{
  "description": "Historical internal multi-hop routing training observations, not a controlled benchmark or a held-out quality evaluation. Extracted 2026-09-08 from recorded training rows; final summary rows excluded. enforce_eager is configuration, not runtime graph-capture evidence. Hardware is omitted because it is not established by these selected artifacts.",
  "source_worktree_commit": "94de10c",
  "runs": [
    {
      "run_id": "1788615027",
      "source_metrics_sha256": "ca5887928320ce54cfc08579ee7638f1c2283423f05b6e5463ed47aca8df0b73",
      "model": "Qwen/Qwen3.6-27B",
      "task": "multi-hop-routing-gym",
      "engine": "vllm",
      "enforce_eager": true,
      "latent": {
        "num_steps": 16,
        "answer_only": true,
        "noise_scale": 0.0,
        "adaptive_stop": false
      },
      "trainer_result": "SUCCESS",
      "steps_recorded": 5,
      "mean_generation_seconds": 437.8930798,
      "mean_step_seconds": 582.6838334140003,
      "rows": [
        {
          "step": 1,
          "generation_time": 678.247263,
          "step_time": 963.0117275200005,
          "reward": 0.041666666666666664,
          "grad_norm": 1.1324882507324219e-06,
          "completions/mean_length": 357.9166666666667,
          "completions/clipped_ratio": 0.08333333333333333
        },
        {
          "step": 2,
          "generation_time": 486.337861,
          "step_time": 604.9699165090005,
          "reward": 0.05359823629260063,
          "grad_norm": 0.14453125,
          "completions/mean_length": 128.5,
          "completions/clipped_ratio": 0.0625
        },
        {
          "step": 3,
          "generation_time": 360.964858,
          "step_time": 431.0270419059998,
          "reward": 0.0625,
          "grad_norm": 0.007293701171875,
          "completions/mean_length": 436.0,
          "completions/clipped_ratio": 0.0
        },
        {
          "step": 4,
          "generation_time": 258.277648,
          "step_time": 379.9356881070007,
          "reward": 0.07222825288772583,
          "grad_norm": 0.015869140625,
          "completions/mean_length": 89.08333333333333,
          "completions/clipped_ratio": 0.0
        },
        {
          "step": 5,
          "generation_time": 405.637769,
          "step_time": 534.474793028,
          "reward": 0.3199015259742737,
          "grad_norm": 0.019287109375,
          "completions/mean_length": 520.5,
          "completions/clipped_ratio": 0.0
        }
      ]
    },
    {
      "run_id": "1788635340",
      "source_metrics_sha256": "72fdc453ac61ed7b72d0d58c5a1dfeae84ad5d64d82923bba27c5999224a712b",
      "model": "Qwen/Qwen3.6-27B",
      "task": "multi-hop-routing-gym",
      "engine": "vllm",
      "enforce_eager": false,
      "latent": {
        "num_steps": 16,
        "answer_only": true,
        "noise_scale": 0.0,
        "adaptive_stop": false
      },
      "trainer_result": "SUCCESS",
      "steps_recorded": 5,
      "mean_generation_seconds": 101.27016400000001,
      "mean_step_seconds": 205.58164714920002,
      "rows": [
        {
          "step": 1,
          "generation_time": 191.101694,
          "step_time": 261.3778865200002,
          "reward": 0.0,
          "grad_norm": 0.0,
          "completions/mean_length": 208.8125,
          "completions/clipped_ratio": 0.1875
        },
        {
          "step": 2,
          "generation_time": 124.850583,
          "step_time": 320.7622503399998,
          "reward": 0.04374999925494194,
          "grad_norm": 3.433342499192804e-11,
          "completions/mean_length": 833.25,
          "completions/clipped_ratio": 0.5
        },
        {
          "step": 3,
          "generation_time": 88.224375,
          "step_time": 185.8081810660001,
          "reward": 0.04374999925494194,
          "grad_norm": 8.86757334228605e-11,
          "completions/mean_length": 553.0,
          "completions/clipped_ratio": 0.25
        },
        {
          "step": 4,
          "generation_time": 31.192981,
          "step_time": 90.98080279099986,
          "reward": 0.0,
          "grad_norm": 0.0,
          "completions/mean_length": 1.0625,
          "completions/clipped_ratio": 0.0
        },
        {
          "step": 5,
          "generation_time": 70.981187,
          "step_time": 168.97911502900024,
          "reward": 0.03125,
          "grad_norm": 8.58306884765625e-06,
          "completions/mean_length": 195.5,
          "completions/clipped_ratio": 0.0
        }
      ]
    },
    {
      "run_id": "1788615660",
      "source_metrics_sha256": "9b5736a62f8301f5c1041f98f25a517a2b421624cadfc294a1058426a6979d9a",
      "model": "Qwen/Qwen3.6-27B",
      "task": "multi-hop-routing-gym",
      "engine": "vllm",
      "enforce_eager": true,
      "latent": {
        "num_steps": 16,
        "answer_only": false,
        "noise_scale": 0.5,
        "adaptive_stop": true
      },
      "trainer_result": "SUCCESS",
      "steps_recorded": 5,
      "mean_generation_seconds": 391.78303900000003,
      "mean_step_seconds": 447.7621936147997,
      "rows": [
        {
          "step": 1,
          "generation_time": 718.041317,
          "step_time": 759.4017434329994,
          "reward": 0.0,
          "grad_norm": 0.0,
          "completions/mean_length": 247.75,
          "completions/clipped_ratio": 0.1875
        },
        {
          "step": 2,
          "generation_time": 505.424118,
          "step_time": 662.0837822300009,
          "reward": 0.03125,
          "grad_norm": 1.765625,
          "completions/mean_length": 282.5,
          "completions/clipped_ratio": 0.125
        },
        {
          "step": 3,
          "generation_time": 539.963793,
          "step_time": 587.9557331139995,
          "reward": 0.0625,
          "grad_norm": 1.7578125,
          "completions/mean_length": 716.5,
          "completions/clipped_ratio": 0.25
        },
        {
          "step": 4,
          "generation_time": 114.311661,
          "step_time": 137.98219719399913,
          "reward": 0.09540529549121857,
          "grad_norm": 17.75,
          "completions/mean_length": 59.5,
          "completions/clipped_ratio": 0.0
        },
        {
          "step": 5,
          "generation_time": 81.174306,
          "step_time": 91.38751210299961,
          "reward": 0.0,
          "grad_norm": 0.0,
          "completions/mean_length": 15.125,
          "completions/clipped_ratio": 0.0
        }
      ]
    },
    {
      "run_id": "1788700311",
      "source_metrics_sha256": "1873c984d6f9428284f568eb5016a24814873be09de9cc65c0ca41147b1da7ee",
      "model": "Qwen/Qwen3.6-27B",
      "task": "multi-hop-routing-gym",
      "engine": "vllm",
      "enforce_eager": false,
      "latent": {
        "num_steps": 16,
        "answer_only": false,
        "noise_scale": 0.5,
        "adaptive_stop": true
      },
      "trainer_result": "SUCCESS",
      "steps_recorded": 50,
      "mean_generation_seconds": 132.71421934,
      "mean_step_seconds": 171.22924674054045,
      "rows": [
        {
          "step": 1,
          "generation_time": 153.907049,
          "step_time": 184.6260055519997,
          "reward": 0.0,
          "grad_norm": 0.0,
          "completions/mean_length": 74.0,
          "completions/clipped_ratio": 0.0625
        },
        {
          "step": 2,
          "generation_time": 143.461285,
          "step_time": 151.71937255000057,
          "reward": 0.0,
          "grad_norm": 0.0,
          "completions/mean_length": 153.625,
          "completions/clipped_ratio": 0.125
        },
        {
          "step": 3,
          "generation_time": 211.765386,
          "step_time": 412.39433006900026,
          "reward": 0.0625,
          "grad_norm": 1.0234375,
          "completions/mean_length": 365.0,
          "completions/clipped_ratio": 0.08333333333333333
        },
        {
          "step": 4,
          "generation_time": 58.570354,
          "step_time": 100.31243031600025,
          "reward": 0.0625,
          "grad_norm": 5.09375,
          "completions/mean_length": 177.5,
          "completions/clipped_ratio": 0.0
        },
        {
          "step": 5,
          "generation_time": 62.355803,
          "step_time": 80.65977063199989,
          "reward": 0.0625,
          "grad_norm": 17.125,
          "completions/mean_length": 314.75,
          "completions/clipped_ratio": 0.25
        },
        {
          "step": 6,
          "generation_time": 70.447478,
          "step_time": 113.36108138400004,
          "reward": 0.03125,
          "grad_norm": 2.71875,
          "completions/mean_length": 145.25,
          "completions/clipped_ratio": 0.0
        },
        {
          "step": 7,
          "generation_time": 147.227502,
          "step_time": 214.14704806200007,
          "reward": 0.490191787481308,
          "grad_norm": 3.890625,
          "completions/mean_length": 765.0,
          "completions/clipped_ratio": 0.0
        },
        {
          "step": 8,
          "generation_time": 205.456099,
          "step_time": 254.20763617000102,
          "reward": 0.125,
          "grad_norm": 0.7265625,
          "completions/mean_length": 1274.0,
          "completions/clipped_ratio": 0.25
        },
        {
          "step": 9,
          "generation_time": 104.708136,
          "step_time": 132.60213805400053,
          "reward": 0.0625,
          "grad_norm": 1.625,
          "completions/mean_length": 591.25,
          "completions/clipped_ratio": 0.25
        },
        {
          "step": 10,
          "generation_time": 142.413447,
          "step_time": 169.50746533599977,
          "reward": 0.125,
          "grad_norm": 22.375,
          "completions/mean_length": 617.25,
          "completions/clipped_ratio": 0.25
        },
        {
          "step": 11,
          "generation_time": 31.944384,
          "step_time": 57.973420811000324,
          "reward": 0.0625,
          "grad_norm": 25.25,
          "completions/mean_length": 75.0,
          "completions/clipped_ratio": 0.0
        },
        {
          "step": 12,
          "generation_time": 72.569138,
          "step_time": 101.2533642569988,
          "reward": 0.03125,
          "grad_norm": 2.59375,
          "completions/mean_length": 161.25,
          "completions/clipped_ratio": 0.0
        },
        {
          "step": 13,
          "generation_time": 181.708616,
          "step_time": 218.17745803900016,
          "reward": 0.18548232316970825,
          "grad_norm": 10.1875,
          "completions/mean_length": 591.625,
          "completions/clipped_ratio": 0.25
        },
        {
          "step": 14,
          "generation_time": 111.671596,
          "step_time": 150.24707495700113,
          "reward": 0.125,
          "grad_norm": 1.5625,
          "completions/mean_length": 719.5,
          "completions/clipped_ratio": 0.25
        },
        {
          "step": 15,
          "generation_time": 191.435669,
          "step_time": 204.91809722800008,
          "reward": 0.014583333084980646,
          "grad_norm": 1.203125,
          "completions/mean_length": 427.0,
          "completions/clipped_ratio": 0.3333333333333333
        },
        {
          "step": 16,
          "generation_time": 173.696934,
          "step_time": 238.74331639200045,
          "reward": 0.1875,
          "grad_norm": 0.8984375,
          "completions/mean_length": 1114.0,
          "completions/clipped_ratio": 0.0
        },
        {
          "step": 17,
          "generation_time": 155.031088,
          "step_time": 165.1811783510011,
          "reward": 0.0625,
          "grad_norm": 8.25,
          "completions/mean_length": 875.25,
          "completions/clipped_ratio": 0.5
        },
        {
          "step": 18,
          "generation_time": 68.36075,
          "step_time": 99.0910901770003,
          "reward": 0.0625,
          "grad_norm": 2.0625,
          "completions/mean_length": 305.0,
          "completions/clipped_ratio": 0.0
        },
        {
          "step": 19,
          "generation_time": 94.681223,
          "step_time": 147.81825519199901,
          "reward": 0.10625000298023224,
          "grad_norm": 1.359375,
          "completions/mean_length": 490.0,
          "completions/clipped_ratio": 0.0
        },
        {
          "step": 20,
          "generation_time": 64.193264,
          "step_time": 80.68161505099943,
          "reward": 0.0625,
          "grad_norm": 52.5,
          "completions/mean_length": 324.0,
          "completions/clipped_ratio": 0.25
        },
        {
          "step": 21,
          "generation_time": 81.131636,
          "step_time": 104.26625990799948,
          "reward": 0.24527229368686676,
          "grad_norm": 4.3125,
          "completions/mean_length": 484.75,
          "completions/clipped_ratio": 0.25
        },
        {
          "step": 22,
          "generation_time": 129.501142,
          "step_time": 171.3919743159995,
          "reward": 0.125,
          "grad_norm": 1.09375,
          "completions/mean_length": 780.75,
          "completions/clipped_ratio": 0.0
        },
        {
          "step": 23,
          "generation_time": 127.864316,
          "step_time": 180.90108112100097,
          "reward": 0.10625000298023224,
          "grad_norm": 0.6640625,
          "completions/mean_length": 734.75,
          "completions/clipped_ratio": 0.0
        },
        {
          "step": 24,
          "generation_time": 66.82585,
          "step_time": 83.32359174299927,
          "reward": 0.0625,
          "grad_norm": 7.15625,
          "completions/mean_length": 349.5,
          "completions/clipped_ratio": 0.25
        },
        {
          "step": 25,
          "generation_time": 166.620807,
          "step_time": 217.8260671829994,
          "reward": 0.125,
          "grad_norm": 1.578125,
          "completions/mean_length": 988.75,
          "completions/clipped_ratio": 0.25
        },
        {
          "step": 26,
          "generation_time": 140.37174,
          "step_time": 188.237178159,
          "reward": 0.2744309902191162,
          "grad_norm": 3.140625,
          "completions/mean_length": 950.5,
          "completions/clipped_ratio": 0.25
        },
        {
          "step": 27,
          "generation_time": 172.786941,
          "step_time": 193.4095703490002,
          "reward": 0.0625,
          "grad_norm": 2.25,
          "completions/mean_length": 1445.0,
          "completions/clipped_ratio": 0.75
        },
        {
          "step": 28,
          "generation_time": 188.196869,
          "step_time": 214.45892536000156,
          "reward": 0.1248939260840416,
          "grad_norm": 8.4375,
          "completions/mean_length": 642.125,
          "completions/clipped_ratio": 0.5
        },
        {
          "step": 29,
          "generation_time": 141.919758,
          "step_time": 181.2195701039991,
          "reward": 0.125,
          "grad_norm": 0.65234375,
          "completions/mean_length": 913.0,
          "completions/clipped_ratio": 0.25
        },
        {
          "step": 30,
          "generation_time": 177.787417,
          "step_time": 223.07321115299965,
          "reward": 0.36422258615493774,
          "grad_norm": 2.3125,
          "completions/mean_length": 1144.0,
          "completions/clipped_ratio": 0.25
        },
        {
          "step": 31,
          "generation_time": 105.966378,
          "step_time": 151.32950187799997,
          "reward": 0.3077574372291565,
          "grad_norm": 6.1875,
          "completions/mean_length": 588.25,
          "completions/clipped_ratio": 0.25
        },
        {
          "step": 32,
          "generation_time": 135.321985,
          "step_time": 153.19137642500027,
          "reward": 0.0625,
          "grad_norm": 22.75,
          "completions/mean_length": 885.5,
          "completions/clipped_ratio": 0.75
        },
        {
          "step": 33,
          "generation_time": 127.053179,
          "step_time": 174.27563132000068,
          "reward": 0.3697729706764221,
          "grad_norm": 7.09375,
          "completions/mean_length": 760.25,
          "completions/clipped_ratio": 0.5
        },
        {
          "step": 34,
          "generation_time": 121.852685,
          "step_time": 159.6496098209991,
          "reward": 0.125,
          "grad_norm": 0.9375,
          "completions/mean_length": 772.75,
          "completions/clipped_ratio": 0.25
        },
        {
          "step": 35,
          "generation_time": 129.210377,
          "step_time": 174.35183388699988,
          "reward": 0.42746734619140625,
          "grad_norm": 8.25,
          "completions/mean_length": 807.75,
          "completions/clipped_ratio": 0.25
        },
        {
          "step": 36,
          "generation_time": 119.355644,
          "step_time": 155.80788011200093,
          "reward": 0.125,
          "grad_norm": 1.28125,
          "completions/mean_length": 912.5,
          "completions/clipped_ratio": 0.25
        },
        {
          "step": 37,
          "generation_time": 150.456708,
          "step_time": 181.39097383100125,
          "reward": 0.125,
          "grad_norm": 3.296875,
          "completions/mean_length": 1047.0,
          "completions/clipped_ratio": 0.5
        },
        {
          "step": 38,
          "generation_time": 335.308943,
          "step_time": 339.5886614249994,
          "reward": 0.035416667660077415,
          "grad_norm": 0.0,
          "completions/mean_length": 817.0,
          "completions/clipped_ratio": 0.6666666666666666
        },
        {
          "step": 39,
          "generation_time": 152.37177,
          "step_time": 195.60635829400235,
          "reward": 0.125,
          "grad_norm": 1.7421875,
          "completions/mean_length": 986.25,
          "completions/clipped_ratio": 0.0
        },
        {
          "step": 40,
          "generation_time": 124.619204,
          "step_time": 167.01287960999798,
          "reward": 0.2811504602432251,
          "grad_norm": 3.234375,
          "completions/mean_length": 732.5,
          "completions/clipped_ratio": 0.0
        },
        {
          "step": 41,
          "generation_time": 155.874303,
          "step_time": 223.8005646970032,
          "reward": 0.3597930669784546,
          "grad_norm": 1.5625,
          "completions/mean_length": 894.75,
          "completions/clipped_ratio": 0.25
        },
        {
          "step": 42,
          "generation_time": 75.550532,
          "step_time": 129.74277494700073,
          "reward": 0.30831724405288696,
          "grad_norm": 11.0,
          "completions/mean_length": 307.0,
          "completions/clipped_ratio": 0.0
        },
        {
          "step": 43,
          "generation_time": 160.359862,
          "step_time": 186.18334202600272,
          "reward": 0.0625,
          "grad_norm": 1.34375,
          "completions/mean_length": 1236.75,
          "completions/clipped_ratio": 0.75
        },
        {
          "step": 44,
          "generation_time": 63.992317,
          "step_time": 102.96383784099999,
          "reward": 0.37260547280311584,
          "grad_norm": 27.875,
          "completions/mean_length": 252.0,
          "completions/clipped_ratio": 0.0
        },
        {
          "step": 45,
          "generation_time": 160.921763,
          "step_time": 209.52406542800236,
          "reward": 0.125,
          "grad_norm": 1.828125,
          "completions/mean_length": 976.75,
          "completions/clipped_ratio": 0.25
        },
        {
          "step": 46,
          "generation_time": 112.784435,
          "step_time": 151.9737875660012,
          "reward": 0.125,
          "grad_norm": 2.125,
          "completions/mean_length": 753.5,
          "completions/clipped_ratio": 0.25
        },
        {
          "step": 47,
          "generation_time": 172.658522,
          "step_time": 196.83646588300326,
          "reward": 0.2000349462032318,
          "grad_norm": 1.5,
          "completions/mean_length": 1255.5,
          "completions/clipped_ratio": 0.75
        },
        {
          "step": 48,
          "generation_time": 127.82151,
          "step_time": 160.54863078400376,
          "reward": 0.21554027497768402,
          "grad_norm": 6.84375,
          "completions/mean_length": 823.5,
          "completions/clipped_ratio": 0.5
        },
        {
          "step": 49,
          "generation_time": 145.397047,
          "step_time": 186.94676491500104,
          "reward": 0.2526083290576935,
          "grad_norm": 20.5,
          "completions/mean_length": 889.75,
          "completions/clipped_ratio": 0.5
        },
        {
          "step": 50,
          "generation_time": 120.222126,
          "step_time": 125.00781836100032,
          "reward": 0.04374999925494194,
          "grad_norm": 34.5,
          "completions/mean_length": 714.25,
          "completions/clipped_ratio": 0.5
        }
      ]
    }
  ]
}