# DEX trainer results and reward datapoints Snapshot: 2026-09-15T01:19:08.614705+00:00 - 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. ## Starting model / no-think / 1789188652 Steps: 0/50. Holdout native: 0.38555043464607947. Successes: 16/40. | Optimizer step | Logged rollout reward | Step seconds | Peak allocated GiB | Peak reserved GiB | | --- | --- | --- | --- | --- | | Family | Seed | Sample | Native | Efficiency | Reliability | Success | | --- | --- | --- | --- | --- | --- | --- | | two_pool_cycle | 900000 | 0 | 0.615643296169515 | 0.33333333333333337 | 1.0 | True | | two_pool_cycle | 900000 | 1 | 0.6156419091142415 | 0.33333333333333337 | 1.0 | True | | two_pool_cycle | 900000 | 2 | 0.6156430169577571 | 0.33333333333333337 | 1.0 | True | | two_pool_cycle | 900000 | 3 | 0.175 | 0.0 | -1.0 | False | | two_pool_cycle | 900001 | 0 | 0.6154821465692866 | 0.33333333333333337 | 1.0 | True | | two_pool_cycle | 900001 | 1 | 0.25 | 0.0 | -1.0 | False | | two_pool_cycle | 900001 | 2 | 0.175 | 0.0 | -1.0 | False | | two_pool_cycle | 900001 | 3 | 0.175 | 0.0 | -1.0 | False | | triangular_cycle | 900000 | 0 | 0.175 | 0.0 | -1.0 | False | | triangular_cycle | 900000 | 1 | 0.5848008209671872 | 0.33333333333333337 | 1.0 | True | | triangular_cycle | 900000 | 2 | 0.175 | 0.0 | -1.0 | False | | triangular_cycle | 900000 | 3 | 0.175 | 0.0 | -1.0 | False | | triangular_cycle | 900001 | 0 | 0.175 | 0.16666666666666663 | 1.0 | False | | triangular_cycle | 900001 | 1 | 0.175 | 0.0 | -1.0 | False | | triangular_cycle | 900001 | 2 | 0.175 | 0.0 | -1.0 | False | | triangular_cycle | 900001 | 3 | 0.175 | 0.0 | -1.0 | False | | competing_cycles | 900000 | 0 | 0.25 | 0.0 | -1.0 | False | | competing_cycles | 900000 | 1 | 0.25 | 0.0 | -1.0 | False | | competing_cycles | 900000 | 2 | 0.25 | 0.0 | -1.0 | False | | competing_cycles | 900000 | 3 | 0.25 | 0.0 | -1.0 | False | | competing_cycles | 900001 | 0 | 0.6192995951616761 | 0.33333333333333337 | 1.0 | True | | competing_cycles | 900001 | 1 | 0.25 | 0.0 | -1.0 | False | | competing_cycles | 900001 | 2 | 0.25 | 0.0 | -1.0 | False | | competing_cycles | 900001 | 3 | 0.6212167624035315 | 0.5 | 1.0 | True | | four_hop_cycle | 900000 | 0 | 0.6161298837365798 | 0.5 | 1.0 | True | | four_hop_cycle | 900000 | 1 | 0.6161282892735525 | 0.5 | 1.0 | True | | four_hop_cycle | 900000 | 2 | 0.6179612389087088 | 0.6666666666666667 | 1.0 | True | | four_hop_cycle | 900000 | 3 | 0.6153297728643924 | 0.33333333333333337 | 1.0 | True | | four_hop_cycle | 900001 | 0 | 0.175 | 0.0 | -1.0 | False | | four_hop_cycle | 900001 | 1 | 0.6163578745967674 | 0.5 | 1.0 | True | | four_hop_cycle | 900001 | 2 | 0.25 | 0.0 | -1.0 | False | | four_hop_cycle | 900001 | 3 | 0.6140284978433885 | 0.33333333333333337 | 1.0 | True | | size_sensitive_cycle | 900000 | 0 | 0.175 | 0.0 | -1.0 | False | | size_sensitive_cycle | 900000 | 1 | 0.8308833579638738 | 0.33333333333333337 | 1.0 | True | | size_sensitive_cycle | 900000 | 2 | 0.175 | 0.0 | -1.0 | False | | size_sensitive_cycle | 900000 | 3 | 0.175 | 0.0 | -1.0 | False | | size_sensitive_cycle | 900001 | 0 | 0.8287354292640952 | 0.33333333333333337 | 1.0 | True | | size_sensitive_cycle | 900001 | 1 | 0.25 | 0.0 | -1.0 | False | | size_sensitive_cycle | 900001 | 2 | 0.8287354940486263 | 0.33333333333333337 | 1.0 | True | | size_sensitive_cycle | 900001 | 3 | 0.25 | 0.0 | -1.0 | False | ## GRPO / no-think / 1789151457 Steps: 50/50. Holdout native: 0.6423985158115345. Successes: 38/40. | Optimizer step | Logged rollout reward | Step seconds | Peak allocated GiB | Peak reserved GiB | | --- | --- | --- | --- | --- | | 1 | 0.4792949855327606 | 352.69348045599986 | — | — | | 2 | — | 82.91500510800006 | — | — | | 3 | 0.6665704846382141 | 258.22600335200013 | — | — | | 4 | — | 73.627856651 | — | — | | 5 | 0.717830240726471 | 292.5127315850002 | — | — | | 6 | — | 74.01661177599931 | — | — | | 7 | 0.51976478099823 | 337.1979487909998 | — | — | | 8 | — | 96.67439458799981 | — | — | | 9 | 0.4425812363624573 | 246.38061882300053 | — | — | | 10 | — | 91.90123079600062 | — | — | | 11 | 0.8345041275024414 | 235.28879666799958 | — | — | | 12 | — | 79.40832818699982 | — | — | | 13 | 0.5374302864074707 | 308.6096049810003 | — | — | | 14 | — | 92.64190135599938 | — | — | | 15 | 0.29846981167793274 | 210.34915412499868 | — | — | | 16 | — | 80.62655930799974 | — | — | | 17 | 0.7304967641830444 | 301.0209074270001 | — | — | | 18 | — | 84.37778744900015 | — | — | | 19 | 0.9400649666786194 | 221.8670776459976 | — | — | | 20 | — | 83.84809216400117 | — | — | | 21 | 0.440828800201416 | 333.5141935849979 | — | — | | 22 | — | 96.38356308099901 | — | — | | 23 | 0.9419509172439575 | 245.29134741899816 | — | — | | 24 | — | 82.59226387000035 | — | — | | 25 | 0.8341889977455139 | 277.141689441999 | — | — | | 26 | — | 88.20175407399893 | — | — | | 27 | 0.9461063146591187 | 251.29658381099944 | — | — | | 28 | — | 87.88071318700077 | — | — | | 29 | 0.5352011919021606 | 316.6319273319996 | — | — | | 30 | — | 92.2031235839977 | — | — | | 31 | 0.9097099304199219 | 290.5032301099991 | — | — | | 32 | — | 75.1522636180016 | — | — | | 33 | 0.9415267705917358 | 263.0930055120007 | — | — | | 34 | — | 83.42563158399844 | — | — | | 35 | 0.246196448802948 | 343.58754661199964 | — | — | | 36 | — | 96.65855460000057 | — | — | | 37 | 0.843722939491272 | 270.79194448 | — | — | | 38 | — | 83.78387444800137 | — | — | | 39 | 0.8493311405181885 | 195.94552210799975 | — | — | | 40 | — | 64.75038468500134 | — | — | | 41 | 0.7462599277496338 | 260.5962830739991 | — | — | | 42 | — | 84.02459510299923 | — | — | | 43 | 0.8476365208625793 | 208.75057098300022 | — | — | | 44 | — | 70.72571368899844 | — | — | | 45 | 0.5456529855728149 | 289.95140555000035 | — | — | | 46 | — | 77.75101602400173 | — | — | | 47 | 0.9542244672775269 | 191.2153643629972 | — | — | | 48 | — | 73.5989647940005 | — | — | | 49 | 0.844783365726471 | 274.59873056799916 | — | — | | 50 | — | 79.95465855900147 | — | — | | Family | Seed | Sample | Native | Efficiency | Reliability | Success | | --- | --- | --- | --- | --- | --- | --- | | two_pool_cycle | 900000 | 0 | 0.6155619431579329 | 0.33333333333333337 | 1.0 | True | | two_pool_cycle | 900000 | 1 | 0.615560891281058 | 0.33333333333333337 | 1.0 | True | | two_pool_cycle | 900000 | 2 | 0.6155606664493392 | 0.33333333333333337 | 1.0 | True | | two_pool_cycle | 900000 | 3 | 0.6164592739162532 | 0.5 | 1.0 | True | | two_pool_cycle | 900001 | 0 | 0.6175072328484663 | 0.33333333333333337 | 1.0 | True | | two_pool_cycle | 900001 | 1 | 0.6175057798087202 | 0.33333333333333337 | 1.0 | True | | two_pool_cycle | 900001 | 2 | 0.6175061121677046 | 0.33333333333333337 | 1.0 | True | | two_pool_cycle | 900001 | 3 | 0.6175069963866499 | 0.33333333333333337 | 1.0 | True | | triangular_cycle | 900000 | 0 | 0.6169302065811881 | 0.5 | 1.0 | True | | triangular_cycle | 900000 | 1 | 0.6184861076618522 | 0.6666666666666667 | 1.0 | True | | triangular_cycle | 900000 | 2 | 0.6184856035464226 | 0.6666666666666667 | 1.0 | True | | triangular_cycle | 900000 | 3 | 0.618486948443024 | 0.6666666666666667 | 1.0 | True | | triangular_cycle | 900001 | 0 | 0.6176954349414234 | 0.5 | 1.0 | True | | triangular_cycle | 900001 | 1 | 0.6176957130694314 | 0.5 | 1.0 | True | | triangular_cycle | 900001 | 2 | 0.6184896583035329 | 0.6666666666666667 | 1.0 | True | | triangular_cycle | 900001 | 3 | 0.6184886158270289 | 0.6666666666666667 | 1.0 | True | | competing_cycles | 900000 | 0 | 0.25 | 0.0 | -1.0 | False | | competing_cycles | 900000 | 1 | 0.6201801608919191 | 0.5 | 1.0 | True | | competing_cycles | 900000 | 2 | 0.25 | 0.0 | -1.0 | False | | competing_cycles | 900000 | 3 | 0.6182885244004738 | 0.33333333333333337 | 1.0 | True | | competing_cycles | 900001 | 0 | 0.6185548230263636 | 0.33333333333333337 | 1.0 | True | | competing_cycles | 900001 | 1 | 0.6207064180086292 | 0.5 | 1.0 | True | | competing_cycles | 900001 | 2 | 0.6207075640003088 | 0.5 | 1.0 | True | | competing_cycles | 900001 | 3 | 0.6207058199470863 | 0.5 | 1.0 | True | | four_hop_cycle | 900000 | 0 | 0.6175192419545681 | 0.6666666666666667 | 1.0 | True | | four_hop_cycle | 900000 | 1 | 0.6167561987402518 | 0.5 | 1.0 | True | | four_hop_cycle | 900000 | 2 | 0.6175174288668923 | 0.6666666666666667 | 1.0 | True | | four_hop_cycle | 900000 | 3 | 0.6167551775519118 | 0.5 | 1.0 | True | | four_hop_cycle | 900001 | 0 | 0.6168364043457519 | 0.6666666666666667 | 1.0 | True | | four_hop_cycle | 900001 | 1 | 0.616834342416455 | 0.6666666666666667 | 1.0 | True | | four_hop_cycle | 900001 | 2 | 0.6168348950551854 | 0.6666666666666667 | 1.0 | True | | four_hop_cycle | 900001 | 3 | 0.6154729421455175 | 0.33333333333333337 | 1.0 | True | | size_sensitive_cycle | 900000 | 0 | 0.8346269127250262 | 0.5 | 1.0 | True | | size_sensitive_cycle | 900000 | 1 | 0.8346271191178777 | 0.5 | 1.0 | True | | size_sensitive_cycle | 900000 | 2 | 0.8346283987501825 | 0.5 | 1.0 | True | | size_sensitive_cycle | 900000 | 3 | 0.8346278914886784 | 0.5 | 1.0 | True | | size_sensitive_cycle | 900001 | 0 | 0.8316970841532554 | 0.5 | 1.0 | True | | size_sensitive_cycle | 900001 | 1 | 0.8316978576102672 | 0.5 | 1.0 | True | | size_sensitive_cycle | 900001 | 2 | 0.8316978253037827 | 0.5 | 1.0 | True | | size_sensitive_cycle | 900001 | 3 | 0.8307404175709702 | 0.33333333333333337 | 1.0 | True | ## DAPO / no-think / 1789152686 Steps: 50/50. Holdout native: 0.646690062246073. Successes: 39/40. | Optimizer step | Logged rollout reward | Step seconds | Peak allocated GiB | Peak reserved GiB | | --- | --- | --- | --- | --- | | 1 | 0.4792949855327606 | 340.52523792299996 | — | — | | 2 | — | 82.57729190400005 | — | — | | 3 | 0.8323149681091309 | 225.7903535270002 | — | — | | 4 | — | 80.15263358700031 | — | — | | 5 | 0.8322824239730835 | 233.43878169599998 | — | — | | 6 | — | 80.14518817899989 | — | — | | 7 | 0.5201954245567322 | 259.96634932300003 | — | — | | 8 | — | 76.91976198899988 | — | — | | 9 | 0.3229309320449829 | 326.76431794399923 | — | — | | 10 | — | 91.97408711800017 | — | — | | 11 | 0.7205997705459595 | 273.963373522 | — | — | | 12 | — | 79.09812102199976 | — | — | | 13 | 0.3326573371887207 | 287.9703269009997 | — | — | | 14 | — | 94.26841234099993 | — | — | | 15 | 0.6651886701583862 | 247.40627943599975 | — | — | | 16 | — | 77.28687519600044 | — | — | | 17 | 0.7395214438438416 | 254.2655020920006 | — | — | | 18 | — | 83.31394233899982 | — | — | | 19 | 0.5656735897064209 | 272.061771614 | — | — | | 20 | — | 84.94119402100023 | — | — | | 21 | 0.5392618775367737 | 277.84228195500054 | — | — | | 22 | — | 89.41478857799848 | — | — | | 23 | 0.8171144723892212 | 269.0507772079991 | — | — | | 24 | — | 82.55980333700063 | — | — | | 25 | 0.8402920961380005 | 246.0319621259996 | — | — | | 26 | — | 87.54783761100043 | — | — | | 27 | 0.4136379063129425 | 434.5469909829999 | — | — | | 28 | — | 90.39110960100152 | — | — | | 29 | 0.5359797477722168 | 309.39598267899873 | — | — | | 30 | — | 94.16109022100045 | — | — | | 31 | 0.7864433526992798 | 290.2619876309991 | — | — | | 32 | — | 82.46991769599936 | — | — | | 33 | 0.8122120499610901 | 271.0022347350023 | — | — | | 34 | — | 78.72165960399889 | — | — | | 35 | 0.14896363019943237 | 312.12456131 | — | — | | 36 | — | 83.60138981300133 | — | — | | 37 | 0.84920334815979 | 212.67323126400152 | — | — | | 38 | — | 82.06627910099905 | — | — | | 39 | 0.8486765623092651 | 158.74121129800187 | — | — | | 40 | — | 61.216063511001266 | — | — | | 41 | 0.7459396123886108 | 257.0219260349986 | — | — | | 42 | — | 84.34784766300072 | — | — | | 43 | 0.7485491037368774 | 248.04164702799972 | — | — | | 44 | — | 78.56714497000394 | — | — | | 45 | 0.5448254346847534 | 291.9126464979945 | — | — | | 46 | — | 89.14526429899888 | — | — | | 47 | 0.7031150460243225 | 243.1568956810006 | — | — | | 48 | — | 90.61540986599903 | — | — | | 49 | 0.9538475275039673 | 194.973901905998 | — | — | | 50 | — | 74.52270755099926 | — | — | | Family | Seed | Sample | Native | Efficiency | Reliability | Success | | --- | --- | --- | --- | --- | --- | --- | | two_pool_cycle | 900000 | 0 | 0.617827415078968 | 0.6666666666666667 | 1.0 | True | | two_pool_cycle | 900000 | 1 | 0.6178268262001205 | 0.6666666666666667 | 1.0 | True | | two_pool_cycle | 900000 | 2 | 0.6178278680985384 | 0.6666666666666667 | 1.0 | True | | two_pool_cycle | 900000 | 3 | 0.6178281013971713 | 0.6666666666666667 | 1.0 | True | | two_pool_cycle | 900001 | 0 | 0.6177086905727631 | 0.6666666666666667 | 1.0 | True | | two_pool_cycle | 900001 | 1 | 0.6177077954531537 | 0.6666666666666667 | 1.0 | True | | two_pool_cycle | 900001 | 2 | 0.6177076480229584 | 0.6666666666666667 | 1.0 | True | | two_pool_cycle | 900001 | 3 | 0.6168370557757319 | 0.33333333333333337 | 1.0 | True | | triangular_cycle | 900000 | 0 | 0.6196029689418718 | 0.6666666666666667 | 1.0 | True | | triangular_cycle | 900000 | 1 | 0.6196020525787467 | 0.6666666666666667 | 1.0 | True | | triangular_cycle | 900000 | 2 | 0.6196037362934148 | 0.6666666666666667 | 1.0 | True | | triangular_cycle | 900000 | 3 | 0.6196031742367349 | 0.6666666666666667 | 1.0 | True | | triangular_cycle | 900001 | 0 | 0.6194053855289625 | 0.5 | 1.0 | True | | triangular_cycle | 900001 | 1 | 0.6200496131721656 | 0.6666666666666667 | 1.0 | True | | triangular_cycle | 900001 | 2 | 0.6200484743316188 | 0.6666666666666667 | 1.0 | True | | triangular_cycle | 900001 | 3 | 0.6200497891838649 | 0.6666666666666667 | 1.0 | True | | competing_cycles | 900000 | 0 | 0.618495692979087 | 0.33333333333333337 | 1.0 | True | | competing_cycles | 900000 | 1 | 0.6207112328656104 | 0.5 | 1.0 | True | | competing_cycles | 900000 | 2 | 0.6207140123110011 | 0.5 | 1.0 | True | | competing_cycles | 900000 | 3 | 0.62071255280854 | 0.5 | 1.0 | True | | competing_cycles | 900001 | 0 | 0.6222763451244105 | 0.5 | 1.0 | True | | competing_cycles | 900001 | 1 | 0.6222766070858753 | 0.5 | 1.0 | True | | competing_cycles | 900001 | 2 | 0.6222768966053871 | 0.5 | 1.0 | True | | competing_cycles | 900001 | 3 | 0.6222726313160121 | 0.5 | 1.0 | True | | four_hop_cycle | 900000 | 0 | 0.6182188046812087 | 0.6666666666666667 | 1.0 | True | | four_hop_cycle | 900000 | 1 | 0.0 | 0.0 | -1.0 | False | | four_hop_cycle | 900000 | 2 | 0.6182191455836307 | 0.6666666666666667 | 1.0 | True | | four_hop_cycle | 900000 | 3 | 0.6182196745883963 | 0.6666666666666667 | 1.0 | True | | four_hop_cycle | 900001 | 0 | 0.6183858537745408 | 0.6666666666666667 | 1.0 | True | | four_hop_cycle | 900001 | 1 | 0.6183862626624119 | 0.6666666666666667 | 1.0 | True | | four_hop_cycle | 900001 | 2 | 0.6183844270508299 | 0.6666666666666667 | 1.0 | True | | four_hop_cycle | 900001 | 3 | 0.6183847253722361 | 0.6666666666666667 | 1.0 | True | | size_sensitive_cycle | 900000 | 0 | 0.8345685745689324 | 0.5 | 1.0 | True | | size_sensitive_cycle | 900000 | 1 | 0.8345688851902214 | 0.5 | 1.0 | True | | size_sensitive_cycle | 900000 | 2 | 0.834567592653991 | 0.5 | 1.0 | True | | size_sensitive_cycle | 900000 | 3 | 0.8352183059394792 | 0.6666666666666667 | 1.0 | True | | size_sensitive_cycle | 900001 | 0 | 0.832439310785638 | 0.5 | 1.0 | True | | size_sensitive_cycle | 900001 | 1 | 0.8333153584278061 | 0.6666666666666667 | 1.0 | True | | size_sensitive_cycle | 900001 | 2 | 0.8333146420131576 | 0.6666666666666667 | 1.0 | True | | size_sensitive_cycle | 900001 | 3 | 0.8324383605877279 | 0.5 | 1.0 | True | ## DAPO-refill / no-think / 1789165288 Steps: 50/50. Holdout native: 0.5839457972260582. Successes: 34/40. | Optimizer step | Logged rollout reward | Step seconds | Peak allocated GiB | Peak reserved GiB | | --- | --- | --- | --- | --- | | 1 | 0.4795018434524536 | 333.8126870860002 | — | — | | 2 | — | 83.20406460700008 | — | — | | 3 | 0.748327225446701 | 422.89352600200027 | — | — | | 4 | — | 80.6114583420001 | — | — | | 5 | 0.7239521145820618 | 258.5455324069999 | — | — | | 6 | — | 73.84239936600011 | — | — | | 7 | 0.16936538740992546 | 574.8180244990001 | — | — | | 8 | — | 96.54974007300007 | — | — | | 9 | 0.4399409890174866 | 565.6080998629991 | — | — | | 10 | — | 92.34179596000013 | — | — | | 11 | 0.8399062752723694 | 218.35304513600022 | — | — | | 12 | — | 79.30637174300045 | — | — | | 13 | 0.4414585630098979 | 501.12077764199876 | — | — | | 14 | — | 93.43431715799761 | — | — | | 15 | 0.42326606810092926 | 318.04057367399855 | — | — | | 16 | — | 80.69576919999872 | — | — | | 17 | 0.40716272592544556 | 360.7394638119995 | — | — | | 18 | — | 89.55805717299972 | — | — | | 19 | 0.6895849704742432 | 231.9451002750011 | — | — | | 20 | — | 83.27529721300107 | — | — | | 21 | 0.43824608623981476 | 495.37745529899803 | — | — | | 22 | — | 96.24851294900054 | — | — | | 23 | 0.9399955868721008 | 204.48686326700226 | — | — | | 24 | — | 83.22265478900044 | — | — | | 25 | 0.8305423855781555 | 519.9165100769987 | — | — | | 26 | — | 89.6818983230014 | — | — | | 27 | 0.8130848407745361 | 248.77518401499856 | — | — | | 28 | — | 79.8105911370003 | — | — | | 29 | 0.44154128432273865 | 555.8580370990003 | — | — | | 30 | — | 94.76554830300029 | — | — | | 31 | 1.046946406364441 | 263.7495171710025 | — | — | | 32 | — | 72.05432988400025 | — | — | | 33 | 0.8970373074213663 | 503.61257829799797 | — | — | | 34 | — | 84.46763328200177 | — | — | | 35 | 0.18153761327266693 | 653.0792657429993 | — | — | | 36 | — | 94.32943631900162 | — | — | | 37 | 0.8432159423828125 | 201.25111297200237 | — | — | | 38 | — | 82.232110277002 | — | — | | 39 | 0.8482327461242676 | 189.21507756300343 | — | — | | 40 | — | 61.378676403999634 | — | — | | 41 | 0.4937981963157654 | 400.15355072100465 | — | — | | 42 | — | 83.41378324700054 | — | — | | 43 | 0.8437418043613434 | 382.6277843600001 | — | — | | 44 | — | 78.79109064600016 | — | — | | 45 | 0.5415472984313965 | 269.876155730999 | — | — | | 46 | — | 87.70896454199828 | — | — | | 47 | 0.9461045265197754 | 216.35137124500216 | — | — | | 48 | — | 78.48169384499852 | — | — | | 49 | 0.942751407623291 | 209.28487726499588 | — | — | | 50 | — | 78.7290867219981 | — | — | | Family | Seed | Sample | Native | Efficiency | Reliability | Success | | --- | --- | --- | --- | --- | --- | --- | | two_pool_cycle | 900000 | 0 | 0.6169376468238315 | 0.33333333333333337 | 1.0 | True | | two_pool_cycle | 900000 | 1 | 0.6169388080321323 | 0.33333333333333337 | 1.0 | True | | two_pool_cycle | 900000 | 2 | 0.6169379690916243 | 0.33333333333333337 | 1.0 | True | | two_pool_cycle | 900000 | 3 | 0.6169384763792611 | 0.33333333333333337 | 1.0 | True | | two_pool_cycle | 900001 | 0 | 0.6172326900050679 | 0.33333333333333337 | 1.0 | True | | two_pool_cycle | 900001 | 1 | 0.6172314366859272 | 0.33333333333333337 | 1.0 | True | | two_pool_cycle | 900001 | 2 | 0.6172326385674898 | 0.33333333333333337 | 1.0 | True | | two_pool_cycle | 900001 | 3 | 0.6172317395034662 | 0.33333333333333337 | 1.0 | True | | triangular_cycle | 900000 | 0 | 0.6168411288846061 | 0.5 | 1.0 | True | | triangular_cycle | 900000 | 1 | 0.618344568706403 | 0.6666666666666667 | 1.0 | True | | triangular_cycle | 900000 | 2 | 0.61834576808361 | 0.6666666666666667 | 1.0 | True | | triangular_cycle | 900000 | 3 | 0.6183447103195866 | 0.6666666666666667 | 1.0 | True | | triangular_cycle | 900001 | 0 | 0.6174224465462672 | 0.5 | 1.0 | True | | triangular_cycle | 900001 | 1 | 0.6174236248934742 | 0.5 | 1.0 | True | | triangular_cycle | 900001 | 2 | 0.6183223973018531 | 0.6666666666666667 | 1.0 | True | | triangular_cycle | 900001 | 3 | 0.6183236836817555 | 0.6666666666666667 | 1.0 | True | | competing_cycles | 900000 | 0 | 0.25 | 0.0 | -1.0 | False | | competing_cycles | 900000 | 1 | 0.25 | 0.0 | -1.0 | False | | competing_cycles | 900000 | 2 | 0.25 | 0.0 | -1.0 | False | | competing_cycles | 900000 | 3 | 0.6212279479623682 | 0.5 | 1.0 | True | | competing_cycles | 900001 | 0 | 0.6189484372109143 | 0.33333333333333337 | 1.0 | True | | competing_cycles | 900001 | 1 | 0.6211425116085901 | 0.5 | 1.0 | True | | competing_cycles | 900001 | 2 | 0.6189508764907581 | 0.33333333333333337 | 1.0 | True | | competing_cycles | 900001 | 3 | 0.6189480592874964 | 0.33333333333333337 | 1.0 | True | | four_hop_cycle | 900000 | 0 | 0.6181305094102322 | 0.6666666666666667 | 1.0 | True | | four_hop_cycle | 900000 | 1 | 0.6173034664748611 | 0.5 | 1.0 | True | | four_hop_cycle | 900000 | 2 | 0.6173049009084548 | 0.5 | 1.0 | True | | four_hop_cycle | 900000 | 3 | 0.6181315855299198 | 0.6666666666666667 | 1.0 | True | | four_hop_cycle | 900001 | 0 | 0.6177471050538963 | 0.33333333333333337 | 1.0 | True | | four_hop_cycle | 900001 | 1 | 0.617748567686201 | 0.33333333333333337 | 1.0 | True | | four_hop_cycle | 900001 | 2 | 0.617746137146201 | 0.33333333333333337 | 1.0 | True | | four_hop_cycle | 900001 | 3 | 0.6177484115209666 | 0.33333333333333337 | 1.0 | True | | size_sensitive_cycle | 900000 | 0 | 0.175 | 0.0 | -1.0 | False | | size_sensitive_cycle | 900000 | 1 | 0.8332660773209832 | 0.33333333333333337 | 1.0 | True | | size_sensitive_cycle | 900000 | 2 | 0.175 | 0.0 | -1.0 | False | | size_sensitive_cycle | 900000 | 3 | 0.8332646993695965 | 0.33333333333333337 | 1.0 | True | | size_sensitive_cycle | 900001 | 0 | 0.175 | 0.0 | -1.0 | False | | size_sensitive_cycle | 900001 | 1 | 0.8317237538429428 | 0.33333333333333337 | 1.0 | True | | size_sensitive_cycle | 900001 | 2 | 0.831724450047142 | 0.33333333333333337 | 1.0 | True | | size_sensitive_cycle | 900001 | 3 | 0.8317246586644467 | 0.33333333333333337 | 1.0 | True | ## GDPO / no-think / 1789166523 Steps: 50/50. Holdout native: 0.6244599752567455. Successes: 36/40. | Optimizer step | Logged rollout reward | Step seconds | Peak allocated GiB | Peak reserved GiB | | --- | --- | --- | --- | --- | | 1 | 0.47949540615081787 | 333.16587591500013 | — | — | | 2 | — | 82.55144402600024 | — | — | | 3 | 0.6760792136192322 | 273.98975242299935 | — | — | | 4 | — | 80.4203468559997 | — | — | | 5 | 0.724294900894165 | 305.7092941430003 | — | — | | 6 | — | 74.60743086900015 | — | — | | 7 | 0.04999999701976776 | 258.437109813 | — | — | | 8 | — | 78.10814379600015 | — | — | | 9 | 0.5357341766357422 | 331.3484992100002 | — | — | | 10 | — | 91.88299751500108 | — | — | | 11 | 0.7235839366912842 | 282.3033622840003 | — | — | | 12 | — | 79.83501741999999 | — | — | | 13 | 0.4408377408981323 | 224.77536741099993 | — | — | | 14 | — | 92.27920960200072 | — | — | | 15 | 0.17396119236946106 | 247.53113513699964 | — | — | | 16 | — | 80.5280956029992 | — | — | | 17 | 0.5361157059669495 | 271.0126385839999 | — | — | | 18 | — | 91.25010780400044 | — | — | | 19 | 0.8132994771003723 | 242.5849810360005 | — | — | | 20 | — | 83.22634736600048 | — | — | | 21 | 0.4406861960887909 | 199.58957194699906 | — | — | | 22 | — | 89.83986751499924 | — | — | | 23 | 0.6862229704856873 | 298.13588110399996 | — | — | | 24 | — | 82.56295539700022 | — | — | | 25 | 0.8312231302261353 | 235.52427597500082 | — | — | | 26 | — | 88.50265450799907 | — | — | | 27 | 0.9402403235435486 | 213.40258446899952 | — | — | | 28 | — | 79.33389681799963 | — | — | | 29 | 0.4404658079147339 | 214.97263195699907 | — | — | | 30 | — | 91.84039551899878 | — | — | | 31 | 1.0470658540725708 | 189.37133719200028 | — | — | | 32 | — | 66.25698401200225 | — | — | | 33 | 0.8158747553825378 | 238.86322008099887 | — | — | | 34 | — | 83.6242890880003 | — | — | | 35 | 0.24660053849220276 | 325.1009769660004 | — | — | | 36 | — | 96.2894305789996 | — | — | | 37 | 0.8385937213897705 | 234.417130584 | — | — | | 38 | — | 84.90608804000021 | — | — | | 39 | 0.8497459888458252 | 171.43775163800092 | — | — | | 40 | — | 61.22994593600197 | — | — | | 41 | 0.643738865852356 | 299.5808600929995 | — | — | | 42 | — | 84.81286498499958 | — | — | | 43 | 0.8469367027282715 | 250.78653634400234 | — | — | | 44 | — | 68.71801683599915 | — | — | | 45 | 0.5450646877288818 | 231.37294147800094 | — | — | | 46 | — | 87.56447466299505 | — | — | | 47 | 0.947074294090271 | 232.82608543000242 | — | — | | 48 | — | 78.5874558070027 | — | — | | 49 | 0.9463198781013489 | 201.91460961000303 | — | — | | 50 | — | 71.03282021299674 | — | — | | Family | Seed | Sample | Native | Efficiency | Reliability | Success | | --- | --- | --- | --- | --- | --- | --- | | two_pool_cycle | 900000 | 0 | 0.6174455440146712 | 0.33333333333333337 | 1.0 | True | | two_pool_cycle | 900000 | 1 | 0.6174466939589095 | 0.33333333333333337 | 1.0 | True | | two_pool_cycle | 900000 | 2 | 0.6174463633467221 | 0.33333333333333337 | 1.0 | True | | two_pool_cycle | 900000 | 3 | 0.617446325597054 | 0.33333333333333337 | 1.0 | True | | two_pool_cycle | 900001 | 0 | 0.6178422403872945 | 0.33333333333333337 | 1.0 | True | | two_pool_cycle | 900001 | 1 | 0.6178429537705367 | 0.33333333333333337 | 1.0 | True | | two_pool_cycle | 900001 | 2 | 0.6178425815094428 | 0.33333333333333337 | 1.0 | True | | two_pool_cycle | 900001 | 3 | 0.6178426923321967 | 0.33333333333333337 | 1.0 | True | | triangular_cycle | 900000 | 0 | 0.6195805269881217 | 0.6666666666666667 | 1.0 | True | | triangular_cycle | 900000 | 1 | 0.6195813048573601 | 0.6666666666666667 | 1.0 | True | | triangular_cycle | 900000 | 2 | 0.6195814813085709 | 0.6666666666666667 | 1.0 | True | | triangular_cycle | 900000 | 3 | 0.6195818791400944 | 0.6666666666666667 | 1.0 | True | | triangular_cycle | 900001 | 0 | 0.6193224283775953 | 0.5 | 1.0 | True | | triangular_cycle | 900001 | 1 | 0.6193222054013648 | 0.5 | 1.0 | True | | triangular_cycle | 900001 | 2 | 0.6199607515405053 | 0.6666666666666667 | 1.0 | True | | triangular_cycle | 900001 | 3 | 0.6199609601615601 | 0.6666666666666667 | 1.0 | True | | competing_cycles | 900000 | 0 | 0.6183800407568019 | 0.5 | 1.0 | True | | competing_cycles | 900000 | 1 | 0.25 | 0.0 | -1.0 | False | | competing_cycles | 900000 | 2 | 0.25 | 0.0 | -1.0 | False | | competing_cycles | 900000 | 3 | 0.25 | 0.0 | -1.0 | False | | competing_cycles | 900001 | 0 | 0.6189366314748597 | 0.33333333333333337 | 1.0 | True | | competing_cycles | 900001 | 1 | 0.6212638547151527 | 0.5 | 1.0 | True | | competing_cycles | 900001 | 2 | 0.25 | 0.0 | -1.0 | False | | competing_cycles | 900001 | 3 | 0.6212623637910315 | 0.5 | 1.0 | True | | four_hop_cycle | 900000 | 0 | 0.618051336943904 | 0.6666666666666667 | 1.0 | True | | four_hop_cycle | 900000 | 1 | 0.6172147132400564 | 0.5 | 1.0 | True | | four_hop_cycle | 900000 | 2 | 0.6180514963968338 | 0.6666666666666667 | 1.0 | True | | four_hop_cycle | 900000 | 3 | 0.6180516145176738 | 0.6666666666666667 | 1.0 | True | | four_hop_cycle | 900001 | 0 | 0.6172641546083886 | 0.33333333333333337 | 1.0 | True | | four_hop_cycle | 900001 | 1 | 0.6172660292773925 | 0.33333333333333337 | 1.0 | True | | four_hop_cycle | 900001 | 2 | 0.6172664365064744 | 0.33333333333333337 | 1.0 | True | | four_hop_cycle | 900001 | 3 | 0.6172659797088377 | 0.33333333333333337 | 1.0 | True | | size_sensitive_cycle | 900000 | 0 | 0.8331320541859051 | 0.33333333333333337 | 1.0 | True | | size_sensitive_cycle | 900000 | 1 | 0.8331336674653191 | 0.33333333333333337 | 1.0 | True | | size_sensitive_cycle | 900000 | 2 | 0.8331322965015887 | 0.33333333333333337 | 1.0 | True | | size_sensitive_cycle | 900000 | 3 | 0.8339143527755926 | 0.5 | 1.0 | True | | size_sensitive_cycle | 900001 | 0 | 0.8314103908148568 | 0.33333333333333337 | 1.0 | True | | size_sensitive_cycle | 900001 | 1 | 0.8325324301229429 | 0.5 | 1.0 | True | | size_sensitive_cycle | 900001 | 2 | 0.8314105724246614 | 0.33333333333333337 | 1.0 | True | | size_sensitive_cycle | 900001 | 3 | 0.8314116613495444 | 0.33333333333333337 | 1.0 | True | ## CISPO / no-think / 1789178251 Steps: 50/50. Holdout native: 0.6190075072883925. Successes: 36/40. | Optimizer step | Logged rollout reward | Step seconds | Peak allocated GiB | Peak reserved GiB | | --- | --- | --- | --- | --- | | 1 | 0.4794864058494568 | 333.7465740409999 | — | — | | 2 | — | 82.6701556299995 | — | — | | 3 | 0.6249271035194397 | 288.856015051 | — | — | | 4 | — | 76.74044861300013 | — | — | | 5 | 0.7239433526992798 | 244.47076902000003 | — | — | | 6 | — | 74.86873130699996 | — | — | | 7 | 0.4221586585044861 | 289.23557070100014 | — | — | | 8 | — | 77.24971800500043 | — | — | | 9 | 0.4408476650714874 | 224.80640497600007 | — | — | | 10 | — | 92.2049963720001 | — | — | | 11 | 0.8311812877655029 | 246.77832825500036 | — | — | | 12 | — | 80.91245346300002 | — | — | | 13 | 0.44084227085113525 | 224.1840341080001 | — | — | | 14 | — | 91.49797025500084 | — | — | | 15 | 0.17429426312446594 | 202.12111372400022 | — | — | | 16 | — | 80.55344558900015 | — | — | | 17 | 0.5385847091674805 | 234.3388837819998 | — | — | | 18 | — | 89.91730763899932 | — | — | | 19 | 0.6886245608329773 | 249.91120985899988 | — | — | | 20 | — | 83.28241945400032 | — | — | | 21 | 0.3097813129425049 | 296.8628074249991 | — | — | | 22 | — | 91.50680776800073 | — | — | | 23 | 0.9405807852745056 | 200.02424894900014 | — | — | | 24 | — | 83.2709051560023 | — | — | | 25 | 0.7329174280166626 | 291.8929681670015 | — | — | | 26 | — | 87.24682994000068 | — | — | | 27 | 0.940902054309845 | 228.7263617339986 | — | — | | 28 | — | 80.99899548400208 | — | — | | 29 | 0.4406677484512329 | 213.71142901199983 | — | — | | 30 | — | 91.9580735519994 | — | — | | 31 | 1.0476813316345215 | 156.64428471800056 | — | — | | 32 | — | 65.24445897799887 | — | — | | 33 | 0.9373488426208496 | 263.94113257299887 | — | — | | 34 | — | 83.02269980999881 | — | — | | 35 | 0.04999999701976776 | 228.89222622999932 | — | — | | 36 | — | 78.71243763899838 | — | — | | 37 | 0.8402534127235413 | 216.00361499799965 | — | — | | 38 | — | 82.63343949400041 | — | — | | 39 | 0.8413939476013184 | 196.96659392399852 | — | — | | 40 | — | 68.90001421400211 | — | — | | 41 | 0.5438093543052673 | 286.34767421000106 | — | — | | 42 | — | 81.9797362069994 | — | — | | 43 | 0.8454612493515015 | 206.41304068699992 | — | — | | 44 | — | 70.62139491300059 | — | — | | 45 | 0.6358797550201416 | 327.8458718129996 | — | — | | 46 | — | 89.92750606700247 | — | — | | 47 | 0.8028456568717957 | 263.3829375919995 | — | — | | 48 | — | 78.81570005100366 | — | — | | 49 | 0.9409310817718506 | 222.13894051999705 | — | — | | 50 | — | 78.57328194000002 | — | — | | Family | Seed | Sample | Native | Efficiency | Reliability | Success | | --- | --- | --- | --- | --- | --- | --- | | two_pool_cycle | 900000 | 0 | 0.6173862425683626 | 0.33333333333333337 | 1.0 | True | | two_pool_cycle | 900000 | 1 | 0.6173873191268587 | 0.33333333333333337 | 1.0 | True | | two_pool_cycle | 900000 | 2 | 0.6173867850859016 | 0.33333333333333337 | 1.0 | True | | two_pool_cycle | 900000 | 3 | 0.6173872431842025 | 0.33333333333333337 | 1.0 | True | | two_pool_cycle | 900001 | 0 | 0.6177540296404975 | 0.33333333333333337 | 1.0 | True | | two_pool_cycle | 900001 | 1 | 0.6177549542583101 | 0.33333333333333337 | 1.0 | True | | two_pool_cycle | 900001 | 2 | 0.61775547305956 | 0.33333333333333337 | 1.0 | True | | two_pool_cycle | 900001 | 3 | 0.6177551720649312 | 0.33333333333333337 | 1.0 | True | | triangular_cycle | 900000 | 0 | 0.6188794449135906 | 0.5 | 1.0 | True | | triangular_cycle | 900000 | 1 | 0.6188811748917741 | 0.5 | 1.0 | True | | triangular_cycle | 900000 | 2 | 0.619441818157028 | 0.6666666666666667 | 1.0 | True | | triangular_cycle | 900000 | 3 | 0.6188807265349773 | 0.5 | 1.0 | True | | triangular_cycle | 900001 | 0 | 0.6193344927032203 | 0.5 | 1.0 | True | | triangular_cycle | 900001 | 1 | 0.6193346308406813 | 0.5 | 1.0 | True | | triangular_cycle | 900001 | 2 | 0.6193351808036109 | 0.5 | 1.0 | True | | triangular_cycle | 900001 | 3 | 0.6193356116082984 | 0.5 | 1.0 | True | | competing_cycles | 900000 | 0 | 0.25 | 0.0 | -1.0 | False | | competing_cycles | 900000 | 1 | 0.25 | 0.0 | -1.0 | False | | competing_cycles | 900000 | 2 | 0.25 | 0.0 | -1.0 | False | | competing_cycles | 900000 | 3 | 0.6208290054711377 | 0.5 | 1.0 | True | | competing_cycles | 900001 | 0 | 0.6188329288659339 | 0.33333333333333337 | 1.0 | True | | competing_cycles | 900001 | 1 | 0.621148871236344 | 0.5 | 1.0 | True | | competing_cycles | 900001 | 2 | 0.6211485251098207 | 0.5 | 1.0 | True | | competing_cycles | 900001 | 3 | 0.621151492362887 | 0.5 | 1.0 | True | | four_hop_cycle | 900000 | 0 | 0.6179127510059353 | 0.6666666666666667 | 1.0 | True | | four_hop_cycle | 900000 | 1 | 0.6172749174515798 | 0.5 | 1.0 | True | | four_hop_cycle | 900000 | 2 | 0.6172738687728689 | 0.5 | 1.0 | True | | four_hop_cycle | 900000 | 3 | 0.617274716620154 | 0.5 | 1.0 | True | | four_hop_cycle | 900001 | 0 | 0.6176202006734276 | 0.33333333333333337 | 1.0 | True | | four_hop_cycle | 900001 | 1 | 0.6176201558103611 | 0.33333333333333337 | 1.0 | True | | four_hop_cycle | 900001 | 2 | 0.6176187713000487 | 0.33333333333333337 | 1.0 | True | | four_hop_cycle | 900001 | 3 | 0.6176187478306736 | 0.33333333333333337 | 1.0 | True | | size_sensitive_cycle | 900000 | 0 | 0.832945101641491 | 0.33333333333333337 | 1.0 | True | | size_sensitive_cycle | 900000 | 1 | 0.8329464305758074 | 0.33333333333333337 | 1.0 | True | | size_sensitive_cycle | 900000 | 2 | 0.8329452279628777 | 0.33333333333333337 | 1.0 | True | | size_sensitive_cycle | 900000 | 3 | 0.833727046042741 | 0.5 | 1.0 | True | | size_sensitive_cycle | 900001 | 0 | 0.25 | 0.0 | -1.0 | False | | size_sensitive_cycle | 900001 | 1 | 0.8311179670606771 | 0.5 | 1.0 | True | | size_sensitive_cycle | 900001 | 2 | 0.8311177434017124 | 0.5 | 1.0 | True | | size_sensitive_cycle | 900001 | 3 | 0.8281855228974155 | 0.33333333333333337 | 1.0 | True | ## DAPO + ADAPO / no-think / 1789190592 Steps: 50/50. Holdout native: 0.6332133409563498. Successes: 37/40. | Optimizer step | Logged rollout reward | Step seconds | Peak allocated GiB | Peak reserved GiB | | --- | --- | --- | --- | --- | | 1 | 0.47944527864456177 | 335.65679702800014 | — | — | | 2 | — | 83.12946474800015 | — | — | | 3 | 0.6158038973808289 | 258.1928802289997 | — | — | | 4 | — | 73.78107018699984 | — | — | | 5 | 0.7251543998718262 | 267.63092621400006 | — | — | | 6 | — | 79.05084699599979 | — | — | | 7 | 0.29813387989997864 | 296.3191180170004 | — | — | | 8 | — | 83.63829929600024 | — | — | | 9 | 0.43826013803482056 | 313.30607508199955 | — | — | | 10 | — | 91.90846876399974 | — | — | | 11 | 0.8324222564697266 | 225.00398189199996 | — | — | | 12 | — | 80.40808471099945 | — | — | | 13 | 0.40813738107681274 | 286.83379497499936 | — | — | | 14 | — | 94.41846370099938 | — | — | | 15 | 0.49560555815696716 | 276.89079451900034 | — | — | | 16 | — | 81.47088516999929 | — | — | | 17 | 0.6016789078712463 | 321.19880535000084 | — | — | | 18 | — | 85.4535766920003 | — | — | | 19 | 0.8141046762466431 | 238.71769372400013 | — | — | | 20 | — | 83.59727125000063 | — | — | | 21 | 0.4383741617202759 | 277.9648794860004 | — | — | | 22 | — | 89.51071626000066 | — | — | | 23 | 0.9401151537895203 | 210.8280203840004 | — | — | | 24 | — | 83.5664421800002 | — | — | | 25 | 0.833543062210083 | 273.73329322100017 | — | — | | 26 | — | 87.7848287859988 | — | — | | 27 | 0.8479599356651306 | 239.18355853600042 | — | — | | 28 | — | 79.72936577699966 | — | — | | 29 | 0.44103488326072693 | 235.61832257200058 | — | — | | 30 | — | 92.13939696100078 | — | — | | 31 | 1.0488786697387695 | 233.9038784129989 | — | — | | 32 | — | 71.2344675839995 | — | — | | 33 | 0.9400039911270142 | 214.05489298000066 | — | — | | 34 | — | 83.12838843599911 | — | — | | 35 | 0.24802881479263306 | 268.3658710079999 | — | — | | 36 | — | 80.57577404099902 | — | — | | 37 | 0.8405643701553345 | 212.81755582700043 | — | — | | 38 | — | 82.24647135800114 | — | — | | 39 | 0.8448494672775269 | 189.7340894450008 | — | — | | 40 | — | 61.10121806999996 | — | — | | 41 | 0.6470715999603271 | 231.81965113800015 | — | — | | 42 | — | 83.21017399399989 | — | — | | 43 | 0.8479927778244019 | 193.35635953499786 | — | — | | 44 | — | 72.92475570000352 | — | — | | 45 | 0.3154837489128113 | 270.28836933099774 | — | — | | 46 | — | 79.52148914399913 | — | — | | 47 | 0.9494844675064087 | 231.68583597400357 | — | — | | 48 | — | 79.13202099600312 | — | — | | 49 | 0.8430052995681763 | 244.29561989499962 | — | — | | 50 | — | 79.96288616399761 | — | — | | Family | Seed | Sample | Native | Efficiency | Reliability | Success | | --- | --- | --- | --- | --- | --- | --- | | two_pool_cycle | 900000 | 0 | 0.6156027462878353 | 0.33333333333333337 | 1.0 | True | | two_pool_cycle | 900000 | 1 | 0.6156035667268978 | 0.33333333333333337 | 1.0 | True | | two_pool_cycle | 900000 | 2 | 0.6156035416276204 | 0.33333333333333337 | 1.0 | True | | two_pool_cycle | 900000 | 3 | 0.6164455951305502 | 0.5 | 1.0 | True | | two_pool_cycle | 900001 | 0 | 0.6176681372465913 | 0.33333333333333337 | 1.0 | True | | two_pool_cycle | 900001 | 1 | 0.6176674544183491 | 0.33333333333333337 | 1.0 | True | | two_pool_cycle | 900001 | 2 | 0.6176677963965522 | 0.33333333333333337 | 1.0 | True | | two_pool_cycle | 900001 | 3 | 0.6176667988512006 | 0.33333333333333337 | 1.0 | True | | triangular_cycle | 900000 | 0 | 0.618972723875739 | 0.5 | 1.0 | True | | triangular_cycle | 900000 | 1 | 0.6189743849813638 | 0.5 | 1.0 | True | | triangular_cycle | 900000 | 2 | 0.6195573225663248 | 0.6666666666666667 | 1.0 | True | | triangular_cycle | 900000 | 3 | 0.6189739411487076 | 0.5 | 1.0 | True | | triangular_cycle | 900001 | 0 | 0.6191941402051148 | 0.5 | 1.0 | True | | triangular_cycle | 900001 | 1 | 0.6191957859660329 | 0.5 | 1.0 | True | | triangular_cycle | 900001 | 2 | 0.6197781923240601 | 0.6666666666666667 | 1.0 | True | | triangular_cycle | 900001 | 3 | 0.6191951726623806 | 0.5 | 1.0 | True | | competing_cycles | 900000 | 0 | 0.25 | 0.0 | -1.0 | False | | competing_cycles | 900000 | 1 | 0.25 | 0.0 | -1.0 | False | | competing_cycles | 900000 | 2 | 0.25 | 0.0 | -1.0 | False | | competing_cycles | 900000 | 3 | 0.6176130096439503 | 0.33333333333333337 | 1.0 | True | | competing_cycles | 900001 | 0 | 0.6190718400809729 | 0.33333333333333337 | 1.0 | True | | competing_cycles | 900001 | 1 | 0.6210487380906995 | 0.5 | 1.0 | True | | competing_cycles | 900001 | 2 | 0.621050450989801 | 0.5 | 1.0 | True | | competing_cycles | 900001 | 3 | 0.6210504229613636 | 0.5 | 1.0 | True | | four_hop_cycle | 900000 | 0 | 0.6178382865214627 | 0.6666666666666667 | 1.0 | True | | four_hop_cycle | 900000 | 1 | 0.6171504838289823 | 0.5 | 1.0 | True | | four_hop_cycle | 900000 | 2 | 0.6178387261247049 | 0.6666666666666667 | 1.0 | True | | four_hop_cycle | 900000 | 3 | 0.6171508614294314 | 0.5 | 1.0 | True | | four_hop_cycle | 900001 | 0 | 0.6169854201482908 | 0.6666666666666667 | 1.0 | True | | four_hop_cycle | 900001 | 1 | 0.6161734524620605 | 0.5 | 1.0 | True | | four_hop_cycle | 900001 | 2 | 0.6161746136479783 | 0.5 | 1.0 | True | | four_hop_cycle | 900001 | 3 | 0.6152915263031933 | 0.33333333333333337 | 1.0 | True | | size_sensitive_cycle | 900000 | 0 | 0.8334000152269989 | 0.5 | 1.0 | True | | size_sensitive_cycle | 900000 | 1 | 0.8323421942212565 | 0.33333333333333337 | 1.0 | True | | size_sensitive_cycle | 900000 | 2 | 0.8323423715483269 | 0.33333333333333337 | 1.0 | True | | size_sensitive_cycle | 900000 | 3 | 0.8333995836483465 | 0.5 | 1.0 | True | | size_sensitive_cycle | 900001 | 0 | 0.8314558652540756 | 0.5 | 1.0 | True | | size_sensitive_cycle | 900001 | 1 | 0.8314559817276693 | 0.5 | 1.0 | True | | size_sensitive_cycle | 900001 | 2 | 0.8314556558584115 | 0.5 | 1.0 | True | | size_sensitive_cycle | 900001 | 3 | 0.8304768381206966 | 0.33333333333333337 | 1.0 | True | ## DAPO + REPO-R / no-think / 1789191744 Steps: 50/50. Holdout native: 0.5724401019920183. Successes: 31/40. | Optimizer step | Logged rollout reward | Step seconds | Peak allocated GiB | Peak reserved GiB | | --- | --- | --- | --- | --- | | 1 | 0.4793113172054291 | 339.4584226249999 | — | — | | 2 | — | 82.7074748819997 | — | — | | 3 | 0.7740623354911804 | 242.9536198360006 | — | — | | 4 | — | 73.88337641899943 | — | — | | 5 | 0.7232417464256287 | 246.90380199200013 | — | — | | 6 | — | 73.95877413000017 | — | — | | 7 | 0.29837942123413086 | 257.80528411099954 | — | — | | 8 | — | 78.26017540400039 | — | — | | 9 | 0.4381065368652344 | 330.50201018900043 | — | — | | 10 | — | 92.07023043800154 | — | — | | 11 | 0.8308002352714539 | 209.886425788 | — | — | | 12 | — | 80.60424775000092 | — | — | | 13 | 0.4407275319099426 | 221.18660950699996 | — | — | | 14 | — | 92.04747875399926 | — | — | | 15 | 0.17419666051864624 | 215.4094084999997 | — | — | | 16 | — | 80.51190537299908 | — | — | | 17 | 0.4404497742652893 | 198.84701016199915 | — | — | | 18 | — | 89.61900508899998 | — | — | | 19 | 0.8132044076919556 | 249.52503382100076 | — | — | | 20 | — | 83.31169498500003 | — | — | | 21 | 0.43981853127479553 | 232.0127544120005 | — | — | | 22 | — | 91.99613114599924 | — | — | | 23 | 0.9395684003829956 | 216.0728001890011 | — | — | | 24 | — | 83.3361953620024 | — | — | | 25 | 0.830710232257843 | 230.28595424500054 | — | — | | 26 | — | 87.43614393299958 | — | — | | 27 | 0.8123317360877991 | 235.536094047 | — | — | | 28 | — | 80.15440302699972 | — | — | | 29 | 0.44026288390159607 | 225.034533508001 | — | — | | 30 | — | 91.9762094639982 | — | — | | 31 | 1.046973466873169 | 169.8875314800007 | — | — | | 32 | — | 66.26559639399966 | — | — | | 33 | 0.9398280382156372 | 207.3118805610011 | — | — | | 34 | — | 83.72439446099997 | — | — | | 35 | 0.01874999701976776 | 222.9520893609988 | — | — | | 36 | — | 70.51143897200109 | — | — | | 37 | 0.735527515411377 | 243.5064796279994 | — | — | | 38 | — | 82.81517247699958 | — | — | | 39 | 0.7249026298522949 | 241.49042306200045 | — | — | | 40 | — | 70.82481263099999 | — | — | | 41 | 0.4414222240447998 | 181.7729249949989 | — | — | | 42 | — | 78.73975231299937 | — | — | | 43 | 0.8320287466049194 | 190.59470718299963 | — | — | | 44 | — | 78.65292478099946 | — | — | | 45 | 0.43970146775245667 | 232.68937776899838 | — | — | | 46 | — | 90.8428927979985 | — | — | | 47 | 0.938776969909668 | 197.25902485600272 | — | — | | 48 | — | 79.22551988899613 | — | — | | 49 | 0.9388787746429443 | 193.7512692649998 | — | — | | 50 | — | 79.27377721599987 | — | — | | Family | Seed | Sample | Native | Efficiency | Reliability | Success | | --- | --- | --- | --- | --- | --- | --- | | two_pool_cycle | 900000 | 0 | 0.6170665707246713 | 0.33333333333333337 | 1.0 | True | | two_pool_cycle | 900000 | 1 | 0.6170664895195931 | 0.33333333333333337 | 1.0 | True | | two_pool_cycle | 900000 | 2 | 0.6170663329743196 | 0.33333333333333337 | 1.0 | True | | two_pool_cycle | 900000 | 3 | 0.6170653436345541 | 0.33333333333333337 | 1.0 | True | | two_pool_cycle | 900001 | 0 | 0.6176353807103607 | 0.33333333333333337 | 1.0 | True | | two_pool_cycle | 900001 | 1 | 0.6176363454881537 | 0.33333333333333337 | 1.0 | True | | two_pool_cycle | 900001 | 2 | 0.617635481460556 | 0.33333333333333337 | 1.0 | True | | two_pool_cycle | 900001 | 3 | 0.6176363608829585 | 0.33333333333333337 | 1.0 | True | | triangular_cycle | 900000 | 0 | 0.6181042530675553 | 0.33333333333333337 | 1.0 | True | | triangular_cycle | 900000 | 1 | 0.6181059146631999 | 0.33333333333333337 | 1.0 | True | | triangular_cycle | 900000 | 2 | 0.61810598559111 | 0.33333333333333337 | 1.0 | True | | triangular_cycle | 900000 | 3 | 0.6181054220639031 | 0.33333333333333337 | 1.0 | True | | triangular_cycle | 900001 | 0 | 0.6188066805953297 | 0.33333333333333337 | 1.0 | True | | triangular_cycle | 900001 | 1 | 0.6188069888191968 | 0.33333333333333337 | 1.0 | True | | triangular_cycle | 900001 | 2 | 0.618805890198689 | 0.33333333333333337 | 1.0 | True | | triangular_cycle | 900001 | 3 | 0.6188056475190211 | 0.33333333333333337 | 1.0 | True | | competing_cycles | 900000 | 0 | 0.25 | 0.0 | -1.0 | False | | competing_cycles | 900000 | 1 | 0.25 | 0.0 | -1.0 | False | | competing_cycles | 900000 | 2 | 0.25 | 0.0 | -1.0 | False | | competing_cycles | 900000 | 3 | 0.25 | 0.0 | -1.0 | False | | competing_cycles | 900001 | 0 | 0.25 | 0.0 | -1.0 | False | | competing_cycles | 900001 | 1 | 0.25 | 0.0 | -1.0 | False | | competing_cycles | 900001 | 2 | 0.25 | 0.0 | -1.0 | False | | competing_cycles | 900001 | 3 | 0.25 | 0.0 | -1.0 | False | | four_hop_cycle | 900000 | 0 | 0.6170014806407986 | 0.33333333333333337 | 1.0 | True | | four_hop_cycle | 900000 | 1 | 0.6170025229249978 | 0.33333333333333337 | 1.0 | True | | four_hop_cycle | 900000 | 2 | 0.6170025763343728 | 0.33333333333333337 | 1.0 | True | | four_hop_cycle | 900000 | 3 | 0.6169999447661306 | 0.33333333333333337 | 1.0 | True | | four_hop_cycle | 900001 | 0 | 0.6170328837317088 | 0.33333333333333337 | 1.0 | True | | four_hop_cycle | 900001 | 1 | 0.617033260678955 | 0.33333333333333337 | 1.0 | True | | four_hop_cycle | 900001 | 2 | 0.6170340618311814 | 0.33333333333333337 | 1.0 | True | | four_hop_cycle | 900001 | 3 | 0.6170338136804978 | 0.33333333333333337 | 1.0 | True | | size_sensitive_cycle | 900000 | 0 | 0.25 | 0.0 | -1.0 | False | | size_sensitive_cycle | 900000 | 1 | 0.8326690123331902 | 0.33333333333333337 | 1.0 | True | | size_sensitive_cycle | 900000 | 2 | 0.8326679764228776 | 0.33333333333333337 | 1.0 | True | | size_sensitive_cycle | 900000 | 3 | 0.8326674563266472 | 0.33333333333333337 | 1.0 | True | | size_sensitive_cycle | 900001 | 0 | 0.831750250372806 | 0.33333333333333337 | 1.0 | True | | size_sensitive_cycle | 900001 | 1 | 0.8317517377728061 | 0.33333333333333337 | 1.0 | True | | size_sensitive_cycle | 900001 | 2 | 0.8317508980261654 | 0.33333333333333337 | 1.0 | True | | size_sensitive_cycle | 900001 | 3 | 0.8317511159244272 | 0.33333333333333337 | 1.0 | True | ## Starting model / thinking / 1789206045 Steps: 0/50. Holdout native: 0.6829987846910179. Successes: 37/40. | Optimizer step | Logged rollout reward | Step seconds | Peak allocated GiB | Peak reserved GiB | | --- | --- | --- | --- | --- | | Family | Seed | Sample | Native | Efficiency | Reliability | Success | | --- | --- | --- | --- | --- | --- | --- | | two_pool_cycle | 900000 | 0 | 0.8591328836248219 | 0.16666666666666663 | 1.0 | True | | two_pool_cycle | 900000 | 1 | 0.6089882858614486 | 0.6666666666666667 | 1.0 | True | | two_pool_cycle | 900000 | 2 | 0.6089882858614486 | 0.6666666666666667 | 1.0 | True | | two_pool_cycle | 900000 | 3 | 0.6089882858614486 | 0.6666666666666667 | 1.0 | True | | two_pool_cycle | 900001 | 0 | 0.6093825392865522 | 0.6666666666666667 | 1.0 | True | | two_pool_cycle | 900001 | 1 | 0.6093825392865522 | 0.6666666666666667 | 1.0 | True | | two_pool_cycle | 900001 | 2 | 0.8523284576243806 | 0.5 | 1.0 | True | | two_pool_cycle | 900001 | 3 | 0.8388160328689895 | 0.16666666666666663 | 1.0 | True | | triangular_cycle | 900000 | 0 | 0.8204653155141224 | 0.33333333333333337 | 1.0 | True | | triangular_cycle | 900000 | 1 | 0.6106601460847819 | 0.6666666666666667 | 1.0 | True | | triangular_cycle | 900000 | 2 | 0.6106601460847819 | 0.6666666666666667 | 1.0 | True | | triangular_cycle | 900000 | 3 | 0.6106601460847819 | 0.6666666666666667 | 1.0 | True | | triangular_cycle | 900001 | 0 | 0.844761535440721 | 0.0 | 1.0 | True | | triangular_cycle | 900001 | 1 | 0.9758783046350626 | 0.16666666666666663 | 1.0 | True | | triangular_cycle | 900001 | 2 | 0.25 | 0.0 | -1.0 | False | | triangular_cycle | 900001 | 3 | 0.6108851248150757 | 0.6666666666666667 | 1.0 | True | | competing_cycles | 900000 | 0 | 0.6093677238672119 | 0.16666666666666663 | 1.0 | True | | competing_cycles | 900000 | 1 | 0.6140552238672119 | 0.6666666666666667 | 1.0 | True | | competing_cycles | 900000 | 2 | 0.7248594440979103 | 0.33333333333333337 | 1.0 | True | | competing_cycles | 900000 | 3 | 0.0 | 0.0 | -1.0 | False | | competing_cycles | 900001 | 0 | 0.175 | 0.5 | 1.0 | False | | competing_cycles | 900001 | 1 | 0.7053628099527464 | 0.33333333333333337 | 1.0 | True | | competing_cycles | 900001 | 2 | 0.6512057439829095 | 0.5 | 1.0 | True | | competing_cycles | 900001 | 3 | 0.6145108982971254 | 0.5 | 1.0 | True | | four_hop_cycle | 900000 | 0 | 0.609376392538787 | 0.6666666666666667 | 1.0 | True | | four_hop_cycle | 900000 | 1 | 0.609376392538787 | 0.6666666666666667 | 1.0 | True | | four_hop_cycle | 900000 | 2 | 0.609376392538787 | 0.6666666666666667 | 1.0 | True | | four_hop_cycle | 900000 | 3 | 0.609376392538787 | 0.6666666666666667 | 1.0 | True | | four_hop_cycle | 900001 | 0 | 0.739978752793518 | 0.5 | 1.0 | True | | four_hop_cycle | 900001 | 1 | 0.8461963597959128 | 0.33333333333333337 | 1.0 | True | | four_hop_cycle | 900001 | 2 | 0.8461963597959128 | 0.33333333333333337 | 1.0 | True | | four_hop_cycle | 900001 | 3 | 0.6095701077161815 | 0.6666666666666667 | 1.0 | True | | size_sensitive_cycle | 900000 | 0 | 0.8262556477476042 | 0.6666666666666667 | 1.0 | True | | size_sensitive_cycle | 900000 | 1 | 0.9870115723152173 | 0.33333333333333337 | 1.0 | True | | size_sensitive_cycle | 900000 | 2 | 0.7421847681444665 | 0.5 | 1.0 | True | | size_sensitive_cycle | 900000 | 3 | 0.8262556477476042 | 0.6666666666666667 | 1.0 | True | | size_sensitive_cycle | 900001 | 0 | 0.8245803237223959 | 0.6666666666666667 | 1.0 | True | | size_sensitive_cycle | 900001 | 1 | 0.8245803237223959 | 0.6666666666666667 | 1.0 | True | | size_sensitive_cycle | 900001 | 2 | 0.9607157572618796 | 0.33333333333333337 | 1.0 | True | | size_sensitive_cycle | 900001 | 3 | 0.8245803237223959 | 0.33333333333333337 | 1.0 | True | ## GRPO / thinking / 1789214712 Steps: 50/50. Holdout native: 0.67982993828857. Successes: 38/40. | Optimizer step | Logged rollout reward | Step seconds | Peak allocated GiB | Peak reserved GiB | | --- | --- | --- | --- | --- | | 1 | 0.7080169916152954 | 1323.271556528 | 46.211081981658936 | 51.38671875 | | 2 | — | 156.92217275399935 | 46.689598083496094 | 51.484375 | | 3 | 0.9316171407699585 | 1171.6926818000006 | 46.689598083496094 | 51.51171875 | | 4 | — | 135.64331109700015 | 46.689598083496094 | 51.51171875 | | 5 | 0.8411365747451782 | 1640.3757725910004 | 48.77573108673096 | 54.79296875 | | 6 | — | 164.45583199400062 | 48.77573108673096 | 54.79296875 | | 7 | 0.7878009080886841 | 2296.4392685740004 | 60.44094133377075 | 66.74609375 | | 8 | — | 240.87562695600081 | 60.44094133377075 | 66.74609375 | | 9 | 0.7606194019317627 | 1397.8388155039956 | 60.44094133377075 | 66.74609375 | | 10 | — | 194.0120085890012 | 60.44094133377075 | 66.74609375 | | 11 | 0.8246521353721619 | 1640.2230279159976 | 60.44094133377075 | 66.74609375 | | 12 | — | 204.0847907480038 | 60.44094133377075 | 66.74609375 | | 13 | 0.7877841591835022 | 1913.5732425510032 | 66.42686176300049 | 73.69921875 | | 14 | — | 286.9272500799998 | 66.42686176300049 | 73.69921875 | | 15 | 0.7849491834640503 | 2235.668606463003 | 66.76106929779053 | 74.265625 | | 16 | — | 287.58674787699965 | 66.76106929779053 | 74.265625 | | 17 | 0.7741709351539612 | 2501.553329219005 | 66.76106929779053 | 74.265625 | | 18 | — | 266.0061489889995 | 66.76106929779053 | 74.265625 | | 19 | 0.5613870024681091 | 1148.603929220004 | 66.76106929779053 | 74.265625 | | 20 | — | 148.67559428100503 | 66.76106929779053 | 74.265625 | | 21 | 0.851814866065979 | 2313.628858852 | 66.76106929779053 | 74.265625 | | 22 | — | 242.45853826799794 | 66.76106929779053 | 74.265625 | | 23 | 0.9088314771652222 | 1439.073491322004 | 66.76106929779053 | 74.265625 | | 24 | — | 228.50225764999777 | 66.76106929779053 | 74.265625 | | 25 | 0.782331109046936 | 1346.7842622450007 | 66.76106929779053 | 74.265625 | | 26 | — | 189.0860957010009 | 66.76106929779053 | 74.265625 | | 27 | 1.0302821397781372 | 1773.342510265993 | 66.76106929779053 | 74.265625 | | 28 | — | 205.62199108300047 | 66.76106929779053 | 74.265625 | | 29 | 0.7447990775108337 | 2090.9175203540035 | 66.76106929779053 | 74.265625 | | 30 | — | 254.11358457000097 | 66.76106929779053 | 74.265625 | | 31 | 1.0804994106292725 | 1802.356435463993 | 66.76106929779053 | 74.265625 | | 32 | — | 211.18617562500003 | 66.76106929779053 | 74.265625 | | 33 | 0.9628927111625671 | 1280.3155918770062 | 66.76106929779053 | 74.265625 | | 34 | — | 199.357629935992 | 66.76106929779053 | 74.265625 | | 35 | 0.8928691148757935 | 2107.7353201419974 | 66.76106929779053 | 74.265625 | | 36 | — | 251.40516569999454 | 66.76106929779053 | 74.265625 | | 37 | 0.7713488340377808 | 1005.5948500160011 | 66.76106929779053 | 74.265625 | | 38 | — | 111.95662548500695 | 66.76106929779053 | 74.265625 | | 39 | 0.8101135492324829 | 1055.1161288180156 | 66.76106929779053 | 74.265625 | | 40 | — | 199.11226085701492 | 66.76106929779053 | 74.265625 | | 41 | 0.797109842300415 | 1965.8781912580016 | 69.07209730148315 | 76.8828125 | | 42 | — | 306.1554447550079 | 69.07209730148315 | 76.8828125 | | 43 | 0.7813174724578857 | 1369.1788975260133 | 69.07209730148315 | 76.8828125 | | 44 | — | 237.64380487699964 | 69.07209730148315 | 76.8828125 | | 45 | 0.8733965158462524 | 1434.0196957680018 | 69.07209730148315 | 76.8828125 | | 46 | — | 200.70870252800523 | 69.07209730148315 | 76.8828125 | | 47 | 0.9806088209152222 | 1442.9635367679948 | 69.07209730148315 | 76.8828125 | | 48 | — | 218.2718387430141 | 69.07209730148315 | 76.8828125 | | 49 | 0.873721182346344 | 1380.251024117999 | 69.07209730148315 | 76.8828125 | | 50 | — | 239.17380329099979 | 69.07209730148315 | 76.8828125 | | Family | Seed | Sample | Native | Efficiency | Reliability | Success | | --- | --- | --- | --- | --- | --- | --- | | two_pool_cycle | 900000 | 0 | 0.6401417781099434 | 0.5 | 1.0 | True | | two_pool_cycle | 900000 | 1 | 0.8638203836248219 | 0.33333333333333337 | 1.0 | True | | two_pool_cycle | 900000 | 2 | 0.6089882858614486 | 0.6666666666666667 | 1.0 | True | | two_pool_cycle | 900000 | 3 | 0.6089882858614486 | 0.6666666666666667 | 1.0 | True | | two_pool_cycle | 900001 | 0 | 0.8476409576243806 | 0.16666666666666663 | 1.0 | True | | two_pool_cycle | 900001 | 1 | 0.6093825392865522 | 0.6666666666666667 | 1.0 | True | | two_pool_cycle | 900001 | 2 | 0.6093825392865522 | 0.6666666666666667 | 1.0 | True | | two_pool_cycle | 900001 | 3 | 0.6093825392865522 | 0.6666666666666667 | 1.0 | True | | triangular_cycle | 900000 | 0 | 0.6106601460847819 | 0.6666666666666667 | 1.0 | True | | triangular_cycle | 900000 | 1 | 0.8522045135672316 | 0.5 | 1.0 | True | | triangular_cycle | 900000 | 2 | 0.6106601460847819 | 0.6666666666666667 | 1.0 | True | | triangular_cycle | 900000 | 3 | 0.6106601460847819 | 0.6666666666666667 | 1.0 | True | | triangular_cycle | 900001 | 0 | 0.6108851248150757 | 0.6666666666666667 | 1.0 | True | | triangular_cycle | 900001 | 1 | 0.6108851248150757 | 0.6666666666666667 | 1.0 | True | | triangular_cycle | 900001 | 2 | 0.7060268919310104 | 0.33333333333333337 | 1.0 | True | | triangular_cycle | 900001 | 3 | 0.9758783046350626 | 0.16666666666666663 | 1.0 | True | | competing_cycles | 900000 | 0 | 0.5924323165105192 | 0.16666666666666663 | 1.0 | True | | competing_cycles | 900000 | 1 | 0.6185941887823871 | 0.33333333333333337 | 1.0 | True | | competing_cycles | 900000 | 2 | 0.25 | 0.0 | -1.0 | False | | competing_cycles | 900000 | 3 | 0.578198213067076 | 0.0 | 1.0 | True | | competing_cycles | 900001 | 0 | 0.8752709263462409 | 0.33333333333333337 | 1.0 | True | | competing_cycles | 900001 | 1 | 0.6145108982971254 | 0.5 | 1.0 | True | | competing_cycles | 900001 | 2 | 0.6145108982971254 | 0.5 | 1.0 | True | | competing_cycles | 900001 | 3 | 0.6512057439829095 | 0.33333333333333337 | 1.0 | True | | four_hop_cycle | 900000 | 0 | 0.609376392538787 | 0.6666666666666667 | 1.0 | True | | four_hop_cycle | 900000 | 1 | 0.8444084478162579 | 0.5 | 1.0 | True | | four_hop_cycle | 900000 | 2 | 0.609376392538787 | 0.6666666666666667 | 1.0 | True | | four_hop_cycle | 900000 | 3 | 0.609376392538787 | 0.6666666666666667 | 1.0 | True | | four_hop_cycle | 900001 | 0 | 0.6095701077161815 | 0.5 | 1.0 | True | | four_hop_cycle | 900001 | 1 | 0.6095701077161815 | 0.6666666666666667 | 1.0 | True | | four_hop_cycle | 900001 | 2 | 0.8461963597959128 | 0.33333333333333337 | 1.0 | True | | four_hop_cycle | 900001 | 3 | 0.739978752793518 | 0.33333333333333337 | 1.0 | True | | size_sensitive_cycle | 900000 | 0 | 0.8262556477476042 | 0.5 | 1.0 | True | | size_sensitive_cycle | 900000 | 1 | 0.0 | 0.0 | -1.0 | False | | size_sensitive_cycle | 900000 | 2 | 0.8215681477476042 | 0.16666666666666663 | 1.0 | True | | size_sensitive_cycle | 900000 | 3 | 0.8262556477476042 | 0.6666666666666667 | 1.0 | True | | size_sensitive_cycle | 900001 | 0 | 0.8245803237223959 | 0.6666666666666667 | 1.0 | True | | size_sensitive_cycle | 900001 | 1 | 0.9872132714354994 | 0.5 | 1.0 | True | | size_sensitive_cycle | 900001 | 2 | 0.8245803237223959 | 0.6666666666666667 | 1.0 | True | | size_sensitive_cycle | 900001 | 3 | 0.8245803237223959 | 0.6666666666666667 | 1.0 | True | ## DAPO / thinking / 1789216886 Steps: 50/50. Holdout native: 0.670334256205925. Successes: 38/40. | Optimizer step | Logged rollout reward | Step seconds | Peak allocated GiB | Peak reserved GiB | | --- | --- | --- | --- | --- | | 1 | 0.7080169916152954 | 1292.0555840519999 | 46.211081981658936 | 51.38671875 | | 2 | — | 156.59421884200037 | 46.689598083496094 | 51.484375 | | 3 | 0.7398985624313354 | 924.362896693 | 46.689598083496094 | 51.494140625 | | 4 | — | 99.59936518700079 | 46.689598083496094 | 51.494140625 | | 5 | 0.7395387291908264 | 1064.131845978999 | 46.689598083496094 | 51.494140625 | | 6 | — | 147.21682394900108 | 46.689598083496094 | 51.494140625 | | 7 | 0.9056540131568909 | 1984.3922040430007 | 57.23006200790405 | 63.291015625 | | 8 | — | 218.58815286200024 | 57.23006200790405 | 63.291015625 | | 9 | 0.6712659001350403 | 2095.5628474430023 | 66.42421674728394 | 73.916015625 | | 10 | — | 284.6669807569997 | 66.42421674728394 | 73.916015625 | | 11 | 0.6725773811340332 | 1532.8431922100008 | 66.42421674728394 | 73.916015625 | | 12 | — | 222.87849550399733 | 66.42421674728394 | 73.916015625 | | 13 | 0.6437807083129883 | 1709.3426463319956 | 66.42421674728394 | 73.916015625 | | 14 | — | 244.61136676099886 | 66.42421674728394 | 73.916015625 | | 15 | 0.8411972522735596 | 1781.2970506860038 | 66.42421674728394 | 73.916015625 | | 16 | — | 222.34257325399994 | 66.42421674728394 | 73.916015625 | | 17 | 0.9005692005157471 | 1728.3003051019969 | 66.42421674728394 | 73.916015625 | | 18 | — | 256.19080749199566 | 66.42421674728394 | 73.916015625 | | 19 | 0.9035705924034119 | 1768.9756165739927 | 66.42421674728394 | 73.916015625 | | 20 | — | 181.6321102799957 | 66.42421674728394 | 73.916015625 | | 21 | 0.7668641805648804 | 1774.4828627140014 | 66.42421674728394 | 73.916015625 | | 22 | — | 206.42375343900858 | 66.42421674728394 | 73.916015625 | | 23 | 0.8813267350196838 | 1365.9782649090012 | 66.42421674728394 | 73.916015625 | | 24 | — | 187.97036160500284 | 66.42421674728394 | 73.916015625 | | 25 | 0.9646377563476562 | 1797.1556359970018 | 66.42421674728394 | 73.916015625 | | 26 | — | 261.40366287499273 | 66.42421674728394 | 73.916015625 | | 27 | 0.9216959476470947 | 1481.860717802003 | 66.42421674728394 | 73.916015625 | | 28 | — | 188.79422012899522 | 66.42421674728394 | 73.916015625 | | 29 | 0.9136285185813904 | 1548.6979675099974 | 66.42421674728394 | 73.916015625 | | 30 | — | 196.82815311599916 | 66.42421674728394 | 73.916015625 | | 31 | 0.7668814659118652 | 2286.557038924999 | 66.42421674728394 | 73.916015625 | | 32 | — | 251.90113309199296 | 66.42421674728394 | 73.916015625 | | 33 | 1.0501649379730225 | 1482.4170789929994 | 66.42421674728394 | 73.916015625 | | 34 | — | 227.86478610999984 | 66.42421674728394 | 73.916015625 | | 35 | 0.7830110192298889 | 2171.7106434480047 | 66.42421674728394 | 73.916015625 | | 36 | — | 273.15497789299116 | 66.42421674728394 | 73.916015625 | | 37 | 0.833756148815155 | 2057.967392379993 | 66.42421674728394 | 73.916015625 | | 38 | — | 263.7077892659945 | 66.42421674728394 | 73.916015625 | | 39 | 0.9829967617988586 | 1684.4528884049869 | 66.42421674728394 | 73.916015625 | | 40 | — | 199.50613806301408 | 66.42421674728394 | 73.916015625 | | 41 | 0.5012402534484863 | 1628.2348596299998 | 66.42421674728394 | 73.916015625 | | 42 | — | 183.32331458000408 | 66.42421674728394 | 73.916015625 | | 43 | 0.7716053128242493 | 1231.5425371530146 | 66.42421674728394 | 73.916015625 | | 44 | — | 172.76562454499071 | 66.42421674728394 | 73.916015625 | | 45 | 0.7913341522216797 | 2018.4625300120024 | 66.51273250579834 | 73.994140625 | | 46 | — | 283.93783065299067 | 66.51273250579834 | 73.994140625 | | 47 | 0.9820054769515991 | 1601.3260191109875 | 66.51273250579834 | 73.994140625 | | 48 | — | 194.61791170100332 | 66.51273250579834 | 73.994140625 | | 49 | 0.4527481198310852 | 1525.5017623179883 | 66.51273250579834 | 73.994140625 | | 50 | — | 211.343559276007 | 66.51273250579834 | 73.994140625 | | Family | Seed | Sample | Native | Efficiency | Reliability | Success | | --- | --- | --- | --- | --- | --- | --- | | two_pool_cycle | 900000 | 0 | 0.8638203836248219 | 0.33333333333333337 | 1.0 | True | | two_pool_cycle | 900000 | 1 | 0.6089882858614486 | 0.6666666666666667 | 1.0 | True | | two_pool_cycle | 900000 | 2 | 0.8638203836248219 | 0.33333333333333337 | 1.0 | True | | two_pool_cycle | 900000 | 3 | 0.8638203836248219 | 0.33333333333333337 | 1.0 | True | | two_pool_cycle | 900001 | 0 | 0.6093825392865522 | 0.6666666666666667 | 1.0 | True | | two_pool_cycle | 900001 | 1 | 0.6093825392865522 | 0.6666666666666667 | 1.0 | True | | two_pool_cycle | 900001 | 2 | 0.6093825392865522 | 0.6666666666666667 | 1.0 | True | | two_pool_cycle | 900001 | 3 | 0.6093825392865522 | 0.6666666666666667 | 1.0 | True | | triangular_cycle | 900000 | 0 | 0.6106601460847819 | 0.6666666666666667 | 1.0 | True | | triangular_cycle | 900000 | 1 | 0.8204653155141224 | 0.5 | 1.0 | True | | triangular_cycle | 900000 | 2 | 0.8204653155141224 | 0.5 | 1.0 | True | | triangular_cycle | 900000 | 3 | 0.0 | 0.0 | -1.0 | False | | triangular_cycle | 900001 | 0 | 0.6108851248150757 | 0.6666666666666667 | 1.0 | True | | triangular_cycle | 900001 | 1 | 0.6108851248150757 | 0.6666666666666667 | 1.0 | True | | triangular_cycle | 900001 | 2 | 0.9711908046350627 | 0.0 | 1.0 | True | | triangular_cycle | 900001 | 3 | 0.8541365354407209 | 0.33333333333333337 | 1.0 | True | | competing_cycles | 900000 | 0 | 0.6140552238672119 | 0.5 | 1.0 | True | | competing_cycles | 900000 | 1 | 0.0 | 0.0 | -1.0 | False | | competing_cycles | 900000 | 2 | 0.6140552238672119 | 0.5 | 1.0 | True | | competing_cycles | 900000 | 3 | 0.5798840889712712 | 0.33333333333333337 | 1.0 | True | | competing_cycles | 900001 | 0 | 0.6098233982971253 | 0.16666666666666663 | 1.0 | True | | competing_cycles | 900001 | 1 | 0.6512057439829095 | 0.33333333333333337 | 1.0 | True | | competing_cycles | 900001 | 2 | 0.7140378233531887 | 0.16666666666666663 | 1.0 | True | | competing_cycles | 900001 | 3 | 0.6145108982971254 | 0.5 | 1.0 | True | | four_hop_cycle | 900000 | 0 | 0.609376392538787 | 0.6666666666666667 | 1.0 | True | | four_hop_cycle | 900000 | 1 | 0.609376392538787 | 0.6666666666666667 | 1.0 | True | | four_hop_cycle | 900000 | 2 | 0.609376392538787 | 0.6666666666666667 | 1.0 | True | | four_hop_cycle | 900000 | 3 | 0.609376392538787 | 0.6666666666666667 | 1.0 | True | | four_hop_cycle | 900001 | 0 | 0.6095701077161815 | 0.6666666666666667 | 1.0 | True | | four_hop_cycle | 900001 | 1 | 0.6095701077161815 | 0.6666666666666667 | 1.0 | True | | four_hop_cycle | 900001 | 2 | 0.6095701077161815 | 0.6666666666666667 | 1.0 | True | | four_hop_cycle | 900001 | 3 | 0.6095701077161815 | 0.6666666666666667 | 1.0 | True | | size_sensitive_cycle | 900000 | 0 | 0.8262556477476042 | 0.33333333333333337 | 1.0 | True | | size_sensitive_cycle | 900000 | 1 | 0.8262556477476042 | 0.6666666666666667 | 1.0 | True | | size_sensitive_cycle | 900000 | 2 | 0.8262556477476042 | 0.6666666666666667 | 1.0 | True | | size_sensitive_cycle | 900000 | 3 | 0.8262556477476042 | 0.6666666666666667 | 1.0 | True | | size_sensitive_cycle | 900001 | 0 | 0.8245803237223959 | 0.5 | 1.0 | True | | size_sensitive_cycle | 900001 | 1 | 0.8245803237223959 | 0.6666666666666667 | 1.0 | True | | size_sensitive_cycle | 900001 | 2 | 0.8245803237223959 | 0.6666666666666667 | 1.0 | True | | size_sensitive_cycle | 900001 | 3 | 0.8245803237223959 | 0.6666666666666667 | 1.0 | True | ## DAPO-refill / thinking / 1789271070 Steps: 50/50. Holdout native: 0.7104741920786344. Successes: 38/40. | Optimizer step | Logged rollout reward | Step seconds | Peak allocated GiB | Peak reserved GiB | | --- | --- | --- | --- | --- | | 1 | 0.7080169916152954 | 1271.8538267409997 | 46.211081981658936 | 51.38671875 | | 2 | — | 156.81330739999976 | 46.689598083496094 | 51.484375 | | 3 | 0.6346210241317749 | 1046.5687405200001 | 48.180649280548096 | 53.22265625 | | 4 | — | 160.3931571160001 | 48.180649280548096 | 53.22265625 | | 5 | 0.9073012471199036 | 1451.4844390130015 | 48.180649280548096 | 53.22265625 | | 6 | — | 155.96452493000015 | 48.180649280548096 | 53.22265625 | | 7 | 0.9194703698158264 | 2205.289949291002 | 66.74047183990479 | 74.21875 | | 8 | — | 286.1488141740001 | 66.74047183990479 | 74.21875 | | 9 | 0.7671912908554077 | 2210.013791121002 | 66.74047183990479 | 74.21875 | | 10 | — | 283.73702334899826 | 66.74047183990479 | 74.21875 | | 11 | 0.8420000076293945 | 1592.7577740629968 | 66.74047183990479 | 74.21875 | | 12 | — | 161.014062000997 | 66.74047183990479 | 74.21875 | | 13 | 0.8281087875366211 | 2020.8756578060038 | 66.74047183990479 | 74.21875 | | 14 | — | 277.22948470600386 | 66.74047183990479 | 74.21875 | | 15 | 0.59874427318573 | 2361.7444208510024 | 66.74047183990479 | 74.21875 | | 16 | — | 261.20766560401535 | 66.74047183990479 | 74.21875 | | 17 | 0.7952899932861328 | 1937.928044410004 | 66.74047183990479 | 74.21875 | | 18 | — | 235.31410456199956 | 66.74047183990479 | 74.21875 | | 19 | 0.9780855178833008 | 1460.218944550994 | 66.74047183990479 | 74.240234375 | | 20 | — | 128.79446079900663 | 66.74047183990479 | 74.240234375 | | 21 | 0.8681157231330872 | 2067.7043030540044 | 66.74047183990479 | 74.240234375 | | 22 | — | 273.1928589199997 | 66.74047183990479 | 74.240234375 | | 23 | 0.9962680339813232 | 1290.6729723049975 | 66.74047183990479 | 74.240234375 | | 24 | — | 158.91810345100748 | 66.74047183990479 | 74.240234375 | | 25 | 0.6849347352981567 | 1867.9587041330087 | 66.74047183990479 | 74.240234375 | | 26 | — | 255.583496878 | 66.74047183990479 | 74.240234375 | | 27 | 0.8178707957267761 | 1640.3296975979974 | 66.74047183990479 | 74.240234375 | | 28 | — | 247.52918564000356 | 66.74047183990479 | 74.240234375 | | 29 | 0.7771891355514526 | 1969.6545821579966 | 66.74047183990479 | 74.240234375 | | 30 | — | 242.06835631800277 | 66.74047183990479 | 74.240234375 | | 31 | 0.8562765121459961 | 2172.747359213983 | 66.74047183990479 | 74.240234375 | | 32 | — | 253.1972244689896 | 66.74047183990479 | 74.240234375 | | 33 | 0.7706488370895386 | 1623.1493520590084 | 66.74047183990479 | 74.240234375 | | 34 | — | 174.2861271330039 | 66.74047183990479 | 74.240234375 | | 35 | 0.7421962022781372 | 2223.7072799999805 | 68.41224241256714 | 76.193359375 | | 36 | — | 299.5672293989919 | 68.41224241256714 | 76.193359375 | | 37 | 0.920289158821106 | 1281.9606787930024 | 68.41224241256714 | 76.193359375 | | 38 | — | 183.18254053301644 | 68.41224241256714 | 76.193359375 | | 39 | 0.970900297164917 | 1925.603753813979 | 68.41224241256714 | 76.193359375 | | 40 | — | 258.53294155299955 | 68.41224241256714 | 76.193359375 | | 41 | 0.8416227102279663 | 2245.0404615720036 | 68.46192026138306 | 76.251953125 | | 42 | — | 299.2247267749917 | 68.46192026138306 | 76.251953125 | | 43 | 0.9652596116065979 | 2151.2987192989967 | 68.46192026138306 | 76.251953125 | | 44 | — | 261.0669290939986 | 68.46192026138306 | 76.251953125 | | 45 | 0.7630787491798401 | 1828.9440834909983 | 68.46192026138306 | 76.251953125 | | 46 | — | 253.96728759899997 | 68.46192026138306 | 76.251953125 | | 47 | 0.7385047674179077 | 1298.146168375999 | 68.46192026138306 | 76.251953125 | | 48 | — | 169.10284090100322 | 68.46192026138306 | 76.251953125 | | 49 | 0.9950897693634033 | 1922.4227069150002 | 68.46192026138306 | 76.251953125 | | 50 | — | 273.44746495200525 | 68.46192026138306 | 76.251953125 | | Family | Seed | Sample | Native | Efficiency | Reliability | Success | | --- | --- | --- | --- | --- | --- | --- | | two_pool_cycle | 900000 | 0 | 0.6089882858614486 | 0.6666666666666667 | 1.0 | True | | two_pool_cycle | 900000 | 1 | 0.8404385754973732 | 0.6666666666666667 | 1.0 | True | | two_pool_cycle | 900000 | 2 | 0.8404385754973732 | 0.5 | 1.0 | True | | two_pool_cycle | 900000 | 3 | 0.6089882858614486 | 0.6666666666666667 | 1.0 | True | | two_pool_cycle | 900001 | 0 | 0.6093825392865522 | 0.6666666666666667 | 1.0 | True | | two_pool_cycle | 900001 | 1 | 0.6093825392865522 | 0.6666666666666667 | 1.0 | True | | two_pool_cycle | 900001 | 2 | 0.0 | 0.0 | -1.0 | False | | two_pool_cycle | 900001 | 3 | 0.6093825392865522 | 0.6666666666666667 | 1.0 | True | | triangular_cycle | 900000 | 0 | 0.8204653155141224 | 0.33333333333333337 | 1.0 | True | | triangular_cycle | 900000 | 1 | 0.6106601460847819 | 0.6666666666666667 | 1.0 | True | | triangular_cycle | 900000 | 2 | 0.8428295135672316 | 0.0 | 1.0 | True | | triangular_cycle | 900000 | 3 | 0.8522045135672316 | 0.5 | 1.0 | True | | triangular_cycle | 900001 | 0 | 0.7060268919310104 | 0.33333333333333337 | 1.0 | True | | triangular_cycle | 900001 | 1 | 0.811786874720879 | 0.5 | 1.0 | True | | triangular_cycle | 900001 | 2 | 0.807099374720879 | 0.16666666666666663 | 1.0 | True | | triangular_cycle | 900001 | 3 | 0.811786874720879 | 0.33333333333333337 | 1.0 | True | | competing_cycles | 900000 | 0 | 0.6185941887823871 | 0.33333333333333337 | 1.0 | True | | competing_cycles | 900000 | 1 | 0.6502053079500402 | 0.33333333333333337 | 1.0 | True | | competing_cycles | 900000 | 2 | 0.8720485688510347 | 0.5 | 1.0 | True | | competing_cycles | 900000 | 3 | 0.6140552238672119 | 0.6666666666666667 | 1.0 | True | | competing_cycles | 900001 | 0 | 0.7053628099527464 | 0.33333333333333337 | 1.0 | True | | competing_cycles | 900001 | 1 | 0.6959878099527463 | 0.0 | 1.0 | True | | competing_cycles | 900001 | 2 | 0.6051358982971253 | 0.0 | 1.0 | True | | competing_cycles | 900001 | 3 | 0.6053090928016737 | 0.16666666666666663 | 1.0 | True | | four_hop_cycle | 900000 | 0 | 0.7456673418171573 | 0.16666666666666663 | 1.0 | True | | four_hop_cycle | 900000 | 1 | 0.6412222518441313 | 0.5 | 1.0 | True | | four_hop_cycle | 900000 | 2 | 0.609376392538787 | 0.6666666666666667 | 1.0 | True | | four_hop_cycle | 900000 | 3 | 0.7503548418171574 | 0.5 | 1.0 | True | | four_hop_cycle | 900001 | 0 | 0.739978752793518 | 0.5 | 1.0 | True | | four_hop_cycle | 900001 | 1 | 0.6095701077161815 | 0.6666666666666667 | 1.0 | True | | four_hop_cycle | 900001 | 2 | 0.25 | 0.0 | -1.0 | False | | four_hop_cycle | 900001 | 3 | 0.6095701077161815 | 0.6666666666666667 | 1.0 | True | | size_sensitive_cycle | 900000 | 0 | 0.962057681977345 | 0.33333333333333337 | 1.0 | True | | size_sensitive_cycle | 900000 | 1 | 0.8262556477476042 | 0.6666666666666667 | 1.0 | True | | size_sensitive_cycle | 900000 | 2 | 0.8262556477476042 | 0.5 | 1.0 | True | | size_sensitive_cycle | 900000 | 3 | 0.8262556477476042 | 0.6666666666666667 | 1.0 | True | | size_sensitive_cycle | 900001 | 0 | 0.9513407572618797 | 0.0 | 1.0 | True | | size_sensitive_cycle | 900001 | 1 | 0.8245803237223959 | 0.6666666666666667 | 1.0 | True | | size_sensitive_cycle | 900001 | 2 | 0.9829802732091317 | 0.16666666666666663 | 1.0 | True | | size_sensitive_cycle | 900001 | 3 | 0.9069421616294167 | 0.5 | 1.0 | True | ## GDPO / thinking / 1789273322 Steps: 50/50. Holdout native: 0.6601590534180904. Successes: 35/40. | Optimizer step | Logged rollout reward | Step seconds | Peak allocated GiB | Peak reserved GiB | | --- | --- | --- | --- | --- | | 1 | 0.7080169916152954 | 1307.445846601 | 46.211081981658936 | 51.38671875 | | 2 | — | 157.2177628310003 | 46.689598083496094 | 51.484375 | | 3 | 0.8029865026473999 | 1627.6266247210006 | 53.68960475921631 | 61.484375 | | 4 | — | 196.10917739700062 | 53.68960475921631 | 61.484375 | | 5 | 0.8990209102630615 | 943.5877459760004 | 53.68960475921631 | 61.4921875 | | 6 | — | 97.99755129099958 | 53.68960475921631 | 61.4921875 | | 7 | 0.7588474750518799 | 1794.1383231250002 | 57.67082643508911 | 64.12890625 | | 8 | — | 223.11134523700093 | 57.67082643508911 | 64.12890625 | | 9 | 0.8736792802810669 | 2104.989630897003 | 63.65529108047485 | 70.6328125 | | 10 | — | 265.4383719210018 | 63.65529108047485 | 70.6328125 | | 11 | 0.6727674603462219 | 1713.9672958210012 | 63.65529108047485 | 70.6328125 | | 12 | — | 182.38993302600284 | 63.65529108047485 | 70.6328125 | | 13 | 0.7504480481147766 | 1381.0262770810004 | 63.65529108047485 | 70.6328125 | | 14 | — | 170.54123297400292 | 63.65529108047485 | 70.6328125 | | 15 | 1.0336999893188477 | 2225.683277173999 | 63.65529108047485 | 70.6328125 | | 16 | — | 237.47537800700047 | 63.65529108047485 | 70.6328125 | | 17 | 0.7887344360351562 | 1791.069795530997 | 63.65529108047485 | 70.6328125 | | 18 | — | 242.6830074339996 | 63.65529108047485 | 70.6328125 | | 19 | 0.9605903625488281 | 1663.0609616120018 | 63.65529108047485 | 70.6328125 | | 20 | — | 224.124938886991 | 63.65529108047485 | 70.6328125 | | 21 | 0.7566317319869995 | 1358.554902425003 | 63.65529108047485 | 70.6328125 | | 22 | — | 186.87984689799487 | 63.65529108047485 | 70.6328125 | | 23 | 0.9423688650131226 | 1482.741087434002 | 63.65529108047485 | 70.6328125 | | 24 | — | 169.07731957399665 | 63.65529108047485 | 70.6328125 | | 25 | 0.8984931111335754 | 1322.410297416005 | 63.65529108047485 | 70.6328125 | | 26 | — | 148.8309440800076 | 63.65529108047485 | 70.6328125 | | 27 | 0.60328608751297 | 1568.3047927020052 | 63.65529108047485 | 70.6328125 | | 28 | — | 251.54640228600329 | 63.65529108047485 | 70.6328125 | | 29 | 0.771971583366394 | 1826.776519818999 | 63.65529108047485 | 70.80859375 | | 30 | — | 264.89417385499473 | 63.65529108047485 | 70.80859375 | | 31 | 1.0023112297058105 | 1455.752558099004 | 63.65529108047485 | 70.80859375 | | 32 | — | 185.55397482400076 | 63.65529108047485 | 70.80859375 | | 33 | 0.8544063568115234 | 1260.1165995250049 | 63.65529108047485 | 70.80859375 | | 34 | — | 157.62631366799906 | 63.65529108047485 | 70.80859375 | | 35 | 0.7379844188690186 | 1710.421675318994 | 68.30920124053955 | 76.08203125 | | 36 | — | 300.1433946209945 | 68.30920124053955 | 76.08203125 | | 37 | 0.770604133605957 | 2042.283405759008 | 68.30920124053955 | 76.08203125 | | 38 | — | 208.07504388699454 | 68.30920124053955 | 76.08203125 | | 39 | 0.924275815486908 | 1623.65240167499 | 68.30920124053955 | 76.08203125 | | 40 | — | 252.00426232099562 | 68.30920124053955 | 76.08203125 | | 41 | 0.6834702491760254 | 2157.995117664992 | 68.30920124053955 | 76.08203125 | | 42 | — | 281.58093438699143 | 68.30920124053955 | 76.08203125 | | 43 | 0.6985449194908142 | 1488.8086183090054 | 68.30920124053955 | 76.08203125 | | 44 | — | 186.98488715499843 | 68.30920124053955 | 76.08203125 | | 45 | 0.7833971381187439 | 2151.2422908390145 | 68.30920124053955 | 76.08203125 | | 46 | — | 274.3883095810015 | 68.30920124053955 | 76.08203125 | | 47 | 0.851170539855957 | 1800.7287981380068 | 68.30920124053955 | 76.08203125 | | 48 | — | 199.12135581698385 | 68.30920124053955 | 76.08203125 | | 49 | 0.9473243355751038 | 1814.5493688019924 | 68.30920124053955 | 76.08203125 | | 50 | — | 214.24098803201196 | 68.30920124053955 | 76.08203125 | | Family | Seed | Sample | Native | Efficiency | Reliability | Success | | --- | --- | --- | --- | --- | --- | --- | | two_pool_cycle | 900000 | 0 | 0.8638203836248219 | 0.5 | 1.0 | True | | two_pool_cycle | 900000 | 1 | 0.9675266661987673 | 0.16666666666666663 | 1.0 | True | | two_pool_cycle | 900000 | 2 | 0.8638203836248219 | 0.5 | 1.0 | True | | two_pool_cycle | 900000 | 3 | 0.8404385754973732 | 0.5 | 1.0 | True | | two_pool_cycle | 900001 | 0 | 0.6093825392865522 | 0.6666666666666667 | 1.0 | True | | two_pool_cycle | 900001 | 1 | 0.6093825392865522 | 0.6666666666666667 | 1.0 | True | | two_pool_cycle | 900001 | 2 | 0.6093825392865522 | 0.6666666666666667 | 1.0 | True | | two_pool_cycle | 900001 | 3 | 0.175 | 0.5 | 1.0 | False | | triangular_cycle | 900000 | 0 | 0.6106601460847819 | 0.6666666666666667 | 1.0 | True | | triangular_cycle | 900000 | 1 | 0.6106601460847819 | 0.6666666666666667 | 1.0 | True | | triangular_cycle | 900000 | 2 | 0.6106601460847819 | 0.6666666666666667 | 1.0 | True | | triangular_cycle | 900000 | 3 | 0.6106601460847819 | 0.6666666666666667 | 1.0 | True | | triangular_cycle | 900001 | 0 | 0.6108851248150757 | 0.6666666666666667 | 1.0 | True | | triangular_cycle | 900001 | 1 | 0.8062415512788537 | 0.16666666666666663 | 1.0 | True | | triangular_cycle | 900001 | 2 | 0.7060268919310104 | 0.5 | 1.0 | True | | triangular_cycle | 900001 | 3 | 0.811786874720879 | 0.5 | 1.0 | True | | competing_cycles | 900000 | 0 | 0.6140552238672119 | 0.5 | 1.0 | True | | competing_cycles | 900000 | 1 | 0.6140552238672119 | 0.6666666666666667 | 1.0 | True | | competing_cycles | 900000 | 2 | 0.9879763779626005 | 0.33333333333333337 | 1.0 | True | | competing_cycles | 900000 | 3 | 0.578198213067076 | 0.0 | 1.0 | True | | competing_cycles | 900001 | 0 | 0.25 | 0.0 | -1.0 | False | | competing_cycles | 900001 | 1 | 0.6145108982971254 | 0.6666666666666667 | 1.0 | True | | competing_cycles | 900001 | 2 | 0.6512057439829095 | 0.33333333333333337 | 1.0 | True | | competing_cycles | 900001 | 3 | 0.6099965928016736 | 0.33333333333333337 | 1.0 | True | | four_hop_cycle | 900000 | 0 | 0.7927032521897649 | 0.0 | 1.0 | True | | four_hop_cycle | 900000 | 1 | 0.609376392538787 | 0.6666666666666667 | 1.0 | True | | four_hop_cycle | 900000 | 2 | 0.7503548418171574 | 0.5 | 1.0 | True | | four_hop_cycle | 900000 | 3 | 0.7005591947284964 | 0.33333333333333337 | 1.0 | True | | four_hop_cycle | 900001 | 0 | 0.739978752793518 | 0.5 | 1.0 | True | | four_hop_cycle | 900001 | 1 | 0.6095701077161815 | 0.6666666666666667 | 1.0 | True | | four_hop_cycle | 900001 | 2 | 0.25 | 0.0 | -1.0 | False | | four_hop_cycle | 900001 | 3 | 0.739978752793518 | 0.5 | 1.0 | True | | size_sensitive_cycle | 900000 | 0 | 0.8262556477476042 | 0.6666666666666667 | 1.0 | True | | size_sensitive_cycle | 900000 | 1 | 0.8262556477476042 | 0.5 | 1.0 | True | | size_sensitive_cycle | 900000 | 2 | 0.175 | 0.33333333333333337 | 1.0 | False | | size_sensitive_cycle | 900000 | 3 | 0.8262556477476042 | 0.6666666666666667 | 1.0 | True | | size_sensitive_cycle | 900001 | 0 | 0.8245803237223959 | 0.5 | 1.0 | True | | size_sensitive_cycle | 900001 | 1 | 0.25 | 0.0 | -1.0 | False | | size_sensitive_cycle | 900001 | 2 | 0.8245803237223959 | 0.6666666666666667 | 1.0 | True | | size_sensitive_cycle | 900001 | 3 | 0.8245803237223959 | 0.5 | 1.0 | True | ## CISPO / thinking / 1789312185 Steps: 50/50. Holdout native: 0.606788222286585. Successes: 33/40. | Optimizer step | Logged rollout reward | Step seconds | Peak allocated GiB | Peak reserved GiB | | --- | --- | --- | --- | --- | | 1 | 0.7080169916152954 | 1432.8779994019997 | 46.211081981658936 | 51.38671875 | | 2 | — | 156.6762321429992 | 46.689598083496094 | 51.484375 | | 3 | 0.9531776905059814 | 1272.7960069709998 | 46.689598083496094 | 51.484375 | | 4 | — | 144.7639433240006 | 46.689598083496094 | 51.484375 | | 5 | 0.8427101373672485 | 984.4018064339998 | 46.689598083496094 | 51.484375 | | 6 | — | 82.75198730199918 | 46.689598083496094 | 51.484375 | | 7 | 0.8339248299598694 | 2108.4935946019996 | 68.21795797348022 | 75.5078125 | | 8 | — | 298.3895189370005 | 68.21795797348022 | 75.5078125 | | 9 | 0.7160741686820984 | 1883.930043686999 | 68.21795797348022 | 75.5078125 | | 10 | — | 210.70282552500066 | 68.21795797348022 | 75.5078125 | | 11 | 0.7871460914611816 | 1801.6967272129987 | 68.21795797348022 | 75.5078125 | | 12 | — | 250.81987606999792 | 68.21795797348022 | 75.5078125 | | 13 | 0.7201732397079468 | 1510.479915453001 | 68.21795797348022 | 75.5078125 | | 14 | — | 185.10257531900243 | 68.21795797348022 | 75.5078125 | | 15 | 0.962762713432312 | 1604.1281318940019 | 68.21795797348022 | 75.5078125 | | 16 | — | 201.4823421400033 | 68.21795797348022 | 75.5078125 | | 17 | 0.6994915008544922 | 2233.6526770439923 | 68.21795797348022 | 75.5078125 | | 18 | — | 290.79190560100324 | 68.21795797348022 | 75.5078125 | | 19 | 0.8835256695747375 | 1453.557944958 | 68.21795797348022 | 75.5078125 | | 20 | — | 204.21084985199923 | 68.21795797348022 | 75.5078125 | | 21 | 0.7769557237625122 | 1420.192871165993 | 68.21795797348022 | 75.5078125 | | 22 | — | 204.32607309300874 | 68.21795797348022 | 75.5078125 | | 23 | 1.0138881206512451 | 1558.686015184001 | 68.21795797348022 | 75.5078125 | | 24 | — | 229.9780768470082 | 68.21795797348022 | 75.5078125 | | 25 | 0.5557511448860168 | 1374.3095642089938 | 68.21795797348022 | 75.5078125 | | 26 | — | 204.0851919899942 | 68.21795797348022 | 75.5078125 | | 27 | 0.7657335996627808 | 1509.8962310990028 | 68.21795797348022 | 75.5078125 | | 28 | — | 199.2965387799959 | 68.21795797348022 | 75.5078125 | | 29 | 0.7987229824066162 | 2059.5521601690016 | 68.21795797348022 | 75.5078125 | | 30 | — | 242.2114704769956 | 68.21795797348022 | 75.5078125 | | 31 | 1.0933129787445068 | 1127.0236637460002 | 68.21795797348022 | 75.5078125 | | 32 | — | 171.6825249080066 | 68.21795797348022 | 75.5078125 | | 33 | 0.8849163055419922 | 1447.7261638749987 | 68.21795797348022 | 75.5078125 | | 34 | — | 155.45952711000064 | 68.21795797348022 | 75.5078125 | | 35 | 0.8709275722503662 | 2078.5321845780054 | 70.00158309936523 | 77.8125 | | 36 | — | 312.09610687001987 | 70.00158309936523 | 77.8125 | | 37 | 0.9378395080566406 | 1434.1412884339952 | 70.00158309936523 | 77.8125 | | 38 | — | 172.46150075998594 | 70.00158309936523 | 77.8125 | | 39 | 0.9736660718917847 | 1806.3814684290046 | 70.00158309936523 | 77.8125 | | 40 | — | 199.77284794001025 | 70.00158309936523 | 77.8125 | | 41 | 0.8000410795211792 | 1882.2396674270058 | 70.00158309936523 | 77.8125 | | 42 | — | 239.25387795001006 | 70.00158309936523 | 77.8125 | | 43 | 0.8927896022796631 | 1635.5146522129944 | 70.00158309936523 | 77.8125 | | 44 | — | 235.15527386898611 | 70.00158309936523 | 77.8125 | | 45 | 0.9811986088752747 | 2391.487052959994 | 70.00158309936523 | 77.8125 | | 46 | — | 275.21273764799844 | 70.00158309936523 | 77.8125 | | 47 | 0.5666121244430542 | 1268.3584206089945 | 70.00158309936523 | 77.8125 | | 48 | — | 191.3939335570103 | 70.00158309936523 | 77.8125 | | 49 | 0.9495829343795776 | 923.0098794490114 | 70.00158309936523 | 77.822265625 | | 50 | — | 108.43975785301882 | 70.00158309936523 | 77.822265625 | | Family | Seed | Sample | Native | Efficiency | Reliability | Success | | --- | --- | --- | --- | --- | --- | --- | | two_pool_cycle | 900000 | 0 | 0.6089882858614486 | 0.6666666666666667 | 1.0 | True | | two_pool_cycle | 900000 | 1 | 0.6089882858614486 | 0.6666666666666667 | 1.0 | True | | two_pool_cycle | 900000 | 2 | 0.6089882858614486 | 0.6666666666666667 | 1.0 | True | | two_pool_cycle | 900000 | 3 | 0.0 | 0.0 | -1.0 | False | | two_pool_cycle | 900001 | 0 | 0.6093825392865522 | 0.6666666666666667 | 1.0 | True | | two_pool_cycle | 900001 | 1 | 0.6093825392865522 | 0.6666666666666667 | 1.0 | True | | two_pool_cycle | 900001 | 2 | 0.8523284576243806 | 0.5 | 1.0 | True | | two_pool_cycle | 900001 | 3 | 0.6093825392865522 | 0.6666666666666667 | 1.0 | True | | triangular_cycle | 900000 | 0 | 0.6106601460847819 | 0.6666666666666667 | 1.0 | True | | triangular_cycle | 900000 | 1 | 0.0 | 0.0 | -1.0 | False | | triangular_cycle | 900000 | 2 | 0.25 | 0.0 | -1.0 | False | | triangular_cycle | 900000 | 3 | 0.8204653155141224 | 0.5 | 1.0 | True | | triangular_cycle | 900001 | 0 | 0.6108851248150757 | 0.6666666666666667 | 1.0 | True | | triangular_cycle | 900001 | 1 | 0.7060268919310104 | 0.5 | 1.0 | True | | triangular_cycle | 900001 | 2 | 0.8541365354407209 | 0.33333333333333337 | 1.0 | True | | triangular_cycle | 900001 | 3 | 0.811786874720879 | 0.33333333333333337 | 1.0 | True | | competing_cycles | 900000 | 0 | 0.8720485688510347 | 0.33333333333333337 | 1.0 | True | | competing_cycles | 900000 | 1 | 0.8720485688510347 | 0.5 | 1.0 | True | | competing_cycles | 900000 | 2 | 0.6502053079500402 | 0.5 | 1.0 | True | | competing_cycles | 900000 | 3 | 0.5875732130670761 | 0.33333333333333337 | 1.0 | True | | competing_cycles | 900001 | 0 | 0.175 | 0.33333333333333337 | 1.0 | False | | competing_cycles | 900001 | 1 | 0.6145108982971254 | 0.6666666666666667 | 1.0 | True | | competing_cycles | 900001 | 2 | 0.6145108982971254 | 0.5 | 1.0 | True | | competing_cycles | 900001 | 3 | 0.7187253233531886 | 0.33333333333333337 | 1.0 | True | | four_hop_cycle | 900000 | 0 | 0.609376392538787 | 0.6666666666666667 | 1.0 | True | | four_hop_cycle | 900000 | 1 | 0.609376392538787 | 0.6666666666666667 | 1.0 | True | | four_hop_cycle | 900000 | 2 | 0.7503548418171574 | 0.5 | 1.0 | True | | four_hop_cycle | 900000 | 3 | 0.609376392538787 | 0.6666666666666667 | 1.0 | True | | four_hop_cycle | 900001 | 0 | 0.6095701077161815 | 0.5 | 1.0 | True | | four_hop_cycle | 900001 | 1 | 0.6095701077161815 | 0.6666666666666667 | 1.0 | True | | four_hop_cycle | 900001 | 2 | 0.6095701077161815 | 0.6666666666666667 | 1.0 | True | | four_hop_cycle | 900001 | 3 | 0.25 | 0.0 | -1.0 | False | | size_sensitive_cycle | 900000 | 0 | 0.962057681977345 | 0.5 | 1.0 | True | | size_sensitive_cycle | 900000 | 1 | 0.8262556477476042 | 0.6666666666666667 | 1.0 | True | | size_sensitive_cycle | 900000 | 2 | 0.25 | 0.0 | -1.0 | False | | size_sensitive_cycle | 900000 | 3 | 0.8262556477476042 | 0.6666666666666667 | 1.0 | True | | size_sensitive_cycle | 900001 | 0 | 0.8245803237223959 | 0.5 | 1.0 | True | | size_sensitive_cycle | 900001 | 1 | 0.0 | 0.0 | -1.0 | False | | size_sensitive_cycle | 900001 | 2 | 0.8245803237223959 | 0.6666666666666667 | 1.0 | True | | size_sensitive_cycle | 900001 | 3 | 0.8245803237223959 | 0.33333333333333337 | 1.0 | True | ## DAPO + ADAPO / thinking / 1789312258 Steps: 50/50. Holdout native: 0.62497194805629. Successes: 34/40. | Optimizer step | Logged rollout reward | Step seconds | Peak allocated GiB | Peak reserved GiB | | --- | --- | --- | --- | --- | | 1 | 0.7080169916152954 | 1387.7308625300004 | 46.211081981658936 | 51.38671875 | | 2 | — | 157.09996648500055 | 46.689598083496094 | 51.484375 | | 3 | 0.7660395503044128 | 1628.453406299 | 47.29379320144653 | 52.34375 | | 4 | — | 144.02420915199946 | 47.29379320144653 | 52.34375 | | 5 | 0.7347538471221924 | 1217.8728956439995 | 47.29379320144653 | 52.361328125 | | 6 | — | 122.98436585900163 | 47.29379320144653 | 52.361328125 | | 7 | 0.7208902835845947 | 1788.0677400999984 | 57.35714864730835 | 63.630859375 | | 8 | — | 239.61313570199945 | 57.35714864730835 | 63.630859375 | | 9 | 0.8042340278625488 | 2218.8322648449966 | 71.76706552505493 | 79.626953125 | | 10 | — | 326.57490212200173 | 71.76706552505493 | 79.626953125 | | 11 | 0.6859537959098816 | 2242.83903173 | 71.76706552505493 | 79.626953125 | | 12 | — | 265.50838019799994 | 71.76706552505493 | 79.626953125 | | 13 | 0.9356735348701477 | 1917.1351157499976 | 71.76706552505493 | 79.626953125 | | 14 | — | 245.83355621399642 | 71.76706552505493 | 79.626953125 | | 15 | 0.6324734687805176 | 1590.8704901529982 | 71.76706552505493 | 79.626953125 | | 16 | — | 241.17334455400123 | 71.76706552505493 | 79.626953125 | | 17 | 0.7066757678985596 | 1761.454811356005 | 71.76706552505493 | 79.626953125 | | 18 | — | 285.45596249799564 | 71.76706552505493 | 79.626953125 | | 19 | 0.7121108174324036 | 1590.5798178500045 | 71.76706552505493 | 79.626953125 | | 20 | — | 206.51005599998825 | 71.76706552505493 | 79.626953125 | | 21 | 0.7445224523544312 | 1799.4436993359996 | 71.76706552505493 | 79.626953125 | | 22 | — | 275.7976030539976 | 71.76706552505493 | 79.626953125 | | 23 | 0.8818289041519165 | 1214.9055495350003 | 71.76706552505493 | 79.626953125 | | 24 | — | 121.5651989240032 | 71.76706552505493 | 79.626953125 | | 25 | 0.7595635652542114 | 1024.8544964600005 | 71.76706552505493 | 79.626953125 | | 26 | — | 152.1805966360007 | 71.76706552505493 | 79.626953125 | | 27 | 1.0317509174346924 | 1710.4444454119985 | 71.76706552505493 | 79.626953125 | | 28 | — | 182.1155030810005 | 71.76706552505493 | 79.626953125 | | 29 | 0.8224276304244995 | 1681.9464596959988 | 71.76706552505493 | 79.626953125 | | 30 | — | 244.9255358420014 | 71.76706552505493 | 79.626953125 | | 31 | 1.090454339981079 | 1001.1734930269995 | 71.76706552505493 | 79.626953125 | | 32 | — | 132.5778156569977 | 71.76706552505493 | 79.626953125 | | 33 | 0.9851870536804199 | 1482.6869959809992 | 71.76706552505493 | 79.626953125 | | 34 | — | 174.89699893899888 | 71.76706552505493 | 79.626953125 | | 35 | 0.6279556751251221 | 1476.4721071010063 | 71.76706552505493 | 79.626953125 | | 36 | — | 198.67015630999958 | 71.76706552505493 | 79.626953125 | | 37 | 0.9181100726127625 | 1063.0183756520055 | 71.76706552505493 | 79.62890625 | | 38 | — | 113.67919772201276 | 71.76706552505493 | 79.62890625 | | 39 | 0.9714346528053284 | 1335.4767555669969 | 71.76706552505493 | 79.62890625 | | 40 | — | 194.4316224219947 | 71.76706552505493 | 79.62890625 | | 41 | 0.8027495741844177 | 1858.2825202210006 | 71.76706552505493 | 79.62890625 | | 42 | — | 250.3050878659924 | 71.76706552505493 | 79.62890625 | | 43 | 0.7570334672927856 | 975.4812457649969 | 71.76706552505493 | 79.62890625 | | 44 | — | 144.64556900199386 | 71.76706552505493 | 79.62890625 | | 45 | 0.9086361527442932 | 2023.3134895429903 | 71.76706552505493 | 79.62890625 | | 46 | — | 265.73495223200007 | 71.76706552505493 | 79.62890625 | | 47 | 0.7729398012161255 | 1709.7197918070233 | 71.76706552505493 | 79.62890625 | | 48 | — | 204.26634985800774 | 71.76706552505493 | 79.62890625 | | 49 | 0.9804160594940186 | 1591.6186869769954 | 71.76706552505493 | 79.62890625 | | 50 | — | 175.35021079199214 | 71.76706552505493 | 79.62890625 | | Family | Seed | Sample | Native | Efficiency | Reliability | Success | | --- | --- | --- | --- | --- | --- | --- | | two_pool_cycle | 900000 | 0 | 0.6089882858614486 | 0.6666666666666667 | 1.0 | True | | two_pool_cycle | 900000 | 1 | 0.6089882858614486 | 0.6666666666666667 | 1.0 | True | | two_pool_cycle | 900000 | 2 | 0.6089882858614486 | 0.6666666666666667 | 1.0 | True | | two_pool_cycle | 900000 | 3 | 0.6089882858614486 | 0.6666666666666667 | 1.0 | True | | two_pool_cycle | 900001 | 0 | 0.8523284576243806 | 0.5 | 1.0 | True | | two_pool_cycle | 900001 | 1 | 0.8435035328689895 | 0.5 | 1.0 | True | | two_pool_cycle | 900001 | 2 | 0.6093825392865522 | 0.6666666666666667 | 1.0 | True | | two_pool_cycle | 900001 | 3 | 0.6093825392865522 | 0.6666666666666667 | 1.0 | True | | triangular_cycle | 900000 | 0 | 0.6106601460847819 | 0.6666666666666667 | 1.0 | True | | triangular_cycle | 900000 | 1 | 0.6106601460847819 | 0.6666666666666667 | 1.0 | True | | triangular_cycle | 900000 | 2 | 0.8522045135672316 | 0.33333333333333337 | 1.0 | True | | triangular_cycle | 900000 | 3 | 0.6106601460847819 | 0.6666666666666667 | 1.0 | True | | triangular_cycle | 900001 | 0 | 0.175 | 0.5 | 1.0 | False | | triangular_cycle | 900001 | 1 | 0.6108851248150757 | 0.6666666666666667 | 1.0 | True | | triangular_cycle | 900001 | 2 | 0.25 | 0.0 | -1.0 | False | | triangular_cycle | 900001 | 3 | 0.811786874720879 | 0.5 | 1.0 | True | | competing_cycles | 900000 | 0 | 0.6140552238672119 | 0.5 | 1.0 | True | | competing_cycles | 900000 | 1 | 0.25 | 0.0 | -1.0 | False | | competing_cycles | 900000 | 2 | 0.25 | 0.0 | -1.0 | False | | competing_cycles | 900000 | 3 | 0.6185941887823871 | 0.5 | 1.0 | True | | competing_cycles | 900001 | 0 | 0.0 | 0.0 | -1.0 | False | | competing_cycles | 900001 | 1 | 0.6145108982971254 | 0.6666666666666667 | 1.0 | True | | competing_cycles | 900001 | 2 | 0.6512057439829095 | 0.5 | 1.0 | True | | competing_cycles | 900001 | 3 | 0.6145108982971254 | 0.5 | 1.0 | True | | four_hop_cycle | 900000 | 0 | 0.7005591947284964 | 0.33333333333333337 | 1.0 | True | | four_hop_cycle | 900000 | 1 | 0.609376392538787 | 0.6666666666666667 | 1.0 | True | | four_hop_cycle | 900000 | 2 | 0.8444084478162579 | 0.33333333333333337 | 1.0 | True | | four_hop_cycle | 900000 | 3 | 0.7456673418171573 | 0.16666666666666663 | 1.0 | True | | four_hop_cycle | 900001 | 0 | 0.7352912527935178 | 0.16666666666666663 | 1.0 | True | | four_hop_cycle | 900001 | 1 | 0.739978752793518 | 0.5 | 1.0 | True | | four_hop_cycle | 900001 | 2 | 0.6095701077161815 | 0.5 | 1.0 | True | | four_hop_cycle | 900001 | 3 | 0.739978752793518 | 0.33333333333333337 | 1.0 | True | | size_sensitive_cycle | 900000 | 0 | 0.8262556477476042 | 0.6666666666666667 | 1.0 | True | | size_sensitive_cycle | 900000 | 1 | 0.8262556477476042 | 0.6666666666666667 | 1.0 | True | | size_sensitive_cycle | 900000 | 2 | 0.8262556477476042 | 0.5 | 1.0 | True | | size_sensitive_cycle | 900000 | 3 | 0.8262556477476042 | 0.5 | 1.0 | True | | size_sensitive_cycle | 900001 | 0 | 0.0 | 0.0 | -1.0 | False | | size_sensitive_cycle | 900001 | 1 | 0.8245803237223959 | 0.6666666666666667 | 1.0 | True | | size_sensitive_cycle | 900001 | 2 | 0.8245803237223959 | 0.6666666666666667 | 1.0 | True | | size_sensitive_cycle | 900001 | 3 | 0.8245803237223959 | 0.33333333333333337 | 1.0 | True | ## DAPO + REPO-R / thinking / 1789312330 Steps: 50/50. Holdout native: 0.646087216010387. Successes: 34/40. | Optimizer step | Logged rollout reward | Step seconds | Peak allocated GiB | Peak reserved GiB | | --- | --- | --- | --- | --- | | 1 | 0.7080169916152954 | 1397.8954549800003 | 46.211081981658936 | 51.38671875 | | 2 | — | 157.18345037700055 | 46.689598083496094 | 51.484375 | | 3 | 0.8999981880187988 | 1548.7683420090002 | 50.21120643615723 | 55.546875 | | 4 | — | 160.3734592299993 | 50.21120643615723 | 55.546875 | | 5 | 0.8725726008415222 | 945.5713407839985 | 50.21120643615723 | 55.55859375 | | 6 | — | 105.7294935389973 | 50.21120643615723 | 55.55859375 | | 7 | 1.0114381313323975 | 2190.706770779001 | 63.816951751708984 | 70.79296875 | | 8 | — | 265.50813235699934 | 63.816951751708984 | 70.79296875 | | 9 | 0.8461290597915649 | 1576.6268515149977 | 63.816951751708984 | 70.79296875 | | 10 | — | 248.70580136699755 | 63.816951751708984 | 70.79296875 | | 11 | 0.7875280380249023 | 1366.433276806998 | 63.816951751708984 | 70.79296875 | | 12 | — | 157.17439058199852 | 63.816951751708984 | 70.79296875 | | 13 | 0.701381266117096 | 2238.979801022 | 63.816951751708984 | 70.79296875 | | 14 | — | 263.99378604999947 | 63.816951751708984 | 70.79296875 | | 15 | 0.876220703125 | 1532.063953914996 | 63.816951751708984 | 70.79296875 | | 16 | — | 237.40580801299984 | 63.816951751708984 | 70.79296875 | | 17 | 0.8561825752258301 | 1719.3857764660006 | 63.816951751708984 | 70.79296875 | | 18 | — | 246.0806536239943 | 63.816951751708984 | 70.79296875 | | 19 | 1.0371909141540527 | 1175.120890283004 | 63.816951751708984 | 70.79296875 | | 20 | — | 155.8805688670036 | 63.816951751708984 | 70.79296875 | | 21 | 0.9169240593910217 | 1894.4528153590036 | 63.816951751708984 | 70.79296875 | | 22 | — | 251.9174536130049 | 63.816951751708984 | 70.79296875 | | 23 | 0.5892045497894287 | 1628.1413750700012 | 63.816951751708984 | 70.79296875 | | 24 | — | 201.867253359007 | 63.816951751708984 | 70.79296875 | | 25 | 0.8015427589416504 | 1220.5425894130021 | 63.816951751708984 | 70.79296875 | | 26 | — | 216.85135657700084 | 63.816951751708984 | 70.79296875 | | 27 | 0.8972219228744507 | 1952.804152615994 | 63.816951751708984 | 70.79296875 | | 28 | — | 217.55828219300383 | 63.816951751708984 | 70.79296875 | | 29 | 0.703647255897522 | 1504.720339291005 | 63.816951751708984 | 70.79296875 | | 30 | — | 246.31147144499482 | 63.816951751708984 | 70.79296875 | | 31 | 0.9867122769355774 | 1519.2641041699972 | 63.816951751708984 | 70.79296875 | | 32 | — | 194.07235350299743 | 63.816951751708984 | 70.79296875 | | 33 | 0.987281322479248 | 1659.7578923590008 | 63.816951751708984 | 70.79296875 | | 34 | — | 222.44440870100516 | 63.816951751708984 | 70.79296875 | | 35 | 0.6315404176712036 | 1771.3755442619877 | 63.816951751708984 | 70.79296875 | | 36 | — | 219.30618613000115 | 63.816951751708984 | 70.79296875 | | 37 | 0.8155384063720703 | 1172.8255403209987 | 63.816951751708984 | 70.79296875 | | 38 | — | 170.8765211380014 | 63.816951751708984 | 70.79296875 | | 39 | 0.8250428438186646 | 1411.5272268420013 | 63.816951751708984 | 70.79296875 | | 40 | — | 182.75106583199522 | 63.816951751708984 | 70.79296875 | | 41 | 0.6421900987625122 | 1962.5831768580101 | 68.00909471511841 | 75.5390625 | | 42 | — | 298.94783146899135 | 68.00909471511841 | 75.5390625 | | 43 | 0.958878219127655 | 1519.0825120040026 | 68.00909471511841 | 75.5390625 | | 44 | — | 221.9907098029944 | 68.00909471511841 | 75.5390625 | | 45 | 0.8257235288619995 | 2097.171696471014 | 68.00909471511841 | 75.5390625 | | 46 | — | 260.7758615089988 | 68.00909471511841 | 75.5390625 | | 47 | 0.7217190265655518 | 1479.5910152529905 | 68.00909471511841 | 75.5390625 | | 48 | — | 194.92949151098583 | 68.00909471511841 | 75.5390625 | | 49 | 0.8118100166320801 | 2294.3973380980024 | 68.00909471511841 | 75.5390625 | | 50 | — | 242.5753946049881 | 68.00909471511841 | 75.5390625 | | Family | Seed | Sample | Native | Efficiency | Reliability | Success | | --- | --- | --- | --- | --- | --- | --- | | two_pool_cycle | 900000 | 0 | 0.8638203836248219 | 0.33333333333333337 | 1.0 | True | | two_pool_cycle | 900000 | 1 | 0.8638203836248219 | 0.5 | 1.0 | True | | two_pool_cycle | 900000 | 2 | 0.8404385754973732 | 0.5 | 1.0 | True | | two_pool_cycle | 900000 | 3 | 0.8591328836248219 | 0.16666666666666663 | 1.0 | True | | two_pool_cycle | 900001 | 0 | 0.6093825392865522 | 0.6666666666666667 | 1.0 | True | | two_pool_cycle | 900001 | 1 | 0.6093825392865522 | 0.6666666666666667 | 1.0 | True | | two_pool_cycle | 900001 | 2 | 0.6093825392865522 | 0.6666666666666667 | 1.0 | True | | two_pool_cycle | 900001 | 3 | 0.6093825392865522 | 0.6666666666666667 | 1.0 | True | | triangular_cycle | 900000 | 0 | 0.6106601460847819 | 0.6666666666666667 | 1.0 | True | | triangular_cycle | 900000 | 1 | 0.8204653155141224 | 0.33333333333333337 | 1.0 | True | | triangular_cycle | 900000 | 2 | 0.7049643139195981 | 0.33333333333333337 | 1.0 | True | | triangular_cycle | 900000 | 3 | 0.6106601460847819 | 0.6666666666666667 | 1.0 | True | | triangular_cycle | 900001 | 0 | 0.6108851248150757 | 0.6666666666666667 | 1.0 | True | | triangular_cycle | 900001 | 1 | 0.8541365354407209 | 0.5 | 1.0 | True | | triangular_cycle | 900001 | 2 | 0.6108851248150757 | 0.6666666666666667 | 1.0 | True | | triangular_cycle | 900001 | 3 | 0.811786874720879 | 0.5 | 1.0 | True | | competing_cycles | 900000 | 0 | 0.0 | 0.0 | -1.0 | False | | competing_cycles | 900000 | 1 | 0.25 | 0.0 | -1.0 | False | | competing_cycles | 900000 | 2 | 0.6455178079500401 | 0.16666666666666663 | 1.0 | True | | competing_cycles | 900000 | 3 | 0.7201719440979103 | 0.16666666666666663 | 1.0 | True | | competing_cycles | 900001 | 0 | 0.175 | 0.33333333333333337 | 1.0 | False | | competing_cycles | 900001 | 1 | 0.25 | 0.0 | -1.0 | False | | competing_cycles | 900001 | 2 | 0.7053628099527464 | 0.33333333333333337 | 1.0 | True | | competing_cycles | 900001 | 3 | 0.5871627452983137 | 0.5 | 1.0 | True | | four_hop_cycle | 900000 | 0 | 0.609376392538787 | 0.6666666666666667 | 1.0 | True | | four_hop_cycle | 900000 | 1 | 0.609376392538787 | 0.6666666666666667 | 1.0 | True | | four_hop_cycle | 900000 | 2 | 0.7503548418171574 | 0.5 | 1.0 | True | | four_hop_cycle | 900000 | 3 | 0.175 | 0.5 | 1.0 | False | | four_hop_cycle | 900001 | 0 | 0.6095701077161815 | 0.6666666666666667 | 1.0 | True | | four_hop_cycle | 900001 | 1 | 0.8461963597959128 | 0.33333333333333337 | 1.0 | True | | four_hop_cycle | 900001 | 2 | 0.6095701077161815 | 0.6666666666666667 | 1.0 | True | | four_hop_cycle | 900001 | 3 | 0.8461963597959128 | 0.5 | 1.0 | True | | size_sensitive_cycle | 900000 | 0 | 0.8262556477476042 | 0.33333333333333337 | 1.0 | True | | size_sensitive_cycle | 900000 | 1 | 0.8262556477476042 | 0.6666666666666667 | 1.0 | True | | size_sensitive_cycle | 900000 | 2 | 0.8262556477476042 | 0.6666666666666667 | 1.0 | True | | size_sensitive_cycle | 900000 | 3 | 0.8262556477476042 | 0.33333333333333337 | 1.0 | True | | size_sensitive_cycle | 900001 | 0 | 0.8245803237223959 | 0.6666666666666667 | 1.0 | True | | size_sensitive_cycle | 900001 | 1 | 0.8245803237223959 | 0.5 | 1.0 | True | | size_sensitive_cycle | 900001 | 2 | 0.25 | 0.0 | -1.0 | False | | size_sensitive_cycle | 900001 | 3 | 0.7512635678492581 | 0.5 | 1.0 | True | ## CISPO + REPO-R w5 / thinking / 1789375757 Steps: 50/50. Holdout native: 0.6243359613981289. Successes: 34/40. | Optimizer step | Logged rollout reward | Step seconds | Peak allocated GiB | Peak reserved GiB | | --- | --- | --- | --- | --- | | 1 | 0.7080169916152954 | 1298.5827015979999 | 46.211081981658936 | 51.38671875 | | 2 | — | 156.98477632500067 | 46.689598083496094 | 51.484375 | | 3 | 0.6858565807342529 | 1804.9482219919996 | 54.23228740692139 | 60.21484375 | | 4 | — | 184.09639192699933 | 54.23228740692139 | 60.21484375 | | 5 | 0.7405804395675659 | 896.6167426029979 | 54.23228740692139 | 60.21484375 | | 6 | — | 77.16678110800058 | 54.23228740692139 | 60.21484375 | | 7 | 0.7085568904876709 | 1950.7574458720019 | 61.0336332321167 | 67.94921875 | | 8 | — | 245.43896711900197 | 61.0336332321167 | 67.94921875 | | 9 | 0.7566396594047546 | 1570.2471904950007 | 61.0336332321167 | 67.94921875 | | 10 | — | 226.77468287900047 | 61.0336332321167 | 67.94921875 | | 11 | 0.8975619077682495 | 1191.7877642379972 | 61.0336332321167 | 67.96875 | | 12 | — | 124.46118573400054 | 61.0336332321167 | 67.96875 | | 13 | 0.8120169043540955 | 1760.5267777860026 | 62.163169384002686 | 68.69140625 | | 14 | — | 252.7138389949996 | 62.163169384002686 | 68.69140625 | | 15 | 0.8079956769943237 | 2033.3330594330037 | 62.163169384002686 | 68.69140625 | | 16 | — | 244.97779498899945 | 62.163169384002686 | 68.69140625 | | 17 | 0.84870445728302 | 1850.8885259510007 | 62.163169384002686 | 68.69140625 | | 18 | — | 215.79665486700105 | 62.163169384002686 | 68.69140625 | | 19 | 1.009394645690918 | 1785.1697965930034 | 63.00174951553345 | 69.9609375 | | 20 | — | 245.27781082600268 | 63.00174951553345 | 69.9609375 | | 21 | 0.9282667636871338 | 2104.7825938940005 | 63.00174951553345 | 69.9609375 | | 22 | — | 258.36378274600065 | 63.00174951553345 | 69.9609375 | | 23 | 1.006659984588623 | 1244.0834711949938 | 63.00174951553345 | 69.9609375 | | 24 | — | 152.53275930000018 | 63.00174951553345 | 69.9609375 | | 25 | 0.658179759979248 | 2135.4792062919987 | 63.00174951553345 | 69.9609375 | | 26 | — | 251.15780723399803 | 63.00174951553345 | 69.9609375 | | 27 | 0.7810078859329224 | 1637.9102483940005 | 63.00174951553345 | 69.9609375 | | 28 | — | 216.23619884599975 | 63.00174951553345 | 69.9609375 | | 29 | 0.7314809560775757 | 1882.0860976619952 | 63.00174951553345 | 69.9609375 | | 30 | — | 229.31584244400437 | 63.00174951553345 | 69.9609375 | | 31 | 0.989723265171051 | 1461.0723787209972 | 63.00174951553345 | 69.9609375 | | 32 | — | 148.00937268200505 | 63.00174951553345 | 69.9609375 | | 33 | 0.9474632143974304 | 1291.4547260530017 | 63.00174951553345 | 69.9609375 | | 34 | — | 121.99650523900709 | 63.00174951553345 | 69.9609375 | | 35 | 0.7480102777481079 | 2276.6506014410006 | 66.8723316192627 | 74.375 | | 36 | — | 290.36681058300746 | 66.8723316192627 | 74.375 | | 37 | 0.5515440702438354 | 1781.7185949529885 | 66.8723316192627 | 74.375 | | 38 | — | 204.5068822660105 | 66.8723316192627 | 74.375 | | 39 | 0.6855406761169434 | 1117.2130076190078 | 66.8723316192627 | 74.375 | | 40 | — | 113.81666151001264 | 66.8723316192627 | 74.375 | | 41 | 0.9058153033256531 | 1863.9089769140119 | 66.8723316192627 | 74.375 | | 42 | — | 231.138219371991 | 66.8723316192627 | 74.375 | | 43 | 0.9385105967521667 | 1633.5453268820056 | 66.8723316192627 | 74.375 | | 44 | — | 180.71860163799283 | 66.8723316192627 | 74.375 | | 45 | 0.7821725606918335 | 2046.1135910660087 | 66.8723316192627 | 74.375 | | 46 | — | 213.78865860099904 | 66.8723316192627 | 74.375 | | 47 | 0.8813004493713379 | 1836.2074433269954 | 66.8723316192627 | 74.375 | | 48 | — | 242.80719270800182 | 66.8723316192627 | 74.375 | | 49 | 0.9014878869056702 | 1400.860066737987 | 66.8723316192627 | 74.375 | | 50 | — | 144.83408938099456 | 66.8723316192627 | 74.375 | | Family | Seed | Sample | Native | Efficiency | Reliability | Success | | --- | --- | --- | --- | --- | --- | --- | | two_pool_cycle | 900000 | 0 | 0.8404385754973732 | 0.5 | 1.0 | True | | two_pool_cycle | 900000 | 1 | 0.6089882858614486 | 0.6666666666666667 | 1.0 | True | | two_pool_cycle | 900000 | 2 | 0.6983308425395819 | 0.33333333333333337 | 1.0 | True | | two_pool_cycle | 900000 | 3 | 0.6089882858614486 | 0.6666666666666667 | 1.0 | True | | two_pool_cycle | 900001 | 0 | 0.6093825392865522 | 0.6666666666666667 | 1.0 | True | | two_pool_cycle | 900001 | 1 | 0.6093825392865522 | 0.6666666666666667 | 1.0 | True | | two_pool_cycle | 900001 | 2 | 0.8435035328689895 | 0.5 | 1.0 | True | | two_pool_cycle | 900001 | 3 | 0.6999976570410972 | 0.5 | 1.0 | True | | triangular_cycle | 900000 | 0 | 0.6106601460847819 | 0.6666666666666667 | 1.0 | True | | triangular_cycle | 900000 | 1 | 0.7049643139195981 | 0.33333333333333337 | 1.0 | True | | triangular_cycle | 900000 | 2 | 0.6106601460847819 | 0.5 | 1.0 | True | | triangular_cycle | 900000 | 3 | 0.6106601460847819 | 0.6666666666666667 | 1.0 | True | | triangular_cycle | 900001 | 0 | 0.0 | 0.0 | -1.0 | False | | triangular_cycle | 900001 | 1 | 0.25 | 0.0 | -1.0 | False | | triangular_cycle | 900001 | 2 | 0.6108851248150757 | 0.6666666666666667 | 1.0 | True | | triangular_cycle | 900001 | 3 | 0.0 | 0.0 | -1.0 | False | | competing_cycles | 900000 | 0 | 0.5875732130670761 | 0.6666666666666667 | 1.0 | True | | competing_cycles | 900000 | 1 | 0.6140552238672119 | 0.6666666666666667 | 1.0 | True | | competing_cycles | 900000 | 2 | 0.8720485688510347 | 0.33333333333333337 | 1.0 | True | | competing_cycles | 900000 | 3 | 0.8720485688510347 | 0.33333333333333337 | 1.0 | True | | competing_cycles | 900001 | 0 | 0.6145108982971254 | 0.6666666666666667 | 1.0 | True | | competing_cycles | 900001 | 1 | 0.0 | 0.0 | -1.0 | False | | competing_cycles | 900001 | 2 | 0.6145108982971254 | 0.6666666666666667 | 1.0 | True | | competing_cycles | 900001 | 3 | 0.6145108982971254 | 0.5 | 1.0 | True | | four_hop_cycle | 900000 | 0 | 0.609376392538787 | 0.33333333333333337 | 1.0 | True | | four_hop_cycle | 900000 | 1 | 0.8397209478162578 | 0.16666666666666663 | 1.0 | True | | four_hop_cycle | 900000 | 2 | 0.609376392538787 | 0.6666666666666667 | 1.0 | True | | four_hop_cycle | 900000 | 3 | 0.7503548418171574 | 0.33333333333333337 | 1.0 | True | | four_hop_cycle | 900001 | 0 | 0.8461963597959128 | 0.33333333333333337 | 1.0 | True | | four_hop_cycle | 900001 | 1 | 0.8415088597959128 | 0.16666666666666663 | 1.0 | True | | four_hop_cycle | 900001 | 2 | 0.7352912527935178 | 0.16666666666666663 | 1.0 | True | | four_hop_cycle | 900001 | 3 | 0.6095701077161815 | 0.6666666666666667 | 1.0 | True | | size_sensitive_cycle | 900000 | 0 | 0.962057681977345 | 0.33333333333333337 | 1.0 | True | | size_sensitive_cycle | 900000 | 1 | 0.175 | 0.33333333333333337 | 1.0 | False | | size_sensitive_cycle | 900000 | 2 | 0.8262556477476042 | 0.6666666666666667 | 1.0 | True | | size_sensitive_cycle | 900000 | 3 | 0.8262556477476042 | 0.6666666666666667 | 1.0 | True | | size_sensitive_cycle | 900001 | 0 | 0.0 | 0.0 | -1.0 | False | | size_sensitive_cycle | 900001 | 1 | 0.8245803237223959 | 0.6666666666666667 | 1.0 | True | | size_sensitive_cycle | 900001 | 2 | 0.8245803237223959 | 0.5 | 1.0 | True | | size_sensitive_cycle | 900001 | 3 | 0.9872132714354994 | 0.5 | 1.0 | True | ## CISPO + REPO-R w5 + refill / thinking / 1789396300 Steps: 41/50. Holdout native: None. Successes: 0/0. | Optimizer step | Logged rollout reward | Step seconds | Peak allocated GiB | Peak reserved GiB | | --- | --- | --- | --- | --- | | 1 | 0.7080169916152954 | 1239.501060564 | 46.211081981658936 | 51.38671875 | | 2 | — | 156.63328702199988 | 46.689598083496094 | 51.484375 | | 3 | 0.750505805015564 | 1574.8926672090001 | 49.95496845245361 | 54.9609375 | | 4 | — | 157.89051116999917 | 49.95496845245361 | 54.9609375 | | 5 | 0.7497185468673706 | 1087.9027467079995 | 49.95496845245361 | 55.15625 | | 6 | — | 171.2098793110008 | 49.95496845245361 | 55.15625 | | 7 | 0.928241491317749 | 1743.7932904109994 | 57.167871952056885 | 63.22265625 | | 8 | — | 217.9013085759998 | 57.167871952056885 | 63.22265625 | | 9 | 0.5923604369163513 | 1398.187785918998 | 61.50002145767212 | 68.22265625 | | 10 | — | 246.8016796869979 | 61.50002145767212 | 68.22265625 | | 11 | 0.940544843673706 | 1629.4883709589976 | 61.50002145767212 | 68.22265625 | | 12 | — | 201.96123017800164 | 61.50002145767212 | 68.22265625 | | 13 | 0.8990108966827393 | 1707.6739688449998 | 61.50002145767212 | 68.22265625 | | 14 | — | 224.84784870800286 | 61.50002145767212 | 68.22265625 | | 15 | 0.6793212890625 | 1931.8897208800026 | 61.644176959991455 | 68.359375 | | 16 | — | 249.5143173660017 | 61.644176959991455 | 68.359375 | | 17 | 0.6942732334136963 | 2454.9009708850044 | 70.26109647750854 | 77.83203125 | | 18 | — | 312.1395660230046 | 70.26109647750854 | 77.83203125 | | 19 | 0.6048398613929749 | 1255.3704401659998 | 70.26109647750854 | 77.83203125 | | 20 | — | 134.67310873000315 | 70.26109647750854 | 77.83203125 | | 21 | 0.7587659955024719 | 1649.0433202029926 | 70.26109647750854 | 77.83203125 | | 22 | — | 213.08353274199908 | 70.26109647750854 | 77.83203125 | | 23 | 0.8505150079727173 | 1193.8290967089888 | 70.26109647750854 | 77.83203125 | | 24 | — | 184.58321064999473 | 70.26109647750854 | 77.83203125 | | 25 | 0.8463578224182129 | 1197.0888584370077 | 70.26109647750854 | 77.83203125 | | 26 | — | 162.0502540699963 | 70.26109647750854 | 77.83203125 | | 27 | 1.0031914710998535 | 1235.7846389290025 | 70.26109647750854 | 77.83203125 | | 28 | — | 189.59208594300435 | 70.26109647750854 | 77.83203125 | | 29 | 0.789441704750061 | 1960.6909793230006 | 70.26109647750854 | 77.83203125 | | 30 | — | 209.2750209980004 | 70.26109647750854 | 77.83203125 | | 31 | 0.9420388340950012 | 1385.4100905860032 | 70.26109647750854 | 77.83203125 | | 32 | — | 203.65134018299796 | 70.26109647750854 | 77.83203125 | | 33 | 0.9690639972686768 | 1414.5700504090055 | 70.26109647750854 | 77.83203125 | | 34 | — | 159.75958859299863 | 70.26109647750854 | 77.83203125 | | 35 | 0.7390574216842651 | 2039.2121586850008 | 70.26109647750854 | 77.83203125 | | 36 | — | 290.5989363379995 | 70.26109647750854 | 77.83203125 | | 37 | 0.90006023645401 | 2265.5598083139994 | 70.26109647750854 | 77.83203125 | | 38 | — | 168.06265335401258 | 70.26109647750854 | 77.83203125 | | 39 | 0.8132238388061523 | 1476.0498073280105 | 70.26109647750854 | 77.83203125 | | 40 | — | 206.3434993449846 | 70.26109647750854 | 77.83203125 | | 41 | 0.44877973198890686 | 1491.0382166269774 | 70.26109647750854 | 77.83203125 | | Family | Seed | Sample | Native | Efficiency | Reliability | Success | | --- | --- | --- | --- | --- | --- | --- |