_rate": 4.2000000000000004e-06, "entropy": 0.05198635067790747, "clip_ratio/low_mean": 0.0005296610179357231, "clip_ratio/low_min": 0.0005296610179357231, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0005296610179357231, "step_time": 91.9762094639982, "entropy_control/entropy": 0.0533380043688536, "entropy_control/target": 0.03175173258437593, "entropy_control/zeta_used": 0.00025, "entropy_control/zeta_next": 0.000125, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.12, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 91.8649, "rollout_overhead_time": 0.0, "reward": null }, { "step": 31, "loss": 0.0007572928443551064, "grad_norm": 0.0031890869140625, "learning_rate": 4.000000000000001e-06, "num_tokens": 451380.0, "completions/mean_length": 283.125, "completions/min_length": 257.0, "completions/max_length": 319.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 283.125, "completions/min_terminated_length": 257.0, "completions/max_terminated_length": 319.0, "tools/call_frequency": 4.0, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/mean": 0.8303067684173584, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/std": 0.000992488581687212, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/mean": 0.3333333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/std": 0.0, "reward": 1.046973466873169, "reward_std": 0.000992488581687212, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.5, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.01986620854586363, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 169.8875314800007, "entropy_control/entropy": 0.019852706463515807, "entropy_control/target": 0.03175173258437593, "entropy_control/zeta_used": 0.000125, "entropy_control/zeta_next": 0.00025, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.124, "adapter_sync_time": 0.0, "generation_time": 87.769804, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 103.579312, "forward_backward_time": 66.214919, "rollout_overhead_time": 15.809508 }, { "step": 32, "loss": 0.0007597359362989664, "grad_norm": 0.0036468505859375, "learning_rate": 3.8000000000000005e-06, "entropy": 0.020172513090074062, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.00044802867341786623, "clip_ratio/high_max": 0.00044802867341786623, "clip_ratio/region_mean": 0.00044802867341786623, "step_time": 66.26559639399966, "entropy_control/entropy": 0.0201377382248116, "entropy_control/target": 0.03175173258437593, "entropy_control/zeta_used": 0.00025, "entropy_control/zeta_next": 0.0005, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.128, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 66.17027, "rollout_overhead_time": 0.0, "reward": null }, { "step": 33, "loss": -0.00016835879068821669, "grad_norm": 0.0024566650390625, "learning_rate": 3.6000000000000003e-06, "num_tokens": 476805.0, "completions/mean_length": 278.625, "completions/min_length": 250.0, "completions/max_length": 316.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 278.625, "completions/min_terminated_length": 250.0, "completions/max_terminated_length": 316.0, "tools/call_frequency": 4.25, "tools/failure_frequency": 0.05882352963089943, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.6154767870903015, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 1.2811532315026852e-06, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.3333333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/mean": 0.8308459520339966, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/std": 9.914219845086336e-07, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/mean": 0.3333333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/std": 0.0, "reward": 0.9398280382156372, "reward_std": 0.11511971056461334, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.5, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.040752315893769264, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 207.3118805610011, "entropy_control/entropy": 0.04239122198632703, "entropy_control/target": 0.03175173258437593, "entropy_control/zeta_used": 0.0005, "entropy_control/zeta_next": 0.00025, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.132, "adapter_sync_time": 0.0, "generation_time": 104.487355, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 123.615215, "forward_backward_time": 83.604137, "rollout_overhead_time": 19.127861 }, { "step": 34, "loss": -0.00016834831330925226, "grad_norm": 0.002471923828125, "learning_rate": 3.4000000000000005e-06, "entropy": 0.0409962716512382, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0008389261784031987, "clip_ratio/high_max": 0.0008389261784031987, "clip_ratio/region_mean": 0.0008389261784031987, "step_time": 83.72439446099997, "entropy_control/entropy": 0.04263105424006609, "entropy_control/target": 0.03175173258437593, "entropy_control/zeta_used": 0.00025, "entropy_control/zeta_next": 0.000125, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.136, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 83.613728, "rollout_overhead_time": 0.0, "reward": null }, { "step": 35, "loss": -0.05533386394381523, "grad_norm": 0.1728515625, "learning_rate": 3.2000000000000003e-06, "num_tokens": 510224.0, "completions/mean_length": 310.375, "completions/min_length": 232.0, "completions/max_length": 636.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 310.375, "completions/min_terminated_length": 232.0, "completions/max_terminated_length": 636.0, "tools/call_frequency": 2.875, "tools/failure_frequency": 0.043478261679410934, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/mean": 0.21875, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/std": 0.0883883461356163, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/mean": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/mean": -1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/std": 0.0, "reward": 0.01874999701976776, "reward_std": 0.0883883461356163, "frac_reward_zero_std": 0.5, "frac_groups_zero_std": 0.5, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 0.5, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.024597404757514596, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 222.9520893609988, "entropy_control/entropy": 0.023220469573852827, "entropy_control/target": 0.03175173258437593, "entropy_control/zeta_used": 0.000125, "entropy_control/zeta_next": 0.00025, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.14, "adapter_sync_time": 0.0, "generation_time": 135.606779, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 152.403962, "forward_backward_time": 70.451005, "rollout_overhead_time": 16.797183 }, { "step": 36, "loss": -0.05530177056789398, "grad_norm": 0.181640625, "learning_rate": 3e-06, "entropy": 0.024223731132224202, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.00019654087373055518, "clip_ratio/high_max": 0.00019654087373055518, "clip_ratio/region_mean": 0.00019654087373055518, "step_time": 70.51143897200109, "entropy_control/entropy": 0.02296708882910044, "entropy_control/target": 0.03175173258437593, "entropy_control/zeta_used": 0.00025, "entropy_control/zeta_next": 0.0005, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.144, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 70.429582, "rollout_overhead_time": 0.0, "reward": null }, { "step": 37, "loss": 0.11811134964227676, "grad_norm": 0.2353515625, "learning_rate": 2.8000000000000003e-06, "num_tokens": 535854.0, "completions/mean_length": 319.75, "completions/min_length": 255.0, "completions/max_length": 483.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 319.75, "completions/min_terminated_length": 255.0, "completions/max_terminated_length": 483.0, "tools/call_frequency": 3.75, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.6132173538208008, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 1.1980384897469776e-06, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.3333333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__native/mean": 0.5245044231414795, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__native/std": 0.18301118910312653, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__efficiency/mean": 0.3333333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__efficiency/std": 0.27216553688049316, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__reliability/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__reliability/std": 1.0, "reward": 0.735527515411377, "reward_std": 0.2770969271659851, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.07394927646964788, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 243.5064796279994, "entropy_control/entropy": 0.0762606649121152, "entropy_control/target": 0.03175173258437593, "entropy_control/zeta_used": 0.0005, "entropy_control/zeta_next": 0.00025, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.148, "adapter_sync_time": 0.0, "generation_time": 141.841758, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 160.718509, "forward_backward_time": 82.700984, "rollout_overhead_time": 18.876751 }, { "step": 38, "loss": 0.11816050857305527, "grad_norm": 0.2177734375, "learning_rate": 2.6e-06, "entropy": 0.07379282952751964, "clip_ratio/low_mean": 0.0002587991766631603, "clip_ratio/low_min": 0.0002587991766631603, "clip_ratio/high_mean": 0.00045454545761458576, "clip_ratio/high_max": 0.00045454545761458576, "clip_ratio/region_mean": 0.0007133446342777461, "step_time": 82.81517247699958, "entropy_control/entropy": 0.07615786743315026, "entropy_control/target": 0.03175173258437593, "entropy_control/zeta_used": 0.00025, "entropy_control/zeta_next": 0.000125, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.152, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 82.713856, "rollout_overhead_time": 0.0, "reward": null }, { "step": 39, "loss": 0.25485917925834656, "grad_norm": 0.189453125, "learning_rate": 2.4000000000000003e-06, "num_tokens": 563256.0, "completions/mean_length": 330.25, "completions/min_length": 239.0, "completions/max_length": 755.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 330.25, "completions/min_terminated_length": 239.0, "completions/max_terminated_length": 755.0, "tools/call_frequency": 3.75, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__native/mean": 0.5582360029220581, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__native/std": 0.15485496819019318, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__efficiency/mean": 0.3333333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__efficiency/std": 0.15430335700511932, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__reliability/mean": 0.75, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__reliability/std": 0.7071067690849304, "reward": 0.7249026298522949, "reward_std": 0.3030456602573395, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.03695832286030054, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 241.49042306200045, "entropy_control/entropy": 0.041021604389103454, "entropy_control/target": 0.03175173258437593, "entropy_control/zeta_used": 0.000125, "entropy_control/zeta_next": -0.0001, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.156, "adapter_sync_time": 0.0, "generation_time": 153.868856, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 170.668723, "forward_backward_time": 70.71586, "rollout_overhead_time": 16.799866 }, { "step": 40, "loss": 0.25534555315971375, "grad_norm": 0.2451171875, "learning_rate": 2.2e-06, "entropy": 0.03707983472850174, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0007852564158383757, "clip_ratio/high_max": 0.0007852564158383757, "clip_ratio/region_mean": 0.0007852564158383757, "step_time": 70.82481263099999, "entropy_control/entropy": 0.041115449063686735, "entropy_control/target": 0.03175173258437593, "entropy_control/zeta_used": -0.0001, "entropy_control/zeta_next": -0.0002, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.16, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 70.720887, "rollout_overhead_time": 0.0, "reward": null }, { "step": 41, "loss": -0.0006246133707463741, "grad_norm": 0.00579833984375, "learning_rate": 2.0000000000000003e-06, "num_tokens": 593164.0, "completions/mean_length": 257.5, "completions/min_length": 234.0, "completions/max_length": 309.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 257.5, "completions/min_terminated_length": 234.0, "completions/max_terminated_length": 309.0, "tools/call_frequency": 3.5, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__native/mean": 0.6161777973175049, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__native/std": 1.2583028592416667e-06, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__efficiency/mean": 0.3333333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__efficiency/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/mean": 0.25, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/mean": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/mean": -1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/std": 0.0, "reward": 0.4414222240447998, "reward_std": 0.41844794154167175, "frac_reward_zero_std": 0.5, "frac_groups_zero_std": 0.5, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.02481665855157189, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 181.7729249949989, "entropy_control/entropy": 0.025124236636757252, "entropy_control/target": 0.03175173258437593, "entropy_control/zeta_used": -0.0002, "entropy_control/zeta_next": -0.0001, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.164, "adapter_sync_time": 0.0, "generation_time": 84.861507, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 103.006375, "forward_backward_time": 78.664182, "rollout_overhead_time": 18.144868 }, { "step": 42, "loss": -0.0006263978430069983, "grad_norm": 0.0020904541015625, "learning_rate": 1.8000000000000001e-06, "entropy": 0.024449583375826478, "clip_ratio/low_mean": 0.00046296295477077365, "clip_ratio/low_min": 0.00046296295477077365, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.00046296295477077365, "step_time": 78.73975231299937, "entropy_control/entropy": 0.024743860042287756, "entropy_control/target": 0.03175173258437593, "entropy_control/zeta_used": -0.0001, "entropy_control/zeta_next": 0.0001, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.168, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 78.642634, "rollout_overhead_time": 0.0, "reward": null }, { "step": 43, "loss": -5.60487387701869e-06, "grad_norm": 0.0013275146484375, "learning_rate": 1.6000000000000001e-06, "num_tokens": 621931.0, "completions/mean_length": 287.875, "completions/min_length": 266.0, "completions/max_length": 349.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 287.875, "completions/min_terminated_length": 266.0, "completions/max_terminated_length": 349.0, "tools/call_frequency": 4.0, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__native/mean": 0.6159943342208862, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__native/std": 1.6671551748004276e-06, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__efficiency/mean": 0.3333333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__efficiency/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/mean": 0.6147299408912659, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/std": 1.0695671335270163e-06, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/mean": 0.3333333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/std": 0.0, "reward": 0.8320287466049194, "reward_std": 0.0006758717936463654, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.5, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.02148130163550377, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 190.59470718299963, "entropy_control/entropy": 0.021568277703800976, "entropy_control/target": 0.03175173258437593, "entropy_control/zeta_used": 0.0001, "entropy_control/zeta_next": 0.0002, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.172, "adapter_sync_time": 0.0, "generation_time": 93.406351, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 111.931507, "forward_backward_time": 78.561703, "rollout_overhead_time": 18.525157 }, { "step": 44, "loss": -4.204513970762491e-06, "grad_norm": 0.00150299072265625, "learning_rate": 1.4000000000000001e-06, "entropy": 0.021757159032858908, "clip_ratio/low_mean": 0.0004448398540262133, "clip_ratio/low_min": 0.0004448398540262133, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0004448398540262133, "step_time": 78.65292478099946, "entropy_control/entropy": 0.021835107669043168, "entropy_control/target": 0.03175173258437593, "entropy_control/zeta_used": 0.0002, "entropy_control/zeta_next": 0.0004, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.176, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 78.555987, "rollout_overhead_time": 0.0, "reward": null }, { "step": 45, "loss": -0.0010895675513893366, "grad_norm": 0.00445556640625, "learning_rate": 1.2000000000000002e-06, "num_tokens": 653773.0, "completions/mean_length": 306.25, "completions/min_length": 214.0, "completions/max_length": 500.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 306.25, "completions/min_terminated_length": 214.0, "completions/max_terminated_length": 500.0, "tools/call_frequency": 3.5, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/mean": 0.25, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/mean": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/mean": -1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/mean": 0.6127362847328186, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/std": 1.9606213754741475e-06, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/mean": 0.3333333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/std": 0.0, "reward": 0.43970146775245667, "reward_std": 0.41660842299461365, "frac_reward_zero_std": 0.5, "frac_groups_zero_std": 0.5, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.020556892210152, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 232.68937776899838, "entropy_control/entropy": 0.020748718833716467, "entropy_control/target": 0.03175173258437593, "entropy_control/zeta_used": 0.0004, "entropy_control/zeta_next": 0.0008, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.18, "adapter_sync_time": 0.0, "generation_time": 121.29158, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 141.831924, "forward_backward_time": 90.746301, "rollout_overhead_time": 20.540344 }, { "step": 46, "loss": -0.0010817635338753462, "grad_norm": 0.0036773681640625, "learning_rate": 1.0000000000000002e-06, "entropy": 0.020918297610478476, "clip_ratio/low_mean": 0.0004310344811528921, "clip_ratio/low_min": 0.0004310344811528921, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0004310344811528921, "step_time": 90.8428927979985, "entropy_control/entropy": 0.021136172758211988, "entropy_control/target": 0.03175173258437593, "entropy_control/zeta_used": 0.0008, "entropy_control/zeta_next": 0.0016, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.184, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 90.731511, "rollout_overhead_time": 0.0, "reward": null }, { "step": 47, "loss": -4.239944973960519e-05, "grad_norm": 0.001312255859375, "learning_rate": 8.000000000000001e-07, "num_tokens": 682413.0, "completions/mean_length": 289.5, "completions/min_length": 251.0, "completions/max_length": 342.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 289.5, "completions/min_terminated_length": 251.0, "completions/max_terminated_length": 342.0, "tools/call_frequency": 4.0, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/mean": 0.6142542362213135, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/std": 7.160831501096254e-07, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/mean": 0.3333333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/mean": 0.8299664258956909, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/std": 8.787053502601339e-07, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/mean": 0.3333333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/std": 0.0, "reward": 0.938776969909668, "reward_std": 0.11530302464962006, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 1.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.016255779890343547, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 197.25902485600272, "entropy_control/entropy": 0.016405128539524992, "entropy_control/target": 0.03175173258437593, "entropy_control/zeta_used": 0.0016, "entropy_control/zeta_next": 0.0032, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.188, "adapter_sync_time": 0.0, "generation_time": 99.347984, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 118.015269, "forward_backward_time": 79.122098, "rollout_overhead_time": 18.667285 }, { "step": 48, "loss": -4.2037107050418854e-05, "grad_norm": 0.001129150390625, "learning_rate": 6.000000000000001e-07, "entropy": 0.016197846503928304, "clip_ratio/low_mean": 0.00048638132284395397, "clip_ratio/low_min": 0.00048638132284395397, "clip_ratio/high_mean": 0.000405844155466184, "clip_ratio/high_max": 0.000405844155466184, "clip_ratio/region_mean": 0.000892225478310138, "step_time": 79.22551988899613, "entropy_control/entropy": 0.0163411147522479, "entropy_control/target": 0.03175173258437593, "entropy_control/zeta_used": 0.0032, "entropy_control/zeta_next": 0.0064, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.192, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 79.109247, "rollout_overhead_time": 0.0, "reward": null }, { "step": 49, "loss": -0.0002504608128219843, "grad_norm": 0.00145721435546875, "learning_rate": 4.0000000000000003e-07, "num_tokens": 711149.0, "completions/mean_length": 297.0, "completions/min_length": 257.0, "completions/max_length": 334.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 297.0, "completions/min_terminated_length": 257.0, "completions/max_terminated_length": 334.0, "tools/call_frequency": 4.0, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/mean": 0.6137466430664062, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/std": 1.442054781364277e-06, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/mean": 0.3333333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/mean": 0.8306776285171509, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/std": 1.1552644991752459e-06, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/mean": 0.3333333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/std": 0.0, "reward": 0.9388787746429443, "reward_std": 0.11595447361469269, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.02201947581488639, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 193.7512692649998, "entropy_control/entropy": 0.022578372963406326, "entropy_control/target": 0.03175173258437593, "entropy_control/zeta_used": 0.0064, "entropy_control/zeta_next": 0.0128, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.196, "adapter_sync_time": 0.0, "generation_time": 95.730815, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 114.432158, "forward_backward_time": 79.207311, "rollout_overhead_time": 18.701343 }, { "step": 50, "loss": -0.0002588336355984211, "grad_norm": 0.0013275146484375, "learning_rate": 2.0000000000000002e-07, "entropy": 0.021800043410621583, "clip_ratio/low_mean": 0.0004139072843827307, "clip_ratio/low_min": 0.0004139072843827307, "clip_ratio/high_mean": 0.000800051202531904, "clip_ratio/high_max": 0.000800051202531904, "clip_ratio/region_mean": 0.0012139584869146347, "step_time": 79.27377721599987, "entropy_control/entropy": 0.022359675930595898, "entropy_control/target": 0.03175173258437593, "entropy_control/zeta_used": 0.0128, "entropy_control/zeta_next": 0.0256, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.2, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 79.174211, "rollout_overhead_time": 0.0, "reward": null } ], "holdout": [ { "seed": 900000, "sample": 0, "native_reward": 0.6170665707246713, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 344, "reward_channels": { "native": 0.6170665707246713, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "two_pool_cycle" }, { "seed": 900000, "sample": 1, "native_reward": 0.6170664895195931, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 288, "reward_channels": { "native": 0.6170664895195931, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "two_pool_cycle" }, { "seed": 900000, "sample": 2, "native_reward": 0.6170663329743196, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 274, "reward_channels": { "native": 0.6170663329743196, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "two_pool_cycle" }, { "seed": 900000, "sample": 3, "native_reward": 0.6170653436345541, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 375, "reward_channels": { "native": 0.6170653436345541, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "two_pool_cycle" }, { "seed": 900001, "sample": 0, "native_reward": 0.6176353807103607, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 272, "reward_channels": { "native": 0.6176353807103607, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "two_pool_cycle" }, { "seed": 900001, "sample": 1, "native_reward": 0.6176363454881537, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 346, "reward_channels": { "native": 0.6176363454881537, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "two_pool_cycle" }, { "seed": 900001, "sample": 2, "native_reward": 0.617635481460556, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 409, "reward_channels": { "native": 0.617635481460556, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "two_pool_cycle" }, { "seed": 900001, "sample": 3, "native_reward": 0.6176363608829585, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 342, "reward_channels": { "native": 0.6176363608829585, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "two_pool_cycle" }, { "seed": 900000, "sample": 0, "native_reward": 0.6181042530675553, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 380, "reward_channels": { "native": 0.6181042530675553, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "triangular_cycle" }, { "seed": 900000, "sample": 1, "native_reward": 0.6181059146631999, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 261, "reward_channels": { "native": 0.6181059146631999, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "triangular_cycle" }, { "seed": 900000, "sample": 2, "native_reward": 0.61810598559111, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 258, "reward_channels": { "native": 0.61810598559111, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "triangular_cycle" }, { "seed": 900000, "sample": 3, "native_reward": 0.6181054220639031, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 259, "reward_channels": { "native": 0.6181054220639031, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "triangular_cycle" }, { "seed": 900001, "sample": 0, "native_reward": 0.6188066805953297, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 247, "reward_channels": { "native": 0.6188066805953297, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "triangular_cycle" }, { "seed": 900001, "sample": 1, "native_reward": 0.6188069888191968, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 264, "reward_channels": { "native": 0.6188069888191968, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "triangular_cycle" }, { "seed": 900001, "sample": 2, "native_reward": 0.618805890198689, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 302, "reward_channels": { "native": 0.618805890198689, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "triangular_cycle" }, { "seed": 900001, "sample": 3, "native_reward": 0.6188056475190211, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 233, "reward_channels": { "native": 0.6188056475190211, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "triangular_cycle" }, { "seed": 900000, "sample": 0, "native_reward": 0.25, "promotion_success": false, "done": false, "errors": 0, "tool_steps": 3, "model_tokens": 357, "reward_channels": { "native": 0.25, "efficiency": 0.0, "reliability": -1.0 }, "verification_level": null, "failed_gates": [ "finalized" ], "family": "competing_cycles" }, { "seed": 900000, "sample": 1, "native_reward": 0.25, "promotion_success": false, "done": false, "errors": 0, "tool_steps": 3, "model_tokens": 243, "reward_channels": { "native": 0.25, "efficiency": 0.0, "reliability": -1.0 }, "verification_level": null, "failed_gates": [ "finalized" ], "family": "competing_cycles" }, { "seed": 900000, "sample": 2, "native_reward": 0.25, "promotion_success": false, "done": false, "errors": 0, "tool_steps": 3, "model_tokens": 278, "reward_channels": { "native": 0.25, "efficiency": 0.0, "reliability": -1.0 }, "verification_level": null, "failed_gates": [ "finalized" ], "family": "competing_cycles" }, { "seed": 900000, "sample": 3, "native_reward": 0.25, "promotion_success": false, "done": false, "errors": 0, "tool_steps": 3, "model_tokens": 284, "reward_channels": { "native": 0.25, "efficiency": 0.0, "reliability": -1.0 }, "verification_level": null, "failed_gates": [ "finalized" ], "family": "competing_cycles" }, { "seed": 900001, "sample": 0, "native_reward": 0.25, "promotion_success": false, "done": false, "errors": 0, "tool_steps": 3, "model_tokens": 238, "reward_channels": { "native": 0.25, "efficiency": 0.0, "reliability": -1.0 }, "verification_level": null, "failed_gates": [ "finalized" ], "family": "competing_cycles" }, { "seed": 900001, "sample": 1, "native_reward": 0.25, "promotion_success": false, "done": false, "errors": 0, "tool_steps": 3, "model_tokens": 256, "reward_channels": { "native": 0.25, "efficiency": 0.0, "reliability": -1.0 }, "verification_level": null, "failed_gates": [ "finalized" ], "family": "competing_cycles" }, { "seed": 900001, "sample": 2, "native_reward": 0.25, "promotion_success": false, "done": false, "errors": 0, "tool_steps": 3, "model_tokens": 284, "reward_channels": { "native": 0.25, "efficiency": 0.0, "reliability": -1.0 }, "verification_level": null, "failed_gates": [ "finalized" ], "family": "competing_cycles" }, { "seed": 900001, "sample": 3, "native_reward": 0.25, "promotion_success": false, "done": false, "errors": 0, "tool_steps": 3, "model_tokens": 251, "reward_channels": { "native": 0.25, "efficiency": 0.0, "reliability": -1.0 }, "verification_level": null, "failed_gates": [ "finalized" ], "family": "competing_cycles" }, { "seed": 900000, "sample": 0, "native_reward": 0.6170014806407986, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 305, "reward_channels": { "native": 0.6170014806407986, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "four_hop_cycle" }, { "seed": 900000, "sample": 1, "native_reward": 0.6170025229249978, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 302, "reward_channels": { "native": 0.6170025229249978, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "four_hop_cycle" }, { "seed": 900000, "sample": 2, "native_reward": 0.6170025763343728, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 319, "reward_channels": { "native": 0.6170025763343728, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "four_hop_cycle" }, { "seed": 900000, "sample": 3, "native_reward": 0.6169999447661306, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 310, "reward_channels": { "native": 0.6169999447661306, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "four_hop_cycle" }, { "seed": 900001, "sample": 0, "native_reward": 0.6170328837317088, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 349, "reward_channels": { "native": 0.6170328837317088, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "four_hop_cycle" }, { "seed": 900001, "sample": 1, "native_reward": 0.617033260678955, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 286, "reward_channels": { "native": 0.617033260678955, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "four_hop_cycle" }, { "seed": 900001, "sample": 2, "native_reward": 0.6170340618311814, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 301, "reward_channels": { "native": 0.6170340618311814, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "four_hop_cycle" }, { "seed": 900001, "sample": 3, "native_reward": 0.6170338136804978, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 323, "reward_channels": { "native": 0.6170338136804978, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "four_hop_cycle" }, { "seed": 900000, "sample": 0, "native_reward": 0.25, "promotion_success": false, "done": false, "errors": 0, "tool_steps": 4, "model_tokens": 395, "reward_channels": { "native": 0.25, "efficiency": 0.0, "reliability": -1.0 }, "verification_level": null, "failed_gates": [ "finalized" ], "family": "size_sensitive_cycle" }, { "seed": 900000, "sample": 1, "native_reward": 0.8326690123331902, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 273, "reward_channels": { "native": 0.8326690123331902, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "size_sensitive_cycle" }, { "seed": 900000, "sample": 2, "native_reward": 0.8326679764228776, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 257, "reward_channels": { "native": 0.8326679764228776, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "size_sensitive_cycle" }, { "seed": 900000, "sample": 3, "native_reward": 0.8326674563266472, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 244, "reward_channels": { "native": 0.8326674563266472, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "size_sensitive_cycle" }, { "seed": 900001, "sample": 0, "native_reward": 0.831750250372806, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 267, "reward_channels": { "native": 0.831750250372806, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "size_sensitive_cycle" }, { "seed": 900001, "sample": 1, "native_reward": 0.8317517377728061, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 279, "reward_channels": { "native": 0.8317517377728061, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "size_sensitive_cycle" }, { "seed": 900001, "sample": 2, "native_reward": 0.8317508980261654, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 295, "reward_channels": { "native": 0.8317508980261654, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "size_sensitive_cycle" }, { "seed": 900001, "sample": 3, "native_reward": 0.8317511159244272, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 317, "reward_channels": { "native": 0.8317511159244272, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "size_sensitive_cycle" } ] }, { "run_id": "1789206045", "mode": "thinking", "method": "initial", "steps": 0, "reward_points": 0, "train_mean": null, "train_last10": null, "holdout_native": 0.6829987846910179, "holdout_episodes": 40, "holdout_successes": 37, "mean_step_seconds": null, "label": "Starting model", "state_at_cutoff": "complete", "step_hours": null, "peak_allocated_gib": null, "peak_reserved_gib": null, "extra_refill_rounds": 0, "groups_refilled": 0, "holdout_component_means": { "native": 0.6829987846910179, "efficiency": 0.47500000000000003, "reliability": 0.9 }, "configuration": { "model": { "id": "Qwen/Qwen3-14B", "revision": "40c069824f4251a91eefaf281ebe4c544efd3e18", "loss": "grpo", "algorithm": "grpo", "no_think": false, "thinking_budget_tokens": 2048, "lora": true, "quantize": true, "lora_r": 32, "lora_alpha": 64, "use_dora": false }, "loss": "grpo", "normalization": "joint", "entropy_control": { "method": "monitor", "epsilon_low": 0.2, "epsilon_high": 0.2, "zeta_initial": 0.001, "zeta_min": 0.0001, "zeta_max": 0.05, "holdout_split": "benchmark", "holdout_seeds_per_family": 2, "num_iterations": 2 }, "dynamic_sampling": false, "sampling": { "group_size": 4, "temperature": 1.0, "max_completion_tokens": 22528, "top_entropy_quantile": 1.0, "temperature_decay_to": null }, "learning_rate": 1e-05, "turn_max_tokens": 3072 }, "source_sha256": { "entropy_holdout_episodes.jsonl": "c7f822f8a8efb2c3874e80c1c84ebe5ed698a1cd94cf7d52a34ddedde00c43de", "resolved_training_config.json": "c3ce95171643c1ef9031ab4b7725f95cbac874cdbd24b5701e0f5d72e7faffdd" }, "training_steps": [], "holdout": [ { "seed": 900000, "sample": 0, "native_reward": 0.8591328836248219, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 5, "model_tokens": 5522, "reward_channels": { "native": 0.8591328836248219, "efficiency": 0.16666666666666663, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "two_pool_cycle" }, { "seed": 900000, "sample": 1, "native_reward": 0.6089882858614486, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 3125, "reward_channels": { "native": 0.6089882858614486, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "two_pool_cycle" }, { "seed": 900000, "sample": 2, "native_reward": 0.6089882858614486, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 2562, "reward_channels": { "native": 0.6089882858614486, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "two_pool_cycle" }, { "seed": 900000, "sample": 3, "native_reward": 0.6089882858614486, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 3074, "reward_channels": { "native": 0.6089882858614486, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "two_pool_cycle" }, { "seed": 900001, "sample": 0, "native_reward": 0.6093825392865522, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 2810, "reward_channels": { "native": 0.6093825392865522, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "two_pool_cycle" }, { "seed": 900001, "sample": 1, "native_reward": 0.6093825392865522, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 2722, "reward_channels": { "native": 0.6093825392865522, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "two_pool_cycle" }, { "seed": 900001, "sample": 2, "native_reward": 0.8523284576243806, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 3, "model_tokens": 3715, "reward_channels": { "native": 0.8523284576243806, "efficiency": 0.5, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "two_pool_cycle" }, { "seed": 900001, "sample": 3, "native_reward": 0.8388160328689895, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 5, "model_tokens": 5088, "reward_channels": { "native": 0.8388160328689895, "efficiency": 0.16666666666666663, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "two_pool_cycle" }, { "seed": 900000, "sample": 0, "native_reward": 0.8204653155141224, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 4035, "reward_channels": { "native": 0.8204653155141224, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "triangular_cycle" }, { "seed": 900000, "sample": 1, "native_reward": 0.6106601460847819, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 2998, "reward_channels": { "native": 0.6106601460847819, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "triangular_cycle" }, { "seed": 900000, "sample": 2, "native_reward": 0.6106601460847819, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 3183, "reward_channels": { "native": 0.6106601460847819, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "triangular_cycle" }, { "seed": 900000, "sample": 3, "native_reward": 0.6106601460847819, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 3891, "reward_channels": { "native": 0.6106601460847819, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "triangular_cycle" }, { "seed": 900001, "sample": 0, "native_reward": 0.844761535440721, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 6, "model_tokens": 5955, "reward_channels": { "native": 0.844761535440721, "efficiency": 0.0, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "triangular_cycle" }, { "seed": 900001, "sample": 1, "native_reward": 0.9758783046350626, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 5, "model_tokens": 4277, "reward_channels": { "native": 0.9758783046350626, "efficiency": 0.16666666666666663, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "triangular_cycle" }, { "seed": 900001, "sample": 2, "native_reward": 0.25, "promotion_success": false, "done": false, "errors": 0, "tool_steps": 6, "model_tokens": 2884, "reward_channels": { "native": 0.25, "efficiency": 0.0, "reliability": -1.0 }, "verification_level": null, "failed_gates": [ "finalized" ], "family": "triangular_cycle" }, { "seed": 900001, "sample": 3, "native_reward": 0.6108851248150757, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 2925, "reward_channels": { "native": 0.6108851248150757, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "triangular_cycle" }, { "seed": 900000, "sample": 0, "native_reward": 0.6093677238672119, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 5, "model_tokens": 7711, "reward_channels": { "native": 0.6093677238672119, "efficiency": 0.16666666666666663, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "competing_cycles" }, { "seed": 900000, "sample": 1, "native_reward": 0.6140552238672119, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 2934, "reward_channels": { "native": 0.6140552238672119, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "competing_cycles" }, { "seed": 900000, "sample": 2, "native_reward": 0.7248594440979103, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 6550, "reward_channels": { "native": 0.7248594440979103, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "competing_cycles" }, { "seed": 900000, "sample": 3, "native_reward": 0.0, "promotion_success": false, "done": false, "errors": 0, "tool_steps": 0, "model_tokens": 0, "reward_channels": { "native": 0.0, "efficiency": 0.0, "reliability": -1.0 }, "verification_level": null, "failed_gates": null, "family": "competing_cycles" }, { "seed": 900001, "sample": 0, "native_reward": 0.175, "promotion_success": false, "done": true, "errors": 0, "tool_steps": 3, "model_tokens": 8190, "reward_channels": { "native": 0.175, "efficiency": 0.5, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [ "positive_net_profit", "stress_positive", "task_complete" ], "family": "competing_cycles" }, { "seed": 900001, "sample": 1, "native_reward": 0.7053628099527464, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 6703, "reward_channels": { "native": 0.7053628099527464, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "competing_cycles" }, { "seed": 900001, "sample": 2, "native_reward": 0.6512057439829095, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 3, "model_tokens": 5598, "reward_channels": { "native": 0.6512057439829095, "efficiency": 0.5, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "competing_cycles" }, { "seed": 900001, "sample": 3, "native_reward": 0.6145108982971254, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 3, "model_tokens": 5775, "reward_channels": { "native": 0.6145108982971254, "efficiency": 0.5, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "competing_cycles" }, { "seed": 900000, "sample": 0, "native_reward": 0.609376392538787, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 1571, "reward_channels": { "native": 0.609376392538787, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "four_hop_cycle" }, { "seed": 900000, "sample": 1, "native_reward": 0.609376392538787, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 1724, "reward_channels": { "native": 0.609376392538787, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "four_hop_cycle" }, { "seed": 900000, "sample": 2, "native_reward": 0.609376392538787, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 3047, "reward_channels": { "native": 0.609376392538787, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "four_hop_cycle" }, { "seed": 900000, "sample": 3, "native_reward": 0.609376392538787, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 3127, "reward_channels": { "native": 0.609376392538787, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "four_hop_cycle" }, { "seed": 900001, "sample": 0, "native_reward": 0.739978752793518, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 3, "model_tokens": 6199, "reward_channels": { "native": 0.739978752793518, "efficiency": 0.5, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "four_hop_cycle" }, { "seed": 900001, "sample": 1, "native_reward": 0.8461963597959128, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 4440, "reward_channels": { "native": 0.8461963597959128, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "four_hop_cycle" }, { "seed": 900001, "sample": 2, "native_reward": 0.8461963597959128, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 3475, "reward_channels": { "native": 0.8461963597959128, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "four_hop_cycle" }, { "seed": 900001, "sample": 3, "native_reward": 0.6095701077161815, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 3138, "reward_channels": { "native": 0.6095701077161815, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "four_hop_cycle" }, { "seed": 900000, "sample": 0, "native_reward": 0.8262556477476042, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 2136, "reward_channels": { "native": 0.8262556477476042, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "size_sensitive_cycle" }, { "seed": 900000, "sample": 1, "native_reward": 0.9870115723152173, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 5056, "reward_channels": { "native": 0.9870115723152173, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "size_sensitive_cycle" }, { "seed": 900000, "sample": 2, "native_reward": 0.7421847681444665, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 3, "model_tokens": 4983, "reward_channels": { "native": 0.7421847681444665, "efficiency": 0.5, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "size_sensitive_cycle" }, { "seed": 900000, "sample": 3, "native_reward": 0.8262556477476042, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 3457, "reward_channels": { "native": 0.8262556477476042, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "size_sensitive_cycle" }, { "seed": 900001, "sample": 0, "native_reward": 0.8245803237223959, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 3160, "reward_channels": { "native": 0.8245803237223959, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "size_sensitive_cycle" }, { "seed": 900001, "sample": 1, "native_reward": 0.8245803237223959, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 3082, "reward_channels": { "native": 0.8245803237223959, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "size_sensitive_cycle" }, { "seed": 900001, "sample": 2, "native_reward": 0.9607157572618796, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 2974, "reward_channels": { "native": 0.9607157572618796, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "size_sensitive_cycle" }, { "seed": 900001, "sample": 3, "native_reward": 0.8245803237223959, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 5269, "reward_channels": { "native": 0.8245803237223959, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "size_sensitive_cycle" } ] }, { "run_id": "1789214712", "mode": "thinking", "method": "grpo", "steps": 50, "reward_points": 25, "train_mean": 0.8361628222465515, "train_last10": 0.8823877453804017, "holdout_native": 0.67982993828857, "holdout_episodes": 40, "holdout_successes": 38, "mean_step_seconds": 929.1267989518816, "label": "GRPO", "state_at_cutoff": "complete", "step_hours": 12.904538874331687, "peak_allocated_gib": 69.07209730148315, "peak_reserved_gib": 76.8828125, "extra_refill_rounds": 0.0, "groups_refilled": 0.0, "holdout_component_means": { "native": 0.67982993828857, "efficiency": 0.47500000000000003, "reliability": 0.9 }, "configuration": { "model": { "id": "Qwen/Qwen3-14B", "revision": "40c069824f4251a91eefaf281ebe4c544efd3e18", "loss": "grpo", "algorithm": "grpo", "no_think": false, "thinking_budget_tokens": 2048, "lora": true, "quantize": true, "lora_r": 32, "lora_alpha": 64, "use_dora": false }, "loss": "grpo", "normalization": "joint", "entropy_control": { "method": "monitor", "epsilon_low": 0.2, "epsilon_high": 0.2, "zeta_initial": 0.001, "zeta_min": 0.0001, "zeta_max": 0.05, "holdout_split": "benchmark", "holdout_seeds_per_family": 2, "num_iterations": 2 }, "dynamic_sampling": false, "sampling": { "group_size": 4, "temperature": 1.0, "max_completion_tokens": 22528, "top_entropy_quantile": 1.0, "temperature_decay_to": null }, "learning_rate": 1e-05, "turn_max_tokens": 3072 }, "source_sha256": { "metrics.jsonl": "1d1057f4bd367542b8b9ba65b0caffe5d8e63ad940b5455a1a2a84bcd6b6fe52", "entropy_holdout_episodes.jsonl": "6c178b921e99405eed0a9e9ac490a957959ba3f659cdd16cbbd274301c59eeb6", "resolved_training_config.json": "0cef43498f7ff664838a990e829bf9c4051c37d27a688a8973e0c7aeec07b616" }, "training_steps": [ { "step": 1, "loss": -0.21574446558952332, "grad_norm": 0.061767578125, "learning_rate": 1e-05, "thinking/tokens_mean": 606.2361111111111, "thinking/budget_hit_fraction": 0.14583333333333334, "num_tokens": 51975.0, "completions/mean_length": 3604.375, "completions/min_length": 2890.0, "completions/max_length": 6014.0, "completions/clipped_ratio": 0.25, "completions/mean_terminated_length": 3782.166748046875, "completions/min_terminated_length": 2890.0, "completions/max_terminated_length": 6014.0, "tools/call_frequency": 2.25, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.7262155413627625, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 0.13481803238391876, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.4583333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.2500000298023224, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/mean": 0.45231834053993225, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/std": 0.526568591594696, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/mean": 0.2916666865348816, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/std": 0.34359216690063477, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/mean": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/std": 1.154700517654419, "reward": 0.7080169916152954, "reward_std": 0.5734051465988159, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 0.5, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.14539151638746262, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 1323.271556528, "entropy_control/entropy": 0.20009788357278327, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "epoch": 0.004, "adapter_sync_time": 0.0, "generation_time": 1131.710895, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1166.291479, "forward_backward_time": 156.873074, "rollout_overhead_time": 34.580584, "peak_allocated_gib": 46.211081981658936, "peak_reserved_gib": 51.38671875 }, { "step": 2, "loss": -0.21584278345108032, "grad_norm": 0.05908203125, "learning_rate": 9.800000000000001e-06, "entropy": 0.14546925388276577, "clip_ratio/low_mean": 0.00041601009434089065, "clip_ratio/low_min": 0.00041601009434089065, "clip_ratio/high_mean": 0.002612101990962401, "clip_ratio/high_max": 0.002612101990962401, "clip_ratio/region_mean": 0.0030281120853032917, "step_time": 156.92217275399935, "entropy_control/entropy": 0.2002554157113976, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "epoch": 0.008, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 156.814344, "rollout_overhead_time": 0.0, "peak_allocated_gib": 46.689598083496094, "peak_reserved_gib": 51.484375, "reward": null }, { "step": 3, "loss": 5.21540641784668e-08, "grad_norm": 0.06884765625, "learning_rate": 9.600000000000001e-06, "thinking/tokens_mean": 679.5833333333333, "thinking/budget_hit_fraction": 0.175, "num_tokens": 102514.0, "completions/mean_length": 3605.375, "completions/min_length": 2236.0, "completions/max_length": 5501.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 3605.375, "completions/min_terminated_length": 2236.0, "completions/max_terminated_length": 5501.0, "tools/call_frequency": 2.5, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.7024504542350769, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 0.12899158895015717, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.5833333730697632, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.12598817050457, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 0.0, "reward": 0.9316171407699585, "reward_std": 0.12343566864728928, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.21422390453517437, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 1171.6926818000006, "entropy_control/entropy": 0.22008486084324633, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "epoch": 0.012, "adapter_sync_time": 0.0, "generation_time": 1006.966434, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1036.049033, "forward_backward_time": 135.534431, "rollout_overhead_time": 29.0826, "peak_allocated_gib": 46.689598083496094, "peak_reserved_gib": 51.51171875 }, { "step": 4, "loss": 9.134411811828613e-05, "grad_norm": 0.064453125, "learning_rate": 9.4e-06, "entropy": 0.2134654000401497, "clip_ratio/low_mean": 0.0019264514558017254, "clip_ratio/low_min": 0.0019264514558017254, "clip_ratio/high_mean": 0.0014562738069798797, "clip_ratio/high_max": 0.0014562738069798797, "clip_ratio/region_mean": 0.003382725262781605, "step_time": 135.64331109700015, "entropy_control/entropy": 0.2193793956353516, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "epoch": 0.016, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 135.532583, "rollout_overhead_time": 0.0, "peak_allocated_gib": 46.689598083496094, "peak_reserved_gib": 51.51171875, "reward": null }, { "step": 5, "loss": -1.4901161193847656e-08, "grad_norm": 0.056884765625, "learning_rate": 9.200000000000002e-06, "thinking/tokens_mean": 591.5535714285714, "thinking/budget_hit_fraction": 0.16071428571428573, "num_tokens": 157826.0, "completions/mean_length": 4209.0, "completions/min_length": 2236.0, "completions/max_length": 7215.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 4209.0, "completions/min_terminated_length": 2236.0, "completions/max_terminated_length": 7215.0, "tools/call_frequency": 3.25, "tools/failure_frequency": 0.19230769574642181, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.6130115389823914, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 0.20563623309135437, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.5625, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.1767766922712326, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 0.0, "reward": 0.8411365747451782, "reward_std": 0.20309090614318848, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.19480029679834843, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 1640.3757725910004, "entropy_control/entropy": 0.19811476079643126, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "epoch": 0.02, "adapter_sync_time": 0.0, "generation_time": 1441.471157, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1475.959354, "forward_backward_time": 164.32011, "rollout_overhead_time": 34.488197, "peak_allocated_gib": 48.77573108673096, "peak_reserved_gib": 54.79296875 }, { "step": 6, "loss": -0.00023845583200454712, "grad_norm": 0.058837890625, "learning_rate": 9e-06, "entropy": 0.19437631219625473, "clip_ratio/low_mean": 0.0007464685040758923, "clip_ratio/low_min": 0.0007464685040758923, "clip_ratio/high_mean": 0.00285797004471533, "clip_ratio/high_max": 0.00285797004471533, "clip_ratio/region_mean": 0.0036044385487912223, "step_time": 164.45583199400062, "entropy_control/entropy": 0.19769418170238393, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "epoch": 0.024, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 164.354975, "rollout_overhead_time": 0.0, "peak_allocated_gib": 48.77573108673096, "peak_reserved_gib": 54.79296875, "reward": null }, { "step": 7, "loss": -2.2351741790771484e-08, "grad_norm": 0.0751953125, "learning_rate": 8.8e-06, "thinking/tokens_mean": 618.3360544217687, "thinking/budget_hit_fraction": 0.1096938775510204, "num_tokens": 223699.0, "completions/mean_length": 4766.625, "completions/min_length": 1827.0, "completions/max_length": 6732.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 4766.625, "completions/min_terminated_length": 1827.0, "completions/max_terminated_length": 6732.0, "tools/call_frequency": 4.375, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/mean": 0.4570293426513672, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/std": 0.2840198874473572, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/mean": 0.125, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/std": 0.15957120060920715, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/mean": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/std": 1.154700517654419, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/mean": 0.8914890885353088, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/std": 0.0759979858994484, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/mean": 0.4166666865348816, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/std": 0.2886751592159271, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/std": 0.0, "reward": 0.7878009080886841, "reward_std": 0.48873358964920044, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.21208408661186695, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 2296.4392685740004, "entropy_control/entropy": 0.2186813504668633, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "epoch": 0.028, "adapter_sync_time": 0.0, "generation_time": 2005.50317, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 2055.451308, "forward_backward_time": 240.881205, "rollout_overhead_time": 49.948137, "peak_allocated_gib": 60.44094133377075, "peak_reserved_gib": 66.74609375 }, { "step": 8, "loss": 2.8386712074279785e-06, "grad_norm": 0.07373046875, "learning_rate": 8.6e-06, "entropy": 0.21231159567832947, "clip_ratio/low_mean": 0.000925836808164604, "clip_ratio/low_min": 0.000925836808164604, "clip_ratio/high_mean": 0.003548822656739503, "clip_ratio/high_max": 0.003548822656739503, "clip_ratio/region_mean": 0.004474659464904107, "step_time": 240.87562695600081, "entropy_control/entropy": 0.21879905398527097, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "epoch": 0.032, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 240.788066, "rollout_overhead_time": 0.0, "peak_allocated_gib": 60.44094133377075, "peak_reserved_gib": 66.74609375, "reward": null }, { "step": 9, "loss": -0.18661607801914215, "grad_norm": 0.052001953125, "learning_rate": 8.400000000000001e-06, "thinking/tokens_mean": 743.3154761904761, "thinking/budget_hit_fraction": 0.23214285714285712, "num_tokens": 280527.0, "completions/mean_length": 3819.5, "completions/min_length": 2234.0, "completions/max_length": 5961.0, "completions/clipped_ratio": 0.125, "completions/mean_terminated_length": 3926.4287109375, "completions/min_terminated_length": 2234.0, "completions/max_terminated_length": 5961.0, "tools/call_frequency": 2.375, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.6691635251045227, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 0.11923417448997498, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.625, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.0833333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/mean": 0.504158616065979, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/std": 0.34144195914268494, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/mean": 0.3333333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/std": 0.27216553688049316, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/std": 1.0, "reward": 0.7606194019317627, "reward_std": 0.3968806862831116, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 0.5, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.17925255000591278, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 1397.8388155039956, "entropy_control/entropy": 0.2161763805416, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "epoch": 0.036, "adapter_sync_time": 0.0, "generation_time": 1162.674265, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1203.702736, "forward_backward_time": 194.042967, "rollout_overhead_time": 41.028471, "peak_allocated_gib": 60.44094133377075, "peak_reserved_gib": 66.74609375 }, { "step": 10, "loss": -0.1866154670715332, "grad_norm": 0.052978515625, "learning_rate": 8.2e-06, "entropy": 0.1790898609906435, "clip_ratio/low_mean": 0.0010788068029796705, "clip_ratio/low_min": 0.0010788068029796705, "clip_ratio/high_mean": 0.0018942377355415374, "clip_ratio/high_max": 0.0018942377355415374, "clip_ratio/region_mean": 0.002973044538521208, "step_time": 194.0120085890012, "entropy_control/entropy": 0.21600092876164814, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "epoch": 0.04, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 193.926184, "rollout_overhead_time": 0.0, "peak_allocated_gib": 60.44094133377075, "peak_reserved_gib": 66.74609375, "reward": null }, { "step": 11, "loss": -7.450580596923828e-08, "grad_norm": 0.060302734375, "learning_rate": 8.000000000000001e-06, "thinking/tokens_mean": 635.5285714285714, "thinking/budget_hit_fraction": 0.17857142857142858, "num_tokens": 345071.0, "completions/mean_length": 4759.0, "completions/min_length": 2510.0, "completions/max_length": 6892.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 4759.0, "completions/min_terminated_length": 2510.0, "completions/max_terminated_length": 6892.0, "tools/call_frequency": 3.375, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__native/mean": 0.6681309938430786, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__native/std": 0.09295085072517395, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__efficiency/mean": 0.4583333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__efficiency/std": 0.15957120060920715, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/mean": 0.6374233365058899, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/std": 0.2812902331352234, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/mean": 0.4166666865348816, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/std": 0.3191424012184143, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/std": 1.0, "reward": 0.8246521353721619, "reward_std": 0.33009234070777893, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.20426424033939838, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 1640.2230279159976, "entropy_control/entropy": 0.20939805002587633, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "epoch": 0.044, "adapter_sync_time": 0.0, "generation_time": 1393.853113, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1436.092919, "forward_backward_time": 204.02604, "rollout_overhead_time": 42.239806, "peak_allocated_gib": 60.44094133377075, "peak_reserved_gib": 66.74609375 }, { "step": 12, "loss": 0.0002558119595050812, "grad_norm": 0.056884765625, "learning_rate": 7.800000000000002e-06, "entropy": 0.2042933963239193, "clip_ratio/low_mean": 0.0008133799565257505, "clip_ratio/low_min": 0.0008133799565257505, "clip_ratio/high_mean": 0.0018578044255264103, "clip_ratio/high_max": 0.0018578044255264103, "clip_ratio/region_mean": 0.002671184382052161, "step_time": 204.0847907480038, "entropy_control/entropy": 0.20943264558814864, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "epoch": 0.048, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 203.983037, "rollout_overhead_time": 0.0, "peak_allocated_gib": 60.44094133377075, "peak_reserved_gib": 66.74609375, "reward": null }, { "step": 13, "loss": -1.2665987014770508e-06, "grad_norm": 0.0556640625, "learning_rate": 7.600000000000001e-06, "thinking/tokens_mean": 696.3249999999999, "thinking/budget_hit_fraction": 0.16071428571428573, "num_tokens": 409316.0, "completions/mean_length": 4745.625, "completions/min_length": 2706.0, "completions/max_length": 6542.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 4745.625, "completions/min_terminated_length": 2706.0, "completions/max_terminated_length": 6542.0, "tools/call_frequency": 3.25, "tools/failure_frequency": 0.07692307978868484, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.6175202131271362, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 0.015931161120533943, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.625, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.0833333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/mean": 0.6163814663887024, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/std": 0.28774014115333557, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/mean": 0.2083333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/std": 0.25, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/std": 1.0, "reward": 0.7877841591835022, "reward_std": 0.3165474534034729, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.21027977019548416, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 1913.5732425510032, "entropy_control/entropy": 0.20998974643007895, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "epoch": 0.052, "adapter_sync_time": 0.0, "generation_time": 1568.566726, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1626.641259, "forward_backward_time": 286.821186, "rollout_overhead_time": 58.074533, "peak_allocated_gib": 66.42686176300049, "peak_reserved_gib": 73.69921875 }, { "step": 14, "loss": 0.0003102123737335205, "grad_norm": 0.0556640625, "learning_rate": 7.4e-06, "entropy": 0.20994763262569904, "clip_ratio/low_mean": 0.0009998756722779945, "clip_ratio/low_min": 0.0009998756722779945, "clip_ratio/high_mean": 0.002319482999155298, "clip_ratio/high_max": 0.002319482999155298, "clip_ratio/region_mean": 0.0033193586714332923, "step_time": 286.9272500799998, "entropy_control/entropy": 0.20963220479580932, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "epoch": 0.056, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 286.812957, "rollout_overhead_time": 0.0, "peak_allocated_gib": 66.42686176300049, "peak_reserved_gib": 73.69921875, "reward": null }, { "step": 15, "loss": -0.10666165500879288, "grad_norm": 0.060546875, "learning_rate": 7.2000000000000005e-06, "thinking/tokens_mean": 781.7341836734694, "thinking/budget_hit_fraction": 0.18367346938775508, "num_tokens": 478171.0, "completions/mean_length": 5123.375, "completions/min_length": 3286.0, "completions/max_length": 8884.0, "completions/clipped_ratio": 0.125, "completions/mean_terminated_length": 5385.857421875, "completions/min_terminated_length": 3309.0, "completions/max_terminated_length": 8884.0, "tools/call_frequency": 3.375, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/mean": 0.45868855714797974, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/std": 0.3439096212387085, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/mean": 0.1666666716337204, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/std": 0.235702246427536, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/mean": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/std": 1.154700517654419, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/mean": 0.8778764009475708, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/std": 0.06689149886369705, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/std": 0.19245010614395142, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/std": 0.0, "reward": 0.7849491834640503, "reward_std": 0.5093115568161011, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.14890395756810904, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 2235.668606463003, "entropy_control/entropy": 0.17151553053826665, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "epoch": 0.06, "adapter_sync_time": 0.0, "generation_time": 1889.736014, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1948.062797, "forward_backward_time": 287.493484, "rollout_overhead_time": 58.326782, "peak_allocated_gib": 66.76106929779053, "peak_reserved_gib": 74.265625 }, { "step": 16, "loss": -0.10667812824249268, "grad_norm": 0.05224609375, "learning_rate": 7e-06, "entropy": 0.14920903742313385, "clip_ratio/low_mean": 0.0006815678352722898, "clip_ratio/low_min": 0.0006815678352722898, "clip_ratio/high_mean": 0.0020250110828783363, "clip_ratio/high_max": 0.0020250110828783363, "clip_ratio/region_mean": 0.002706578918150626, "step_time": 287.58674787699965, "entropy_control/entropy": 0.1718210866764176, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "epoch": 0.064, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 287.47271, "rollout_overhead_time": 0.0, "peak_allocated_gib": 66.76106929779053, "peak_reserved_gib": 74.265625, "reward": null }, { "step": 17, "loss": 2.7939677238464355e-07, "grad_norm": 0.056640625, "learning_rate": 6.800000000000001e-06, "thinking/tokens_mean": 718.342857142857, "thinking/budget_hit_fraction": 0.14285714285714285, "num_tokens": 550619.0, "completions/mean_length": 5365.0, "completions/min_length": 2833.0, "completions/max_length": 7500.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 5365.0, "completions/min_terminated_length": 2833.0, "completions/max_terminated_length": 7500.0, "tools/call_frequency": 4.125, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/mean": 0.5120359063148499, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/std": 0.22543221712112427, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/mean": 0.25, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/std": 0.21516574919223785, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/std": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/mean": 0.7050560116767883, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/std": 0.1191813126206398, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/mean": 0.375, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/std": 0.34359216690063477, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/std": 0.0, "reward": 0.7741709351539612, "reward_std": 0.3329608142375946, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.18250237219035625, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 2501.553329219005, "entropy_control/entropy": 0.17966567875644504, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "epoch": 0.068, "adapter_sync_time": 0.0, "generation_time": 2181.071163, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 2235.57908, "forward_backward_time": 265.860784, "rollout_overhead_time": 54.507917, "peak_allocated_gib": 66.76106929779053, "peak_reserved_gib": 74.265625 }, { "step": 18, "loss": 5.6877732276916504e-05, "grad_norm": 0.05419921875, "learning_rate": 6.600000000000001e-06, "entropy": 0.1821766123175621, "clip_ratio/low_mean": 0.0008115740347420797, "clip_ratio/low_min": 0.0008115740347420797, "clip_ratio/high_mean": 0.0024077362031675875, "clip_ratio/high_max": 0.0024077362031675875, "clip_ratio/region_mean": 0.003219310237909667, "step_time": 266.0061489889995, "entropy_control/entropy": 0.1793118971963076, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "epoch": 0.072, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 265.896535, "rollout_overhead_time": 0.0, "peak_allocated_gib": 66.76106929779053, "peak_reserved_gib": 74.265625, "reward": null }, { "step": 19, "loss": -0.316281795501709, "grad_norm": 0.05078125, "learning_rate": 6.4000000000000006e-06, "thinking/tokens_mean": 499.7142857142857, "thinking/budget_hit_fraction": 0.10714285714285714, "num_tokens": 600620.0, "completions/mean_length": 3359.625, "completions/min_length": 1970.0, "completions/max_length": 4846.0, "completions/clipped_ratio": 0.25, "completions/mean_terminated_length": 3455.83349609375, "completions/min_terminated_length": 1970.0, "completions/max_terminated_length": 4846.0, "tools/call_frequency": 2.375, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.4572652578353882, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 0.3048435151576996, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.3333333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/mean": 0.5300920009613037, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/std": 0.4792630672454834, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/mean": 0.2083333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/std": 0.25, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/mean": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/std": 1.154700517654419, "reward": 0.5613870024681091, "reward_std": 0.5816197991371155, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 0.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.14770432747900486, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 1148.603929220004, "entropy_control/entropy": 0.1939812108135933, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "epoch": 0.076, "adapter_sync_time": 0.0, "generation_time": 967.942453, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 999.918835, "forward_backward_time": 148.579143, "rollout_overhead_time": 31.976382, "peak_allocated_gib": 66.76106929779053, "peak_reserved_gib": 74.265625 }, { "step": 20, "loss": -0.31621694564819336, "grad_norm": 0.06689453125, "learning_rate": 6.200000000000001e-06, "entropy": 0.14768492057919502, "clip_ratio/low_mean": 0.00045045046135783195, "clip_ratio/low_min": 0.00045045046135783195, "clip_ratio/high_mean": 0.0038841489295009524, "clip_ratio/high_max": 0.0038841489295009524, "clip_ratio/region_mean": 0.004334599390858784, "step_time": 148.67559428100503, "entropy_control/entropy": 0.19398792214377067, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "epoch": 0.08, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 148.56869, "rollout_overhead_time": 0.0, "peak_allocated_gib": 66.76106929779053, "peak_reserved_gib": 74.265625, "reward": null }, { "step": 21, "loss": -1.4156103134155273e-07, "grad_norm": 0.06982421875, "learning_rate": 6e-06, "thinking/tokens_mean": 643.1464285714285, "thinking/budget_hit_fraction": 0.07142857142857142, "num_tokens": 671578.0, "completions/mean_length": 5170.75, "completions/min_length": 3667.0, "completions/max_length": 7026.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 5170.75, "completions/min_terminated_length": 3667.0, "completions/max_terminated_length": 7026.0, "tools/call_frequency": 4.25, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/mean": 0.5778571367263794, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/std": 0.22537729144096375, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/mean": 0.2916666865348816, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/std": 0.25, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/std": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/mean": 0.7966060638427734, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/std": 0.0518668107688427, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/mean": 0.2916666865348816, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/std": 0.15957118570804596, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/std": 0.0, "reward": 0.851814866065979, "reward_std": 0.33230361342430115, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.19175784476101398, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 2313.628858852, "entropy_control/entropy": 0.19590833289032367, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "epoch": 0.084, "adapter_sync_time": 0.0, "generation_time": 2021.380127, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 2071.198077, "forward_backward_time": 242.313947, "rollout_overhead_time": 49.81795, "peak_allocated_gib": 66.76106929779053, "peak_reserved_gib": 74.265625 }, { "step": 22, "loss": 0.00027307868003845215, "grad_norm": 0.0673828125, "learning_rate": 5.8e-06, "entropy": 0.19180474057793617, "clip_ratio/low_mean": 0.0010520591167733073, "clip_ratio/low_min": 0.0010520591167733073, "clip_ratio/high_mean": 0.002789115416817367, "clip_ratio/high_max": 0.002789115416817367, "clip_ratio/region_mean": 0.0038411745335906744, "step_time": 242.45853826799794, "entropy_control/entropy": 0.19600329028751037, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "epoch": 0.088, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 242.338485, "rollout_overhead_time": 0.0, "peak_allocated_gib": 66.76106929779053, "peak_reserved_gib": 74.265625, "reward": null }, { "step": 23, "loss": -0.18633608520030975, "grad_norm": 0.057373046875, "learning_rate": 5.600000000000001e-06, "thinking/tokens_mean": 625.1433673469388, "thinking/budget_hit_fraction": 0.125, "num_tokens": 727140.0, "completions/mean_length": 4059.75, "completions/min_length": 2731.0, "completions/max_length": 6460.0, "completions/clipped_ratio": 0.125, "completions/mean_terminated_length": 4201.0, "completions/min_terminated_length": 2731.0, "completions/max_terminated_length": 6460.0, "tools/call_frequency": 2.875, "tools/failure_frequency": 0.043478261679410934, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.8164325952529907, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 0.14929784834384918, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.3750000298023224, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.28463754057884216, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/mean": 0.6595636606216431, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/std": 0.4462873339653015, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/mean": 0.4583333730697632, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/std": 0.3154948949813843, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/std": 1.0, "reward": 0.9088314771652222, "reward_std": 0.4608263373374939, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 0.5, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.16791206691414118, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 1439.073491322004, "entropy_control/entropy": 0.19424822948180198, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "epoch": 0.092, "adapter_sync_time": 0.0, "generation_time": 1164.343427, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1210.54189, "forward_backward_time": 228.425522, "rollout_overhead_time": 46.198463, "peak_allocated_gib": 66.76106929779053, "peak_reserved_gib": 74.265625 }, { "step": 24, "loss": -0.186231330037117, "grad_norm": 0.055908203125, "learning_rate": 5.400000000000001e-06, "entropy": 0.16807778552174568, "clip_ratio/low_mean": 0.00022885389626026154, "clip_ratio/low_min": 0.00022885389626026154, "clip_ratio/high_mean": 0.003561089630238712, "clip_ratio/high_max": 0.003561089630238712, "clip_ratio/region_mean": 0.0037899435264989734, "step_time": 228.50225764999777, "entropy_control/entropy": 0.1943999188217632, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "epoch": 0.096, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 228.405244, "rollout_overhead_time": 0.0, "peak_allocated_gib": 66.76106929779053, "peak_reserved_gib": 74.265625, "reward": null }, { "step": 25, "loss": -0.18551751971244812, "grad_norm": 0.058349609375, "learning_rate": 5.2e-06, "thinking/tokens_mean": 582.3472222222223, "thinking/budget_hit_fraction": 0.08333333333333333, "num_tokens": 778744.0, "completions/mean_length": 3360.5, "completions/min_length": 1891.0, "completions/max_length": 5791.0, "completions/clipped_ratio": 0.125, "completions/mean_terminated_length": 3401.857177734375, "completions/min_terminated_length": 1891.0, "completions/max_terminated_length": 5791.0, "tools/call_frequency": 2.75, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.6715044975280762, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 0.12447243928909302, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.5833333730697632, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.1666666716337204, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/mean": 0.5514910221099854, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/std": 0.3796755373477936, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/mean": 0.25, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/std": 0.2886751592159271, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/std": 1.0, "reward": 0.782331109046936, "reward_std": 0.4086287319660187, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 0.5, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.1782005950808525, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 1346.7842622450007, "entropy_control/entropy": 0.1990801086140548, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "epoch": 0.1, "adapter_sync_time": 0.0, "generation_time": 1117.828152, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1157.694675, "forward_backward_time": 188.982338, "rollout_overhead_time": 39.866522, "peak_allocated_gib": 66.76106929779053, "peak_reserved_gib": 74.265625 }, { "step": 26, "loss": -0.18542535603046417, "grad_norm": 0.057373046875, "learning_rate": 5e-06, "entropy": 0.17806172370910645, "clip_ratio/low_mean": 0.001531724352389574, "clip_ratio/low_min": 0.001531724352389574, "clip_ratio/high_mean": 0.0026588996406644583, "clip_ratio/high_max": 0.0026588996406644583, "clip_ratio/region_mean": 0.004190623993054032, "step_time": 189.0860957010009, "entropy_control/entropy": 0.198946864873294, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "epoch": 0.104, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 188.981787, "rollout_overhead_time": 0.0, "peak_allocated_gib": 66.76106929779053, "peak_reserved_gib": 74.265625, "reward": null }, { "step": 27, "loss": 3.427267074584961e-07, "grad_norm": 0.05810546875, "learning_rate": 4.800000000000001e-06, "thinking/tokens_mean": 752.9513888888888, "thinking/budget_hit_fraction": 0.14583333333333334, "num_tokens": 844036.0, "completions/mean_length": 4859.0, "completions/min_length": 2609.0, "completions/max_length": 6061.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 4859.0, "completions/min_terminated_length": 2609.0, "completions/max_terminated_length": 6061.0, "tools/call_frequency": 3.5, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/mean": 0.6995120048522949, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/std": 0.11226262152194977, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/std": 0.235702246427536, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/mean": 0.9193856716156006, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/std": 0.07096325606107712, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/mean": 0.3333333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/std": 0.13608276844024658, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/std": 0.0, "reward": 1.0302821397781372, "reward_std": 0.13856662809848785, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.19423643127083778, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 1773.342510265993, "entropy_control/entropy": 0.1916025433434907, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "epoch": 0.108, "adapter_sync_time": 0.0, "generation_time": 1524.595898, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1567.692393, "forward_backward_time": 205.54728, "rollout_overhead_time": 43.096495, "peak_allocated_gib": 66.76106929779053, "peak_reserved_gib": 74.265625 }, { "step": 28, "loss": 3.993511199951172e-06, "grad_norm": 0.058349609375, "learning_rate": 4.600000000000001e-06, "entropy": 0.19427297078073025, "clip_ratio/low_mean": 0.0007872360074543394, "clip_ratio/low_min": 0.0007872360074543394, "clip_ratio/high_mean": 0.002591346390545368, "clip_ratio/high_max": 0.002591346390545368, "clip_ratio/region_mean": 0.0033785823979997076, "step_time": 205.62199108300047, "entropy_control/entropy": 0.19155715934777284, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "epoch": 0.112, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 205.527939, "rollout_overhead_time": 0.0, "peak_allocated_gib": 66.76106929779053, "peak_reserved_gib": 74.265625, "reward": null }, { "step": 29, "loss": -3.725290298461914e-08, "grad_norm": 0.095703125, "learning_rate": 4.4e-06, "thinking/tokens_mean": 692.3154761904761, "thinking/budget_hit_fraction": 0.10714285714285714, "num_tokens": 911325.0, "completions/mean_length": 5128.625, "completions/min_length": 2756.0, "completions/max_length": 6932.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 5128.625, "completions/min_terminated_length": 2756.0, "completions/max_terminated_length": 6932.0, "tools/call_frequency": 4.125, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.735346257686615, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 0.14634527266025543, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.19245010614395142, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/mean": 0.5230018496513367, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/std": 0.32128164172172546, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/mean": 0.125, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/std": 0.15957120060920715, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/mean": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/std": 1.154700517654419, "reward": 0.7447990775108337, "reward_std": 0.4406716227531433, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.20591633208096027, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 2090.9175203540035, "entropy_control/entropy": 0.203890279917937, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "epoch": 0.116, "adapter_sync_time": 0.0, "generation_time": 1784.526045, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1836.761119, "forward_backward_time": 254.058516, "rollout_overhead_time": 52.235075, "peak_allocated_gib": 66.76106929779053, "peak_reserved_gib": 74.265625 }, { "step": 30, "loss": 0.00021416693925857544, "grad_norm": 0.058349609375, "learning_rate": 4.2000000000000004e-06, "entropy": 0.2061260398477316, "clip_ratio/low_mean": 0.0010790508677018806, "clip_ratio/low_min": 0.0010790508677018806, "clip_ratio/high_mean": 0.002600030682515353, "clip_ratio/high_max": 0.002600030682515353, "clip_ratio/region_mean": 0.0036790815502172336, "step_time": 254.11358457000097, "entropy_control/entropy": 0.2040441510170926, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "epoch": 0.12, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 254.021738, "rollout_overhead_time": 0.0, "peak_allocated_gib": 66.76106929779053, "peak_reserved_gib": 74.265625, "reward": null }, { "step": 31, "loss": -4.470348358154297e-08, "grad_norm": 0.06982421875, "learning_rate": 4.000000000000001e-06, "thinking/tokens_mean": 724.2827380952381, "thinking/budget_hit_fraction": 0.16666666666666666, "num_tokens": 974705.0, "completions/mean_length": 4839.0, "completions/min_length": 3435.0, "completions/max_length": 7366.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 4839.0, "completions/min_terminated_length": 3435.0, "completions/max_terminated_length": 7366.0, "tools/call_frequency": 3.875, "tools/failure_frequency": 0.06451612710952759, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/mean": 0.8607076406478882, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/std": 0.06753762811422348, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/mean": 0.3958333730697632, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/std": 0.21707837283611298, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/std": 0.0, "reward": 1.0804994106292725, "reward_std": 0.061188261955976486, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.16757085174322128, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 1802.356435463993, "entropy_control/entropy": 0.16794919033265657, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "epoch": 0.124, "adapter_sync_time": 0.0, "generation_time": 1547.406371, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1591.136157, "forward_backward_time": 211.123917, "rollout_overhead_time": 43.729786, "peak_allocated_gib": 66.76106929779053, "peak_reserved_gib": 74.265625 }, { "step": 32, "loss": 7.340312004089355e-05, "grad_norm": 0.0654296875, "learning_rate": 3.8000000000000005e-06, "entropy": 0.16734056919813156, "clip_ratio/low_mean": 0.0014973652141634375, "clip_ratio/low_min": 0.0014973652141634375, "clip_ratio/high_mean": 0.00299794934107922, "clip_ratio/high_max": 0.00299794934107922, "clip_ratio/region_mean": 0.004495314555242658, "step_time": 211.18617562500003, "entropy_control/entropy": 0.16773713178710217, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "epoch": 0.128, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 211.081678, "rollout_overhead_time": 0.0, "peak_allocated_gib": 66.76106929779053, "peak_reserved_gib": 74.265625, "reward": null }, { "step": 33, "loss": -5.364418029785156e-07, "grad_norm": 0.040771484375, "learning_rate": 3.6000000000000003e-06, "thinking/tokens_mean": 641.875, "thinking/budget_hit_fraction": 0.14583333333333334, "num_tokens": 1026740.0, "completions/mean_length": 3608.875, "completions/min_length": 1955.0, "completions/max_length": 6699.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 3608.875, "completions/min_terminated_length": 1955.0, "completions/max_terminated_length": 6699.0, "tools/call_frequency": 2.5, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.6095701456069946, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.6666666865348816, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/mean": 0.8578818440437317, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/std": 0.06588447093963623, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/std": 0.235702246427536, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/std": 0.0, "reward": 0.9628927111625671, "reward_std": 0.13321620225906372, "frac_reward_zero_std": 0.5, "frac_groups_zero_std": 0.5, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.18777185678482056, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 1280.3155918770062, "entropy_control/entropy": 0.19526047992356108, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "epoch": 0.132, "adapter_sync_time": 0.0, "generation_time": 1039.468172, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1080.972167, "forward_backward_time": 199.239674, "rollout_overhead_time": 41.503995, "peak_allocated_gib": 66.76106929779053, "peak_reserved_gib": 74.265625 }, { "step": 34, "loss": 0.00017583370208740234, "grad_norm": 0.037109375, "learning_rate": 3.4000000000000005e-06, "entropy": 0.18781633861362934, "clip_ratio/low_mean": 0.0009418768458999693, "clip_ratio/low_min": 0.0009418768458999693, "clip_ratio/high_mean": 0.0004291685181669891, "clip_ratio/high_max": 0.0004291685181669891, "clip_ratio/region_mean": 0.0013710453640669584, "step_time": 199.357629935992, "entropy_control/entropy": 0.19529972474706533, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "epoch": 0.136, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 199.256936, "rollout_overhead_time": 0.0, "peak_allocated_gib": 66.76106929779053, "peak_reserved_gib": 74.265625, "reward": null }, { "step": 35, "loss": -2.3096799850463867e-07, "grad_norm": 0.0712890625, "learning_rate": 3.2000000000000003e-06, "thinking/tokens_mean": 873.9464285714286, "thinking/budget_hit_fraction": 0.25, "num_tokens": 1103774.0, "completions/mean_length": 5766.25, "completions/min_length": 2389.0, "completions/max_length": 7394.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 5766.25, "completions/min_terminated_length": 2389.0, "completions/max_terminated_length": 7394.0, "tools/call_frequency": 3.75, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/mean": 0.6741191744804382, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/std": 0.05711929872632027, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/mean": 0.3750000298023224, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/std": 0.17251639068126678, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/std": 0.0, "reward": 0.8928691148757935, "reward_std": 0.0518985390663147, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.2012415211647749, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 2107.7353201419974, "entropy_control/entropy": 0.1924563170504371, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "epoch": 0.14, "adapter_sync_time": 0.0, "generation_time": 1804.905781, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1856.314271, "forward_backward_time": 251.322372, "rollout_overhead_time": 51.40849, "peak_allocated_gib": 66.76106929779053, "peak_reserved_gib": 74.265625 }, { "step": 36, "loss": -2.5227665901184082e-05, "grad_norm": 0.06689453125, "learning_rate": 3e-06, "entropy": 0.20116382464766502, "clip_ratio/low_mean": 0.0009656167385401204, "clip_ratio/low_min": 0.0009656167385401204, "clip_ratio/high_mean": 0.0029716521385125816, "clip_ratio/high_max": 0.0029716521385125816, "clip_ratio/region_mean": 0.003937268877052702, "step_time": 251.40516569999454, "entropy_control/entropy": 0.1923181859802659, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "epoch": 0.144, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 251.302596, "rollout_overhead_time": 0.0, "peak_allocated_gib": 66.76106929779053, "peak_reserved_gib": 74.265625, "reward": null }, { "step": 37, "loss": -0.18385322391986847, "grad_norm": 0.0625, "learning_rate": 2.8000000000000003e-06, "thinking/tokens_mean": 827.7901785714286, "thinking/budget_hit_fraction": 0.2544642857142857, "num_tokens": 1154849.0, "completions/mean_length": 3504.375, "completions/min_length": 2713.0, "completions/max_length": 4893.0, "completions/clipped_ratio": 0.125, "completions/mean_terminated_length": 3566.285888671875, "completions/min_terminated_length": 2713.0, "completions/max_terminated_length": 4893.0, "tools/call_frequency": 2.0, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.6686257123947144, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 0.11784413456916809, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.625, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.0833333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__native/mean": 0.5199053883552551, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__native/std": 0.3654637634754181, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__efficiency/mean": 0.4583333730697632, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__efficiency/std": 0.3154948949813843, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__reliability/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__reliability/std": 1.0, "reward": 0.7713488340377808, "reward_std": 0.4063173830509186, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 0.5, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.1783539429306984, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 1005.5948500160011, "entropy_control/entropy": 0.2105146749393795, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "epoch": 0.148, "adapter_sync_time": 0.0, "generation_time": 868.938258, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 893.617108, "forward_backward_time": 111.879814, "rollout_overhead_time": 24.678851, "peak_allocated_gib": 66.76106929779053, "peak_reserved_gib": 74.265625 }, { "step": 38, "loss": -0.18384769558906555, "grad_norm": 0.047607421875, "learning_rate": 2.6e-06, "entropy": 0.17824725061655045, "clip_ratio/low_mean": 0.0005780552310170606, "clip_ratio/low_min": 0.0005780552310170606, "clip_ratio/high_mean": 0.0020121496345382184, "clip_ratio/high_max": 0.0020121496345382184, "clip_ratio/region_mean": 0.002590204865555279, "step_time": 111.95662548500695, "entropy_control/entropy": 0.21051747600180507, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "epoch": 0.152, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 111.863829, "rollout_overhead_time": 0.0, "peak_allocated_gib": 66.76106929779053, "peak_reserved_gib": 74.265625, "reward": null }, { "step": 39, "loss": -1.2852251529693604e-07, "grad_norm": 0.072265625, "learning_rate": 2.4000000000000003e-06, "thinking/tokens_mean": 447.7023809523809, "thinking/budget_hit_fraction": 0.08928571428571429, "num_tokens": 1205242.0, "completions/mean_length": 3208.125, "completions/min_length": 1977.0, "completions/max_length": 4848.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 3208.125, "completions/min_terminated_length": 1977.0, "completions/max_terminated_length": 4848.0, "tools/call_frequency": 3.125, "tools/failure_frequency": 0.03999999910593033, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__native/mean": 0.6351134777069092, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__native/std": 0.20579981803894043, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__efficiency/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__efficiency/std": 0.25197634100914, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__reliability/mean": 0.75, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__reliability/std": 0.7071067690849304, "reward": 0.8101135492324829, "reward_std": 0.3326144516468048, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.18824629299342632, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 1055.1161288180156, "entropy_control/entropy": 0.18568970518463374, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "epoch": 0.156, "adapter_sync_time": 0.0, "generation_time": 814.417533, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 856.038607, "forward_backward_time": 198.978442, "rollout_overhead_time": 41.621073, "peak_allocated_gib": 66.76106929779053, "peak_reserved_gib": 74.265625 }, { "step": 40, "loss": 5.204230546951294e-05, "grad_norm": 0.0810546875, "learning_rate": 2.2e-06, "entropy": 0.1873590387403965, "clip_ratio/low_mean": 0.001982754183700308, "clip_ratio/low_min": 0.001982754183700308, "clip_ratio/high_mean": 0.0030543222092092037, "clip_ratio/high_max": 0.0030543222092092037, "clip_ratio/region_mean": 0.005037076392909512, "step_time": 199.11226085701492, "entropy_control/entropy": 0.18497066177414806, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "epoch": 0.16, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 199.009806, "rollout_overhead_time": 0.0, "peak_allocated_gib": 66.76106929779053, "peak_reserved_gib": 74.265625, "reward": null }, { "step": 41, "loss": -0.1802990734577179, "grad_norm": 0.0634765625, "learning_rate": 2.0000000000000003e-06, "thinking/tokens_mean": 610.0527210884354, "thinking/budget_hit_fraction": 0.12755102040816327, "num_tokens": 1267954.0, "completions/mean_length": 4362.0, "completions/min_length": 2725.0, "completions/max_length": 7129.0, "completions/clipped_ratio": 0.125, "completions/mean_terminated_length": 4546.4287109375, "completions/min_terminated_length": 2725.0, "completions/max_terminated_length": 7129.0, "tools/call_frequency": 3.5, "tools/failure_frequency": 0.0357142873108387, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__native/mean": 0.7108664512634277, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__native/std": 0.11576669663190842, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__efficiency/mean": 0.4583333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__efficiency/std": 0.2500000298023224, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/mean": 0.5521032810211182, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/std": 0.40590348839759827, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/mean": 0.1666666716337204, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/std": 0.235702246427536, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/std": 1.0, "reward": 0.797109842300415, "reward_std": 0.42240777611732483, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 0.5, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.16891460679471493, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 1965.8781912580016, "entropy_control/entropy": 0.1970858858765876, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "epoch": 0.164, "adapter_sync_time": 0.0, "generation_time": 1598.035267, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1659.726971, "forward_backward_time": 306.050515, "rollout_overhead_time": 61.691704, "peak_allocated_gib": 69.07209730148315, "peak_reserved_gib": 76.8828125 }, { "step": 42, "loss": -0.18048246204853058, "grad_norm": 0.055419921875, "learning_rate": 1.8000000000000001e-06, "entropy": 0.16886846721172333, "clip_ratio/low_mean": 0.0005234681011643261, "clip_ratio/low_min": 0.0005234681011643261, "clip_ratio/high_mean": 0.00373097465489991, "clip_ratio/high_max": 0.00373097465489991, "clip_ratio/region_mean": 0.004254442756064236, "step_time": 306.1554447550079, "entropy_control/entropy": 0.1969915345860917, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "epoch": 0.168, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 306.054867, "rollout_overhead_time": 0.0, "peak_allocated_gib": 69.07209730148315, "peak_reserved_gib": 76.8828125, "reward": null }, { "step": 43, "loss": -0.1847926378250122, "grad_norm": 0.08056640625, "learning_rate": 1.6000000000000001e-06, "thinking/tokens_mean": 497.8384353741497, "thinking/budget_hit_fraction": 0.10714285714285714, "num_tokens": 1321735.0, "completions/mean_length": 3418.625, "completions/min_length": 1189.0, "completions/max_length": 5326.0, "completions/clipped_ratio": 0.125, "completions/mean_terminated_length": 3468.285888671875, "completions/min_terminated_length": 1189.0, "completions/max_terminated_length": 5326.0, "tools/call_frequency": 2.625, "tools/failure_frequency": 0.0476190485060215, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__native/mean": 0.516849160194397, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__native/std": 0.3618776798248291, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__efficiency/mean": 0.3333333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__efficiency/std": 0.38490021228790283, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__reliability/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__reliability/std": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/mean": 0.6999523639678955, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/std": 0.1142006516456604, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/mean": 0.5833333730697632, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/std": 0.09622505307197571, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/std": 0.0, "reward": 0.7813174724578857, "reward_std": 0.40783098340034485, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 0.5, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.15408311132341623, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 1369.1788975260133, "entropy_control/entropy": 0.18647065540602978, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "epoch": 0.172, "adapter_sync_time": 0.0, "generation_time": 1083.46247, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1131.564535, "forward_backward_time": 237.515248, "rollout_overhead_time": 48.102065, "peak_allocated_gib": 69.07209730148315, "peak_reserved_gib": 76.8828125 }, { "step": 44, "loss": -0.18476806581020355, "grad_norm": 0.06787109375, "learning_rate": 1.4000000000000001e-06, "entropy": 0.15387438237667084, "clip_ratio/low_mean": 0.0006578518805326894, "clip_ratio/low_min": 0.0006578518805326894, "clip_ratio/high_mean": 0.0032520195527467877, "clip_ratio/high_max": 0.0032520195527467877, "clip_ratio/region_mean": 0.003909871433279477, "step_time": 237.64380487699964, "entropy_control/entropy": 0.18629113557588436, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "epoch": 0.176, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 237.543046, "rollout_overhead_time": 0.0, "peak_allocated_gib": 69.07209730148315, "peak_reserved_gib": 76.8828125, "reward": null }, { "step": 45, "loss": -2.682209014892578e-07, "grad_norm": 0.0390625, "learning_rate": 1.2000000000000002e-06, "thinking/tokens_mean": 725.63, "thinking/budget_hit_fraction": 0.2, "num_tokens": 1382070.0, "completions/mean_length": 3871.875, "completions/min_length": 1227.0, "completions/max_length": 6431.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 3871.875, "completions/min_terminated_length": 1227.0, "completions/max_terminated_length": 6431.0, "tools/call_frequency": 2.75, "tools/failure_frequency": 0.04545454680919647, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/mean": 0.6812279224395752, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/std": 0.05867519602179527, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/mean": 0.4583333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/std": 0.0833333283662796, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/mean": 0.609315037727356, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/mean": 0.6666666865348816, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/std": 0.0, "reward": 0.8733965158462524, "reward_std": 0.05135876685380936, "frac_reward_zero_std": 0.5, "frac_groups_zero_std": 0.5, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.2036501243710518, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 1434.0196957680018, "entropy_control/entropy": 0.2165004215226099, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "epoch": 0.18, "adapter_sync_time": 0.0, "generation_time": 1191.264096, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1233.33525, "forward_backward_time": 200.578298, "rollout_overhead_time": 42.071154, "peak_allocated_gib": 69.07209730148315, "peak_reserved_gib": 76.8828125 }, { "step": 46, "loss": 5.476176738739014e-05, "grad_norm": 0.037841796875, "learning_rate": 1.0000000000000002e-06, "entropy": 0.2035566046833992, "clip_ratio/low_mean": 0.00040974977309815586, "clip_ratio/low_min": 0.00040974977309815586, "clip_ratio/high_mean": 0.001062688126694411, "clip_ratio/high_max": 0.001062688126694411, "clip_ratio/region_mean": 0.001472437899792567, "step_time": 200.70870252800523, "entropy_control/entropy": 0.21648923488084992, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "epoch": 0.184, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 200.608904, "rollout_overhead_time": 0.0, "peak_allocated_gib": 69.07209730148315, "peak_reserved_gib": 76.8828125, "reward": null }, { "step": 47, "loss": -4.172325134277344e-07, "grad_norm": 0.046630859375, "learning_rate": 8.000000000000001e-07, "thinking/tokens_mean": 683.1875, "thinking/budget_hit_fraction": 0.10416666666666667, "num_tokens": 1438712.0, "completions/mean_length": 3793.75, "completions/min_length": 1546.0, "completions/max_length": 6494.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 3793.75, "completions/min_terminated_length": 1546.0, "completions/max_terminated_length": 6494.0, "tools/call_frequency": 2.75, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/mean": 0.6095432043075562, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/mean": 0.6666666865348816, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/mean": 0.8975076675415039, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/std": 0.08410575240850449, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/mean": 0.4166666865348816, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/std": 0.28867512941360474, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/std": 0.0, "reward": 0.9806088209152222, "reward_std": 0.15417182445526123, "frac_reward_zero_std": 0.5, "frac_groups_zero_std": 0.5, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.21144754625856876, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 1442.9635367679948, "entropy_control/entropy": 0.21850325503297707, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "epoch": 0.188, "adapter_sync_time": 0.0, "generation_time": 1180.318585, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1224.659429, "forward_backward_time": 218.205439, "rollout_overhead_time": 44.340844, "peak_allocated_gib": 69.07209730148315, "peak_reserved_gib": 76.8828125 }, { "step": 48, "loss": 6.0051679611206055e-05, "grad_norm": 0.052978515625, "learning_rate": 6.000000000000001e-07, "entropy": 0.2115715704858303, "clip_ratio/low_mean": 0.0006678483041469008, "clip_ratio/low_min": 0.0006678483041469008, "clip_ratio/high_mean": 0.0013030875124968588, "clip_ratio/high_max": 0.0013030875124968588, "clip_ratio/region_mean": 0.0019709358166437596, "step_time": 218.2718387430141, "entropy_control/entropy": 0.21860735752902272, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "epoch": 0.192, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 218.176116, "rollout_overhead_time": 0.0, "peak_allocated_gib": 69.07209730148315, "peak_reserved_gib": 76.8828125, "reward": null }, { "step": 49, "loss": -1.7136335372924805e-07, "grad_norm": 0.11279296875, "learning_rate": 4.0000000000000003e-07, "thinking/tokens_mean": 580.9702380952381, "thinking/budget_hit_fraction": 0.10714285714285714, "num_tokens": 1497081.0, "completions/mean_length": 4005.125, "completions/min_length": 2847.0, "completions/max_length": 5514.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 4005.125, "completions/min_terminated_length": 2847.0, "completions/max_terminated_length": 5514.0, "tools/call_frequency": 3.625, "tools/failure_frequency": 0.03448275849223137, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/mean": 0.5439562201499939, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/std": 0.20148904621601105, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/mean": 0.375, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/std": 0.34359216690063477, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/std": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/mean": 0.8639028072357178, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/std": 0.07629269361495972, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/mean": 0.4166666865348816, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/std": 0.3191424012184143, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/std": 0.0, "reward": 0.873721182346344, "reward_std": 0.3530108332633972, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.20466671139001846, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 1380.251024117999, "entropy_control/entropy": 0.2025214993560043, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "epoch": 0.196, "adapter_sync_time": 0.0, "generation_time": 1092.512068, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1141.032596, "forward_backward_time": 239.117184, "rollout_overhead_time": 48.520527, "peak_allocated_gib": 69.07209730148315, "peak_reserved_gib": 76.8828125 }, { "step": 50, "loss": 0.00019299983978271484, "grad_norm": 0.07568359375, "learning_rate": 2.0000000000000002e-07, "entropy": 0.20437165908515453, "clip_ratio/low_mean": 0.0019550809520296752, "clip_ratio/low_min": 0.0019550809520296752, "clip_ratio/high_mean": 0.0027724673855118454, "clip_ratio/high_max": 0.0027724673855118454, "clip_ratio/region_mean": 0.004727548337541521, "step_time": 239.17380329099979, "entropy_control/entropy": 0.20221694821805103, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "epoch": 0.2, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 239.072435, "rollout_overhead_time": 0.0, "peak_allocated_gib": 69.07209730148315, "peak_reserved_gib": 76.8828125, "reward": null } ], "holdout": [ { "seed": 900000, "sample": 0, "native_reward": 0.6401417781099434, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 3, "model_tokens": 4513, "reward_channels": { "native": 0.6401417781099434, "efficiency": 0.5, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "two_pool_cycle" }, { "seed": 900000, "sample": 1, "native_reward": 0.8638203836248219, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 5475, "reward_channels": { "native": 0.8638203836248219, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "two_pool_cycle" }, { "seed": 900000, "sample": 2, "native_reward": 0.6089882858614486, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 3327, "reward_channels": { "native": 0.6089882858614486, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "two_pool_cycle" }, { "seed": 900000, "sample": 3, "native_reward": 0.6089882858614486, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 3393, "reward_channels": { "native": 0.6089882858614486, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "two_pool_cycle" }, { "seed": 900001, "sample": 0, "native_reward": 0.8476409576243806, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 5, "model_tokens": 5539, "reward_channels": { "native": 0.8476409576243806, "efficiency": 0.16666666666666663, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "two_pool_cycle" }, { "seed": 900001, "sample": 1, "native_reward": 0.6093825392865522, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 2962, "reward_channels": { "native": 0.6093825392865522, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "two_pool_cycle" }, { "seed": 900001, "sample": 2, "native_reward": 0.6093825392865522, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 1839, "reward_channels": { "native": 0.6093825392865522, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "two_pool_cycle" }, { "seed": 900001, "sample": 3, "native_reward": 0.6093825392865522, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 1841, "reward_channels": { "native": 0.6093825392865522, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "two_pool_cycle" }, { "seed": 900000, "sample": 0, "native_reward": 0.6106601460847819, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 3304, "reward_channels": { "native": 0.6106601460847819, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "triangular_cycle" }, { "seed": 900000, "sample": 1, "native_reward": 0.8522045135672316, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 3, "model_tokens": 4942, "reward_channels": { "native": 0.8522045135672316, "efficiency": 0.5, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "triangular_cycle" }, { "seed": 900000, "sample": 2, "native_reward": 0.6106601460847819, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 5134, "reward_channels": { "native": 0.6106601460847819, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "triangular_cycle" }, { "seed": 900000, "sample": 3, "native_reward": 0.6106601460847819, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 3022, "reward_channels": { "native": 0.6106601460847819, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "triangular_cycle" }, { "seed": 900001, "sample": 0, "native_reward": 0.6108851248150757, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 3377, "reward_channels": { "native": 0.6108851248150757, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "triangular_cycle" }, { "seed": 900001, "sample": 1, "native_reward": 0.6108851248150757, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 3254, "reward_channels": { "native": 0.6108851248150757, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "triangular_cycle" }, { "seed": 900001, "sample": 2, "native_reward": 0.7060268919310104, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 2787, "reward_channels": { "native": 0.7060268919310104, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "triangular_cycle" }, { "seed": 900001, "sample": 3, "native_reward": 0.9758783046350626, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 5, "model_tokens": 6212, "reward_channels": { "native": 0.9758783046350626, "efficiency": 0.16666666666666663, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "triangular_cycle" }, { "seed": 900000, "sample": 0, "native_reward": 0.5924323165105192, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 5, "model_tokens": 6394, "reward_channels": { "native": 0.5924323165105192, "efficiency": 0.16666666666666663, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "competing_cycles" }, { "seed": 900000, "sample": 1, "native_reward": 0.6185941887823871, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 5493, "reward_channels": { "native": 0.6185941887823871, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "competing_cycles" }, { "seed": 900000, "sample": 2, "native_reward": 0.25, "promotion_success": false, "done": false, "errors": 0, "tool_steps": 6, "model_tokens": 5714, "reward_channels": { "native": 0.25, "efficiency": 0.0, "reliability": -1.0 }, "verification_level": null, "failed_gates": [ "finalized" ], "family": "competing_cycles" }, { "seed": 900000, "sample": 3, "native_reward": 0.578198213067076, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 6, "model_tokens": 5651, "reward_channels": { "native": 0.578198213067076, "efficiency": 0.0, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "competing_cycles" }, { "seed": 900001, "sample": 0, "native_reward": 0.8752709263462409, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 6882, "reward_channels": { "native": 0.8752709263462409, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "competing_cycles" }, { "seed": 900001, "sample": 1, "native_reward": 0.6145108982971254, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 3, "model_tokens": 3306, "reward_channels": { "native": 0.6145108982971254, "efficiency": 0.5, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "competing_cycles" }, { "seed": 900001, "sample": 2, "native_reward": 0.6145108982971254, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 3, "model_tokens": 3741, "reward_channels": { "native": 0.6145108982971254, "efficiency": 0.5, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "competing_cycles" }, { "seed": 900001, "sample": 3, "native_reward": 0.6512057439829095, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 5034, "reward_channels": { "native": 0.6512057439829095, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "competing_cycles" }, { "seed": 900000, "sample": 0, "native_reward": 0.609376392538787, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 1829, "reward_channels": { "native": 0.609376392538787, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "four_hop_cycle" }, { "seed": 900000, "sample": 1, "native_reward": 0.8444084478162579, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 3, "model_tokens": 2601, "reward_channels": { "native": 0.8444084478162579, "efficiency": 0.5, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "four_hop_cycle" }, { "seed": 900000, "sample": 2, "native_reward": 0.609376392538787, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 3139, "reward_channels": { "native": 0.609376392538787, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "four_hop_cycle" }, { "seed": 900000, "sample": 3, "native_reward": 0.609376392538787, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 2258, "reward_channels": { "native": 0.609376392538787, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "four_hop_cycle" }, { "seed": 900001, "sample": 0, "native_reward": 0.6095701077161815, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 3, "model_tokens": 4289, "reward_channels": { "native": 0.6095701077161815, "efficiency": 0.5, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "four_hop_cycle" }, { "seed": 900001, "sample": 1, "native_reward": 0.6095701077161815, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 2965, "reward_channels": { "native": 0.6095701077161815, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "four_hop_cycle" }, { "seed": 900001, "sample": 2, "native_reward": 0.8461963597959128, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 3414, "reward_channels": { "native": 0.8461963597959128, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "four_hop_cycle" }, { "seed": 900001, "sample": 3, "native_reward": 0.739978752793518, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 4013, "reward_channels": { "native": 0.739978752793518, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "four_hop_cycle" }, { "seed": 900000, "sample": 0, "native_reward": 0.8262556477476042, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 3, "model_tokens": 3954, "reward_channels": { "native": 0.8262556477476042, "efficiency": 0.5, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "size_sensitive_cycle" }, { "seed": 900000, "sample": 1, "native_reward": 0.0, "promotion_success": false, "done": false, "errors": 0, "tool_steps": 0, "model_tokens": 0, "reward_channels": { "native": 0.0, "efficiency": 0.0, "reliability": -1.0 }, "verification_level": null, "failed_gates": null, "family": "size_sensitive_cycle" }, { "seed": 900000, "sample": 2, "native_reward": 0.8215681477476042, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 5, "model_tokens": 9514, "reward_channels": { "native": 0.8215681477476042, "efficiency": 0.16666666666666663, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "size_sensitive_cycle" }, { "seed": 900000, "sample": 3, "native_reward": 0.8262556477476042, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 3702, "reward_channels": { "native": 0.8262556477476042, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "size_sensitive_cycle" }, { "seed": 900001, "sample": 0, "native_reward": 0.8245803237223959, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 3163, "reward_channels": { "native": 0.8245803237223959, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "size_sensitive_cycle" }, { "seed": 900001, "sample": 1, "native_reward": 0.9872132714354994, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 3, "model_tokens": 6563, "reward_channels": { "native": 0.9872132714354994, "efficiency": 0.5, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "size_sensitive_cycle" }, { "seed": 900001, "sample": 2, "native_reward": 0.8245803237223959, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 2838, "reward_channels": { "native": 0.8245803237223959, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "size_sensitive_cycle" }, { "seed": 900001, "sample": 3, "native_reward": 0.8245803237223959, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 3290, "reward_channels": { "native": 0.8245803237223959, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "size_sensitive_cycle" } ] }, { "run_id": "1789216886", "mode": "thinking", "method": "dapo", "steps": 50, "reward_points": 25, "train_mean": 0.8035986447334289, "train_last10": 0.791574364900589, "holdout_native": 0.670334256205925, "holdout_episodes": 40, "holdout_successes": 38, "mean_step_seconds": 937.509529993039, "label": "DAPO", "state_at_cutoff": "complete", "step_hours": 13.020965694347764, "peak_allocated_gib": 66.51273250579834, "peak_reserved_gib": 73.994140625, "extra_refill_rounds": 0.0, "groups_refilled": 0.0, "holdout_component_means": { "native": 0.670334256205925, "efficiency": 0.5083333333333334, "reliability": 0.9 }, "configuration": { "model": { "id": "Qwen/Qwen3-14B", "revision": "40c069824f4251a91eefaf281ebe4c544efd3e18", "loss": "dapo", "algorithm": "grpo", "no_think": false, "thinking_budget_tokens": 2048, "lora": true, "quantize": true, "lora_r": 32, "lora_alpha": 64, "use_dora": false }, "loss": "dapo", "normalization": "joint", "entropy_control": { "method": "monitor", "epsilon_low": 0.2, "epsilon_high": 0.28, "zeta_initial": 0.001, "zeta_min": 0.0001, "zeta_max": 0.05, "holdout_split": "benchmark", "holdout_seeds_per_family": 2, "num_iterations": 2 }, "dynamic_sampling": false, "sampling": { "group_size": 4, "temperature": 1.0, "max_completion_tokens": 22528, "top_entropy_quantile": 1.0, "temperature_decay_to": null }, "learning_rate": 1e-05, "turn_max_tokens": 3072 }, "source_sha256": { "metrics.jsonl": "c2042e6432234a101cef6be102e9ef2c79580b8b92ad8b9f7602bc39b3e1421e", "entropy_holdout_episodes.jsonl": "0276883c2f9922d2f9e877c478c6ae2ca44274aea2f492150e6ba5cbc3899f20", "resolved_training_config.json": "c926eec6dc33d3247dac92087f27e3f5926bd0a21a488fc60fda4ea9afcd61a4" }, "training_steps": [ { "step": 1, "loss": -0.32585203647613525, "grad_norm": 0.061279296875, "learning_rate": 1e-05, "thinking/tokens_mean": 606.2361111111111, "thinking/budget_hit_fraction": 0.14583333333333334, "num_tokens": 51975.0, "completions/mean_length": 3604.375, "completions/min_length": 2890.0, "completions/max_length": 6014.0, "completions/clipped_ratio": 0.25, "completions/mean_terminated_length": 3782.166748046875, "completions/min_terminated_length": 2890.0, "completions/max_terminated_length": 6014.0, "tools/call_frequency": 2.25, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.7262155413627625, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 0.13481803238391876, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.4583333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.2500000298023224, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/mean": 0.45231834053993225, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/std": 0.526568591594696, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/mean": 0.2916666865348816, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/std": 0.34359216690063477, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/mean": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/std": 1.154700517654419, "reward": 0.7080169916152954, "reward_std": 0.5734051465988159, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 0.5, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.14539151638746262, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 1292.0555840519999, "entropy_control/entropy": 0.20009788357278327, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.004, "adapter_sync_time": 0.0, "generation_time": 1101.067592, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1135.401235, "forward_backward_time": 156.55792, "rollout_overhead_time": 34.333643, "peak_allocated_gib": 46.211081981658936, "peak_reserved_gib": 51.38671875 }, { "step": 2, "loss": -0.32607361674308777, "grad_norm": 0.06396484375, "learning_rate": 9.800000000000001e-06, "entropy": 0.14505659975111485, "clip_ratio/low_mean": 0.0004901283973595127, "clip_ratio/low_min": 0.0004901283973595127, "clip_ratio/high_mean": 0.0006680102596874349, "clip_ratio/high_max": 0.0006680102596874349, "clip_ratio/region_mean": 0.0011581386570469476, "step_time": 156.59421884200037, "entropy_control/entropy": 0.1996898505403954, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.008, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 156.489999, "rollout_overhead_time": 0.0, "peak_allocated_gib": 46.689598083496094, "peak_reserved_gib": 51.484375, "reward": null }, { "step": 3, "loss": -0.2275024652481079, "grad_norm": 0.12109375, "learning_rate": 9.600000000000001e-06, "thinking/tokens_mean": 659.40625, "thinking/budget_hit_fraction": 0.15625, "num_tokens": 97103.0, "completions/mean_length": 2929.0, "completions/min_length": 1322.0, "completions/max_length": 4748.0, "completions/clipped_ratio": 0.125, "completions/mean_terminated_length": 2908.71435546875, "completions/min_terminated_length": 1322.0, "completions/max_terminated_length": 4748.0, "tools/call_frequency": 1.875, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.5617735385894775, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 0.2409123331308365, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.5625, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.23464767634868622, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 0.75, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 0.7071067690849304, "reward": 0.7398985624313354, "reward_std": 0.38766199350357056, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 0.5, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.16785027459263802, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 924.362896693, "entropy_control/entropy": 0.19792324347928317, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.012, "adapter_sync_time": 0.0, "generation_time": 802.354936, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 824.776987, "forward_backward_time": 99.481161, "rollout_overhead_time": 22.422051, "peak_allocated_gib": 46.689598083496094, "peak_reserved_gib": 51.494140625 }, { "step": 4, "loss": -0.22761014103889465, "grad_norm": 0.06298828125, "learning_rate": 9.4e-06, "entropy": 0.1675509177148342, "clip_ratio/low_mean": 0.0007690747879678383, "clip_ratio/low_min": 0.0007690747879678383, "clip_ratio/high_mean": 0.0011651632230496034, "clip_ratio/high_max": 0.0011651632230496034, "clip_ratio/region_mean": 0.0019342380110174417, "step_time": 99.59936518700079, "entropy_control/entropy": 0.1976198080155136, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.016, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 99.49852, "rollout_overhead_time": 0.0, "peak_allocated_gib": 46.689598083496094, "peak_reserved_gib": 51.494140625, "reward": null }, { "step": 5, "loss": -0.1114804595708847, "grad_norm": 0.03857421875, "learning_rate": 9.200000000000002e-06, "thinking/tokens_mean": 771.9714285714286, "thinking/budget_hit_fraction": 0.2571428571428571, "num_tokens": 147204.0, "completions/mean_length": 3557.625, "completions/min_length": 2221.0, "completions/max_length": 5795.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 3557.625, "completions/min_terminated_length": 2221.0, "completions/max_terminated_length": 5795.0, "tools/call_frequency": 2.0, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.562455415725708, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 0.24158011376857758, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.5416666865348816, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.24800795316696167, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 0.75, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 0.7071067690849304, "reward": 0.7395387291908264, "reward_std": 0.38718098402023315, "frac_reward_zero_std": 0.5, "frac_groups_zero_std": 0.5, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 0.5, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.20901422388851643, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 1064.131845978999, "entropy_control/entropy": 0.21309731888319075, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.02, "adapter_sync_time": 0.0, "generation_time": 885.447733, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 916.920766, "forward_backward_time": 147.10745, "rollout_overhead_time": 31.473034, "peak_allocated_gib": 46.689598083496094, "peak_reserved_gib": 51.494140625 }, { "step": 6, "loss": -0.11165289580821991, "grad_norm": 0.03955078125, "learning_rate": 9e-06, "entropy": 0.20917844772338867, "clip_ratio/low_mean": 0.0002251238183816895, "clip_ratio/low_min": 0.0002251238183816895, "clip_ratio/high_mean": 0.00043359493793104775, "clip_ratio/high_max": 0.00043359493793104775, "clip_ratio/region_mean": 0.0006587187563127372, "step_time": 147.21682394900108, "entropy_control/entropy": 0.2132870137415681, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.024, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 147.10944, "rollout_overhead_time": 0.0, "peak_allocated_gib": 46.689598083496094, "peak_reserved_gib": 51.494140625, "reward": null }, { "step": 7, "loss": -0.051447492092847824, "grad_norm": 0.061279296875, "learning_rate": 8.8e-06, "thinking/tokens_mean": 600.5586734693878, "thinking/budget_hit_fraction": 0.16071428571428573, "num_tokens": 213243.0, "completions/mean_length": 4787.375, "completions/min_length": 2629.0, "completions/max_length": 7505.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 4787.375, "completions/min_terminated_length": 2629.0, "completions/max_terminated_length": 7505.0, "tools/call_frequency": 4.0, "tools/failure_frequency": 0.03125, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/mean": 0.5704948902130127, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/std": 0.21831679344177246, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/mean": 0.25, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/std": 0.21516574919223785, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/std": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/mean": 0.9053964614868164, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/std": 0.09198641031980515, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/mean": 0.4583333730697632, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/std": 0.20971761643886566, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/std": 0.0, "reward": 0.9056540131568909, "reward_std": 0.3700742721557617, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.18943717889487743, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 1984.3922040430007, "entropy_control/entropy": 0.1950053953556034, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.028, "adapter_sync_time": 0.0, "generation_time": 1720.094763, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1765.746887, "forward_backward_time": 218.534155, "rollout_overhead_time": 45.652124, "peak_allocated_gib": 57.23006200790405, "peak_reserved_gib": 63.291015625 }, { "step": 8, "loss": -0.051317404955625534, "grad_norm": 0.056396484375, "learning_rate": 8.6e-06, "entropy": 0.18944677151739597, "clip_ratio/low_mean": 0.0008207407663576305, "clip_ratio/low_min": 0.0008207407663576305, "clip_ratio/high_mean": 0.0008121527353068814, "clip_ratio/high_max": 0.0008121527353068814, "clip_ratio/region_mean": 0.0016328935016645119, "step_time": 218.58815286200024, "entropy_control/entropy": 0.19497450350938475, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.032, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 218.490703, "rollout_overhead_time": 0.0, "peak_allocated_gib": 57.23006200790405, "peak_reserved_gib": 63.291015625, "reward": null }, { "step": 9, "loss": -0.022280991077423096, "grad_norm": 0.053955078125, "learning_rate": 8.400000000000001e-06, "thinking/tokens_mean": 815.9724489795918, "thinking/budget_hit_fraction": 0.22448979591836735, "num_tokens": 280425.0, "completions/mean_length": 5113.75, "completions/min_length": 2215.0, "completions/max_length": 7902.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 5113.75, "completions/min_terminated_length": 2215.0, "completions/max_terminated_length": 7902.0, "tools/call_frequency": 3.125, "tools/failure_frequency": 0.03999999910593033, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.5167769193649292, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 0.36239445209503174, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.4166666865348816, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.3191424012184143, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/mean": 0.5882549285888672, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/std": 0.2566491961479187, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/mean": 0.3333333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/std": 0.2357022762298584, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/std": 1.0, "reward": 0.6712659001350403, "reward_std": 0.47680771350860596, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 0.5, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.2029896266758442, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 2095.5628474430023, "entropy_control/entropy": 0.20581572082624197, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.036, "adapter_sync_time": 0.0, "generation_time": 1753.213057, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1810.901782, "forward_backward_time": 284.561784, "rollout_overhead_time": 57.688725, "peak_allocated_gib": 66.42421674728394, "peak_reserved_gib": 73.916015625 }, { "step": 10, "loss": -0.02226249873638153, "grad_norm": 0.05419921875, "learning_rate": 8.2e-06, "entropy": 0.20288954116404057, "clip_ratio/low_mean": 0.0005352675216272473, "clip_ratio/low_min": 0.0005352675216272473, "clip_ratio/high_mean": 0.0006748229934601113, "clip_ratio/high_max": 0.0006748229934601113, "clip_ratio/region_mean": 0.0012100905150873587, "step_time": 284.6669807569997, "entropy_control/entropy": 0.20570136341666023, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.04, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 284.568336, "rollout_overhead_time": 0.0, "peak_allocated_gib": 66.42421674728394, "peak_reserved_gib": 73.916015625, "reward": null }, { "step": 11, "loss": 0.0351431667804718, "grad_norm": 0.185546875, "learning_rate": 8.000000000000001e-06, "thinking/tokens_mean": 498.4304421768707, "thinking/budget_hit_fraction": 0.05612244897959184, "num_tokens": 337676.0, "completions/mean_length": 3847.375, "completions/min_length": 1208.0, "completions/max_length": 6759.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 3847.375, "completions/min_terminated_length": 1208.0, "completions/max_terminated_length": 6759.0, "tools/call_frequency": 3.625, "tools/failure_frequency": 0.17241379618644714, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__native/mean": 0.5858154892921448, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__native/std": 0.2379951924085617, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__efficiency/mean": 0.375, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__efficiency/std": 0.28463754057884216, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__reliability/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__reliability/std": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/mean": 0.5218393802642822, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/std": 0.354369193315506, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/mean": 0.375, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/std": 0.25, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/std": 1.0, "reward": 0.6725773811340332, "reward_std": 0.4711439609527588, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 0.5, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.1955870036035776, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 1532.8431922100008, "entropy_control/entropy": 0.1943371465758466, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.044, "adapter_sync_time": 0.0, "generation_time": 1264.596421, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1309.967144, "forward_backward_time": 222.770879, "rollout_overhead_time": 45.370723, "peak_allocated_gib": 66.42421674728394, "peak_reserved_gib": 73.916015625 }, { "step": 12, "loss": 0.03544866666197777, "grad_norm": 0.068359375, "learning_rate": 7.800000000000002e-06, "entropy": 0.19627046585083008, "clip_ratio/low_mean": 0.0003918922448065132, "clip_ratio/low_min": 0.0003918922448065132, "clip_ratio/high_mean": 0.0010237462411168963, "clip_ratio/high_max": 0.0010237462411168963, "clip_ratio/region_mean": 0.0014156384859234095, "step_time": 222.87849550399733, "entropy_control/entropy": 0.19470662796741972, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.048, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 222.775793, "rollout_overhead_time": 0.0, "peak_allocated_gib": 66.42421674728394, "peak_reserved_gib": 73.916015625, "reward": null }, { "step": 13, "loss": -0.10882490873336792, "grad_norm": 0.052734375, "learning_rate": 7.600000000000001e-06, "thinking/tokens_mean": 602.0, "thinking/budget_hit_fraction": 0.16326530612244897, "num_tokens": 397832.0, "completions/mean_length": 4234.5, "completions/min_length": 2811.0, "completions/max_length": 6909.0, "completions/clipped_ratio": 0.125, "completions/mean_terminated_length": 4400.71435546875, "completions/min_terminated_length": 2811.0, "completions/max_terminated_length": 6909.0, "tools/call_frequency": 2.875, "tools/failure_frequency": 0.043478261679410934, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.6766901016235352, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 0.11402305960655212, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.5416666865348816, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.15957120060920715, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/mean": 0.36920472979545593, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/std": 0.29988542199134827, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/mean": 0.2916666865348816, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/std": 0.34359216690063477, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/mean": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/std": 1.154700517654419, "reward": 0.6437807083129883, "reward_std": 0.4548361301422119, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 0.5, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.18049024604260921, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 1709.3426463319956, "entropy_control/entropy": 0.2073216940034155, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.052, "adapter_sync_time": 0.0, "generation_time": 1414.950437, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1464.73315, "forward_backward_time": 244.505207, "rollout_overhead_time": 49.782713, "peak_allocated_gib": 66.42421674728394, "peak_reserved_gib": 73.916015625 }, { "step": 14, "loss": -0.10887935757637024, "grad_norm": 0.052490234375, "learning_rate": 7.4e-06, "entropy": 0.18070777505636215, "clip_ratio/low_mean": 0.000984702754067257, "clip_ratio/low_min": 0.000984702754067257, "clip_ratio/high_mean": 0.0005047562153777108, "clip_ratio/high_max": 0.0005047562153777108, "clip_ratio/region_mean": 0.0014894589694449678, "step_time": 244.61136676099886, "entropy_control/entropy": 0.20755062122581508, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.056, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 244.510147, "rollout_overhead_time": 0.0, "peak_allocated_gib": 66.42421674728394, "peak_reserved_gib": 73.916015625, "reward": null }, { "step": 15, "loss": -0.16348621249198914, "grad_norm": 0.04931640625, "learning_rate": 7.2000000000000005e-06, "thinking/tokens_mean": 658.561507936508, "thinking/budget_hit_fraction": 0.1726190476190476, "num_tokens": 458862.0, "completions/mean_length": 4145.25, "completions/min_length": 2335.0, "completions/max_length": 6232.0, "completions/clipped_ratio": 0.125, "completions/mean_terminated_length": 4298.71435546875, "completions/min_terminated_length": 2335.0, "completions/max_terminated_length": 6232.0, "tools/call_frequency": 2.875, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/mean": 0.4855157732963562, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/std": 0.3275722563266754, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/mean": 0.2916666865348816, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/std": 0.28463754057884216, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/std": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/mean": 0.8572953343391418, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/std": 0.06820380687713623, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/std": 0.0, "reward": 0.8411972522735596, "reward_std": 0.4375230073928833, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 0.5, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.16464687511324883, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 1781.2970506860038, "entropy_control/entropy": 0.18991162269140954, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.06, "adapter_sync_time": 0.0, "generation_time": 1512.930426, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1558.960194, "forward_backward_time": 222.230419, "rollout_overhead_time": 46.029767, "peak_allocated_gib": 66.42421674728394, "peak_reserved_gib": 73.916015625 }, { "step": 16, "loss": -0.16334208846092224, "grad_norm": 0.047119140625, "learning_rate": 7e-06, "entropy": 0.1645781807601452, "clip_ratio/low_mean": 0.000881033978657797, "clip_ratio/low_min": 0.000881033978657797, "clip_ratio/high_mean": 0.0003130473469354911, "clip_ratio/high_max": 0.0003130473469354911, "clip_ratio/region_mean": 0.001194081325593288, "step_time": 222.34257325399994, "entropy_control/entropy": 0.1899335232739352, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.064, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 222.237415, "rollout_overhead_time": 0.0, "peak_allocated_gib": 66.42421674728394, "peak_reserved_gib": 73.916015625, "reward": null }, { "step": 17, "loss": -0.2446298897266388, "grad_norm": 0.07373046875, "learning_rate": 6.800000000000001e-06, "thinking/tokens_mean": 652.5625, "thinking/budget_hit_fraction": 0.14583333333333334, "num_tokens": 520614.0, "completions/mean_length": 4028.0, "completions/min_length": 2127.0, "completions/max_length": 6822.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 4028.0, "completions/min_terminated_length": 2127.0, "completions/max_terminated_length": 6822.0, "tools/call_frequency": 2.875, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/mean": 0.7041756510734558, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/std": 0.12205106765031815, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/std": 0.13608276844024658, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/mean": 0.6448794603347778, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/std": 0.07191970944404602, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/mean": 0.5416666865348816, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/std": 0.25, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/std": 0.0, "reward": 0.9005692005157471, "reward_std": 0.09277623146772385, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.2022705264389515, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 1728.3003051019969, "entropy_control/entropy": 0.2114883457479949, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.068, "adapter_sync_time": 0.0, "generation_time": 1420.438021, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1472.093508, "forward_backward_time": 256.108781, "rollout_overhead_time": 51.655487, "peak_allocated_gib": 66.42421674728394, "peak_reserved_gib": 73.916015625 }, { "step": 18, "loss": -0.24461257457733154, "grad_norm": 0.0751953125, "learning_rate": 6.600000000000001e-06, "entropy": 0.20262296870350838, "clip_ratio/low_mean": 0.0012727567518595606, "clip_ratio/low_min": 0.0012727567518595606, "clip_ratio/high_mean": 0.000507575947267469, "clip_ratio/high_max": 0.000507575947267469, "clip_ratio/region_mean": 0.0017803326991270296, "step_time": 256.19080749199566, "entropy_control/entropy": 0.21171698209385525, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.072, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 256.088687, "rollout_overhead_time": 0.0, "peak_allocated_gib": 66.42421674728394, "peak_reserved_gib": 73.916015625, "reward": null }, { "step": 19, "loss": -0.234040305018425, "grad_norm": 0.046630859375, "learning_rate": 6.4000000000000006e-06, "thinking/tokens_mean": 734.444246031746, "thinking/budget_hit_fraction": 0.1488095238095238, "num_tokens": 579366.0, "completions/mean_length": 4453.5, "completions/min_length": 3071.0, "completions/max_length": 5717.0, "completions/clipped_ratio": 0.125, "completions/mean_terminated_length": 4651.0, "completions/min_terminated_length": 3686.0, "completions/max_terminated_length": 5717.0, "tools/call_frequency": 3.0, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.8162115812301636, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 0.15060126781463623, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.3333333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.2357022762298584, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/mean": 0.6534296274185181, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/std": 0.440314918756485, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/mean": 0.4166666865348816, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/std": 0.2886751592159271, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/std": 1.0, "reward": 0.9035705924034119, "reward_std": 0.4581960141658783, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 0.5, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.17671867087483406, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 1768.9756165739927, "entropy_control/entropy": 0.20007666941078414, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.076, "adapter_sync_time": 0.0, "generation_time": 1549.589429, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1587.335027, "forward_backward_time": 181.544711, "rollout_overhead_time": 37.745598, "peak_allocated_gib": 66.42421674728394, "peak_reserved_gib": 73.916015625 }, { "step": 20, "loss": -0.23420467972755432, "grad_norm": 0.042236328125, "learning_rate": 6.200000000000001e-06, "entropy": 0.17669420316815376, "clip_ratio/low_mean": 0.00015516385610681027, "clip_ratio/low_min": 0.00015516385610681027, "clip_ratio/high_mean": 0.000958040873229038, "clip_ratio/high_max": 0.000958040873229038, "clip_ratio/region_mean": 0.0011132047293358482, "step_time": 181.6321102799957, "entropy_control/entropy": 0.2000932475076012, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.08, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 181.533869, "rollout_overhead_time": 0.0, "peak_allocated_gib": 66.42421674728394, "peak_reserved_gib": 73.916015625, "reward": null }, { "step": 21, "loss": 0.15066727995872498, "grad_norm": 0.059814453125, "learning_rate": 6e-06, "thinking/tokens_mean": 873.0083333333334, "thinking/budget_hit_fraction": 0.275, "num_tokens": 647029.0, "completions/mean_length": 4758.875, "completions/min_length": 1426.0, "completions/max_length": 6865.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 4758.875, "completions/min_terminated_length": 1426.0, "completions/max_terminated_length": 6865.0, "tools/call_frequency": 3.0, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/mean": 0.5967466831207275, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/std": 0.3003218472003937, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/mean": 0.4166666865348816, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/std": 0.09622503817081451, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/mean": 0.49323171377182007, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/std": 0.33581945300102234, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/mean": 0.4583333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/std": 0.3154948949813843, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/std": 0.0, "reward": 0.7668641805648804, "reward_std": 0.30521443486213684, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 0.5, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.18542677350342274, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 1774.4828627140014, "entropy_control/entropy": 0.1942894017855657, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.084, "adapter_sync_time": 0.0, "generation_time": 1524.824504, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1568.056851, "forward_backward_time": 206.329051, "rollout_overhead_time": 43.232347, "peak_allocated_gib": 66.42421674728394, "peak_reserved_gib": 73.916015625 }, { "step": 22, "loss": 0.1506814956665039, "grad_norm": 0.08056640625, "learning_rate": 5.8e-06, "entropy": 0.18549648858606815, "clip_ratio/low_mean": 0.00032234430545940995, "clip_ratio/low_min": 0.00032234430545940995, "clip_ratio/high_mean": 0.0012016383097943617, "clip_ratio/high_max": 0.0012016383097943617, "clip_ratio/region_mean": 0.0015239826152537717, "step_time": 206.42375343900858, "entropy_control/entropy": 0.19437401801192536, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.088, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 206.3231, "rollout_overhead_time": 0.0, "peak_allocated_gib": 66.42421674728394, "peak_reserved_gib": 73.916015625, "reward": null }, { "step": 23, "loss": 0.1105230450630188, "grad_norm": 0.0732421875, "learning_rate": 5.600000000000001e-06, "thinking/tokens_mean": 549.9583333333334, "thinking/budget_hit_fraction": 0.125, "num_tokens": 702486.0, "completions/mean_length": 4046.625, "completions/min_length": 2702.0, "completions/max_length": 6388.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 4046.625, "completions/min_terminated_length": 2702.0, "completions/max_terminated_length": 6388.0, "tools/call_frequency": 3.5, "tools/failure_frequency": 0.0714285746216774, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.735243558883667, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 0.12735584378242493, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.4583333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.15957120060920715, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/mean": 0.6815767288208008, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/std": 0.2877177894115448, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/mean": 0.4583333730697632, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/std": 0.3154948949813843, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/std": 1.0, "reward": 0.8813267350196838, "reward_std": 0.3490362763404846, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.1694607101380825, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 1365.9782649090012, "entropy_control/entropy": 0.16855090199562944, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.092, "adapter_sync_time": 0.0, "generation_time": 1138.350714, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1178.007235, "forward_backward_time": 187.867558, "rollout_overhead_time": 39.65652, "peak_allocated_gib": 66.42421674728394, "peak_reserved_gib": 73.916015625 }, { "step": 24, "loss": 0.11080944538116455, "grad_norm": 0.078125, "learning_rate": 5.400000000000001e-06, "entropy": 0.16979407519102097, "clip_ratio/low_mean": 0.0008367355476366356, "clip_ratio/low_min": 0.0008367355476366356, "clip_ratio/high_mean": 0.0008138977864291519, "clip_ratio/high_max": 0.0008138977864291519, "clip_ratio/region_mean": 0.0016506333340657875, "step_time": 187.97036160500284, "entropy_control/entropy": 0.16893767577369181, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.096, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 187.866728, "rollout_overhead_time": 0.0, "peak_allocated_gib": 66.42421674728394, "peak_reserved_gib": 73.916015625, "reward": null }, { "step": 25, "loss": -0.16266947984695435, "grad_norm": 0.059326171875, "learning_rate": 5.2e-06, "thinking/tokens_mean": 591.782142857143, "thinking/budget_hit_fraction": 0.125, "num_tokens": 761774.0, "completions/mean_length": 4321.0, "completions/min_length": 2464.0, "completions/max_length": 7546.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 4321.0, "completions/min_terminated_length": 2464.0, "completions/max_terminated_length": 7546.0, "tools/call_frequency": 3.5, "tools/failure_frequency": 0.0357142873108387, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.6715044975280762, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 0.12447243928909302, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.5833333730697632, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.1666666716337204, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/mean": 0.8140208721160889, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/std": 0.15102744102478027, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/mean": 0.2916666865348816, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/std": 0.28463754057884216, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/std": 0.0, "reward": 0.9646377563476562, "reward_std": 0.1373920887708664, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.21843799017369747, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 1797.1556359970018, "entropy_control/entropy": 0.2188761932152463, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.1, "adapter_sync_time": 0.0, "generation_time": 1483.105186, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1535.716725, "forward_backward_time": 261.324729, "rollout_overhead_time": 52.611539, "peak_allocated_gib": 66.42421674728394, "peak_reserved_gib": 73.916015625 }, { "step": 26, "loss": -0.16288836300373077, "grad_norm": 0.05712890625, "learning_rate": 5e-06, "entropy": 0.2184695415198803, "clip_ratio/low_mean": 0.0011264146414760035, "clip_ratio/low_min": 0.0011264146414760035, "clip_ratio/high_mean": 0.0004859872642555274, "clip_ratio/high_max": 0.0004859872642555274, "clip_ratio/region_mean": 0.0016124019057315309, "step_time": 261.40366287499273, "entropy_control/entropy": 0.2188503763701188, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.104, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 261.296295, "rollout_overhead_time": 0.0, "peak_allocated_gib": 66.42421674728394, "peak_reserved_gib": 73.916015625, "reward": null }, { "step": 27, "loss": 0.0430377721786499, "grad_norm": 0.06689453125, "learning_rate": 4.800000000000001e-06, "thinking/tokens_mean": 558.719387755102, "thinking/budget_hit_fraction": 0.125, "num_tokens": 822946.0, "completions/mean_length": 4344.0, "completions/min_length": 1426.0, "completions/max_length": 5702.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 4344.0, "completions/min_terminated_length": 1426.0, "completions/max_terminated_length": 5702.0, "tools/call_frequency": 3.375, "tools/failure_frequency": 0.03703703731298447, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/mean": 0.7350375652313232, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/std": 0.00234377384185791, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/mean": 0.4166666865348816, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/std": 0.1666666567325592, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/mean": 0.6625210046768188, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/std": 0.32501402497291565, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/std": 0.13608276844024658, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/std": 0.0, "reward": 0.9216959476470947, "reward_std": 0.21604451537132263, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.2035625521093607, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 1481.860717802003, "entropy_control/entropy": 0.20417728463592227, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.108, "adapter_sync_time": 0.0, "generation_time": 1253.208215, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1293.051743, "forward_backward_time": 188.70501, "rollout_overhead_time": 39.843528, "peak_allocated_gib": 66.42421674728394, "peak_reserved_gib": 73.916015625 }, { "step": 28, "loss": 0.04259427264332771, "grad_norm": 0.0625, "learning_rate": 4.600000000000001e-06, "entropy": 0.20379056967794895, "clip_ratio/low_mean": 0.0006127892702352256, "clip_ratio/low_min": 0.0006127892702352256, "clip_ratio/high_mean": 0.0014242628822103143, "clip_ratio/high_max": 0.0014242628822103143, "clip_ratio/region_mean": 0.00203705215244554, "step_time": 188.79422012899522, "entropy_control/entropy": 0.20432824070388717, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.112, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 188.689139, "rollout_overhead_time": 0.0, "peak_allocated_gib": 66.42421674728394, "peak_reserved_gib": 73.916015625, "reward": null }, { "step": 29, "loss": -0.06107598915696144, "grad_norm": 0.05810546875, "learning_rate": 4.4e-06, "thinking/tokens_mean": 881.7666666666667, "thinking/budget_hit_fraction": 0.225, "num_tokens": 887927.0, "completions/mean_length": 4840.125, "completions/min_length": 3344.0, "completions/max_length": 6239.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 4840.125, "completions/min_terminated_length": 3344.0, "completions/max_terminated_length": 6239.0, "tools/call_frequency": 3.125, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.735346257686615, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 0.14634527266025543, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.19245010614395142, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/mean": 0.6439940333366394, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/std": 0.05961908400058746, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/mean": 0.4583333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/std": 0.0833333283662796, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/std": 0.0, "reward": 0.9136285185813904, "reward_std": 0.10964539647102356, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.2220097780227661, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 1548.6979675099974, "entropy_control/entropy": 0.2205613230403864, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.116, "adapter_sync_time": 0.0, "generation_time": 1310.158924, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1351.849943, "forward_backward_time": 196.738621, "rollout_overhead_time": 41.69102, "peak_allocated_gib": 66.42421674728394, "peak_reserved_gib": 73.916015625 }, { "step": 30, "loss": -0.06128338724374771, "grad_norm": 0.07568359375, "learning_rate": 4.2000000000000004e-06, "entropy": 0.2213791236281395, "clip_ratio/low_mean": 0.001440455875126645, "clip_ratio/low_min": 0.001440455875126645, "clip_ratio/high_mean": 0.0003812400027527474, "clip_ratio/high_max": 0.0003812400027527474, "clip_ratio/region_mean": 0.0018216958778793924, "step_time": 196.82815311599916, "entropy_control/entropy": 0.2199646374989032, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.12, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 196.720821, "rollout_overhead_time": 0.0, "peak_allocated_gib": 66.42421674728394, "peak_reserved_gib": 73.916015625, "reward": null }, { "step": 31, "loss": 0.11386636644601822, "grad_norm": 0.0966796875, "learning_rate": 4.000000000000001e-06, "thinking/tokens_mean": 698.3647959183673, "thinking/budget_hit_fraction": 0.14285714285714285, "num_tokens": 959110.0, "completions/mean_length": 5814.375, "completions/min_length": 4186.0, "completions/max_length": 8945.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 5814.375, "completions/min_terminated_length": 4186.0, "completions/max_terminated_length": 8945.0, "tools/call_frequency": 4.875, "tools/failure_frequency": 0.025641025975346565, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/mean": 0.6564647555351257, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/std": 0.3634983003139496, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/mean": 0.2083333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/std": 0.19416078925132751, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/std": 0.9258201122283936, "reward": 0.7668814659118652, "reward_std": 0.5157641172409058, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.1981115248054266, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 2286.557038924999, "entropy_control/entropy": 0.205436449117097, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.124, "adapter_sync_time": 0.0, "generation_time": 1983.012707, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 2034.660368, "forward_backward_time": 251.791192, "rollout_overhead_time": 51.647661, "peak_allocated_gib": 66.42421674728394, "peak_reserved_gib": 73.916015625 }, { "step": 32, "loss": 0.11421447992324829, "grad_norm": 0.054443359375, "learning_rate": 3.8000000000000005e-06, "entropy": 0.19822406955063343, "clip_ratio/low_mean": 0.0005815364493173547, "clip_ratio/low_min": 0.0005815364493173547, "clip_ratio/high_mean": 0.0007530661969212815, "clip_ratio/high_max": 0.0007530661969212815, "clip_ratio/region_mean": 0.0013346026462386362, "step_time": 251.90113309199296, "entropy_control/entropy": 0.20553707141040561, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.128, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 251.799768, "rollout_overhead_time": 0.0, "peak_allocated_gib": 66.42421674728394, "peak_reserved_gib": 73.916015625, "reward": null }, { "step": 33, "loss": -0.14891234040260315, "grad_norm": 0.0595703125, "learning_rate": 3.6000000000000003e-06, "thinking/tokens_mean": 619.3869047619047, "thinking/budget_hit_fraction": 0.125, "num_tokens": 1016174.0, "completions/mean_length": 4237.5, "completions/min_length": 2360.0, "completions/max_length": 6402.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 4237.5, "completions/min_terminated_length": 2360.0, "completions/max_terminated_length": 6402.0, "tools/call_frequency": 3.375, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.7185772657394409, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 0.14374896883964539, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.5416666865348816, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.0833333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/mean": 0.9380024671554565, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/std": 0.07614614069461823, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/mean": 0.3333333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/std": 0.27216553688049316, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/std": 0.0, "reward": 1.0501649379730225, "reward_std": 0.15139304101467133, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.19579855911433697, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 1482.4170789929994, "entropy_control/entropy": 0.2025952334944647, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.132, "adapter_sync_time": 0.0, "generation_time": 1207.180918, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1254.546605, "forward_backward_time": 227.772787, "rollout_overhead_time": 47.365687, "peak_allocated_gib": 66.42421674728394, "peak_reserved_gib": 73.916015625 }, { "step": 34, "loss": -0.14863863587379456, "grad_norm": 0.0927734375, "learning_rate": 3.4000000000000005e-06, "entropy": 0.19608349725604057, "clip_ratio/low_mean": 0.0007156495121307671, "clip_ratio/low_min": 0.0007156495121307671, "clip_ratio/high_mean": 0.0005583660204138141, "clip_ratio/high_max": 0.0005583660204138141, "clip_ratio/region_mean": 0.0012740155325445812, "step_time": 227.86478610999984, "entropy_control/entropy": 0.202869317321604, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.136, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 227.768728, "rollout_overhead_time": 0.0, "peak_allocated_gib": 66.42421674728394, "peak_reserved_gib": 73.916015625, "reward": null }, { "step": 35, "loss": 0.04887096956372261, "grad_norm": 0.05712890625, "learning_rate": 3.2000000000000003e-06, "thinking/tokens_mean": 712.0357142857143, "thinking/budget_hit_fraction": 0.125, "num_tokens": 1087844.0, "completions/mean_length": 5095.75, "completions/min_length": 3593.0, "completions/max_length": 6793.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 5095.75, "completions/min_terminated_length": 3593.0, "completions/max_terminated_length": 6793.0, "tools/call_frequency": 4.0, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/mean": 0.6163443326950073, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/std": 0.17312179505825043, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/mean": 0.3333333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/std": 0.29546844959259033, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/mean": 0.75, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/std": 0.7071067690849304, "reward": 0.7830110192298889, "reward_std": 0.3088133931159973, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.21758883446455002, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 2171.7106434480047, "entropy_control/entropy": 0.2166911986716076, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.14, "adapter_sync_time": 0.0, "generation_time": 1843.603763, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1898.55651, "forward_backward_time": 273.057726, "rollout_overhead_time": 54.952747, "peak_allocated_gib": 66.42421674728394, "peak_reserved_gib": 73.916015625 }, { "step": 36, "loss": 0.048965927213430405, "grad_norm": 0.05908203125, "learning_rate": 3e-06, "entropy": 0.2172426339238882, "clip_ratio/low_mean": 0.001170090981759131, "clip_ratio/low_min": 0.001170090981759131, "clip_ratio/high_mean": 0.0005635659981635399, "clip_ratio/high_max": 0.0005635659981635399, "clip_ratio/region_mean": 0.0017336569799226709, "step_time": 273.15497789299116, "entropy_control/entropy": 0.2163397982679518, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.144, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 273.059198, "rollout_overhead_time": 0.0, "peak_allocated_gib": 66.42421674728394, "peak_reserved_gib": 73.916015625, "reward": null }, { "step": 37, "loss": 0.2859470844268799, "grad_norm": 0.072265625, "learning_rate": 2.8000000000000003e-06, "thinking/tokens_mean": 703.547619047619, "thinking/budget_hit_fraction": 0.17857142857142858, "num_tokens": 1146538.0, "completions/mean_length": 4456.75, "completions/min_length": 2812.0, "completions/max_length": 10325.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 4456.75, "completions/min_terminated_length": 2812.0, "completions/max_terminated_length": 10325.0, "tools/call_frequency": 3.125, "tools/failure_frequency": 0.03999999910593033, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.6686257123947144, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 0.11784413456916809, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.5416666865348816, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.15957120060920715, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__native/mean": 0.5488865971565247, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__native/std": 0.2643485367298126, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__efficiency/mean": 0.4583333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__efficiency/std": 0.2500000298023224, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__reliability/std": 0.0, "reward": 0.833756148815155, "reward_std": 0.20288874208927155, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.19427929259836674, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 2057.967392379993, "entropy_control/entropy": 0.19999829083888732, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.148, "adapter_sync_time": 0.0, "generation_time": 1740.651301, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1794.257522, "forward_backward_time": 263.603011, "rollout_overhead_time": 53.606221, "peak_allocated_gib": 66.42421674728394, "peak_reserved_gib": 73.916015625 }, { "step": 38, "loss": 0.2857789993286133, "grad_norm": 0.0673828125, "learning_rate": 2.6e-06, "entropy": 0.19363762624561787, "clip_ratio/low_mean": 0.0006089746602810919, "clip_ratio/low_min": 0.0006089746602810919, "clip_ratio/high_mean": 0.0009969057573471218, "clip_ratio/high_max": 0.0009969057573471218, "clip_ratio/region_mean": 0.0016058804176282138, "step_time": 263.7077892659945, "entropy_control/entropy": 0.19954557126813618, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.152, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 263.605404, "rollout_overhead_time": 0.0, "peak_allocated_gib": 66.42421674728394, "peak_reserved_gib": 73.916015625, "reward": null }, { "step": 39, "loss": -0.18318499624729156, "grad_norm": 0.064453125, "learning_rate": 2.4000000000000003e-06, "thinking/tokens_mean": 547.4107142857143, "thinking/budget_hit_fraction": 0.10714285714285714, "num_tokens": 1203755.0, "completions/mean_length": 4061.125, "completions/min_length": 2074.0, "completions/max_length": 7569.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 4061.125, "completions/min_terminated_length": 2074.0, "completions/max_terminated_length": 7569.0, "tools/call_frequency": 3.625, "tools/failure_frequency": 0.03448275849223137, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__native/mean": 0.7621634006500244, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__native/std": 0.16878987848758698, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__efficiency/mean": 0.4166666865348816, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__efficiency/std": 0.2357022762298584, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__reliability/std": 0.0, "reward": 0.9829967617988586, "reward_std": 0.15770334005355835, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.17808955907821655, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 1684.4528884049869, "entropy_control/entropy": 0.18310575210556304, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.156, "adapter_sync_time": 0.0, "generation_time": 1443.224573, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1484.945297, "forward_backward_time": 199.402443, "rollout_overhead_time": 41.720724, "peak_allocated_gib": 66.42421674728394, "peak_reserved_gib": 73.916015625 }, { "step": 40, "loss": -0.1832675337791443, "grad_norm": 0.064453125, "learning_rate": 2.2e-06, "entropy": 0.17827879264950752, "clip_ratio/low_mean": 0.0013926073152106255, "clip_ratio/low_min": 0.0013926073152106255, "clip_ratio/high_mean": 0.0004915288263873663, "clip_ratio/high_max": 0.0004915288263873663, "clip_ratio/region_mean": 0.0018841361415979918, "step_time": 199.50613806301408, "entropy_control/entropy": 0.18329289456727577, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.16, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 199.400434, "rollout_overhead_time": 0.0, "peak_allocated_gib": 66.42421674728394, "peak_reserved_gib": 73.916015625, "reward": null }, { "step": 41, "loss": -0.22436562180519104, "grad_norm": 0.064453125, "learning_rate": 2.0000000000000003e-06, "thinking/tokens_mean": 816.255, "thinking/budget_hit_fraction": 0.19, "num_tokens": 1262273.0, "completions/mean_length": 3837.75, "completions/min_length": 2457.0, "completions/max_length": 6153.0, "completions/clipped_ratio": 0.375, "completions/mean_terminated_length": 4297.80029296875, "completions/min_terminated_length": 2457.0, "completions/max_terminated_length": 6153.0, "tools/call_frequency": 2.0, "tools/failure_frequency": 0.0625, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__native/mean": 0.7159771919250488, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__native/std": 0.08374388515949249, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__efficiency/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__efficiency/std": 0.13608276844024658, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/mean": 0.1552533656358719, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/std": 0.3105067312717438, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/mean": 0.125, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/std": 0.25, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/mean": -0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/std": 1.0, "reward": 0.5012402534484863, "reward_std": 0.5839705467224121, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 0.5, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.12630949169397354, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 1628.2348596299998, "entropy_control/entropy": 0.21240435895114057, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.164, "adapter_sync_time": 0.0, "generation_time": 1406.627649, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1444.889648, "forward_backward_time": 183.246335, "rollout_overhead_time": 38.261999, "peak_allocated_gib": 66.42421674728394, "peak_reserved_gib": 73.916015625 }, { "step": 42, "loss": -0.2244964838027954, "grad_norm": 0.0634765625, "learning_rate": 1.8000000000000001e-06, "entropy": 0.1261419951915741, "clip_ratio/low_mean": 0.0007117298082448542, "clip_ratio/low_min": 0.0007117298082448542, "clip_ratio/high_mean": 0.0006165484446682967, "clip_ratio/high_max": 0.0006165484446682967, "clip_ratio/region_mean": 0.001328278252913151, "step_time": 183.32331458000408, "entropy_control/entropy": 0.2120536675808387, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.168, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 183.226289, "rollout_overhead_time": 0.0, "peak_allocated_gib": 66.42421674728394, "peak_reserved_gib": 73.916015625, "reward": null }, { "step": 43, "loss": -0.11457764357328415, "grad_norm": 0.06396484375, "learning_rate": 1.6000000000000001e-06, "thinking/tokens_mean": 770.5678571428572, "thinking/budget_hit_fraction": 0.2, "num_tokens": 1321139.0, "completions/mean_length": 4054.25, "completions/min_length": 2686.0, "completions/max_length": 6287.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 4054.25, "completions/min_terminated_length": 2686.0, "completions/max_terminated_length": 6287.0, "tools/call_frequency": 2.625, "tools/failure_frequency": 0.0476190485060215, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__native/mean": 0.4665586054325104, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__native/std": 0.31143802404403687, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__efficiency/mean": 0.4166666865348816, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__efficiency/std": 0.3191424012184143, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__reliability/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__reliability/std": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/mean": 0.7308187484741211, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/std": 0.09706426411867142, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/std": 0.13608276844024658, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/std": 0.0, "reward": 0.7716053128242493, "reward_std": 0.40083083510398865, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 0.5, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.20480629988014698, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 1231.5425371530146, "entropy_control/entropy": 0.20772306326729478, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.172, "adapter_sync_time": 0.0, "generation_time": 1022.520319, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1058.76146, "forward_backward_time": 172.681923, "rollout_overhead_time": 36.241141, "peak_allocated_gib": 66.42421674728394, "peak_reserved_gib": 73.916015625 }, { "step": 44, "loss": -0.11447513103485107, "grad_norm": 0.061279296875, "learning_rate": 1.4000000000000001e-06, "entropy": 0.2048952467739582, "clip_ratio/low_mean": 0.0006981215992709622, "clip_ratio/low_min": 0.0006981215992709622, "clip_ratio/high_mean": 0.0007196775623015128, "clip_ratio/high_max": 0.0007196775623015128, "clip_ratio/region_mean": 0.001417799161572475, "step_time": 172.76562454499071, "entropy_control/entropy": 0.20774115002095983, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.176, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 172.667739, "rollout_overhead_time": 0.0, "peak_allocated_gib": 66.42421674728394, "peak_reserved_gib": 73.916015625, "reward": null }, { "step": 45, "loss": -0.08469386398792267, "grad_norm": 0.064453125, "learning_rate": 1.2000000000000002e-06, "thinking/tokens_mean": 623.4416666666666, "thinking/budget_hit_fraction": 0.14285714285714285, "num_tokens": 1387005.0, "completions/mean_length": 4563.25, "completions/min_length": 1985.0, "completions/max_length": 6768.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 4563.25, "completions/min_terminated_length": 1985.0, "completions/max_terminated_length": 6768.0, "tools/call_frequency": 3.75, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/mean": 0.6128880977630615, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/std": 0.2622775435447693, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/mean": 0.1666666716337204, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/std": 0.235702246427536, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/std": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/mean": 0.6322802305221558, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/std": 0.045930415391922, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/mean": 0.5833333730697632, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/std": 0.1666666716337204, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/std": 0.0, "reward": 0.7913341522216797, "reward_std": 0.3092081546783447, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.18784888088703156, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 2018.4625300120024, "entropy_control/entropy": 0.18692615238509233, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.18, "adapter_sync_time": 0.0, "generation_time": 1677.31727, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1734.556493, "forward_backward_time": 283.79525, "rollout_overhead_time": 57.239222, "peak_allocated_gib": 66.51273250579834, "peak_reserved_gib": 73.994140625 }, { "step": 46, "loss": -0.08453282713890076, "grad_norm": 0.06201171875, "learning_rate": 1.0000000000000002e-06, "entropy": 0.18786270171403885, "clip_ratio/low_mean": 0.001475626733736135, "clip_ratio/low_min": 0.001475626733736135, "clip_ratio/high_mean": 0.0003618006085162051, "clip_ratio/high_max": 0.0003618006085162051, "clip_ratio/region_mean": 0.0018374273422523402, "step_time": 283.93783065299067, "entropy_control/entropy": 0.18700232878490297, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.184, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 283.817246, "rollout_overhead_time": 0.0, "peak_allocated_gib": 66.51273250579834, "peak_reserved_gib": 73.994140625, "reward": null }, { "step": 47, "loss": -0.17486082017421722, "grad_norm": 0.07373046875, "learning_rate": 8.000000000000001e-07, "thinking/tokens_mean": 694.2222222222222, "thinking/budget_hit_fraction": 0.16666666666666666, "num_tokens": 1444322.0, "completions/mean_length": 3878.125, "completions/min_length": 1925.0, "completions/max_length": 5465.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 3878.125, "completions/min_terminated_length": 1925.0, "completions/max_terminated_length": 5465.0, "tools/call_frequency": 2.625, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/mean": 0.6431146860122681, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/std": 0.0671430230140686, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/mean": 0.5833333730697632, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/std": 0.09622505307197571, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/mean": 0.8646462559700012, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/std": 0.07877790927886963, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/mean": 0.5416666865348816, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/std": 0.25, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/std": 0.0, "reward": 0.9820054769515991, "reward_std": 0.1320001631975174, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.19449527747929096, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 1601.3260191109875, "entropy_control/entropy": 0.19548352027935895, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.188, "adapter_sync_time": 0.0, "generation_time": 1366.158659, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1406.71528, "forward_backward_time": 194.496634, "rollout_overhead_time": 40.556621, "peak_allocated_gib": 66.51273250579834, "peak_reserved_gib": 73.994140625 }, { "step": 48, "loss": -0.17461475729942322, "grad_norm": 0.07470703125, "learning_rate": 6.000000000000001e-07, "entropy": 0.19404620304703712, "clip_ratio/low_mean": 0.0018585657380754128, "clip_ratio/low_min": 0.0018585657380754128, "clip_ratio/high_mean": 0.00026816325771505944, "clip_ratio/high_max": 0.00026816325771505944, "clip_ratio/region_mean": 0.0021267289957904723, "step_time": 194.61791170100332, "entropy_control/entropy": 0.19509070221750152, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.192, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 194.493542, "rollout_overhead_time": 0.0, "peak_allocated_gib": 66.51273250579834, "peak_reserved_gib": 73.994140625, "reward": null }, { "step": 49, "loss": -0.1855476051568985, "grad_norm": 0.06201171875, "learning_rate": 4.0000000000000003e-07, "thinking/tokens_mean": 498.35714285714283, "thinking/budget_hit_fraction": 0.07142857142857142, "num_tokens": 1498068.0, "completions/mean_length": 3427.25, "completions/min_length": 1215.0, "completions/max_length": 5800.0, "completions/clipped_ratio": 0.25, "completions/mean_terminated_length": 3546.0, "completions/min_terminated_length": 1215.0, "completions/max_terminated_length": 5800.0, "tools/call_frequency": 2.875, "tools/failure_frequency": 0.043478261679410934, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/mean": 0.3669130802154541, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/std": 0.2973985970020294, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/mean": 0.3333333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/std": 0.38490021228790283, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/mean": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/std": 1.154700517654419, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/mean": 0.5094164609909058, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/std": 0.45884740352630615, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/mean": 0.25, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/std": 0.28867512941360474, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/mean": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/std": 1.154700517654419, "reward": 0.4527481198310852, "reward_std": 0.5826615691184998, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 0.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.14470533467829227, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 1525.5017623179883, "entropy_control/entropy": 0.20231896153053505, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.196, "adapter_sync_time": 0.0, "generation_time": 1270.81918, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1314.141797, "forward_backward_time": 211.25236, "rollout_overhead_time": 43.322618, "peak_allocated_gib": 66.51273250579834, "peak_reserved_gib": 73.994140625 }, { "step": 50, "loss": -0.18560901284217834, "grad_norm": 0.06396484375, "learning_rate": 2.0000000000000002e-07, "entropy": 0.14430952444672585, "clip_ratio/low_mean": 0.0010318411368643865, "clip_ratio/low_min": 0.0010318411368643865, "clip_ratio/high_mean": 0.000781740018283017, "clip_ratio/high_max": 0.000781740018283017, "clip_ratio/region_mean": 0.0018135811551474035, "step_time": 211.343559276007, "entropy_control/entropy": 0.20206067844582837, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.2, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 211.237227, "rollout_overhead_time": 0.0, "peak_allocated_gib": 66.51273250579834, "peak_reserved_gib": 73.994140625, "reward": null } ], "holdout": [ { "seed": 900000, "sample": 0, "native_reward": 0.8638203836248219, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 5833, "reward_channels": { "native": 0.8638203836248219, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "two_pool_cycle" }, { "seed": 900000, "sample": 1, "native_reward": 0.6089882858614486, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 2844, "reward_channels": { "native": 0.6089882858614486, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "two_pool_cycle" }, { "seed": 900000, "sample": 2, "native_reward": 0.8638203836248219, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 3989, "reward_channels": { "native": 0.8638203836248219, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "two_pool_cycle" }, { "seed": 900000, "sample": 3, "native_reward": 0.8638203836248219, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 5767, "reward_channels": { "native": 0.8638203836248219, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "two_pool_cycle" }, { "seed": 900001, "sample": 0, "native_reward": 0.6093825392865522, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 2860, "reward_channels": { "native": 0.6093825392865522, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "two_pool_cycle" }, { "seed": 900001, "sample": 1, "native_reward": 0.6093825392865522, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 4003, "reward_channels": { "native": 0.6093825392865522, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "two_pool_cycle" }, { "seed": 900001, "sample": 2, "native_reward": 0.6093825392865522, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 2868, "reward_channels": { "native": 0.6093825392865522, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "two_pool_cycle" }, { "seed": 900001, "sample": 3, "native_reward": 0.6093825392865522, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 3397, "reward_channels": { "native": 0.6093825392865522, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "two_pool_cycle" }, { "seed": 900000, "sample": 0, "native_reward": 0.6106601460847819, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 2803, "reward_channels": { "native": 0.6106601460847819, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "triangular_cycle" }, { "seed": 900000, "sample": 1, "native_reward": 0.8204653155141224, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 3, "model_tokens": 5287, "reward_channels": { "native": 0.8204653155141224, "efficiency": 0.5, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "triangular_cycle" }, { "seed": 900000, "sample": 2, "native_reward": 0.8204653155141224, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 3, "model_tokens": 5805, "reward_channels": { "native": 0.8204653155141224, "efficiency": 0.5, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "triangular_cycle" }, { "seed": 900000, "sample": 3, "native_reward": 0.0, "promotion_success": false, "done": false, "errors": 0, "tool_steps": 0, "model_tokens": 0, "reward_channels": { "native": 0.0, "efficiency": 0.0, "reliability": -1.0 }, "verification_level": null, "failed_gates": null, "family": "triangular_cycle" }, { "seed": 900001, "sample": 0, "native_reward": 0.6108851248150757, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 5849, "reward_channels": { "native": 0.6108851248150757, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "triangular_cycle" }, { "seed": 900001, "sample": 1, "native_reward": 0.6108851248150757, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 2883, "reward_channels": { "native": 0.6108851248150757, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "triangular_cycle" }, { "seed": 900001, "sample": 2, "native_reward": 0.9711908046350627, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 6, "model_tokens": 5434, "reward_channels": { "native": 0.9711908046350627, "efficiency": 0.0, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "triangular_cycle" }, { "seed": 900001, "sample": 3, "native_reward": 0.8541365354407209, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 6366, "reward_channels": { "native": 0.8541365354407209, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "triangular_cycle" }, { "seed": 900000, "sample": 0, "native_reward": 0.6140552238672119, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 3, "model_tokens": 4679, "reward_channels": { "native": 0.6140552238672119, "efficiency": 0.5, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "competing_cycles" }, { "seed": 900000, "sample": 1, "native_reward": 0.0, "promotion_success": false, "done": false, "errors": 0, "tool_steps": 0, "model_tokens": 0, "reward_channels": { "native": 0.0, "efficiency": 0.0, "reliability": -1.0 }, "verification_level": null, "failed_gates": null, "family": "competing_cycles" }, { "seed": 900000, "sample": 2, "native_reward": 0.6140552238672119, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 3, "model_tokens": 4880, "reward_channels": { "native": 0.6140552238672119, "efficiency": 0.5, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "competing_cycles" }, { "seed": 900000, "sample": 3, "native_reward": 0.5798840889712712, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 7203, "reward_channels": { "native": 0.5798840889712712, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "competing_cycles" }, { "seed": 900001, "sample": 0, "native_reward": 0.6098233982971253, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 5, "model_tokens": 8430, "reward_channels": { "native": 0.6098233982971253, "efficiency": 0.16666666666666663, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "competing_cycles" }, { "seed": 900001, "sample": 1, "native_reward": 0.6512057439829095, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 6466, "reward_channels": { "native": 0.6512057439829095, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "competing_cycles" }, { "seed": 900001, "sample": 2, "native_reward": 0.7140378233531887, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 5, "model_tokens": 7727, "reward_channels": { "native": 0.7140378233531887, "efficiency": 0.16666666666666663, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "competing_cycles" }, { "seed": 900001, "sample": 3, "native_reward": 0.6145108982971254, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 3, "model_tokens": 4801, "reward_channels": { "native": 0.6145108982971254, "efficiency": 0.5, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "competing_cycles" }, { "seed": 900000, "sample": 0, "native_reward": 0.609376392538787, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 1426, "reward_channels": { "native": 0.609376392538787, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "four_hop_cycle" }, { "seed": 900000, "sample": 1, "native_reward": 0.609376392538787, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 2317, "reward_channels": { "native": 0.609376392538787, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "four_hop_cycle" }, { "seed": 900000, "sample": 2, "native_reward": 0.609376392538787, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 4209, "reward_channels": { "native": 0.609376392538787, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "four_hop_cycle" }, { "seed": 900000, "sample": 3, "native_reward": 0.609376392538787, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 2922, "reward_channels": { "native": 0.609376392538787, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "four_hop_cycle" }, { "seed": 900001, "sample": 0, "native_reward": 0.6095701077161815, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 3818, "reward_channels": { "native": 0.6095701077161815, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "four_hop_cycle" }, { "seed": 900001, "sample": 1, "native_reward": 0.6095701077161815, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 1444, "reward_channels": { "native": 0.6095701077161815, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "four_hop_cycle" }, { "seed": 900001, "sample": 2, "native_reward": 0.6095701077161815, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 1463, "reward_channels": { "native": 0.6095701077161815, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "four_hop_cycle" }, { "seed": 900001, "sample": 3, "native_reward": 0.6095701077161815, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 2396, "reward_channels": { "native": 0.6095701077161815, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "four_hop_cycle" }, { "seed": 900000, "sample": 0, "native_reward": 0.8262556477476042, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 5561, "reward_channels": { "native": 0.8262556477476042, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "size_sensitive_cycle" }, { "seed": 900000, "sample": 1, "native_reward": 0.8262556477476042, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 2562, "reward_channels": { "native": 0.8262556477476042, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "size_sensitive_cycle" }, { "seed": 900000, "sample": 2, "native_reward": 0.8262556477476042, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 2742, "reward_channels": { "native": 0.8262556477476042, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "size_sensitive_cycle" }, { "seed": 900000, "sample": 3, "native_reward": 0.8262556477476042, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 3740, "reward_channels": { "native": 0.8262556477476042, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "size_sensitive_cycle" }, { "seed": 900001, "sample": 0, "native_reward": 0.8245803237223959, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 3, "model_tokens": 4736, "reward_channels": { "native": 0.8245803237223959, "efficiency": 0.5, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "size_sensitive_cycle" }, { "seed": 900001, "sample": 1, "native_reward": 0.8245803237223959, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 2434, "reward_channels": { "native": 0.8245803237223959, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "size_sensitive_cycle" }, { "seed": 900001, "sample": 2, "native_reward": 0.8245803237223959, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 3641, "reward_channels": { "native": 0.8245803237223959, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "size_sensitive_cycle" }, { "seed": 900001, "sample": 3, "native_reward": 0.8245803237223959, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 2804, "reward_channels": { "native": 0.8245803237223959, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "size_sensitive_cycle" } ] }, { "run_id": "1789271070", "mode": "thinking", "method": "dapo-refill", "steps": 50, "reward_points": 25, "train_mean": 0.8274829816818238, "train_last10": 0.856386661529541, "holdout_native": 0.7104741920786344, "holdout_episodes": 40, "holdout_successes": 38, "mean_step_seconds": 1015.5512065880799, "label": "DAPO-refill", "state_at_cutoff": "complete", "step_hours": 14.104877869278889, "peak_allocated_gib": 68.46192026138306, "peak_reserved_gib": 76.251953125, "extra_refill_rounds": 0.0, "groups_refilled": 0.0, "holdout_component_means": { "native": 0.7104741920786344, "efficiency": 0.42083333333333334, "reliability": 0.9 }, "configuration": { "model": { "id": "Qwen/Qwen3-14B", "revision": "40c069824f4251a91eefaf281ebe4c544efd3e18", "loss": "dapo", "algorithm": "grpo", "no_think": false, "thinking_budget_tokens": 2048, "lora": true, "quantize": true, "lora_r": 32, "lora_alpha": 64, "use_dora": false }, "loss": "dapo", "normalization": "joint", "entropy_control": { "method": "monitor", "epsilon_low": 0.2, "epsilon_high": 0.28, "zeta_initial": 0.001, "zeta_min": 0.0001, "zeta_max": 0.05, "holdout_split": "benchmark", "holdout_seeds_per_family": 2, "num_iterations": 2 }, "dynamic_sampling": true, "sampling": { "group_size": 4, "temperature": 1.0, "max_completion_tokens": 22528, "top_entropy_quantile": 1.0, "temperature_decay_to": null }, "learning_rate": 1e-05, "turn_max_tokens": 3072 }, "source_sha256": { "metrics.jsonl": "10e45bd635ffe9b8c4953bfec377bbc2fc2c91edde6f226721bd4e2fb040caf6", "entropy_holdout_episodes.jsonl": "ebd77943f778dee4dac2d22e67ca07bf1117880221e5965ea787ec886e0d5d56", "resolved_training_config.json": "f94229355981ffb1a7c5d059a74bd06ded0fa98c4eddc5a9bb321408770f26d4" }, "training_steps": [ { "step": 1, "loss": -0.41404587030410767, "grad_norm": 0.078125, "learning_rate": 1e-05, "thinking/tokens_mean": 606.2361111111111, "thinking/budget_hit_fraction": 0.14583333333333334, "num_tokens": 51975.0, "completions/mean_length": 3604.375, "completions/min_length": 2890.0, "completions/max_length": 6014.0, "completions/clipped_ratio": 0.25, "completions/mean_terminated_length": 3782.166748046875, "completions/min_terminated_length": 2890.0, "completions/max_terminated_length": 6014.0, "tools/call_frequency": 2.25, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.7262155413627625, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 0.13481803238391876, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.4583333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.2500000298023224, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/mean": 0.45231834053993225, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/std": 0.526568591594696, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/mean": 0.2916666865348816, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/std": 0.34359216690063477, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/mean": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/std": 1.154700517654419, "reward": 0.7080169916152954, "reward_std": 0.5734051465988159, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 0.5, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.14539151638746262, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 1271.8538267409997, "entropy_control/entropy": 0.20009788357278327, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.004, "adapter_sync_time": 0.0, "generation_time": 1080.446999, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1114.965437, "forward_backward_time": 156.7967, "rollout_overhead_time": 34.518438, "peak_allocated_gib": 46.211081981658936, "peak_reserved_gib": 51.38671875 }, { "step": 2, "loss": -0.4143061637878418, "grad_norm": 0.07763671875, "learning_rate": 9.800000000000001e-06, "entropy": 0.14520721696317196, "clip_ratio/low_mean": 0.0005725046212319285, "clip_ratio/low_min": 0.0005725046212319285, "clip_ratio/high_mean": 0.0005652937870763708, "clip_ratio/high_max": 0.0005652937870763708, "clip_ratio/region_mean": 0.0011377984083082993, "step_time": 156.81330739999976, "entropy_control/entropy": 0.19984217459147838, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.008, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 156.724333, "rollout_overhead_time": 0.0, "peak_allocated_gib": 46.689598083496094, "peak_reserved_gib": 51.484375, "reward": null }, { "step": 3, "loss": -0.5281946063041687, "grad_norm": 0.059814453125, "learning_rate": 9.600000000000001e-06, "thinking/tokens_mean": 468.8392857142857, "thinking/budget_hit_fraction": 0.08928571428571429, "num_tokens": 97984.0, "completions/mean_length": 3039.125, "completions/min_length": 1441.0, "completions/max_length": 4823.0, "completions/clipped_ratio": 0.25, "completions/mean_terminated_length": 3028.5, "completions/min_terminated_length": 1441.0, "completions/max_terminated_length": 4823.0, "tools/call_frequency": 2.375, "tools/failure_frequency": 0.10526315867900848, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.5148292779922485, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 0.33359119296073914, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.3958333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.3078019917011261, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 0.9258201122283936, "reward": 0.6346210241317749, "reward_std": 0.5233859419822693, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 0.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.13246088474988937, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 1046.5687405200001, "entropy_control/entropy": 0.1794875835345148, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.012, "adapter_sync_time": 0.0, "generation_time": 852.482957, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 886.186538, "forward_backward_time": 160.289038, "rollout_overhead_time": 33.703581, "peak_allocated_gib": 48.180649280548096, "peak_reserved_gib": 53.22265625 }, { "step": 4, "loss": -0.52854323387146, "grad_norm": 0.057861328125, "learning_rate": 9.4e-06, "entropy": 0.13232495449483395, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0015268891293089837, "clip_ratio/high_max": 0.0015268891293089837, "clip_ratio/region_mean": 0.0015268891293089837, "step_time": 160.3931571160001, "entropy_control/entropy": 0.17936813542256924, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.016, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 160.299952, "rollout_overhead_time": 0.0, "peak_allocated_gib": 48.180649280548096, "peak_reserved_gib": 53.22265625, "reward": null }, { "step": 5, "loss": -0.05710165947675705, "grad_norm": 0.07080078125, "learning_rate": 9.200000000000002e-06, "thinking/tokens_mean": 668.375, "thinking/budget_hit_fraction": 0.20833333333333334, "num_tokens": 152522.0, "completions/mean_length": 4112.25, "completions/min_length": 2754.0, "completions/max_length": 5956.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 4112.25, "completions/min_terminated_length": 2754.0, "completions/max_terminated_length": 5956.0, "tools/call_frequency": 2.75, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.680217981338501, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 0.10803307592868805, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.5416666865348816, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.17251639068126678, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 0.0, "reward": 0.9073012471199036, "reward_std": 0.10156676918268204, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.20104767382144928, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 1451.4844390130015, "entropy_control/entropy": 0.204112548446505, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.02, "adapter_sync_time": 0.0, "generation_time": 1262.375617, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1295.511743, "forward_backward_time": 155.879775, "rollout_overhead_time": 33.136126, "peak_allocated_gib": 48.180649280548096, "peak_reserved_gib": 53.22265625 }, { "step": 6, "loss": -0.057032063603401184, "grad_norm": 0.0673828125, "learning_rate": 9e-06, "entropy": 0.20102406851947308, "clip_ratio/low_mean": 0.0012988815578864887, "clip_ratio/low_min": 0.0012988815578864887, "clip_ratio/high_mean": 0.0004859589898842387, "clip_ratio/high_max": 0.0004859589898842387, "clip_ratio/region_mean": 0.0017848405477707274, "step_time": 155.96452493000015, "entropy_control/entropy": 0.20395353653773732, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.024, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 155.87106, "rollout_overhead_time": 0.0, "peak_allocated_gib": 48.180649280548096, "peak_reserved_gib": 53.22265625, "reward": null }, { "step": 7, "loss": -0.0018136687576770782, "grad_norm": 0.056884765625, "learning_rate": 8.8e-06, "thinking/tokens_mean": 773.0738095238095, "thinking/budget_hit_fraction": 0.14285714285714285, "num_tokens": 226566.0, "completions/mean_length": 5788.0, "completions/min_length": 3740.0, "completions/max_length": 7932.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 5788.0, "completions/min_terminated_length": 3740.0, "completions/max_terminated_length": 7932.0, "tools/call_frequency": 4.125, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/mean": 0.6478960514068604, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/std": 0.2928895652294159, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/mean": 0.0833333358168602, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/std": 0.09622505307197571, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/std": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/mean": 0.8597946166992188, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/std": 0.06807643175125122, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/mean": 0.5416666865348816, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/std": 0.15957120060920715, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/std": 0.0, "reward": 0.9194703698158264, "reward_std": 0.3659683167934418, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.19893752969801426, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 2205.289949291002, "entropy_control/entropy": 0.19834686563335419, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.028, "adapter_sync_time": 0.0, "generation_time": 1860.865076, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1919.055288, "forward_backward_time": 286.123635, "rollout_overhead_time": 58.190212, "peak_allocated_gib": 66.74047183990479, "peak_reserved_gib": 74.21875 }, { "step": 8, "loss": -0.0017804503440856934, "grad_norm": 0.05517578125, "learning_rate": 8.6e-06, "entropy": 0.1987463440746069, "clip_ratio/low_mean": 0.0012409407499944791, "clip_ratio/low_min": 0.0012409407499944791, "clip_ratio/high_mean": 0.0006047607457730919, "clip_ratio/high_max": 0.0006047607457730919, "clip_ratio/region_mean": 0.001845701495767571, "step_time": 286.1488141740001, "entropy_control/entropy": 0.19819089188407168, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.032, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 286.05478, "rollout_overhead_time": 0.0, "peak_allocated_gib": 66.74047183990479, "peak_reserved_gib": 74.21875, "reward": null }, { "step": 9, "loss": -0.39284372329711914, "grad_norm": 0.05810546875, "learning_rate": 8.400000000000001e-06, "thinking/tokens_mean": 834.1726190476189, "thinking/budget_hit_fraction": 0.12797619047619047, "num_tokens": 286150.0, "completions/mean_length": 4164.0, "completions/min_length": 2423.0, "completions/max_length": 7522.0, "completions/clipped_ratio": 0.125, "completions/mean_terminated_length": 4320.14306640625, "completions/min_terminated_length": 2423.0, "completions/max_terminated_length": 7522.0, "tools/call_frequency": 2.875, "tools/failure_frequency": 0.043478261679410934, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.6691635251045227, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 0.11923417448997498, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.5833333730697632, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.1666666716337204, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/mean": 0.5235523581504822, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/std": 0.37063276767730713, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/mean": 0.25, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/std": 0.3191424012184143, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/std": 1.0, "reward": 0.7671912908554077, "reward_std": 0.4060142934322357, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 0.5, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.1782013326883316, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 2210.013791121002, "entropy_control/entropy": 0.20875006682111244, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.036, "adapter_sync_time": 0.0, "generation_time": 1868.719707, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1926.261481, "forward_backward_time": 283.632344, "rollout_overhead_time": 57.541774, "peak_allocated_gib": 66.74047183990479, "peak_reserved_gib": 74.21875 }, { "step": 10, "loss": -0.3931627869606018, "grad_norm": 0.0546875, "learning_rate": 8.2e-06, "entropy": 0.17814059741795063, "clip_ratio/low_mean": 0.0006339414248941466, "clip_ratio/low_min": 0.0006339414248941466, "clip_ratio/high_mean": 0.0006241699738893658, "clip_ratio/high_max": 0.0006241699738893658, "clip_ratio/region_mean": 0.0012581113987835124, "step_time": 283.73702334899826, "entropy_control/entropy": 0.208622635800752, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.04, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 283.621717, "rollout_overhead_time": 0.0, "peak_allocated_gib": 66.74047183990479, "peak_reserved_gib": 74.21875, "reward": null }, { "step": 11, "loss": 0.005036424845457077, "grad_norm": 0.060302734375, "learning_rate": 8.000000000000001e-06, "thinking/tokens_mean": 696.7291666666666, "thinking/budget_hit_fraction": 0.16666666666666666, "num_tokens": 347781.0, "completions/mean_length": 4394.875, "completions/min_length": 3100.0, "completions/max_length": 6185.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 4394.875, "completions/min_terminated_length": 3100.0, "completions/max_terminated_length": 6185.0, "tools/call_frequency": 2.875, "tools/failure_frequency": 0.08695652335882187, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__native/mean": 0.5260480642318726, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__native/std": 0.23834127187728882, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__efficiency/mean": 0.5416666865348816, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__efficiency/std": 0.15957120060920715, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/mean": 0.7017019987106323, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/std": 0.11484745144844055, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/mean": 0.5833333730697632, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/std": 0.09622505307197571, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/std": 0.0, "reward": 0.8420000076293945, "reward_std": 0.19645576179027557, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.21366984024643898, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 1592.7577740629968, "entropy_control/entropy": 0.21534860302940878, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.044, "adapter_sync_time": 0.0, "generation_time": 1397.723132, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1431.749457, "forward_backward_time": 160.897133, "rollout_overhead_time": 34.026325, "peak_allocated_gib": 66.74047183990479, "peak_reserved_gib": 74.21875 }, { "step": 12, "loss": 0.005168553441762924, "grad_norm": 0.05859375, "learning_rate": 7.800000000000002e-06, "entropy": 0.21324916929006577, "clip_ratio/low_mean": 0.0005457193256006576, "clip_ratio/low_min": 0.0005457193256006576, "clip_ratio/high_mean": 0.0005808843488921411, "clip_ratio/high_max": 0.0005808843488921411, "clip_ratio/region_mean": 0.0011266036744927987, "step_time": 161.014062000997, "entropy_control/entropy": 0.21491973479821413, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.048, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 160.90025, "rollout_overhead_time": 0.0, "peak_allocated_gib": 66.74047183990479, "peak_reserved_gib": 74.21875, "reward": null }, { "step": 13, "loss": 0.09634605795145035, "grad_norm": 0.054443359375, "learning_rate": 7.600000000000001e-06, "thinking/tokens_mean": 853.588888888889, "thinking/budget_hit_fraction": 0.1875, "num_tokens": 415399.0, "completions/mean_length": 5167.25, "completions/min_length": 3056.0, "completions/max_length": 8148.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 5167.25, "completions/min_terminated_length": 3056.0, "completions/max_terminated_length": 8148.0, "tools/call_frequency": 3.125, "tools/failure_frequency": 0.03999999910593033, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.5009159445762634, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 0.21727728843688965, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.625, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.0833333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/mean": 0.7053016424179077, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/std": 0.12215055525302887, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/mean": 0.375, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/std": 0.15957118570804596, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/std": 0.0, "reward": 0.8281087875366211, "reward_std": 0.19517196714878082, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.2082820888608694, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 2020.8756578060038, "entropy_control/entropy": 0.21198962241283884, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.052, "adapter_sync_time": 0.0, "generation_time": 1687.421936, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1743.634275, "forward_backward_time": 277.123339, "rollout_overhead_time": 56.212338, "peak_allocated_gib": 66.74047183990479, "peak_reserved_gib": 74.21875 }, { "step": 14, "loss": 0.09629321098327637, "grad_norm": 0.053955078125, "learning_rate": 7.4e-06, "entropy": 0.20798952691257, "clip_ratio/low_mean": 0.0007254107185872272, "clip_ratio/low_min": 0.0007254107185872272, "clip_ratio/high_mean": 0.000983088102657348, "clip_ratio/high_max": 0.000983088102657348, "clip_ratio/region_mean": 0.001708498821244575, "step_time": 277.22948470600386, "entropy_control/entropy": 0.21172358444850858, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.056, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 277.117598, "rollout_overhead_time": 0.0, "peak_allocated_gib": 66.74047183990479, "peak_reserved_gib": 74.21875, "reward": null }, { "step": 15, "loss": -0.11289608478546143, "grad_norm": 0.1220703125, "learning_rate": 7.2000000000000005e-06, "thinking/tokens_mean": 749.7755102040817, "thinking/budget_hit_fraction": 0.18367346938775508, "num_tokens": 486302.0, "completions/mean_length": 5379.375, "completions/min_length": 3071.0, "completions/max_length": 6517.0, "completions/clipped_ratio": 0.125, "completions/mean_terminated_length": 5709.14306640625, "completions/min_terminated_length": 3665.0, "completions/max_terminated_length": 6517.0, "tools/call_frequency": 3.875, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/mean": 0.6016470193862915, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/std": 0.2629297971725464, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/mean": 0.25, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/std": 0.28867512941360474, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/std": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/mean": 0.4729248285293579, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/std": 0.4145148992538452, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/mean": 0.2083333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/std": 0.31549492478370667, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/mean": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/std": 1.154700517654419, "reward": 0.59874427318573, "reward_std": 0.5368551015853882, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 0.5, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.17465469613671303, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 2361.7444208510024, "entropy_control/entropy": 0.19701933693096052, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.06, "adapter_sync_time": 0.0, "generation_time": 2046.550667, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 2100.549428, "forward_backward_time": 261.095244, "rollout_overhead_time": 53.99876, "peak_allocated_gib": 66.74047183990479, "peak_reserved_gib": 74.21875 }, { "step": 16, "loss": -0.11281206458806992, "grad_norm": 0.06103515625, "learning_rate": 7e-06, "entropy": 0.1747240461409092, "clip_ratio/low_mean": 0.0004409278044477105, "clip_ratio/low_min": 0.0004409278044477105, "clip_ratio/high_mean": 0.000910558803298045, "clip_ratio/high_max": 0.000910558803298045, "clip_ratio/region_mean": 0.0013514866077457555, "step_time": 261.20766560401535, "entropy_control/entropy": 0.19709243253552391, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.064, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 261.095538, "rollout_overhead_time": 0.0, "peak_allocated_gib": 66.74047183990479, "peak_reserved_gib": 74.21875, "reward": null }, { "step": 17, "loss": -0.19307515025138855, "grad_norm": 0.0673828125, "learning_rate": 6.800000000000001e-06, "thinking/tokens_mean": 528.0034013605442, "thinking/budget_hit_fraction": 0.09183673469387754, "num_tokens": 547833.0, "completions/mean_length": 4000.375, "completions/min_length": 2373.0, "completions/max_length": 6010.0, "completions/clipped_ratio": 0.125, "completions/mean_terminated_length": 4133.14306640625, "completions/min_terminated_length": 2373.0, "completions/max_terminated_length": 6010.0, "tools/call_frequency": 3.375, "tools/failure_frequency": 0.03703703731298447, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/mean": 0.7049051523208618, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/std": 0.12004125118255615, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/mean": 0.25, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/std": 0.09622505307197571, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/mean": 0.5523415207862854, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/std": 0.38071539998054504, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/mean": 0.4166666865348816, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/std": 0.2886751592159271, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/std": 1.0, "reward": 0.7952899932861328, "reward_std": 0.4147125780582428, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 0.5, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.18406094796955585, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 1937.928044410004, "entropy_control/entropy": 0.20794299019123733, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.068, "adapter_sync_time": 0.0, "generation_time": 1653.924235, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1702.590113, "forward_backward_time": 235.219332, "rollout_overhead_time": 48.665879, "peak_allocated_gib": 66.74047183990479, "peak_reserved_gib": 74.21875 }, { "step": 18, "loss": -0.1930653303861618, "grad_norm": 0.0693359375, "learning_rate": 6.600000000000001e-06, "entropy": 0.1838545147329569, "clip_ratio/low_mean": 0.0007282581645995378, "clip_ratio/low_min": 0.0007282581645995378, "clip_ratio/high_mean": 0.0007973067622515373, "clip_ratio/high_max": 0.0007973067622515373, "clip_ratio/region_mean": 0.0015255649268510751, "step_time": 235.31410456199956, "entropy_control/entropy": 0.2076618631034234, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.072, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 235.201377, "rollout_overhead_time": 0.0, "peak_allocated_gib": 66.74047183990479, "peak_reserved_gib": 74.21875, "reward": null }, { "step": 19, "loss": -0.07311653345823288, "grad_norm": 0.06884765625, "learning_rate": 6.4000000000000006e-06, "thinking/tokens_mean": 667.5, "thinking/budget_hit_fraction": 0.15, "num_tokens": 600513.0, "completions/mean_length": 3694.5, "completions/min_length": 1339.0, "completions/max_length": 5164.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 3694.5, "completions/min_terminated_length": 1339.0, "completions/max_terminated_length": 5164.0, "tools/call_frequency": 2.875, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.6101748943328857, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 0.025337064638733864, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.5416666865348816, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.15957120060920715, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/mean": 0.8939128518104553, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/std": 0.07854292541742325, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/std": 0.13608276844024658, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/std": 0.0, "reward": 0.9780855178833008, "reward_std": 0.15884076058864594, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.19092376343905926, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 1460.218944550994, "entropy_control/entropy": 0.19918622018720106, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.076, "adapter_sync_time": 0.0, "generation_time": 1302.994384, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1331.426936, "forward_backward_time": 128.675874, "rollout_overhead_time": 28.432553, "peak_allocated_gib": 66.74047183990479, "peak_reserved_gib": 74.240234375 }, { "step": 20, "loss": -0.0731862410902977, "grad_norm": 0.06787109375, "learning_rate": 6.200000000000001e-06, "entropy": 0.1917433701455593, "clip_ratio/low_mean": 0.0005687122466042638, "clip_ratio/low_min": 0.0005687122466042638, "clip_ratio/high_mean": 0.0010158204386243597, "clip_ratio/high_max": 0.0010158204386243597, "clip_ratio/region_mean": 0.0015845326852286234, "step_time": 128.79446079900663, "entropy_control/entropy": 0.20005662801063512, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.08, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 128.68168, "rollout_overhead_time": 0.0, "peak_allocated_gib": 66.74047183990479, "peak_reserved_gib": 74.240234375, "reward": null }, { "step": 21, "loss": -0.13835109770298004, "grad_norm": 0.05908203125, "learning_rate": 6e-06, "thinking/tokens_mean": 742.0208333333334, "thinking/budget_hit_fraction": 0.20833333333333334, "num_tokens": 668260.0, "completions/mean_length": 4769.375, "completions/min_length": 2701.0, "completions/max_length": 7890.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 4769.375, "completions/min_terminated_length": 2701.0, "completions/max_terminated_length": 7890.0, "tools/call_frequency": 3.25, "tools/failure_frequency": 0.07692307978868484, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/mean": 0.6688806414604187, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/std": 0.06329096108675003, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/mean": 0.375, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/std": 0.15957118570804596, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/mean": 0.6173508167266846, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/std": 0.015867799520492554, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/mean": 0.625, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/std": 0.0833333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/std": 0.0, "reward": 0.8681157231330872, "reward_std": 0.043037109076976776, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.19520873576402664, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 2067.7043030540044, "entropy_control/entropy": 0.20241929057839275, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.084, "adapter_sync_time": 0.0, "generation_time": 1738.580876, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1794.518201, "forward_backward_time": 273.082422, "rollout_overhead_time": 55.937325, "peak_allocated_gib": 66.74047183990479, "peak_reserved_gib": 74.240234375 }, { "step": 22, "loss": -0.13812217116355896, "grad_norm": 0.059814453125, "learning_rate": 5.8e-06, "entropy": 0.19527459517121315, "clip_ratio/low_mean": 0.0012276367633603513, "clip_ratio/low_min": 0.0012276367633603513, "clip_ratio/high_mean": 0.0003591680942918174, "clip_ratio/high_max": 0.0003591680942918174, "clip_ratio/region_mean": 0.0015868048576521687, "step_time": 273.1928589199997, "entropy_control/entropy": 0.20245840756560973, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.088, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 273.088669, "rollout_overhead_time": 0.0, "peak_allocated_gib": 66.74047183990479, "peak_reserved_gib": 74.240234375, "reward": null }, { "step": 23, "loss": -0.2540090084075928, "grad_norm": 0.07177734375, "learning_rate": 5.600000000000001e-06, "thinking/tokens_mean": 695.7666666666667, "thinking/budget_hit_fraction": 0.15, "num_tokens": 721754.0, "completions/mean_length": 3801.25, "completions/min_length": 2753.0, "completions/max_length": 6114.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 3801.25, "completions/min_terminated_length": 2753.0, "completions/max_terminated_length": 6114.0, "tools/call_frequency": 2.5, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.6682802438735962, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 0.11743566393852234, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.5833333730697632, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.1666666716337204, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/mean": 0.8659225702285767, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/std": 0.08097398281097412, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/mean": 0.5833333730697632, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/std": 0.09622505307197571, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/std": 0.0, "reward": 0.9962680339813232, "reward_std": 0.1374383568763733, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.1908215843141079, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 1290.6729723049975, "entropy_control/entropy": 0.18894690473745326, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.092, "adapter_sync_time": 0.0, "generation_time": 1098.338165, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1131.757219, "forward_backward_time": 158.801012, "rollout_overhead_time": 33.419053, "peak_allocated_gib": 66.74047183990479, "peak_reserved_gib": 74.240234375 }, { "step": 24, "loss": -0.25433793663978577, "grad_norm": 0.0693359375, "learning_rate": 5.400000000000001e-06, "entropy": 0.1907775029540062, "clip_ratio/low_mean": 0.0015442325093317777, "clip_ratio/low_min": 0.0015442325093317777, "clip_ratio/high_mean": 0.0001226692838827148, "clip_ratio/high_max": 0.0001226692838827148, "clip_ratio/region_mean": 0.0016669017932144925, "step_time": 158.91810345100748, "entropy_control/entropy": 0.18890588621257712, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.096, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 158.800509, "rollout_overhead_time": 0.0, "peak_allocated_gib": 66.74047183990479, "peak_reserved_gib": 74.240234375, "reward": null }, { "step": 25, "loss": -0.3781495988368988, "grad_norm": 0.0703125, "learning_rate": 5.2e-06, "thinking/tokens_mean": 569.3003401360544, "thinking/budget_hit_fraction": 0.07142857142857142, "num_tokens": 779184.0, "completions/mean_length": 4088.75, "completions/min_length": 1462.0, "completions/max_length": 6761.0, "completions/clipped_ratio": 0.125, "completions/mean_terminated_length": 4234.14306640625, "completions/min_terminated_length": 1462.0, "completions/max_terminated_length": 6761.0, "tools/call_frequency": 3.625, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.7392415404319763, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 0.13263945281505585, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.3750000298023224, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.28463754057884216, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/mean": 0.40354466438293457, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/std": 0.34268611669540405, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/mean": 0.1666666716337204, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/std": 0.3333333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/mean": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/std": 1.154700517654419, "reward": 0.6849347352981567, "reward_std": 0.48197078704833984, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 0.5, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.16409864276647568, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 1867.9587041330087, "entropy_control/entropy": 0.19068812964328846, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.1, "adapter_sync_time": 0.0, "generation_time": 1560.219861, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1612.384982, "forward_backward_time": 255.480551, "rollout_overhead_time": 52.165122, "peak_allocated_gib": 66.74047183990479, "peak_reserved_gib": 74.240234375 }, { "step": 26, "loss": -0.37836676836013794, "grad_norm": 0.06396484375, "learning_rate": 5e-06, "entropy": 0.16466358676552773, "clip_ratio/low_mean": 0.0010647143644746393, "clip_ratio/low_min": 0.0010647143644746393, "clip_ratio/high_mean": 0.0006297622821875848, "clip_ratio/high_max": 0.0006297622821875848, "clip_ratio/region_mean": 0.001694476646662224, "step_time": 255.583496878, "entropy_control/entropy": 0.19107552944040063, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.104, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 255.474175, "rollout_overhead_time": 0.0, "peak_allocated_gib": 66.74047183990479, "peak_reserved_gib": 74.240234375, "reward": null }, { "step": 27, "loss": -0.25493109226226807, "grad_norm": 0.07666015625, "learning_rate": 4.800000000000001e-06, "thinking/tokens_mean": 594.0025510204081, "thinking/budget_hit_fraction": 0.10714285714285714, "num_tokens": 835601.0, "completions/mean_length": 3749.625, "completions/min_length": 1408.0, "completions/max_length": 5843.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 3749.625, "completions/min_terminated_length": 1408.0, "completions/max_terminated_length": 5843.0, "tools/call_frequency": 2.875, "tools/failure_frequency": 0.08695652335882187, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/mean": 0.6414066553115845, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/std": 0.06320184469223022, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/mean": 0.5833333730697632, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/std": 0.1666666716337204, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/mean": 0.6505849361419678, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/std": 0.4378522038459778, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/mean": 0.2916666865348816, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/std": 0.34359216690063477, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/std": 1.0, "reward": 0.8178707957267761, "reward_std": 0.4275791347026825, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 0.5, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.17324734199792147, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 1640.3296975979974, "entropy_control/entropy": 0.17818871078912338, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.108, "adapter_sync_time": 0.0, "generation_time": 1342.781024, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1392.775115, "forward_backward_time": 247.44297, "rollout_overhead_time": 49.994092, "peak_allocated_gib": 66.74047183990479, "peak_reserved_gib": 74.240234375 }, { "step": 28, "loss": -0.2548600435256958, "grad_norm": 0.06884765625, "learning_rate": 4.600000000000001e-06, "entropy": 0.1727202758193016, "clip_ratio/low_mean": 0.0009054355468833819, "clip_ratio/low_min": 0.0009054355468833819, "clip_ratio/high_mean": 0.000712330816895701, "clip_ratio/high_max": 0.000712330816895701, "clip_ratio/region_mean": 0.001617766363779083, "step_time": 247.52918564000356, "entropy_control/entropy": 0.17790918098791764, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.112, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 247.419597, "rollout_overhead_time": 0.0, "peak_allocated_gib": 66.74047183990479, "peak_reserved_gib": 74.240234375, "reward": null }, { "step": 29, "loss": -0.07595027983188629, "grad_norm": 0.059326171875, "learning_rate": 4.4e-06, "thinking/tokens_mean": 665.0119047619047, "thinking/budget_hit_fraction": 0.19642857142857142, "num_tokens": 902322.0, "completions/mean_length": 5057.625, "completions/min_length": 2928.0, "completions/max_length": 6352.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 5057.625, "completions/min_terminated_length": 2928.0, "completions/max_terminated_length": 6352.0, "tools/call_frequency": 3.875, "tools/failure_frequency": 0.06451612710952759, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.6719769239425659, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 0.12673872709274292, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.5416666865348816, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.15957120060920715, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/mean": 0.5428180694580078, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/std": 0.2034512609243393, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/mean": 0.25, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/std": 0.21516574919223785, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/std": 1.0, "reward": 0.7771891355514526, "reward_std": 0.30675458908081055, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.22062398493289948, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 1969.6545821579966, "entropy_control/entropy": 0.2188746276922711, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.116, "adapter_sync_time": 0.0, "generation_time": 1677.420093, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1727.592026, "forward_backward_time": 241.949094, "rollout_overhead_time": 50.171933, "peak_allocated_gib": 66.74047183990479, "peak_reserved_gib": 74.240234375 }, { "step": 30, "loss": -0.07589748501777649, "grad_norm": 0.0595703125, "learning_rate": 4.2000000000000004e-06, "entropy": 0.22067385911941528, "clip_ratio/low_mean": 0.0010641067347023636, "clip_ratio/low_min": 0.0010641067347023636, "clip_ratio/high_mean": 0.0005229517410043627, "clip_ratio/high_max": 0.0005229517410043627, "clip_ratio/region_mean": 0.0015870584757067263, "step_time": 242.06835631800277, "entropy_control/entropy": 0.21898990172933744, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.12, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 241.95705, "rollout_overhead_time": 0.0, "peak_allocated_gib": 66.74047183990479, "peak_reserved_gib": 74.240234375, "reward": null }, { "step": 31, "loss": -0.05214519053697586, "grad_norm": 0.078125, "learning_rate": 4.000000000000001e-06, "thinking/tokens_mean": 667.8375850340136, "thinking/budget_hit_fraction": 0.125, "num_tokens": 967693.0, "completions/mean_length": 5087.875, "completions/min_length": 2771.0, "completions/max_length": 7788.0, "completions/clipped_ratio": 0.125, "completions/mean_terminated_length": 4702.14306640625, "completions/min_terminated_length": 2771.0, "completions/max_terminated_length": 7074.0, "tools/call_frequency": 4.0, "tools/failure_frequency": 0.03125, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/mean": 0.6885681748390198, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/std": 0.29825878143310547, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/mean": 0.3541666865348816, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/std": 0.2077372521162033, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/mean": 0.75, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/std": 0.7071067690849304, "reward": 0.8562765121459961, "reward_std": 0.4053352475166321, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.1631258949637413, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 2172.747359213983, "entropy_control/entropy": 0.18656705000969062, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.124, "adapter_sync_time": 0.0, "generation_time": 1868.67497, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1919.549486, "forward_backward_time": 253.084808, "rollout_overhead_time": 50.874516, "peak_allocated_gib": 66.74047183990479, "peak_reserved_gib": 74.240234375 }, { "step": 32, "loss": -0.05206791311502457, "grad_norm": 0.0625, "learning_rate": 3.8000000000000005e-06, "entropy": 0.1631073895841837, "clip_ratio/low_mean": 0.0007781426684232429, "clip_ratio/low_min": 0.0007781426684232429, "clip_ratio/high_mean": 0.0006902424938743934, "clip_ratio/high_max": 0.0006902424938743934, "clip_ratio/region_mean": 0.0014683851622976363, "step_time": 253.1972244689896, "entropy_control/entropy": 0.18660941503527725, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.128, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 253.085928, "rollout_overhead_time": 0.0, "peak_allocated_gib": 66.74047183990479, "peak_reserved_gib": 74.240234375, "reward": null }, { "step": 33, "loss": 0.1136540025472641, "grad_norm": 0.0625, "learning_rate": 3.6000000000000003e-06, "thinking/tokens_mean": 555.125, "thinking/budget_hit_fraction": 0.125, "num_tokens": 1022886.0, "completions/mean_length": 4003.625, "completions/min_length": 2284.0, "completions/max_length": 6209.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 4003.625, "completions/min_terminated_length": 2284.0, "completions/max_terminated_length": 6209.0, "tools/call_frequency": 3.375, "tools/failure_frequency": 0.07407407462596893, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.6746612787246704, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 0.10988383740186691, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.235702246427536, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/mean": 0.5187197923660278, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/std": 0.3549153804779053, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/mean": 0.4583333730697632, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/std": 0.3154948949813843, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/std": 1.0, "reward": 0.7706488370895386, "reward_std": 0.3620922863483429, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.19170449674129486, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 1623.1493520590084, "entropy_control/entropy": 0.1938683924069639, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.132, "adapter_sync_time": 0.0, "generation_time": 1411.504947, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1448.883133, "forward_backward_time": 174.159315, "rollout_overhead_time": 37.378186, "peak_allocated_gib": 66.74047183990479, "peak_reserved_gib": 74.240234375 }, { "step": 34, "loss": 0.11371944844722748, "grad_norm": 0.0673828125, "learning_rate": 3.4000000000000005e-06, "entropy": 0.19164535589516163, "clip_ratio/low_mean": 0.0013614757772302255, "clip_ratio/low_min": 0.0013614757772302255, "clip_ratio/high_mean": 0.0004331821473897435, "clip_ratio/high_max": 0.0004331821473897435, "clip_ratio/region_mean": 0.001794657924619969, "step_time": 174.2861271330039, "entropy_control/entropy": 0.1939651095572516, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.136, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 174.170299, "rollout_overhead_time": 0.0, "peak_allocated_gib": 66.74047183990479, "peak_reserved_gib": 74.240234375, "reward": null }, { "step": 35, "loss": -0.32272768020629883, "grad_norm": 0.058837890625, "learning_rate": 3.2000000000000003e-06, "thinking/tokens_mean": 756.1493197278912, "thinking/budget_hit_fraction": 0.14795918367346936, "num_tokens": 1093376.0, "completions/mean_length": 4948.25, "completions/min_length": 2220.0, "completions/max_length": 7897.0, "completions/clipped_ratio": 0.125, "completions/mean_terminated_length": 5216.4287109375, "completions/min_terminated_length": 2220.0, "completions/max_terminated_length": 7897.0, "tools/call_frequency": 3.375, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/mean": 0.5765712261199951, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/std": 0.25050315260887146, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/mean": 0.3125, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/std": 0.3012869358062744, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/mean": 0.75, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/std": 0.7071067690849304, "reward": 0.7421962022781372, "reward_std": 0.39045342803001404, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 0.5, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.17574194259941578, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 2223.7072799999805, "entropy_control/entropy": 0.2012081827944982, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.14, "adapter_sync_time": 0.0, "generation_time": 1863.796411, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1924.116497, "forward_backward_time": 299.476099, "rollout_overhead_time": 60.320086, "peak_allocated_gib": 68.41224241256714, "peak_reserved_gib": 76.193359375 }, { "step": 36, "loss": -0.3228744566440582, "grad_norm": 0.050048828125, "learning_rate": 3e-06, "entropy": 0.17565468326210976, "clip_ratio/low_mean": 0.0007310346118174493, "clip_ratio/low_min": 0.0007310346118174493, "clip_ratio/high_mean": 0.0003735209902515635, "clip_ratio/high_max": 0.0003735209902515635, "clip_ratio/region_mean": 0.0011045556020690128, "step_time": 299.5672293989919, "entropy_control/entropy": 0.201147767312505, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.144, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 299.46275, "rollout_overhead_time": 0.0, "peak_allocated_gib": 68.41224241256714, "peak_reserved_gib": 76.193359375, "reward": null }, { "step": 37, "loss": -0.17068979144096375, "grad_norm": 0.07861328125, "learning_rate": 2.8000000000000003e-06, "thinking/tokens_mean": 513.6964285714286, "thinking/budget_hit_fraction": 0.14285714285714285, "num_tokens": 1145984.0, "completions/mean_length": 3696.0, "completions/min_length": 1549.0, "completions/max_length": 5708.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 3696.0, "completions/min_terminated_length": 1549.0, "completions/max_terminated_length": 5708.0, "tools/call_frequency": 2.875, "tools/failure_frequency": 0.08695652335882187, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.7275477647781372, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 0.13607469201087952, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.5833333730697632, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.09622505307197571, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__native/mean": 0.656780481338501, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__native/std": 0.0910348892211914, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__efficiency/mean": 0.5416666865348816, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__efficiency/std": 0.25, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__reliability/std": 0.0, "reward": 0.920289158821106, "reward_std": 0.1074429303407669, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.20684580318629742, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 1281.9606787930024, "entropy_control/entropy": 0.2127551682690933, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.148, "adapter_sync_time": 0.0, "generation_time": 1059.999023, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1098.757953, "forward_backward_time": 183.087473, "rollout_overhead_time": 38.75893, "peak_allocated_gib": 68.41224241256714, "peak_reserved_gib": 76.193359375 }, { "step": 38, "loss": -0.17084544897079468, "grad_norm": 0.07421875, "learning_rate": 2.6e-06, "entropy": 0.207224877551198, "clip_ratio/low_mean": 0.0018302314565517008, "clip_ratio/low_min": 0.0018302314565517008, "clip_ratio/high_mean": 0.0003249518049415201, "clip_ratio/high_max": 0.0003249518049415201, "clip_ratio/region_mean": 0.002155183261493221, "step_time": 183.18254053301644, "entropy_control/entropy": 0.21309143610341108, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.152, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 183.082426, "rollout_overhead_time": 0.0, "peak_allocated_gib": 68.41224241256714, "peak_reserved_gib": 76.193359375, "reward": null }, { "step": 39, "loss": -0.09496024250984192, "grad_norm": 0.046875, "learning_rate": 2.4000000000000003e-06, "thinking/tokens_mean": 690.9642857142857, "thinking/budget_hit_fraction": 0.125, "num_tokens": 1211456.0, "completions/mean_length": 5093.0, "completions/min_length": 3470.0, "completions/max_length": 7826.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 5093.0, "completions/min_terminated_length": 3470.0, "completions/max_terminated_length": 7826.0, "tools/call_frequency": 3.75, "tools/failure_frequency": 0.03333333507180214, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__native/mean": 0.7511086463928223, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__native/std": 0.0961092934012413, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__efficiency/mean": 0.3958333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__efficiency/std": 0.23464767634868622, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__reliability/std": 0.0, "reward": 0.970900297164917, "reward_std": 0.0874413251876831, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.21844188123941422, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 1925.603753813979, "entropy_control/entropy": 0.2150383363867733, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.156, "adapter_sync_time": 0.0, "generation_time": 1614.123999, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1667.01614, "forward_backward_time": 258.482786, "rollout_overhead_time": 52.892141, "peak_allocated_gib": 68.41224241256714, "peak_reserved_gib": 76.193359375 }, { "step": 40, "loss": -0.0950980931520462, "grad_norm": 0.046142578125, "learning_rate": 2.2e-06, "entropy": 0.21833787113428116, "clip_ratio/low_mean": 0.0007938431372167543, "clip_ratio/low_min": 0.0007938431372167543, "clip_ratio/high_mean": 0.0005773864177172072, "clip_ratio/high_max": 0.0005773864177172072, "clip_ratio/region_mean": 0.0013712295549339615, "step_time": 258.53294155299955, "entropy_control/entropy": 0.2149255913309748, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.16, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 258.42222, "rollout_overhead_time": 0.0, "peak_allocated_gib": 68.41224241256714, "peak_reserved_gib": 76.193359375, "reward": null }, { "step": 41, "loss": 0.1262494921684265, "grad_norm": 0.057373046875, "learning_rate": 2.0000000000000003e-06, "thinking/tokens_mean": 679.2714285714285, "thinking/budget_hit_fraction": 0.10714285714285714, "num_tokens": 1279842.0, "completions/mean_length": 5071.25, "completions/min_length": 2665.0, "completions/max_length": 8500.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 5071.25, "completions/min_terminated_length": 2665.0, "completions/max_terminated_length": 8500.0, "tools/call_frequency": 4.125, "tools/failure_frequency": 0.03030303120613098, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__native/mean": 0.643683910369873, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__native/std": 0.34630754590034485, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__efficiency/mean": 0.4583333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__efficiency/std": 0.2500000298023224, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/mean": 0.7062281370162964, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/std": 0.34980955719947815, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/mean": 0.2083333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/std": 0.3154948949813843, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/std": 1.0, "reward": 0.8416227102279663, "reward_std": 0.4144892990589142, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.19727547280490398, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 2245.0404615720036, "entropy_control/entropy": 0.20188952450310807, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.164, "adapter_sync_time": 0.0, "generation_time": 1885.256483, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1945.812313, "forward_backward_time": 299.114605, "rollout_overhead_time": 60.555831, "peak_allocated_gib": 68.46192026138306, "peak_reserved_gib": 76.251953125 }, { "step": 42, "loss": 0.12635992467403412, "grad_norm": 0.05859375, "learning_rate": 1.8000000000000001e-06, "entropy": 0.1972111500799656, "clip_ratio/low_mean": 0.0007892334251664579, "clip_ratio/low_min": 0.0007892334251664579, "clip_ratio/high_mean": 0.0006546082731802016, "clip_ratio/high_max": 0.0006546082731802016, "clip_ratio/region_mean": 0.0014438416983466595, "step_time": 299.2247267749917, "entropy_control/entropy": 0.2018908316259224, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.168, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 299.11161, "rollout_overhead_time": 0.0, "peak_allocated_gib": 68.46192026138306, "peak_reserved_gib": 76.251953125, "reward": null }, { "step": 43, "loss": -0.05368418991565704, "grad_norm": 0.061767578125, "learning_rate": 1.6000000000000001e-06, "thinking/tokens_mean": 576.9719387755102, "thinking/budget_hit_fraction": 0.05357142857142857, "num_tokens": 1342394.0, "completions/mean_length": 4515.0, "completions/min_length": 2961.0, "completions/max_length": 7979.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 4515.0, "completions/min_terminated_length": 2961.0, "completions/max_terminated_length": 7979.0, "tools/call_frequency": 3.875, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__native/mean": 0.7894160747528076, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__native/std": 0.09969109296798706, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__efficiency/mean": 0.3333333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__efficiency/std": 0.13608276844024658, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/mean": 0.7056864500045776, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/std": 0.10593900829553604, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/mean": 0.375, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/std": 0.28463754057884216, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/std": 0.0, "reward": 0.9652596116065979, "reward_std": 0.10003283619880676, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.19226796738803387, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 2151.2987192989967, "entropy_control/entropy": 0.19243019402606767, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.172, "adapter_sync_time": 0.0, "generation_time": 1836.32729, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1890.224049, "forward_backward_time": 260.972588, "rollout_overhead_time": 53.896759, "peak_allocated_gib": 68.46192026138306, "peak_reserved_gib": 76.251953125 }, { "step": 44, "loss": -0.053502753376960754, "grad_norm": 0.06103515625, "learning_rate": 1.4000000000000001e-06, "entropy": 0.19260794669389725, "clip_ratio/low_mean": 0.0012998088932363316, "clip_ratio/low_min": 0.0012998088932363316, "clip_ratio/high_mean": 0.0004901139982393943, "clip_ratio/high_max": 0.0004901139982393943, "clip_ratio/region_mean": 0.001789922891475726, "step_time": 261.0669290939986, "entropy_control/entropy": 0.19283728370264044, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.176, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 260.963191, "rollout_overhead_time": 0.0, "peak_allocated_gib": 68.46192026138306, "peak_reserved_gib": 76.251953125, "reward": null }, { "step": 45, "loss": -0.355094313621521, "grad_norm": 0.056640625, "learning_rate": 1.2000000000000002e-06, "thinking/tokens_mean": 628.6498015873016, "thinking/budget_hit_fraction": 0.1755952380952381, "num_tokens": 1405039.0, "completions/mean_length": 4160.625, "completions/min_length": 2181.0, "completions/max_length": 6990.0, "completions/clipped_ratio": 0.125, "completions/mean_terminated_length": 4316.2861328125, "completions/min_terminated_length": 2181.0, "completions/max_terminated_length": 6990.0, "tools/call_frequency": 3.375, "tools/failure_frequency": 0.07407407462596893, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/mean": 0.5175503492355347, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/std": 0.34920167922973633, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/mean": 0.2916666865348816, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/std": 0.25, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/std": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/mean": 0.6731905341148376, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/std": 0.07375707477331161, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/mean": 0.4166666865348816, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/std": 0.28867512941360474, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/std": 0.0, "reward": 0.7630787491798401, "reward_std": 0.3925502300262451, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 0.5, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.14445575140416622, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 1828.9440834909983, "entropy_control/entropy": 0.1795795040736851, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.18, "adapter_sync_time": 0.0, "generation_time": 1522.479237, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1574.986152, "forward_backward_time": 253.863638, "rollout_overhead_time": 52.506915, "peak_allocated_gib": 68.46192026138306, "peak_reserved_gib": 76.251953125 }, { "step": 46, "loss": -0.3554255962371826, "grad_norm": 0.058837890625, "learning_rate": 1.0000000000000002e-06, "entropy": 0.14436963014304638, "clip_ratio/low_mean": 0.0007786930655129254, "clip_ratio/low_min": 0.0007786930655129254, "clip_ratio/high_mean": 0.0010567690624156967, "clip_ratio/high_max": 0.0010567690624156967, "clip_ratio/region_mean": 0.001835462127928622, "step_time": 253.96728759899997, "entropy_control/entropy": 0.17943839526184938, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.184, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 253.87415, "rollout_overhead_time": 0.0, "peak_allocated_gib": 68.46192026138306, "peak_reserved_gib": 76.251953125, "reward": null }, { "step": 47, "loss": -0.2797643542289734, "grad_norm": 0.07373046875, "learning_rate": 8.000000000000001e-07, "thinking/tokens_mean": 706.6880952380952, "thinking/budget_hit_fraction": 0.20357142857142857, "num_tokens": 1460341.0, "completions/mean_length": 3626.25, "completions/min_length": 2777.0, "completions/max_length": 5604.0, "completions/clipped_ratio": 0.125, "completions/mean_terminated_length": 3705.571533203125, "completions/min_terminated_length": 2777.0, "completions/max_terminated_length": 5604.0, "tools/call_frequency": 2.25, "tools/failure_frequency": 0.0555555559694767, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/mean": 0.6685474514961243, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/std": 0.11800849437713623, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/mean": 0.5833333730697632, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/std": 0.1666666716337204, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/mean": 0.45637863874435425, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/std": 0.4318944215774536, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/mean": 0.4583333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/std": 0.3154948949813843, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/std": 1.0, "reward": 0.7385047674179077, "reward_std": 0.4371601641178131, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 0.5, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.16909927129745483, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 1298.146168375999, "entropy_control/entropy": 0.19756781492478873, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.188, "adapter_sync_time": 0.0, "generation_time": 1092.833734, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1129.068102, "forward_backward_time": 168.976474, "rollout_overhead_time": 36.234368, "peak_allocated_gib": 68.46192026138306, "peak_reserved_gib": 76.251953125 }, { "step": 48, "loss": -0.27966615557670593, "grad_norm": 0.06982421875, "learning_rate": 6.000000000000001e-07, "entropy": 0.1692350823432207, "clip_ratio/low_mean": 0.001093564831535332, "clip_ratio/low_min": 0.001093564831535332, "clip_ratio/high_mean": 0.00047645412269048393, "clip_ratio/high_max": 0.00047645412269048393, "clip_ratio/region_mean": 0.0015700189542258158, "step_time": 169.10284090100322, "entropy_control/entropy": 0.1977788357877354, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.192, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 169.004128, "rollout_overhead_time": 0.0, "peak_allocated_gib": 68.46192026138306, "peak_reserved_gib": 76.251953125, "reward": null }, { "step": 49, "loss": 0.017329305410385132, "grad_norm": 0.06884765625, "learning_rate": 4.0000000000000003e-07, "thinking/tokens_mean": 594.172619047619, "thinking/budget_hit_fraction": 0.14285714285714285, "num_tokens": 1520047.0, "completions/mean_length": 4172.25, "completions/min_length": 1384.0, "completions/max_length": 7241.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 4172.25, "completions/min_terminated_length": 1384.0, "completions/max_terminated_length": 7241.0, "tools/call_frequency": 3.5, "tools/failure_frequency": 0.0357142873108387, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/mean": 0.7230167388916016, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/std": 0.13191144168376923, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/mean": 0.4166666865348816, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/std": 0.3191424012184143, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/mean": 0.8234126567840576, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/std": 0.004687488079071045, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/mean": 0.4583333730697632, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/std": 0.3154948949813843, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/std": 0.0, "reward": 0.9950897693634033, "reward_std": 0.09558027237653732, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.18376885168254375, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 1922.4227069150002, "entropy_control/entropy": 0.19315520377796397, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.196, "adapter_sync_time": 0.0, "generation_time": 1594.478549, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1648.977091, "forward_backward_time": 273.338029, "rollout_overhead_time": 54.498542, "peak_allocated_gib": 68.46192026138306, "peak_reserved_gib": 76.251953125 }, { "step": 50, "loss": 0.017506033182144165, "grad_norm": 0.072265625, "learning_rate": 2.0000000000000002e-07, "entropy": 0.18357139453291893, "clip_ratio/low_mean": 0.0009963576740119606, "clip_ratio/low_min": 0.0009963576740119606, "clip_ratio/high_mean": 0.0009350554882985307, "clip_ratio/high_max": 0.0009350554882985307, "clip_ratio/region_mean": 0.0019314131623104913, "step_time": 273.44746495200525, "entropy_control/entropy": 0.19305243055411617, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.2, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 273.334569, "rollout_overhead_time": 0.0, "peak_allocated_gib": 68.46192026138306, "peak_reserved_gib": 76.251953125, "reward": null } ], "holdout": [ { "seed": 900000, "sample": 0, "native_reward": 0.6089882858614486, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 2542, "reward_channels": { "native": 0.6089882858614486, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "two_pool_cycle" }, { "seed": 900000, "sample": 1, "native_reward": 0.8404385754973732, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 3328, "reward_channels": { "native": 0.8404385754973732, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "two_pool_cycle" }, { "seed": 900000, "sample": 2, "native_reward": 0.8404385754973732, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 3, "model_tokens": 4721, "reward_channels": { "native": 0.8404385754973732, "efficiency": 0.5, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "two_pool_cycle" }, { "seed": 900000, "sample": 3, "native_reward": 0.6089882858614486, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 2496, "reward_channels": { "native": 0.6089882858614486, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "two_pool_cycle" }, { "seed": 900001, "sample": 0, "native_reward": 0.6093825392865522, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 3384, "reward_channels": { "native": 0.6093825392865522, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "two_pool_cycle" }, { "seed": 900001, "sample": 1, "native_reward": 0.6093825392865522, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 3318, "reward_channels": { "native": 0.6093825392865522, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "two_pool_cycle" }, { "seed": 900001, "sample": 2, "native_reward": 0.0, "promotion_success": false, "done": false, "errors": 0, "tool_steps": 0, "model_tokens": 911, "reward_channels": { "native": 0.0, "efficiency": 0.0, "reliability": -1.0 }, "verification_level": null, "failed_gates": null, "family": "two_pool_cycle" }, { "seed": 900001, "sample": 3, "native_reward": 0.6093825392865522, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 3254, "reward_channels": { "native": 0.6093825392865522, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "two_pool_cycle" }, { "seed": 900000, "sample": 0, "native_reward": 0.8204653155141224, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 2445, "reward_channels": { "native": 0.8204653155141224, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "triangular_cycle" }, { "seed": 900000, "sample": 1, "native_reward": 0.6106601460847819, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 3217, "reward_channels": { "native": 0.6106601460847819, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "triangular_cycle" }, { "seed": 900000, "sample": 2, "native_reward": 0.8428295135672316, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 6, "model_tokens": 6975, "reward_channels": { "native": 0.8428295135672316, "efficiency": 0.0, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "triangular_cycle" }, { "seed": 900000, "sample": 3, "native_reward": 0.8522045135672316, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 3, "model_tokens": 3920, "reward_channels": { "native": 0.8522045135672316, "efficiency": 0.5, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "triangular_cycle" }, { "seed": 900001, "sample": 0, "native_reward": 0.7060268919310104, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 5330, "reward_channels": { "native": 0.7060268919310104, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "triangular_cycle" }, { "seed": 900001, "sample": 1, "native_reward": 0.811786874720879, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 3, "model_tokens": 1901, "reward_channels": { "native": 0.811786874720879, "efficiency": 0.5, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "triangular_cycle" }, { "seed": 900001, "sample": 2, "native_reward": 0.807099374720879, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 5, "model_tokens": 4850, "reward_channels": { "native": 0.807099374720879, "efficiency": 0.16666666666666663, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "triangular_cycle" }, { "seed": 900001, "sample": 3, "native_reward": 0.811786874720879, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 5110, "reward_channels": { "native": 0.811786874720879, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "triangular_cycle" }, { "seed": 900000, "sample": 0, "native_reward": 0.6185941887823871, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 7543, "reward_channels": { "native": 0.6185941887823871, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "competing_cycles" }, { "seed": 900000, "sample": 1, "native_reward": 0.6502053079500402, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 4105, "reward_channels": { "native": 0.6502053079500402, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "competing_cycles" }, { "seed": 900000, "sample": 2, "native_reward": 0.8720485688510347, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 3, "model_tokens": 5804, "reward_channels": { "native": 0.8720485688510347, "efficiency": 0.5, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "competing_cycles" }, { "seed": 900000, "sample": 3, "native_reward": 0.6140552238672119, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 4088, "reward_channels": { "native": 0.6140552238672119, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "competing_cycles" }, { "seed": 900001, "sample": 0, "native_reward": 0.7053628099527464, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 5401, "reward_channels": { "native": 0.7053628099527464, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "competing_cycles" }, { "seed": 900001, "sample": 1, "native_reward": 0.6959878099527463, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 6, "model_tokens": 7482, "reward_channels": { "native": 0.6959878099527463, "efficiency": 0.0, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "competing_cycles" }, { "seed": 900001, "sample": 2, "native_reward": 0.6051358982971253, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 6, "model_tokens": 6933, "reward_channels": { "native": 0.6051358982971253, "efficiency": 0.0, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "competing_cycles" }, { "seed": 900001, "sample": 3, "native_reward": 0.6053090928016737, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 5, "model_tokens": 8546, "reward_channels": { "native": 0.6053090928016737, "efficiency": 0.16666666666666663, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "competing_cycles" }, { "seed": 900000, "sample": 0, "native_reward": 0.7456673418171573, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 5, "model_tokens": 6103, "reward_channels": { "native": 0.7456673418171573, "efficiency": 0.16666666666666663, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "four_hop_cycle" }, { "seed": 900000, "sample": 1, "native_reward": 0.6412222518441313, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 3, "model_tokens": 5299, "reward_channels": { "native": 0.6412222518441313, "efficiency": 0.5, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "four_hop_cycle" }, { "seed": 900000, "sample": 2, "native_reward": 0.609376392538787, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 1392, "reward_channels": { "native": 0.609376392538787, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "four_hop_cycle" }, { "seed": 900000, "sample": 3, "native_reward": 0.7503548418171574, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 3, "model_tokens": 3231, "reward_channels": { "native": 0.7503548418171574, "efficiency": 0.5, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "four_hop_cycle" }, { "seed": 900001, "sample": 0, "native_reward": 0.739978752793518, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 3, "model_tokens": 5252, "reward_channels": { "native": 0.739978752793518, "efficiency": 0.5, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "four_hop_cycle" }, { "seed": 900001, "sample": 1, "native_reward": 0.6095701077161815, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 3015, "reward_channels": { "native": 0.6095701077161815, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "four_hop_cycle" }, { "seed": 900001, "sample": 2, "native_reward": 0.25, "promotion_success": false, "done": false, "errors": 0, "tool_steps": 6, "model_tokens": 5040, "reward_channels": { "native": 0.25, "efficiency": 0.0, "reliability": -1.0 }, "verification_level": null, "failed_gates": [ "finalized" ], "family": "four_hop_cycle" }, { "seed": 900001, "sample": 3, "native_reward": 0.6095701077161815, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 4438, "reward_channels": { "native": 0.6095701077161815, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "four_hop_cycle" }, { "seed": 900000, "sample": 0, "native_reward": 0.962057681977345, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 5392, "reward_channels": { "native": 0.962057681977345, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "size_sensitive_cycle" }, { "seed": 900000, "sample": 1, "native_reward": 0.8262556477476042, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 4431, "reward_channels": { "native": 0.8262556477476042, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "size_sensitive_cycle" }, { "seed": 900000, "sample": 2, "native_reward": 0.8262556477476042, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 3, "model_tokens": 4198, "reward_channels": { "native": 0.8262556477476042, "efficiency": 0.5, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "size_sensitive_cycle" }, { "seed": 900000, "sample": 3, "native_reward": 0.8262556477476042, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 3091, "reward_channels": { "native": 0.8262556477476042, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "size_sensitive_cycle" }, { "seed": 900001, "sample": 0, "native_reward": 0.9513407572618797, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 6, "model_tokens": 6078, "reward_channels": { "native": 0.9513407572618797, "efficiency": 0.0, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "size_sensitive_cycle" }, { "seed": 900001, "sample": 1, "native_reward": 0.8245803237223959, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 3898, "reward_channels": { "native": 0.8245803237223959, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "size_sensitive_cycle" }, { "seed": 900001, "sample": 2, "native_reward": 0.9829802732091317, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 5, "model_tokens": 4624, "reward_channels": { "native": 0.9829802732091317, "efficiency": 0.16666666666666663, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "size_sensitive_cycle" }, { "seed": 900001, "sample": 3, "native_reward": 0.9069421616294167, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 3, "model_tokens": 6983, "reward_channels": { "native": 0.9069421616294167, "efficiency": 0.5, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "size_sensitive_cycle" } ] }, { "run_id": "1789273322", "mode": "thinking", "method": "gdpo", "steps": 50, "reward_points": 25, "train_mean": 0.8190012788772583, "train_last10": 0.8253489136695862, "holdout_native": 0.6601590534180904, "holdout_episodes": 40, "holdout_successes": 35, "mean_step_seconds": 937.89942124912, "label": "GDPO", "state_at_cutoff": "complete", "step_hours": 13.026380850682221, "peak_allocated_gib": 68.30920124053955, "peak_reserved_gib": 76.08203125, "extra_refill_rounds": 0.0, "groups_refilled": 0.0, "holdout_component_means": { "native": 0.6601590534180904, "efficiency": 0.4625, "reliability": 0.85 }, "configuration": { "model": { "id": "Qwen/Qwen3-14B", "revision": "40c069824f4251a91eefaf281ebe4c544efd3e18", "loss": "dapo", "algorithm": "grpo", "no_think": false, "thinking_budget_tokens": 2048, "lora": true, "quantize": true, "lora_r": 32, "lora_alpha": 64, "use_dora": false }, "loss": "dapo", "normalization": "gdpo", "entropy_control": { "method": "monitor", "epsilon_low": 0.2, "epsilon_high": 0.28, "zeta_initial": 0.001, "zeta_min": 0.0001, "zeta_max": 0.05, "holdout_split": "benchmark", "holdout_seeds_per_family": 2, "num_iterations": 2 }, "dynamic_sampling": false, "sampling": { "group_size": 4, "temperature": 1.0, "max_completion_tokens": 22528, "top_entropy_quantile": 1.0, "temperature_decay_to": null }, "learning_rate": 1e-05, "turn_max_tokens": 3072 }, "source_sha256": { "metrics.jsonl": "2c412be9b59e031efa49d7d0802b2c45d2f0a0173d593018df248c13770c9742", "entropy_holdout_episodes.jsonl": "19c1153be62ef5b60bdfa4d9b235db785888e26e652783f95a5ebcd6b1348ed6", "resolved_training_config.json": "77d3d9071a4f6d59362223862c015a4cf5c477ca1247fca7eac9140caa5ae09d" }, "training_steps": [ { "step": 1, "loss": -0.36769813299179077, "grad_norm": 0.0634765625, "learning_rate": 1e-05, "thinking/tokens_mean": 606.2361111111111, "thinking/budget_hit_fraction": 0.14583333333333334, "num_tokens": 51975.0, "completions/mean_length": 3604.375, "completions/min_length": 2890.0, "completions/max_length": 6014.0, "completions/clipped_ratio": 0.25, "completions/mean_terminated_length": 3782.166748046875, "completions/min_terminated_length": 2890.0, "completions/max_terminated_length": 6014.0, "tools/call_frequency": 2.25, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.7262155413627625, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 0.13481803238391876, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.4583333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.2500000298023224, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/mean": 0.45231834053993225, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/std": 0.526568591594696, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/mean": 0.2916666865348816, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/std": 0.34359216690063477, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/mean": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/std": 1.154700517654419, "reward": 0.7080169916152954, "reward_std": 0.5734051465988159, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 0.5, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.14539151638746262, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 1307.445846601, "entropy_control/entropy": 0.20009788357278327, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.004, "adapter_sync_time": 0.0, "generation_time": 1115.571375, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1150.240427, "forward_backward_time": 157.103401, "rollout_overhead_time": 34.669052, "peak_allocated_gib": 46.211081981658936, "peak_reserved_gib": 51.38671875 }, { "step": 2, "loss": -0.36766910552978516, "grad_norm": 0.0615234375, "learning_rate": 9.800000000000001e-06, "entropy": 0.14550552144646645, "clip_ratio/low_mean": 0.0006198861810844392, "clip_ratio/low_min": 0.0006198861810844392, "clip_ratio/high_mean": 0.0006267015924095176, "clip_ratio/high_max": 0.0006267015924095176, "clip_ratio/region_mean": 0.0012465877734939568, "step_time": 157.2177628310003, "entropy_control/entropy": 0.20025774703837768, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.008, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 157.113298, "rollout_overhead_time": 0.0, "peak_allocated_gib": 46.689598083496094, "peak_reserved_gib": 51.484375, "reward": null }, { "step": 3, "loss": 0.15657348930835724, "grad_norm": 0.080078125, "learning_rate": 9.600000000000001e-06, "thinking/tokens_mean": 770.8214285714286, "thinking/budget_hit_fraction": 0.125, "num_tokens": 104114.0, "completions/mean_length": 3805.375, "completions/min_length": 2013.0, "completions/max_length": 7910.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 3805.375, "completions/min_terminated_length": 2013.0, "completions/max_terminated_length": 7910.0, "tools/call_frequency": 2.75, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.6259031295776367, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 0.18856824934482574, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.5416666865348816, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.2314550280570984, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 0.75, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 0.7071067690849304, "reward": 0.8029865026473999, "reward_std": 0.3228488862514496, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.19134098012000322, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 1627.6266247210006, "entropy_control/entropy": 0.19150208340759745, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.012, "adapter_sync_time": 0.0, "generation_time": 1391.188849, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1431.552137, "forward_backward_time": 195.959858, "rollout_overhead_time": 40.363288, "peak_allocated_gib": 53.68960475921631, "peak_reserved_gib": 61.484375 }, { "step": 4, "loss": 0.15650466084480286, "grad_norm": 0.080078125, "learning_rate": 9.4e-06, "entropy": 0.19041317608207464, "clip_ratio/low_mean": 0.0009439951099921018, "clip_ratio/low_min": 0.0009439951099921018, "clip_ratio/high_mean": 0.00044810882354795467, "clip_ratio/high_max": 0.00044810882354795467, "clip_ratio/region_mean": 0.0013921039335400565, "step_time": 196.10917739700062, "entropy_control/entropy": 0.1905926968047838, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.016, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 195.992883, "rollout_overhead_time": 0.0, "peak_allocated_gib": 53.68960475921631, "peak_reserved_gib": 61.484375, "reward": null }, { "step": 5, "loss": -0.20414340496063232, "grad_norm": 0.09912109375, "learning_rate": 9.200000000000002e-06, "thinking/tokens_mean": 653.15625, "thinking/budget_hit_fraction": 0.125, "num_tokens": 148180.0, "completions/mean_length": 2803.25, "completions/min_length": 1984.0, "completions/max_length": 4650.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 2803.25, "completions/min_terminated_length": 1984.0, "completions/max_terminated_length": 4650.0, "tools/call_frequency": 2.25, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.6677709817886353, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 0.10812530666589737, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.625, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.07715168595314026, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 0.0, "reward": 0.8990209102630615, "reward_std": 0.10426770895719528, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.17905353382229805, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 943.5877459760004, "entropy_control/entropy": 0.18634943474164326, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.02, "adapter_sync_time": 0.0, "generation_time": 823.17774, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 845.614013, "forward_backward_time": 97.867061, "rollout_overhead_time": 22.436273, "peak_allocated_gib": 53.68960475921631, "peak_reserved_gib": 61.4921875 }, { "step": 6, "loss": -0.20495037734508514, "grad_norm": 0.09814453125, "learning_rate": 9e-06, "entropy": 0.17913627997040749, "clip_ratio/low_mean": 0.0017803673981688917, "clip_ratio/low_min": 0.0017803673981688917, "clip_ratio/high_mean": 0.00029758658638456836, "clip_ratio/high_max": 0.00029758658638456836, "clip_ratio/region_mean": 0.00207795398455346, "step_time": 97.99755129099958, "entropy_control/entropy": 0.186299497104382, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.024, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 97.886359, "rollout_overhead_time": 0.0, "peak_allocated_gib": 53.68960475921631, "peak_reserved_gib": 61.4921875, "reward": null }, { "step": 7, "loss": -0.13794663548469543, "grad_norm": 0.054931640625, "learning_rate": 8.8e-06, "thinking/tokens_mean": 649.7137755102041, "thinking/budget_hit_fraction": 0.125, "num_tokens": 214624.0, "completions/mean_length": 4838.0, "completions/min_length": 3071.0, "completions/max_length": 5855.0, "completions/clipped_ratio": 0.125, "completions/mean_terminated_length": 5090.4287109375, "completions/min_terminated_length": 4301.0, "completions/max_terminated_length": 5855.0, "tools/call_frequency": 3.875, "tools/failure_frequency": 0.032258063554763794, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/mean": 0.7492401599884033, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/std": 0.1634497493505478, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/mean": 0.25, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/std": 0.09622505307197571, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/mean": 0.543454647064209, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/std": 0.4938512444496155, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/mean": 0.25, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/std": 0.28867512941360474, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/mean": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/std": 1.154700517654419, "reward": 0.7588474750518799, "reward_std": 0.5387411713600159, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 0.5, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.16913630813360214, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 1794.1383231250002, "entropy_control/entropy": 0.19205359679628978, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.028, "adapter_sync_time": 0.0, "generation_time": 1524.395877, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1571.036354, "forward_backward_time": 222.998039, "rollout_overhead_time": 46.640476, "peak_allocated_gib": 57.67082643508911, "peak_reserved_gib": 64.12890625 }, { "step": 8, "loss": -0.13806083798408508, "grad_norm": 0.056396484375, "learning_rate": 8.6e-06, "entropy": 0.16933406330645084, "clip_ratio/low_mean": 0.0014047705917619169, "clip_ratio/low_min": 0.0014047705917619169, "clip_ratio/high_mean": 0.0003195654026058037, "clip_ratio/high_max": 0.0003195654026058037, "clip_ratio/region_mean": 0.0017243359943677206, "step_time": 223.11134523700093, "entropy_control/entropy": 0.19230884432662818, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.032, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 222.991656, "rollout_overhead_time": 0.0, "peak_allocated_gib": 57.67082643508911, "peak_reserved_gib": 64.12890625, "reward": null }, { "step": 9, "loss": -0.038439735770225525, "grad_norm": 0.06103515625, "learning_rate": 8.400000000000001e-06, "thinking/tokens_mean": 833.625, "thinking/budget_hit_fraction": 0.25, "num_tokens": 279413.0, "completions/mean_length": 4814.625, "completions/min_length": 3024.0, "completions/max_length": 7275.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 4814.625, "completions/min_terminated_length": 3024.0, "completions/max_terminated_length": 7275.0, "tools/call_frequency": 3.125, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.6691635251045227, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 0.11923417448997498, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.625, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.0833333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/mean": 0.6302784085273743, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/std": 0.021266307681798935, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/mean": 0.3333333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/std": 0.19245009124279022, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/std": 0.0, "reward": 0.8736792802810669, "reward_std": 0.08214110136032104, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.20976854860782623, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 2104.989630897003, "entropy_control/entropy": 0.20675822542460923, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.036, "adapter_sync_time": 0.0, "generation_time": 1785.25502, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1839.428141, "forward_backward_time": 265.438223, "rollout_overhead_time": 54.173121, "peak_allocated_gib": 63.65529108047485, "peak_reserved_gib": 70.6328125 }, { "step": 10, "loss": -0.0384441539645195, "grad_norm": 0.0625, "learning_rate": 8.2e-06, "entropy": 0.209997046738863, "clip_ratio/low_mean": 0.0014056539366720244, "clip_ratio/low_min": 0.0014056539366720244, "clip_ratio/high_mean": 0.00027252716972725466, "clip_ratio/high_max": 0.00027252716972725466, "clip_ratio/region_mean": 0.001678181106399279, "step_time": 265.4383719210018, "entropy_control/entropy": 0.206922773582336, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.04, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 265.318509, "rollout_overhead_time": 0.0, "peak_allocated_gib": 63.65529108047485, "peak_reserved_gib": 70.6328125, "reward": null }, { "step": 11, "loss": -0.1930418461561203, "grad_norm": 0.0625, "learning_rate": 8.000000000000001e-06, "thinking/tokens_mean": 538.2755102040817, "thinking/budget_hit_fraction": 0.07397959183673468, "num_tokens": 334052.0, "completions/mean_length": 3520.875, "completions/min_length": 1712.0, "completions/max_length": 6269.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 3520.875, "completions/min_terminated_length": 1712.0, "completions/max_terminated_length": 6269.0, "tools/call_frequency": 3.0, "tools/failure_frequency": 0.0416666679084301, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__native/mean": 0.5194672346115112, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__native/std": 0.36497068405151367, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__efficiency/mean": 0.4166666865348816, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__efficiency/std": 0.3191424012184143, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__reliability/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__reliability/std": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/mean": 0.5864843726158142, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/std": 0.23310863971710205, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/mean": 0.375, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/std": 0.28463754057884216, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/std": 1.0, "reward": 0.6727674603462219, "reward_std": 0.47311633825302124, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 0.5, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.17794616241008043, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 1713.9672958210012, "entropy_control/entropy": 0.18543802125086845, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.044, "adapter_sync_time": 0.0, "generation_time": 1493.682371, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1531.594649, "forward_backward_time": 182.250001, "rollout_overhead_time": 37.912278, "peak_allocated_gib": 63.65529108047485, "peak_reserved_gib": 70.6328125 }, { "step": 12, "loss": -0.19339165091514587, "grad_norm": 0.06103515625, "learning_rate": 7.800000000000002e-06, "entropy": 0.17742812866345048, "clip_ratio/low_mean": 0.00016134236648213118, "clip_ratio/low_min": 0.00016134236648213118, "clip_ratio/high_mean": 0.0008200673182727769, "clip_ratio/high_max": 0.0008200673182727769, "clip_ratio/region_mean": 0.000981409684754908, "step_time": 182.38993302600284, "entropy_control/entropy": 0.18502110753660309, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.048, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 182.271258, "rollout_overhead_time": 0.0, "peak_allocated_gib": 63.65529108047485, "peak_reserved_gib": 70.6328125, "reward": null }, { "step": 13, "loss": -0.3882444202899933, "grad_norm": 0.0546875, "learning_rate": 7.600000000000001e-06, "thinking/tokens_mean": 654.7202380952381, "thinking/budget_hit_fraction": 0.14583333333333334, "num_tokens": 390022.0, "completions/mean_length": 3711.25, "completions/min_length": 2549.0, "completions/max_length": 5202.0, "completions/clipped_ratio": 0.125, "completions/mean_terminated_length": 3802.71435546875, "completions/min_terminated_length": 2549.0, "completions/max_terminated_length": 5202.0, "tools/call_frequency": 2.375, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.6672567129135132, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 0.11540427803993225, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.5416666865348816, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.25, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/mean": 0.48572278022766113, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/std": 0.3274116814136505, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/mean": 0.4166666865348816, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/std": 0.2886751592159271, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/std": 1.0, "reward": 0.7504480481147766, "reward_std": 0.39120444655418396, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 0.5, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.1802157610654831, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 1381.0262770810004, "entropy_control/entropy": 0.20467618746349558, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.052, "adapter_sync_time": 0.0, "generation_time": 1174.97906, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1210.476384, "forward_backward_time": 170.431939, "rollout_overhead_time": 35.497324, "peak_allocated_gib": 63.65529108047485, "peak_reserved_gib": 70.6328125 }, { "step": 14, "loss": -0.3882194757461548, "grad_norm": 0.0546875, "learning_rate": 7.4e-06, "entropy": 0.18055002391338348, "clip_ratio/low_mean": 0.000992554210824892, "clip_ratio/low_min": 0.000992554210824892, "clip_ratio/high_mean": 0.000505655276356265, "clip_ratio/high_max": 0.000505655276356265, "clip_ratio/region_mean": 0.0014982094871811569, "step_time": 170.54123297400292, "entropy_control/entropy": 0.20512804829279208, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.056, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 170.422358, "rollout_overhead_time": 0.0, "peak_allocated_gib": 63.65529108047485, "peak_reserved_gib": 70.6328125, "reward": null }, { "step": 15, "loss": -0.09555818140506744, "grad_norm": 0.054443359375, "learning_rate": 7.2000000000000005e-06, "thinking/tokens_mean": 858.0555555555555, "thinking/budget_hit_fraction": 0.22916666666666666, "num_tokens": 462068.0, "completions/mean_length": 5522.25, "completions/min_length": 2731.0, "completions/max_length": 8343.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 5522.25, "completions/min_terminated_length": 2731.0, "completions/max_terminated_length": 8343.0, "tools/call_frequency": 3.75, "tools/failure_frequency": 0.03333333507180214, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/mean": 0.730453610420227, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/std": 0.1056305542588234, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/mean": 0.3333333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/std": 0.13608276844024658, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/mean": 0.8973628878593445, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/std": 0.08619581162929535, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/mean": 0.4583333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/std": 0.2500000298023224, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/std": 0.0, "reward": 1.0336999893188477, "reward_std": 0.12287236750125885, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.17676966823637486, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 2225.683277173999, "entropy_control/entropy": 0.17954016063507616, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.06, "adapter_sync_time": 0.0, "generation_time": 1939.05255, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1988.181629, "forward_backward_time": 237.386505, "rollout_overhead_time": 49.129078, "peak_allocated_gib": 63.65529108047485, "peak_reserved_gib": 70.6328125 }, { "step": 16, "loss": -0.09546507149934769, "grad_norm": 0.053466796875, "learning_rate": 7e-06, "entropy": 0.17667382024228573, "clip_ratio/low_mean": 0.0014553236178471707, "clip_ratio/low_min": 0.0014553236178471707, "clip_ratio/high_mean": 0.0003516139186103828, "clip_ratio/high_max": 0.0003516139186103828, "clip_ratio/region_mean": 0.0018069375364575535, "step_time": 237.47537800700047, "entropy_control/entropy": 0.1795387570793141, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.064, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 237.355971, "rollout_overhead_time": 0.0, "peak_allocated_gib": 63.65529108047485, "peak_reserved_gib": 70.6328125, "reward": null }, { "step": 17, "loss": -0.33884182572364807, "grad_norm": 0.05419921875, "learning_rate": 6.800000000000001e-06, "thinking/tokens_mean": 480.4959183673469, "thinking/budget_hit_fraction": 0.08928571428571429, "num_tokens": 520469.0, "completions/mean_length": 3609.125, "completions/min_length": 1630.0, "completions/max_length": 6829.0, "completions/clipped_ratio": 0.125, "completions/mean_terminated_length": 3686.000244140625, "completions/min_terminated_length": 1630.0, "completions/max_terminated_length": 6829.0, "tools/call_frequency": 3.25, "tools/failure_frequency": 0.03846153989434242, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/mean": 0.5374807715415955, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/std": 0.35832566022872925, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/mean": 0.125, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/std": 0.15957120060920715, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/std": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/mean": 0.7045713663101196, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/std": 0.1158401221036911, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/mean": 0.5833333730697632, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/std": 0.09622505307197571, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/std": 0.0, "reward": 0.7887344360351562, "reward_std": 0.4061516523361206, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 0.5, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.1640746984630823, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 1791.069795530997, "entropy_control/entropy": 0.1958823486718863, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.068, "adapter_sync_time": 0.0, "generation_time": 1497.803051, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1548.337111, "forward_backward_time": 242.613345, "rollout_overhead_time": 50.53406, "peak_allocated_gib": 63.65529108047485, "peak_reserved_gib": 70.6328125 }, { "step": 18, "loss": -0.33923161029815674, "grad_norm": 0.053466796875, "learning_rate": 6.600000000000001e-06, "entropy": 0.16425933130085468, "clip_ratio/low_mean": 0.0006798447866458446, "clip_ratio/low_min": 0.0006798447866458446, "clip_ratio/high_mean": 0.0010638363746693358, "clip_ratio/high_max": 0.0010638363746693358, "clip_ratio/region_mean": 0.0017436811613151804, "step_time": 242.6830074339996, "entropy_control/entropy": 0.19605819728218557, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.072, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 242.570539, "rollout_overhead_time": 0.0, "peak_allocated_gib": 63.65529108047485, "peak_reserved_gib": 70.6328125, "reward": null }, { "step": 19, "loss": 0.012447401881217957, "grad_norm": 0.08935546875, "learning_rate": 6.4000000000000006e-06, "thinking/tokens_mean": 674.6964285714286, "thinking/budget_hit_fraction": 0.10714285714285714, "num_tokens": 575269.0, "completions/mean_length": 3959.5, "completions/min_length": 2473.0, "completions/max_length": 6220.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 3959.5, "completions/min_terminated_length": 2473.0, "completions/max_terminated_length": 6220.0, "tools/call_frequency": 3.0, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.607343316078186, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 0.004687517881393433, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.3333333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/mean": 0.8638374209403992, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/std": 0.07588943839073181, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/std": 0.3333333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/std": 0.0, "reward": 0.9605903625488281, "reward_std": 0.14316250383853912, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.19665975868701935, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 1663.0609616120018, "entropy_control/entropy": 0.1936969557636714, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.076, "adapter_sync_time": 0.0, "generation_time": 1392.212287, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1438.877575, "forward_backward_time": 224.066708, "rollout_overhead_time": 46.665287, "peak_allocated_gib": 63.65529108047485, "peak_reserved_gib": 70.6328125 }, { "step": 20, "loss": 0.012685425579547882, "grad_norm": 0.08349609375, "learning_rate": 6.200000000000001e-06, "entropy": 0.19716499000787735, "clip_ratio/low_mean": 0.0015608484100084752, "clip_ratio/low_min": 0.0015608484100084752, "clip_ratio/high_mean": 0.0006244356918614358, "clip_ratio/high_max": 0.0006244356918614358, "clip_ratio/region_mean": 0.002185284101869911, "step_time": 224.124938886991, "entropy_control/entropy": 0.19413256457173864, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.08, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 224.010219, "rollout_overhead_time": 0.0, "peak_allocated_gib": 63.65529108047485, "peak_reserved_gib": 70.6328125, "reward": null }, { "step": 21, "loss": -0.25645989179611206, "grad_norm": 0.058837890625, "learning_rate": 6e-06, "thinking/tokens_mean": 739.2238095238096, "thinking/budget_hit_fraction": 0.2571428571428571, "num_tokens": 636136.0, "completions/mean_length": 3909.375, "completions/min_length": 2254.0, "completions/max_length": 5996.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 3909.375, "completions/min_terminated_length": 2254.0, "completions/max_terminated_length": 5996.0, "tools/call_frequency": 2.375, "tools/failure_frequency": 0.10526315867900848, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/mean": 0.4605706036090851, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/std": 0.3070470690727234, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/mean": 0.4583333730697632, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/std": 0.3154948949813843, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/std": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/mean": 0.700609564781189, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/std": 0.1823853850364685, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/mean": 0.5833333730697632, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/std": 0.1666666716337204, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/std": 0.0, "reward": 0.7566317319869995, "reward_std": 0.40518611669540405, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 0.5, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.2014899179339409, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 1358.554902425003, "entropy_control/entropy": 0.20816090601034415, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.084, "adapter_sync_time": 0.0, "generation_time": 1133.133636, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1171.653807, "forward_backward_time": 186.799026, "rollout_overhead_time": 38.520171, "peak_allocated_gib": 63.65529108047485, "peak_reserved_gib": 70.6328125 }, { "step": 22, "loss": -0.25666821002960205, "grad_norm": 0.06591796875, "learning_rate": 5.8e-06, "entropy": 0.2016368806362152, "clip_ratio/low_mean": 0.0011461442772997543, "clip_ratio/low_min": 0.0011461442772997543, "clip_ratio/high_mean": 0.0003696301246236544, "clip_ratio/high_max": 0.0003696301246236544, "clip_ratio/region_mean": 0.0015157744019234087, "step_time": 186.87984689799487, "entropy_control/entropy": 0.20827655213029939, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.088, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 186.76527, "rollout_overhead_time": 0.0, "peak_allocated_gib": 63.65529108047485, "peak_reserved_gib": 70.6328125, "reward": null }, { "step": 23, "loss": 0.2094455063343048, "grad_norm": 0.072265625, "learning_rate": 5.600000000000001e-06, "thinking/tokens_mean": 690.1527777777778, "thinking/budget_hit_fraction": 0.14583333333333334, "num_tokens": 692023.0, "completions/mean_length": 4100.375, "completions/min_length": 2039.0, "completions/max_length": 6449.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 4100.375, "completions/min_terminated_length": 2039.0, "completions/max_terminated_length": 6449.0, "tools/call_frequency": 2.875, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.6095624566078186, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.6666666865348816, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/mean": 0.8230918645858765, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/std": 0.0027063225861638784, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/mean": 0.375, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/std": 0.25, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/std": 0.0, "reward": 0.9423688650131226, "reward_std": 0.10680093616247177, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.5, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.1906361961737275, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 1482.741087434002, "entropy_control/entropy": 0.19142697965461175, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.092, "adapter_sync_time": 0.0, "generation_time": 1277.44327, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1313.633773, "forward_backward_time": 168.986496, "rollout_overhead_time": 36.190503, "peak_allocated_gib": 63.65529108047485, "peak_reserved_gib": 70.6328125 }, { "step": 24, "loss": 0.20957893133163452, "grad_norm": 0.07275390625, "learning_rate": 5.400000000000001e-06, "entropy": 0.19064043555408716, "clip_ratio/low_mean": 0.0005921031697653234, "clip_ratio/low_min": 0.0005921031697653234, "clip_ratio/high_mean": 0.0009996965964091942, "clip_ratio/high_max": 0.0009996965964091942, "clip_ratio/region_mean": 0.0015917997661745176, "step_time": 169.07731957399665, "entropy_control/entropy": 0.19152861339121893, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.096, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 168.956209, "rollout_overhead_time": 0.0, "peak_allocated_gib": 63.65529108047485, "peak_reserved_gib": 70.6328125, "reward": null }, { "step": 25, "loss": -0.14349018037319183, "grad_norm": 0.08203125, "learning_rate": 5.2e-06, "thinking/tokens_mean": 712.575, "thinking/budget_hit_fraction": 0.125, "num_tokens": 745634.0, "completions/mean_length": 3611.375, "completions/min_length": 2148.0, "completions/max_length": 4889.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 3611.375, "completions/min_terminated_length": 2148.0, "completions/max_terminated_length": 4889.0, "tools/call_frequency": 2.5, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.6715044975280762, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 0.12447243928909302, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.5833333730697632, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.1666666716337204, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/mean": 0.6671483516693115, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/std": 0.11642485857009888, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/mean": 0.5833333730697632, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/std": 0.1666666716337204, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/std": 0.0, "reward": 0.8984931111335754, "reward_std": 0.1038915365934372, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.19667992927134037, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 1322.410297416005, "entropy_control/entropy": 0.20309221561600888, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.1, "adapter_sync_time": 0.0, "generation_time": 1141.403989, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1173.571129, "forward_backward_time": 148.727847, "rollout_overhead_time": 32.167139, "peak_allocated_gib": 63.65529108047485, "peak_reserved_gib": 70.6328125 }, { "step": 26, "loss": -0.14316785335540771, "grad_norm": 0.078125, "learning_rate": 5e-06, "entropy": 0.1969421859830618, "clip_ratio/low_mean": 0.0017875201592687517, "clip_ratio/low_min": 0.0017875201592687517, "clip_ratio/high_mean": 0.0003260288794990629, "clip_ratio/high_max": 0.0003260288794990629, "clip_ratio/region_mean": 0.0021135490387678146, "step_time": 148.8309440800076, "entropy_control/entropy": 0.20345788274826315, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.104, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 148.727129, "rollout_overhead_time": 0.0, "peak_allocated_gib": 63.65529108047485, "peak_reserved_gib": 70.6328125, "reward": null }, { "step": 27, "loss": -0.06274816393852234, "grad_norm": 0.0693359375, "learning_rate": 4.800000000000001e-06, "thinking/tokens_mean": 572.0399659863946, "thinking/budget_hit_fraction": 0.125, "num_tokens": 804774.0, "completions/mean_length": 4090.0, "completions/min_length": 2118.0, "completions/max_length": 5898.0, "completions/clipped_ratio": 0.125, "completions/mean_terminated_length": 4235.57177734375, "completions/min_terminated_length": 2118.0, "completions/max_terminated_length": 5898.0, "tools/call_frequency": 3.375, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/mean": 0.5661505460739136, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/std": 0.22813044488430023, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/mean": 0.3333333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/std": 0.38490021228790283, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/std": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/mean": 0.5091716647148132, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/std": 0.45859766006469727, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/mean": 0.2916666865348816, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/std": 0.34359216690063477, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/mean": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/std": 1.154700517654419, "reward": 0.60328608751297, "reward_std": 0.544246256351471, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 0.5, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.14777466095983982, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 1568.3047927020052, "entropy_control/entropy": 0.16951632124726934, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.108, "adapter_sync_time": 0.0, "generation_time": 1265.271209, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1316.730336, "forward_backward_time": 251.466472, "rollout_overhead_time": 51.459127, "peak_allocated_gib": 63.65529108047485, "peak_reserved_gib": 70.6328125 }, { "step": 28, "loss": -0.06255805492401123, "grad_norm": 0.0673828125, "learning_rate": 4.600000000000001e-06, "entropy": 0.14784383215010166, "clip_ratio/low_mean": 0.0007143398688640445, "clip_ratio/low_min": 0.0007143398688640445, "clip_ratio/high_mean": 0.0007623398123541847, "clip_ratio/high_max": 0.0007623398123541847, "clip_ratio/region_mean": 0.0014766796812182292, "step_time": 251.54640228600329, "entropy_control/entropy": 0.16952065923173681, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.112, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 251.439412, "rollout_overhead_time": 0.0, "peak_allocated_gib": 63.65529108047485, "peak_reserved_gib": 70.6328125, "reward": null }, { "step": 29, "loss": 0.05750003829598427, "grad_norm": 0.072265625, "learning_rate": 4.4e-06, "thinking/tokens_mean": 673.2035714285714, "thinking/budget_hit_fraction": 0.17857142857142858, "num_tokens": 869160.0, "completions/mean_length": 4765.75, "completions/min_length": 1572.0, "completions/max_length": 7415.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 4765.75, "completions/min_terminated_length": 1572.0, "completions/max_terminated_length": 7415.0, "tools/call_frequency": 3.5, "tools/failure_frequency": 0.1071428582072258, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.6086075305938721, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.6666666865348816, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/mean": 0.5874189138412476, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/std": 0.25575050711631775, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/mean": 0.2916666865348816, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/std": 0.20971763134002686, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/std": 1.0, "reward": 0.771971583366394, "reward_std": 0.3044925630092621, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.5, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.19841952435672283, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 1826.776519818999, "entropy_control/entropy": 0.20271056372889976, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.116, "adapter_sync_time": 0.0, "generation_time": 1508.021128, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1561.827371, "forward_backward_time": 264.8428, "rollout_overhead_time": 53.806243, "peak_allocated_gib": 63.65529108047485, "peak_reserved_gib": 70.80859375 }, { "step": 30, "loss": 0.05753684788942337, "grad_norm": 0.068359375, "learning_rate": 4.2000000000000004e-06, "entropy": 0.19805839471518993, "clip_ratio/low_mean": 0.00011800404899986461, "clip_ratio/low_min": 0.00011800404899986461, "clip_ratio/high_mean": 0.000237524276599288, "clip_ratio/high_max": 0.000237524276599288, "clip_ratio/region_mean": 0.0003555283255991526, "step_time": 264.89417385499473, "entropy_control/entropy": 0.20232579636542353, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.12, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 264.791987, "rollout_overhead_time": 0.0, "peak_allocated_gib": 63.65529108047485, "peak_reserved_gib": 70.80859375, "reward": null }, { "step": 31, "loss": 0.08018381893634796, "grad_norm": 0.08837890625, "learning_rate": 4.000000000000001e-06, "thinking/tokens_mean": 719.7777777777778, "thinking/budget_hit_fraction": 0.16666666666666666, "num_tokens": 930599.0, "completions/mean_length": 4596.375, "completions/min_length": 3064.0, "completions/max_length": 6267.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 4596.375, "completions/min_terminated_length": 3064.0, "completions/max_terminated_length": 6267.0, "tools/call_frequency": 3.125, "tools/failure_frequency": 0.03999999910593033, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/mean": 0.7773112058639526, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/std": 0.2507764399051666, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/std": 0.15430335700511932, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/std": 0.0, "reward": 1.0023112297058105, "reward_std": 0.2496374100446701, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.19933259673416615, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 1455.752558099004, "entropy_control/entropy": 0.2018930165597738, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.124, "adapter_sync_time": 0.0, "generation_time": 1230.853753, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1270.152366, "forward_backward_time": 185.492087, "rollout_overhead_time": 39.298613, "peak_allocated_gib": 63.65529108047485, "peak_reserved_gib": 70.80859375 }, { "step": 32, "loss": 0.08029289543628693, "grad_norm": 0.07421875, "learning_rate": 3.8000000000000005e-06, "entropy": 0.1996234729886055, "clip_ratio/low_mean": 0.0007656209054403007, "clip_ratio/low_min": 0.0007656209054403007, "clip_ratio/high_mean": 0.0005647374273394234, "clip_ratio/high_max": 0.0005647374273394234, "clip_ratio/region_mean": 0.0013303583327797242, "step_time": 185.55397482400076, "entropy_control/entropy": 0.2021787957564968, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.128, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 185.449541, "rollout_overhead_time": 0.0, "peak_allocated_gib": 63.65529108047485, "peak_reserved_gib": 70.80859375, "reward": null }, { "step": 33, "loss": -0.17429736256599426, "grad_norm": 0.0537109375, "learning_rate": 3.6000000000000003e-06, "thinking/tokens_mean": 635.0119047619047, "thinking/budget_hit_fraction": 0.14583333333333334, "num_tokens": 984979.0, "completions/mean_length": 3902.0, "completions/min_length": 2860.0, "completions/max_length": 5233.0, "completions/clipped_ratio": 0.125, "completions/mean_terminated_length": 4020.71435546875, "completions/min_terminated_length": 2860.0, "completions/max_terminated_length": 5233.0, "tools/call_frequency": 2.5, "tools/failure_frequency": 0.05000000074505806, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.5030142068862915, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 0.3381158411502838, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.4166666865348816, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.2886751592159271, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/mean": 0.8578818440437317, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/std": 0.06588447093963623, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/mean": 0.5416666865348816, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/std": 0.25, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/std": 0.0, "reward": 0.8544063568115234, "reward_std": 0.43808114528656006, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 0.5, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.16345028206706047, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 1260.1165995250049, "entropy_control/entropy": 0.1901272417645771, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.132, "adapter_sync_time": 0.0, "generation_time": 1068.786813, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1102.486461, "forward_backward_time": 157.532617, "rollout_overhead_time": 33.699647, "peak_allocated_gib": 63.65529108047485, "peak_reserved_gib": 70.80859375 }, { "step": 34, "loss": -0.17431220412254333, "grad_norm": 0.052490234375, "learning_rate": 3.4000000000000005e-06, "entropy": 0.16317202150821686, "clip_ratio/low_mean": 0.0009362715936731547, "clip_ratio/low_min": 0.0009362715936731547, "clip_ratio/high_mean": 0.0008944949295255356, "clip_ratio/high_max": 0.0008944949295255356, "clip_ratio/region_mean": 0.0018307665231986903, "step_time": 157.62631366799906, "entropy_control/entropy": 0.1897910693214771, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.136, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 157.523362, "rollout_overhead_time": 0.0, "peak_allocated_gib": 63.65529108047485, "peak_reserved_gib": 70.80859375, "reward": null }, { "step": 35, "loss": -0.1423398107290268, "grad_norm": 0.062255859375, "learning_rate": 3.2000000000000003e-06, "thinking/tokens_mean": 662.85, "thinking/budget_hit_fraction": 0.16071428571428573, "num_tokens": 1051659.0, "completions/mean_length": 4472.0, "completions/min_length": 2264.0, "completions/max_length": 7392.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 4472.0, "completions/min_terminated_length": 2264.0, "completions/max_terminated_length": 7392.0, "tools/call_frequency": 2.875, "tools/failure_frequency": 0.043478261679410934, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/mean": 0.5661094188690186, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/std": 0.23527806997299194, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/mean": 0.4375000298023224, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/std": 0.2808363437652588, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/mean": 0.75, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/std": 0.7071067690849304, "reward": 0.7379844188690186, "reward_std": 0.38319867849349976, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 0.5, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.21335223503410816, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 1710.421675318994, "entropy_control/entropy": 0.21247092390897898, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.14, "adapter_sync_time": 0.0, "generation_time": 1349.928758, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1410.28788, "forward_backward_time": 300.030776, "rollout_overhead_time": 60.359122, "peak_allocated_gib": 68.30920124053955, "peak_reserved_gib": 76.08203125 }, { "step": 36, "loss": -0.14226660132408142, "grad_norm": 0.0576171875, "learning_rate": 3e-06, "entropy": 0.21380090340971947, "clip_ratio/low_mean": 0.0013880024635000154, "clip_ratio/low_min": 0.0013880024635000154, "clip_ratio/high_mean": 0.0004817337612621486, "clip_ratio/high_max": 0.0004817337612621486, "clip_ratio/region_mean": 0.001869736224762164, "step_time": 300.1433946209945, "entropy_control/entropy": 0.21280985767545538, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.144, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 300.032993, "rollout_overhead_time": 0.0, "peak_allocated_gib": 68.30920124053955, "peak_reserved_gib": 76.08203125, "reward": null }, { "step": 37, "loss": 0.06676673144102097, "grad_norm": 0.064453125, "learning_rate": 2.8000000000000003e-06, "thinking/tokens_mean": 720.0892857142857, "thinking/budget_hit_fraction": 0.16071428571428573, "num_tokens": 1116140.0, "completions/mean_length": 5180.125, "completions/min_length": 2771.0, "completions/max_length": 7024.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 5180.125, "completions/min_terminated_length": 2771.0, "completions/max_terminated_length": 7024.0, "tools/call_frequency": 4.125, "tools/failure_frequency": 0.06060606241226196, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.6381549835205078, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 0.3154814541339874, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.2916666567325592, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.25, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__native/mean": 0.5676366090774536, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__native/std": 0.22932963073253632, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__efficiency/mean": 0.4166666865348816, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__efficiency/std": 0.3191424012184143, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__reliability/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__reliability/std": 1.0, "reward": 0.770604133605957, "reward_std": 0.3584084212779999, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.20501251704990864, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 2042.283405759008, "entropy_control/entropy": 0.2084450076313529, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.148, "adapter_sync_time": 0.0, "generation_time": 1791.217466, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1834.225008, "forward_backward_time": 207.952013, "rollout_overhead_time": 43.007542, "peak_allocated_gib": 68.30920124053955, "peak_reserved_gib": 76.08203125 }, { "step": 38, "loss": 0.06687550246715546, "grad_norm": 0.064453125, "learning_rate": 2.6e-06, "entropy": 0.2049033436924219, "clip_ratio/low_mean": 0.0004826112708542496, "clip_ratio/low_min": 0.0004826112708542496, "clip_ratio/high_mean": 0.0006639463208557572, "clip_ratio/high_max": 0.0006639463208557572, "clip_ratio/region_mean": 0.0011465575917100068, "step_time": 208.07504388699454, "entropy_control/entropy": 0.20831680852214052, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.152, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 207.96491, "rollout_overhead_time": 0.0, "peak_allocated_gib": 68.30920124053955, "peak_reserved_gib": 76.08203125, "reward": null }, { "step": 39, "loss": -0.1414264291524887, "grad_norm": 0.0732421875, "learning_rate": 2.4000000000000003e-06, "thinking/tokens_mean": 614.2261904761905, "thinking/budget_hit_fraction": 0.125, "num_tokens": 1173866.0, "completions/mean_length": 4124.75, "completions/min_length": 2546.0, "completions/max_length": 6556.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 4124.75, "completions/min_terminated_length": 2546.0, "completions/max_terminated_length": 6556.0, "tools/call_frequency": 3.25, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__native/mean": 0.7013591527938843, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__native/std": 0.24371957778930664, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__efficiency/mean": 0.4583333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__efficiency/std": 0.2136233150959015, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__reliability/std": 0.0, "reward": 0.924275815486908, "reward_std": 0.238555446267128, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.19241669587790966, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 1623.65240167499, "entropy_control/entropy": 0.19407293279736726, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.156, "adapter_sync_time": 0.0, "generation_time": 1320.964607, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1371.632097, "forward_backward_time": 251.920862, "rollout_overhead_time": 50.66749, "peak_allocated_gib": 68.30920124053955, "peak_reserved_gib": 76.08203125 }, { "step": 40, "loss": -0.14135050773620605, "grad_norm": 0.07421875, "learning_rate": 2.2e-06, "entropy": 0.19248690083622932, "clip_ratio/low_mean": 0.0005736229577451013, "clip_ratio/low_min": 0.0005736229577451013, "clip_ratio/high_mean": 0.0008601539011579007, "clip_ratio/high_max": 0.0008601539011579007, "clip_ratio/region_mean": 0.001433776858903002, "step_time": 252.00426232099562, "entropy_control/entropy": 0.194069118367981, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.16, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 251.903949, "rollout_overhead_time": 0.0, "peak_allocated_gib": 68.30920124053955, "peak_reserved_gib": 76.08203125, "reward": null }, { "step": 41, "loss": -0.4750300943851471, "grad_norm": 0.037841796875, "learning_rate": 2.0000000000000003e-06, "thinking/tokens_mean": 909.4222222222221, "thinking/budget_hit_fraction": 0.22777777777777777, "num_tokens": 1243351.0, "completions/mean_length": 5208.625, "completions/min_length": 3071.0, "completions/max_length": 8981.0, "completions/clipped_ratio": 0.25, "completions/mean_terminated_length": 5921.1669921875, "completions/min_terminated_length": 4462.0, "completions/max_terminated_length": 8981.0, "tools/call_frequency": 3.0, "tools/failure_frequency": 0.0416666679084301, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__native/mean": 0.6518133282661438, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__native/std": 0.4416067898273468, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__efficiency/mean": 0.2916666865348816, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__efficiency/std": 0.20971763134002686, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__reliability/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__reliability/std": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/mean": 0.4880439043045044, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/std": 0.32949313521385193, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/mean": 0.25, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/std": 0.3191424012184143, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/std": 1.0, "reward": 0.6834702491760254, "reward_std": 0.5577273964881897, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 0.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.17076805979013443, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 2157.995117664992, "entropy_control/entropy": 0.2206734436527375, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.164, "adapter_sync_time": 0.0, "generation_time": 1819.711307, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1876.314375, "forward_backward_time": 281.556777, "rollout_overhead_time": 56.603067, "peak_allocated_gib": 68.30920124053955, "peak_reserved_gib": 76.08203125 }, { "step": 42, "loss": -0.4752137362957001, "grad_norm": 0.031494140625, "learning_rate": 1.8000000000000001e-06, "entropy": 0.17083237878978252, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0007528942805947736, "clip_ratio/high_max": 0.0007528942805947736, "clip_ratio/region_mean": 0.0007528942805947736, "step_time": 281.58093438699143, "entropy_control/entropy": 0.22086477739931318, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.168, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 281.458833, "rollout_overhead_time": 0.0, "peak_allocated_gib": 68.30920124053955, "peak_reserved_gib": 76.08203125, "reward": null }, { "step": 43, "loss": 0.03924206644296646, "grad_norm": 0.0810546875, "learning_rate": 1.6000000000000001e-06, "thinking/tokens_mean": 493.9948979591837, "thinking/budget_hit_fraction": 0.07142857142857142, "num_tokens": 1296289.0, "completions/mean_length": 3313.25, "completions/min_length": 2179.0, "completions/max_length": 5413.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 3313.25, "completions/min_terminated_length": 2179.0, "completions/max_terminated_length": 5413.0, "tools/call_frequency": 2.875, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__native/mean": 0.5816929340362549, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__native/std": 0.24924789369106293, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__efficiency/mean": 0.4583333730697632, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__efficiency/std": 0.3154948949813843, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__reliability/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__reliability/std": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/mean": 0.5737302899360657, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/std": 0.22210049629211426, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/mean": 0.375, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/std": 0.28463754057884216, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/std": 1.0, "reward": 0.6985449194908142, "reward_std": 0.40811148285865784, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.19458864256739616, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 1488.8086183090054, "entropy_control/entropy": 0.19670942085366805, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.172, "adapter_sync_time": 0.0, "generation_time": 1263.270646, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1301.816107, "forward_backward_time": 186.859284, "rollout_overhead_time": 38.545462, "peak_allocated_gib": 68.30920124053955, "peak_reserved_gib": 76.08203125 }, { "step": 44, "loss": 0.039154961705207825, "grad_norm": 0.078125, "learning_rate": 1.4000000000000001e-06, "entropy": 0.1948654819279909, "clip_ratio/low_mean": 0.0006667591078439727, "clip_ratio/low_min": 0.0006667591078439727, "clip_ratio/high_mean": 0.0008239384260377847, "clip_ratio/high_max": 0.0008239384260377847, "clip_ratio/region_mean": 0.0014906975338817574, "step_time": 186.98488715499843, "entropy_control/entropy": 0.19700822117510752, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.176, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 186.85173, "rollout_overhead_time": 0.0, "peak_allocated_gib": 68.30920124053955, "peak_reserved_gib": 76.08203125, "reward": null }, { "step": 45, "loss": -0.12898582220077515, "grad_norm": 0.0693359375, "learning_rate": 1.2000000000000002e-06, "thinking/tokens_mean": 689.8571428571429, "thinking/budget_hit_fraction": 0.17857142857142858, "num_tokens": 1366177.0, "completions/mean_length": 5066.0, "completions/min_length": 1853.0, "completions/max_length": 7091.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 5066.0, "completions/min_terminated_length": 1853.0, "completions/max_terminated_length": 7091.0, "tools/call_frequency": 3.625, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/mean": 0.561121940612793, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/std": 0.21304446458816528, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/mean": 0.2916666865348816, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/std": 0.20971763134002686, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/std": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/mean": 0.6660890579223633, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/std": 0.11354804039001465, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/std": 0.3333333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/std": 0.0, "reward": 0.7833971381187439, "reward_std": 0.30454862117767334, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.20160364359617233, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 2151.2422908390145, "entropy_control/entropy": 0.20532171599279905, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.18, "adapter_sync_time": 0.0, "generation_time": 1822.015512, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1876.851635, "forward_backward_time": 274.269095, "rollout_overhead_time": 54.836124, "peak_allocated_gib": 68.30920124053955, "peak_reserved_gib": 76.08203125 }, { "step": 46, "loss": -0.12891356647014618, "grad_norm": 0.08544921875, "learning_rate": 1.0000000000000002e-06, "entropy": 0.20192984119057655, "clip_ratio/low_mean": 0.0010274413361912593, "clip_ratio/low_min": 0.0010274413361912593, "clip_ratio/high_mean": 0.00024982879040180705, "clip_ratio/high_max": 0.00024982879040180705, "clip_ratio/region_mean": 0.0012772701265930664, "step_time": 274.3883095810015, "entropy_control/entropy": 0.20559768318617128, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.184, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 274.276248, "rollout_overhead_time": 0.0, "peak_allocated_gib": 68.30920124053955, "peak_reserved_gib": 76.08203125, "reward": null }, { "step": 47, "loss": -0.09007535874843597, "grad_norm": 0.07763671875, "learning_rate": 8.000000000000001e-07, "thinking/tokens_mean": 856.5649999999999, "thinking/budget_hit_fraction": 0.275, "num_tokens": 1430710.0, "completions/mean_length": 4780.125, "completions/min_length": 1957.0, "completions/max_length": 7289.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 4780.125, "completions/min_terminated_length": 1957.0, "completions/max_terminated_length": 7289.0, "tools/call_frequency": 3.125, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/mean": 0.6431146860122681, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/std": 0.0671430230140686, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/std": 0.19245010614395142, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/mean": 0.7154762744903564, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/std": 0.3168836534023285, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/mean": 0.3750000298023224, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/std": 0.28463754057884216, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/std": 1.0, "reward": 0.851170539855957, "reward_std": 0.34725460410118103, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.19823584705591202, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 1800.7287981380068, "entropy_control/entropy": 0.20250650673586154, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.188, "adapter_sync_time": 0.0, "generation_time": 1560.024169, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1601.645599, "forward_backward_time": 198.97692, "rollout_overhead_time": 41.62143, "peak_allocated_gib": 68.30920124053955, "peak_reserved_gib": 76.08203125 }, { "step": 48, "loss": -0.08989251405000687, "grad_norm": 0.06787109375, "learning_rate": 6.000000000000001e-07, "entropy": 0.19837375171482563, "clip_ratio/low_mean": 0.001343853582511656, "clip_ratio/low_min": 0.001343853582511656, "clip_ratio/high_mean": 0.0006475899936049245, "clip_ratio/high_max": 0.0006475899936049245, "clip_ratio/region_mean": 0.0019914435761165805, "step_time": 199.12135581698385, "entropy_control/entropy": 0.2027051213315856, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.192, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 198.999611, "rollout_overhead_time": 0.0, "peak_allocated_gib": 68.30920124053955, "peak_reserved_gib": 76.08203125, "reward": null }, { "step": 49, "loss": 0.04273369163274765, "grad_norm": 0.072265625, "learning_rate": 4.0000000000000003e-07, "thinking/tokens_mean": 715.2916666666666, "thinking/budget_hit_fraction": 0.14583333333333334, "num_tokens": 1493808.0, "completions/mean_length": 4596.25, "completions/min_length": 2868.0, "completions/max_length": 6596.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 4596.25, "completions/min_terminated_length": 2868.0, "completions/max_terminated_length": 6596.0, "tools/call_frequency": 3.375, "tools/failure_frequency": 0.07407407462596893, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/mean": 0.7445147633552551, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/std": 0.09768664091825485, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/mean": 0.4583333730697632, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/std": 0.15957120060920715, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/mean": 0.7022172212600708, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/std": 0.35914701223373413, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/std": 0.235702246427536, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/std": 0.0, "reward": 0.9473243355751038, "reward_std": 0.24201710522174835, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.20186464861035347, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 1814.5493688019924, "entropy_control/entropy": 0.20290295311124665, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.196, "adapter_sync_time": 0.0, "generation_time": 1556.482902, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1600.315104, "forward_backward_time": 214.110287, "rollout_overhead_time": 43.832202, "peak_allocated_gib": 68.30920124053955, "peak_reserved_gib": 76.08203125 }, { "step": 50, "loss": 0.04274625703692436, "grad_norm": 0.0693359375, "learning_rate": 2.0000000000000002e-07, "entropy": 0.20205847918987274, "clip_ratio/low_mean": 0.000517784254043363, "clip_ratio/low_min": 0.000517784254043363, "clip_ratio/high_mean": 0.0008554057894798461, "clip_ratio/high_max": 0.0008554057894798461, "clip_ratio/region_mean": 0.0013731900435232092, "step_time": 214.24098803201196, "entropy_control/entropy": 0.20315267632682507, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.2, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 214.113101, "rollout_overhead_time": 0.0, "peak_allocated_gib": 68.30920124053955, "peak_reserved_gib": 76.08203125, "reward": null } ], "holdout": [ { "seed": 900000, "sample": 0, "native_reward": 0.8638203836248219, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 3, "model_tokens": 4058, "reward_channels": { "native": 0.8638203836248219, "efficiency": 0.5, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "two_pool_cycle" }, { "seed": 900000, "sample": 1, "native_reward": 0.9675266661987673, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 5, "model_tokens": 8233, "reward_channels": { "native": 0.9675266661987673, "efficiency": 0.16666666666666663, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "two_pool_cycle" }, { "seed": 900000, "sample": 2, "native_reward": 0.8638203836248219, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 3, "model_tokens": 3996, "reward_channels": { "native": 0.8638203836248219, "efficiency": 0.5, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "two_pool_cycle" }, { "seed": 900000, "sample": 3, "native_reward": 0.8404385754973732, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 3, "model_tokens": 5137, "reward_channels": { "native": 0.8404385754973732, "efficiency": 0.5, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "two_pool_cycle" }, { "seed": 900001, "sample": 0, "native_reward": 0.6093825392865522, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 3603, "reward_channels": { "native": 0.6093825392865522, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "two_pool_cycle" }, { "seed": 900001, "sample": 1, "native_reward": 0.6093825392865522, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 4155, "reward_channels": { "native": 0.6093825392865522, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "two_pool_cycle" }, { "seed": 900001, "sample": 2, "native_reward": 0.6093825392865522, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 3214, "reward_channels": { "native": 0.6093825392865522, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "two_pool_cycle" }, { "seed": 900001, "sample": 3, "native_reward": 0.175, "promotion_success": false, "done": true, "errors": 0, "tool_steps": 3, "model_tokens": 5817, "reward_channels": { "native": 0.175, "efficiency": 0.5, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [ "positive_net_profit", "stress_positive", "task_complete" ], "family": "two_pool_cycle" }, { "seed": 900000, "sample": 0, "native_reward": 0.6106601460847819, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 2626, "reward_channels": { "native": 0.6106601460847819, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "triangular_cycle" }, { "seed": 900000, "sample": 1, "native_reward": 0.6106601460847819, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 2805, "reward_channels": { "native": 0.6106601460847819, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "triangular_cycle" }, { "seed": 900000, "sample": 2, "native_reward": 0.6106601460847819, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 4228, "reward_channels": { "native": 0.6106601460847819, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "triangular_cycle" }, { "seed": 900000, "sample": 3, "native_reward": 0.6106601460847819, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 2997, "reward_channels": { "native": 0.6106601460847819, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "triangular_cycle" }, { "seed": 900001, "sample": 0, "native_reward": 0.6108851248150757, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 2936, "reward_channels": { "native": 0.6108851248150757, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "triangular_cycle" }, { "seed": 900001, "sample": 1, "native_reward": 0.8062415512788537, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 5, "model_tokens": 5012, "reward_channels": { "native": 0.8062415512788537, "efficiency": 0.16666666666666663, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "triangular_cycle" }, { "seed": 900001, "sample": 2, "native_reward": 0.7060268919310104, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 3, "model_tokens": 2620, "reward_channels": { "native": 0.7060268919310104, "efficiency": 0.5, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "triangular_cycle" }, { "seed": 900001, "sample": 3, "native_reward": 0.811786874720879, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 3, "model_tokens": 4790, "reward_channels": { "native": 0.811786874720879, "efficiency": 0.5, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "triangular_cycle" }, { "seed": 900000, "sample": 0, "native_reward": 0.6140552238672119, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 3, "model_tokens": 6332, "reward_channels": { "native": 0.6140552238672119, "efficiency": 0.5, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "competing_cycles" }, { "seed": 900000, "sample": 1, "native_reward": 0.6140552238672119, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 3325, "reward_channels": { "native": 0.6140552238672119, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "competing_cycles" }, { "seed": 900000, "sample": 2, "native_reward": 0.9879763779626005, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 6128, "reward_channels": { "native": 0.9879763779626005, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "competing_cycles" }, { "seed": 900000, "sample": 3, "native_reward": 0.578198213067076, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 6, "model_tokens": 7521, "reward_channels": { "native": 0.578198213067076, "efficiency": 0.0, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "competing_cycles" }, { "seed": 900001, "sample": 0, "native_reward": 0.25, "promotion_success": false, "done": false, "errors": 0, "tool_steps": 6, "model_tokens": 7015, "reward_channels": { "native": 0.25, "efficiency": 0.0, "reliability": -1.0 }, "verification_level": null, "failed_gates": [ "finalized" ], "family": "competing_cycles" }, { "seed": 900001, "sample": 1, "native_reward": 0.6145108982971254, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 4539, "reward_channels": { "native": 0.6145108982971254, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "competing_cycles" }, { "seed": 900001, "sample": 2, "native_reward": 0.6512057439829095, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 4658, "reward_channels": { "native": 0.6512057439829095, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "competing_cycles" }, { "seed": 900001, "sample": 3, "native_reward": 0.6099965928016736, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 6365, "reward_channels": { "native": 0.6099965928016736, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "competing_cycles" }, { "seed": 900000, "sample": 0, "native_reward": 0.7927032521897649, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 6, "model_tokens": 3955, "reward_channels": { "native": 0.7927032521897649, "efficiency": 0.0, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "four_hop_cycle" }, { "seed": 900000, "sample": 1, "native_reward": 0.609376392538787, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 3825, "reward_channels": { "native": 0.609376392538787, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "four_hop_cycle" }, { "seed": 900000, "sample": 2, "native_reward": 0.7503548418171574, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 3, "model_tokens": 5119, "reward_channels": { "native": 0.7503548418171574, "efficiency": 0.5, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "four_hop_cycle" }, { "seed": 900000, "sample": 3, "native_reward": 0.7005591947284964, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 4116, "reward_channels": { "native": 0.7005591947284964, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "four_hop_cycle" }, { "seed": 900001, "sample": 0, "native_reward": 0.739978752793518, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 3, "model_tokens": 3746, "reward_channels": { "native": 0.739978752793518, "efficiency": 0.5, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "four_hop_cycle" }, { "seed": 900001, "sample": 1, "native_reward": 0.6095701077161815, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 2042, "reward_channels": { "native": 0.6095701077161815, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "four_hop_cycle" }, { "seed": 900001, "sample": 2, "native_reward": 0.25, "promotion_success": false, "done": false, "errors": 0, "tool_steps": 6, "model_tokens": 4521, "reward_channels": { "native": 0.25, "efficiency": 0.0, "reliability": -1.0 }, "verification_level": null, "failed_gates": [ "finalized" ], "family": "four_hop_cycle" }, { "seed": 900001, "sample": 3, "native_reward": 0.739978752793518, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 3, "model_tokens": 4345, "reward_channels": { "native": 0.739978752793518, "efficiency": 0.5, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "four_hop_cycle" }, { "seed": 900000, "sample": 0, "native_reward": 0.8262556477476042, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 3160, "reward_channels": { "native": 0.8262556477476042, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "size_sensitive_cycle" }, { "seed": 900000, "sample": 1, "native_reward": 0.8262556477476042, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 3, "model_tokens": 5310, "reward_channels": { "native": 0.8262556477476042, "efficiency": 0.5, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "size_sensitive_cycle" }, { "seed": 900000, "sample": 2, "native_reward": 0.175, "promotion_success": false, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 7909, "reward_channels": { "native": 0.175, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [ "positive_net_profit", "stress_positive", "task_complete" ], "family": "size_sensitive_cycle" }, { "seed": 900000, "sample": 3, "native_reward": 0.8262556477476042, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 3915, "reward_channels": { "native": 0.8262556477476042, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "size_sensitive_cycle" }, { "seed": 900001, "sample": 0, "native_reward": 0.8245803237223959, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 3, "model_tokens": 3958, "reward_channels": { "native": 0.8245803237223959, "efficiency": 0.5, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "size_sensitive_cycle" }, { "seed": 900001, "sample": 1, "native_reward": 0.25, "promotion_success": false, "done": false, "errors": 0, "tool_steps": 3, "model_tokens": 5047, "reward_channels": { "native": 0.25, "efficiency": 0.0, "reliability": -1.0 }, "verification_level": null, "failed_gates": [ "finalized" ], "family": "size_sensitive_cycle" }, { "seed": 900001, "sample": 2, "native_reward": 0.8245803237223959, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 2468, "reward_channels": { "native": 0.8245803237223959, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "size_sensitive_cycle" }, { "seed": 900001, "sample": 3, "native_reward": 0.8245803237223959, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 3, "model_tokens": 4523, "reward_channels": { "native": 0.8245803237223959, "efficiency": 0.5, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "size_sensitive_cycle" } ] }, { "run_id": "1789312185", "mode": "thinking", "method": "cispo", "steps": 50, "reward_points": 25, "train_mean": 0.8387576556205749, "train_last10": 0.8950886785984039, "holdout_native": 0.606788222286585, "holdout_episodes": 40, "holdout_successes": 33, "mean_step_seconds": 907.3916725414009, "label": "CISPO", "state_at_cutoff": "complete", "step_hours": 12.602662118630569, "peak_allocated_gib": 70.00158309936523, "peak_reserved_gib": 77.822265625, "extra_refill_rounds": 0.0, "groups_refilled": 0.0, "holdout_component_means": { "native": 0.606788222286585, "efficiency": 0.46666666666666673, "reliability": 0.7 }, "configuration": { "model": { "id": "Qwen/Qwen3-14B", "revision": "40c069824f4251a91eefaf281ebe4c544efd3e18", "loss": "cispo", "algorithm": "grpo", "no_think": false, "thinking_budget_tokens": 2048, "lora": true, "quantize": true, "lora_r": 32, "lora_alpha": 64, "use_dora": false }, "loss": "cispo", "normalization": "joint", "entropy_control": { "method": "monitor", "epsilon_low": 0.2, "epsilon_high": 0.2, "zeta_initial": 0.001, "zeta_min": 0.0001, "zeta_max": 0.05, "holdout_split": "benchmark", "holdout_seeds_per_family": 2, "num_iterations": 2 }, "dynamic_sampling": false, "sampling": { "group_size": 4, "temperature": 1.0, "max_completion_tokens": 22528, "top_entropy_quantile": 1.0, "temperature_decay_to": null }, "learning_rate": 1e-05, "turn_max_tokens": 3072 }, "source_sha256": { "metrics.jsonl": "5194bee8a98c562cbef72b829d9d573e4442fe9697418446b63b2cd43cc2b9a9", "entropy_holdout_episodes.jsonl": "091926046554bb56cd696d89472414dfa8fbae3281b14b732027b945aca9d900", "resolved_training_config.json": "025bb90f7e7c1dd0afc8438d21ddb6dfd229d8bef4df7fd62f09d66d92f579e4" }, "training_steps": [ { "step": 1, "loss": 0.05959773063659668, "grad_norm": 0.061279296875, "learning_rate": 1e-05, "thinking/tokens_mean": 606.2361111111111, "thinking/budget_hit_fraction": 0.14583333333333334, "num_tokens": 51975.0, "completions/mean_length": 3604.375, "completions/min_length": 2890.0, "completions/max_length": 6014.0, "completions/clipped_ratio": 0.25, "completions/mean_terminated_length": 3782.166748046875, "completions/min_terminated_length": 2890.0, "completions/max_terminated_length": 6014.0, "tools/call_frequency": 2.25, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.7262155413627625, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 0.13481803238391876, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.4583333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.2500000298023224, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/mean": 0.45231834053993225, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/std": 0.526568591594696, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/mean": 0.2916666865348816, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/std": 0.34359216690063477, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/mean": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/std": 1.154700517654419, "reward": 0.7080169916152954, "reward_std": 0.5734051465988159, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 0.5, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.14539151638746262, "cispo_clip_ratio": 0.0, "step_time": 1432.8779994019997, "entropy_control/entropy": 0.20009788357278327, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.004, "adapter_sync_time": 0.0, "generation_time": 1241.577735, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1276.206926, "forward_backward_time": 156.563045, "rollout_overhead_time": 34.629191, "peak_allocated_gib": 46.211081981658936, "peak_reserved_gib": 51.38671875 }, { "step": 2, "loss": 0.059736188501119614, "grad_norm": 0.061767578125, "learning_rate": 9.800000000000001e-06, "entropy": 0.14504320733249187, "cispo_clip_ratio": 0.0025359369174111634, "step_time": 156.6762321429992, "entropy_control/entropy": 0.19971678657231526, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.008, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 156.555146, "rollout_overhead_time": 0.0, "peak_allocated_gib": 46.689598083496094, "peak_reserved_gib": 51.484375, "reward": null }, { "step": 3, "loss": 0.07673680037260056, "grad_norm": 0.0673828125, "learning_rate": 9.600000000000001e-06, "thinking/tokens_mean": 675.96, "thinking/budget_hit_fraction": 0.15, "num_tokens": 103240.0, "completions/mean_length": 3696.125, "completions/min_length": 1260.0, "completions/max_length": 5569.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 3696.125, "completions/min_terminated_length": 1260.0, "completions/max_terminated_length": 5569.0, "tools/call_frequency": 3.0, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.7281777262687683, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 0.12758977711200714, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.15430335700511932, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 0.0, "reward": 0.9531776905059814, "reward_std": 0.120913565158844, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.19031991995871067, "cispo_clip_ratio": 0.0, "step_time": 1272.7960069709998, "entropy_control/entropy": 0.20376649588502274, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.012, "adapter_sync_time": 0.0, "generation_time": 1097.50989, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1128.21115, "forward_backward_time": 144.460731, "rollout_overhead_time": 30.70126, "peak_allocated_gib": 46.689598083496094, "peak_reserved_gib": 51.484375 }, { "step": 4, "loss": 0.07664913684129715, "grad_norm": 0.06689453125, "learning_rate": 9.4e-06, "entropy": 0.19033740274608135, "cispo_clip_ratio": 0.0025286943127866834, "step_time": 144.7639433240006, "entropy_control/entropy": 0.20369696012972682, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.016, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 144.645296, "rollout_overhead_time": 0.0, "peak_allocated_gib": 46.689598083496094, "peak_reserved_gib": 51.484375, "reward": null }, { "step": 5, "loss": 0.0, "grad_norm": 0.0, "learning_rate": 9.200000000000002e-06, "thinking/tokens_mean": 673.34375, "thinking/budget_hit_fraction": 0.21875, "num_tokens": 148302.0, "completions/mean_length": 2927.75, "completions/min_length": 1462.0, "completions/max_length": 4656.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 2927.75, "completions/min_terminated_length": 1462.0, "completions/max_terminated_length": 4656.0, "tools/call_frequency": 2.25, "tools/failure_frequency": 0.1111111119389534, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.6093767881393433, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 4.6674944314872846e-05, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.6666666865348816, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 0.0, "reward": 0.8427101373672485, "reward_std": 4.6674944314872846e-05, "frac_reward_zero_std": 1.0, "frac_groups_zero_std": 1.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.18442012928426266, "cispo_clip_ratio": 0.0, "step_time": 984.4018064339998, "entropy_control/entropy": 0.19255304304690926, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.02, "adapter_sync_time": 0.0, "generation_time": 882.771041, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 901.64074, "forward_backward_time": 82.650254, "rollout_overhead_time": 18.869699, "peak_allocated_gib": 46.689598083496094, "peak_reserved_gib": 51.484375 }, { "step": 6, "loss": 0.0, "grad_norm": 0.0, "learning_rate": 9e-06, "entropy": 0.18489300087094307, "cispo_clip_ratio": 0.0, "step_time": 82.75198730199918, "entropy_control/entropy": 0.19300358485280206, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.024, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 82.640343, "rollout_overhead_time": 0.0, "peak_allocated_gib": 46.689598083496094, "peak_reserved_gib": 51.484375, "reward": null }, { "step": 7, "loss": 0.051492560654878616, "grad_norm": 0.044677734375, "learning_rate": 8.8e-06, "thinking/tokens_mean": 661.6743197278912, "thinking/budget_hit_fraction": 0.16581632653061223, "num_tokens": 214410.0, "completions/mean_length": 4796.0, "completions/min_length": 3071.0, "completions/max_length": 7083.0, "completions/clipped_ratio": 0.125, "completions/mean_terminated_length": 5042.4287109375, "completions/min_terminated_length": 3251.0, "completions/max_terminated_length": 7083.0, "tools/call_frequency": 3.375, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/mean": 0.5317510962486267, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/std": 0.35458892583847046, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/mean": 0.125, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/std": 0.15957120060920715, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/std": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/mean": 0.8048485517501831, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/std": 0.04181578755378723, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/std": 0.13608276844024658, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/std": 0.0, "reward": 0.8339248299598694, "reward_std": 0.4224560856819153, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 0.5, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.16800322569906712, "cispo_clip_ratio": 0.0, "step_time": 2108.4935946019996, "entropy_control/entropy": 0.19551524949746044, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.028, "adapter_sync_time": 0.0, "generation_time": 1749.503862, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1809.978419, "forward_backward_time": 298.392886, "rollout_overhead_time": 60.474557, "peak_allocated_gib": 68.21795797348022, "peak_reserved_gib": 75.5078125 }, { "step": 8, "loss": 0.05173943191766739, "grad_norm": 0.047119140625, "learning_rate": 8.6e-06, "entropy": 0.1678721159696579, "cispo_clip_ratio": 0.00378150338656269, "step_time": 298.3895189370005, "entropy_control/entropy": 0.19539698056410154, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.032, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 298.282565, "rollout_overhead_time": 0.0, "peak_allocated_gib": 68.21795797348022, "peak_reserved_gib": 75.5078125, "reward": null }, { "step": 9, "loss": 0.023579686880111694, "grad_norm": 0.052978515625, "learning_rate": 8.400000000000001e-06, "thinking/tokens_mean": 933.9440476190475, "thinking/budget_hit_fraction": 0.26071428571428573, "num_tokens": 277293.0, "completions/mean_length": 4576.375, "completions/min_length": 2726.0, "completions/max_length": 7056.0, "completions/clipped_ratio": 0.125, "completions/mean_terminated_length": 4791.4287109375, "completions/min_terminated_length": 2726.0, "completions/max_terminated_length": 7056.0, "tools/call_frequency": 2.5, "tools/failure_frequency": 0.05000000074505806, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.6683154702186584, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 0.11753809452056885, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.625, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.0833333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/mean": 0.41591623425483704, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/std": 0.40022072196006775, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/mean": 0.3333333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/std": 0.27216553688049316, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/std": 1.0, "reward": 0.7160741686820984, "reward_std": 0.4268726706504822, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 0.5, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.1945186909288168, "cispo_clip_ratio": 0.0, "step_time": 1883.930043686999, "entropy_control/entropy": 0.22380286454406373, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.036, "adapter_sync_time": 0.0, "generation_time": 1629.017555, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1673.223345, "forward_backward_time": 210.581205, "rollout_overhead_time": 44.20579, "peak_allocated_gib": 68.21795797348022, "peak_reserved_gib": 75.5078125 }, { "step": 10, "loss": 0.023223545402288437, "grad_norm": 0.052001953125, "learning_rate": 8.2e-06, "entropy": 0.19437705539166927, "cispo_clip_ratio": 0.0015916558913886547, "step_time": 210.70282552500066, "entropy_control/entropy": 0.2235291455100569, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.04, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 210.58093, "rollout_overhead_time": 0.0, "peak_allocated_gib": 68.21795797348022, "peak_reserved_gib": 75.5078125, "reward": null }, { "step": 11, "loss": 0.06816785037517548, "grad_norm": 0.057373046875, "learning_rate": 8.000000000000001e-06, "thinking/tokens_mean": 563.6964285714286, "thinking/budget_hit_fraction": 0.08928571428571429, "num_tokens": 333240.0, "completions/mean_length": 3684.375, "completions/min_length": 1718.0, "completions/max_length": 7003.0, "completions/clipped_ratio": 0.125, "completions/mean_terminated_length": 3772.000244140625, "completions/min_terminated_length": 1718.0, "completions/max_terminated_length": 7003.0, "tools/call_frequency": 2.75, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__native/mean": 0.5645434856414795, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__native/std": 0.3898860514163971, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__efficiency/mean": 0.2083333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__efficiency/std": 0.31549492478370667, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__reliability/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__reliability/std": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/mean": 0.6680818796157837, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/std": 0.11817905306816101, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/mean": 0.625, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/std": 0.0833333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/std": 0.0, "reward": 0.7871460914611816, "reward_std": 0.4111788272857666, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 0.5, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.17892753519117832, "cispo_clip_ratio": 0.0, "step_time": 1801.6967272129987, "entropy_control/entropy": 0.20602153379997154, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.044, "adapter_sync_time": 0.0, "generation_time": 1499.890518, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1550.885347, "forward_backward_time": 250.697833, "rollout_overhead_time": 50.994829, "peak_allocated_gib": 68.21795797348022, "peak_reserved_gib": 75.5078125 }, { "step": 12, "loss": 0.06817689538002014, "grad_norm": 0.057373046875, "learning_rate": 7.800000000000002e-06, "entropy": 0.1793829184025526, "cispo_clip_ratio": 0.0026480586966499686, "step_time": 250.81987606999792, "entropy_control/entropy": 0.20650926472102377, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.048, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 250.708816, "rollout_overhead_time": 0.0, "peak_allocated_gib": 68.21795797348022, "peak_reserved_gib": 75.5078125, "reward": null }, { "step": 13, "loss": 0.05013948678970337, "grad_norm": 0.046142578125, "learning_rate": 7.600000000000001e-06, "thinking/tokens_mean": 884.5952380952382, "thinking/budget_hit_fraction": 0.2857142857142857, "num_tokens": 395227.0, "completions/mean_length": 4463.375, "completions/min_length": 2704.0, "completions/max_length": 6607.0, "completions/clipped_ratio": 0.125, "completions/mean_terminated_length": 4662.2861328125, "completions/min_terminated_length": 2704.0, "completions/max_terminated_length": 6607.0, "tools/call_frequency": 2.375, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.6323727369308472, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 0.04563635587692261, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.5833333730697632, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.1666666716337204, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/mean": 0.4600570797920227, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/std": 0.30670472979545593, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/mean": 0.375, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/std": 0.28463754057884216, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/std": 1.0, "reward": 0.7201732397079468, "reward_std": 0.37280964851379395, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 0.5, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.17529656179249287, "cispo_clip_ratio": 0.0, "step_time": 1510.479915453001, "entropy_control/entropy": 0.2079731292202752, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.052, "adapter_sync_time": 0.0, "generation_time": 1285.932623, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1325.368707, "forward_backward_time": 185.008122, "rollout_overhead_time": 39.436085, "peak_allocated_gib": 68.21795797348022, "peak_reserved_gib": 75.5078125 }, { "step": 14, "loss": 0.04992172494530678, "grad_norm": 0.045654296875, "learning_rate": 7.4e-06, "entropy": 0.1745990291237831, "cispo_clip_ratio": 0.001840055309003219, "step_time": 185.10257531900243, "entropy_control/entropy": 0.2071626262243796, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.056, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 184.996985, "rollout_overhead_time": 0.0, "peak_allocated_gib": 68.21795797348022, "peak_reserved_gib": 75.5078125, "reward": null }, { "step": 15, "loss": -0.00855071097612381, "grad_norm": 0.059814453125, "learning_rate": 7.2000000000000005e-06, "thinking/tokens_mean": 816.6666666666666, "thinking/budget_hit_fraction": 0.14583333333333334, "num_tokens": 460157.0, "completions/mean_length": 4632.75, "completions/min_length": 3274.0, "completions/max_length": 7270.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 4632.75, "completions/min_terminated_length": 3274.0, "completions/max_terminated_length": 7270.0, "tools/call_frequency": 2.875, "tools/failure_frequency": 0.043478261679410934, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/mean": 0.6072787046432495, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/std": 0.013663023710250854, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/std": 0.19245010614395142, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/mean": 0.8640799522399902, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/std": 0.08177304267883301, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/mean": 0.5833333730697632, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/std": 0.1666666716337204, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/std": 0.0, "reward": 0.962762713432312, "reward_std": 0.1477309763431549, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.21425736043602228, "cispo_clip_ratio": 0.0, "step_time": 1604.1281318940019, "entropy_control/entropy": 0.22083469835945016, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.06, "adapter_sync_time": 0.0, "generation_time": 1360.390406, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1402.625146, "forward_backward_time": 201.401124, "rollout_overhead_time": 42.23474, "peak_allocated_gib": 68.21795797348022, "peak_reserved_gib": 75.5078125 }, { "step": 16, "loss": -0.008626427501440048, "grad_norm": 0.053955078125, "learning_rate": 7e-06, "entropy": 0.2144729858264327, "cispo_clip_ratio": 0.0010139655496459454, "step_time": 201.4823421400033, "entropy_control/entropy": 0.22101901031865806, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.064, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 201.376088, "rollout_overhead_time": 0.0, "peak_allocated_gib": 68.21795797348022, "peak_reserved_gib": 75.5078125, "reward": null }, { "step": 17, "loss": -0.022877300158143044, "grad_norm": 0.057373046875, "learning_rate": 6.800000000000001e-06, "thinking/tokens_mean": 753.3971088435374, "thinking/budget_hit_fraction": 0.21428571428571427, "num_tokens": 532896.0, "completions/mean_length": 5401.375, "completions/min_length": 2972.0, "completions/max_length": 8553.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 5401.375, "completions/min_terminated_length": 2972.0, "completions/max_terminated_length": 8553.0, "tools/call_frequency": 3.875, "tools/failure_frequency": 0.032258063554763794, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/mean": 0.4319382905960083, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/std": 0.2100842446088791, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/mean": 0.2916666865348816, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/std": 0.34359216690063477, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/mean": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/std": 1.154700517654419, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/mean": 0.7295446395874023, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/std": 0.09698600322008133, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/mean": 0.4583333730697632, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/std": 0.20971761643886566, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/std": 0.0, "reward": 0.6994915008544922, "reward_std": 0.40824830532073975, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.18524269945919514, "cispo_clip_ratio": 0.0, "step_time": 2233.6526770439923, "entropy_control/entropy": 0.1851913962170918, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.068, "adapter_sync_time": 0.0, "generation_time": 1883.715355, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1942.822108, "forward_backward_time": 290.71654, "rollout_overhead_time": 59.106752, "peak_allocated_gib": 68.21795797348022, "peak_reserved_gib": 75.5078125 }, { "step": 18, "loss": -0.022642826661467552, "grad_norm": 0.052978515625, "learning_rate": 6.600000000000001e-06, "entropy": 0.18491178378462791, "cispo_clip_ratio": 0.002408703847322613, "step_time": 290.79190560100324, "entropy_control/entropy": 0.18482873933896737, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.072, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 290.683148, "rollout_overhead_time": 0.0, "peak_allocated_gib": 68.21795797348022, "peak_reserved_gib": 75.5078125, "reward": null }, { "step": 19, "loss": -0.027121691033244133, "grad_norm": 0.0556640625, "learning_rate": 6.4000000000000006e-06, "thinking/tokens_mean": 573.375, "thinking/budget_hit_fraction": 0.17857142857142858, "num_tokens": 587462.0, "completions/mean_length": 3930.25, "completions/min_length": 1782.0, "completions/max_length": 6024.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 3930.25, "completions/min_terminated_length": 1782.0, "completions/max_terminated_length": 6024.0, "tools/call_frequency": 2.875, "tools/failure_frequency": 0.08695652335882187, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.6096870303153992, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.6666666865348816, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/mean": 0.7011142373085022, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/std": 0.35796624422073364, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/mean": 0.4583333730697632, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/std": 0.3154948949813843, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/std": 0.0, "reward": 0.8835256695747375, "reward_std": 0.2356216162443161, "frac_reward_zero_std": 0.5, "frac_groups_zero_std": 0.5, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.1949994508177042, "cispo_clip_ratio": 0.0, "step_time": 1453.557944958, "entropy_control/entropy": 0.2024309433565279, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.076, "adapter_sync_time": 0.0, "generation_time": 1206.809612, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1249.328148, "forward_backward_time": 204.126315, "rollout_overhead_time": 42.518535, "peak_allocated_gib": 68.21795797348022, "peak_reserved_gib": 75.5078125 }, { "step": 20, "loss": -0.027196794748306274, "grad_norm": 0.048583984375, "learning_rate": 6.200000000000001e-06, "entropy": 0.19486972130835056, "cispo_clip_ratio": 0.0017310466500930488, "step_time": 204.21084985199923, "entropy_control/entropy": 0.20227207682142498, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.08, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 204.108367, "rollout_overhead_time": 0.0, "peak_allocated_gib": 68.21795797348022, "peak_reserved_gib": 75.5078125, "reward": null }, { "step": 21, "loss": -0.0015484169125556946, "grad_norm": 0.0712890625, "learning_rate": 6e-06, "thinking/tokens_mean": 555.3928571428571, "thinking/budget_hit_fraction": 0.125, "num_tokens": 650873.0, "completions/mean_length": 4227.375, "completions/min_length": 1598.0, "completions/max_length": 5516.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 4227.375, "completions/min_terminated_length": 1598.0, "completions/max_terminated_length": 5516.0, "tools/call_frequency": 3.375, "tools/failure_frequency": 0.07407407462596893, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/mean": 0.6140941381454468, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/mean": 0.5833333730697632, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/std": 0.09622505307197571, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/mean": 0.5919007062911987, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/std": 0.2379184514284134, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/mean": 0.3750000298023224, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/std": 0.28463754057884216, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/std": 1.0, "reward": 0.7769557237625122, "reward_std": 0.2997303307056427, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.19899269193410873, "cispo_clip_ratio": 0.0, "step_time": 1420.192871165993, "entropy_control/entropy": 0.20385769032553278, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.084, "adapter_sync_time": 0.0, "generation_time": 1173.81351, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1215.831439, "forward_backward_time": 204.259396, "rollout_overhead_time": 42.017929, "peak_allocated_gib": 68.21795797348022, "peak_reserved_gib": 75.5078125 }, { "step": 22, "loss": -0.0017351359128952026, "grad_norm": 0.06689453125, "learning_rate": 5.8e-06, "entropy": 0.19878941774368286, "cispo_clip_ratio": 0.003685811476316303, "step_time": 204.32607309300874, "entropy_control/entropy": 0.203702783959607, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.088, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 204.224661, "rollout_overhead_time": 0.0, "peak_allocated_gib": 68.21795797348022, "peak_reserved_gib": 75.5078125, "reward": null }, { "step": 23, "loss": 0.027916889637708664, "grad_norm": 0.07373046875, "learning_rate": 5.600000000000001e-06, "thinking/tokens_mean": 585.8214285714286, "thinking/budget_hit_fraction": 0.125, "num_tokens": 705962.0, "completions/mean_length": 4000.625, "completions/min_length": 1995.0, "completions/max_length": 7082.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 4000.625, "completions/min_terminated_length": 1995.0, "completions/max_terminated_length": 7082.0, "tools/call_frequency": 3.25, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.7249159812927246, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 0.13324211537837982, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.4583333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.3154948949813843, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/mean": 0.857026994228363, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/std": 0.06957332044839859, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/mean": 0.4583333730697632, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/std": 0.3154948949813843, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/std": 0.0, "reward": 1.0138881206512451, "reward_std": 0.11696893721818924, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.17570242285728455, "cispo_clip_ratio": 0.0, "step_time": 1558.686015184001, "entropy_control/entropy": 0.1775421606632509, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.092, "adapter_sync_time": 0.0, "generation_time": 1281.633612, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1328.705847, "forward_backward_time": 229.87527, "rollout_overhead_time": 47.072235, "peak_allocated_gib": 68.21795797348022, "peak_reserved_gib": 75.5078125 }, { "step": 24, "loss": 0.0279421154409647, "grad_norm": 0.07763671875, "learning_rate": 5.400000000000001e-06, "entropy": 0.1764808613806963, "cispo_clip_ratio": 0.001917229441460222, "step_time": 229.9780768470082, "entropy_control/entropy": 0.1782901138241022, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.096, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 229.876599, "rollout_overhead_time": 0.0, "peak_allocated_gib": 68.21795797348022, "peak_reserved_gib": 75.5078125, "reward": null }, { "step": 25, "loss": 0.020461905747652054, "grad_norm": 0.058349609375, "learning_rate": 5.2e-06, "thinking/tokens_mean": 499.8494897959184, "thinking/budget_hit_fraction": 0.08928571428571429, "num_tokens": 755267.0, "completions/mean_length": 3073.125, "completions/min_length": 1262.0, "completions/max_length": 7150.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 3073.125, "completions/min_terminated_length": 1262.0, "completions/max_terminated_length": 7150.0, "tools/call_frequency": 2.875, "tools/failure_frequency": 0.08695652335882187, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.609268307685852, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.6666666865348816, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/mean": 0.25848397612571716, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/std": 0.2560417056083679, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/mean": 0.2083333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/std": 0.3154948949813843, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/mean": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/std": 1.154700517654419, "reward": 0.5557511448860168, "reward_std": 0.4255915880203247, "frac_reward_zero_std": 0.5, "frac_groups_zero_std": 0.5, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 0.5, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.17823073640465736, "cispo_clip_ratio": 0.0, "step_time": 1374.3095642089938, "entropy_control/entropy": 0.19470474385825953, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.1, "adapter_sync_time": 0.0, "generation_time": 1127.34155, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1170.193834, "forward_backward_time": 204.009631, "rollout_overhead_time": 42.852285, "peak_allocated_gib": 68.21795797348022, "peak_reserved_gib": 75.5078125 }, { "step": 26, "loss": 0.02035672590136528, "grad_norm": 0.05615234375, "learning_rate": 5e-06, "entropy": 0.17873864714056253, "cispo_clip_ratio": 0.0013546488480642438, "step_time": 204.0851919899942, "entropy_control/entropy": 0.19520758576970662, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.104, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 203.98208, "rollout_overhead_time": 0.0, "peak_allocated_gib": 68.21795797348022, "peak_reserved_gib": 75.5078125, "reward": null }, { "step": 27, "loss": -0.014896320179104805, "grad_norm": 0.061279296875, "learning_rate": 4.800000000000001e-06, "thinking/tokens_mean": 691.5535714285714, "thinking/budget_hit_fraction": 0.16964285714285712, "num_tokens": 813017.0, "completions/mean_length": 3916.25, "completions/min_length": 2443.0, "completions/max_length": 7757.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 3916.25, "completions/min_terminated_length": 2443.0, "completions/max_terminated_length": 7757.0, "tools/call_frequency": 2.625, "tools/failure_frequency": 0.0476190485060215, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/mean": 0.4889552593231201, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/std": 0.3313716948032379, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/mean": 0.4166666865348816, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/std": 0.3191424012184143, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/std": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/mean": 0.696678638458252, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/std": 0.35369956493377686, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/std": 0.13608276844024658, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/std": 0.0, "reward": 0.7657335996627808, "reward_std": 0.4571972191333771, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 0.5, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.21122080087661743, "cispo_clip_ratio": 0.0, "step_time": 1509.8962310990028, "entropy_control/entropy": 0.205215673983445, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.108, "adapter_sync_time": 0.0, "generation_time": 1269.259398, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1310.572853, "forward_backward_time": 199.217093, "rollout_overhead_time": 41.313454, "peak_allocated_gib": 68.21795797348022, "peak_reserved_gib": 75.5078125 }, { "step": 28, "loss": -0.014891263097524643, "grad_norm": 0.0673828125, "learning_rate": 4.600000000000001e-06, "entropy": 0.21169204264879227, "cispo_clip_ratio": 0.003810049034655094, "step_time": 199.2965387799959, "entropy_control/entropy": 0.20553460569612514, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.112, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 199.192826, "rollout_overhead_time": 0.0, "peak_allocated_gib": 68.21795797348022, "peak_reserved_gib": 75.5078125, "reward": null }, { "step": 29, "loss": -0.004603184759616852, "grad_norm": 0.08349609375, "learning_rate": 4.4e-06, "thinking/tokens_mean": 661.1130952380953, "thinking/budget_hit_fraction": 0.125, "num_tokens": 874891.0, "completions/mean_length": 4451.75, "completions/min_length": 2944.0, "completions/max_length": 6588.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 4451.75, "completions/min_terminated_length": 2944.0, "completions/max_terminated_length": 6588.0, "tools/call_frequency": 3.375, "tools/failure_frequency": 0.03703703731298447, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.6719769239425659, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 0.12673872709274292, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.625, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.0833333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/mean": 0.579635739326477, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/std": 0.2551511824131012, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/mean": 0.2916666865348816, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/std": 0.28463754057884216, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/std": 1.0, "reward": 0.7987229824066162, "reward_std": 0.32421475648880005, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.20279635302722454, "cispo_clip_ratio": 0.0, "step_time": 2059.5521601690016, "entropy_control/entropy": 0.20050150044775508, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.116, "adapter_sync_time": 0.0, "generation_time": 1768.197283, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1817.310238, "forward_backward_time": 242.137128, "rollout_overhead_time": 49.112956, "peak_allocated_gib": 68.21795797348022, "peak_reserved_gib": 75.5078125 }, { "step": 30, "loss": -0.004749072249978781, "grad_norm": 0.0673828125, "learning_rate": 4.2000000000000004e-06, "entropy": 0.20302190445363522, "cispo_clip_ratio": 0.0029782248311676085, "step_time": 242.2114704769956, "entropy_control/entropy": 0.20068203646619256, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.12, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 242.104717, "rollout_overhead_time": 0.0, "peak_allocated_gib": 68.21795797348022, "peak_reserved_gib": 75.5078125, "reward": null }, { "step": 31, "loss": -0.013075809925794601, "grad_norm": 0.051513671875, "learning_rate": 4.000000000000001e-06, "thinking/tokens_mean": 557.8333333333334, "thinking/budget_hit_fraction": 0.14583333333333334, "num_tokens": 928197.0, "completions/mean_length": 3579.75, "completions/min_length": 2682.0, "completions/max_length": 4490.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 3579.75, "completions/min_terminated_length": 2682.0, "completions/max_terminated_length": 4490.0, "tools/call_frequency": 2.625, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/mean": 0.8651880025863647, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/std": 0.07400666177272797, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/mean": 0.5625, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/std": 0.19795581698417664, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/std": 0.0, "reward": 1.0933129787445068, "reward_std": 0.06444501876831055, "frac_reward_zero_std": 0.5, "frac_groups_zero_std": 0.5, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.2001099605113268, "cispo_clip_ratio": 0.0, "step_time": 1127.0236637460002, "entropy_control/entropy": 0.19981127089133777, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.124, "adapter_sync_time": 0.0, "generation_time": 919.242119, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 955.346876, "forward_backward_time": 171.570368, "rollout_overhead_time": 36.104757, "peak_allocated_gib": 68.21795797348022, "peak_reserved_gib": 75.5078125 }, { "step": 32, "loss": -0.013320034369826317, "grad_norm": 0.048583984375, "learning_rate": 3.8000000000000005e-06, "entropy": 0.1999568399041891, "cispo_clip_ratio": 0.0010547005804255605, "step_time": 171.6825249080066, "entropy_control/entropy": 0.19959165633013187, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.128, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 171.574778, "rollout_overhead_time": 0.0, "peak_allocated_gib": 68.21795797348022, "peak_reserved_gib": 75.5078125, "reward": null }, { "step": 33, "loss": -0.07050250470638275, "grad_norm": 0.05712890625, "learning_rate": 3.6000000000000003e-06, "thinking/tokens_mean": 520.4097222222223, "thinking/budget_hit_fraction": 0.16666666666666666, "num_tokens": 978403.0, "completions/mean_length": 3380.25, "completions/min_length": 2344.0, "completions/max_length": 6682.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 3380.25, "completions/min_terminated_length": 2344.0, "completions/max_terminated_length": 6682.0, "tools/call_frequency": 2.875, "tools/failure_frequency": 0.1304347813129425, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.6095701456069946, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.6666666865348816, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/mean": 0.7019290924072266, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/std": 0.3590852916240692, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/std": 0.235702246427536, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/std": 0.0, "reward": 0.8849163055419922, "reward_std": 0.237877756357193, "frac_reward_zero_std": 0.5, "frac_groups_zero_std": 0.5, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.19189164973795414, "cispo_clip_ratio": 0.0, "step_time": 1447.7261638749987, "entropy_control/entropy": 0.20349327952755297, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.132, "adapter_sync_time": 0.0, "generation_time": 1258.473654, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1292.24578, "forward_backward_time": 155.374454, "rollout_overhead_time": 33.772126, "peak_allocated_gib": 68.21795797348022, "peak_reserved_gib": 75.5078125 }, { "step": 34, "loss": -0.07099034637212753, "grad_norm": 0.056884765625, "learning_rate": 3.4000000000000005e-06, "entropy": 0.1921864915639162, "cispo_clip_ratio": 0.0021364886197261512, "step_time": 155.45952711000064, "entropy_control/entropy": 0.20378273496889898, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.136, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 155.353332, "rollout_overhead_time": 0.0, "peak_allocated_gib": 68.21795797348022, "peak_reserved_gib": 75.5078125, "reward": null }, { "step": 35, "loss": 0.003738459199666977, "grad_norm": 0.05810546875, "learning_rate": 3.2000000000000003e-06, "thinking/tokens_mean": 696.1147959183673, "thinking/budget_hit_fraction": 0.125, "num_tokens": 1050726.0, "completions/mean_length": 5177.375, "completions/min_length": 3816.0, "completions/max_length": 7994.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 5177.375, "completions/min_terminated_length": 3816.0, "completions/max_terminated_length": 7994.0, "tools/call_frequency": 3.875, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/mean": 0.7032192945480347, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/std": 0.20619390904903412, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/mean": 0.3541666865348816, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/std": 0.24295634031295776, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/mean": 0.75, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/std": 0.7071067690849304, "reward": 0.8709275722503662, "reward_std": 0.34340617060661316, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.2363110724836588, "cispo_clip_ratio": 0.0, "step_time": 2078.5321845780054, "entropy_control/entropy": 0.23738854010738542, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.14, "adapter_sync_time": 0.0, "generation_time": 1703.779782, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1766.373098, "forward_backward_time": 312.054464, "rollout_overhead_time": 62.593315, "peak_allocated_gib": 70.00158309936523, "peak_reserved_gib": 77.8125 }, { "step": 36, "loss": 0.002995230257511139, "grad_norm": 0.0595703125, "learning_rate": 3e-06, "entropy": 0.23690798692405224, "cispo_clip_ratio": 0.0029003190575167537, "step_time": 312.09610687001987, "entropy_control/entropy": 0.23801668714223417, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.144, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 311.992477, "rollout_overhead_time": 0.0, "peak_allocated_gib": 70.00158309936523, "peak_reserved_gib": 77.8125, "reward": null }, { "step": 37, "loss": 0.038784705102443695, "grad_norm": 0.083984375, "learning_rate": 2.8000000000000003e-06, "thinking/tokens_mean": 623.4791666666666, "thinking/budget_hit_fraction": 0.125, "num_tokens": 1105921.0, "completions/mean_length": 4019.375, "completions/min_length": 1065.0, "completions/max_length": 6134.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 4019.375, "completions/min_terminated_length": 1065.0, "completions/max_terminated_length": 6134.0, "tools/call_frequency": 3.375, "tools/failure_frequency": 0.03703703731298447, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.7263758182525635, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 0.1347351223230362, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.4583333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.2500000298023224, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__native/mean": 0.7034698128700256, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__native/std": 0.1064884141087532, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__efficiency/mean": 0.4583333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__efficiency/std": 0.2500000298023224, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__reliability/std": 0.0, "reward": 0.9378395080566406, "reward_std": 0.10221715271472931, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.18527615256607533, "cispo_clip_ratio": 0.0, "step_time": 1434.1412884339952, "entropy_control/entropy": 0.19221650355896527, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.148, "adapter_sync_time": 0.0, "generation_time": 1225.456016, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1261.67663, "forward_backward_time": 172.355831, "rollout_overhead_time": 36.220614, "peak_allocated_gib": 70.00158309936523, "peak_reserved_gib": 77.8125 }, { "step": 38, "loss": 0.03889436274766922, "grad_norm": 0.06787109375, "learning_rate": 2.6e-06, "entropy": 0.18586737290024757, "cispo_clip_ratio": 0.00231436311150901, "step_time": 172.46150075998594, "entropy_control/entropy": 0.19268527457717816, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.152, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 172.353799, "rollout_overhead_time": 0.0, "peak_allocated_gib": 70.00158309936523, "peak_reserved_gib": 77.8125, "reward": null }, { "step": 39, "loss": 0.03874988853931427, "grad_norm": 0.058837890625, "learning_rate": 2.4000000000000003e-06, "thinking/tokens_mean": 595.4499999999999, "thinking/budget_hit_fraction": 0.125, "num_tokens": 1165864.0, "completions/mean_length": 4401.875, "completions/min_length": 2761.0, "completions/max_length": 6486.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 4401.875, "completions/min_terminated_length": 2761.0, "completions/max_terminated_length": 6486.0, "tools/call_frequency": 3.75, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__native/mean": 0.7549160122871399, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__native/std": 0.1321459859609604, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__efficiency/mean": 0.375, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__efficiency/std": 0.2781743109226227, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__reliability/std": 0.0, "reward": 0.9736660718917847, "reward_std": 0.11896516382694244, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.18749587051570415, "cispo_clip_ratio": 0.0, "step_time": 1806.3814684290046, "entropy_control/entropy": 0.18776935633964761, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.156, "adapter_sync_time": 0.0, "generation_time": 1564.981242, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1606.591353, "forward_backward_time": 199.681867, "rollout_overhead_time": 41.610111, "peak_allocated_gib": 70.00158309936523, "peak_reserved_gib": 77.8125 }, { "step": 40, "loss": 0.03878646716475487, "grad_norm": 0.0595703125, "learning_rate": 2.2e-06, "entropy": 0.187082689255476, "cispo_clip_ratio": 0.0027922270237468183, "step_time": 199.77284794001025, "entropy_control/entropy": 0.1873935803470218, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.16, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 199.669119, "rollout_overhead_time": 0.0, "peak_allocated_gib": 70.00158309936523, "peak_reserved_gib": 77.8125, "reward": null }, { "step": 41, "loss": 0.02279401570558548, "grad_norm": 0.0556640625, "learning_rate": 2.0000000000000003e-06, "thinking/tokens_mean": 700.3630952380953, "thinking/budget_hit_fraction": 0.19642857142857142, "num_tokens": 1235159.0, "completions/mean_length": 5184.875, "completions/min_length": 2162.0, "completions/max_length": 7152.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 5184.875, "completions/min_terminated_length": 2162.0, "completions/max_terminated_length": 7152.0, "tools/call_frequency": 3.75, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__native/mean": 0.7096945643424988, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__native/std": 0.11446171998977661, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__efficiency/mean": 0.4166666865348816, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__efficiency/std": 0.3191424012184143, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/mean": 0.5528876781463623, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/std": 0.2096613198518753, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/mean": 0.3333333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/std": 0.2357022762298584, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/std": 1.0, "reward": 0.8000410795211792, "reward_std": 0.31318575143814087, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.20578624680638313, "cispo_clip_ratio": 0.0, "step_time": 1882.2396674270058, "entropy_control/entropy": 0.21090759849034313, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.164, "adapter_sync_time": 0.0, "generation_time": 1593.493517, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1642.967185, "forward_backward_time": 239.166147, "rollout_overhead_time": 49.473668, "peak_allocated_gib": 70.00158309936523, "peak_reserved_gib": 77.8125 }, { "step": 42, "loss": 0.023024503141641617, "grad_norm": 0.052001953125, "learning_rate": 1.8000000000000001e-06, "entropy": 0.20608649775385857, "cispo_clip_ratio": 0.002797896770061925, "step_time": 239.25387795001006, "entropy_control/entropy": 0.21126767682563613, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.168, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 239.152955, "rollout_overhead_time": 0.0, "peak_allocated_gib": 70.00158309936523, "peak_reserved_gib": 77.8125, "reward": null }, { "step": 43, "loss": 0.03825816512107849, "grad_norm": 0.07861328125, "learning_rate": 1.6000000000000001e-06, "thinking/tokens_mean": 572.6785714285714, "thinking/budget_hit_fraction": 0.07142857142857142, "num_tokens": 1292693.0, "completions/mean_length": 3887.75, "completions/min_length": 2811.0, "completions/max_length": 5795.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 3887.75, "completions/min_terminated_length": 2811.0, "completions/max_terminated_length": 5795.0, "tools/call_frequency": 3.25, "tools/failure_frequency": 0.03846153989434242, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__native/mean": 0.6662607192993164, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__native/std": 0.08975932747125626, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__efficiency/mean": 0.4583333730697632, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__efficiency/std": 0.3154948949813843, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/mean": 0.6714017987251282, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/std": 0.07128286361694336, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/std": 0.19245010614395142, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/std": 0.0, "reward": 0.8927896022796631, "reward_std": 0.06319531798362732, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.18599931336939335, "cispo_clip_ratio": 0.0, "step_time": 1635.5146522129944, "entropy_control/entropy": 0.18642948896211642, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.172, "adapter_sync_time": 0.0, "generation_time": 1353.151486, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1400.387877, "forward_backward_time": 235.021513, "rollout_overhead_time": 47.236391, "peak_allocated_gib": 70.00158309936523, "peak_reserved_gib": 77.8125 }, { "step": 44, "loss": 0.038326285779476166, "grad_norm": 0.07373046875, "learning_rate": 1.4000000000000001e-06, "entropy": 0.18594970554113388, "cispo_clip_ratio": 0.0018617352470755577, "step_time": 235.15527386898611, "entropy_control/entropy": 0.18642186386322818, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.176, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 235.049395, "rollout_overhead_time": 0.0, "peak_allocated_gib": 70.00158309936523, "peak_reserved_gib": 77.8125, "reward": null }, { "step": 45, "loss": 0.04289444908499718, "grad_norm": 0.0703125, "learning_rate": 1.2000000000000002e-06, "thinking/tokens_mean": 651.801530612245, "thinking/budget_hit_fraction": 0.08928571428571429, "num_tokens": 1363121.0, "completions/mean_length": 5133.5, "completions/min_length": 2004.0, "completions/max_length": 8072.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 5133.5, "completions/min_terminated_length": 2004.0, "completions/max_terminated_length": 8072.0, "tools/call_frequency": 4.375, "tools/failure_frequency": 0.05714285746216774, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/mean": 0.7988388538360596, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/std": 0.15891478955745697, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/mean": 0.25, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/std": 0.1666666567325592, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/mean": 0.7323082685470581, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/std": 0.09677081555128098, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/mean": 0.3750000298023224, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/std": 0.28463754057884216, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/std": 0.0, "reward": 0.9811986088752747, "reward_std": 0.12073472887277603, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.18510518595576286, "cispo_clip_ratio": 0.0, "step_time": 2391.487052959994, "entropy_control/entropy": 0.18973849059490716, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.18, "adapter_sync_time": 0.0, "generation_time": 2060.356795, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 2116.187572, "forward_backward_time": 275.197723, "rollout_overhead_time": 55.830777, "peak_allocated_gib": 70.00158309936523, "peak_reserved_gib": 77.8125 }, { "step": 46, "loss": 0.04262041300535202, "grad_norm": 0.059814453125, "learning_rate": 1.0000000000000002e-06, "entropy": 0.18503615446388721, "cispo_clip_ratio": 0.0022268831671681255, "step_time": 275.21273764799844, "entropy_control/entropy": 0.18960556835069395, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.184, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 275.111284, "rollout_overhead_time": 0.0, "peak_allocated_gib": 70.00158309936523, "peak_reserved_gib": 77.8125, "reward": null }, { "step": 47, "loss": 0.06568296998739243, "grad_norm": 0.05224609375, "learning_rate": 8.000000000000001e-07, "thinking/tokens_mean": 576.6964285714286, "thinking/budget_hit_fraction": 0.125, "num_tokens": 1420422.0, "completions/mean_length": 3876.125, "completions/min_length": 2954.0, "completions/max_length": 5240.0, "completions/clipped_ratio": 0.25, "completions/mean_terminated_length": 4144.5, "completions/min_terminated_length": 2954.0, "completions/max_terminated_length": 5240.0, "tools/call_frequency": 2.5, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/mean": 0.4907289147377014, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/std": 0.33322080969810486, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/mean": 0.4166666865348816, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/std": 0.3191424012184143, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/std": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/mean": 0.5091619491577148, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/std": 0.4585621654987335, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/mean": 0.25, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/std": 0.3191424012184143, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/mean": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/std": 1.154700517654419, "reward": 0.5666121244430542, "reward_std": 0.5813743472099304, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 0.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.16945160552859306, "cispo_clip_ratio": 0.0, "step_time": 1268.3584206089945, "entropy_control/entropy": 0.22861190022942993, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.188, "adapter_sync_time": 0.0, "generation_time": 1037.530062, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1076.969644, "forward_backward_time": 191.278391, "rollout_overhead_time": 39.439582, "peak_allocated_gib": 70.00158309936523, "peak_reserved_gib": 77.8125 }, { "step": 48, "loss": 0.06532293558120728, "grad_norm": 0.05126953125, "learning_rate": 6.000000000000001e-07, "entropy": 0.16923405788838863, "cispo_clip_ratio": 0.0031085670343600214, "step_time": 191.3939335570103, "entropy_control/entropy": 0.22838055826677772, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.192, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 191.284028, "rollout_overhead_time": 0.0, "peak_allocated_gib": 70.00158309936523, "peak_reserved_gib": 77.8125, "reward": null }, { "step": 49, "loss": 0.010915946215391159, "grad_norm": 0.052978515625, "learning_rate": 4.0000000000000003e-07, "thinking/tokens_mean": 725.46875, "thinking/budget_hit_fraction": 0.1875, "num_tokens": 1472574.0, "completions/mean_length": 3228.0, "completions/min_length": 1529.0, "completions/max_length": 4703.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 3228.0, "completions/min_terminated_length": 1529.0, "completions/max_terminated_length": 4703.0, "tools/call_frequency": 2.125, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/mean": 0.6088261604309082, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/mean": 0.6666666865348816, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/mean": 0.8257564306259155, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/mean": 0.625, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/std": 0.0833333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/std": 0.0, "reward": 0.9495829343795776, "reward_std": 0.11487291008234024, "frac_reward_zero_std": 0.5, "frac_groups_zero_std": 0.5, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.20170070603489876, "cispo_clip_ratio": 0.0, "step_time": 923.0098794490114, "entropy_control/entropy": 0.2068337830312001, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.196, "adapter_sync_time": 0.0, "generation_time": 790.636338, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 814.570624, "forward_backward_time": 108.329651, "rollout_overhead_time": 23.934286, "peak_allocated_gib": 70.00158309936523, "peak_reserved_gib": 77.822265625 }, { "step": 50, "loss": 0.01088174432516098, "grad_norm": 0.05517578125, "learning_rate": 2.0000000000000002e-07, "entropy": 0.20143523439764977, "cispo_clip_ratio": 0.0016747722693253309, "step_time": 108.43975785301882, "entropy_control/entropy": 0.20652793239727366, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.2, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 108.34081, "rollout_overhead_time": 0.0, "peak_allocated_gib": 70.00158309936523, "peak_reserved_gib": 77.822265625, "reward": null } ], "holdout": [ { "seed": 900000, "sample": 0, "native_reward": 0.6089882858614486, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 3348, "reward_channels": { "native": 0.6089882858614486, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "two_pool_cycle" }, { "seed": 900000, "sample": 1, "native_reward": 0.6089882858614486, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 1698, "reward_channels": { "native": 0.6089882858614486, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "two_pool_cycle" }, { "seed": 900000, "sample": 2, "native_reward": 0.6089882858614486, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 3442, "reward_channels": { "native": 0.6089882858614486, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "two_pool_cycle" }, { "seed": 900000, "sample": 3, "native_reward": 0.0, "promotion_success": false, "done": false, "errors": 0, "tool_steps": 0, "model_tokens": 2274, "reward_channels": { "native": 0.0, "efficiency": 0.0, "reliability": -1.0 }, "verification_level": null, "failed_gates": null, "family": "two_pool_cycle" }, { "seed": 900001, "sample": 0, "native_reward": 0.6093825392865522, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 2420, "reward_channels": { "native": 0.6093825392865522, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "two_pool_cycle" }, { "seed": 900001, "sample": 1, "native_reward": 0.6093825392865522, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 3006, "reward_channels": { "native": 0.6093825392865522, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "two_pool_cycle" }, { "seed": 900001, "sample": 2, "native_reward": 0.8523284576243806, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 3, "model_tokens": 4574, "reward_channels": { "native": 0.8523284576243806, "efficiency": 0.5, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "two_pool_cycle" }, { "seed": 900001, "sample": 3, "native_reward": 0.6093825392865522, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 2625, "reward_channels": { "native": 0.6093825392865522, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "two_pool_cycle" }, { "seed": 900000, "sample": 0, "native_reward": 0.6106601460847819, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 1448, "reward_channels": { "native": 0.6106601460847819, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "triangular_cycle" }, { "seed": 900000, "sample": 1, "native_reward": 0.0, "promotion_success": false, "done": false, "errors": 0, "tool_steps": 0, "model_tokens": 0, "reward_channels": { "native": 0.0, "efficiency": 0.0, "reliability": -1.0 }, "verification_level": null, "failed_gates": null, "family": "triangular_cycle" }, { "seed": 900000, "sample": 2, "native_reward": 0.25, "promotion_success": false, "done": false, "errors": 0, "tool_steps": 5, "model_tokens": 6009, "reward_channels": { "native": 0.25, "efficiency": 0.0, "reliability": -1.0 }, "verification_level": null, "failed_gates": [ "finalized" ], "family": "triangular_cycle" }, { "seed": 900000, "sample": 3, "native_reward": 0.8204653155141224, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 3, "model_tokens": 4220, "reward_channels": { "native": 0.8204653155141224, "efficiency": 0.5, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "triangular_cycle" }, { "seed": 900001, "sample": 0, "native_reward": 0.6108851248150757, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 2719, "reward_channels": { "native": 0.6108851248150757, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "triangular_cycle" }, { "seed": 900001, "sample": 1, "native_reward": 0.7060268919310104, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 3, "model_tokens": 2680, "reward_channels": { "native": 0.7060268919310104, "efficiency": 0.5, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "triangular_cycle" }, { "seed": 900001, "sample": 2, "native_reward": 0.8541365354407209, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 4571, "reward_channels": { "native": 0.8541365354407209, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "triangular_cycle" }, { "seed": 900001, "sample": 3, "native_reward": 0.811786874720879, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 5487, "reward_channels": { "native": 0.811786874720879, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "triangular_cycle" }, { "seed": 900000, "sample": 0, "native_reward": 0.8720485688510347, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 6363, "reward_channels": { "native": 0.8720485688510347, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "competing_cycles" }, { "seed": 900000, "sample": 1, "native_reward": 0.8720485688510347, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 3, "model_tokens": 4101, "reward_channels": { "native": 0.8720485688510347, "efficiency": 0.5, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "competing_cycles" }, { "seed": 900000, "sample": 2, "native_reward": 0.6502053079500402, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 3, "model_tokens": 6425, "reward_channels": { "native": 0.6502053079500402, "efficiency": 0.5, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "competing_cycles" }, { "seed": 900000, "sample": 3, "native_reward": 0.5875732130670761, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 8098, "reward_channels": { "native": 0.5875732130670761, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "competing_cycles" }, { "seed": 900001, "sample": 0, "native_reward": 0.175, "promotion_success": false, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 8758, "reward_channels": { "native": 0.175, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [ "positive_net_profit", "stress_positive", "task_complete" ], "family": "competing_cycles" }, { "seed": 900001, "sample": 1, "native_reward": 0.6145108982971254, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 4275, "reward_channels": { "native": 0.6145108982971254, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "competing_cycles" }, { "seed": 900001, "sample": 2, "native_reward": 0.6145108982971254, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 3, "model_tokens": 5657, "reward_channels": { "native": 0.6145108982971254, "efficiency": 0.5, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "competing_cycles" }, { "seed": 900001, "sample": 3, "native_reward": 0.7187253233531886, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 5633, "reward_channels": { "native": 0.7187253233531886, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "competing_cycles" }, { "seed": 900000, "sample": 0, "native_reward": 0.609376392538787, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 1860, "reward_channels": { "native": 0.609376392538787, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "four_hop_cycle" }, { "seed": 900000, "sample": 1, "native_reward": 0.609376392538787, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 2708, "reward_channels": { "native": 0.609376392538787, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "four_hop_cycle" }, { "seed": 900000, "sample": 2, "native_reward": 0.7503548418171574, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 3, "model_tokens": 2149, "reward_channels": { "native": 0.7503548418171574, "efficiency": 0.5, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "four_hop_cycle" }, { "seed": 900000, "sample": 3, "native_reward": 0.609376392538787, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 2085, "reward_channels": { "native": 0.609376392538787, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "four_hop_cycle" }, { "seed": 900001, "sample": 0, "native_reward": 0.6095701077161815, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 3, "model_tokens": 2984, "reward_channels": { "native": 0.6095701077161815, "efficiency": 0.5, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "four_hop_cycle" }, { "seed": 900001, "sample": 1, "native_reward": 0.6095701077161815, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 2433, "reward_channels": { "native": 0.6095701077161815, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "four_hop_cycle" }, { "seed": 900001, "sample": 2, "native_reward": 0.6095701077161815, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 1642, "reward_channels": { "native": 0.6095701077161815, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "four_hop_cycle" }, { "seed": 900001, "sample": 3, "native_reward": 0.25, "promotion_success": false, "done": false, "errors": 0, "tool_steps": 6, "model_tokens": 6308, "reward_channels": { "native": 0.25, "efficiency": 0.0, "reliability": -1.0 }, "verification_level": null, "failed_gates": [ "finalized" ], "family": "four_hop_cycle" }, { "seed": 900000, "sample": 0, "native_reward": 0.962057681977345, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 3, "model_tokens": 2907, "reward_channels": { "native": 0.962057681977345, "efficiency": 0.5, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "size_sensitive_cycle" }, { "seed": 900000, "sample": 1, "native_reward": 0.8262556477476042, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 2561, "reward_channels": { "native": 0.8262556477476042, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "size_sensitive_cycle" }, { "seed": 900000, "sample": 2, "native_reward": 0.25, "promotion_success": false, "done": false, "errors": 0, "tool_steps": 6, "model_tokens": 5922, "reward_channels": { "native": 0.25, "efficiency": 0.0, "reliability": -1.0 }, "verification_level": null, "failed_gates": [ "finalized" ], "family": "size_sensitive_cycle" }, { "seed": 900000, "sample": 3, "native_reward": 0.8262556477476042, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 3879, "reward_channels": { "native": 0.8262556477476042, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "size_sensitive_cycle" }, { "seed": 900001, "sample": 0, "native_reward": 0.8245803237223959, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 3, "model_tokens": 5265, "reward_channels": { "native": 0.8245803237223959, "efficiency": 0.5, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "size_sensitive_cycle" }, { "seed": 900001, "sample": 1, "native_reward": 0.0, "promotion_success": false, "done": false, "errors": 0, "tool_steps": 0, "model_tokens": 0, "reward_channels": { "native": 0.0, "efficiency": 0.0, "reliability": -1.0 }, "verification_level": null, "failed_gates": null, "family": "size_sensitive_cycle" }, { "seed": 900001, "sample": 2, "native_reward": 0.8245803237223959, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 2790, "reward_channels": { "native": 0.8245803237223959, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "size_sensitive_cycle" }, { "seed": 900001, "sample": 3, "native_reward": 0.8245803237223959, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 5800, "reward_channels": { "native": 0.8245803237223959, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "size_sensitive_cycle" } ] }, { "run_id": "1789312258", "mode": "thinking", "method": "adapo", "steps": 50, "reward_points": 25, "train_mean": 0.8184732961654663, "train_last10": 0.8814916849136353, "holdout_native": 0.62497194805629, "holdout_episodes": 40, "holdout_successes": 34, "mean_step_seconds": 887.1718965091598, "label": "DAPO + ADAPO", "state_at_cutoff": "complete", "step_hours": 12.321831895960553, "peak_allocated_gib": 71.76706552505493, "peak_reserved_gib": 79.62890625, "extra_refill_rounds": 0.0, "groups_refilled": 0.0, "holdout_component_means": { "native": 0.62497194805629, "efficiency": 0.4666666666666667, "reliability": 0.75 }, "configuration": { "model": { "id": "Qwen/Qwen3-14B", "revision": "40c069824f4251a91eefaf281ebe4c544efd3e18", "loss": "dapo", "algorithm": "grpo", "no_think": false, "thinking_budget_tokens": 2048, "lora": true, "quantize": true, "lora_r": 32, "lora_alpha": 64, "use_dora": false }, "loss": "dapo", "normalization": "joint", "entropy_control": { "method": "adapo", "epsilon_low": 0.2, "epsilon_high": 0.28, "zeta_initial": 0.001, "zeta_min": 0.0001, "zeta_max": 0.05, "holdout_split": "benchmark", "holdout_seeds_per_family": 2, "num_iterations": 2 }, "dynamic_sampling": false, "sampling": { "group_size": 4, "temperature": 1.0, "max_completion_tokens": 22528, "top_entropy_quantile": 1.0, "temperature_decay_to": null }, "learning_rate": 1e-05, "turn_max_tokens": 3072 }, "source_sha256": { "metrics.jsonl": "dda0c6a01e93f1c0a7dc456724f5163caa61bb27098b308e5789d53649401a11", "entropy_holdout_episodes.jsonl": "fa9bf442b10ade0aee6c14bd1457868244b3c6aaaa2abf139febaa0dacf51a5f", "resolved_training_config.json": "a67d445b7631e9af0b4c7d78adf37bccb739a86923cb099854980fc46d5202ab" }, "training_steps": [ { "step": 1, "loss": -0.32585203647613525, "grad_norm": 0.061279296875, "learning_rate": 1e-05, "thinking/tokens_mean": 606.2361111111111, "thinking/budget_hit_fraction": 0.14583333333333334, "num_tokens": 51975.0, "completions/mean_length": 3604.375, "completions/min_length": 2890.0, "completions/max_length": 6014.0, "completions/clipped_ratio": 0.25, "completions/mean_terminated_length": 3782.166748046875, "completions/min_terminated_length": 2890.0, "completions/max_terminated_length": 6014.0, "tools/call_frequency": 2.25, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.7262155413627625, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 0.13481803238391876, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.4583333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.2500000298023224, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/mean": 0.45231834053993225, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/std": 0.526568591594696, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/mean": 0.2916666865348816, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/std": 0.34359216690063477, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/mean": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/std": 1.154700517654419, "reward": 0.7080169916152954, "reward_std": 0.5734051465988159, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 0.5, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.14539151638746262, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 1387.7308625300004, "entropy_control/entropy": 0.20009788357278327, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.004, "adapter_sync_time": 0.0, "generation_time": 1195.690378, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1230.57605, "forward_backward_time": 157.037873, "rollout_overhead_time": 34.885672, "peak_allocated_gib": 46.211081981658936, "peak_reserved_gib": 51.38671875 }, { "step": 2, "loss": -0.32616713643074036, "grad_norm": 0.0634765625, "learning_rate": 9.800000000000001e-06, "entropy": 0.14562873356044292, "clip_ratio/low_mean": 0.0007928965496830642, "clip_ratio/low_min": 0.0007928965496830642, "clip_ratio/high_mean": 0.0009503347682766616, "clip_ratio/high_max": 0.0009503347682766616, "clip_ratio/region_mean": 0.0017432313179597259, "step_time": 157.09996648500055, "entropy_control/entropy": 0.2004543321347234, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.266, "epoch": 0.008, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 156.985637, "rollout_overhead_time": 0.0, "peak_allocated_gib": 46.689598083496094, "peak_reserved_gib": 51.484375, "reward": null }, { "step": 3, "loss": -0.22225600481033325, "grad_norm": 0.064453125, "learning_rate": 9.600000000000001e-06, "thinking/tokens_mean": 519.0739795918367, "thinking/budget_hit_fraction": 0.13010204081632654, "num_tokens": 102698.0, "completions/mean_length": 3628.375, "completions/min_length": 2853.0, "completions/max_length": 6269.0, "completions/clipped_ratio": 0.125, "completions/mean_terminated_length": 3708.000244140625, "completions/min_terminated_length": 2853.0, "completions/max_terminated_length": 6269.0, "tools/call_frequency": 2.5, "tools/failure_frequency": 0.10000000149011612, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.5910395383834839, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 0.26091644167900085, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.2357022762298584, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 0.75, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 0.7071067690849304, "reward": 0.7660395503044128, "reward_std": 0.4036981463432312, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 0.5, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.1909853182733059, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 1628.453406299, "entropy_control/entropy": 0.2153525839184378, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.266, "entropy_control/epsilon_high_next": 0.2527, "epoch": 0.012, "adapter_sync_time": 0.0, "generation_time": 1453.213702, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1484.423931, "forward_backward_time": 143.911268, "rollout_overhead_time": 31.210229, "peak_allocated_gib": 47.29379320144653, "peak_reserved_gib": 52.34375 }, { "step": 4, "loss": -0.22221553325653076, "grad_norm": 0.052490234375, "learning_rate": 9.4e-06, "entropy": 0.19090092182159424, "clip_ratio/low_mean": 0.00041991088073700666, "clip_ratio/low_min": 0.00041991088073700666, "clip_ratio/high_mean": 0.0013236310769570991, "clip_ratio/high_max": 0.0013236310769570991, "clip_ratio/region_mean": 0.0017435419576941058, "step_time": 144.02420915199946, "entropy_control/entropy": 0.21517221245654633, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2527, "entropy_control/epsilon_high_next": 0.24006499999999997, "epoch": 0.016, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 143.909696, "rollout_overhead_time": 0.0, "peak_allocated_gib": 47.29379320144653, "peak_reserved_gib": 52.34375, "reward": null }, { "step": 5, "loss": -0.07189202308654785, "grad_norm": 0.06787109375, "learning_rate": 9.200000000000002e-06, "thinking/tokens_mean": 691.5464285714286, "thinking/budget_hit_fraction": 0.20357142857142857, "num_tokens": 152094.0, "completions/mean_length": 3469.5, "completions/min_length": 2229.0, "completions/max_length": 5309.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 3469.5, "completions/min_terminated_length": 2229.0, "completions/max_terminated_length": 5309.0, "tools/call_frequency": 2.125, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.5587122440338135, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 0.24098125100135803, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.5208333730697632, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.24295634031295776, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 0.75, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 0.7071067690849304, "reward": 0.7347538471221924, "reward_std": 0.38595905900001526, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 0.5, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.2072792425751686, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 1217.8728956439995, "entropy_control/entropy": 0.20896115617756297, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.24006499999999997, "entropy_control/epsilon_high_next": 0.22806174999999995, "epoch": 0.02, "adapter_sync_time": 0.0, "generation_time": 1068.010088, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1094.917518, "forward_backward_time": 122.849557, "rollout_overhead_time": 26.907429, "peak_allocated_gib": 47.29379320144653, "peak_reserved_gib": 52.361328125 }, { "step": 6, "loss": -0.07212880253791809, "grad_norm": 0.06396484375, "learning_rate": 9e-06, "entropy": 0.20731810480356216, "clip_ratio/low_mean": 0.0014326409436762333, "clip_ratio/low_min": 0.0014326409436762333, "clip_ratio/high_mean": 0.0015884223394095898, "clip_ratio/high_max": 0.0015884223394095898, "clip_ratio/region_mean": 0.003021063283085823, "step_time": 122.98436585900163, "entropy_control/entropy": 0.2088881614850532, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.22806174999999995, "entropy_control/epsilon_high_next": 0.21665866249999993, "epoch": 0.024, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 122.869361, "rollout_overhead_time": 0.0, "peak_allocated_gib": 47.29379320144653, "peak_reserved_gib": 52.361328125, "reward": null }, { "step": 7, "loss": -0.07044494152069092, "grad_norm": 0.06103515625, "learning_rate": 8.8e-06, "thinking/tokens_mean": 633.5765306122448, "thinking/budget_hit_fraction": 0.13010204081632654, "num_tokens": 215292.0, "completions/mean_length": 4432.25, "completions/min_length": 2256.0, "completions/max_length": 5572.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 4432.25, "completions/min_terminated_length": 2256.0, "completions/max_terminated_length": 5572.0, "tools/call_frequency": 3.5, "tools/failure_frequency": 0.0357142873108387, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/mean": 0.4606367349624634, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/std": 0.3070911467075348, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/mean": 0.3333333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/std": 0.2357022762298584, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/std": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/mean": 0.7498937845230103, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/std": 0.3393867611885071, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/mean": 0.2916666865348816, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/std": 0.20971763134002686, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/std": 1.0, "reward": 0.7208902835845947, "reward_std": 0.5163019299507141, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 0.5, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.20314441993832588, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 1788.0677400999984, "entropy_control/entropy": 0.20750716747612813, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.21665866249999993, "entropy_control/epsilon_high_next": 0.20582572937499993, "epoch": 0.028, "adapter_sync_time": 0.0, "generation_time": 1500.15402, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1548.457852, "forward_backward_time": 239.495424, "rollout_overhead_time": 48.303831, "peak_allocated_gib": 57.35714864730835, "peak_reserved_gib": 63.630859375 }, { "step": 8, "loss": -0.07007166743278503, "grad_norm": 0.059814453125, "learning_rate": 8.6e-06, "entropy": 0.20309989154338837, "clip_ratio/low_mean": 0.0004893832956440747, "clip_ratio/low_min": 0.0004893832956440747, "clip_ratio/high_mean": 0.003914013446774334, "clip_ratio/high_max": 0.003914013446774334, "clip_ratio/region_mean": 0.004403396742418408, "step_time": 239.61313570199945, "entropy_control/entropy": 0.20740975057343733, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.20582572937499993, "entropy_control/epsilon_high_next": 0.2, "epoch": 0.032, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 239.502459, "rollout_overhead_time": 0.0, "peak_allocated_gib": 57.35714864730835, "peak_reserved_gib": 63.630859375, "reward": null }, { "step": 9, "loss": -0.1880415380001068, "grad_norm": 0.05224609375, "learning_rate": 8.400000000000001e-06, "thinking/tokens_mean": 653.8630952380953, "thinking/budget_hit_fraction": 0.16581632653061223, "num_tokens": 278150.0, "completions/mean_length": 4573.25, "completions/min_length": 2480.0, "completions/max_length": 8169.0, "completions/clipped_ratio": 0.125, "completions/mean_terminated_length": 4787.857421875, "completions/min_terminated_length": 2480.0, "completions/max_terminated_length": 8169.0, "tools/call_frequency": 3.125, "tools/failure_frequency": 0.07999999821186066, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.4571598172187805, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 0.304773211479187, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.3333333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/mean": 0.8117247819900513, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/std": 0.12323763966560364, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/mean": 0.2916666865348816, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/std": 0.25, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/std": 0.0, "reward": 0.8042340278625488, "reward_std": 0.4230191707611084, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 0.5, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.1853284016251564, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 2218.8322648449966, "entropy_control/entropy": 0.2157486272186285, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "epoch": 0.036, "adapter_sync_time": 0.0, "generation_time": 1826.641978, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1892.237148, "forward_backward_time": 326.47602, "rollout_overhead_time": 65.59517, "peak_allocated_gib": 71.76706552505493, "peak_reserved_gib": 79.626953125 }, { "step": 10, "loss": -0.18819811940193176, "grad_norm": 0.051513671875, "learning_rate": 8.2e-06, "entropy": 0.1851575542241335, "clip_ratio/low_mean": 0.0006237653142306954, "clip_ratio/low_min": 0.0006237653142306954, "clip_ratio/high_mean": 0.003308254206785932, "clip_ratio/high_max": 0.003308254206785932, "clip_ratio/region_mean": 0.0039320195210166276, "step_time": 326.57490212200173, "entropy_control/entropy": 0.2156876346170331, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "epoch": 0.04, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 326.455564, "rollout_overhead_time": 0.0, "peak_allocated_gib": 71.76706552505493, "peak_reserved_gib": 79.626953125, "reward": null }, { "step": 11, "loss": 0.10778550058603287, "grad_norm": 0.052734375, "learning_rate": 8.000000000000001e-06, "thinking/tokens_mean": 615.225, "thinking/budget_hit_fraction": 0.1357142857142857, "num_tokens": 335818.0, "completions/mean_length": 3899.5, "completions/min_length": 1089.0, "completions/max_length": 9936.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 3899.5, "completions/min_terminated_length": 1089.0, "completions/max_terminated_length": 9936.0, "tools/call_frequency": 3.25, "tools/failure_frequency": 0.07692307978868484, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__native/mean": 0.6111614108085632, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__native/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__efficiency/mean": 0.6666666865348816, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__efficiency/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/mean": 0.4107462167739868, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/std": 0.2309536188840866, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/mean": 0.3333333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/std": 0.3042903244495392, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/std": 1.0, "reward": 0.6859537959098816, "reward_std": 0.30306175351142883, "frac_reward_zero_std": 0.5, "frac_groups_zero_std": 0.5, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.2010525595396757, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 2242.83903173, "entropy_control/entropy": 0.19583375305031564, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.21000000000000002, "epoch": 0.044, "adapter_sync_time": 0.0, "generation_time": 1923.133504, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1977.342599, "forward_backward_time": 265.39298, "rollout_overhead_time": 54.209095, "peak_allocated_gib": 71.76706552505493, "peak_reserved_gib": 79.626953125 }, { "step": 12, "loss": 0.10798507183790207, "grad_norm": 0.049560546875, "learning_rate": 7.800000000000002e-06, "entropy": 0.20115671306848526, "clip_ratio/low_mean": 0.0006724394479533657, "clip_ratio/low_min": 0.0006724394479533657, "clip_ratio/high_mean": 0.0015257851337082684, "clip_ratio/high_max": 0.0015257851337082684, "clip_ratio/region_mean": 0.002198224581661634, "step_time": 265.50838019799994, "entropy_control/entropy": 0.19599206063965324, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.21000000000000002, "entropy_control/epsilon_high_next": 0.22050000000000003, "epoch": 0.048, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 265.400663, "rollout_overhead_time": 0.0, "peak_allocated_gib": 71.76706552505493, "peak_reserved_gib": 79.626953125, "reward": null }, { "step": 13, "loss": -0.12117783725261688, "grad_norm": 0.06884765625, "learning_rate": 7.600000000000001e-06, "thinking/tokens_mean": 556.0612244897959, "thinking/budget_hit_fraction": 0.14285714285714285, "num_tokens": 394460.0, "completions/mean_length": 4045.25, "completions/min_length": 1176.0, "completions/max_length": 6310.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 4045.25, "completions/min_terminated_length": 1176.0, "completions/max_terminated_length": 6310.0, "tools/call_frequency": 3.625, "tools/failure_frequency": 0.03448275849223137, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.725550651550293, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 0.1339954286813736, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.4583333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.3154948949813843, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/mean": 0.704129695892334, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/std": 0.1182098463177681, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/mean": 0.375, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/std": 0.25, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/std": 0.0, "reward": 0.9356735348701477, "reward_std": 0.10756239295005798, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.19102087058126926, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 1917.1351157499976, "entropy_control/entropy": 0.20026604458137717, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.22050000000000003, "entropy_control/epsilon_high_next": 0.20947500000000002, "epoch": 0.052, "adapter_sync_time": 0.0, "generation_time": 1620.659739, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1671.291614, "forward_backward_time": 245.733358, "rollout_overhead_time": 50.631875, "peak_allocated_gib": 71.76706552505493, "peak_reserved_gib": 79.626953125 }, { "step": 14, "loss": -0.12116202712059021, "grad_norm": 0.0625, "learning_rate": 7.4e-06, "entropy": 0.19114640541374683, "clip_ratio/low_mean": 0.0014320093032438308, "clip_ratio/low_min": 0.0014320093032438308, "clip_ratio/high_mean": 0.002373186231125146, "clip_ratio/high_max": 0.002373186231125146, "clip_ratio/region_mean": 0.003805195534368977, "step_time": 245.83355621399642, "entropy_control/entropy": 0.20042587934555045, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.20947500000000002, "entropy_control/epsilon_high_next": 0.2, "epoch": 0.056, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 245.73045, "rollout_overhead_time": 0.0, "peak_allocated_gib": 71.76706552505493, "peak_reserved_gib": 79.626953125, "reward": null }, { "step": 15, "loss": -0.24541530013084412, "grad_norm": 0.06298828125, "learning_rate": 7.2000000000000005e-06, "thinking/tokens_mean": 580.2142857142857, "thinking/budget_hit_fraction": 0.1369047619047619, "num_tokens": 451538.0, "completions/mean_length": 3651.25, "completions/min_length": 1024.0, "completions/max_length": 6859.0, "completions/clipped_ratio": 0.125, "completions/mean_terminated_length": 4026.571533203125, "completions/min_terminated_length": 2063.0, "completions/max_terminated_length": 6859.0, "tools/call_frequency": 2.375, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/mean": 0.32948943972587585, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/std": 0.38222140073776245, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/mean": 0.25, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/std": 0.3191424012184143, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/mean": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/std": 1.154700517654419, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/mean": 0.7000407576560974, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/std": 0.3576284945011139, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/mean": 0.4583333730697632, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/std": 0.3154948949813843, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/std": 0.0, "reward": 0.6324734687805176, "reward_std": 0.5627145767211914, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 0.5, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.18075690045952797, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 1590.8704901529982, "entropy_control/entropy": 0.201307438755561, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "epoch": 0.06, "adapter_sync_time": 0.0, "generation_time": 1300.560626, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1349.674936, "forward_backward_time": 241.091042, "rollout_overhead_time": 49.11431, "peak_allocated_gib": 71.76706552505493, "peak_reserved_gib": 79.626953125 }, { "step": 16, "loss": -0.24518853425979614, "grad_norm": 0.059814453125, "learning_rate": 7e-06, "entropy": 0.1812347825616598, "clip_ratio/low_mean": 0.0004967123823007569, "clip_ratio/low_min": 0.0004967123823007569, "clip_ratio/high_mean": 0.002570429176557809, "clip_ratio/high_max": 0.002570429176557809, "clip_ratio/region_mean": 0.003067141558858566, "step_time": 241.17334455400123, "entropy_control/entropy": 0.2017987437893315, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "epoch": 0.064, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 241.068307, "rollout_overhead_time": 0.0, "peak_allocated_gib": 71.76706552505493, "peak_reserved_gib": 79.626953125, "reward": null }, { "step": 17, "loss": -0.5237339735031128, "grad_norm": 0.061279296875, "learning_rate": 6.800000000000001e-06, "thinking/tokens_mean": 544.2440476190476, "thinking/budget_hit_fraction": 0.13095238095238096, "num_tokens": 512227.0, "completions/mean_length": 3895.125, "completions/min_length": 1265.0, "completions/max_length": 6062.0, "completions/clipped_ratio": 0.25, "completions/mean_terminated_length": 4169.83349609375, "completions/min_terminated_length": 1265.0, "completions/max_terminated_length": 6062.0, "tools/call_frequency": 3.0, "tools/failure_frequency": 0.0416666679084301, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/mean": 0.3977721333503723, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/std": 0.46304839849472046, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/mean": 0.1666666716337204, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/std": 0.235702246427536, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/mean": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/std": 1.154700517654419, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/mean": 0.7884960174560547, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/std": 0.20735859870910645, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/mean": 0.375, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/std": 0.34359216690063477, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/std": 0.0, "reward": 0.7066757678985596, "reward_std": 0.5743606090545654, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 0.5, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.13210453558713198, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 1761.454811356005, "entropy_control/entropy": 0.17669977463671643, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.21000000000000002, "epoch": 0.068, "adapter_sync_time": 0.0, "generation_time": 1418.497371, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1476.009611, "forward_backward_time": 285.340178, "rollout_overhead_time": 57.51224, "peak_allocated_gib": 71.76706552505493, "peak_reserved_gib": 79.626953125 }, { "step": 18, "loss": -0.5241345167160034, "grad_norm": 0.056884765625, "learning_rate": 6.600000000000001e-06, "entropy": 0.13225947506725788, "clip_ratio/low_mean": 0.0009709621663205326, "clip_ratio/low_min": 0.0009709621663205326, "clip_ratio/high_mean": 0.0020131196943111718, "clip_ratio/high_max": 0.0020131196943111718, "clip_ratio/region_mean": 0.0029840818606317043, "step_time": 285.45596249799564, "entropy_control/entropy": 0.17665275834403102, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.21000000000000002, "entropy_control/epsilon_high_next": 0.22050000000000003, "epoch": 0.072, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 285.352041, "rollout_overhead_time": 0.0, "peak_allocated_gib": 71.76706552505493, "peak_reserved_gib": 79.626953125, "reward": null }, { "step": 19, "loss": 0.11047493666410446, "grad_norm": 0.04833984375, "learning_rate": 6.4000000000000006e-06, "thinking/tokens_mean": 675.4285714285714, "thinking/budget_hit_fraction": 0.14285714285714285, "num_tokens": 571537.0, "completions/mean_length": 4523.25, "completions/min_length": 1649.0, "completions/max_length": 7480.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 4523.25, "completions/min_terminated_length": 1649.0, "completions/max_terminated_length": 7480.0, "tools/call_frequency": 3.5, "tools/failure_frequency": 0.0357142873108387, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.6096870303153992, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.6666666865348816, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/mean": 0.5707845687866211, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/std": 0.3742716312408447, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/mean": 0.2083333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/std": 0.31549492478370667, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/mean": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/std": 1.154700517654419, "reward": 0.7121108174324036, "reward_std": 0.4261879026889801, "frac_reward_zero_std": 0.5, "frac_groups_zero_std": 0.5, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.1934636514633894, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 1590.5798178500045, "entropy_control/entropy": 0.1948017890068756, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.22050000000000003, "entropy_control/epsilon_high_next": 0.23152500000000004, "epoch": 0.076, "adapter_sync_time": 0.0, "generation_time": 1340.513125, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1384.070744, "forward_backward_time": 206.407363, "rollout_overhead_time": 43.557619, "peak_allocated_gib": 71.76706552505493, "peak_reserved_gib": 79.626953125 }, { "step": 20, "loss": 0.11036525666713715, "grad_norm": 0.043701171875, "learning_rate": 6.200000000000001e-06, "entropy": 0.19397841580212116, "clip_ratio/low_mean": 0.00043536111479625106, "clip_ratio/low_min": 0.00043536111479625106, "clip_ratio/high_mean": 0.0008276761218439788, "clip_ratio/high_max": 0.0008276761218439788, "clip_ratio/region_mean": 0.0012630372366402298, "step_time": 206.51005599998825, "entropy_control/entropy": 0.19524820448367877, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.23152500000000004, "entropy_control/epsilon_high_next": 0.24310125000000005, "epoch": 0.08, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 206.406227, "rollout_overhead_time": 0.0, "peak_allocated_gib": 71.76706552505493, "peak_reserved_gib": 79.626953125, "reward": null }, { "step": 21, "loss": -0.38374578952789307, "grad_norm": 0.057861328125, "learning_rate": 6e-06, "thinking/tokens_mean": 689.5102040816327, "thinking/budget_hit_fraction": 0.08928571428571429, "num_tokens": 632020.0, "completions/mean_length": 3861.375, "completions/min_length": 1494.0, "completions/max_length": 5863.0, "completions/clipped_ratio": 0.125, "completions/mean_terminated_length": 3974.285888671875, "completions/min_terminated_length": 1494.0, "completions/max_terminated_length": 5863.0, "tools/call_frequency": 2.625, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/mean": 0.47962427139282227, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/std": 0.32456180453300476, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/mean": 0.4166666865348816, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/std": 0.2886751592159271, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/std": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/mean": 0.6656705737113953, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/std": 0.11250734329223633, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/mean": 0.4583333730697632, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/std": 0.3154948949813843, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/std": 0.0, "reward": 0.7445224523544312, "reward_std": 0.3886309862136841, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 0.5, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.17417041957378387, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 1799.4436993359996, "entropy_control/entropy": 0.20476358123306698, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.24310125000000005, "entropy_control/epsilon_high_next": 0.23094618750000004, "epoch": 0.084, "adapter_sync_time": 0.0, "generation_time": 1468.484256, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1523.710427, "forward_backward_time": 275.628949, "rollout_overhead_time": 55.226171, "peak_allocated_gib": 71.76706552505493, "peak_reserved_gib": 79.626953125 }, { "step": 22, "loss": -0.38404008746147156, "grad_norm": 0.060546875, "learning_rate": 5.8e-06, "entropy": 0.17441867664456367, "clip_ratio/low_mean": 0.001117232270189561, "clip_ratio/low_min": 0.001117232270189561, "clip_ratio/high_mean": 0.0018263663514517248, "clip_ratio/high_max": 0.0018263663514517248, "clip_ratio/region_mean": 0.0029435986216412857, "step_time": 275.7976030539976, "entropy_control/entropy": 0.2047466029763731, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.23094618750000004, "entropy_control/epsilon_high_next": 0.21939887812500003, "epoch": 0.088, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 275.694136, "rollout_overhead_time": 0.0, "peak_allocated_gib": 71.76706552505493, "peak_reserved_gib": 79.626953125, "reward": null }, { "step": 23, "loss": -0.017500929534435272, "grad_norm": 0.07080078125, "learning_rate": 5.600000000000001e-06, "thinking/tokens_mean": 857.5, "thinking/budget_hit_fraction": 0.125, "num_tokens": 684836.0, "completions/mean_length": 3716.5, "completions/min_length": 2290.0, "completions/max_length": 5510.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 3716.5, "completions/min_terminated_length": 2290.0, "completions/max_terminated_length": 5510.0, "tools/call_frequency": 2.5, "tools/failure_frequency": 0.05000000074505806, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.6404144763946533, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 0.04312657564878464, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.5833333730697632, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.09622505307197571, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/mean": 0.662826657295227, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/std": 0.3252177834510803, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/mean": 0.625, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/std": 0.0833333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/std": 0.0, "reward": 0.8818289041519165, "reward_std": 0.21396346390247345, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.18027674406766891, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 1214.9055495350003, "entropy_control/entropy": 0.1839382771645593, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.21939887812500003, "entropy_control/epsilon_high_next": 0.23036882203125003, "epoch": 0.092, "adapter_sync_time": 0.0, "generation_time": 1066.810888, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1093.354588, "forward_backward_time": 121.449762, "rollout_overhead_time": 26.5437, "peak_allocated_gib": 71.76706552505493, "peak_reserved_gib": 79.626953125 }, { "step": 24, "loss": -0.01762533187866211, "grad_norm": 0.07080078125, "learning_rate": 5.400000000000001e-06, "entropy": 0.18072561360895634, "clip_ratio/low_mean": 0.0012508091749623418, "clip_ratio/low_min": 0.0012508091749623418, "clip_ratio/high_mean": 0.001962777489097789, "clip_ratio/high_max": 0.001962777489097789, "clip_ratio/region_mean": 0.0032135866640601307, "step_time": 121.5651989240032, "entropy_control/entropy": 0.1844771598448184, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.23036882203125003, "entropy_control/epsilon_high_next": 0.24188726313281253, "epoch": 0.096, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 121.463173, "rollout_overhead_time": 0.0, "peak_allocated_gib": 71.76706552505493, "peak_reserved_gib": 79.626953125, "reward": null }, { "step": 25, "loss": -0.201101154088974, "grad_norm": 0.06689453125, "learning_rate": 5.2e-06, "thinking/tokens_mean": 651.0571428571428, "thinking/budget_hit_fraction": 0.15, "num_tokens": 736632.0, "completions/mean_length": 3384.5, "completions/min_length": 1382.0, "completions/max_length": 5307.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 3384.5, "completions/min_terminated_length": 1382.0, "completions/max_terminated_length": 5307.0, "tools/call_frequency": 2.375, "tools/failure_frequency": 0.10526315867900848, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.5191874504089355, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 0.3654783070087433, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.4166666865348816, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.3191424012184143, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/mean": 0.6478563547134399, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/std": 0.07784092426300049, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/mean": 0.625, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/std": 0.0833333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/std": 0.0, "reward": 0.7595635652542114, "reward_std": 0.3977477550506592, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 0.5, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.20006831921637058, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 1024.8544964600005, "entropy_control/entropy": 0.21341266139776957, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.24188726313281253, "entropy_control/epsilon_high_next": 0.2297928999761719, "epoch": 0.1, "adapter_sync_time": 0.0, "generation_time": 840.236188, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 872.671843, "forward_backward_time": 152.082744, "rollout_overhead_time": 32.435655, "peak_allocated_gib": 71.76706552505493, "peak_reserved_gib": 79.626953125 }, { "step": 26, "loss": -0.2011522650718689, "grad_norm": 0.06591796875, "learning_rate": 5e-06, "entropy": 0.19986664317548275, "clip_ratio/low_mean": 0.0013427212252281606, "clip_ratio/low_min": 0.0013427212252281606, "clip_ratio/high_mean": 0.0019352009403519332, "clip_ratio/high_max": 0.0019352009403519332, "clip_ratio/region_mean": 0.003277922165580094, "step_time": 152.1805966360007, "entropy_control/entropy": 0.2131841554288399, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2297928999761719, "entropy_control/epsilon_high_next": 0.21830325497736328, "epoch": 0.104, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 152.079885, "rollout_overhead_time": 0.0, "peak_allocated_gib": 71.76706552505493, "peak_reserved_gib": 79.626953125, "reward": null }, { "step": 27, "loss": -0.16143891215324402, "grad_norm": 0.06787109375, "learning_rate": 4.800000000000001e-06, "thinking/tokens_mean": 646.45, "thinking/budget_hit_fraction": 0.10833333333333332, "num_tokens": 792291.0, "completions/mean_length": 3654.875, "completions/min_length": 1506.0, "completions/max_length": 6260.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 3654.875, "completions/min_terminated_length": 1506.0, "completions/max_terminated_length": 6260.0, "tools/call_frequency": 3.125, "tools/failure_frequency": 0.03999999910593033, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/mean": 0.673007607460022, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/std": 0.07297920435667038, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/mean": 0.5833333730697632, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/std": 0.09622505307197571, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/mean": 0.9404942393302917, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/std": 0.07794613391160965, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/mean": 0.4166666865348816, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/std": 0.1666666716337204, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/std": 0.0, "reward": 1.0317509174346924, "reward_std": 0.15252529084682465, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.18070382438600063, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 1710.4444454119985, "entropy_control/entropy": 0.17121478947137753, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.21830325497736328, "entropy_control/epsilon_high_next": 0.22921841772623144, "epoch": 0.108, "adapter_sync_time": 0.0, "generation_time": 1490.457345, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1528.317853, "forward_backward_time": 182.026363, "rollout_overhead_time": 37.860508, "peak_allocated_gib": 71.76706552505493, "peak_reserved_gib": 79.626953125 }, { "step": 28, "loss": -0.16135908663272858, "grad_norm": 0.08056640625, "learning_rate": 4.600000000000001e-06, "entropy": 0.18029101006686687, "clip_ratio/low_mean": 0.0012838473194278777, "clip_ratio/low_min": 0.0012838473194278777, "clip_ratio/high_mean": 0.0024870873312465847, "clip_ratio/high_max": 0.0024870873312465847, "clip_ratio/region_mean": 0.0037709346506744623, "step_time": 182.1155030810005, "entropy_control/entropy": 0.171024284979791, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.22921841772623144, "entropy_control/epsilon_high_next": 0.24067933861254304, "epoch": 0.112, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 182.008493, "rollout_overhead_time": 0.0, "peak_allocated_gib": 71.76706552505493, "peak_reserved_gib": 79.626953125, "reward": null }, { "step": 29, "loss": -0.07088714092969894, "grad_norm": 0.06201171875, "learning_rate": 4.4e-06, "thinking/tokens_mean": 569.4464285714286, "thinking/budget_hit_fraction": 0.125, "num_tokens": 853868.0, "completions/mean_length": 4414.625, "completions/min_length": 3128.0, "completions/max_length": 6527.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 4414.625, "completions/min_terminated_length": 3128.0, "completions/max_terminated_length": 6527.0, "tools/call_frequency": 3.625, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.6391764879226685, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 0.2832109332084656, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.3333333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.3042903244495392, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/mean": 0.6660953760147095, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/std": 0.060347460210323334, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/mean": 0.4583333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/std": 0.2500000298023224, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/std": 0.0, "reward": 0.8224276304244995, "reward_std": 0.3251701891422272, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.191108962520957, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 1681.9464596959988, "entropy_control/entropy": 0.19075258099255643, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.24067933861254304, "entropy_control/epsilon_high_next": 0.2527133055431702, "epoch": 0.116, "adapter_sync_time": 0.0, "generation_time": 1386.857225, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1437.03397, "forward_backward_time": 244.809965, "rollout_overhead_time": 50.176745, "peak_allocated_gib": 71.76706552505493, "peak_reserved_gib": 79.626953125 }, { "step": 30, "loss": -0.07075826823711395, "grad_norm": 0.060302734375, "learning_rate": 4.2000000000000004e-06, "entropy": 0.19117097929120064, "clip_ratio/low_mean": 0.0006441471632570028, "clip_ratio/low_min": 0.0006441471632570028, "clip_ratio/high_mean": 0.0010533519234741107, "clip_ratio/high_max": 0.0010533519234741107, "clip_ratio/region_mean": 0.0016974990867311135, "step_time": 244.9255358420014, "entropy_control/entropy": 0.1908541491110437, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2527133055431702, "entropy_control/epsilon_high_next": 0.2653489708203287, "epoch": 0.12, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 244.819153, "rollout_overhead_time": 0.0, "peak_allocated_gib": 71.76706552505493, "peak_reserved_gib": 79.626953125, "reward": null }, { "step": 31, "loss": -0.0906260535120964, "grad_norm": 0.08154296875, "learning_rate": 4.000000000000001e-06, "thinking/tokens_mean": 550.4, "thinking/budget_hit_fraction": 0.15, "num_tokens": 903122.0, "completions/mean_length": 3073.25, "completions/min_length": 1155.0, "completions/max_length": 4391.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 3073.25, "completions/min_terminated_length": 1155.0, "completions/max_terminated_length": 4391.0, "tools/call_frequency": 2.75, "tools/failure_frequency": 0.04545454680919647, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/mean": 0.862329363822937, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/std": 0.06961258500814438, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/mean": 0.5625, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/std": 0.12400397658348083, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/std": 0.0, "reward": 1.090454339981079, "reward_std": 0.06799826771020889, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.17278917878866196, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 1001.1734930269995, "entropy_control/entropy": 0.1790868654981576, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2653489708203287, "entropy_control/epsilon_high_next": 0.2786164193613452, "epoch": 0.124, "adapter_sync_time": 0.0, "generation_time": 839.902116, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 868.566563, "forward_backward_time": 132.502458, "rollout_overhead_time": 28.664448, "peak_allocated_gib": 71.76706552505493, "peak_reserved_gib": 79.626953125 }, { "step": 32, "loss": -0.09048309922218323, "grad_norm": 0.08203125, "learning_rate": 3.8000000000000005e-06, "entropy": 0.17249196767807007, "clip_ratio/low_mean": 0.002061832492472604, "clip_ratio/low_min": 0.002061832492472604, "clip_ratio/high_mean": 0.0002366217304370366, "clip_ratio/high_max": 0.0002366217304370366, "clip_ratio/region_mean": 0.0022984542229096405, "step_time": 132.5778156569977, "entropy_control/entropy": 0.17875137752195996, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2786164193613452, "entropy_control/epsilon_high_next": 0.29254724032941243, "epoch": 0.128, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 132.475599, "rollout_overhead_time": 0.0, "peak_allocated_gib": 71.76706552505493, "peak_reserved_gib": 79.626953125, "reward": null }, { "step": 33, "loss": -0.11338338255882263, "grad_norm": 0.06689453125, "learning_rate": 3.6000000000000003e-06, "thinking/tokens_mean": 678.3152777777779, "thinking/budget_hit_fraction": 0.14583333333333334, "num_tokens": 958967.0, "completions/mean_length": 4085.125, "completions/min_length": 2616.0, "completions/max_length": 6136.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 4085.125, "completions/min_terminated_length": 2616.0, "completions/max_terminated_length": 6136.0, "tools/call_frequency": 3.25, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.6678308844566345, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 0.11652150750160217, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.625, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.0833333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/mean": 0.8567099571228027, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/std": 0.06670232862234116, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/mean": 0.2916666865348816, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/std": 0.15957118570804596, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/std": 0.0, "reward": 0.9851870536804199, "reward_std": 0.12486284971237183, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.20331044867634773, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 1482.6869959809992, "entropy_control/entropy": 0.20108881576114443, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.29254724032941243, "entropy_control/epsilon_high_next": 0.2779198783129418, "epoch": 0.132, "adapter_sync_time": 0.0, "generation_time": 1270.347497, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1307.78921, "forward_backward_time": 174.792873, "rollout_overhead_time": 37.441713, "peak_allocated_gib": 71.76706552505493, "peak_reserved_gib": 79.626953125 }, { "step": 34, "loss": -0.11310680210590363, "grad_norm": 0.07568359375, "learning_rate": 3.4000000000000005e-06, "entropy": 0.20380816981196404, "clip_ratio/low_mean": 0.0015240056563925464, "clip_ratio/low_min": 0.0015240056563925464, "clip_ratio/high_mean": 0.000278479463304393, "clip_ratio/high_max": 0.000278479463304393, "clip_ratio/region_mean": 0.0018024851196969394, "step_time": 174.89699893899888, "entropy_control/entropy": 0.20159983308152926, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2779198783129418, "entropy_control/epsilon_high_next": 0.2640238843972947, "epoch": 0.136, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 174.794567, "rollout_overhead_time": 0.0, "peak_allocated_gib": 71.76706552505493, "peak_reserved_gib": 79.626953125, "reward": null }, { "step": 35, "loss": -0.32082635164260864, "grad_norm": 0.04638671875, "learning_rate": 3.2000000000000003e-06, "thinking/tokens_mean": 843.0123809523809, "thinking/budget_hit_fraction": 0.22857142857142856, "num_tokens": 1024576.0, "completions/mean_length": 4338.125, "completions/min_length": 2393.0, "completions/max_length": 6076.0, "completions/clipped_ratio": 0.125, "completions/mean_terminated_length": 4519.14306640625, "completions/min_terminated_length": 2393.0, "completions/max_terminated_length": 6076.0, "tools/call_frequency": 2.5, "tools/failure_frequency": 0.05000000074505806, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/mean": 0.50920569896698, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/std": 0.24889644980430603, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/mean": 0.375, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/std": 0.263523131608963, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/std": 0.9258201122283936, "reward": 0.6279556751251221, "reward_std": 0.44035667181015015, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 0.5, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.18321112729609013, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 1476.4721071010063, "entropy_control/entropy": 0.2186302428886329, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2640238843972947, "entropy_control/epsilon_high_next": 0.2508226901774299, "epoch": 0.14, "adapter_sync_time": 0.0, "generation_time": 1236.165505, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1277.813583, "forward_backward_time": 198.553303, "rollout_overhead_time": 41.648078, "peak_allocated_gib": 71.76706552505493, "peak_reserved_gib": 79.626953125 }, { "step": 36, "loss": -0.3209863007068634, "grad_norm": 0.0439453125, "learning_rate": 3e-06, "entropy": 0.18317675963044167, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0012614787701750174, "clip_ratio/high_max": 0.0012614787701750174, "clip_ratio/region_mean": 0.0012614787701750174, "step_time": 198.67015630999958, "entropy_control/entropy": 0.218624029132364, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2508226901774299, "entropy_control/epsilon_high_next": 0.2382815556685584, "epoch": 0.144, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 198.558416, "rollout_overhead_time": 0.0, "peak_allocated_gib": 71.76706552505493, "peak_reserved_gib": 79.626953125, "reward": null }, { "step": 37, "loss": -0.054933544248342514, "grad_norm": 0.07470703125, "learning_rate": 2.8000000000000003e-06, "thinking/tokens_mean": 864.75, "thinking/budget_hit_fraction": 0.21875, "num_tokens": 1077659.0, "completions/mean_length": 3755.375, "completions/min_length": 2668.0, "completions/max_length": 4991.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 3755.375, "completions/min_terminated_length": 2668.0, "completions/max_terminated_length": 4991.0, "tools/call_frequency": 2.5, "tools/failure_frequency": 0.05000000074505806, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.6097036004066467, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.6666666865348816, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__native/mean": 0.7660998106002808, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__native/std": 0.10691505670547485, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__efficiency/mean": 0.5416666865348816, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__efficiency/std": 0.0833333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__reliability/std": 0.0, "reward": 0.9181100726127625, "reward_std": 0.10477986186742783, "frac_reward_zero_std": 0.5, "frac_groups_zero_std": 0.5, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.2165136467665434, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 1063.0183756520055, "entropy_control/entropy": 0.22111522776930506, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2382815556685584, "entropy_control/epsilon_high_next": 0.22636747788513048, "epoch": 0.148, "adapter_sync_time": 0.0, "generation_time": 923.867443, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 949.333476, "forward_backward_time": 113.572721, "rollout_overhead_time": 25.466033, "peak_allocated_gib": 71.76706552505493, "peak_reserved_gib": 79.62890625 }, { "step": 38, "loss": -0.05470426380634308, "grad_norm": 0.130859375, "learning_rate": 2.6e-06, "entropy": 0.21626891754567623, "clip_ratio/low_mean": 0.0002055357617791742, "clip_ratio/low_min": 0.0002055357617791742, "clip_ratio/high_mean": 0.0013210960605647415, "clip_ratio/high_max": 0.0013210960605647415, "clip_ratio/region_mean": 0.0015266318223439157, "step_time": 113.67919772201276, "entropy_control/entropy": 0.22098230880714823, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.22636747788513048, "entropy_control/epsilon_high_next": 0.21504910399087396, "epoch": 0.152, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 113.55989, "rollout_overhead_time": 0.0, "peak_allocated_gib": 71.76706552505493, "peak_reserved_gib": 79.62890625, "reward": null }, { "step": 39, "loss": -0.26859793066978455, "grad_norm": 0.06689453125, "learning_rate": 2.4000000000000003e-06, "thinking/tokens_mean": 549.5583333333333, "thinking/budget_hit_fraction": 0.125, "num_tokens": 1131453.0, "completions/mean_length": 3633.25, "completions/min_length": 1174.0, "completions/max_length": 5464.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 3633.25, "completions/min_terminated_length": 1174.0, "completions/max_terminated_length": 5464.0, "tools/call_frequency": 3.125, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__native/mean": 0.74747633934021, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__native/std": 0.11491075158119202, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__efficiency/mean": 0.4791666865348816, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__efficiency/std": 0.18766526877880096, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__reliability/std": 0.0, "reward": 0.9714346528053284, "reward_std": 0.1079513430595398, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.18086103163659573, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 1335.4767555669969, "entropy_control/entropy": 0.1890181002301508, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.21504910399087396, "entropy_control/epsilon_high_next": 0.22580155919041767, "epoch": 0.156, "adapter_sync_time": 0.0, "generation_time": 1100.428583, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1141.04147, "forward_backward_time": 194.333393, "rollout_overhead_time": 40.612888, "peak_allocated_gib": 71.76706552505493, "peak_reserved_gib": 79.62890625 }, { "step": 40, "loss": -0.26867029070854187, "grad_norm": 0.0673828125, "learning_rate": 2.2e-06, "entropy": 0.18077907338738441, "clip_ratio/low_mean": 0.0008553312727599405, "clip_ratio/low_min": 0.0008553312727599405, "clip_ratio/high_mean": 0.00214733014581725, "clip_ratio/high_max": 0.00214733014581725, "clip_ratio/region_mean": 0.0030026614185771905, "step_time": 194.4316224219947, "entropy_control/entropy": 0.18876167534824811, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.22580155919041767, "entropy_control/epsilon_high_next": 0.23709163714993856, "epoch": 0.16, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 194.31877, "rollout_overhead_time": 0.0, "peak_allocated_gib": 71.76706552505493, "peak_reserved_gib": 79.62890625, "reward": null }, { "step": 41, "loss": -0.1159663125872612, "grad_norm": 0.046630859375, "learning_rate": 2.0000000000000003e-06, "thinking/tokens_mean": 817.8630952380953, "thinking/budget_hit_fraction": 0.2619047619047619, "num_tokens": 1200270.0, "completions/mean_length": 5125.125, "completions/min_length": 3071.0, "completions/max_length": 6806.0, "completions/clipped_ratio": 0.125, "completions/mean_terminated_length": 5418.57177734375, "completions/min_terminated_length": 3133.0, "completions/max_terminated_length": 6806.0, "tools/call_frequency": 3.0, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__native/mean": 0.5503982305526733, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__native/std": 0.40622156858444214, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__efficiency/mean": 0.4166666865348816, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__efficiency/std": 0.3191424012184143, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__reliability/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__reliability/std": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/mean": 0.71760094165802, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/std": 0.11173579841852188, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/mean": 0.3333333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/std": 0.13608276844024658, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/std": 0.0, "reward": 0.8027495741844177, "reward_std": 0.4265289008617401, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 0.5, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.20343837141990662, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 1858.2825202210006, "entropy_control/entropy": 0.2303477592256566, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.23709163714993856, "entropy_control/epsilon_high_next": 0.22523705529244162, "epoch": 0.164, "adapter_sync_time": 0.0, "generation_time": 1556.781741, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1607.940264, "forward_backward_time": 250.235761, "rollout_overhead_time": 51.158524, "peak_allocated_gib": 71.76706552505493, "peak_reserved_gib": 79.62890625 }, { "step": 42, "loss": -0.11591409891843796, "grad_norm": 0.04248046875, "learning_rate": 1.8000000000000001e-06, "entropy": 0.2033819518983364, "clip_ratio/low_mean": 0.0005881242686882615, "clip_ratio/low_min": 0.0005881242686882615, "clip_ratio/high_mean": 0.0019081929058302194, "clip_ratio/high_max": 0.0019081929058302194, "clip_ratio/region_mean": 0.002496317174518481, "step_time": 250.3050878659924, "entropy_control/entropy": 0.2303099920568688, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.22523705529244162, "entropy_control/epsilon_high_next": 0.2139752025278195, "epoch": 0.168, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 250.196043, "rollout_overhead_time": 0.0, "peak_allocated_gib": 71.76706552505493, "peak_reserved_gib": 79.62890625, "reward": null }, { "step": 43, "loss": -0.09850254654884338, "grad_norm": 0.05126953125, "learning_rate": 1.6000000000000001e-06, "thinking/tokens_mean": 637.2821428571428, "thinking/budget_hit_fraction": 0.1, "num_tokens": 1250821.0, "completions/mean_length": 3014.875, "completions/min_length": 1675.0, "completions/max_length": 4096.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 3014.875, "completions/min_terminated_length": 1675.0, "completions/max_terminated_length": 4096.0, "tools/call_frequency": 2.125, "tools/failure_frequency": 0.05882352963089943, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__native/mean": 0.610934853553772, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__native/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__efficiency/mean": 0.6666666865348816, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__efficiency/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/mean": 0.548965334892273, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/std": 0.40477925539016724, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/mean": 0.4166666865348816, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/std": 0.3191424012184143, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/std": 1.0, "reward": 0.7570334672927856, "reward_std": 0.40556252002716064, "frac_reward_zero_std": 0.5, "frac_groups_zero_std": 0.5, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 0.5, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.21251165866851807, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 975.4812457649969, "entropy_control/entropy": 0.21498346984857497, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2139752025278195, "entropy_control/epsilon_high_next": 0.20327644240142853, "epoch": 0.172, "adapter_sync_time": 0.0, "generation_time": 799.469618, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 830.851928, "forward_backward_time": 144.525428, "rollout_overhead_time": 31.38231, "peak_allocated_gib": 71.76706552505493, "peak_reserved_gib": 79.62890625 }, { "step": 44, "loss": -0.09850792586803436, "grad_norm": 0.051025390625, "learning_rate": 1.4000000000000001e-06, "entropy": 0.212388776242733, "clip_ratio/low_mean": 0.00038839978515170515, "clip_ratio/low_min": 0.00038839978515170515, "clip_ratio/high_mean": 0.0023090573376975954, "clip_ratio/high_max": 0.0023090573376975954, "clip_ratio/region_mean": 0.0026974571228493005, "step_time": 144.64556900199386, "entropy_control/entropy": 0.21487993369050687, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.20327644240142853, "entropy_control/epsilon_high_next": 0.2, "epoch": 0.176, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 144.538271, "rollout_overhead_time": 0.0, "peak_allocated_gib": 71.76706552505493, "peak_reserved_gib": 79.62890625, "reward": null }, { "step": 45, "loss": 0.01590845175087452, "grad_norm": 0.05810546875, "learning_rate": 1.2000000000000002e-06, "thinking/tokens_mean": 751.4166666666666, "thinking/budget_hit_fraction": 0.125, "num_tokens": 1318636.0, "completions/mean_length": 4806.875, "completions/min_length": 1452.0, "completions/max_length": 7509.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 4806.875, "completions/min_terminated_length": 1452.0, "completions/max_terminated_length": 7509.0, "tools/call_frequency": 3.75, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/mean": 0.7373476624488831, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/std": 0.12540532648563385, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/mean": 0.1666666716337204, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/mean": 0.6424247026443481, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/std": 0.0662192702293396, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/mean": 0.5833333730697632, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/std": 0.09622505307197571, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/std": 0.0, "reward": 0.9086361527442932, "reward_std": 0.10019896179437637, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.20502127707004547, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 2023.3134895429903, "entropy_control/entropy": 0.2044678493637871, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "epoch": 0.18, "adapter_sync_time": 0.0, "generation_time": 1702.871336, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1757.588775, "forward_backward_time": 265.61732, "rollout_overhead_time": 54.717439, "peak_allocated_gib": 71.76706552505493, "peak_reserved_gib": 79.62890625 }, { "step": 46, "loss": 0.016000889241695404, "grad_norm": 0.060302734375, "learning_rate": 1.0000000000000002e-06, "entropy": 0.20487543009221554, "clip_ratio/low_mean": 0.001967767399037257, "clip_ratio/low_min": 0.001967767399037257, "clip_ratio/high_mean": 0.0016679588879924268, "clip_ratio/high_max": 0.0016679588879924268, "clip_ratio/region_mean": 0.0036357262870296836, "step_time": 265.73495223200007, "entropy_control/entropy": 0.20430856135558217, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "epoch": 0.184, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 265.628295, "rollout_overhead_time": 0.0, "peak_allocated_gib": 71.76706552505493, "peak_reserved_gib": 79.62890625, "reward": null }, { "step": 47, "loss": 0.09188379347324371, "grad_norm": 0.0771484375, "learning_rate": 8.000000000000001e-07, "thinking/tokens_mean": 510.4059523809524, "thinking/budget_hit_fraction": 0.08928571428571429, "num_tokens": 1376992.0, "completions/mean_length": 4008.0, "completions/min_length": 1883.0, "completions/max_length": 6183.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 4008.0, "completions/min_terminated_length": 1883.0, "completions/max_terminated_length": 6183.0, "tools/call_frequency": 4.125, "tools/failure_frequency": 0.03030303120613098, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/mean": 0.48760396242141724, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/std": 0.2901604175567627, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/mean": 0.2083333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/std": 0.31549492478370667, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/mean": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/std": 1.154700517654419, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/mean": 0.8249422311782837, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/std": 0.06404384970664978, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/mean": 0.4583333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/std": 0.20971764624118805, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/std": 0.0, "reward": 0.7729398012161255, "reward_std": 0.4533771872520447, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.16416428238153458, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 1709.7197918070233, "entropy_control/entropy": 0.1681721646306333, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.21000000000000002, "epoch": 0.188, "adapter_sync_time": 0.0, "generation_time": 1462.965742, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1505.441768, "forward_backward_time": 204.17351, "rollout_overhead_time": 42.476026, "peak_allocated_gib": 71.76706552505493, "peak_reserved_gib": 79.62890625 }, { "step": 48, "loss": 0.09163554012775421, "grad_norm": 0.06396484375, "learning_rate": 6.000000000000001e-07, "entropy": 0.16428864933550358, "clip_ratio/low_mean": 0.0009273621544707566, "clip_ratio/low_min": 0.0009273621544707566, "clip_ratio/high_mean": 0.0036222787166479975, "clip_ratio/high_max": 0.0036222787166479975, "clip_ratio/region_mean": 0.004549640871118754, "step_time": 204.26634985800774, "entropy_control/entropy": 0.16840625487843042, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.21000000000000002, "entropy_control/epsilon_high_next": 0.22050000000000003, "epoch": 0.192, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 204.162215, "rollout_overhead_time": 0.0, "peak_allocated_gib": 71.76706552505493, "peak_reserved_gib": 79.62890625, "reward": null }, { "step": 49, "loss": -0.040280066430568695, "grad_norm": 0.064453125, "learning_rate": 4.0000000000000003e-07, "thinking/tokens_mean": 694.0875, "thinking/budget_hit_fraction": 0.14583333333333334, "num_tokens": 1436680.0, "completions/mean_length": 4170.0, "completions/min_length": 2453.0, "completions/max_length": 6379.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 4170.0, "completions/min_terminated_length": 2453.0, "completions/max_terminated_length": 6379.0, "tools/call_frequency": 3.0, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/mean": 0.6862475872039795, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/std": 0.08939856290817261, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/mean": 0.5833333730697632, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/std": 0.09622505307197571, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/mean": 0.8245845437049866, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/std": 0.0023437440395355225, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/mean": 0.4166666865348816, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/std": 0.21516574919223785, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/std": 0.0, "reward": 0.9804160594940186, "reward_std": 0.08924110233783722, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.19738993979990482, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 1591.6186869769954, "entropy_control/entropy": 0.20175876097369605, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.22050000000000003, "entropy_control/epsilon_high_next": 0.20947500000000002, "epoch": 0.196, "adapter_sync_time": 0.0, "generation_time": 1379.444516, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1416.252849, "forward_backward_time": 175.254696, "rollout_overhead_time": 36.808333, "peak_allocated_gib": 71.76706552505493, "peak_reserved_gib": 79.62890625 }, { "step": 50, "loss": -0.040061138570308685, "grad_norm": 0.062255859375, "learning_rate": 2.0000000000000002e-07, "entropy": 0.19751964323222637, "clip_ratio/low_mean": 0.0013465700903907418, "clip_ratio/low_min": 0.0013465700903907418, "clip_ratio/high_mean": 0.0019430930260568857, "clip_ratio/high_max": 0.0019430930260568857, "clip_ratio/region_mean": 0.0032896631164476275, "step_time": 175.35021079199214, "entropy_control/entropy": 0.20191867984365963, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.20947500000000002, "entropy_control/epsilon_high_next": 0.2, "epoch": 0.2, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 175.242225, "rollout_overhead_time": 0.0, "peak_allocated_gib": 71.76706552505493, "peak_reserved_gib": 79.62890625, "reward": null } ], "holdout": [ { "seed": 900000, "sample": 0, "native_reward": 0.6089882858614486, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 2224, "reward_channels": { "native": 0.6089882858614486, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "two_pool_cycle" }, { "seed": 900000, "sample": 1, "native_reward": 0.6089882858614486, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 3204, "reward_channels": { "native": 0.6089882858614486, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "two_pool_cycle" }, { "seed": 900000, "sample": 2, "native_reward": 0.6089882858614486, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 3807, "reward_channels": { "native": 0.6089882858614486, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "two_pool_cycle" }, { "seed": 900000, "sample": 3, "native_reward": 0.6089882858614486, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 2653, "reward_channels": { "native": 0.6089882858614486, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "two_pool_cycle" }, { "seed": 900001, "sample": 0, "native_reward": 0.8523284576243806, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 3, "model_tokens": 4987, "reward_channels": { "native": 0.8523284576243806, "efficiency": 0.5, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "two_pool_cycle" }, { "seed": 900001, "sample": 1, "native_reward": 0.8435035328689895, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 3, "model_tokens": 4440, "reward_channels": { "native": 0.8435035328689895, "efficiency": 0.5, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "two_pool_cycle" }, { "seed": 900001, "sample": 2, "native_reward": 0.6093825392865522, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 3825, "reward_channels": { "native": 0.6093825392865522, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "two_pool_cycle" }, { "seed": 900001, "sample": 3, "native_reward": 0.6093825392865522, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 2831, "reward_channels": { "native": 0.6093825392865522, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "two_pool_cycle" }, { "seed": 900000, "sample": 0, "native_reward": 0.6106601460847819, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 3001, "reward_channels": { "native": 0.6106601460847819, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "triangular_cycle" }, { "seed": 900000, "sample": 1, "native_reward": 0.6106601460847819, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 3157, "reward_channels": { "native": 0.6106601460847819, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "triangular_cycle" }, { "seed": 900000, "sample": 2, "native_reward": 0.8522045135672316, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 5328, "reward_channels": { "native": 0.8522045135672316, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "triangular_cycle" }, { "seed": 900000, "sample": 3, "native_reward": 0.6106601460847819, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 3224, "reward_channels": { "native": 0.6106601460847819, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "triangular_cycle" }, { "seed": 900001, "sample": 0, "native_reward": 0.175, "promotion_success": false, "done": true, "errors": 0, "tool_steps": 3, "model_tokens": 5461, "reward_channels": { "native": 0.175, "efficiency": 0.5, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [ "positive_net_profit", "stress_positive", "task_complete" ], "family": "triangular_cycle" }, { "seed": 900001, "sample": 1, "native_reward": 0.6108851248150757, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 3109, "reward_channels": { "native": 0.6108851248150757, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "triangular_cycle" }, { "seed": 900001, "sample": 2, "native_reward": 0.25, "promotion_success": false, "done": false, "errors": 0, "tool_steps": 6, "model_tokens": 5912, "reward_channels": { "native": 0.25, "efficiency": 0.0, "reliability": -1.0 }, "verification_level": null, "failed_gates": [ "finalized" ], "family": "triangular_cycle" }, { "seed": 900001, "sample": 3, "native_reward": 0.811786874720879, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 3, "model_tokens": 4856, "reward_channels": { "native": 0.811786874720879, "efficiency": 0.5, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "triangular_cycle" }, { "seed": 900000, "sample": 0, "native_reward": 0.6140552238672119, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 3, "model_tokens": 4881, "reward_channels": { "native": 0.6140552238672119, "efficiency": 0.5, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "competing_cycles" }, { "seed": 900000, "sample": 1, "native_reward": 0.25, "promotion_success": false, "done": false, "errors": 0, "tool_steps": 6, "model_tokens": 5108, "reward_channels": { "native": 0.25, "efficiency": 0.0, "reliability": -1.0 }, "verification_level": null, "failed_gates": [ "finalized" ], "family": "competing_cycles" }, { "seed": 900000, "sample": 2, "native_reward": 0.25, "promotion_success": false, "done": false, "errors": 0, "tool_steps": 6, "model_tokens": 8927, "reward_channels": { "native": 0.25, "efficiency": 0.0, "reliability": -1.0 }, "verification_level": null, "failed_gates": [ "finalized" ], "family": "competing_cycles" }, { "seed": 900000, "sample": 3, "native_reward": 0.6185941887823871, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 3, "model_tokens": 5438, "reward_channels": { "native": 0.6185941887823871, "efficiency": 0.5, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "competing_cycles" }, { "seed": 900001, "sample": 0, "native_reward": 0.0, "promotion_success": false, "done": false, "errors": 0, "tool_steps": 1, "model_tokens": 0, "reward_channels": { "native": 0.0, "efficiency": 0.0, "reliability": -1.0 }, "verification_level": null, "failed_gates": [ "source_safe", "route_valid", "amount_bounded", "execution_success", "accounting_conserved", "positive_net_profit", "stress_positive", "native_replay", "verified_protocol", "evidence_bound", "finalized", "effective_work", "task_complete" ], "family": "competing_cycles" }, { "seed": 900001, "sample": 1, "native_reward": 0.6145108982971254, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 3860, "reward_channels": { "native": 0.6145108982971254, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "competing_cycles" }, { "seed": 900001, "sample": 2, "native_reward": 0.6512057439829095, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 3, "model_tokens": 3868, "reward_channels": { "native": 0.6512057439829095, "efficiency": 0.5, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "competing_cycles" }, { "seed": 900001, "sample": 3, "native_reward": 0.6145108982971254, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 3, "model_tokens": 3557, "reward_channels": { "native": 0.6145108982971254, "efficiency": 0.5, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "competing_cycles" }, { "seed": 900000, "sample": 0, "native_reward": 0.7005591947284964, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 3734, "reward_channels": { "native": 0.7005591947284964, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "four_hop_cycle" }, { "seed": 900000, "sample": 1, "native_reward": 0.609376392538787, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 1411, "reward_channels": { "native": 0.609376392538787, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "four_hop_cycle" }, { "seed": 900000, "sample": 2, "native_reward": 0.8444084478162579, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 5142, "reward_channels": { "native": 0.8444084478162579, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "four_hop_cycle" }, { "seed": 900000, "sample": 3, "native_reward": 0.7456673418171573, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 5, "model_tokens": 3540, "reward_channels": { "native": 0.7456673418171573, "efficiency": 0.16666666666666663, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "four_hop_cycle" }, { "seed": 900001, "sample": 0, "native_reward": 0.7352912527935178, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 5, "model_tokens": 5458, "reward_channels": { "native": 0.7352912527935178, "efficiency": 0.16666666666666663, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "four_hop_cycle" }, { "seed": 900001, "sample": 1, "native_reward": 0.739978752793518, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 3, "model_tokens": 2244, "reward_channels": { "native": 0.739978752793518, "efficiency": 0.5, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "four_hop_cycle" }, { "seed": 900001, "sample": 2, "native_reward": 0.6095701077161815, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 3, "model_tokens": 1872, "reward_channels": { "native": 0.6095701077161815, "efficiency": 0.5, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "four_hop_cycle" }, { "seed": 900001, "sample": 3, "native_reward": 0.739978752793518, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 2585, "reward_channels": { "native": 0.739978752793518, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "four_hop_cycle" }, { "seed": 900000, "sample": 0, "native_reward": 0.8262556477476042, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 3192, "reward_channels": { "native": 0.8262556477476042, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "size_sensitive_cycle" }, { "seed": 900000, "sample": 1, "native_reward": 0.8262556477476042, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 3638, "reward_channels": { "native": 0.8262556477476042, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "size_sensitive_cycle" }, { "seed": 900000, "sample": 2, "native_reward": 0.8262556477476042, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 3, "model_tokens": 4892, "reward_channels": { "native": 0.8262556477476042, "efficiency": 0.5, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "size_sensitive_cycle" }, { "seed": 900000, "sample": 3, "native_reward": 0.8262556477476042, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 3, "model_tokens": 4492, "reward_channels": { "native": 0.8262556477476042, "efficiency": 0.5, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "size_sensitive_cycle" }, { "seed": 900001, "sample": 0, "native_reward": 0.0, "promotion_success": false, "done": false, "errors": 0, "tool_steps": 0, "model_tokens": 0, "reward_channels": { "native": 0.0, "efficiency": 0.0, "reliability": -1.0 }, "verification_level": null, "failed_gates": null, "family": "size_sensitive_cycle" }, { "seed": 900001, "sample": 1, "native_reward": 0.8245803237223959, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 2218, "reward_channels": { "native": 0.8245803237223959, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "size_sensitive_cycle" }, { "seed": 900001, "sample": 2, "native_reward": 0.8245803237223959, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 3203, "reward_channels": { "native": 0.8245803237223959, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "size_sensitive_cycle" }, { "seed": 900001, "sample": 3, "native_reward": 0.8245803237223959, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 5074, "reward_channels": { "native": 0.8245803237223959, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "size_sensitive_cycle" } ] }, { "run_id": "1789312330", "mode": "thinking", "method": "repo-r", "steps": 50, "reward_points": 25, "train_mean": 0.8284654068946838, "train_last10": 0.8206436157226562, "holdout_native": 0.646087216010387, "holdout_episodes": 40, "holdout_successes": 34, "mean_step_seconds": 922.4200192468194, "label": "DAPO + REPO-R", "state_at_cutoff": "complete", "step_hours": 12.811389156205825, "peak_allocated_gib": 68.00909471511841, "peak_reserved_gib": 75.5390625, "extra_refill_rounds": 0.0, "groups_refilled": 0.0, "holdout_component_means": { "native": 0.646087216010387, "efficiency": 0.4541666666666667, "reliability": 0.8 }, "configuration": { "model": { "id": "Qwen/Qwen3-14B", "revision": "40c069824f4251a91eefaf281ebe4c544efd3e18", "loss": "dapo", "algorithm": "grpo", "no_think": false, "thinking_budget_tokens": 2048, "lora": true, "quantize": true, "lora_r": 32, "lora_alpha": 64, "use_dora": false }, "loss": "dapo", "normalization": "joint", "entropy_control": { "method": "repo_r", "epsilon_low": 0.2, "epsilon_high": 0.28, "zeta_initial": 0.001, "zeta_min": 0.0001, "zeta_max": 0.05, "holdout_split": "benchmark", "holdout_seeds_per_family": 2, "num_iterations": 2 }, "dynamic_sampling": false, "sampling": { "group_size": 4, "temperature": 1.0, "max_completion_tokens": 22528, "top_entropy_quantile": 1.0, "temperature_decay_to": null }, "learning_rate": 1e-05, "turn_max_tokens": 3072 }, "source_sha256": { "metrics.jsonl": "174e16c95109bd9333850af284778a331544bc8625e06a3b8c45112ddf77edd1", "entropy_holdout_episodes.jsonl": "6c53816ae490ab1264b6b65fd11d5705dbb56deeb7f6d48aeec071864bd28c3f", "resolved_training_config.json": "c770506d6ac33f690596f71c4b137202b3ac566a82db8a820bcfb2b2cfdd1769" }, "training_steps": [ { "step": 1, "loss": -0.3259884715080261, "grad_norm": 0.061279296875, "learning_rate": 1e-05, "thinking/tokens_mean": 606.2361111111111, "thinking/budget_hit_fraction": 0.14583333333333334, "num_tokens": 51975.0, "completions/mean_length": 3604.375, "completions/min_length": 2890.0, "completions/max_length": 6014.0, "completions/clipped_ratio": 0.25, "completions/mean_terminated_length": 3782.166748046875, "completions/min_terminated_length": 2890.0, "completions/max_terminated_length": 6014.0, "tools/call_frequency": 2.25, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.7262155413627625, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 0.13481803238391876, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.4583333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.2500000298023224, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/mean": 0.45231834053993225, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/std": 0.526568591594696, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/mean": 0.2916666865348816, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/std": 0.34359216690063477, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/mean": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/std": 1.154700517654419, "reward": 0.7080169916152954, "reward_std": 0.5734051465988159, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 0.5, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.14539151638746262, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 1397.8954549800003, "entropy_control/entropy": 0.20009788357278327, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.001, "entropy_control/zeta_next": 0.001, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.004, "adapter_sync_time": 0.0, "generation_time": 1205.825887, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1240.642983, "forward_backward_time": 157.147441, "rollout_overhead_time": 34.817096, "peak_allocated_gib": 46.211081981658936, "peak_reserved_gib": 51.38671875 }, { "step": 2, "loss": -0.3258787989616394, "grad_norm": 0.06298828125, "learning_rate": 9.800000000000001e-06, "entropy": 0.14511961489915848, "clip_ratio/low_mean": 0.0005848915170645341, "clip_ratio/low_min": 0.0005848915170645341, "clip_ratio/high_mean": 0.0006545017204189207, "clip_ratio/high_max": 0.0006545017204189207, "clip_ratio/region_mean": 0.0012393932374834549, "step_time": 157.18345037700055, "entropy_control/entropy": 0.1997818389141923, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.001, "entropy_control/zeta_next": 0.002, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.008, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 157.072636, "rollout_overhead_time": 0.0, "peak_allocated_gib": 46.689598083496094, "peak_reserved_gib": 51.484375, "reward": null }, { "step": 3, "loss": -0.3502673804759979, "grad_norm": 0.07763671875, "learning_rate": 9.600000000000001e-06, "thinking/tokens_mean": 656.7291666666666, "thinking/budget_hit_fraction": 0.14583333333333334, "num_tokens": 103035.0, "completions/mean_length": 3670.5, "completions/min_length": 1836.0, "completions/max_length": 7313.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 3670.5, "completions/min_terminated_length": 1836.0, "completions/max_terminated_length": 7313.0, "tools/call_frequency": 2.625, "tools/failure_frequency": 0.0476190485060215, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.6708315014839172, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 0.11447126418352127, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.5833333730697632, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.15430335700511932, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 0.0, "reward": 0.8999981880187988, "reward_std": 0.10676517337560654, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.20492193195968866, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 1548.7683420090002, "entropy_control/entropy": 0.2165817376038305, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.002, "entropy_control/zeta_next": 0.001, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.012, "adapter_sync_time": 0.0, "generation_time": 1353.74487, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1388.408148, "forward_backward_time": 160.264247, "rollout_overhead_time": 34.663278, "peak_allocated_gib": 50.21120643615723, "peak_reserved_gib": 55.546875 }, { "step": 4, "loss": -0.35036006569862366, "grad_norm": 0.0810546875, "learning_rate": 9.4e-06, "entropy": 0.20442917477339506, "clip_ratio/low_mean": 0.0018149314128095284, "clip_ratio/low_min": 0.0018149314128095284, "clip_ratio/high_mean": 0.00011673664994305, "clip_ratio/high_max": 0.00011673664994305, "clip_ratio/region_mean": 0.0019316680627525784, "step_time": 160.3734592299993, "entropy_control/entropy": 0.2161768356278206, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.001, "entropy_control/zeta_next": 0.0005, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.016, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 160.274561, "rollout_overhead_time": 0.0, "peak_allocated_gib": 50.21120643615723, "peak_reserved_gib": 55.546875, "reward": null }, { "step": 5, "loss": -0.1864038109779358, "grad_norm": 0.06787109375, "learning_rate": 9.200000000000002e-06, "thinking/tokens_mean": 645.15625, "thinking/budget_hit_fraction": 0.1875, "num_tokens": 147211.0, "completions/mean_length": 2817.0, "completions/min_length": 1370.0, "completions/max_length": 5240.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 2817.0, "completions/min_terminated_length": 1370.0, "completions/max_terminated_length": 5240.0, "tools/call_frequency": 2.25, "tools/failure_frequency": 0.0555555559694767, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.6402809619903564, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 0.08739244937896729, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.6458333730697632, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.0589255727827549, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 0.0, "reward": 0.8725726008415222, "reward_std": 0.08444617688655853, "frac_reward_zero_std": 0.5, "frac_groups_zero_std": 0.5, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.19345584325492382, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 945.5713407839985, "entropy_control/entropy": 0.20152860386475172, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0005, "entropy_control/zeta_next": 0.00025, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.02, "adapter_sync_time": 0.0, "generation_time": 815.934001, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 839.828299, "forward_backward_time": 105.638448, "rollout_overhead_time": 23.894298, "peak_allocated_gib": 50.21120643615723, "peak_reserved_gib": 55.55859375 }, { "step": 6, "loss": -0.18664570152759552, "grad_norm": 0.060791015625, "learning_rate": 9e-06, "entropy": 0.19353977404534817, "clip_ratio/low_mean": 0.0012533892004285008, "clip_ratio/low_min": 0.0012533892004285008, "clip_ratio/high_mean": 0.00021469464991241693, "clip_ratio/high_max": 0.00021469464991241693, "clip_ratio/region_mean": 0.0014680838503409177, "step_time": 105.7294935389973, "entropy_control/entropy": 0.20151033951591435, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.00025, "entropy_control/zeta_next": 0.000125, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.024, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 105.629041, "rollout_overhead_time": 0.0, "peak_allocated_gib": 50.21120643615723, "peak_reserved_gib": 55.55859375, "reward": null }, { "step": 7, "loss": -0.08715154230594635, "grad_norm": 0.056884765625, "learning_rate": 8.8e-06, "thinking/tokens_mean": 712.6023809523809, "thinking/budget_hit_fraction": 0.17857142857142858, "num_tokens": 217300.0, "completions/mean_length": 5293.625, "completions/min_length": 3090.0, "completions/max_length": 7621.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 5293.625, "completions/min_terminated_length": 3090.0, "completions/max_terminated_length": 7621.0, "tools/call_frequency": 4.0, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/mean": 0.6874937415122986, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/std": 0.12660378217697144, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/mean": 0.375, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/std": 0.20971763134002686, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/mean": 0.9020491242408752, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/std": 0.08809521794319153, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/mean": 0.2916666865348816, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/std": 0.34359216690063477, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/std": 0.0, "reward": 1.0114381313323975, "reward_std": 0.14690396189689636, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.19253043085336685, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 2190.706770779001, "entropy_control/entropy": 0.19304759234163033, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.000125, "entropy_control/zeta_next": 0.00025, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.028, "adapter_sync_time": 0.0, "generation_time": 1870.915505, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1925.203813, "forward_backward_time": 265.406044, "rollout_overhead_time": 54.288308, "peak_allocated_gib": 63.816951751708984, "peak_reserved_gib": 70.79296875 }, { "step": 8, "loss": -0.08721963316202164, "grad_norm": 0.054931640625, "learning_rate": 8.6e-06, "entropy": 0.19271664507687092, "clip_ratio/low_mean": 0.0016498293407494202, "clip_ratio/low_min": 0.0016498293407494202, "clip_ratio/high_mean": 0.0003511721297400072, "clip_ratio/high_max": 0.0003511721297400072, "clip_ratio/region_mean": 0.0020010014704894274, "step_time": 265.50813235699934, "entropy_control/entropy": 0.1932519039943739, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.00025, "entropy_control/zeta_next": 0.0005, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.032, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 265.40423, "rollout_overhead_time": 0.0, "peak_allocated_gib": 63.816951751708984, "peak_reserved_gib": 70.79296875, "reward": null }, { "step": 9, "loss": 0.05990079045295715, "grad_norm": 0.0654296875, "learning_rate": 8.400000000000001e-06, "thinking/tokens_mean": 740.2291666666666, "thinking/budget_hit_fraction": 0.16666666666666666, "num_tokens": 277354.0, "completions/mean_length": 4222.75, "completions/min_length": 3070.0, "completions/max_length": 7225.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 4222.75, "completions/min_terminated_length": 3070.0, "completions/max_terminated_length": 7225.0, "tools/call_frequency": 3.0, "tools/failure_frequency": 0.0416666679084301, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.609546422958374, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.625, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.0833333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/mean": 0.6306283473968506, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/std": 0.05577770620584488, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/mean": 0.4166666865348816, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/std": 0.21516573429107666, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/std": 0.0, "reward": 0.8461290597915649, "reward_std": 0.03769681602716446, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.1999448798596859, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 1576.6268515149977, "entropy_control/entropy": 0.19911247380742922, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0005, "entropy_control/zeta_next": 0.001, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.036, "adapter_sync_time": 0.0, "generation_time": 1276.454583, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1327.900362, "forward_backward_time": 248.626321, "rollout_overhead_time": 51.445779, "peak_allocated_gib": 63.816951751708984, "peak_reserved_gib": 70.79296875 }, { "step": 10, "loss": 0.06006348878145218, "grad_norm": 0.06689453125, "learning_rate": 8.2e-06, "entropy": 0.19951391220092773, "clip_ratio/low_mean": 0.0013390503881964833, "clip_ratio/low_min": 0.0013390503881964833, "clip_ratio/high_mean": 0.0006475002883234993, "clip_ratio/high_max": 0.0006475002883234993, "clip_ratio/region_mean": 0.0019865506765199825, "step_time": 248.70580136699755, "entropy_control/entropy": 0.1987466156907593, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.001, "entropy_control/zeta_next": 0.002, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.04, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 248.601753, "rollout_overhead_time": 0.0, "peak_allocated_gib": 63.816951751708984, "peak_reserved_gib": 70.79296875, "reward": null }, { "step": 11, "loss": -0.33186206221580505, "grad_norm": 0.060302734375, "learning_rate": 8.000000000000001e-06, "thinking/tokens_mean": 661.9261904761904, "thinking/budget_hit_fraction": 0.15, "num_tokens": 332088.0, "completions/mean_length": 3532.75, "completions/min_length": 1457.0, "completions/max_length": 4803.0, "completions/clipped_ratio": 0.125, "completions/mean_terminated_length": 3598.71435546875, "completions/min_terminated_length": 1457.0, "completions/max_terminated_length": 4803.0, "tools/call_frequency": 2.375, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__native/mean": 0.5509068965911865, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__native/std": 0.4066128134727478, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__efficiency/mean": 0.4166666865348816, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__efficiency/std": 0.3191424012184143, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__reliability/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__reliability/std": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/mean": 0.6762324571609497, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/std": 0.07764213532209396, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/mean": 0.5416666865348816, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/std": 0.15957120060920715, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/std": 0.0, "reward": 0.7875280380249023, "reward_std": 0.41725409030914307, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 0.5, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.15983807668089867, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 1366.433276806998, "entropy_control/entropy": 0.18736514871922605, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.002, "entropy_control/zeta_next": 0.004, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.044, "adapter_sync_time": 0.0, "generation_time": 1175.752347, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1209.253643, "forward_backward_time": 157.068599, "rollout_overhead_time": 33.501296, "peak_allocated_gib": 63.816951751708984, "peak_reserved_gib": 70.79296875 }, { "step": 12, "loss": -0.33221426606178284, "grad_norm": 0.0595703125, "learning_rate": 7.800000000000002e-06, "entropy": 0.1599844992160797, "clip_ratio/low_mean": 0.0007805810600984842, "clip_ratio/low_min": 0.0007805810600984842, "clip_ratio/high_mean": 0.0007289346867764834, "clip_ratio/high_max": 0.0007289346867764834, "clip_ratio/region_mean": 0.0015095157468749676, "step_time": 157.17439058199852, "entropy_control/entropy": 0.18762194995238382, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.004, "entropy_control/zeta_next": 0.008, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.048, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 157.0646, "rollout_overhead_time": 0.0, "peak_allocated_gib": 63.816951751708984, "peak_reserved_gib": 70.79296875, "reward": null }, { "step": 13, "loss": -0.2631528377532959, "grad_norm": 0.05419921875, "learning_rate": 7.600000000000001e-06, "thinking/tokens_mean": 638.756462585034, "thinking/budget_hit_fraction": 0.2122448979591837, "num_tokens": 394315.0, "completions/mean_length": 4493.375, "completions/min_length": 3071.0, "completions/max_length": 7951.0, "completions/clipped_ratio": 0.125, "completions/mean_terminated_length": 4696.57177734375, "completions/min_terminated_length": 3585.0, "completions/max_terminated_length": 7951.0, "tools/call_frequency": 3.0, "tools/failure_frequency": 0.0416666679084301, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.7278944253921509, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 0.1366470754146576, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.19245010614395142, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/mean": 0.44153478741645813, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/std": 0.3896979093551636, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/mean": 0.1666666716337204, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/std": 0.235702246427536, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/mean": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/std": 1.154700517654419, "reward": 0.701381266117096, "reward_std": 0.49392881989479065, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 0.5, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.18411841057240963, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 2238.979801022, "entropy_control/entropy": 0.20485563017215608, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.008, "entropy_control/zeta_next": 0.004, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.052, "adapter_sync_time": 0.0, "generation_time": 1921.36754, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1974.973927, "forward_backward_time": 263.896289, "rollout_overhead_time": 53.606387, "peak_allocated_gib": 63.816951751708984, "peak_reserved_gib": 70.79296875 }, { "step": 14, "loss": -0.2626897394657135, "grad_norm": 0.051025390625, "learning_rate": 7.4e-06, "entropy": 0.18377662170678377, "clip_ratio/low_mean": 0.0008020004752324894, "clip_ratio/low_min": 0.0008020004752324894, "clip_ratio/high_mean": 0.0006139878314570524, "clip_ratio/high_max": 0.0006139878314570524, "clip_ratio/region_mean": 0.0014159883066895418, "step_time": 263.99378604999947, "entropy_control/entropy": 0.20453749518624145, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.004, "entropy_control/zeta_next": 0.002, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.056, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 263.88747, "rollout_overhead_time": 0.0, "peak_allocated_gib": 63.816951751708984, "peak_reserved_gib": 70.79296875, "reward": null }, { "step": 15, "loss": -0.2147071659564972, "grad_norm": 0.05712890625, "learning_rate": 7.2000000000000005e-06, "thinking/tokens_mean": 570.465306122449, "thinking/budget_hit_fraction": 0.14540816326530612, "num_tokens": 454805.0, "completions/mean_length": 4077.75, "completions/min_length": 2950.0, "completions/max_length": 4939.0, "completions/clipped_ratio": 0.125, "completions/mean_terminated_length": 4221.57177734375, "completions/min_terminated_length": 2950.0, "completions/max_terminated_length": 4939.0, "tools/call_frequency": 3.0, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/mean": 0.5258879661560059, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/std": 0.3715890347957611, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/mean": 0.4166666865348816, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/std": 0.3191424012184143, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/std": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/mean": 0.889053463935852, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/std": 0.07614489644765854, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/mean": 0.3333333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/std": 0.2357022762298584, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/std": 0.0, "reward": 0.876220703125, "reward_std": 0.45209410786628723, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 0.5, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.18128307536244392, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 1532.063953914996, "entropy_control/entropy": 0.20716967831794803, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.002, "entropy_control/zeta_next": 0.001, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.06, "adapter_sync_time": 0.0, "generation_time": 1246.226957, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1294.656645, "forward_backward_time": 237.296763, "rollout_overhead_time": 48.429688, "peak_allocated_gib": 63.816951751708984, "peak_reserved_gib": 70.79296875 }, { "step": 16, "loss": -0.21472179889678955, "grad_norm": 0.058837890625, "learning_rate": 7e-06, "entropy": 0.18125001713633537, "clip_ratio/low_mean": 0.0005973542865831405, "clip_ratio/low_min": 0.0005973542865831405, "clip_ratio/high_mean": 0.0006045860354788601, "clip_ratio/high_max": 0.0006045860354788601, "clip_ratio/region_mean": 0.0012019403220620006, "step_time": 237.40580801299984, "entropy_control/entropy": 0.2071025205107389, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.001, "entropy_control/zeta_next": 0.0005, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.064, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 237.299794, "rollout_overhead_time": 0.0, "peak_allocated_gib": 63.816951751708984, "peak_reserved_gib": 70.79296875, "reward": null }, { "step": 17, "loss": -0.05078861117362976, "grad_norm": 0.0712890625, "learning_rate": 6.800000000000001e-06, "thinking/tokens_mean": 504.6464285714286, "thinking/budget_hit_fraction": 0.07142857142857142, "num_tokens": 514341.0, "completions/mean_length": 3751.0, "completions/min_length": 2136.0, "completions/max_length": 5259.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 3751.0, "completions/min_terminated_length": 2136.0, "completions/max_terminated_length": 5259.0, "tools/call_frequency": 3.375, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/mean": 0.5663875341415405, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/std": 0.2860962152481079, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/mean": 0.4166666865348816, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/std": 0.2886751592159271, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/mean": 0.7022275924682617, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/std": 0.11214153468608856, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/mean": 0.4583333730697632, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/std": 0.3154948949813843, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/std": 0.0, "reward": 0.8561825752258301, "reward_std": 0.2066703736782074, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.21092794835567474, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 1719.3857764660006, "entropy_control/entropy": 0.2120397044521445, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0005, "entropy_control/zeta_next": 0.00025, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.068, "adapter_sync_time": 0.0, "generation_time": 1422.643111, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1473.245488, "forward_backward_time": 246.033984, "rollout_overhead_time": 50.602377, "peak_allocated_gib": 63.816951751708984, "peak_reserved_gib": 70.79296875 }, { "step": 18, "loss": -0.050790995359420776, "grad_norm": 0.0732421875, "learning_rate": 6.600000000000001e-06, "entropy": 0.21118208207190037, "clip_ratio/low_mean": 0.0008529162441845983, "clip_ratio/low_min": 0.0008529162441845983, "clip_ratio/high_mean": 0.0011459748639026657, "clip_ratio/high_max": 0.0011459748639026657, "clip_ratio/region_mean": 0.001998891108087264, "step_time": 246.0806536239943, "entropy_control/entropy": 0.2123507976546537, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.00025, "entropy_control/zeta_next": 0.000125, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.072, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 245.97729, "rollout_overhead_time": 0.0, "peak_allocated_gib": 63.816951751708984, "peak_reserved_gib": 70.79296875, "reward": null }, { "step": 19, "loss": 0.029592961072921753, "grad_norm": 0.068359375, "learning_rate": 6.4000000000000006e-06, "thinking/tokens_mean": 714.2850000000001, "thinking/budget_hit_fraction": 0.175, "num_tokens": 568836.0, "completions/mean_length": 3921.375, "completions/min_length": 1926.0, "completions/max_length": 5923.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 3921.375, "completions/min_terminated_length": 1926.0, "completions/max_terminated_length": 5923.0, "tools/call_frequency": 2.75, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.7603124976158142, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 0.1818402260541916, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.5416666865348816, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.15957120060920715, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/mean": 0.8599027395248413, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/std": 0.06802016496658325, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/mean": 0.5416666865348816, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/std": 0.0833333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/std": 0.0, "reward": 1.0371909141540527, "reward_std": 0.1342398226261139, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.2082794103771448, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 1175.120890283004, "entropy_control/entropy": 0.21819758630772168, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.000125, "entropy_control/zeta_next": -0.0001, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.076, "adapter_sync_time": 0.0, "generation_time": 986.266377, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1019.209362, "forward_backward_time": 155.805796, "rollout_overhead_time": 32.942985, "peak_allocated_gib": 63.816951751708984, "peak_reserved_gib": 70.79296875 }, { "step": 20, "loss": 0.029754847288131714, "grad_norm": 0.06396484375, "learning_rate": 6.200000000000001e-06, "entropy": 0.20808766968548298, "clip_ratio/low_mean": 0.0012133791606174782, "clip_ratio/low_min": 0.0012133791606174782, "clip_ratio/high_mean": 0.0005690024336217903, "clip_ratio/high_max": 0.0005690024336217903, "clip_ratio/region_mean": 0.0017823815942392685, "step_time": 155.8805688670036, "entropy_control/entropy": 0.21820074963448827, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": -0.0001, "entropy_control/zeta_next": -0.0002, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.08, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 155.780796, "rollout_overhead_time": 0.0, "peak_allocated_gib": 63.816951751708984, "peak_reserved_gib": 70.79296875, "reward": null }, { "step": 21, "loss": -0.31540295481681824, "grad_norm": 0.06494140625, "learning_rate": 6e-06, "thinking/tokens_mean": 724.3333333333334, "thinking/budget_hit_fraction": 0.10416666666666667, "num_tokens": 633899.0, "completions/mean_length": 4433.875, "completions/min_length": 1723.0, "completions/max_length": 7881.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 4433.875, "completions/min_terminated_length": 1723.0, "completions/max_terminated_length": 7881.0, "tools/call_frequency": 3.25, "tools/failure_frequency": 0.03846153989434242, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/mean": 0.7041764259338379, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/std": 0.11985881626605988, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/mean": 0.4166666865348816, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/std": 0.28867512941360474, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/mean": 0.6817549467086792, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/std": 0.0835287794470787, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/mean": 0.5416666865348816, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/std": 0.15957120060920715, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/std": 0.0, "reward": 0.9169240593910217, "reward_std": 0.08680310100317001, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.20554296672344208, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 1894.4528153590036, "entropy_control/entropy": 0.21422870491736745, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": -0.0002, "entropy_control/zeta_next": -0.0004, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.084, "adapter_sync_time": 0.0, "generation_time": 1591.184369, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1642.459084, "forward_backward_time": 251.89198, "rollout_overhead_time": 51.274715, "peak_allocated_gib": 63.816951751708984, "peak_reserved_gib": 70.79296875 }, { "step": 22, "loss": -0.31544703245162964, "grad_norm": 0.0625, "learning_rate": 5.8e-06, "entropy": 0.20528246276080608, "clip_ratio/low_mean": 0.0021731581655330956, "clip_ratio/low_min": 0.0021731581655330956, "clip_ratio/high_mean": 0.0003061480056203436, "clip_ratio/high_max": 0.0003061480056203436, "clip_ratio/region_mean": 0.002479306171153439, "step_time": 251.9174536130049, "entropy_control/entropy": 0.21406269566679018, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": -0.0004, "entropy_control/zeta_next": -0.0008, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.088, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 251.817624, "rollout_overhead_time": 0.0, "peak_allocated_gib": 63.816951751708984, "peak_reserved_gib": 70.79296875, "reward": null }, { "step": 23, "loss": -0.12085682153701782, "grad_norm": 0.0625, "learning_rate": 5.600000000000001e-06, "thinking/tokens_mean": 656.0714285714286, "thinking/budget_hit_fraction": 0.17261904761904762, "num_tokens": 691397.0, "completions/mean_length": 4301.75, "completions/min_length": 3071.0, "completions/max_length": 6698.0, "completions/clipped_ratio": 0.25, "completions/mean_terminated_length": 4712.0, "completions/min_terminated_length": 3097.0, "completions/max_terminated_length": 6698.0, "tools/call_frequency": 2.5, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.3634990453720093, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 0.4305455982685089, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.25, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.3191424012184143, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 1.154700517654419, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/mean": 0.6815767288208008, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/std": 0.2877177894115448, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/mean": 0.4166666865348816, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/std": 0.3191424012184143, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/std": 1.0, "reward": 0.5892045497894287, "reward_std": 0.5939663052558899, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 0.5, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.1555847767740488, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 1628.1413750700012, "entropy_control/entropy": 0.21338015065453458, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": -0.0008, "entropy_control/zeta_next": -0.0016, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.092, "adapter_sync_time": 0.0, "generation_time": 1384.163753, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1426.234478, "forward_backward_time": 201.796162, "rollout_overhead_time": 42.070724, "peak_allocated_gib": 63.816951751708984, "peak_reserved_gib": 70.79296875 }, { "step": 24, "loss": -0.12096354365348816, "grad_norm": 0.0595703125, "learning_rate": 5.400000000000001e-06, "entropy": 0.15518717840313911, "clip_ratio/low_mean": 0.0003359211841598153, "clip_ratio/low_min": 0.0003359211841598153, "clip_ratio/high_mean": 0.0006267605058383197, "clip_ratio/high_max": 0.0006267605058383197, "clip_ratio/region_mean": 0.000962681689998135, "step_time": 201.867253359007, "entropy_control/entropy": 0.2128140368689202, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": -0.0016, "entropy_control/zeta_next": -0.0032, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.096, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 201.768838, "rollout_overhead_time": 0.0, "peak_allocated_gib": 63.816951751708984, "peak_reserved_gib": 70.79296875, "reward": null }, { "step": 25, "loss": 0.2103634476661682, "grad_norm": 0.10888671875, "learning_rate": 5.2e-06, "thinking/tokens_mean": 435.86309523809524, "thinking/budget_hit_fraction": 0.05357142857142857, "num_tokens": 738777.0, "completions/mean_length": 2832.5, "completions/min_length": 1469.0, "completions/max_length": 6308.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 2832.5, "completions/min_terminated_length": 1469.0, "completions/max_terminated_length": 6308.0, "tools/call_frequency": 2.875, "tools/failure_frequency": 0.043478261679410934, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.6715044975280762, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 0.12447243928909302, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.625, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.0833333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/mean": 0.5774143934249878, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/std": 0.24432186782360077, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/mean": 0.4583333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/std": 0.3154948949813843, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/std": 1.0, "reward": 0.8015427589416504, "reward_std": 0.32135501503944397, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.19548141490668058, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 1220.5425894130021, "entropy_control/entropy": 0.21258601135593283, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": -0.0032, "entropy_control/zeta_next": -0.0064, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.1, "adapter_sync_time": 0.0, "generation_time": 958.455504, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1003.67661, "forward_backward_time": 216.763821, "rollout_overhead_time": 45.221106, "peak_allocated_gib": 63.816951751708984, "peak_reserved_gib": 70.79296875 }, { "step": 26, "loss": 0.21094360947608948, "grad_norm": 0.09814453125, "learning_rate": 5e-06, "entropy": 0.19584557320922613, "clip_ratio/low_mean": 0.001290573156438768, "clip_ratio/low_min": 0.001290573156438768, "clip_ratio/high_mean": 0.0009297065698774531, "clip_ratio/high_max": 0.0009297065698774531, "clip_ratio/region_mean": 0.002220279726316221, "step_time": 216.85135657700084, "entropy_control/entropy": 0.2129269318505866, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": -0.0064, "entropy_control/zeta_next": -0.0128, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.104, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 216.746383, "rollout_overhead_time": 0.0, "peak_allocated_gib": 63.816951751708984, "peak_reserved_gib": 70.79296875, "reward": null }, { "step": 27, "loss": 0.09279686212539673, "grad_norm": 0.0634765625, "learning_rate": 4.800000000000001e-06, "thinking/tokens_mean": 602.5680272108845, "thinking/budget_hit_fraction": 0.17857142857142858, "num_tokens": 800262.0, "completions/mean_length": 4383.125, "completions/min_length": 2033.0, "completions/max_length": 8346.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 4383.125, "completions/min_terminated_length": 2033.0, "completions/max_terminated_length": 8346.0, "tools/call_frequency": 3.5, "tools/failure_frequency": 0.1071428582072258, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/mean": 0.673007607460022, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/std": 0.07297920435667038, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/mean": 0.5833333730697632, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/std": 0.09622505307197571, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/mean": 0.6735195517539978, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/std": 0.3504052758216858, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/mean": 0.375, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/std": 0.15957118570804596, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/std": 0.0, "reward": 0.8972219228744507, "reward_std": 0.23320473730564117, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.19499888643622398, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 1952.804152615994, "entropy_control/entropy": 0.20484359934442195, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": -0.0128, "entropy_control/zeta_next": -0.0256, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.108, "adapter_sync_time": 0.0, "generation_time": 1690.824962, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1735.214541, "forward_backward_time": 217.490981, "rollout_overhead_time": 44.38958, "peak_allocated_gib": 63.816951751708984, "peak_reserved_gib": 70.79296875 }, { "step": 28, "loss": 0.09532079100608826, "grad_norm": 0.08056640625, "learning_rate": 4.600000000000001e-06, "entropy": 0.19523677229881287, "clip_ratio/low_mean": 0.0011541051499079913, "clip_ratio/low_min": 0.0011541051499079913, "clip_ratio/high_mean": 0.0008292896673083305, "clip_ratio/high_max": 0.0008292896673083305, "clip_ratio/region_mean": 0.001983394817216322, "step_time": 217.55828219300383, "entropy_control/entropy": 0.20515715360251255, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": -0.0256, "entropy_control/zeta_next": -0.05, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.112, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 217.457063, "rollout_overhead_time": 0.0, "peak_allocated_gib": 63.816951751708984, "peak_reserved_gib": 70.79296875, "reward": null }, { "step": 29, "loss": -0.18449796736240387, "grad_norm": 0.06298828125, "learning_rate": 4.4e-06, "thinking/tokens_mean": 517.373299319728, "thinking/budget_hit_fraction": 0.14285714285714285, "num_tokens": 857417.0, "completions/mean_length": 3861.875, "completions/min_length": 1429.0, "completions/max_length": 6376.0, "completions/clipped_ratio": 0.125, "completions/mean_terminated_length": 3944.857421875, "completions/min_terminated_length": 1429.0, "completions/max_terminated_length": 6376.0, "tools/call_frequency": 3.375, "tools/failure_frequency": 0.03703703731298447, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.6719769239425659, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 0.12673872709274292, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.625, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.0833333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/mean": 0.5019842386245728, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/std": 0.30418699979782104, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/mean": 0.0416666679084301, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/std": 0.0833333358168602, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/mean": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/std": 1.154700517654419, "reward": 0.703647255897522, "reward_std": 0.41580674052238464, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.16537689603865147, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 1504.720339291005, "entropy_control/entropy": 0.18967879363501303, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": -0.05, "entropy_control/zeta_next": -0.025, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.116, "adapter_sync_time": 0.0, "generation_time": 1207.691629, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1258.410011, "forward_backward_time": 246.204273, "rollout_overhead_time": 50.718382, "peak_allocated_gib": 63.816951751708984, "peak_reserved_gib": 70.79296875 }, { "step": 30, "loss": -0.1881130039691925, "grad_norm": 0.058837890625, "learning_rate": 4.2000000000000004e-06, "entropy": 0.16569842398166656, "clip_ratio/low_mean": 0.001011697924695909, "clip_ratio/low_min": 0.001011697924695909, "clip_ratio/high_mean": 0.0004280114735593088, "clip_ratio/high_max": 0.0004280114735593088, "clip_ratio/region_mean": 0.0014397093982552178, "step_time": 246.31147144499482, "entropy_control/entropy": 0.18999628573423533, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": -0.025, "entropy_control/zeta_next": -0.0125, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.12, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 246.208792, "rollout_overhead_time": 0.0, "peak_allocated_gib": 63.816951751708984, "peak_reserved_gib": 70.79296875, "reward": null }, { "step": 31, "loss": 0.11245188862085342, "grad_norm": 0.060302734375, "learning_rate": 4.000000000000001e-06, "thinking/tokens_mean": 794.2, "thinking/budget_hit_fraction": 0.25, "num_tokens": 915998.0, "completions/mean_length": 4239.125, "completions/min_length": 2550.0, "completions/max_length": 7442.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 4239.125, "completions/min_terminated_length": 2550.0, "completions/max_terminated_length": 7442.0, "tools/call_frequency": 2.875, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/mean": 0.7606706023216248, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/std": 0.24138176441192627, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/mean": 0.5208333730697632, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/std": 0.13908717036247253, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/std": 0.0, "reward": 0.9867122769355774, "reward_std": 0.24179385602474213, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.1865224912762642, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 1519.2641041699972, "entropy_control/entropy": 0.19128284683263885, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": -0.0125, "entropy_control/zeta_next": -0.00625, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.124, "adapter_sync_time": 0.0, "generation_time": 1284.900645, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1325.123069, "forward_backward_time": 194.033613, "rollout_overhead_time": 40.222424, "peak_allocated_gib": 63.816951751708984, "peak_reserved_gib": 70.79296875 }, { "step": 32, "loss": 0.11187267303466797, "grad_norm": 0.061279296875, "learning_rate": 3.8000000000000005e-06, "entropy": 0.18621982634067535, "clip_ratio/low_mean": 0.0012468731147237122, "clip_ratio/low_min": 0.0012468731147237122, "clip_ratio/high_mean": 0.000771423103287816, "clip_ratio/high_max": 0.000771423103287816, "clip_ratio/region_mean": 0.0020182962180115283, "step_time": 194.07235350299743, "entropy_control/entropy": 0.19088988400191137, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": -0.00625, "entropy_control/zeta_next": -0.003125, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.128, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 193.96996, "rollout_overhead_time": 0.0, "peak_allocated_gib": 63.816951751708984, "peak_reserved_gib": 70.79296875, "reward": null }, { "step": 33, "loss": -0.1797182261943817, "grad_norm": 0.0712890625, "learning_rate": 3.6000000000000003e-06, "thinking/tokens_mean": 686.7797619047618, "thinking/budget_hit_fraction": 0.125, "num_tokens": 973826.0, "completions/mean_length": 4333.0, "completions/min_length": 2403.0, "completions/max_length": 7804.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 4333.0, "completions/min_terminated_length": 2403.0, "completions/max_terminated_length": 7804.0, "tools/call_frequency": 3.25, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.6666589975357056, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 0.11417773365974426, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.5416666865348816, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.25, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/mean": 0.8620702028274536, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/std": 0.0774177834391594, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/mean": 0.375, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/std": 0.34359216690063477, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/std": 0.0, "reward": 0.987281322479248, "reward_std": 0.1277439296245575, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.17658358626067638, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 1659.7578923590008, "entropy_control/entropy": 0.18745651616632011, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": -0.003125, "entropy_control/zeta_next": -0.0015625, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.132, "adapter_sync_time": 0.0, "generation_time": 1391.378185, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1437.295868, "forward_backward_time": 222.362118, "rollout_overhead_time": 45.917683, "peak_allocated_gib": 63.816951751708984, "peak_reserved_gib": 70.79296875 }, { "step": 34, "loss": -0.18020114302635193, "grad_norm": 0.06884765625, "learning_rate": 3.4000000000000005e-06, "entropy": 0.1765458695590496, "clip_ratio/low_mean": 0.0018103036418324336, "clip_ratio/low_min": 0.0018103036418324336, "clip_ratio/high_mean": 0.00023717802832834423, "clip_ratio/high_max": 0.00023717802832834423, "clip_ratio/region_mean": 0.002047481670160778, "step_time": 222.44440870100516, "entropy_control/entropy": 0.18732343816424013, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": -0.0015625, "entropy_control/zeta_next": -0.00078125, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.136, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 222.341553, "rollout_overhead_time": 0.0, "peak_allocated_gib": 63.816951751708984, "peak_reserved_gib": 70.79296875, "reward": null }, { "step": 35, "loss": -0.3731302320957184, "grad_norm": 0.05419921875, "learning_rate": 3.2000000000000003e-06, "thinking/tokens_mean": 865.725, "thinking/budget_hit_fraction": 0.2569444444444445, "num_tokens": 1043625.0, "completions/mean_length": 4861.875, "completions/min_length": 3071.0, "completions/max_length": 7109.0, "completions/clipped_ratio": 0.25, "completions/mean_terminated_length": 5458.83349609375, "completions/min_terminated_length": 4089.0, "completions/max_terminated_length": 7109.0, "tools/call_frequency": 2.5, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/mean": 0.514873743057251, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/std": 0.3303550183773041, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/mean": 0.3333333134651184, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/std": 0.25197634100914, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/std": 0.9258201122283936, "reward": 0.6315404176712036, "reward_std": 0.520806074142456, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 0.5, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.16334183886647224, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 1771.3755442619877, "entropy_control/entropy": 0.2196462359248013, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": -0.00078125, "entropy_control/zeta_next": -0.0015625, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.14, "adapter_sync_time": 0.0, "generation_time": 1506.987091, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1551.99472, "forward_backward_time": 219.277062, "rollout_overhead_time": 45.007628, "peak_allocated_gib": 63.816951751708984, "peak_reserved_gib": 70.79296875 }, { "step": 36, "loss": -0.37306493520736694, "grad_norm": 0.052734375, "learning_rate": 3e-06, "entropy": 0.1635054349899292, "clip_ratio/low_mean": 0.0009762152621988207, "clip_ratio/low_min": 0.0009762152621988207, "clip_ratio/high_mean": 0.0003261847305111587, "clip_ratio/high_max": 0.0003261847305111587, "clip_ratio/region_mean": 0.0013023999927099794, "step_time": 219.30618613000115, "entropy_control/entropy": 0.21990908590362296, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": -0.0015625, "entropy_control/zeta_next": -0.003125, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.144, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 219.204383, "rollout_overhead_time": 0.0, "peak_allocated_gib": 63.816951751708984, "peak_reserved_gib": 70.79296875, "reward": null }, { "step": 37, "loss": -0.23768571019172668, "grad_norm": 0.053955078125, "learning_rate": 2.8000000000000003e-06, "thinking/tokens_mean": 508.3265306122449, "thinking/budget_hit_fraction": 0.14285714285714285, "num_tokens": 1097018.0, "completions/mean_length": 3794.125, "completions/min_length": 2939.0, "completions/max_length": 5140.0, "completions/clipped_ratio": 0.125, "completions/mean_terminated_length": 3897.4287109375, "completions/min_terminated_length": 2939.0, "completions/max_terminated_length": 5140.0, "tools/call_frequency": 3.125, "tools/failure_frequency": 0.11999999731779099, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.6686257123947144, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 0.11784413456916809, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.5833333730697632, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.1666666716337204, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__native/mean": 0.6207844018936157, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__native/std": 0.42944246530532837, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__efficiency/mean": 0.25, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__efficiency/std": 0.21516574919223785, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__reliability/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__reliability/std": 1.0, "reward": 0.8155384063720703, "reward_std": 0.42998006939888, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 0.5, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.16481151804327965, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 1172.8255403209987, "entropy_control/entropy": 0.1861763239960755, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": -0.003125, "entropy_control/zeta_next": -0.0015625, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.148, "adapter_sync_time": 0.0, "generation_time": 965.961993, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1001.937108, "forward_backward_time": 170.784393, "rollout_overhead_time": 35.975114, "peak_allocated_gib": 63.816951751708984, "peak_reserved_gib": 70.79296875 }, { "step": 38, "loss": -0.23787617683410645, "grad_norm": 0.05078125, "learning_rate": 2.6e-06, "entropy": 0.1651085540652275, "clip_ratio/low_mean": 0.000696542949299328, "clip_ratio/low_min": 0.000696542949299328, "clip_ratio/high_mean": 0.000684014150465373, "clip_ratio/high_max": 0.000684014150465373, "clip_ratio/region_mean": 0.001380557099764701, "step_time": 170.8765211380014, "entropy_control/entropy": 0.18647694617464183, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": -0.0015625, "entropy_control/zeta_next": -0.00078125, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.152, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 170.775489, "rollout_overhead_time": 0.0, "peak_allocated_gib": 63.816951751708984, "peak_reserved_gib": 70.79296875, "reward": null }, { "step": 39, "loss": -0.3025710880756378, "grad_norm": 0.05517578125, "learning_rate": 2.4000000000000003e-06, "thinking/tokens_mean": 641.109126984127, "thinking/budget_hit_fraction": 0.14583333333333334, "num_tokens": 1154511.0, "completions/mean_length": 4095.625, "completions/min_length": 2089.0, "completions/max_length": 5795.0, "completions/clipped_ratio": 0.125, "completions/mean_terminated_length": 4242.0, "completions/min_terminated_length": 2089.0, "completions/max_terminated_length": 5795.0, "tools/call_frequency": 3.125, "tools/failure_frequency": 0.03999999910593033, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__native/mean": 0.6562927961349487, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__native/std": 0.283261239528656, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__efficiency/mean": 0.3750000298023224, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__efficiency/std": 0.23145504295825958, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__reliability/mean": 0.75, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__reliability/std": 0.7071067690849304, "reward": 0.8250428438186646, "reward_std": 0.424506813287735, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 0.5, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.16697922721505165, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 1411.5272268420013, "entropy_control/entropy": 0.20049225201811438, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": -0.00078125, "entropy_control/zeta_next": -0.0015625, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.156, "adapter_sync_time": 0.0, "generation_time": 1190.893998, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1228.681028, "forward_backward_time": 182.742885, "rollout_overhead_time": 37.78703, "peak_allocated_gib": 63.816951751708984, "peak_reserved_gib": 70.79296875 }, { "step": 40, "loss": -0.30246925354003906, "grad_norm": 0.08251953125, "learning_rate": 2.2e-06, "entropy": 0.16671635583043098, "clip_ratio/low_mean": 0.00044798235467169434, "clip_ratio/low_min": 0.00044798235467169434, "clip_ratio/high_mean": 0.0006984352257859427, "clip_ratio/high_max": 0.0006984352257859427, "clip_ratio/region_mean": 0.001146417580457637, "step_time": 182.75106583199522, "entropy_control/entropy": 0.20023103373452278, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": -0.0015625, "entropy_control/zeta_next": -0.003125, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.16, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 182.649481, "rollout_overhead_time": 0.0, "peak_allocated_gib": 63.816951751708984, "peak_reserved_gib": 70.79296875, "reward": null }, { "step": 41, "loss": -0.1452902853488922, "grad_norm": 0.0634765625, "learning_rate": 2.0000000000000003e-06, "thinking/tokens_mean": 712.2908163265307, "thinking/budget_hit_fraction": 0.16581632653061223, "num_tokens": 1220760.0, "completions/mean_length": 4804.125, "completions/min_length": 3071.0, "completions/max_length": 8339.0, "completions/clipped_ratio": 0.125, "completions/mean_terminated_length": 5051.71435546875, "completions/min_terminated_length": 3120.0, "completions/max_terminated_length": 8339.0, "tools/call_frequency": 3.25, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__native/mean": 0.6815719604492188, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__native/std": 0.09391827881336212, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__efficiency/mean": 0.375, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__efficiency/std": 0.34359216690063477, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/mean": 0.3694747984409332, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/std": 0.3001787066459656, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/mean": 0.2916666865348816, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/std": 0.34359216690063477, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/mean": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/std": 1.154700517654419, "reward": 0.6421900987625122, "reward_std": 0.4515507221221924, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 0.5, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.16480917483568192, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 1962.5831768580101, "entropy_control/entropy": 0.186033986646413, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": -0.003125, "entropy_control/zeta_next": -0.0015625, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.164, "adapter_sync_time": 0.0, "generation_time": 1603.018019, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1663.572094, "forward_backward_time": 298.905308, "rollout_overhead_time": 60.554075, "peak_allocated_gib": 68.00909471511841, "peak_reserved_gib": 75.5390625 }, { "step": 42, "loss": -0.14536669850349426, "grad_norm": 0.08642578125, "learning_rate": 1.8000000000000001e-06, "entropy": 0.1649950724095106, "clip_ratio/low_mean": 0.0010179237433476374, "clip_ratio/low_min": 0.0010179237433476374, "clip_ratio/high_mean": 0.0004604757486958988, "clip_ratio/high_max": 0.0004604757486958988, "clip_ratio/region_mean": 0.0014783994920435362, "step_time": 298.94783146899135, "entropy_control/entropy": 0.1863435752899098, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": -0.0015625, "entropy_control/zeta_next": -0.00078125, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.168, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 298.845444, "rollout_overhead_time": 0.0, "peak_allocated_gib": 68.00909471511841, "peak_reserved_gib": 75.5390625, "reward": null }, { "step": 43, "loss": -0.20788943767547607, "grad_norm": 0.07177734375, "learning_rate": 1.6000000000000001e-06, "thinking/tokens_mean": 542.8666666666667, "thinking/budget_hit_fraction": 0.10416666666666667, "num_tokens": 1276636.0, "completions/mean_length": 3680.5, "completions/min_length": 1751.0, "completions/max_length": 7050.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 3680.5, "completions/min_terminated_length": 1751.0, "completions/max_terminated_length": 7050.0, "tools/call_frequency": 3.25, "tools/failure_frequency": 0.07692307978868484, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__native/mean": 0.7095590829849243, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__native/std": 0.11388145387172699, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__efficiency/mean": 0.5416666865348816, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__efficiency/std": 0.15957120060920715, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/mean": 0.7581974267959595, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/std": 0.11220034211874008, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/mean": 0.4583333730697632, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/std": 0.20971761643886566, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/std": 0.0, "reward": 0.958878219127655, "reward_std": 0.10066084563732147, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.18887744471430779, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 1519.0825120040026, "entropy_control/entropy": 0.19871808623602302, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": -0.00078125, "entropy_control/zeta_next": -0.000390625, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.172, "adapter_sync_time": 0.0, "generation_time": 1251.182353, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1297.051325, "forward_backward_time": 221.923109, "rollout_overhead_time": 45.868972, "peak_allocated_gib": 68.00909471511841, "peak_reserved_gib": 75.5390625 }, { "step": 44, "loss": -0.20786413550376892, "grad_norm": 0.07275390625, "learning_rate": 1.4000000000000001e-06, "entropy": 0.18883494287729263, "clip_ratio/low_mean": 0.001120483866543509, "clip_ratio/low_min": 0.001120483866543509, "clip_ratio/high_mean": 0.0005872241526958533, "clip_ratio/high_max": 0.0005872241526958533, "clip_ratio/region_mean": 0.0017077080192393623, "step_time": 221.9907098029944, "entropy_control/entropy": 0.19874537778473034, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": -0.000390625, "entropy_control/zeta_next": -0.0001953125, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.176, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 221.8927, "rollout_overhead_time": 0.0, "peak_allocated_gib": 68.00909471511841, "peak_reserved_gib": 75.5390625, "reward": null }, { "step": 45, "loss": 0.06324189901351929, "grad_norm": 0.060791015625, "learning_rate": 1.2000000000000002e-06, "thinking/tokens_mean": 618.2857142857143, "thinking/budget_hit_fraction": 0.07142857142857142, "num_tokens": 1345351.0, "completions/mean_length": 4919.375, "completions/min_length": 3344.0, "completions/max_length": 7296.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 4919.375, "completions/min_terminated_length": 3344.0, "completions/max_terminated_length": 7296.0, "tools/call_frequency": 4.5, "tools/failure_frequency": 0.0555555559694767, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/mean": 0.530854344367981, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/std": 0.18811559677124023, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/mean": 0.2916666567325592, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/std": 0.28463754057884216, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/std": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/mean": 0.7914260625839233, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/std": 0.06288611888885498, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/mean": 0.2916666865348816, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/std": 0.20971763134002686, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/std": 0.0, "reward": 0.8257235288619995, "reward_std": 0.326738178730011, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.18930352851748466, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 2097.171696471014, "entropy_control/entropy": 0.1893670189901332, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": -0.0001953125, "entropy_control/zeta_next": 0.0001, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.18, "adapter_sync_time": 0.0, "generation_time": 1782.842777, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1836.412046, "forward_backward_time": 260.660002, "rollout_overhead_time": 53.569268, "peak_allocated_gib": 68.00909471511841, "peak_reserved_gib": 75.5390625 }, { "step": 46, "loss": 0.06325738877058029, "grad_norm": 0.062255859375, "learning_rate": 1.0000000000000002e-06, "entropy": 0.18961226567626, "clip_ratio/low_mean": 0.0007533775205956772, "clip_ratio/low_min": 0.0007533775205956772, "clip_ratio/high_mean": 0.0007244420266943052, "clip_ratio/high_max": 0.0007244420266943052, "clip_ratio/region_mean": 0.0014778195472899824, "step_time": 260.7758615089988, "entropy_control/entropy": 0.1897029001935754, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0001, "entropy_control/zeta_next": 0.0002, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.184, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 260.67105, "rollout_overhead_time": 0.0, "peak_allocated_gib": 68.00909471511841, "peak_reserved_gib": 75.5390625, "reward": null }, { "step": 47, "loss": -0.1152733638882637, "grad_norm": 0.060302734375, "learning_rate": 8.000000000000001e-07, "thinking/tokens_mean": 591.780612244898, "thinking/budget_hit_fraction": 0.08928571428571429, "num_tokens": 1400370.0, "completions/mean_length": 3590.875, "completions/min_length": 2267.0, "completions/max_length": 4957.0, "completions/clipped_ratio": 0.125, "completions/mean_terminated_length": 3665.14306640625, "completions/min_terminated_length": 2267.0, "completions/max_terminated_length": 4957.0, "tools/call_frequency": 2.5, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/mean": 0.5786616802215576, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/std": 0.24573741853237152, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/mean": 0.4166666865348816, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/std": 0.3191424012184143, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/std": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/mean": 0.6189429759979248, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/std": 0.4126286804676056, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/std": 0.3333333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/std": 1.0, "reward": 0.7217190265655518, "reward_std": 0.5051907300949097, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 0.5, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.15867376886308193, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 1479.5910152529905, "entropy_control/entropy": 0.17867388151881314, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0002, "entropy_control/zeta_next": 0.0004, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.188, "adapter_sync_time": 0.0, "generation_time": 1243.889404, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1284.66303, "forward_backward_time": 194.826885, "rollout_overhead_time": 40.773626, "peak_allocated_gib": 68.00909471511841, "peak_reserved_gib": 75.5390625 }, { "step": 48, "loss": -0.11526072770357132, "grad_norm": 0.06201171875, "learning_rate": 6.000000000000001e-07, "entropy": 0.15877635776996613, "clip_ratio/low_mean": 0.0002521686546970159, "clip_ratio/low_min": 0.0002521686546970159, "clip_ratio/high_mean": 0.0007957203561090864, "clip_ratio/high_max": 0.0007957203561090864, "clip_ratio/region_mean": 0.0010478890108061023, "step_time": 194.92949151098583, "entropy_control/entropy": 0.1788228330419891, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0004, "entropy_control/zeta_next": 0.0008, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.192, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 194.827027, "rollout_overhead_time": 0.0, "peak_allocated_gib": 68.00909471511841, "peak_reserved_gib": 75.5390625, "reward": null }, { "step": 49, "loss": 0.00038623809814453125, "grad_norm": 0.06591796875, "learning_rate": 4.0000000000000003e-07, "thinking/tokens_mean": 608.9505102040815, "thinking/budget_hit_fraction": 0.12755102040816327, "num_tokens": 1462626.0, "completions/mean_length": 4491.0, "completions/min_length": 1555.0, "completions/max_length": 8626.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 4491.0, "completions/min_terminated_length": 1555.0, "completions/max_terminated_length": 8626.0, "tools/call_frequency": 3.875, "tools/failure_frequency": 0.09677419066429138, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/mean": 0.6475368738174438, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/std": 0.07742142677307129, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/std": 0.19245010614395142, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/mean": 0.5344164371490479, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/std": 0.41872403025627136, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/mean": 0.3333333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/std": 0.13608276844024658, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/std": 0.0, "reward": 0.8118100166320801, "reward_std": 0.2883836030960083, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.17468035593628883, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/high_max": 0.0, "clip_ratio/region_mean": 0.0, "step_time": 2294.3973380980024, "entropy_control/entropy": 0.17883179927992465, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0008, "entropy_control/zeta_next": 0.0016, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.196, "adapter_sync_time": 0.0, "generation_time": 2002.705896, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 2051.794896, "forward_backward_time": 242.502166, "rollout_overhead_time": 49.089, "peak_allocated_gib": 68.00909471511841, "peak_reserved_gib": 75.5390625 }, { "step": 50, "loss": 0.00040015578269958496, "grad_norm": 0.0654296875, "learning_rate": 2.0000000000000002e-07, "entropy": 0.1747208870947361, "clip_ratio/low_mean": 0.0015571884578093886, "clip_ratio/low_min": 0.0015571884578093886, "clip_ratio/high_mean": 0.0004357036086730659, "clip_ratio/high_max": 0.0004357036086730659, "clip_ratio/region_mean": 0.0019928920664824545, "step_time": 242.5753946049881, "entropy_control/entropy": 0.17884242490666238, "entropy_control/target": 0.20009788357278327, "entropy_control/zeta_used": 0.0016, "entropy_control/zeta_next": 0.0032, "entropy_control/epsilon_high_used": 0.28, "entropy_control/epsilon_high_next": 0.28, "epoch": 0.2, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 242.471038, "rollout_overhead_time": 0.0, "peak_allocated_gib": 68.00909471511841, "peak_reserved_gib": 75.5390625, "reward": null } ], "holdout": [ { "seed": 900000, "sample": 0, "native_reward": 0.8638203836248219, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 3216, "reward_channels": { "native": 0.8638203836248219, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "two_pool_cycle" }, { "seed": 900000, "sample": 1, "native_reward": 0.8638203836248219, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 3, "model_tokens": 5453, "reward_channels": { "native": 0.8638203836248219, "efficiency": 0.5, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "two_pool_cycle" }, { "seed": 900000, "sample": 2, "native_reward": 0.8404385754973732, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 3, "model_tokens": 4772, "reward_channels": { "native": 0.8404385754973732, "efficiency": 0.5, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "two_pool_cycle" }, { "seed": 900000, "sample": 3, "native_reward": 0.8591328836248219, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 5, "model_tokens": 4542, "reward_channels": { "native": 0.8591328836248219, "efficiency": 0.16666666666666663, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "two_pool_cycle" }, { "seed": 900001, "sample": 0, "native_reward": 0.6093825392865522, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 3848, "reward_channels": { "native": 0.6093825392865522, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "two_pool_cycle" }, { "seed": 900001, "sample": 1, "native_reward": 0.6093825392865522, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 3149, "reward_channels": { "native": 0.6093825392865522, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "two_pool_cycle" }, { "seed": 900001, "sample": 2, "native_reward": 0.6093825392865522, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 3178, "reward_channels": { "native": 0.6093825392865522, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "two_pool_cycle" }, { "seed": 900001, "sample": 3, "native_reward": 0.6093825392865522, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 3020, "reward_channels": { "native": 0.6093825392865522, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "two_pool_cycle" }, { "seed": 900000, "sample": 0, "native_reward": 0.6106601460847819, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 2891, "reward_channels": { "native": 0.6106601460847819, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "triangular_cycle" }, { "seed": 900000, "sample": 1, "native_reward": 0.8204653155141224, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 4826, "reward_channels": { "native": 0.8204653155141224, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "triangular_cycle" }, { "seed": 900000, "sample": 2, "native_reward": 0.7049643139195981, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 5670, "reward_channels": { "native": 0.7049643139195981, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "triangular_cycle" }, { "seed": 900000, "sample": 3, "native_reward": 0.6106601460847819, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 2974, "reward_channels": { "native": 0.6106601460847819, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "triangular_cycle" }, { "seed": 900001, "sample": 0, "native_reward": 0.6108851248150757, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 3040, "reward_channels": { "native": 0.6108851248150757, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "triangular_cycle" }, { "seed": 900001, "sample": 1, "native_reward": 0.8541365354407209, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 3, "model_tokens": 3943, "reward_channels": { "native": 0.8541365354407209, "efficiency": 0.5, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "triangular_cycle" }, { "seed": 900001, "sample": 2, "native_reward": 0.6108851248150757, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 2252, "reward_channels": { "native": 0.6108851248150757, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "triangular_cycle" }, { "seed": 900001, "sample": 3, "native_reward": 0.811786874720879, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 3, "model_tokens": 4686, "reward_channels": { "native": 0.811786874720879, "efficiency": 0.5, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "triangular_cycle" }, { "seed": 900000, "sample": 0, "native_reward": 0.0, "promotion_success": false, "done": false, "errors": 0, "tool_steps": 0, "model_tokens": 0, "reward_channels": { "native": 0.0, "efficiency": 0.0, "reliability": -1.0 }, "verification_level": null, "failed_gates": null, "family": "competing_cycles" }, { "seed": 900000, "sample": 1, "native_reward": 0.25, "promotion_success": false, "done": false, "errors": 0, "tool_steps": 6, "model_tokens": 5983, "reward_channels": { "native": 0.25, "efficiency": 0.0, "reliability": -1.0 }, "verification_level": null, "failed_gates": [ "finalized" ], "family": "competing_cycles" }, { "seed": 900000, "sample": 2, "native_reward": 0.6455178079500401, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 5, "model_tokens": 7567, "reward_channels": { "native": 0.6455178079500401, "efficiency": 0.16666666666666663, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "competing_cycles" }, { "seed": 900000, "sample": 3, "native_reward": 0.7201719440979103, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 5, "model_tokens": 8462, "reward_channels": { "native": 0.7201719440979103, "efficiency": 0.16666666666666663, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "competing_cycles" }, { "seed": 900001, "sample": 0, "native_reward": 0.175, "promotion_success": false, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 7579, "reward_channels": { "native": 0.175, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [ "positive_net_profit", "stress_positive", "task_complete" ], "family": "competing_cycles" }, { "seed": 900001, "sample": 1, "native_reward": 0.25, "promotion_success": false, "done": false, "errors": 0, "tool_steps": 6, "model_tokens": 5007, "reward_channels": { "native": 0.25, "efficiency": 0.0, "reliability": -1.0 }, "verification_level": null, "failed_gates": [ "finalized" ], "family": "competing_cycles" }, { "seed": 900001, "sample": 2, "native_reward": 0.7053628099527464, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 5674, "reward_channels": { "native": 0.7053628099527464, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "competing_cycles" }, { "seed": 900001, "sample": 3, "native_reward": 0.5871627452983137, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 3, "model_tokens": 4448, "reward_channels": { "native": 0.5871627452983137, "efficiency": 0.5, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "competing_cycles" }, { "seed": 900000, "sample": 0, "native_reward": 0.609376392538787, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 2135, "reward_channels": { "native": 0.609376392538787, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "four_hop_cycle" }, { "seed": 900000, "sample": 1, "native_reward": 0.609376392538787, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 3092, "reward_channels": { "native": 0.609376392538787, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "four_hop_cycle" }, { "seed": 900000, "sample": 2, "native_reward": 0.7503548418171574, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 3, "model_tokens": 4223, "reward_channels": { "native": 0.7503548418171574, "efficiency": 0.5, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "four_hop_cycle" }, { "seed": 900000, "sample": 3, "native_reward": 0.175, "promotion_success": false, "done": true, "errors": 0, "tool_steps": 3, "model_tokens": 6244, "reward_channels": { "native": 0.175, "efficiency": 0.5, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [ "positive_net_profit", "stress_positive", "task_complete" ], "family": "four_hop_cycle" }, { "seed": 900001, "sample": 0, "native_reward": 0.6095701077161815, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 3626, "reward_channels": { "native": 0.6095701077161815, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "four_hop_cycle" }, { "seed": 900001, "sample": 1, "native_reward": 0.8461963597959128, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 5285, "reward_channels": { "native": 0.8461963597959128, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "four_hop_cycle" }, { "seed": 900001, "sample": 2, "native_reward": 0.6095701077161815, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 1475, "reward_channels": { "native": 0.6095701077161815, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "four_hop_cycle" }, { "seed": 900001, "sample": 3, "native_reward": 0.8461963597959128, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 3, "model_tokens": 5229, "reward_channels": { "native": 0.8461963597959128, "efficiency": 0.5, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "four_hop_cycle" }, { "seed": 900000, "sample": 0, "native_reward": 0.8262556477476042, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 5138, "reward_channels": { "native": 0.8262556477476042, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "size_sensitive_cycle" }, { "seed": 900000, "sample": 1, "native_reward": 0.8262556477476042, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 3120, "reward_channels": { "native": 0.8262556477476042, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "size_sensitive_cycle" }, { "seed": 900000, "sample": 2, "native_reward": 0.8262556477476042, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 3146, "reward_channels": { "native": 0.8262556477476042, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "size_sensitive_cycle" }, { "seed": 900000, "sample": 3, "native_reward": 0.8262556477476042, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 5066, "reward_channels": { "native": 0.8262556477476042, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "size_sensitive_cycle" }, { "seed": 900001, "sample": 0, "native_reward": 0.8245803237223959, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 4085, "reward_channels": { "native": 0.8245803237223959, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "size_sensitive_cycle" }, { "seed": 900001, "sample": 1, "native_reward": 0.8245803237223959, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 3, "model_tokens": 3507, "reward_channels": { "native": 0.8245803237223959, "efficiency": 0.5, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "size_sensitive_cycle" }, { "seed": 900001, "sample": 2, "native_reward": 0.25, "promotion_success": false, "done": false, "errors": 0, "tool_steps": 6, "model_tokens": 3594, "reward_channels": { "native": 0.25, "efficiency": 0.0, "reliability": -1.0 }, "verification_level": null, "failed_gates": [ "finalized" ], "family": "size_sensitive_cycle" }, { "seed": 900001, "sample": 3, "native_reward": 0.7512635678492581, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 3, "model_tokens": 4819, "reward_channels": { "native": 0.7512635678492581, "efficiency": 0.5, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "size_sensitive_cycle" } ] }, { "run_id": "1789375757", "mode": "thinking", "method": "cispo-repo-r-window5", "steps": 50, "reward_points": 25, "train_mean": 0.8164995121955871, "train_last10": 0.8331568300724029, "holdout_native": 0.6243359613981289, "holdout_episodes": 40, "holdout_successes": 34, "mean_step_seconds": 936.5844565406206, "label": "CISPO + REPO-R w5", "state_at_cutoff": "complete", "step_hours": 13.008117451953064, "peak_allocated_gib": 66.8723316192627, "peak_reserved_gib": 74.375, "extra_refill_rounds": 0.0, "groups_refilled": 0.0, "holdout_component_means": { "native": 0.6243359613981289, "efficiency": 0.4416666666666667, "reliability": 0.75 }, "configuration": { "model": { "id": "Qwen/Qwen3-14B", "revision": "40c069824f4251a91eefaf281ebe4c544efd3e18", "loss": "cispo", "algorithm": "grpo", "no_think": false, "thinking_budget_tokens": 2048, "lora": true, "quantize": true, "lora_r": 32, "lora_alpha": 64, "use_dora": false }, "loss": "cispo", "normalization": "joint", "entropy_control": { "method": "repo_r", "epsilon_low": 0.2, "epsilon_high": 0.2, "zeta_initial": 0.001, "zeta_min": 0.0001, "zeta_max": 0.05, "zeta_nonnegative": true, "target_window_steps": 5, "target_entropy": null, "holdout_split": "benchmark", "holdout_seeds_per_family": 2, "num_iterations": 2 }, "dynamic_sampling": false, "sampling": { "group_size": 4, "temperature": 1.0, "max_completion_tokens": 22528, "top_entropy_quantile": 1.0, "temperature_decay_to": null }, "learning_rate": 1e-05, "turn_max_tokens": 3072 }, "source_sha256": { "metrics.jsonl": "aba98f8765bf953712762b5f8032c304291aee52ab063e4068ee57d8c02ce60b", "entropy_holdout_episodes.jsonl": "62a9b75bed62ca558b18fe8fac7ee0a41c431d2cb5517f36250c1ff941bdd1bd", "resolved_training_config.json": "efca1b48973d92e27a151d9a6a73a87272cc2a3d4c801104dad5f1d53aee1e8a" }, "training_steps": [ { "step": 1, "loss": 0.05959773063659668, "grad_norm": 0.061279296875, "learning_rate": 1e-05, "thinking/tokens_mean": 606.2361111111111, "thinking/budget_hit_fraction": 0.14583333333333334, "num_tokens": 51975.0, "completions/mean_length": 3604.375, "completions/min_length": 2890.0, "completions/max_length": 6014.0, "completions/clipped_ratio": 0.25, "completions/mean_terminated_length": 3782.166748046875, "completions/min_terminated_length": 2890.0, "completions/max_terminated_length": 6014.0, "tools/call_frequency": 2.25, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.7262155413627625, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 0.13481803238391876, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.4583333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.2500000298023224, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/mean": 0.45231834053993225, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/std": 0.526568591594696, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/mean": 0.2916666865348816, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/std": 0.34359216690063477, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/mean": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/std": 1.154700517654419, "reward": 0.7080169916152954, "reward_std": 0.5734051465988159, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 0.5, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.14539151638746262, "cispo_clip_ratio": 0.0, "step_time": 1298.5827015979999, "entropy_control/entropy": 0.20009788357278327, "entropy_control/target": 0.20009788357278327, "entropy_control/target_ready": 0.0, "entropy_control/target_samples": 1.0, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.004, "adapter_sync_time": 0.0, "generation_time": 1106.899504, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1141.579841, "forward_backward_time": 156.896813, "rollout_overhead_time": 34.680337, "peak_allocated_gib": 46.211081981658936, "peak_reserved_gib": 51.38671875 }, { "step": 2, "loss": 0.05981137603521347, "grad_norm": 0.0634765625, "learning_rate": 9.800000000000001e-06, "entropy": 0.1451741699129343, "cispo_clip_ratio": 0.002398241456830874, "step_time": 156.98477632500067, "entropy_control/entropy": 0.19986809630015606, "entropy_control/target": 0.19998298993646968, "entropy_control/target_ready": 0.0, "entropy_control/target_samples": 2.0, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.008, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 156.871494, "rollout_overhead_time": 0.0, "peak_allocated_gib": 46.689598083496094, "peak_reserved_gib": 51.484375, "reward": null }, { "step": 3, "loss": -0.027710001915693283, "grad_norm": 0.056640625, "learning_rate": 9.600000000000001e-06, "thinking/tokens_mean": 802.625, "thinking/budget_hit_fraction": 0.27678571428571425, "num_tokens": 104821.0, "completions/mean_length": 3893.75, "completions/min_length": 2023.0, "completions/max_length": 8946.0, "completions/clipped_ratio": 0.125, "completions/mean_terminated_length": 4011.285888671875, "completions/min_terminated_length": 2023.0, "completions/max_terminated_length": 8946.0, "tools/call_frequency": 2.375, "tools/failure_frequency": 0.05263157933950424, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.5108565092086792, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 0.27986934781074524, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.2519763112068176, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 0.75, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 0.7071067690849304, "reward": 0.6858565807342529, "reward_std": 0.40558508038520813, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 0.5, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.16930530779063702, "cispo_clip_ratio": 0.0, "step_time": 1804.9482219919996, "entropy_control/entropy": 0.20253203749190046, "entropy_control/target": 0.2008326724549466, "entropy_control/target_ready": 0.0, "entropy_control/target_samples": 3.0, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.012, "adapter_sync_time": 0.0, "generation_time": 1581.966809, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1620.83914, "forward_backward_time": 183.99845, "rollout_overhead_time": 38.872331, "peak_allocated_gib": 54.23228740692139, "peak_reserved_gib": 60.21484375 }, { "step": 4, "loss": -0.027300480753183365, "grad_norm": 0.056884765625, "learning_rate": 9.4e-06, "entropy": 0.16915154829621315, "cispo_clip_ratio": 0.0035100987879559398, "step_time": 184.09639192699933, "entropy_control/entropy": 0.2024014143879632, "entropy_control/target": 0.20122485793820075, "entropy_control/target_ready": 0.0, "entropy_control/target_samples": 4.0, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.016, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 183.99314, "rollout_overhead_time": 0.0, "peak_allocated_gib": 54.23228740692139, "peak_reserved_gib": 60.21484375, "reward": null }, { "step": 5, "loss": 0.010351837612688541, "grad_norm": 0.05517578125, "learning_rate": 9.200000000000002e-06, "thinking/tokens_mean": 604.78125, "thinking/budget_hit_fraction": 0.1919642857142857, "num_tokens": 147129.0, "completions/mean_length": 2583.5, "completions/min_length": 1535.0, "completions/max_length": 4630.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 2583.5, "completions/min_terminated_length": 1535.0, "completions/max_terminated_length": 4630.0, "tools/call_frequency": 2.125, "tools/failure_frequency": 0.11764705926179886, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.562455415725708, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 0.24158011376857758, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.5625, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.23464767634868622, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 0.75, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 0.7071067690849304, "reward": 0.7405804395675659, "reward_std": 0.38817667961120605, "frac_reward_zero_std": 0.5, "frac_groups_zero_std": 0.5, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 0.5, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.2071676291525364, "cispo_clip_ratio": 0.0, "step_time": 896.6167426029979, "entropy_control/entropy": 0.21937658947497593, "entropy_control/target": 0.20485520424555576, "entropy_control/target_ready": 1.0, "entropy_control/target_samples": 5.0, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.001, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.02, "adapter_sync_time": 0.0, "generation_time": 801.331517, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 819.376388, "forward_backward_time": 77.119401, "rollout_overhead_time": 18.044871, "peak_allocated_gib": 54.23228740692139, "peak_reserved_gib": 60.21484375 }, { "step": 6, "loss": 0.010341729037463665, "grad_norm": 0.05224609375, "learning_rate": 9e-06, "entropy": 0.2073134146630764, "cispo_clip_ratio": 0.002112114743795246, "step_time": 77.16678110800058, "entropy_control/entropy": 0.21951732770760654, "entropy_control/target": 0.20485520424555576, "entropy_control/target_ready": 1.0, "entropy_control/target_samples": 5.0, "entropy_control/zeta_used": 0.001, "entropy_control/zeta_next": 0.0005, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.024, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 77.058135, "rollout_overhead_time": 0.0, "peak_allocated_gib": 54.23228740692139, "peak_reserved_gib": 60.21484375, "reward": null }, { "step": 7, "loss": 0.0929565355181694, "grad_norm": 0.0311279296875, "learning_rate": 8.8e-06, "thinking/tokens_mean": 837.4722222222222, "thinking/budget_hit_fraction": 0.17361111111111113, "num_tokens": 211265.0, "completions/mean_length": 4549.5, "completions/min_length": 2944.0, "completions/max_length": 6572.0, "completions/clipped_ratio": 0.25, "completions/mean_terminated_length": 5042.33349609375, "completions/min_terminated_length": 2944.0, "completions/max_terminated_length": 6572.0, "tools/call_frequency": 2.5, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/mean": 0.5304248929023743, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/std": 0.35364797711372375, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/mean": 0.25, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/std": 0.21516574919223785, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/std": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/mean": 0.6533555388450623, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/std": 0.44027379155158997, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/mean": 0.4166666865348816, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/std": 0.3191424012184143, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/std": 1.0, "reward": 0.7085568904876709, "reward_std": 0.5677993297576904, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 0.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.16554154455661774, "cispo_clip_ratio": 0.0, "step_time": 1950.7574458720019, "entropy_control/entropy": 0.2220573221806916, "entropy_control/target": 0.20485520424555576, "entropy_control/target_ready": 1.0, "entropy_control/target_samples": 5.0, "entropy_control/zeta_used": 0.0005, "entropy_control/zeta_next": 0.00025, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.028, "adapter_sync_time": 0.0, "generation_time": 1655.079841, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1705.294165, "forward_backward_time": 245.354171, "rollout_overhead_time": 50.214325, "peak_allocated_gib": 61.0336332321167, "peak_reserved_gib": 67.94921875 }, { "step": 8, "loss": 0.09264954924583435, "grad_norm": 0.03173828125, "learning_rate": 8.6e-06, "entropy": 0.16547657921910286, "cispo_clip_ratio": 0.003446480754064396, "step_time": 245.43896711900197, "entropy_control/entropy": 0.22195000362938813, "entropy_control/target": 0.20485520424555576, "entropy_control/target_ready": 1.0, "entropy_control/target_samples": 5.0, "entropy_control/zeta_used": 0.00025, "entropy_control/zeta_next": 0.000125, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.032, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 245.334972, "rollout_overhead_time": 0.0, "peak_allocated_gib": 61.0336332321167, "peak_reserved_gib": 67.94921875, "reward": null }, { "step": 9, "loss": 0.0013452228158712387, "grad_norm": 0.05908203125, "learning_rate": 8.400000000000001e-06, "thinking/tokens_mean": 592.6428571428571, "thinking/budget_hit_fraction": 0.125, "num_tokens": 267115.0, "completions/mean_length": 3697.25, "completions/min_length": 2668.0, "completions/max_length": 5423.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 3697.25, "completions/min_terminated_length": 2668.0, "completions/max_terminated_length": 5423.0, "tools/call_frequency": 2.625, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.609546422958374, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.6666666865348816, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/mean": 0.5474828481674194, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/std": 0.2035110741853714, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/mean": 0.4583333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/std": 0.3154948949813843, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/std": 1.0, "reward": 0.7566396594047546, "reward_std": 0.2873321771621704, "frac_reward_zero_std": 0.5, "frac_groups_zero_std": 0.5, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.20267178118228912, "cispo_clip_ratio": 0.0, "step_time": 1570.2471904950007, "entropy_control/entropy": 0.20323615243261975, "entropy_control/target": 0.20485520424555576, "entropy_control/target_ready": 1.0, "entropy_control/target_samples": 5.0, "entropy_control/zeta_used": 0.000125, "entropy_control/zeta_next": 0.00025, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.036, "adapter_sync_time": 0.0, "generation_time": 1296.699372, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1343.478417, "forward_backward_time": 226.666845, "rollout_overhead_time": 46.779044, "peak_allocated_gib": 61.0336332321167, "peak_reserved_gib": 67.94921875 }, { "step": 10, "loss": 0.0013723522424697876, "grad_norm": 0.0712890625, "learning_rate": 8.2e-06, "entropy": 0.20214428007602692, "cispo_clip_ratio": 0.0015592083218507469, "step_time": 226.77468287900047, "entropy_control/entropy": 0.2026540667355008, "entropy_control/target": 0.20485520424555576, "entropy_control/target_ready": 1.0, "entropy_control/target_samples": 5.0, "entropy_control/zeta_used": 0.00025, "entropy_control/zeta_next": 0.0005, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.04, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 226.668711, "rollout_overhead_time": 0.0, "peak_allocated_gib": 61.0336332321167, "peak_reserved_gib": 67.94921875, "reward": null }, { "step": 11, "loss": 0.020669659599661827, "grad_norm": 0.0888671875, "learning_rate": 8.000000000000001e-06, "thinking/tokens_mean": 526.9166666666667, "thinking/budget_hit_fraction": 0.1, "num_tokens": 317201.0, "completions/mean_length": 2951.75, "completions/min_length": 1729.0, "completions/max_length": 4615.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 2951.75, "completions/min_terminated_length": 1729.0, "completions/max_terminated_length": 4615.0, "tools/call_frequency": 2.625, "tools/failure_frequency": 0.0476190485060215, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__native/mean": 0.6350884437561035, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__native/std": 0.04785403609275818, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__efficiency/mean": 0.5833333730697632, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__efficiency/std": 0.1666666716337204, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/mean": 0.7017019987106323, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/std": 0.11484745144844055, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/mean": 0.5833333730697632, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/std": 0.09622505307197571, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/std": 0.0, "reward": 0.8975619077682495, "reward_std": 0.08460775762796402, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.20168504863977432, "cispo_clip_ratio": 0.0, "step_time": 1191.7877642379972, "entropy_control/entropy": 0.20284171005910367, "entropy_control/target": 0.20485520424555576, "entropy_control/target_ready": 1.0, "entropy_control/target_samples": 5.0, "entropy_control/zeta_used": 0.0005, "entropy_control/zeta_next": 0.001, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.044, "adapter_sync_time": 0.0, "generation_time": 1039.792731, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1067.322069, "forward_backward_time": 124.356247, "rollout_overhead_time": 27.529338, "peak_allocated_gib": 61.0336332321167, "peak_reserved_gib": 67.96875 }, { "step": 12, "loss": 0.02032589167356491, "grad_norm": 0.087890625, "learning_rate": 7.800000000000002e-06, "entropy": 0.2027429062873125, "cispo_clip_ratio": 0.002254660299513489, "step_time": 124.46118573400054, "entropy_control/entropy": 0.20381974235789269, "entropy_control/target": 0.20485520424555576, "entropy_control/target_ready": 1.0, "entropy_control/target_samples": 5.0, "entropy_control/zeta_used": 0.001, "entropy_control/zeta_next": 0.002, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.048, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 124.358736, "rollout_overhead_time": 0.0, "peak_allocated_gib": 61.0336332321167, "peak_reserved_gib": 67.96875, "reward": null }, { "step": 13, "loss": -0.0024953139945864677, "grad_norm": 0.06787109375, "learning_rate": 7.600000000000001e-06, "thinking/tokens_mean": 610.6904761904761, "thinking/budget_hit_fraction": 0.10714285714285714, "num_tokens": 376359.0, "completions/mean_length": 4109.75, "completions/min_length": 2108.0, "completions/max_length": 7360.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 4109.75, "completions/min_terminated_length": 2108.0, "completions/max_terminated_length": 7360.0, "tools/call_frequency": 3.25, "tools/failure_frequency": 0.03846153989434242, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.7278944253921509, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 0.1366470754146576, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.19245010614395142, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/mean": 0.5482227802276611, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/std": 0.20462064445018768, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/mean": 0.4583333730697632, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/std": 0.3154948949813843, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/std": 1.0, "reward": 0.8120169043540955, "reward_std": 0.32238584756851196, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.18441541120409966, "cispo_clip_ratio": 0.0, "step_time": 1760.5267777860026, "entropy_control/entropy": 0.1900139432682572, "entropy_control/target": 0.20485520424555576, "entropy_control/target_ready": 1.0, "entropy_control/target_samples": 5.0, "entropy_control/zeta_used": 0.002, "entropy_control/zeta_next": 0.004, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.052, "adapter_sync_time": 0.0, "generation_time": 1456.105935, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1507.811771, "forward_backward_time": 252.613774, "rollout_overhead_time": 51.705836, "peak_allocated_gib": 62.163169384002686, "peak_reserved_gib": 68.69140625 }, { "step": 14, "loss": -0.0014013415202498436, "grad_norm": 0.06787109375, "learning_rate": 7.4e-06, "entropy": 0.1845948975533247, "cispo_clip_ratio": 0.004129788081627339, "step_time": 252.7138389949996, "entropy_control/entropy": 0.19006710776945815, "entropy_control/target": 0.20485520424555576, "entropy_control/target_ready": 1.0, "entropy_control/target_samples": 5.0, "entropy_control/zeta_used": 0.004, "entropy_control/zeta_next": 0.008, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.056, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 252.610297, "rollout_overhead_time": 0.0, "peak_allocated_gib": 62.163169384002686, "peak_reserved_gib": 68.69140625, "reward": null }, { "step": 15, "loss": -0.020297521725296974, "grad_norm": 0.050537109375, "learning_rate": 7.2000000000000005e-06, "thinking/tokens_mean": 758.5357142857143, "thinking/budget_hit_fraction": 0.19642857142857142, "num_tokens": 450060.0, "completions/mean_length": 5729.125, "completions/min_length": 4202.0, "completions/max_length": 6894.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 5729.125, "completions/min_terminated_length": 4202.0, "completions/max_terminated_length": 6894.0, "tools/call_frequency": 4.125, "tools/failure_frequency": 0.03030303120613098, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/mean": 0.7659832835197449, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/std": 0.1288890242576599, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/mean": 0.375, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/std": 0.15957118570804596, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/mean": 0.516674816608429, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/std": 0.3525680899620056, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/mean": 0.2916666567325592, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/std": 0.25, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/std": 1.0, "reward": 0.8079956769943237, "reward_std": 0.38156166672706604, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.1951613910496235, "cispo_clip_ratio": 0.0, "step_time": 2033.3330594330037, "entropy_control/entropy": 0.1983596866809788, "entropy_control/target": 0.20485520424555576, "entropy_control/target_ready": 1.0, "entropy_control/target_samples": 5.0, "entropy_control/zeta_used": 0.008, "entropy_control/zeta_next": 0.016, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.06, "adapter_sync_time": 0.0, "generation_time": 1737.74738, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1788.325392, "forward_backward_time": 244.896591, "rollout_overhead_time": 50.578012, "peak_allocated_gib": 62.163169384002686, "peak_reserved_gib": 68.69140625 }, { "step": 16, "loss": -0.01815813221037388, "grad_norm": 0.051513671875, "learning_rate": 7e-06, "entropy": 0.19524658098816872, "cispo_clip_ratio": 0.002333718177396804, "step_time": 244.97779498899945, "entropy_control/entropy": 0.19852210625771613, "entropy_control/target": 0.20485520424555576, "entropy_control/target_ready": 1.0, "entropy_control/target_samples": 5.0, "entropy_control/zeta_used": 0.016, "entropy_control/zeta_next": 0.032, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.064, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 244.863429, "rollout_overhead_time": 0.0, "peak_allocated_gib": 62.163169384002686, "peak_reserved_gib": 68.69140625, "reward": null }, { "step": 17, "loss": 0.03386733680963516, "grad_norm": 0.055908203125, "learning_rate": 6.800000000000001e-06, "thinking/tokens_mean": 688.922619047619, "thinking/budget_hit_fraction": 0.16071428571428573, "num_tokens": 519466.0, "completions/mean_length": 4984.75, "completions/min_length": 2691.0, "completions/max_length": 7173.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 4984.75, "completions/min_terminated_length": 2691.0, "completions/max_terminated_length": 7173.0, "tools/call_frequency": 3.5, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/mean": 0.7994208335876465, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/std": 0.1643586903810501, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/mean": 0.375, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/std": 0.0833333283662796, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/mean": 0.556321382522583, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/std": 0.21588364243507385, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/mean": 0.4583333730697632, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/std": 0.3154948949813843, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/std": 1.0, "reward": 0.84870445728302, "reward_std": 0.34821492433547974, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.2243624310940504, "cispo_clip_ratio": 0.0, "step_time": 1850.8885259510007, "entropy_control/entropy": 0.22530296515731843, "entropy_control/target": 0.20485520424555576, "entropy_control/target_ready": 1.0, "entropy_control/target_samples": 5.0, "entropy_control/zeta_used": 0.032, "entropy_control/zeta_next": 0.016, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.068, "adapter_sync_time": 0.0, "generation_time": 1590.128505, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1635.09638, "forward_backward_time": 215.673129, "rollout_overhead_time": 44.967875, "peak_allocated_gib": 62.163169384002686, "peak_reserved_gib": 68.69140625 }, { "step": 18, "loss": 0.02839519828557968, "grad_norm": 0.054931640625, "learning_rate": 6.600000000000001e-06, "entropy": 0.22481019608676434, "cispo_clip_ratio": 0.003336210415000096, "step_time": 215.79665486700105, "entropy_control/entropy": 0.22577012742912464, "entropy_control/target": 0.20485520424555576, "entropy_control/target_ready": 1.0, "entropy_control/target_samples": 5.0, "entropy_control/zeta_used": 0.016, "entropy_control/zeta_next": 0.008, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.072, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 215.671206, "rollout_overhead_time": 0.0, "peak_allocated_gib": 62.163169384002686, "peak_reserved_gib": 68.69140625, "reward": null }, { "step": 19, "loss": 0.06444666534662247, "grad_norm": 0.0693359375, "learning_rate": 6.4000000000000006e-06, "thinking/tokens_mean": 622.8214285714286, "thinking/budget_hit_fraction": 0.14285714285714285, "num_tokens": 575231.0, "completions/mean_length": 4080.125, "completions/min_length": 2023.0, "completions/max_length": 7409.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 4080.125, "completions/min_terminated_length": 2023.0, "completions/max_terminated_length": 7409.0, "tools/call_frequency": 3.125, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.6687541007995605, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 0.11813405156135559, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.625, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.0833333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/mean": 0.902118444442749, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/std": 0.08801792562007904, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/mean": 0.3333333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/std": 0.38490021228790283, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/std": 0.0, "reward": 1.009394645690918, "reward_std": 0.14584001898765564, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.19914155080914497, "cispo_clip_ratio": 0.0, "step_time": 1785.1697965930034, "entropy_control/entropy": 0.2037297005870767, "entropy_control/target": 0.20485520424555576, "entropy_control/target_ready": 1.0, "entropy_control/target_samples": 5.0, "entropy_control/zeta_used": 0.008, "entropy_control/zeta_next": 0.016, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.076, "adapter_sync_time": 0.0, "generation_time": 1489.654155, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1539.884745, "forward_backward_time": 245.159618, "rollout_overhead_time": 50.23059, "peak_allocated_gib": 63.00174951553345, "peak_reserved_gib": 69.9609375 }, { "step": 20, "loss": 0.06706183403730392, "grad_norm": 0.08056640625, "learning_rate": 6.200000000000001e-06, "entropy": 0.1990173440426588, "cispo_clip_ratio": 0.0015563149063382298, "step_time": 245.27781082600268, "entropy_control/entropy": 0.2036310221385924, "entropy_control/target": 0.20485520424555576, "entropy_control/target_ready": 1.0, "entropy_control/target_samples": 5.0, "entropy_control/zeta_used": 0.016, "entropy_control/zeta_next": 0.032, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.08, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 245.159189, "rollout_overhead_time": 0.0, "peak_allocated_gib": 63.00174951553345, "peak_reserved_gib": 69.9609375, "reward": null }, { "step": 21, "loss": 0.01057123951613903, "grad_norm": 0.059814453125, "learning_rate": 6e-06, "thinking/tokens_mean": 682.9277210884354, "thinking/budget_hit_fraction": 0.17857142857142858, "num_tokens": 645527.0, "completions/mean_length": 5088.0, "completions/min_length": 3594.0, "completions/max_length": 7083.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 5088.0, "completions/min_terminated_length": 3594.0, "completions/max_terminated_length": 7083.0, "tools/call_frequency": 3.75, "tools/failure_frequency": 0.13333334028720856, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/mean": 0.6140941381454468, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/mean": 0.5833333730697632, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/std": 0.09622505307197571, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/mean": 0.7966060638427734, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/std": 0.05463671311736107, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/mean": 0.3333333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/std": 0.2357022762298584, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/std": 0.0, "reward": 0.9282667636871338, "reward_std": 0.09932444244623184, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.21079818531870842, "cispo_clip_ratio": 0.0, "step_time": 2104.7825938940005, "entropy_control/entropy": 0.20509145250836291, "entropy_control/target": 0.20485520424555576, "entropy_control/target_ready": 1.0, "entropy_control/target_samples": 5.0, "entropy_control/zeta_used": 0.032, "entropy_control/zeta_next": 0.016, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.084, "adapter_sync_time": 0.0, "generation_time": 1793.906321, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1846.446981, "forward_backward_time": 258.226986, "rollout_overhead_time": 52.54066, "peak_allocated_gib": 63.00174951553345, "peak_reserved_gib": 69.9609375 }, { "step": 22, "loss": 0.005065605044364929, "grad_norm": 0.05712890625, "learning_rate": 5.8e-06, "entropy": 0.21090805158019066, "cispo_clip_ratio": 0.0024892244837246835, "step_time": 258.36378274600065, "entropy_control/entropy": 0.2051834969096493, "entropy_control/target": 0.20485520424555576, "entropy_control/target_ready": 1.0, "entropy_control/target_samples": 5.0, "entropy_control/zeta_used": 0.016, "entropy_control/zeta_next": 0.008, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.088, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 258.25595, "rollout_overhead_time": 0.0, "peak_allocated_gib": 63.00174951553345, "peak_reserved_gib": 69.9609375, "reward": null }, { "step": 23, "loss": 0.06523437798023224, "grad_norm": 0.0751953125, "learning_rate": 5.600000000000001e-06, "thinking/tokens_mean": 687.2166666666667, "thinking/budget_hit_fraction": 0.15, "num_tokens": 697274.0, "completions/mean_length": 3582.875, "completions/min_length": 2165.0, "completions/max_length": 5790.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 3582.875, "completions/min_terminated_length": 2165.0, "completions/max_terminated_length": 5790.0, "tools/call_frequency": 2.625, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.6682802438735962, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 0.11743566393852234, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.625, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.0833333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/mean": 0.8887895345687866, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/std": 0.07352567464113235, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/std": 0.19245010614395142, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/std": 0.0, "reward": 1.006659984588623, "reward_std": 0.1427164375782013, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.19497286900877953, "cispo_clip_ratio": 0.0, "step_time": 1244.0834711949938, "entropy_control/entropy": 0.20386093177247533, "entropy_control/target": 0.20485520424555576, "entropy_control/target_ready": 1.0, "entropy_control/target_samples": 5.0, "entropy_control/zeta_used": 0.008, "entropy_control/zeta_next": 0.016, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.092, "adapter_sync_time": 0.0, "generation_time": 1058.673756, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1091.555687, "forward_backward_time": 152.427777, "rollout_overhead_time": 32.881931, "peak_allocated_gib": 63.00174951553345, "peak_reserved_gib": 69.9609375 }, { "step": 24, "loss": 0.06809286773204803, "grad_norm": 0.1025390625, "learning_rate": 5.400000000000001e-06, "entropy": 0.19526192545890808, "cispo_clip_ratio": 0.0016663168498780578, "step_time": 152.53275930000018, "entropy_control/entropy": 0.2041409603798984, "entropy_control/target": 0.20485520424555576, "entropy_control/target_ready": 1.0, "entropy_control/target_samples": 5.0, "entropy_control/zeta_used": 0.016, "entropy_control/zeta_next": 0.032, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.096, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 152.430022, "rollout_overhead_time": 0.0, "peak_allocated_gib": 63.00174951553345, "peak_reserved_gib": 69.9609375, "reward": null }, { "step": 25, "loss": 0.054781217128038406, "grad_norm": 0.06298828125, "learning_rate": 5.2e-06, "thinking/tokens_mean": 681.1017006802721, "thinking/budget_hit_fraction": 0.14795918367346936, "num_tokens": 758753.0, "completions/mean_length": 4594.875, "completions/min_length": 2531.0, "completions/max_length": 7613.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 4594.875, "completions/min_terminated_length": 2531.0, "completions/max_terminated_length": 7613.0, "tools/call_frequency": 3.25, "tools/failure_frequency": 0.07692307978868484, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.6674500703811646, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 0.1163635328412056, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.625, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.0833333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/mean": 0.4114094376564026, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/std": 0.35037869215011597, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/mean": 0.125, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/std": 0.25, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/mean": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/std": 1.154700517654419, "reward": 0.658179759979248, "reward_std": 0.46525639295578003, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 0.5, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.1866621244698763, "cispo_clip_ratio": 0.0, "step_time": 2135.4792062919987, "entropy_control/entropy": 0.18523013277229308, "entropy_control/target": 0.20485520424555576, "entropy_control/target_ready": 1.0, "entropy_control/target_samples": 5.0, "entropy_control/zeta_used": 0.032, "entropy_control/zeta_next": 0.05, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.1, "adapter_sync_time": 0.0, "generation_time": 1832.699307, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1884.293525, "forward_backward_time": 251.07364, "rollout_overhead_time": 51.594217, "peak_allocated_gib": 63.00174951553345, "peak_reserved_gib": 69.9609375 }, { "step": 26, "loss": 0.06057373806834221, "grad_norm": 0.0673828125, "learning_rate": 5e-06, "entropy": 0.18667729198932648, "cispo_clip_ratio": 0.0015251196455210447, "step_time": 251.15780723399803, "entropy_control/entropy": 0.18529547279088113, "entropy_control/target": 0.20485520424555576, "entropy_control/target_ready": 1.0, "entropy_control/target_samples": 5.0, "entropy_control/zeta_used": 0.05, "entropy_control/zeta_next": 0.05, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.104, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 251.049987, "rollout_overhead_time": 0.0, "peak_allocated_gib": 63.00174951553345, "peak_reserved_gib": 69.9609375, "reward": null }, { "step": 27, "loss": -0.00391291081905365, "grad_norm": 0.07763671875, "learning_rate": 4.800000000000001e-06, "thinking/tokens_mean": 517.8619047619047, "thinking/budget_hit_fraction": 0.07142857142857142, "num_tokens": 816208.0, "completions/mean_length": 3879.375, "completions/min_length": 2891.0, "completions/max_length": 4980.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 3879.375, "completions/min_terminated_length": 2891.0, "completions/max_terminated_length": 4980.0, "tools/call_frequency": 3.625, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/mean": 0.6411614418029785, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/std": 0.26617082953453064, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/mean": 0.3333333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/std": 0.3042903244495392, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/std": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/mean": 0.681270956993103, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/std": 0.28751400113105774, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/mean": 0.4583333730697632, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/std": 0.3154948949813843, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/std": 1.0, "reward": 0.7810078859329224, "reward_std": 0.45390626788139343, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.17713535204529762, "cispo_clip_ratio": 0.0, "step_time": 1637.9102483940005, "entropy_control/entropy": 0.17816867897465358, "entropy_control/target": 0.20485520424555576, "entropy_control/target_ready": 1.0, "entropy_control/target_samples": 5.0, "entropy_control/zeta_used": 0.05, "entropy_control/zeta_next": 0.05, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.108, "adapter_sync_time": 0.0, "generation_time": 1376.580796, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1421.6892, "forward_backward_time": 216.112246, "rollout_overhead_time": 45.108404, "peak_allocated_gib": 63.00174951553345, "peak_reserved_gib": 69.9609375 }, { "step": 28, "loss": -0.003446172922849655, "grad_norm": 0.07568359375, "learning_rate": 4.600000000000001e-06, "entropy": 0.17688164487481117, "cispo_clip_ratio": 0.004137578274821863, "step_time": 216.23619884599975, "entropy_control/entropy": 0.1778971547315059, "entropy_control/target": 0.20485520424555576, "entropy_control/target_ready": 1.0, "entropy_control/target_samples": 5.0, "entropy_control/zeta_used": 0.05, "entropy_control/zeta_next": 0.05, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.112, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 216.122646, "rollout_overhead_time": 0.0, "peak_allocated_gib": 63.00174951553345, "peak_reserved_gib": 69.9609375, "reward": null }, { "step": 29, "loss": 0.07119542360305786, "grad_norm": 0.458984375, "learning_rate": 4.4e-06, "thinking/tokens_mean": 809.1194444444445, "thinking/budget_hit_fraction": 0.15178571428571427, "num_tokens": 880463.0, "completions/mean_length": 4749.375, "completions/min_length": 3071.0, "completions/max_length": 7169.0, "completions/clipped_ratio": 0.125, "completions/mean_terminated_length": 4989.14306640625, "completions/min_terminated_length": 3259.0, "completions/max_terminated_length": 7169.0, "tools/call_frequency": 2.875, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.41142311692237854, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 0.39460644125938416, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.4166666865348816, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.2886751592159271, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/mean": 0.7119555473327637, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/std": 0.11502954363822937, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/mean": 0.375, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/std": 0.20971763134002686, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/std": 0.0, "reward": 0.7314809560775757, "reward_std": 0.43306535482406616, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 0.5, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.19427763298153877, "cispo_clip_ratio": 0.0, "step_time": 1882.0860976619952, "entropy_control/entropy": 0.2256108444236916, "entropy_control/target": 0.20485520424555576, "entropy_control/target_ready": 1.0, "entropy_control/target_samples": 5.0, "entropy_control/zeta_used": 0.05, "entropy_control/zeta_next": 0.025, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.116, "adapter_sync_time": 0.0, "generation_time": 1605.322095, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1652.758794, "forward_backward_time": 229.217038, "rollout_overhead_time": 47.436699, "peak_allocated_gib": 63.00174951553345, "peak_reserved_gib": 69.9609375 }, { "step": 30, "loss": 0.06267890334129333, "grad_norm": 0.0634765625, "learning_rate": 4.2000000000000004e-06, "entropy": 0.19474821910262108, "cispo_clip_ratio": 0.002466427569743246, "step_time": 229.31584244400437, "entropy_control/entropy": 0.22615603252093686, "entropy_control/target": 0.20485520424555576, "entropy_control/target_ready": 1.0, "entropy_control/target_samples": 5.0, "entropy_control/zeta_used": 0.025, "entropy_control/zeta_next": 0.0125, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.12, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 229.206883, "rollout_overhead_time": 0.0, "peak_allocated_gib": 63.00174951553345, "peak_reserved_gib": 69.9609375, "reward": null }, { "step": 31, "loss": 0.004682481288909912, "grad_norm": 0.11083984375, "learning_rate": 4.000000000000001e-06, "thinking/tokens_mean": 1023.53125, "thinking/budget_hit_fraction": 0.28125, "num_tokens": 940497.0, "completions/mean_length": 4420.75, "completions/min_length": 2259.0, "completions/max_length": 7356.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 4420.75, "completions/min_terminated_length": 2259.0, "completions/max_terminated_length": 7356.0, "tools/call_frequency": 2.5, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/mean": 0.7605566382408142, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/std": 0.2414233684539795, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/mean": 0.5833333730697632, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/std": 0.0890870913863182, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/std": 0.0, "reward": 0.989723265171051, "reward_std": 0.24271216988563538, "frac_reward_zero_std": 0.5, "frac_groups_zero_std": 0.5, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.1941378451883793, "cispo_clip_ratio": 0.0, "step_time": 1461.0723787209972, "entropy_control/entropy": 0.20328535505643772, "entropy_control/target": 0.20485520424555576, "entropy_control/target_ready": 1.0, "entropy_control/target_samples": 5.0, "entropy_control/zeta_used": 0.0125, "entropy_control/zeta_next": 0.025, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.124, "adapter_sync_time": 0.0, "generation_time": 1281.218686, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1313.171535, "forward_backward_time": 147.795266, "rollout_overhead_time": 31.952849, "peak_allocated_gib": 63.00174951553345, "peak_reserved_gib": 69.9609375 }, { "step": 32, "loss": 0.007850755006074905, "grad_norm": 0.047119140625, "learning_rate": 3.8000000000000005e-06, "entropy": 0.19449243694543839, "cispo_clip_ratio": 0.0018772510520648211, "step_time": 148.00937268200505, "entropy_control/entropy": 0.203654636444159, "entropy_control/target": 0.20485520424555576, "entropy_control/target_ready": 1.0, "entropy_control/target_samples": 5.0, "entropy_control/zeta_used": 0.025, "entropy_control/zeta_next": 0.05, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.128, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 147.896003, "rollout_overhead_time": 0.0, "peak_allocated_gib": 63.00174951553345, "peak_reserved_gib": 69.9609375, "reward": null }, { "step": 33, "loss": -0.017484482377767563, "grad_norm": 0.07080078125, "learning_rate": 3.6000000000000003e-06, "thinking/tokens_mean": 870.40625, "thinking/budget_hit_fraction": 0.28125, "num_tokens": 994370.0, "completions/mean_length": 3838.625, "completions/min_length": 2761.0, "completions/max_length": 5601.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 3838.625, "completions/min_terminated_length": 2761.0, "completions/max_terminated_length": 5601.0, "tools/call_frequency": 2.375, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.6095701456069946, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.625, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.0833333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/mean": 0.8249396085739136, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/mean": 0.5833333730697632, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/std": 0.09622505307197571, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/std": 0.0, "reward": 0.9474632143974304, "reward_std": 0.11408235877752304, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.20731479860842228, "cispo_clip_ratio": 0.0, "step_time": 1291.4547260530017, "entropy_control/entropy": 0.20557631052707728, "entropy_control/target": 0.20485520424555576, "entropy_control/target_ready": 1.0, "entropy_control/target_samples": 5.0, "entropy_control/zeta_used": 0.05, "entropy_control/zeta_next": 0.025, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.132, "adapter_sync_time": 0.0, "generation_time": 1142.442221, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1169.44514, "forward_backward_time": 121.904203, "rollout_overhead_time": 27.002919, "peak_allocated_gib": 63.00174951553345, "peak_reserved_gib": 69.9609375 }, { "step": 34, "loss": -0.02583444118499756, "grad_norm": 0.06787109375, "learning_rate": 3.4000000000000005e-06, "entropy": 0.20729834772646427, "cispo_clip_ratio": 0.0020852537418249995, "step_time": 121.99650523900709, "entropy_control/entropy": 0.20556026268745775, "entropy_control/target": 0.20485520424555576, "entropy_control/target_ready": 1.0, "entropy_control/target_samples": 5.0, "entropy_control/zeta_used": 0.025, "entropy_control/zeta_next": 0.0125, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.136, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 121.890285, "rollout_overhead_time": 0.0, "peak_allocated_gib": 63.00174951553345, "peak_reserved_gib": 69.9609375, "reward": null }, { "step": 35, "loss": 0.06391463428735733, "grad_norm": 0.046630859375, "learning_rate": 3.2000000000000003e-06, "thinking/tokens_mean": 736.1457482993197, "thinking/budget_hit_fraction": 0.1683673469387755, "num_tokens": 1065792.0, "completions/mean_length": 5064.75, "completions/min_length": 3071.0, "completions/max_length": 8135.0, "completions/clipped_ratio": 0.125, "completions/mean_terminated_length": 5349.57177734375, "completions/min_terminated_length": 3388.0, "completions/max_terminated_length": 8135.0, "tools/call_frequency": 3.75, "tools/failure_frequency": 0.03333333507180214, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/mean": 0.5844686031341553, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/std": 0.252232164144516, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/mean": 0.2708333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/std": 0.294627845287323, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/mean": 0.75, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/std": 0.7071067690849304, "reward": 0.7480102777481079, "reward_std": 0.3923490345478058, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 0.5, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.17276779748499393, "cispo_clip_ratio": 0.0, "step_time": 2276.6506014410006, "entropy_control/entropy": 0.19601446939552125, "entropy_control/target": 0.20485520424555576, "entropy_control/target_ready": 1.0, "entropy_control/target_samples": 5.0, "entropy_control/zeta_used": 0.0125, "entropy_control/zeta_next": 0.025, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.14, "adapter_sync_time": 0.0, "generation_time": 1927.29938, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1986.253628, "forward_backward_time": 290.28521, "rollout_overhead_time": 58.954248, "peak_allocated_gib": 66.8723316192627, "peak_reserved_gib": 74.375 }, { "step": 36, "loss": 0.06670805811882019, "grad_norm": 0.07177734375, "learning_rate": 3e-06, "entropy": 0.1726672165095806, "cispo_clip_ratio": 0.0018082544847857207, "step_time": 290.36681058300746, "entropy_control/entropy": 0.19595297285575342, "entropy_control/target": 0.20485520424555576, "entropy_control/target_ready": 1.0, "entropy_control/target_samples": 5.0, "entropy_control/zeta_used": 0.025, "entropy_control/zeta_next": 0.05, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.144, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 290.252721, "rollout_overhead_time": 0.0, "peak_allocated_gib": 66.8723316192627, "peak_reserved_gib": 74.375, "reward": null }, { "step": 37, "loss": -0.0004202350974082947, "grad_norm": 0.06591796875, "learning_rate": 2.8000000000000003e-06, "thinking/tokens_mean": 637.6488095238095, "thinking/budget_hit_fraction": 0.14285714285714285, "num_tokens": 1123206.0, "completions/mean_length": 4296.75, "completions/min_length": 1682.0, "completions/max_length": 7214.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 4296.75, "completions/min_terminated_length": 1682.0, "completions/max_terminated_length": 7214.0, "tools/call_frequency": 3.25, "tools/failure_frequency": 0.03846153989434242, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.5197777152061462, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 0.17985181510448456, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.3333333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__native/mean": 0.3499770760536194, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__native/std": 0.41055575013160706, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__efficiency/mean": 0.1666666716337204, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__efficiency/std": 0.3333333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__reliability/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__reliability/std": 1.0, "reward": 0.5515440702438354, "reward_std": 0.45853644609451294, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 0.5, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.2026723064482212, "cispo_clip_ratio": 0.0, "step_time": 1781.7185949529885, "entropy_control/entropy": 0.20316437525962658, "entropy_control/target": 0.20485520424555576, "entropy_control/target_ready": 1.0, "entropy_control/target_samples": 5.0, "entropy_control/zeta_used": 0.05, "entropy_control/zeta_next": 0.05, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.148, "adapter_sync_time": 0.0, "generation_time": 1534.11242, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1577.213866, "forward_backward_time": 204.400545, "rollout_overhead_time": 43.101447, "peak_allocated_gib": 66.8723316192627, "peak_reserved_gib": 74.375 }, { "step": 38, "loss": -0.00047516636550426483, "grad_norm": 0.06982421875, "learning_rate": 2.6e-06, "entropy": 0.20305371284484863, "cispo_clip_ratio": 0.0031116618192754686, "step_time": 204.5068822660105, "entropy_control/entropy": 0.20365197121687217, "entropy_control/target": 0.20485520424555576, "entropy_control/target_ready": 1.0, "entropy_control/target_samples": 5.0, "entropy_control/zeta_used": 0.05, "entropy_control/zeta_next": 0.05, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.152, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 204.398001, "rollout_overhead_time": 0.0, "peak_allocated_gib": 66.8723316192627, "peak_reserved_gib": 74.375, "reward": null }, { "step": 39, "loss": 0.07694607228040695, "grad_norm": 0.052001953125, "learning_rate": 2.4000000000000003e-06, "thinking/tokens_mean": 806.8958333333334, "thinking/budget_hit_fraction": 0.21875, "num_tokens": 1175392.0, "completions/mean_length": 3432.25, "completions/min_length": 2260.0, "completions/max_length": 4888.0, "completions/clipped_ratio": 0.125, "completions/mean_terminated_length": 3483.857177734375, "completions/min_terminated_length": 2260.0, "completions/max_terminated_length": 4888.0, "tools/call_frequency": 2.0, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__native/mean": 0.5647073984146118, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__native/std": 0.3610021471977234, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__efficiency/mean": 0.4166666865348816, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__efficiency/std": 0.267261266708374, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__reliability/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__reliability/std": 0.9258201122283936, "reward": 0.6855406761169434, "reward_std": 0.5539965033531189, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 0.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.17503059096634388, "cispo_clip_ratio": 0.0, "step_time": 1117.2130076190078, "entropy_control/entropy": 0.2098017228977747, "entropy_control/target": 0.20485520424555576, "entropy_control/target_ready": 1.0, "entropy_control/target_samples": 5.0, "entropy_control/zeta_used": 0.05, "entropy_control/zeta_next": 0.025, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.156, "adapter_sync_time": 0.0, "generation_time": 977.977568, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1003.401329, "forward_backward_time": 113.705469, "rollout_overhead_time": 25.423761, "peak_allocated_gib": 66.8723316192627, "peak_reserved_gib": 74.375 }, { "step": 40, "loss": 0.07094622403383255, "grad_norm": 0.05224609375, "learning_rate": 2.2e-06, "entropy": 0.1750783771276474, "cispo_clip_ratio": 0.003981826244853437, "step_time": 113.81666151001264, "entropy_control/entropy": 0.20988456722929072, "entropy_control/target": 0.20485520424555576, "entropy_control/target_ready": 1.0, "entropy_control/target_samples": 5.0, "entropy_control/zeta_used": 0.025, "entropy_control/zeta_next": 0.0125, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.16, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 113.710954, "rollout_overhead_time": 0.0, "peak_allocated_gib": 66.8723316192627, "peak_reserved_gib": 74.375, "reward": null }, { "step": 41, "loss": 0.0182107612490654, "grad_norm": 0.06494140625, "learning_rate": 2.0000000000000003e-06, "thinking/tokens_mean": 651.6333333333333, "thinking/budget_hit_fraction": 0.1875, "num_tokens": 1237863.0, "completions/mean_length": 4331.875, "completions/min_length": 2074.0, "completions/max_length": 6552.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 4331.875, "completions/min_terminated_length": 2074.0, "completions/max_terminated_length": 6552.0, "tools/call_frequency": 3.25, "tools/failure_frequency": 0.03846153989434242, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__native/mean": 0.7120382785797119, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__native/std": 0.11710403859615326, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__efficiency/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__efficiency/std": 0.19245010614395142, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/mean": 0.6516755819320679, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/std": 0.05423208698630333, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/mean": 0.4583333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/std": 0.2500000298023224, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/std": 0.0, "reward": 0.9058153033256531, "reward_std": 0.08276018500328064, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.20266657881438732, "cispo_clip_ratio": 0.0, "step_time": 1863.9089769140119, "entropy_control/entropy": 0.20727977515112125, "entropy_control/target": 0.20485520424555576, "entropy_control/target_ready": 1.0, "entropy_control/target_samples": 5.0, "entropy_control/zeta_used": 0.0125, "entropy_control/zeta_next": 0.00625, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.164, "adapter_sync_time": 0.0, "generation_time": 1584.546704, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1632.753558, "forward_backward_time": 231.041445, "rollout_overhead_time": 48.206854, "peak_allocated_gib": 66.8723316192627, "peak_reserved_gib": 74.375 }, { "step": 42, "loss": 0.016232745721936226, "grad_norm": 0.062255859375, "learning_rate": 1.8000000000000001e-06, "entropy": 0.20269928686320782, "cispo_clip_ratio": 0.0014816199545748532, "step_time": 231.138219371991, "entropy_control/entropy": 0.20735147211712707, "entropy_control/target": 0.20485520424555576, "entropy_control/target_ready": 1.0, "entropy_control/target_samples": 5.0, "entropy_control/zeta_used": 0.00625, "entropy_control/zeta_next": 0.003125, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.168, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 231.032407, "rollout_overhead_time": 0.0, "peak_allocated_gib": 66.8723316192627, "peak_reserved_gib": 74.375, "reward": null }, { "step": 43, "loss": 0.05386475473642349, "grad_norm": 0.0703125, "learning_rate": 1.6000000000000001e-06, "thinking/tokens_mean": 579.7458333333333, "thinking/budget_hit_fraction": 0.10416666666666667, "num_tokens": 1294305.0, "completions/mean_length": 3751.25, "completions/min_length": 820.0, "completions/max_length": 5090.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 3751.25, "completions/min_terminated_length": 820.0, "completions/max_terminated_length": 5090.0, "tools/call_frequency": 3.0, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__native/mean": 0.7556194067001343, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__native/std": 0.1185450553894043, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__efficiency/mean": 0.4583333730697632, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__efficiency/std": 0.20971761643886566, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/mean": 0.6714017987251282, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/std": 0.07128286361694336, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/mean": 0.5416666865348816, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/std": 0.15957120060920715, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/std": 0.0, "reward": 0.9385105967521667, "reward_std": 0.09417950361967087, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.19579153321683407, "cispo_clip_ratio": 0.0, "step_time": 1633.5453268820056, "entropy_control/entropy": 0.20656649488264806, "entropy_control/target": 0.20485520424555576, "entropy_control/target_ready": 1.0, "entropy_control/target_samples": 5.0, "entropy_control/zeta_used": 0.003125, "entropy_control/zeta_next": 0.0015625, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.172, "adapter_sync_time": 0.0, "generation_time": 1415.327464, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1452.835355, "forward_backward_time": 180.602873, "rollout_overhead_time": 37.50789, "peak_allocated_gib": 66.8723316192627, "peak_reserved_gib": 74.375 }, { "step": 44, "loss": 0.05319508910179138, "grad_norm": 0.07080078125, "learning_rate": 1.4000000000000001e-06, "entropy": 0.195509048178792, "cispo_clip_ratio": 0.0021582006593234837, "step_time": 180.71860163799283, "entropy_control/entropy": 0.20638041637880689, "entropy_control/target": 0.20485520424555576, "entropy_control/target_ready": 1.0, "entropy_control/target_samples": 5.0, "entropy_control/zeta_used": 0.0015625, "entropy_control/zeta_next": 0.00078125, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.176, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 180.61375, "rollout_overhead_time": 0.0, "peak_allocated_gib": 66.8723316192627, "peak_reserved_gib": 74.375, "reward": null }, { "step": 45, "loss": -0.008169680833816528, "grad_norm": 0.06494140625, "learning_rate": 1.2000000000000002e-06, "thinking/tokens_mean": 559.6845238095237, "thinking/budget_hit_fraction": 0.08928571428571429, "num_tokens": 1358171.0, "completions/mean_length": 4313.25, "completions/min_length": 1751.0, "completions/max_length": 7750.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 4313.25, "completions/min_terminated_length": 1751.0, "completions/max_terminated_length": 7750.0, "tools/call_frequency": 3.625, "tools/failure_frequency": 0.03448275849223137, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/mean": 0.553654670715332, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/std": 0.20938174426555634, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/mean": 0.3333333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/std": 0.2357022762298584, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/std": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/mean": 0.6731905341148376, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/std": 0.07375707477331161, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/mean": 0.4166666865348816, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/std": 0.28867512941360474, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/std": 0.0, "reward": 0.7821725606918335, "reward_std": 0.30025607347488403, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.18156417924910784, "cispo_clip_ratio": 0.0, "step_time": 2046.1135910660087, "entropy_control/entropy": 0.18522452547652773, "entropy_control/target": 0.20485520424555576, "entropy_control/target_ready": 1.0, "entropy_control/target_samples": 5.0, "entropy_control/zeta_used": 0.00078125, "entropy_control/zeta_next": 0.0015625, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.18, "adapter_sync_time": 0.0, "generation_time": 1787.038514, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1832.34899, "forward_backward_time": 213.655498, "rollout_overhead_time": 45.310476, "peak_allocated_gib": 66.8723316192627, "peak_reserved_gib": 74.375 }, { "step": 46, "loss": -0.0083540640771389, "grad_norm": 0.08154296875, "learning_rate": 1.0000000000000002e-06, "entropy": 0.18177321273833513, "cispo_clip_ratio": 0.003489038615953177, "step_time": 213.78865860099904, "entropy_control/entropy": 0.18546149322949249, "entropy_control/target": 0.20485520424555576, "entropy_control/target_ready": 1.0, "entropy_control/target_samples": 5.0, "entropy_control/zeta_used": 0.0015625, "entropy_control/zeta_next": 0.003125, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.184, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 213.666836, "rollout_overhead_time": 0.0, "peak_allocated_gib": 66.8723316192627, "peak_reserved_gib": 74.375, "reward": null }, { "step": 47, "loss": 0.011561974883079529, "grad_norm": 0.07421875, "learning_rate": 8.000000000000001e-07, "thinking/tokens_mean": 588.3035714285714, "thinking/budget_hit_fraction": 0.16071428571428573, "num_tokens": 1417345.0, "completions/mean_length": 4110.25, "completions/min_length": 1555.0, "completions/max_length": 7501.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 4110.25, "completions/min_terminated_length": 1555.0, "completions/max_terminated_length": 7501.0, "tools/call_frequency": 3.25, "tools/failure_frequency": 0.07692307978868484, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/mean": 0.6431146860122681, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/std": 0.0671430230140686, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/mean": 0.625, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/std": 0.0833333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/mean": 0.6694862246513367, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/std": 0.32990628480911255, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/mean": 0.375, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/std": 0.28463754057884216, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/std": 0.0, "reward": 0.8813004493713379, "reward_std": 0.22074399888515472, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.18419698625802994, "cispo_clip_ratio": 0.0, "step_time": 1836.2074433269954, "entropy_control/entropy": 0.1833051982737068, "entropy_control/target": 0.20485520424555576, "entropy_control/target_ready": 1.0, "entropy_control/target_samples": 5.0, "entropy_control/zeta_used": 0.003125, "entropy_control/zeta_next": 0.00625, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.188, "adapter_sync_time": 0.0, "generation_time": 1544.144557, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1593.399081, "forward_backward_time": 242.694666, "rollout_overhead_time": 49.254524, "peak_allocated_gib": 66.8723316192627, "peak_reserved_gib": 74.375 }, { "step": 48, "loss": 0.012772273272275925, "grad_norm": 0.0673828125, "learning_rate": 6.000000000000001e-07, "entropy": 0.18411483988165855, "cispo_clip_ratio": 0.0024980568268802017, "step_time": 242.80719270800182, "entropy_control/entropy": 0.18331591951209228, "entropy_control/target": 0.20485520424555576, "entropy_control/target_ready": 1.0, "entropy_control/target_samples": 5.0, "entropy_control/zeta_used": 0.00625, "entropy_control/zeta_next": 0.0125, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.192, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 242.690787, "rollout_overhead_time": 0.0, "peak_allocated_gib": 66.8723316192627, "peak_reserved_gib": 74.375, "reward": null }, { "step": 49, "loss": 0.009742127731442451, "grad_norm": 0.0732421875, "learning_rate": 4.0000000000000003e-07, "thinking/tokens_mean": 691.875, "thinking/budget_hit_fraction": 0.2, "num_tokens": 1474326.0, "completions/mean_length": 3831.625, "completions/min_length": 1485.0, "completions/max_length": 6148.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 3831.625, "completions/min_terminated_length": 1485.0, "completions/max_terminated_length": 6148.0, "tools/call_frequency": 3.0, "tools/failure_frequency": 0.1666666716337204, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/mean": 0.6475368738174438, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/std": 0.07742142677307129, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/mean": 0.625, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/std": 0.0833333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/mean": 0.6971055269241333, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/std": 0.353938490152359, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/mean": 0.5416666865348816, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/std": 0.0833333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/std": 0.0, "reward": 0.9014878869056702, "reward_std": 0.23851774632930756, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.2077923733741045, "cispo_clip_ratio": 0.0, "step_time": 1400.860066737987, "entropy_control/entropy": 0.21604512623340566, "entropy_control/target": 0.20485520424555576, "entropy_control/target_ready": 1.0, "entropy_control/target_samples": 5.0, "entropy_control/zeta_used": 0.0125, "entropy_control/zeta_next": 0.00625, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.196, "adapter_sync_time": 0.0, "generation_time": 1225.281319, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1256.01869, "forward_backward_time": 144.732648, "rollout_overhead_time": 30.737371, "peak_allocated_gib": 66.8723316192627, "peak_reserved_gib": 74.375 }, { "step": 50, "loss": 0.007500660140067339, "grad_norm": 0.078125, "learning_rate": 2.0000000000000002e-07, "entropy": 0.20776817575097084, "cispo_clip_ratio": 0.002450441155815497, "step_time": 144.83408938099456, "entropy_control/entropy": 0.21604681538305054, "entropy_control/target": 0.20485520424555576, "entropy_control/target_ready": 1.0, "entropy_control/target_samples": 5.0, "entropy_control/zeta_used": 0.00625, "entropy_control/zeta_next": 0.003125, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.2, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 144.727252, "rollout_overhead_time": 0.0, "peak_allocated_gib": 66.8723316192627, "peak_reserved_gib": 74.375, "reward": null } ], "holdout": [ { "seed": 900000, "sample": 0, "native_reward": 0.8404385754973732, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 3, "model_tokens": 1804, "reward_channels": { "native": 0.8404385754973732, "efficiency": 0.5, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "two_pool_cycle" }, { "seed": 900000, "sample": 1, "native_reward": 0.6089882858614486, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 3002, "reward_channels": { "native": 0.6089882858614486, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "two_pool_cycle" }, { "seed": 900000, "sample": 2, "native_reward": 0.6983308425395819, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 2102, "reward_channels": { "native": 0.6983308425395819, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "two_pool_cycle" }, { "seed": 900000, "sample": 3, "native_reward": 0.6089882858614486, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 2387, "reward_channels": { "native": 0.6089882858614486, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "two_pool_cycle" }, { "seed": 900001, "sample": 0, "native_reward": 0.6093825392865522, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 2008, "reward_channels": { "native": 0.6093825392865522, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "two_pool_cycle" }, { "seed": 900001, "sample": 1, "native_reward": 0.6093825392865522, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 2709, "reward_channels": { "native": 0.6093825392865522, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "two_pool_cycle" }, { "seed": 900001, "sample": 2, "native_reward": 0.8435035328689895, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 3, "model_tokens": 3491, "reward_channels": { "native": 0.8435035328689895, "efficiency": 0.5, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "two_pool_cycle" }, { "seed": 900001, "sample": 3, "native_reward": 0.6999976570410972, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 3, "model_tokens": 5609, "reward_channels": { "native": 0.6999976570410972, "efficiency": 0.5, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "two_pool_cycle" }, { "seed": 900000, "sample": 0, "native_reward": 0.6106601460847819, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 2139, "reward_channels": { "native": 0.6106601460847819, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "triangular_cycle" }, { "seed": 900000, "sample": 1, "native_reward": 0.7049643139195981, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 4642, "reward_channels": { "native": 0.7049643139195981, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "triangular_cycle" }, { "seed": 900000, "sample": 2, "native_reward": 0.6106601460847819, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 3, "model_tokens": 4728, "reward_channels": { "native": 0.6106601460847819, "efficiency": 0.5, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "triangular_cycle" }, { "seed": 900000, "sample": 3, "native_reward": 0.6106601460847819, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 3870, "reward_channels": { "native": 0.6106601460847819, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "triangular_cycle" }, { "seed": 900001, "sample": 0, "native_reward": 0.0, "promotion_success": false, "done": false, "errors": 0, "tool_steps": 0, "model_tokens": 0, "reward_channels": { "native": 0.0, "efficiency": 0.0, "reliability": -1.0 }, "verification_level": null, "failed_gates": null, "family": "triangular_cycle" }, { "seed": 900001, "sample": 1, "native_reward": 0.25, "promotion_success": false, "done": false, "errors": 0, "tool_steps": 1, "model_tokens": 3362, "reward_channels": { "native": 0.25, "efficiency": 0.0, "reliability": -1.0 }, "verification_level": null, "failed_gates": [ "finalized" ], "family": "triangular_cycle" }, { "seed": 900001, "sample": 2, "native_reward": 0.6108851248150757, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 3622, "reward_channels": { "native": 0.6108851248150757, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "triangular_cycle" }, { "seed": 900001, "sample": 3, "native_reward": 0.0, "promotion_success": false, "done": false, "errors": 0, "tool_steps": 0, "model_tokens": 0, "reward_channels": { "native": 0.0, "efficiency": 0.0, "reliability": -1.0 }, "verification_level": null, "failed_gates": null, "family": "triangular_cycle" }, { "seed": 900000, "sample": 0, "native_reward": 0.5875732130670761, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 3741, "reward_channels": { "native": 0.5875732130670761, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "competing_cycles" }, { "seed": 900000, "sample": 1, "native_reward": 0.6140552238672119, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 3861, "reward_channels": { "native": 0.6140552238672119, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "competing_cycles" }, { "seed": 900000, "sample": 2, "native_reward": 0.8720485688510347, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 4494, "reward_channels": { "native": 0.8720485688510347, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "competing_cycles" }, { "seed": 900000, "sample": 3, "native_reward": 0.8720485688510347, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 5776, "reward_channels": { "native": 0.8720485688510347, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "competing_cycles" }, { "seed": 900001, "sample": 0, "native_reward": 0.6145108982971254, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 4590, "reward_channels": { "native": 0.6145108982971254, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "competing_cycles" }, { "seed": 900001, "sample": 1, "native_reward": 0.0, "promotion_success": false, "done": false, "errors": 0, "tool_steps": 0, "model_tokens": 0, "reward_channels": { "native": 0.0, "efficiency": 0.0, "reliability": -1.0 }, "verification_level": null, "failed_gates": null, "family": "competing_cycles" }, { "seed": 900001, "sample": 2, "native_reward": 0.6145108982971254, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 2279, "reward_channels": { "native": 0.6145108982971254, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "competing_cycles" }, { "seed": 900001, "sample": 3, "native_reward": 0.6145108982971254, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 3, "model_tokens": 5650, "reward_channels": { "native": 0.6145108982971254, "efficiency": 0.5, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "competing_cycles" }, { "seed": 900000, "sample": 0, "native_reward": 0.609376392538787, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 4780, "reward_channels": { "native": 0.609376392538787, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "four_hop_cycle" }, { "seed": 900000, "sample": 1, "native_reward": 0.8397209478162578, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 5, "model_tokens": 6375, "reward_channels": { "native": 0.8397209478162578, "efficiency": 0.16666666666666663, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "four_hop_cycle" }, { "seed": 900000, "sample": 2, "native_reward": 0.609376392538787, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 1846, "reward_channels": { "native": 0.609376392538787, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "four_hop_cycle" }, { "seed": 900000, "sample": 3, "native_reward": 0.7503548418171574, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 5100, "reward_channels": { "native": 0.7503548418171574, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "four_hop_cycle" }, { "seed": 900001, "sample": 0, "native_reward": 0.8461963597959128, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 6286, "reward_channels": { "native": 0.8461963597959128, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "four_hop_cycle" }, { "seed": 900001, "sample": 1, "native_reward": 0.8415088597959128, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 5, "model_tokens": 4017, "reward_channels": { "native": 0.8415088597959128, "efficiency": 0.16666666666666663, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "four_hop_cycle" }, { "seed": 900001, "sample": 2, "native_reward": 0.7352912527935178, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 5, "model_tokens": 3469, "reward_channels": { "native": 0.7352912527935178, "efficiency": 0.16666666666666663, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "four_hop_cycle" }, { "seed": 900001, "sample": 3, "native_reward": 0.6095701077161815, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 3076, "reward_channels": { "native": 0.6095701077161815, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "four_hop_cycle" }, { "seed": 900000, "sample": 0, "native_reward": 0.962057681977345, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 5182, "reward_channels": { "native": 0.962057681977345, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "size_sensitive_cycle" }, { "seed": 900000, "sample": 1, "native_reward": 0.175, "promotion_success": false, "done": true, "errors": 0, "tool_steps": 4, "model_tokens": 8525, "reward_channels": { "native": 0.175, "efficiency": 0.33333333333333337, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [ "positive_net_profit", "stress_positive", "task_complete" ], "family": "size_sensitive_cycle" }, { "seed": 900000, "sample": 2, "native_reward": 0.8262556477476042, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 3254, "reward_channels": { "native": 0.8262556477476042, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "size_sensitive_cycle" }, { "seed": 900000, "sample": 3, "native_reward": 0.8262556477476042, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 3609, "reward_channels": { "native": 0.8262556477476042, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "size_sensitive_cycle" }, { "seed": 900001, "sample": 0, "native_reward": 0.0, "promotion_success": false, "done": false, "errors": 0, "tool_steps": 0, "model_tokens": 0, "reward_channels": { "native": 0.0, "efficiency": 0.0, "reliability": -1.0 }, "verification_level": null, "failed_gates": null, "family": "size_sensitive_cycle" }, { "seed": 900001, "sample": 1, "native_reward": 0.8245803237223959, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 2, "model_tokens": 3045, "reward_channels": { "native": 0.8245803237223959, "efficiency": 0.6666666666666667, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "size_sensitive_cycle" }, { "seed": 900001, "sample": 2, "native_reward": 0.8245803237223959, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 3, "model_tokens": 4197, "reward_channels": { "native": 0.8245803237223959, "efficiency": 0.5, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "size_sensitive_cycle" }, { "seed": 900001, "sample": 3, "native_reward": 0.9872132714354994, "promotion_success": true, "done": true, "errors": 0, "tool_steps": 3, "model_tokens": 3949, "reward_channels": { "native": 0.9872132714354994, "efficiency": 0.5, "reliability": 1.0 }, "verification_level": "foundry_revm_arithmetic_replay", "failed_gates": [], "family": "size_sensitive_cycle" } ] }, { "run_id": "1789396300", "mode": "thinking", "method": "cispo-repo-r-window5-refill", "steps": 41, "reward_points": 21, "train_mean": 0.7908252122856322, "train_last10": 0.8301730066537857, "holdout_native": null, "holdout_episodes": 0, "holdout_successes": 0, "mean_step_seconds": 912.0109977337314, "label": "CISPO + REPO-R w5 + refill", "state_at_cutoff": "partial", "step_hours": 10.386791918634163, "peak_allocated_gib": 70.26109647750854, "peak_reserved_gib": 77.83203125, "extra_refill_rounds": 1.0, "groups_refilled": 1.0, "holdout_component_means": null, "configuration": { "model": { "id": "Qwen/Qwen3-14B", "revision": "40c069824f4251a91eefaf281ebe4c544efd3e18", "loss": "cispo", "algorithm": "grpo", "no_think": false, "thinking_budget_tokens": 2048, "lora": true, "quantize": true, "lora_r": 32, "lora_alpha": 64, "use_dora": false }, "loss": "cispo", "normalization": "joint", "entropy_control": { "method": "repo_r", "epsilon_low": 0.2, "epsilon_high": 0.2, "zeta_initial": 0.001, "zeta_min": 0.0001, "zeta_max": 0.05, "zeta_nonnegative": true, "target_window_steps": 5, "target_entropy": null, "holdout_split": "benchmark", "holdout_seeds_per_family": 2, "num_iterations": 2 }, "dynamic_sampling": true, "sampling": { "group_size": 4, "temperature": 1.0, "max_completion_tokens": 22528, "top_entropy_quantile": 1.0, "temperature_decay_to": null }, "learning_rate": 1e-05, "turn_max_tokens": 3072 }, "source_sha256": { "metrics.jsonl": "5635a7af48bd6216438037cd6e2b997645be9eb7b68e4b67a8d3251fc5f8847b", "resolved_training_config.json": "84afef7ace2e4a86ce01314335f765b25f45ed5c38912f99f8e2930869783359" }, "training_steps": [ { "step": 1, "loss": 0.07572822272777557, "grad_norm": 0.078125, "learning_rate": 1e-05, "thinking/tokens_mean": 606.2361111111111, "thinking/budget_hit_fraction": 0.14583333333333334, "num_tokens": 51975.0, "completions/mean_length": 3604.375, "completions/min_length": 2890.0, "completions/max_length": 6014.0, "completions/clipped_ratio": 0.25, "completions/mean_terminated_length": 3782.166748046875, "completions/min_terminated_length": 2890.0, "completions/max_terminated_length": 6014.0, "tools/call_frequency": 2.25, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.7262155413627625, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 0.13481803238391876, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.4583333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.2500000298023224, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/mean": 0.45231834053993225, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/std": 0.526568591594696, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/mean": 0.2916666865348816, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/std": 0.34359216690063477, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/mean": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/std": 1.154700517654419, "reward": 0.7080169916152954, "reward_std": 0.5734051465988159, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 0.5, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.14539151638746262, "cispo_clip_ratio": 0.0, "step_time": 1239.501060564, "entropy_control/entropy": 0.20009788357278327, "entropy_control/target": 0.20009788357278327, "entropy_control/target_ready": 0.0, "entropy_control/target_samples": 1.0, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/update_skipped": 0.0, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.004, "adapter_sync_time": 0.0, "generation_time": 1048.524445, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1082.869805, "forward_backward_time": 156.543807, "rollout_overhead_time": 34.345361, "peak_allocated_gib": 46.211081981658936, "peak_reserved_gib": 51.38671875 }, { "step": 2, "loss": 0.07578127086162567, "grad_norm": 0.08349609375, "learning_rate": 9.800000000000001e-06, "entropy": 0.14508133009076118, "cispo_clip_ratio": 0.0023914053745102137, "step_time": 156.63328702199988, "entropy_control/entropy": 0.1997213626651036, "entropy_control/target": 0.19990962311894345, "entropy_control/target_ready": 0.0, "entropy_control/target_samples": 2.0, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/update_skipped": 0.0, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.008, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 156.532347, "rollout_overhead_time": 0.0, "peak_allocated_gib": 46.689598083496094, "peak_reserved_gib": 51.484375, "reward": null }, { "step": 3, "loss": 0.09946434199810028, "grad_norm": 0.0654296875, "learning_rate": 9.600000000000001e-06, "thinking/tokens_mean": 642.3095238095237, "thinking/budget_hit_fraction": 0.1726190476190476, "num_tokens": 103527.0, "completions/mean_length": 3732.0, "completions/min_length": 1736.0, "completions/max_length": 5934.0, "completions/clipped_ratio": 0.125, "completions/mean_terminated_length": 3826.4287109375, "completions/min_terminated_length": 1736.0, "completions/max_terminated_length": 5934.0, "tools/call_frequency": 2.625, "tools/failure_frequency": 0.095238097012043, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.5765475034713745, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 0.2458232194185257, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.4791666865348816, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.25877460837364197, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 0.75, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 0.7071067690849304, "reward": 0.750505805015564, "reward_std": 0.39041367173194885, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 0.5, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.18255372159183025, "cispo_clip_ratio": 0.0, "step_time": 1574.8926672090001, "entropy_control/entropy": 0.2241447887375959, "entropy_control/target": 0.20798801165849426, "entropy_control/target_ready": 0.0, "entropy_control/target_samples": 3.0, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/update_skipped": 0.0, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.012, "adapter_sync_time": 0.0, "generation_time": 1383.022687, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1416.991309, "forward_backward_time": 157.795754, "rollout_overhead_time": 33.968622, "peak_allocated_gib": 49.95496845245361, "peak_reserved_gib": 54.9609375 }, { "step": 4, "loss": 0.10000739991664886, "grad_norm": 0.06689453125, "learning_rate": 9.4e-06, "entropy": 0.18299074657261372, "cispo_clip_ratio": 0.0030211390694603324, "step_time": 157.89051116999917, "entropy_control/entropy": 0.22478497224881575, "entropy_control/target": 0.21218725180607464, "entropy_control/target_ready": 0.0, "entropy_control/target_samples": 4.0, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.0, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/update_skipped": 0.0, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.016, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 157.792978, "rollout_overhead_time": 0.0, "peak_allocated_gib": 49.95496845245361, "peak_reserved_gib": 54.9609375, "reward": null }, { "step": 5, "loss": 0.013005992397665977, "grad_norm": 0.06396484375, "learning_rate": 9.200000000000002e-06, "thinking/tokens_mean": 628.0238095238095, "thinking/budget_hit_fraction": 0.19047619047619047, "num_tokens": 153323.0, "completions/mean_length": 3519.5, "completions/min_length": 2199.0, "completions/max_length": 5461.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 3519.5, "completions/min_terminated_length": 2199.0, "completions/max_terminated_length": 5461.0, "tools/call_frequency": 2.375, "tools/failure_frequency": 0.05263157933950424, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.5747185349464417, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 0.2655002176761627, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.26726123690605164, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 0.75, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 0.7071067690849304, "reward": 0.7497185468673706, "reward_std": 0.4022914469242096, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 0.5, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.2027073223143816, "cispo_clip_ratio": 0.0, "step_time": 1087.9027467079995, "entropy_control/entropy": 0.2071854305056443, "entropy_control/target": 0.21118688754598858, "entropy_control/target_ready": 1.0, "entropy_control/target_samples": 5.0, "entropy_control/zeta_used": 0.0, "entropy_control/zeta_next": 0.001, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/update_skipped": 0.0, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.02, "adapter_sync_time": 0.0, "generation_time": 880.958177, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 916.705705, "forward_backward_time": 171.102261, "rollout_overhead_time": 35.747528, "peak_allocated_gib": 49.95496845245361, "peak_reserved_gib": 55.15625 }, { "step": 6, "loss": 0.012761129066348076, "grad_norm": 0.06396484375, "learning_rate": 9e-06, "entropy": 0.20208648033440113, "cispo_clip_ratio": 0.003421459812670946, "step_time": 171.2098793110008, "entropy_control/entropy": 0.20650391197626736, "entropy_control/target": 0.21118688754598858, "entropy_control/target_ready": 1.0, "entropy_control/target_samples": 5.0, "entropy_control/zeta_used": 0.001, "entropy_control/zeta_next": 0.002, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/update_skipped": 0.0, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.024, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 171.105973, "rollout_overhead_time": 0.0, "peak_allocated_gib": 49.95496845245361, "peak_reserved_gib": 55.15625, "reward": null }, { "step": 7, "loss": -0.01953023113310337, "grad_norm": 0.06103515625, "learning_rate": 8.8e-06, "thinking/tokens_mean": 816.4416666666667, "thinking/budget_hit_fraction": 0.225, "num_tokens": 216848.0, "completions/mean_length": 4473.125, "completions/min_length": 2059.0, "completions/max_length": 7562.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 4473.125, "completions/min_terminated_length": 2059.0, "completions/max_terminated_length": 7562.0, "tools/call_frequency": 3.125, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/mean": 0.5696794986724854, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/std": 0.2904992997646332, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/mean": 0.4583333730697632, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/std": 0.15957120060920715, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/mean": 0.8388867378234863, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/std": 0.090998575091362, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/std": 0.13608276844024658, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/std": 0.0, "reward": 0.928241491317749, "reward_std": 0.24654604494571686, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.18779405020177364, "cispo_clip_ratio": 0.0, "step_time": 1743.7932904109994, "entropy_control/entropy": 0.19367323781161366, "entropy_control/target": 0.21118688754598858, "entropy_control/target_ready": 1.0, "entropy_control/target_samples": 5.0, "entropy_control/zeta_used": 0.002, "entropy_control/zeta_next": 0.004, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/update_skipped": 0.0, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.028, "adapter_sync_time": 0.0, "generation_time": 1480.588623, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1525.898319, "forward_backward_time": 217.784316, "rollout_overhead_time": 45.309696, "peak_allocated_gib": 57.167871952056885, "peak_reserved_gib": 63.22265625 }, { "step": 8, "loss": -0.01878420077264309, "grad_norm": 0.06494140625, "learning_rate": 8.6e-06, "entropy": 0.18777861446142197, "cispo_clip_ratio": 0.002460553514538333, "step_time": 217.9013085759998, "entropy_control/entropy": 0.19359376347444454, "entropy_control/target": 0.21118688754598858, "entropy_control/target_ready": 1.0, "entropy_control/target_samples": 5.0, "entropy_control/zeta_used": 0.004, "entropy_control/zeta_next": 0.008, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/update_skipped": 0.0, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.032, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 217.797755, "rollout_overhead_time": 0.0, "peak_allocated_gib": 57.167871952056885, "peak_reserved_gib": 63.22265625, "reward": null }, { "step": 9, "loss": 0.11859361827373505, "grad_norm": 0.042236328125, "learning_rate": 8.400000000000001e-06, "thinking/tokens_mean": 780.75, "thinking/budget_hit_fraction": 0.25, "num_tokens": 274906.0, "completions/mean_length": 3973.25, "completions/min_length": 2801.0, "completions/max_length": 6710.0, "completions/clipped_ratio": 0.25, "completions/mean_terminated_length": 4274.0, "completions/min_terminated_length": 2801.0, "completions/max_terminated_length": 6710.0, "tools/call_frequency": 1.875, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.4571598172187805, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 0.304773211479187, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.3333333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/mean": 0.48381099104881287, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/std": 0.3254546523094177, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/mean": 0.375, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/std": 0.34359216690063477, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/std": 1.0, "reward": 0.5923604369163513, "reward_std": 0.4896566569805145, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 0.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.17063424549996853, "cispo_clip_ratio": 0.0, "step_time": 1398.187785918998, "entropy_control/entropy": 0.230657749842964, "entropy_control/target": 0.21118688754598858, "entropy_control/target_ready": 1.0, "entropy_control/target_samples": 5.0, "entropy_control/zeta_used": 0.008, "entropy_control/zeta_next": 0.004, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/update_skipped": 0.0, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.036, "adapter_sync_time": 0.0, "generation_time": 1100.928635, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1151.379982, "forward_backward_time": 246.699895, "rollout_overhead_time": 50.451347, "peak_allocated_gib": 61.50002145767212, "peak_reserved_gib": 68.22265625 }, { "step": 10, "loss": 0.11739887297153473, "grad_norm": 0.042724609375, "learning_rate": 8.2e-06, "entropy": 0.17027848213911057, "cispo_clip_ratio": 0.0032983401615638286, "step_time": 246.8016796869979, "entropy_control/entropy": 0.23019734439087136, "entropy_control/target": 0.21118688754598858, "entropy_control/target_ready": 1.0, "entropy_control/target_samples": 5.0, "entropy_control/zeta_used": 0.004, "entropy_control/zeta_next": 0.002, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/update_skipped": 0.0, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.04, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 246.70414, "rollout_overhead_time": 0.0, "peak_allocated_gib": 61.50002145767212, "peak_reserved_gib": 68.22265625, "reward": null }, { "step": 11, "loss": 0.06222832575440407, "grad_norm": 0.06591796875, "learning_rate": 8.000000000000001e-06, "thinking/tokens_mean": 617.4642857142857, "thinking/budget_hit_fraction": 0.14285714285714285, "num_tokens": 334616.0, "completions/mean_length": 4154.75, "completions/min_length": 1584.0, "completions/max_length": 7545.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 4154.75, "completions/min_terminated_length": 1584.0, "completions/max_terminated_length": 7545.0, "tools/call_frequency": 3.375, "tools/failure_frequency": 0.07407407462596893, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__native/mean": 0.7071732878684998, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__native/std": 0.1108814924955368, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__efficiency/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__efficiency/std": 0.235702246427536, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/mean": 0.7259995937347412, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/std": 0.13512180745601654, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/mean": 0.4583333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/std": 0.2500000298023224, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/std": 0.0, "reward": 0.940544843673706, "reward_std": 0.10492833703756332, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.19184984639286995, "cispo_clip_ratio": 0.0, "step_time": 1629.4883709589976, "entropy_control/entropy": 0.19991044346835185, "entropy_control/target": 0.21118688754598858, "entropy_control/target_ready": 1.0, "entropy_control/target_samples": 5.0, "entropy_control/zeta_used": 0.002, "entropy_control/zeta_next": 0.004, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/update_skipped": 0.0, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.044, "adapter_sync_time": 0.0, "generation_time": 1385.120445, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1427.554151, "forward_backward_time": 201.838395, "rollout_overhead_time": 42.433705, "peak_allocated_gib": 61.50002145767212, "peak_reserved_gib": 68.22265625 }, { "step": 12, "loss": 0.06212514638900757, "grad_norm": 0.0625, "learning_rate": 7.800000000000002e-06, "entropy": 0.1914893463253975, "cispo_clip_ratio": 0.0025537287001498044, "step_time": 201.96123017800164, "entropy_control/entropy": 0.19939919574821227, "entropy_control/target": 0.21118688754598858, "entropy_control/target_ready": 1.0, "entropy_control/target_samples": 5.0, "entropy_control/zeta_used": 0.004, "entropy_control/zeta_next": 0.008, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/update_skipped": 0.0, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.048, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 201.852485, "rollout_overhead_time": 0.0, "peak_allocated_gib": 61.50002145767212, "peak_reserved_gib": 68.22265625, "reward": null }, { "step": 13, "loss": -0.0030301217921078205, "grad_norm": 0.062255859375, "learning_rate": 7.600000000000001e-06, "thinking/tokens_mean": 684.6785714285714, "thinking/budget_hit_fraction": 0.125, "num_tokens": 397446.0, "completions/mean_length": 4568.75, "completions/min_length": 2830.0, "completions/max_length": 7770.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 4568.75, "completions/min_terminated_length": 2830.0, "completions/max_terminated_length": 7770.0, "tools/call_frequency": 3.25, "tools/failure_frequency": 0.07692307978868484, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.668724536895752, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 0.11833984404802322, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.5833333730697632, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.1666666716337204, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/mean": 0.6792972087860107, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/std": 0.12803438305854797, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/mean": 0.4166666865348816, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/std": 0.1666666716337204, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/std": 0.0, "reward": 0.8990108966827393, "reward_std": 0.10922113060951233, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.19204666838049889, "cispo_clip_ratio": 0.0, "step_time": 1707.6739688449998, "entropy_control/entropy": 0.1843919927225112, "entropy_control/target": 0.21118688754598858, "entropy_control/target_ready": 1.0, "entropy_control/target_samples": 5.0, "entropy_control/zeta_used": 0.008, "entropy_control/zeta_next": 0.016, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/update_skipped": 0.0, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.052, "adapter_sync_time": 0.0, "generation_time": 1435.974728, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1482.796125, "forward_backward_time": 224.78145, "rollout_overhead_time": 46.821397, "peak_allocated_gib": 61.50002145767212, "peak_reserved_gib": 68.22265625 }, { "step": 14, "loss": -0.0010479791089892387, "grad_norm": 0.06005859375, "learning_rate": 7.4e-06, "entropy": 0.1921336855739355, "cispo_clip_ratio": 0.0009056136186700314, "step_time": 224.84784870800286, "entropy_control/entropy": 0.18457037096193685, "entropy_control/target": 0.21118688754598858, "entropy_control/target_ready": 1.0, "entropy_control/target_samples": 5.0, "entropy_control/zeta_used": 0.016, "entropy_control/zeta_next": 0.032, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/update_skipped": 0.0, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.056, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 224.742291, "rollout_overhead_time": 0.0, "peak_allocated_gib": 61.50002145767212, "peak_reserved_gib": 68.22265625, "reward": null }, { "step": 15, "loss": 0.03000808134675026, "grad_norm": 0.053955078125, "learning_rate": 7.2000000000000005e-06, "thinking/tokens_mean": 814.0982142857142, "thinking/budget_hit_fraction": 0.1488095238095238, "num_tokens": 464841.0, "completions/mean_length": 4940.875, "completions/min_length": 3071.0, "completions/max_length": 8198.0, "completions/clipped_ratio": 0.125, "completions/mean_terminated_length": 5208.0, "completions/min_terminated_length": 3398.0, "completions/max_terminated_length": 8198.0, "tools/call_frequency": 3.25, "tools/failure_frequency": 0.07692307978868484, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/mean": 0.40374743938446045, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/std": 0.2645609676837921, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/mean": 0.3333333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/std": 0.13608276844024658, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/mean": 0.6173950433731079, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/std": 0.41159671545028687, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/mean": 0.4166666865348816, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/std": 0.2886751592159271, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/std": 1.0, "reward": 0.6793212890625, "reward_std": 0.4502282738685608, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 0.5, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.18340378440916538, "cispo_clip_ratio": 0.0, "step_time": 1931.8897208800026, "entropy_control/entropy": 0.20570165910858026, "entropy_control/target": 0.21118688754598858, "entropy_control/target_ready": 1.0, "entropy_control/target_samples": 5.0, "entropy_control/zeta_used": 0.032, "entropy_control/zeta_next": 0.05, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/update_skipped": 0.0, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.06, "adapter_sync_time": 0.0, "generation_time": 1631.464454, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1682.396464, "forward_backward_time": 249.397272, "rollout_overhead_time": 50.93201, "peak_allocated_gib": 61.644176959991455, "peak_reserved_gib": 68.359375 }, { "step": 16, "loss": 0.03547542169690132, "grad_norm": 0.05419921875, "learning_rate": 7e-06, "entropy": 0.1832587793469429, "cispo_clip_ratio": 0.0033472176000941545, "step_time": 249.5143173660017, "entropy_control/entropy": 0.20542467953660967, "entropy_control/target": 0.21118688754598858, "entropy_control/target_ready": 1.0, "entropy_control/target_samples": 5.0, "entropy_control/zeta_used": 0.05, "entropy_control/zeta_next": 0.05, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/update_skipped": 0.0, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.064, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 249.418405, "rollout_overhead_time": 0.0, "peak_allocated_gib": 61.644176959991455, "peak_reserved_gib": 68.359375, "reward": null }, { "step": 17, "loss": 0.0017249472439289093, "grad_norm": 0.053955078125, "learning_rate": 6.800000000000001e-06, "thinking/tokens_mean": 714.3630952380953, "thinking/budget_hit_fraction": 0.14285714285714285, "num_tokens": 535380.0, "completions/mean_length": 5126.375, "completions/min_length": 1699.0, "completions/max_length": 9937.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 5126.375, "completions/min_terminated_length": 1699.0, "completions/max_terminated_length": 9937.0, "tools/call_frequency": 3.75, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/mean": 0.5947250723838806, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/std": 0.25756439566612244, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/mean": 0.2916666865348816, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/std": 0.25, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/std": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/mean": 0.556321382522583, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/std": 0.21588364243507385, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/mean": 0.4583333730697632, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/std": 0.3154948949813843, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/std": 1.0, "reward": 0.6942732334136963, "reward_std": 0.40823739767074585, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.18763641081750393, "cispo_clip_ratio": 0.0, "step_time": 2454.9009708850044, "entropy_control/entropy": 0.19494015099851772, "entropy_control/target": 0.21118688754598858, "entropy_control/target_ready": 1.0, "entropy_control/target_samples": 5.0, "entropy_control/zeta_used": 0.05, "entropy_control/zeta_next": 0.05, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/update_skipped": 0.0, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.068, "adapter_sync_time": 0.0, "generation_time": 2079.896167, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 2142.749997, "forward_backward_time": 312.045877, "rollout_overhead_time": 62.853829, "peak_allocated_gib": 70.26109647750854, "peak_reserved_gib": 77.83203125 }, { "step": 18, "loss": 0.001457948237657547, "grad_norm": 0.06005859375, "learning_rate": 6.600000000000001e-06, "entropy": 0.1876405104994774, "cispo_clip_ratio": 0.0045884831342846155, "step_time": 312.1395660230046, "entropy_control/entropy": 0.19493996098454794, "entropy_control/target": 0.21118688754598858, "entropy_control/target_ready": 1.0, "entropy_control/target_samples": 5.0, "entropy_control/zeta_used": 0.05, "entropy_control/zeta_next": 0.05, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/update_skipped": 0.0, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.072, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 312.038325, "rollout_overhead_time": 0.0, "peak_allocated_gib": 70.26109647750854, "peak_reserved_gib": 77.83203125, "reward": null }, { "step": 19, "loss": 0.05143576115369797, "grad_norm": 0.083984375, "learning_rate": 6.4000000000000006e-06, "thinking/tokens_mean": 711.5166666666667, "thinking/budget_hit_fraction": 0.16666666666666666, "num_tokens": 585154.0, "completions/mean_length": 3331.25, "completions/min_length": 1047.0, "completions/max_length": 5969.0, "completions/clipped_ratio": 0.125, "completions/mean_terminated_length": 3368.4287109375, "completions/min_terminated_length": 1047.0, "completions/max_terminated_length": 5969.0, "tools/call_frequency": 2.0, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.34859350323677063, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 0.30983424186706543, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.4583333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.3154948949813843, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/mean": 0.619419515132904, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/std": 0.41294634342193604, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/mean": 0.375, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/std": 0.28463754057884216, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/std": 1.0, "reward": 0.6048398613929749, "reward_std": 0.5408963561058044, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 0.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.16084277536720037, "cispo_clip_ratio": 0.0, "step_time": 1255.3704401659998, "entropy_control/entropy": 0.1911744343746652, "entropy_control/target": 0.21118688754598858, "entropy_control/target_ready": 1.0, "entropy_control/target_samples": 5.0, "entropy_control/zeta_used": 0.05, "entropy_control/zeta_next": 0.05, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/update_skipped": 0.0, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.076, "adapter_sync_time": 0.0, "generation_time": 1091.244395, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1120.723201, "forward_backward_time": 134.54555, "rollout_overhead_time": 29.478805, "peak_allocated_gib": 70.26109647750854, "peak_reserved_gib": 77.83203125 }, { "step": 20, "loss": 0.05150822177529335, "grad_norm": 0.0654296875, "learning_rate": 6.200000000000001e-06, "entropy": 0.1609591143205762, "cispo_clip_ratio": 0.003984676877735183, "step_time": 134.67310873000315, "entropy_control/entropy": 0.1913696210057772, "entropy_control/target": 0.21118688754598858, "entropy_control/target_ready": 1.0, "entropy_control/target_samples": 5.0, "entropy_control/zeta_used": 0.05, "entropy_control/zeta_next": 0.05, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/update_skipped": 0.0, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.08, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 134.565871, "rollout_overhead_time": 0.0, "peak_allocated_gib": 70.26109647750854, "peak_reserved_gib": 77.83203125, "reward": null }, { "step": 21, "loss": 0.07471884787082672, "grad_norm": 0.0654296875, "learning_rate": 6e-06, "thinking/tokens_mean": 714.1078571428571, "thinking/budget_hit_fraction": 0.10357142857142856, "num_tokens": 645727.0, "completions/mean_length": 3872.625, "completions/min_length": 1854.0, "completions/max_length": 7092.0, "completions/clipped_ratio": 0.125, "completions/mean_terminated_length": 3987.14306640625, "completions/min_terminated_length": 1854.0, "completions/max_terminated_length": 7092.0, "tools/call_frequency": 2.625, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/mean": 0.4695986807346344, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/std": 0.31352829933166504, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/mean": 0.375, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/std": 0.28463754057884216, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/std": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/mean": 0.7041833400726318, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/std": 0.1153934895992279, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/std": 0.13608276844024658, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/std": 0.0, "reward": 0.7587659955024719, "reward_std": 0.3966803252696991, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 0.5, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.19051527045667171, "cispo_clip_ratio": 0.0, "step_time": 1649.0433202029926, "entropy_control/entropy": 0.22611796571848017, "entropy_control/target": 0.21118688754598858, "entropy_control/target_ready": 1.0, "entropy_control/target_samples": 5.0, "entropy_control/zeta_used": 0.05, "entropy_control/zeta_next": 0.025, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/update_skipped": 0.0, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.084, "adapter_sync_time": 0.0, "generation_time": 1391.315053, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1435.932255, "forward_backward_time": 213.003064, "rollout_overhead_time": 44.617201, "peak_allocated_gib": 70.26109647750854, "peak_reserved_gib": 77.83203125 }, { "step": 22, "loss": 0.06720726937055588, "grad_norm": 0.0654296875, "learning_rate": 5.8e-06, "entropy": 0.19100352749228477, "cispo_clip_ratio": 0.00360257487045601, "step_time": 213.08353274199908, "entropy_control/entropy": 0.22664861035666894, "entropy_control/target": 0.21118688754598858, "entropy_control/target_ready": 1.0, "entropy_control/target_samples": 5.0, "entropy_control/zeta_used": 0.025, "entropy_control/zeta_next": 0.0125, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/update_skipped": 0.0, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.088, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 212.975308, "rollout_overhead_time": 0.0, "peak_allocated_gib": 70.26109647750854, "peak_reserved_gib": 77.83203125, "reward": null }, { "step": 23, "loss": -0.03625545650720596, "grad_norm": 0.0849609375, "learning_rate": 5.600000000000001e-06, "thinking/tokens_mean": 447.9523809523809, "thinking/budget_hit_fraction": 0.10714285714285714, "num_tokens": 695184.0, "completions/mean_length": 3296.625, "completions/min_length": 1980.0, "completions/max_length": 6147.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 3296.625, "completions/min_terminated_length": 1980.0, "completions/max_terminated_length": 6147.0, "tools/call_frequency": 3.125, "tools/failure_frequency": 0.07999999821186066, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.6673699617385864, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 0.11561504751443863, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.5416666865348816, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.25, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/mean": 0.6815767288208008, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/std": 0.2877177894115448, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/std": 0.3333333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/std": 1.0, "reward": 0.8505150079727173, "reward_std": 0.3402842581272125, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.1736058034002781, "cispo_clip_ratio": 0.0, "step_time": 1193.8290967089888, "entropy_control/entropy": 0.17548780286473117, "entropy_control/target": 0.21118688754598858, "entropy_control/target_ready": 1.0, "entropy_control/target_samples": 5.0, "entropy_control/zeta_used": 0.0125, "entropy_control/zeta_next": 0.025, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/update_skipped": 0.0, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.092, "adapter_sync_time": 0.0, "generation_time": 970.043981, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1009.256144, "forward_backward_time": 184.471185, "rollout_overhead_time": 39.212163, "peak_allocated_gib": 70.26109647750854, "peak_reserved_gib": 77.83203125 }, { "step": 24, "loss": -0.032149478793144226, "grad_norm": 0.08984375, "learning_rate": 5.400000000000001e-06, "entropy": 0.17358879279345274, "cispo_clip_ratio": 0.002499209367670119, "step_time": 184.58321064999473, "entropy_control/entropy": 0.17547268927231588, "entropy_control/target": 0.21118688754598858, "entropy_control/target_ready": 1.0, "entropy_control/target_samples": 5.0, "entropy_control/zeta_used": 0.025, "entropy_control/zeta_next": 0.05, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/update_skipped": 0.0, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.096, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 184.494663, "rollout_overhead_time": 0.0, "peak_allocated_gib": 70.26109647750854, "peak_reserved_gib": 77.83203125, "reward": null }, { "step": 25, "loss": 0.0003781411796808243, "grad_norm": 0.07373046875, "learning_rate": 5.2e-06, "thinking/tokens_mean": 570.8857142857142, "thinking/budget_hit_fraction": 0.125, "num_tokens": 751556.0, "completions/mean_length": 3956.5, "completions/min_length": 3004.0, "completions/max_length": 4727.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 3956.5, "completions/min_terminated_length": 3004.0, "completions/max_terminated_length": 4727.0, "tools/call_frequency": 3.0, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.7263808250427246, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 0.15270063281059265, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.5416666865348816, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.0833333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/mean": 0.6163347959518433, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/std": 0.2627241015434265, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/mean": 0.4583333730697632, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/std": 0.3154948949813843, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/std": 1.0, "reward": 0.8463578224182129, "reward_std": 0.3424694538116455, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.20230159349739552, "cispo_clip_ratio": 0.0, "step_time": 1197.0888584370077, "entropy_control/entropy": 0.20091232362194608, "entropy_control/target": 0.21118688754598858, "entropy_control/target_ready": 1.0, "entropy_control/target_samples": 5.0, "entropy_control/zeta_used": 0.05, "entropy_control/zeta_next": 0.05, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/update_skipped": 0.0, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.1, "adapter_sync_time": 0.0, "generation_time": 1000.48221, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1035.028993, "forward_backward_time": 161.951689, "rollout_overhead_time": 34.546783, "peak_allocated_gib": 70.26109647750854, "peak_reserved_gib": 77.83203125 }, { "step": 26, "loss": 0.00022291205823421478, "grad_norm": 0.07177734375, "learning_rate": 5e-06, "entropy": 0.20256466418504715, "cispo_clip_ratio": 0.0030252868600655347, "step_time": 162.0502540699963, "entropy_control/entropy": 0.20121698974361912, "entropy_control/target": 0.21118688754598858, "entropy_control/target_ready": 1.0, "entropy_control/target_samples": 5.0, "entropy_control/zeta_used": 0.05, "entropy_control/zeta_next": 0.05, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/update_skipped": 0.0, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.104, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 161.945519, "rollout_overhead_time": 0.0, "peak_allocated_gib": 70.26109647750854, "peak_reserved_gib": 77.83203125, "reward": null }, { "step": 27, "loss": 0.00848444178700447, "grad_norm": 0.0712890625, "learning_rate": 4.800000000000001e-06, "thinking/tokens_mean": 505.85714285714283, "thinking/budget_hit_fraction": 0.07142857142857142, "num_tokens": 807779.0, "completions/mean_length": 3725.375, "completions/min_length": 2768.0, "completions/max_length": 5253.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 3725.375, "completions/min_terminated_length": 2768.0, "completions/max_terminated_length": 5253.0, "tools/call_frequency": 3.375, "tools/failure_frequency": 0.18518517911434174, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/mean": 0.727188229560852, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__native/std": 0.13555511832237244, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__efficiency/std": 0.235702246427536, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__four_hop_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/mean": 0.8250280022621155, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/mean": 0.5833333730697632, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/std": 0.1666666716337204, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/std": 0.0, "reward": 1.0031914710998535, "reward_std": 0.09914910048246384, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.1896395180374384, "cispo_clip_ratio": 0.0, "step_time": 1235.7846389290025, "entropy_control/entropy": 0.18774464511480862, "entropy_control/target": 0.21118688754598858, "entropy_control/target_ready": 1.0, "entropy_control/target_samples": 5.0, "entropy_control/zeta_used": 0.05, "entropy_control/zeta_next": 0.05, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/update_skipped": 0.0, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.108, "adapter_sync_time": 0.0, "generation_time": 1006.205013, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1046.201583, "forward_backward_time": 189.476789, "rollout_overhead_time": 39.99657, "peak_allocated_gib": 70.26109647750854, "peak_reserved_gib": 77.83203125 }, { "step": 28, "loss": 0.008762340992689133, "grad_norm": 0.0693359375, "learning_rate": 4.600000000000001e-06, "entropy": 0.18963168933987617, "cispo_clip_ratio": 0.003457594779320061, "step_time": 189.59208594300435, "entropy_control/entropy": 0.18770631076593217, "entropy_control/target": 0.21118688754598858, "entropy_control/target_ready": 1.0, "entropy_control/target_samples": 5.0, "entropy_control/zeta_used": 0.05, "entropy_control/zeta_next": 0.05, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/update_skipped": 0.0, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.112, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 189.485055, "rollout_overhead_time": 0.0, "peak_allocated_gib": 70.26109647750854, "peak_reserved_gib": 77.83203125, "reward": null }, { "step": 29, "loss": 0.08357852697372437, "grad_norm": 0.050048828125, "learning_rate": 4.4e-06, "thinking/tokens_mean": 854.6934523809523, "thinking/budget_hit_fraction": 0.16964285714285712, "num_tokens": 872932.0, "completions/mean_length": 4861.625, "completions/min_length": 3071.0, "completions/max_length": 7547.0, "completions/clipped_ratio": 0.125, "completions/mean_terminated_length": 5117.4287109375, "completions/min_terminated_length": 3180.0, "completions/max_terminated_length": 7547.0, "tools/call_frequency": 2.875, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.7234823703765869, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 0.13265986740589142, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.4583333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.2500000298023224, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/mean": 0.5158176422119141, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/std": 0.3478271961212158, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/mean": 0.3333333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/std": 0.27216553688049316, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/std": 1.0, "reward": 0.789441704750061, "reward_std": 0.4089658856391907, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 0.5, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.18881307914853096, "cispo_clip_ratio": 0.0, "step_time": 1960.6909793230006, "entropy_control/entropy": 0.2095442864458711, "entropy_control/target": 0.21118688754598858, "entropy_control/target_ready": 1.0, "entropy_control/target_samples": 5.0, "entropy_control/zeta_used": 0.05, "entropy_control/zeta_next": 0.05, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/update_skipped": 0.0, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.116, "adapter_sync_time": 0.0, "generation_time": 1707.781295, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1751.396552, "forward_backward_time": 209.185927, "rollout_overhead_time": 43.615257, "peak_allocated_gib": 70.26109647750854, "peak_reserved_gib": 77.83203125 }, { "step": 30, "loss": 0.0832657739520073, "grad_norm": 0.052734375, "learning_rate": 4.2000000000000004e-06, "entropy": 0.18904010578989983, "cispo_clip_ratio": 0.0023134484654292464, "step_time": 209.2750209980004, "entropy_control/entropy": 0.2097417175741633, "entropy_control/target": 0.21118688754598858, "entropy_control/target_ready": 1.0, "entropy_control/target_samples": 5.0, "entropy_control/zeta_used": 0.05, "entropy_control/zeta_next": 0.05, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/update_skipped": 0.0, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.12, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 209.177107, "rollout_overhead_time": 0.0, "peak_allocated_gib": 70.26109647750854, "peak_reserved_gib": 77.83203125, "reward": null }, { "step": 31, "loss": 0.001579820178449154, "grad_norm": 0.0771484375, "learning_rate": 4.000000000000001e-06, "thinking/tokens_mean": 627.3452380952382, "thinking/budget_hit_fraction": 0.14285714285714285, "num_tokens": 933130.0, "completions/mean_length": 4441.25, "completions/min_length": 3212.0, "completions/max_length": 6325.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 4441.25, "completions/min_terminated_length": 3212.0, "completions/max_terminated_length": 6325.0, "tools/call_frequency": 3.375, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/mean": 0.7701637744903564, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/std": 0.2155686467885971, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/mean": 0.4375, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/std": 0.21707837283611298, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/mean": 0.75, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/std": 0.7071067690849304, "reward": 0.9420388340950012, "reward_std": 0.36321714520454407, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.20752463303506374, "cispo_clip_ratio": 0.0, "step_time": 1385.4100905860032, "entropy_control/entropy": 0.20660830092974508, "entropy_control/target": 0.21118688754598858, "entropy_control/target_ready": 1.0, "entropy_control/target_samples": 5.0, "entropy_control/zeta_used": 0.05, "entropy_control/zeta_next": 0.05, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/update_skipped": 0.0, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.124, "adapter_sync_time": 0.0, "generation_time": 1139.629887, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1181.757595, "forward_backward_time": 203.548728, "rollout_overhead_time": 42.127708, "peak_allocated_gib": 70.26109647750854, "peak_reserved_gib": 77.83203125 }, { "step": 32, "loss": 0.0015498446300625801, "grad_norm": 0.0654296875, "learning_rate": 3.8000000000000005e-06, "entropy": 0.2078357059508562, "cispo_clip_ratio": 0.0028741663554683328, "step_time": 203.65134018299796, "entropy_control/entropy": 0.20693874506030674, "entropy_control/target": 0.21118688754598858, "entropy_control/target_ready": 1.0, "entropy_control/target_samples": 5.0, "entropy_control/zeta_used": 0.05, "entropy_control/zeta_next": 0.05, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/update_skipped": 0.0, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.128, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 203.544703, "rollout_overhead_time": 0.0, "peak_allocated_gib": 70.26109647750854, "peak_reserved_gib": 77.83203125, "reward": null }, { "step": 33, "loss": 0.050470951944589615, "grad_norm": 0.0888671875, "learning_rate": 3.6000000000000003e-06, "thinking/tokens_mean": 595.4236111111111, "thinking/budget_hit_fraction": 0.14583333333333334, "num_tokens": 986589.0, "completions/mean_length": 3786.875, "completions/min_length": 2352.0, "completions/max_length": 6311.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 3786.875, "completions/min_terminated_length": 2352.0, "completions/max_terminated_length": 6311.0, "tools/call_frequency": 3.0, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.6095701456069946, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.625, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.0833333432674408, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/mean": 0.8785579204559326, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__native/std": 0.06635081022977829, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/mean": 0.375, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__efficiency/std": 0.20971763134002686, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__size_sensitive_cycle__reliability/std": 0.0, "reward": 0.9690639972686768, "reward_std": 0.14291220903396606, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.20151151157915592, "cispo_clip_ratio": 0.0, "step_time": 1414.5700504090055, "entropy_control/entropy": 0.2087628544028924, "entropy_control/target": 0.21118688754598858, "entropy_control/target_ready": 1.0, "entropy_control/target_samples": 5.0, "entropy_control/zeta_used": 0.05, "entropy_control/zeta_next": 0.05, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/update_skipped": 0.0, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.132, "adapter_sync_time": 0.0, "generation_time": 1220.409603, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1254.805871, "forward_backward_time": 159.672002, "rollout_overhead_time": 34.396268, "peak_allocated_gib": 70.26109647750854, "peak_reserved_gib": 77.83203125 }, { "step": 34, "loss": 0.05107717216014862, "grad_norm": 0.08056640625, "learning_rate": 3.4000000000000005e-06, "entropy": 0.20190092362463474, "cispo_clip_ratio": 0.0029951603792142123, "step_time": 159.75958859299863, "entropy_control/entropy": 0.20911363313327067, "entropy_control/target": 0.21118688754598858, "entropy_control/target_ready": 1.0, "entropy_control/target_samples": 5.0, "entropy_control/zeta_used": 0.05, "entropy_control/zeta_next": 0.05, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/update_skipped": 0.0, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.136, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 159.665871, "rollout_overhead_time": 0.0, "peak_allocated_gib": 70.26109647750854, "peak_reserved_gib": 77.83203125, "reward": null }, { "step": 35, "loss": 0.046646419912576675, "grad_norm": 0.0556640625, "learning_rate": 3.2000000000000003e-06, "thinking/tokens_mean": 765.7163265306123, "thinking/budget_hit_fraction": 0.14285714285714285, "num_tokens": 1056494.0, "completions/mean_length": 4875.125, "completions/min_length": 3071.0, "completions/max_length": 7862.0, "completions/clipped_ratio": 0.125, "completions/mean_terminated_length": 5132.857421875, "completions/min_terminated_length": 3939.0, "completions/max_terminated_length": 7862.0, "tools/call_frequency": 3.125, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/mean": 0.5713491439819336, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/std": 0.23723171651363373, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/mean": 0.3541666865348816, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/std": 0.27368009090423584, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/mean": 0.75, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/std": 0.7071067690849304, "reward": 0.7390574216842651, "reward_std": 0.3830585777759552, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 0.5, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.1837924774736166, "cispo_clip_ratio": 0.0, "step_time": 2039.2121586850008, "entropy_control/entropy": 0.2052942728653329, "entropy_control/target": 0.21118688754598858, "entropy_control/target_ready": 1.0, "entropy_control/target_samples": 5.0, "entropy_control/zeta_used": 0.05, "entropy_control/zeta_next": 0.05, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/update_skipped": 0.0, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.14, "adapter_sync_time": 0.0, "generation_time": 1689.638788, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1748.581729, "forward_backward_time": 290.520022, "rollout_overhead_time": 58.942942, "peak_allocated_gib": 70.26109647750854, "peak_reserved_gib": 77.83203125 }, { "step": 36, "loss": 0.046330079436302185, "grad_norm": 0.0517578125, "learning_rate": 3e-06, "entropy": 0.1836686972528696, "cispo_clip_ratio": 0.0017699807649478316, "step_time": 290.5989363379995, "entropy_control/entropy": 0.20518178344367957, "entropy_control/target": 0.21118688754598858, "entropy_control/target_ready": 1.0, "entropy_control/target_samples": 5.0, "entropy_control/zeta_used": 0.05, "entropy_control/zeta_next": 0.05, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/update_skipped": 0.0, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.144, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 290.498635, "rollout_overhead_time": 0.0, "peak_allocated_gib": 70.26109647750854, "peak_reserved_gib": 77.83203125, "reward": null }, { "step": 37, "loss": 0.06093544885516167, "grad_norm": 0.07763671875, "learning_rate": 2.8000000000000003e-06, "thinking/tokens_mean": 531.2395833333334, "thinking/budget_hit_fraction": 0.13541666666666666, "num_tokens": 1156159.0, "completions/mean_length": 3332.0625, "completions/min_length": 1997.0, "completions/max_length": 5808.5, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 3332.0625, "completions/min_terminated_length": 1997.0, "completions/max_terminated_length": 5808.5, "tools/call_frequency": 2.625, "tools/failure_frequency": 0.11590909399092197, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/mean": 0.6096699237823486, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__native/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/mean": 0.6666666865348816, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__efficiency/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__two_pool_cycle__reliability/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__native/mean": 0.7289921641349792, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__native/std": 0.13682812452316284, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__efficiency/mean": 0.5625000298023224, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__efficiency/std": 0.12789812684059143, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__reliability/std": 0.0, "reward": 0.90006023645401, "reward_std": 0.10524261742830276, "frac_reward_zero_std": 0.5, "frac_groups_zero_std": 0.5, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 1.0, "n_refills": 1.0, "entropy": 0.24642304703593254, "cispo_clip_ratio": 0.001176156336441636, "step_time": 2265.5598083139994, "entropy_control/entropy": 0.24700052264976258, "entropy_control/target": 0.21118688754598858, "entropy_control/target_ready": 1.0, "entropy_control/target_samples": 5.0, "entropy_control/zeta_used": 0.05, "entropy_control/zeta_next": 0.025, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/update_skipped": 0.0, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.148, "adapter_sync_time": 0.0, "generation_time": 2036.460413, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 2097.484425, "forward_backward_time": 167.972686, "rollout_overhead_time": 61.024012, "peak_allocated_gib": 70.26109647750854, "peak_reserved_gib": 77.83203125 }, { "step": 38, "loss": 0.05009032040834427, "grad_norm": 0.0830078125, "learning_rate": 2.6e-06, "entropy": 0.24709582887589931, "cispo_clip_ratio": 0.0028627690044231713, "step_time": 168.06265335401258, "entropy_control/entropy": 0.24758377868775422, "entropy_control/target": 0.21118688754598858, "entropy_control/target_ready": 1.0, "entropy_control/target_samples": 5.0, "entropy_control/zeta_used": 0.025, "entropy_control/zeta_next": 0.0125, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/update_skipped": 0.0, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.152, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 167.951618, "rollout_overhead_time": 0.0, "peak_allocated_gib": 70.26109647750854, "peak_reserved_gib": 77.83203125, "reward": null }, { "step": 39, "loss": 0.006497297435998917, "grad_norm": 0.06103515625, "learning_rate": 2.4000000000000003e-06, "thinking/tokens_mean": 634.1535714285714, "thinking/budget_hit_fraction": 0.16071428571428573, "num_tokens": 1216950.0, "completions/mean_length": 4507.875, "completions/min_length": 2730.0, "completions/max_length": 5361.0, "completions/clipped_ratio": 0.0, "completions/mean_terminated_length": 4507.875, "completions/min_terminated_length": 2730.0, "completions/max_terminated_length": 5361.0, "tools/call_frequency": 3.375, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__native/mean": 0.6413488388061523, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__native/std": 0.21211744844913483, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__efficiency/mean": 0.4375, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__efficiency/std": 0.2509901225566864, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__reliability/mean": 0.75, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__reliability/std": 0.7071067690849304, "reward": 0.8132238388061523, "reward_std": 0.33593258261680603, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 1.0, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.2179273311048746, "cispo_clip_ratio": 0.0, "step_time": 1476.0498073280105, "entropy_control/entropy": 0.21493800920526132, "entropy_control/target": 0.21118688754598858, "entropy_control/target_ready": 1.0, "entropy_control/target_samples": 5.0, "entropy_control/zeta_used": 0.0125, "entropy_control/zeta_next": 0.00625, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/update_skipped": 0.0, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.156, "adapter_sync_time": 0.0, "generation_time": 1226.948879, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1269.687226, "forward_backward_time": 206.248885, "rollout_overhead_time": 42.738347, "peak_allocated_gib": 70.26109647750854, "peak_reserved_gib": 77.83203125 }, { "step": 40, "loss": 0.00487714447081089, "grad_norm": 0.059814453125, "learning_rate": 2.2e-06, "entropy": 0.21740558557212353, "cispo_clip_ratio": 0.0022549554123543203, "step_time": 206.3434993449846, "entropy_control/entropy": 0.2144068599828818, "entropy_control/target": 0.21118688754598858, "entropy_control/target_ready": 1.0, "entropy_control/target_samples": 5.0, "entropy_control/zeta_used": 0.00625, "entropy_control/zeta_next": 0.003125, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/update_skipped": 0.0, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.16, "adapter_sync_time": 0.0, "generation_time": 0.0, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 0.0, "forward_backward_time": 206.236138, "rollout_overhead_time": 0.0, "peak_allocated_gib": 70.26109647750854, "peak_reserved_gib": 77.83203125, "reward": null }, { "step": 41, "loss": 0.08561861515045166, "grad_norm": 0.08251953125, "learning_rate": 2.0000000000000003e-06, "thinking/tokens_mean": 980.8799999999999, "thinking/budget_hit_fraction": 0.27999999999999997, "num_tokens": 1280030.0, "completions/mean_length": 4408.0, "completions/min_length": 3071.0, "completions/max_length": 6718.0, "completions/clipped_ratio": 0.375, "completions/mean_terminated_length": 5210.2001953125, "completions/min_terminated_length": 3783.0, "completions/max_terminated_length": 6718.0, "tools/call_frequency": 2.0, "tools/failure_frequency": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__native/mean": 0.16097012162208557, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__native/std": 0.32194024324417114, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__efficiency/mean": 0.0833333358168602, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__efficiency/std": 0.1666666716337204, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__reliability/mean": -0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__triangular_cycle__reliability/std": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/mean": 0.6074227094650269, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__native/std": 0.013053804636001587, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/mean": 0.5, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__efficiency/std": 0.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/mean": 1.0, "rewards/reward_env__skillhub__dex-arbitrage-routing-gym__competing_cycles__reliability/std": 0.0, "reward": 0.44877973198890686, "reward_std": 0.5373927354812622, "frac_reward_zero_std": 0.0, "frac_groups_zero_std": 0.0, "frac_groups_all_zero": 0.0, "frac_groups_all_one": 0.5, "n_groups_dropped": 0.0, "n_groups_refilled": 0.0, "n_refills": 0.0, "entropy": 0.1304418332874775, "cispo_clip_ratio": 0.0, "step_time": 1491.0382166269774, "entropy_control/entropy": 0.21580574484545761, "entropy_control/target": 0.21118688754598858, "entropy_control/target_ready": 1.0, "entropy_control/target_samples": 5.0, "entropy_control/zeta_used": 0.003125, "entropy_control/zeta_next": 0.0015625, "entropy_control/epsilon_high_used": 0.2, "entropy_control/epsilon_high_next": 0.2, "entropy_control/update_skipped": 0.0, "entropy_control/importance_weight_cap": 1.2, "epoch": 0.164, "adapter_sync_time": 0.0, "generation_time": 1290.075829, "base_reward_time": 0.0, "jepa_reward_time": 0.0, "rollout_prepare_time": 1325.141617, "forward_backward_time": 165.807732, "rollout_overhead_time": 35.065787, "peak_allocated_gib": 70.26109647750854, "peak_reserved_gib": 77.83203125 } ], "holdout": [] } ], "refill_log_audit": [ { "run_id": "1789165288", "method": "dapo-refill", "mode": "no-think", "step": 3, "logged_reward": 0.748327225446701, "parquet_mean": 0.7221509153023362, "delta": -0.026176310144364834, "n_refills": 1.0 }, { "run_id": "1789165288", "method": "dapo-refill", "mode": "no-think", "step": 7, "logged_reward": 0.16936538740992546, "parquet_mean": 0.04062499962747096, "delta": -0.1287403877824545, "n_refills": 1.0 }, { "run_id": "1789165288", "method": "dapo-refill", "mode": "no-think", "step": 9, "logged_reward": 0.4399409890174866, "parquet_mean": 0.4408177591860294, "delta": 0.0008767701685428175, "n_refills": 2.0 }, { "run_id": "1789165288", "method": "dapo-refill", "mode": "no-think", "step": 13, "logged_reward": 0.4414585630098979, "parquet_mean": 0.4426821870729327, "delta": 0.0012236240630348338, "n_refills": 2.0 }, { "run_id": "1789165288", "method": "dapo-refill", "mode": "no-think", "step": 15, "logged_reward": 0.42326606810092926, "parquet_mean": 0.4236089544370771, "delta": 0.0003428863361478185, "n_refills": 1.0 }, { "run_id": "1789165288", "method": "dapo-refill", "mode": "no-think", "step": 21, "logged_reward": 0.43824608623981476, "parquet_mean": 0.4380193211138248, "delta": -0.00022676512598995835, "n_refills": 1.0 }, { "run_id": "1789165288", "method": "dapo-refill", "mode": "no-think", "step": 25, "logged_reward": 0.8305423855781555, "parquet_mean": 0.8293203920125961, "delta": -0.0012219935655594094, "n_refills": 2.0 }, { "run_id": "1789165288", "method": "dapo-refill", "mode": "no-think", "step": 29, "logged_reward": 0.44154128432273865, "parquet_mean": 0.4409255467355251, "delta": -0.0006157375872135606, "n_refills": 2.0 }, { "run_id": "1789165288", "method": "dapo-refill", "mode": "no-think", "step": 31, "logged_reward": 1.046946406364441, "parquet_mean": 1.046293979883194, "delta": -0.0006524264812468594, "n_refills": 1.0 }, { "run_id": "1789165288", "method": "dapo-refill", "mode": "no-think", "step": 33, "logged_reward": 0.8970373074213663, "parquet_mean": 0.8130628922954202, "delta": -0.08397441512594617, "n_refills": 2.0 }, { "run_id": "1789165288", "method": "dapo-refill", "mode": "no-think", "step": 35, "logged_reward": 0.18153761327266693, "parquet_mean": 0.3456965064629912, "delta": 0.16415889319032428, "n_refills": 2.0 }, { "run_id": "1789165288", "method": "dapo-refill", "mode": "no-think", "step": 41, "logged_reward": 0.4937981963157654, "parquet_mean": 0.5411768492311239, "delta": 0.04737865291535848, "n_refills": 1.0 }, { "run_id": "1789165288", "method": "dapo-refill", "mode": "no-think", "step": 43, "logged_reward": 0.8437418043613434, "parquet_mean": 0.8403472529724241, "delta": -0.003394551388919309, "n_refills": 1.0 }, { "run_id": "1789396300", "method": "cispo-repo-r-window5-refill", "mode": "thinking", "step": 37, "logged_reward": 0.90006023645401, "parquet_mean": 0.8969290001317859, "delta": -0.003131236322224118, "n_refills": 1.0 } ] }