{
  "best_global_step": null,
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 3.989247311827957,
  "eval_steps": 500,
  "global_step": 1300,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "completion_length": 26.546875,
      "epoch": 0.0030721966205837174,
      "grad_norm": 10.181303977966309,
      "kl": 0.0,
      "learning_rate": 1e-06,
      "loss": 0.6527,
      "reward": 0.4175647348165512,
      "reward_std": 0.12526048347353935,
      "reward_std_all": 0.2330886423587799,
      "rewards-max/forensics_sinkhorn_reward": 0.8773394525051117,
      "rewards-max/forensics_strict_edge_reward": 0.04458222679386381,
      "rewards-max/forensics_tiou_reward": 0.7230861485004425,
      "rewards/forensics_sinkhorn_reward": 0.30053071677684784,
      "rewards/forensics_strict_edge_reward": 0.00185797936705967,
      "rewards/forensics_tiou_reward": 0.05685761943459511,
      "step": 1
    },
    {
      "completion_length": 25.234375,
      "epoch": 0.006144393241167435,
      "grad_norm": 7.063746452331543,
      "kl": 0.0002772808074951172,
      "learning_rate": 9.992307692307693e-07,
      "loss": 0.6002,
      "reward": 0.45084571838378906,
      "reward_std": 0.1289445050060749,
      "reward_std_all": 0.23008975386619568,
      "rewards-max/forensics_sinkhorn_reward": 0.8444178700447083,
      "rewards-max/forensics_strict_edge_reward": 5.6263077567564324e-05,
      "rewards-max/forensics_tiou_reward": 0.6500846743583679,
      "rewards/forensics_sinkhorn_reward": 0.3190772384405136,
      "rewards/forensics_strict_edge_reward": 2.3447844910151616e-06,
      "rewards/forensics_tiou_reward": 0.0764623861759901,
      "step": 2
    },
    {
      "completion_length": 26.078125,
      "epoch": 0.009216589861751152,
      "grad_norm": 10.094746589660645,
      "kl": 0.001651763916015625,
      "learning_rate": 9.984615384615384e-07,
      "loss": 0.5623,
      "reward": 0.4529549479484558,
      "reward_std": 0.17076516151428223,
      "reward_std_all": 0.2456374391913414,
      "rewards-max/forensics_sinkhorn_reward": 0.7467035055160522,
      "rewards-max/forensics_strict_edge_reward": 0.02429219105397351,
      "rewards-max/forensics_tiou_reward": 0.4745758920907974,
      "rewards/forensics_sinkhorn_reward": 0.3113248199224472,
      "rewards/forensics_strict_edge_reward": 0.0021508840354727,
      "rewards/forensics_tiou_reward": 0.08370377868413925,
      "step": 3
    },
    {
      "completion_length": 25.421875,
      "epoch": 0.01228878648233487,
      "grad_norm": 6.371845245361328,
      "kl": 0.00335693359375,
      "learning_rate": 9.976923076923077e-07,
      "loss": 0.5469,
      "reward": 0.5212883651256561,
      "reward_std": 0.1719500795006752,
      "reward_std_all": 0.21727631241083145,
      "rewards-max/forensics_sinkhorn_reward": 0.8983431160449982,
      "rewards-max/forensics_strict_edge_reward": 0.009770949371159077,
      "rewards-max/forensics_tiou_reward": 0.7666391432285309,
      "rewards/forensics_sinkhorn_reward": 0.39878085255622864,
      "rewards/forensics_strict_edge_reward": 0.0005705242510885,
      "rewards/forensics_tiou_reward": 0.13626059144735336,
      "step": 4
    },
    {
      "completion_length": 21.0625,
      "epoch": 0.015360983102918587,
      "grad_norm": 6.704594135284424,
      "kl": 0.010650634765625,
      "learning_rate": 9.969230769230768e-07,
      "loss": 0.3954,
      "reward": 0.5023650974035263,
      "reward_std": 0.14931586384773254,
      "reward_std_all": 0.2190636470913887,
      "rewards-max/forensics_sinkhorn_reward": 0.8877105116844177,
      "rewards-max/forensics_strict_edge_reward": 0.00625885453337105,
      "rewards-max/forensics_tiou_reward": 0.7388273179531097,
      "rewards/forensics_sinkhorn_reward": 0.38746675848960876,
      "rewards/forensics_strict_edge_reward": 0.00047593131193934823,
      "rewards/forensics_tiou_reward": 0.1161554604768753,
      "step": 5
    },
    {
      "completion_length": 24.359375,
      "epoch": 0.018433179723502304,
      "grad_norm": 9.140335083007812,
      "kl": 0.0206298828125,
      "learning_rate": 9.961538461538461e-07,
      "loss": 0.7206,
      "reward": 0.48833276331424713,
      "reward_std": 0.11277982965111732,
      "reward_std_all": 0.19997645914554596,
      "rewards-max/forensics_sinkhorn_reward": 0.8953557014465332,
      "rewards-max/forensics_strict_edge_reward": 0.013407269863819238,
      "rewards-max/forensics_tiou_reward": 0.7592588067054749,
      "rewards/forensics_sinkhorn_reward": 0.36255085468292236,
      "rewards/forensics_strict_edge_reward": 0.0007411377191601787,
      "rewards/forensics_tiou_reward": 0.11288484185934067,
      "step": 6
    },
    {
      "completion_length": 23.640625,
      "epoch": 0.021505376344086023,
      "grad_norm": 7.054891586303711,
      "kl": 0.02447509765625,
      "learning_rate": 9.953846153846154e-07,
      "loss": 0.6783,
      "reward": 0.44893379509449005,
      "reward_std": 0.07625889033079147,
      "reward_std_all": 0.20036489516496658,
      "rewards-max/forensics_sinkhorn_reward": 0.6171925514936447,
      "rewards-max/forensics_strict_edge_reward": 0.04511191777068291,
      "rewards-max/forensics_tiou_reward": 0.3841990604996681,
      "rewards/forensics_sinkhorn_reward": 0.30690382421016693,
      "rewards/forensics_strict_edge_reward": 0.0020620080947182373,
      "rewards/forensics_tiou_reward": 0.06819903291761875,
      "step": 7
    },
    {
      "completion_length": 23.78125,
      "epoch": 0.02457757296466974,
      "grad_norm": 7.491725444793701,
      "kl": 0.0416259765625,
      "learning_rate": 9.946153846153846e-07,
      "loss": 0.4335,
      "reward": 0.48561161756515503,
      "reward_std": 0.1650330126285553,
      "reward_std_all": 0.22568687051534653,
      "rewards-max/forensics_sinkhorn_reward": 0.8380758762359619,
      "rewards-max/forensics_strict_edge_reward": 0.006157445033750264,
      "rewards-max/forensics_tiou_reward": 0.6283012479543686,
      "rewards/forensics_sinkhorn_reward": 0.362634539604187,
      "rewards/forensics_strict_edge_reward": 0.00031603009074387955,
      "rewards/forensics_tiou_reward": 0.1046195924282074,
      "step": 8
    },
    {
      "completion_length": 24.34375,
      "epoch": 0.027649769585253458,
      "grad_norm": 8.823945045471191,
      "kl": 0.05322265625,
      "learning_rate": 9.938461538461539e-07,
      "loss": 0.794,
      "reward": 0.526919037103653,
      "reward_std": 0.11319463327527046,
      "reward_std_all": 0.18281669169664383,
      "rewards-max/forensics_sinkhorn_reward": 0.791563481092453,
      "rewards-max/forensics_strict_edge_reward": 0.0034101101009582635,
      "rewards-max/forensics_tiou_reward": 0.6242441833019257,
      "rewards/forensics_sinkhorn_reward": 0.3832026273012161,
      "rewards/forensics_strict_edge_reward": 0.00019610764957178617,
      "rewards/forensics_tiou_reward": 0.1147181298583746,
      "step": 9
    },
    {
      "completion_length": 23.78125,
      "epoch": 0.030721966205837174,
      "grad_norm": 7.129749774932861,
      "kl": 0.121826171875,
      "learning_rate": 9.93076923076923e-07,
      "loss": 0.5442,
      "reward": 0.5444976687431335,
      "reward_std": 0.11373293772339821,
      "reward_std_all": 0.17797772586345673,
      "rewards-max/forensics_sinkhorn_reward": 0.9591047763824463,
      "rewards-max/forensics_strict_edge_reward": 0.4961976185441017,
      "rewards-max/forensics_tiou_reward": 0.9274976551532745,
      "rewards/forensics_sinkhorn_reward": 0.417828232049942,
      "rewards/forensics_strict_edge_reward": 0.020731696509756148,
      "rewards/forensics_tiou_reward": 0.17099995911121368,
      "step": 10
    },
    {
      "completion_length": 21.703125,
      "epoch": 0.03379416282642089,
      "grad_norm": 7.5534749031066895,
      "kl": 0.31103515625,
      "learning_rate": 9.923076923076923e-07,
      "loss": 0.5313,
      "reward": 0.6101964116096497,
      "reward_std": 0.14328953623771667,
      "reward_std_all": 0.168047696352005,
      "rewards-max/forensics_sinkhorn_reward": 0.953924298286438,
      "rewards-max/forensics_strict_edge_reward": 0.3550787717103958,
      "rewards-max/forensics_tiou_reward": 0.8898778855800629,
      "rewards/forensics_sinkhorn_reward": 0.4808973968029022,
      "rewards/forensics_strict_edge_reward": 0.017716040834784508,
      "rewards/forensics_tiou_reward": 0.19245732575654984,
      "step": 11
    },
    {
      "completion_length": 23.75,
      "epoch": 0.03686635944700461,
      "grad_norm": 7.000489711761475,
      "kl": 0.16748046875,
      "learning_rate": 9.915384615384616e-07,
      "loss": 0.7356,
      "reward": 0.560980498790741,
      "reward_std": 0.09331471472978592,
      "reward_std_all": 0.1564619541168213,
      "rewards-max/forensics_sinkhorn_reward": 0.6973535418510437,
      "rewards-max/forensics_strict_edge_reward": 0.023956687564350432,
      "rewards-max/forensics_tiou_reward": 0.4378853738307953,
      "rewards/forensics_sinkhorn_reward": 0.41837045550346375,
      "rewards/forensics_strict_edge_reward": 0.001963268199290269,
      "rewards/forensics_tiou_reward": 0.08936677128076553,
      "step": 12
    },
    {
      "completion_length": 25.015625,
      "epoch": 0.039938556067588324,
      "grad_norm": 5.876563549041748,
      "kl": 0.16943359375,
      "learning_rate": 9.907692307692307e-07,
      "loss": 0.5086,
      "reward": 0.5669562220573425,
      "reward_std": 0.0855768546462059,
      "reward_std_all": 0.1572589948773384,
      "rewards-max/forensics_sinkhorn_reward": 0.8043360114097595,
      "rewards-max/forensics_strict_edge_reward": 0.0025827323631801846,
      "rewards-max/forensics_tiou_reward": 0.5905881226062775,
      "rewards/forensics_sinkhorn_reward": 0.42309512197971344,
      "rewards/forensics_strict_edge_reward": 0.00013842033641253693,
      "rewards/forensics_tiou_reward": 0.15300129354000092,
      "step": 13
    },
    {
      "completion_length": 25.765625,
      "epoch": 0.043010752688172046,
      "grad_norm": 6.544926166534424,
      "kl": 0.16162109375,
      "learning_rate": 9.9e-07,
      "loss": 0.8566,
      "reward": 0.5827749669551849,
      "reward_std": 0.06780358403921127,
      "reward_std_all": 0.1348160095512867,
      "rewards-max/forensics_sinkhorn_reward": 0.8511152863502502,
      "rewards-max/forensics_strict_edge_reward": 0.013832542579621077,
      "rewards-max/forensics_tiou_reward": 0.6825975775718689,
      "rewards/forensics_sinkhorn_reward": 0.4274456799030304,
      "rewards/forensics_strict_edge_reward": 0.0005803228523291182,
      "rewards/forensics_tiou_reward": 0.1316841021180153,
      "step": 14
    },
    {
      "completion_length": 25.625,
      "epoch": 0.04608294930875576,
      "grad_norm": 6.044095993041992,
      "kl": 0.19775390625,
      "learning_rate": 9.892307692307692e-07,
      "loss": 0.655,
      "reward": 0.5967158079147339,
      "reward_std": 0.12730014696717262,
      "reward_std_all": 0.16367751359939575,
      "rewards-max/forensics_sinkhorn_reward": 0.8360993266105652,
      "rewards-max/forensics_strict_edge_reward": 0.00401440831956279,
      "rewards-max/forensics_tiou_reward": 0.6678624749183655,
      "rewards/forensics_sinkhorn_reward": 0.4720650911331177,
      "rewards/forensics_strict_edge_reward": 0.00022310909943712431,
      "rewards/forensics_tiou_reward": 0.15156809985637665,
      "step": 15
    },
    {
      "completion_length": 24.15625,
      "epoch": 0.04915514592933948,
      "grad_norm": 6.591726779937744,
      "kl": 0.2001953125,
      "learning_rate": 9.884615384615385e-07,
      "loss": 0.416,
      "reward": 0.6616182327270508,
      "reward_std": 0.11177250742912292,
      "reward_std_all": 0.12506632506847382,
      "rewards-max/forensics_sinkhorn_reward": 0.926959902048111,
      "rewards-max/forensics_strict_edge_reward": 0.01775249093770981,
      "rewards-max/forensics_tiou_reward": 0.8330679535865784,
      "rewards/forensics_sinkhorn_reward": 0.5375887453556061,
      "rewards/forensics_strict_edge_reward": 0.0007852011040085927,
      "rewards/forensics_tiou_reward": 0.19157405197620392,
      "step": 16
    },
    {
      "completion_length": 26.234375,
      "epoch": 0.05222734254992319,
      "grad_norm": 5.98685359954834,
      "kl": 0.302734375,
      "learning_rate": 9.876923076923076e-07,
      "loss": 0.5373,
      "reward": 0.5528596937656403,
      "reward_std": 0.12573399394750595,
      "reward_std_all": 0.1760127693414688,
      "rewards-max/forensics_sinkhorn_reward": 0.7151473760604858,
      "rewards-max/forensics_strict_edge_reward": 0.003048383955915668,
      "rewards-max/forensics_tiou_reward": 0.42483802139759064,
      "rewards/forensics_sinkhorn_reward": 0.38804756104946136,
      "rewards/forensics_strict_edge_reward": 0.00013632609280966879,
      "rewards/forensics_tiou_reward": 0.10191615670919418,
      "step": 17
    },
    {
      "completion_length": 21.546875,
      "epoch": 0.055299539170506916,
      "grad_norm": 5.725204944610596,
      "kl": 0.3330078125,
      "learning_rate": 9.86923076923077e-07,
      "loss": 0.4861,
      "reward": 0.6317689716815948,
      "reward_std": 0.09890483319759369,
      "reward_std_all": 0.12607019394636154,
      "rewards-max/forensics_sinkhorn_reward": 0.9281022846698761,
      "rewards-max/forensics_strict_edge_reward": 0.0024264330795631395,
      "rewards-max/forensics_tiou_reward": 0.8192909359931946,
      "rewards/forensics_sinkhorn_reward": 0.4942435175180435,
      "rewards/forensics_strict_edge_reward": 0.00010194740747238029,
      "rewards/forensics_tiou_reward": 0.16946857422590256,
      "step": 18
    },
    {
      "completion_length": 23.671875,
      "epoch": 0.05837173579109063,
      "grad_norm": 6.7422614097595215,
      "kl": 0.3017578125,
      "learning_rate": 9.86153846153846e-07,
      "loss": 0.8064,
      "reward": 0.5687305629253387,
      "reward_std": 0.11123369634151459,
      "reward_std_all": 0.16970990598201752,
      "rewards-max/forensics_sinkhorn_reward": 0.9491944313049316,
      "rewards-max/forensics_strict_edge_reward": 0.022110791062004864,
      "rewards-max/forensics_tiou_reward": 0.8838862776756287,
      "rewards/forensics_sinkhorn_reward": 0.4555470645427704,
      "rewards/forensics_strict_edge_reward": 0.0009621047356631607,
      "rewards/forensics_tiou_reward": 0.14237740635871887,
      "step": 19
    },
    {
      "completion_length": 28.6875,
      "epoch": 0.06144393241167435,
      "grad_norm": 7.299766540527344,
      "kl": 0.2099609375,
      "learning_rate": 9.853846153846153e-07,
      "loss": 0.4992,
      "reward": 0.6233664751052856,
      "reward_std": 0.10396116226911545,
      "reward_std_all": 0.13686279207468033,
      "rewards-max/forensics_sinkhorn_reward": 0.9438132345676422,
      "rewards-max/forensics_strict_edge_reward": 0.021894400008022785,
      "rewards-max/forensics_tiou_reward": 0.8807677626609802,
      "rewards/forensics_sinkhorn_reward": 0.4970535337924957,
      "rewards/forensics_strict_edge_reward": 0.0012106046197004616,
      "rewards/forensics_tiou_reward": 0.24965795129537582,
      "step": 20
    },
    {
      "completion_length": 24.046875,
      "epoch": 0.06451612903225806,
      "grad_norm": 7.139297008514404,
      "kl": 0.2998046875,
      "learning_rate": 9.846153846153847e-07,
      "loss": -0.0131,
      "reward": 0.6586631238460541,
      "reward_std": 0.1062769740819931,
      "reward_std_all": 0.12149158865213394,
      "rewards-max/forensics_sinkhorn_reward": 0.964886486530304,
      "rewards-max/forensics_strict_edge_reward": 0.1766245341859758,
      "rewards-max/forensics_tiou_reward": 0.9188539683818817,
      "rewards/forensics_sinkhorn_reward": 0.5501957535743713,
      "rewards/forensics_strict_edge_reward": 0.007477786566596478,
      "rewards/forensics_tiou_reward": 0.24234233796596527,
      "step": 21
    },
    {
      "completion_length": 23.46875,
      "epoch": 0.06758832565284179,
      "grad_norm": 6.483025074005127,
      "kl": 0.28955078125,
      "learning_rate": 9.838461538461538e-07,
      "loss": 0.3289,
      "reward": 0.6759262979030609,
      "reward_std": 0.08503086119890213,
      "reward_std_all": 0.10729056969285011,
      "rewards-max/forensics_sinkhorn_reward": 0.9063804149627686,
      "rewards-max/forensics_strict_edge_reward": 0.0003864885451889677,
      "rewards-max/forensics_tiou_reward": 0.761429101228714,
      "rewards/forensics_sinkhorn_reward": 0.5479381084442139,
      "rewards/forensics_strict_edge_reward": 1.6141156912397037e-05,
      "rewards/forensics_tiou_reward": 0.20186502486467361,
      "step": 22
    },
    {
      "completion_length": 23.328125,
      "epoch": 0.0706605222734255,
      "grad_norm": 5.880183696746826,
      "kl": 0.28515625,
      "learning_rate": 9.83076923076923e-07,
      "loss": 0.363,
      "reward": 0.7220751643180847,
      "reward_std": 0.05348847806453705,
      "reward_std_all": 0.06794365867972374,
      "rewards-max/forensics_sinkhorn_reward": 0.9494934678077698,
      "rewards-max/forensics_strict_edge_reward": 0.16529452800750732,
      "rewards-max/forensics_tiou_reward": 0.9020977914333344,
      "rewards/forensics_sinkhorn_reward": 0.6040627360343933,
      "rewards/forensics_strict_edge_reward": 0.007507501926738769,
      "rewards/forensics_tiou_reward": 0.2536662667989731,
      "step": 23
    },
    {
      "completion_length": 23.15625,
      "epoch": 0.07373271889400922,
      "grad_norm": 6.305806636810303,
      "kl": 0.302734375,
      "learning_rate": 9.823076923076922e-07,
      "loss": 0.8104,
      "reward": 0.6401253640651703,
      "reward_std": 0.07592717185616493,
      "reward_std_all": 0.1150936596095562,
      "rewards-max/forensics_sinkhorn_reward": 0.810362309217453,
      "rewards-max/forensics_strict_edge_reward": 6.662893883913057e-05,
      "rewards-max/forensics_tiou_reward": 0.6069632768630981,
      "rewards/forensics_sinkhorn_reward": 0.49749135971069336,
      "rewards/forensics_strict_edge_reward": 3.9373969329403735e-06,
      "rewards/forensics_tiou_reward": 0.12988242506980896,
      "step": 24
    },
    {
      "completion_length": 22.140625,
      "epoch": 0.07680491551459294,
      "grad_norm": 8.357145309448242,
      "kl": 0.42578125,
      "learning_rate": 9.815384615384615e-07,
      "loss": 0.1251,
      "reward": 0.6412210464477539,
      "reward_std": 0.0948207788169384,
      "reward_std_all": 0.12226106226444244,
      "rewards-max/forensics_sinkhorn_reward": 0.9021228551864624,
      "rewards-max/forensics_strict_edge_reward": 0.010894212410903492,
      "rewards-max/forensics_tiou_reward": 0.7668349146842957,
      "rewards/forensics_sinkhorn_reward": 0.5101267099380493,
      "rewards/forensics_strict_edge_reward": 0.0004539408283363855,
      "rewards/forensics_tiou_reward": 0.13821794465184212,
      "step": 25
    },
    {
      "completion_length": 24.84375,
      "epoch": 0.07987711213517665,
      "grad_norm": 6.222749710083008,
      "kl": 0.24560546875,
      "learning_rate": 9.807692307692306e-07,
      "loss": 0.3333,
      "reward": 0.6788141131401062,
      "reward_std": 0.0705704353749752,
      "reward_std_all": 0.09481744840741158,
      "rewards-max/forensics_sinkhorn_reward": 0.9629515111446381,
      "rewards-max/forensics_strict_edge_reward": 0.1902864659205079,
      "rewards-max/forensics_tiou_reward": 0.9174692630767822,
      "rewards/forensics_sinkhorn_reward": 0.5429506152868271,
      "rewards/forensics_strict_edge_reward": 0.007975005137268454,
      "rewards/forensics_tiou_reward": 0.19742275029420853,
      "step": 26
    },
    {
      "completion_length": 24.421875,
      "epoch": 0.08294930875576037,
      "grad_norm": 7.171803951263428,
      "kl": 0.4501953125,
      "learning_rate": 9.8e-07,
      "loss": 0.5984,
      "reward": 0.684033215045929,
      "reward_std": 0.061292197555303574,
      "reward_std_all": 0.08682284876704216,
      "rewards-max/forensics_sinkhorn_reward": 0.8571203649044037,
      "rewards-max/forensics_strict_edge_reward": 0.0005996643362777831,
      "rewards-max/forensics_tiou_reward": 0.6667603850364685,
      "rewards/forensics_sinkhorn_reward": 0.5174548923969269,
      "rewards/forensics_strict_edge_reward": 2.5041582350837643e-05,
      "rewards/forensics_tiou_reward": 0.18299618363380432,
      "step": 27
    },
    {
      "completion_length": 22.796875,
      "epoch": 0.08602150537634409,
      "grad_norm": 5.411977767944336,
      "kl": 0.4091796875,
      "learning_rate": 9.792307692307693e-07,
      "loss": 0.405,
      "reward": 0.6591670513153076,
      "reward_std": 0.10444765165448189,
      "reward_std_all": 0.12249496206641197,
      "rewards-max/forensics_sinkhorn_reward": 0.8964106738567352,
      "rewards-max/forensics_strict_edge_reward": 0.0067554491688497365,
      "rewards-max/forensics_tiou_reward": 0.7765367329120636,
      "rewards/forensics_sinkhorn_reward": 0.539239227771759,
      "rewards/forensics_strict_edge_reward": 0.0003427133124205284,
      "rewards/forensics_tiou_reward": 0.2027970477938652,
      "step": 28
    },
    {
      "completion_length": 21.8125,
      "epoch": 0.0890937019969278,
      "grad_norm": 7.126173496246338,
      "kl": 0.4208984375,
      "learning_rate": 9.784615384615384e-07,
      "loss": 0.551,
      "reward": 0.643310934305191,
      "reward_std": 0.05009162425994873,
      "reward_std_all": 0.10359115898609161,
      "rewards-max/forensics_sinkhorn_reward": 0.8852065205574036,
      "rewards-max/forensics_strict_edge_reward": 0.03374008230457548,
      "rewards-max/forensics_tiou_reward": 0.7399689257144928,
      "rewards/forensics_sinkhorn_reward": 0.5136191993951797,
      "rewards/forensics_strict_edge_reward": 0.001865106095920055,
      "rewards/forensics_tiou_reward": 0.22111177444458008,
      "step": 29
    },
    {
      "completion_length": 27.734375,
      "epoch": 0.09216589861751152,
      "grad_norm": 5.707650661468506,
      "kl": 0.287109375,
      "learning_rate": 9.776923076923077e-07,
      "loss": 0.487,
      "reward": 0.6659985184669495,
      "reward_std": 0.08380122855305672,
      "reward_std_all": 0.1037074439227581,
      "rewards-max/forensics_sinkhorn_reward": 0.9229583144187927,
      "rewards-max/forensics_strict_edge_reward": 0.1354072503745556,
      "rewards-max/forensics_tiou_reward": 0.8095270097255707,
      "rewards/forensics_sinkhorn_reward": 0.5169288367033005,
      "rewards/forensics_strict_edge_reward": 0.005840569909196347,
      "rewards/forensics_tiou_reward": 0.22921697795391083,
      "step": 30
    },
    {
      "completion_length": 24.40625,
      "epoch": 0.09523809523809523,
      "grad_norm": 6.24884557723999,
      "kl": 0.30078125,
      "learning_rate": 9.769230769230768e-07,
      "loss": 0.6589,
      "reward": 0.6392602920532227,
      "reward_std": 0.10717959702014923,
      "reward_std_all": 0.13465750962495804,
      "rewards-max/forensics_sinkhorn_reward": 0.8982238471508026,
      "rewards-max/forensics_strict_edge_reward": 0.1520347185432911,
      "rewards-max/forensics_tiou_reward": 0.7740064561367035,
      "rewards/forensics_sinkhorn_reward": 0.48238660395145416,
      "rewards/forensics_strict_edge_reward": 0.006496764370240271,
      "rewards/forensics_tiou_reward": 0.17804523557424545,
      "step": 31
    },
    {
      "completion_length": 25.171875,
      "epoch": 0.09831029185867896,
      "grad_norm": 5.17914342880249,
      "kl": 0.29296875,
      "learning_rate": 9.761538461538461e-07,
      "loss": 0.6322,
      "reward": 0.6674498617649078,
      "reward_std": 0.08872529491782188,
      "reward_std_all": 0.10853579640388489,
      "rewards-max/forensics_sinkhorn_reward": 0.857435554265976,
      "rewards-max/forensics_strict_edge_reward": 0.004058904480189085,
      "rewards-max/forensics_tiou_reward": 0.6862230598926544,
      "rewards/forensics_sinkhorn_reward": 0.49986782670021057,
      "rewards/forensics_strict_edge_reward": 0.00021553279657382518,
      "rewards/forensics_tiou_reward": 0.20623653382062912,
      "step": 32
    },
    {
      "completion_length": 27.515625,
      "epoch": 0.10138248847926268,
      "grad_norm": 6.148991107940674,
      "kl": 0.22998046875,
      "learning_rate": 9.753846153846154e-07,
      "loss": 0.5572,
      "reward": 0.633224755525589,
      "reward_std": 0.08855130523443222,
      "reward_std_all": 0.12229292467236519,
      "rewards-max/forensics_sinkhorn_reward": 0.7760298848152161,
      "rewards-max/forensics_strict_edge_reward": 0.03997922859252867,
      "rewards-max/forensics_tiou_reward": 0.5560178458690643,
      "rewards/forensics_sinkhorn_reward": 0.4773750752210617,
      "rewards/forensics_strict_edge_reward": 0.0016658852145496894,
      "rewards/forensics_tiou_reward": 0.15759941190481186,
      "step": 33
    },
    {
      "completion_length": 26.140625,
      "epoch": 0.10445468509984639,
      "grad_norm": 5.861257076263428,
      "kl": 0.3173828125,
      "learning_rate": 9.746153846153845e-07,
      "loss": 0.7267,
      "reward": 0.6759535074234009,
      "reward_std": 0.08192706853151321,
      "reward_std_all": 0.10154104977846146,
      "rewards-max/forensics_sinkhorn_reward": 0.9064610302448273,
      "rewards-max/forensics_strict_edge_reward": 0.15391016006469727,
      "rewards-max/forensics_tiou_reward": 0.789433479309082,
      "rewards/forensics_sinkhorn_reward": 0.5525744259357452,
      "rewards/forensics_strict_edge_reward": 0.006553542334586382,
      "rewards/forensics_tiou_reward": 0.22910284250974655,
      "step": 34
    },
    {
      "completion_length": 28.46875,
      "epoch": 0.10752688172043011,
      "grad_norm": 5.253672122955322,
      "kl": 0.341796875,
      "learning_rate": 9.738461538461539e-07,
      "loss": 0.7219,
      "reward": 0.6554085314273834,
      "reward_std": 0.08356896042823792,
      "reward_std_all": 0.11276204138994217,
      "rewards-max/forensics_sinkhorn_reward": 0.8726183772087097,
      "rewards-max/forensics_strict_edge_reward": 0.1139159738086164,
      "rewards-max/forensics_tiou_reward": 0.7354656159877777,
      "rewards/forensics_sinkhorn_reward": 0.5200017988681793,
      "rewards/forensics_strict_edge_reward": 0.005085679833428003,
      "rewards/forensics_tiou_reward": 0.19769610464572906,
      "step": 35
    },
    {
      "completion_length": 24.859375,
      "epoch": 0.11059907834101383,
      "grad_norm": 6.155404567718506,
      "kl": 0.2509765625,
      "learning_rate": 9.73076923076923e-07,
      "loss": 0.5363,
      "reward": 0.6615864932537079,
      "reward_std": 0.08276910707354546,
      "reward_std_all": 0.11047735810279846,
      "rewards-max/forensics_sinkhorn_reward": 0.8711915910243988,
      "rewards-max/forensics_strict_edge_reward": 0.008918786784334998,
      "rewards-max/forensics_tiou_reward": 0.6862347722053528,
      "rewards/forensics_sinkhorn_reward": 0.5253790616989136,
      "rewards/forensics_strict_edge_reward": 0.0010099554525506704,
      "rewards/forensics_tiou_reward": 0.23180187493562698,
      "step": 36
    },
    {
      "completion_length": 26.453125,
      "epoch": 0.11367127496159754,
      "grad_norm": 19.496318817138672,
      "kl": 0.7119140625,
      "learning_rate": 9.723076923076923e-07,
      "loss": 0.2753,
      "reward": 0.5864214599132538,
      "reward_std": 0.11982946842908859,
      "reward_std_all": 0.1613617092370987,
      "rewards-max/forensics_sinkhorn_reward": 0.9459239542484283,
      "rewards-max/forensics_strict_edge_reward": 0.0702162609086372,
      "rewards-max/forensics_tiou_reward": 0.8783794641494751,
      "rewards/forensics_sinkhorn_reward": 0.44671396911144257,
      "rewards/forensics_strict_edge_reward": 0.0036847776791546494,
      "rewards/forensics_tiou_reward": 0.24288442730903625,
      "step": 37
    },
    {
      "completion_length": 23.6875,
      "epoch": 0.11674347158218126,
      "grad_norm": 6.401200771331787,
      "kl": 0.29541015625,
      "learning_rate": 9.715384615384616e-07,
      "loss": 0.5456,
      "reward": 0.6527158915996552,
      "reward_std": 0.09162716194987297,
      "reward_std_all": 0.11847542971372604,
      "rewards-max/forensics_sinkhorn_reward": 0.9495659172534943,
      "rewards-max/forensics_strict_edge_reward": 0.08923090994358063,
      "rewards-max/forensics_tiou_reward": 0.8893253207206726,
      "rewards/forensics_sinkhorn_reward": 0.5373213291168213,
      "rewards/forensics_strict_edge_reward": 0.004500956740230322,
      "rewards/forensics_tiou_reward": 0.23232361674308777,
      "step": 38
    },
    {
      "completion_length": 23.546875,
      "epoch": 0.11981566820276497,
      "grad_norm": 10.622596740722656,
      "kl": 0.4453125,
      "learning_rate": 9.707692307692307e-07,
      "loss": 0.5559,
      "reward": 0.6706763505935669,
      "reward_std": 0.09497852250933647,
      "reward_std_all": 0.11335716769099236,
      "rewards-max/forensics_sinkhorn_reward": 0.939858078956604,
      "rewards-max/forensics_strict_edge_reward": 0.3083941712975502,
      "rewards-max/forensics_tiou_reward": 0.8612769544124603,
      "rewards/forensics_sinkhorn_reward": 0.543131947517395,
      "rewards/forensics_strict_edge_reward": 0.01349797286093235,
      "rewards/forensics_tiou_reward": 0.23636147379875183,
      "step": 39
    },
    {
      "completion_length": 24.65625,
      "epoch": 0.1228878648233487,
      "grad_norm": 6.277389049530029,
      "kl": 0.3837890625,
      "learning_rate": 9.7e-07,
      "loss": 0.3945,
      "reward": 0.6784734427928925,
      "reward_std": 0.097993653267622,
      "reward_std_all": 0.11353708431124687,
      "rewards-max/forensics_sinkhorn_reward": 0.9555149674415588,
      "rewards-max/forensics_strict_edge_reward": 0.04507712088525295,
      "rewards-max/forensics_tiou_reward": 0.8976238667964935,
      "rewards/forensics_sinkhorn_reward": 0.5628388822078705,
      "rewards/forensics_strict_edge_reward": 0.0036174305132590234,
      "rewards/forensics_tiou_reward": 0.2606813684105873,
      "step": 40
    },
    {
      "completion_length": 24.6875,
      "epoch": 0.1259600614439324,
      "grad_norm": 8.096061706542969,
      "kl": 0.36328125,
      "learning_rate": 9.692307692307691e-07,
      "loss": 0.4699,
      "reward": 0.6631449162960052,
      "reward_std": 0.08584506437182426,
      "reward_std_all": 0.10654018074274063,
      "rewards-max/forensics_sinkhorn_reward": 0.9432768523693085,
      "rewards-max/forensics_strict_edge_reward": 0.10412253066897392,
      "rewards-max/forensics_tiou_reward": 0.8744350671768188,
      "rewards/forensics_sinkhorn_reward": 0.5111621916294098,
      "rewards/forensics_strict_edge_reward": 0.004357942088972777,
      "rewards/forensics_tiou_reward": 0.16751857101917267,
      "step": 41
    },
    {
      "completion_length": 22.296875,
      "epoch": 0.12903225806451613,
      "grad_norm": 7.419539928436279,
      "kl": 0.2412109375,
      "learning_rate": 9.684615384615385e-07,
      "loss": 0.3993,
      "reward": 0.6866852343082428,
      "reward_std": 0.07638628780841827,
      "reward_std_all": 0.09273902326822281,
      "rewards-max/forensics_sinkhorn_reward": 0.961083710193634,
      "rewards-max/forensics_strict_edge_reward": 0.33575673773884773,
      "rewards-max/forensics_tiou_reward": 0.9295032024383545,
      "rewards/forensics_sinkhorn_reward": 0.5891121029853821,
      "rewards/forensics_strict_edge_reward": 0.022084905300289392,
      "rewards/forensics_tiou_reward": 0.25935982167720795,
      "step": 42
    },
    {
      "completion_length": 22.625,
      "epoch": 0.13210445468509985,
      "grad_norm": 7.2399396896362305,
      "kl": 0.25341796875,
      "learning_rate": 9.676923076923078e-07,
      "loss": 0.4348,
      "reward": 0.6682320833206177,
      "reward_std": 0.0873880460858345,
      "reward_std_all": 0.10813819989562035,
      "rewards-max/forensics_sinkhorn_reward": 0.9312325716018677,
      "rewards-max/forensics_strict_edge_reward": 0.03405768459197134,
      "rewards-max/forensics_tiou_reward": 0.8343399167060852,
      "rewards/forensics_sinkhorn_reward": 0.5443432629108429,
      "rewards/forensics_strict_edge_reward": 0.0030078487761784345,
      "rewards/forensics_tiou_reward": 0.2235982045531273,
      "step": 43
    },
    {
      "completion_length": 22.203125,
      "epoch": 0.13517665130568357,
      "grad_norm": 6.037975788116455,
      "kl": 0.3251953125,
      "learning_rate": 9.669230769230769e-07,
      "loss": 0.1983,
      "reward": 0.6867565214633942,
      "reward_std": 0.10166280344128609,
      "reward_std_all": 0.110755305737257,
      "rewards-max/forensics_sinkhorn_reward": 0.9545769989490509,
      "rewards-max/forensics_strict_edge_reward": 0.33302006497979164,
      "rewards-max/forensics_tiou_reward": 0.9118772745132446,
      "rewards/forensics_sinkhorn_reward": 0.5683097839355469,
      "rewards/forensics_strict_edge_reward": 0.013962602475658059,
      "rewards/forensics_tiou_reward": 0.20916517823934555,
      "step": 44
    },
    {
      "completion_length": 21.28125,
      "epoch": 0.1382488479262673,
      "grad_norm": 7.623114109039307,
      "kl": 0.25146484375,
      "learning_rate": 9.66153846153846e-07,
      "loss": 0.4376,
      "reward": 0.7046921253204346,
      "reward_std": 0.06085069477558136,
      "reward_std_all": 0.0784127339720726,
      "rewards-max/forensics_sinkhorn_reward": 0.9118042588233948,
      "rewards-max/forensics_strict_edge_reward": 0.0011696447854774306,
      "rewards-max/forensics_tiou_reward": 0.7917759418487549,
      "rewards/forensics_sinkhorn_reward": 0.605652391910553,
      "rewards/forensics_strict_edge_reward": 4.890115911848625e-05,
      "rewards/forensics_tiou_reward": 0.21470049023628235,
      "step": 45
    },
    {
      "completion_length": 21.828125,
      "epoch": 0.141321044546851,
      "grad_norm": 7.08945369720459,
      "kl": 0.29345703125,
      "learning_rate": 9.653846153846153e-07,
      "loss": 0.4035,
      "reward": 0.6469593346118927,
      "reward_std": 0.07776941731572151,
      "reward_std_all": 0.11385097354650497,
      "rewards-max/forensics_sinkhorn_reward": 0.8843746781349182,
      "rewards-max/forensics_strict_edge_reward": 0.04826308800329571,
      "rewards-max/forensics_tiou_reward": 0.6820308864116669,
      "rewards/forensics_sinkhorn_reward": 0.545612096786499,
      "rewards/forensics_strict_edge_reward": 0.0025375268768357273,
      "rewards/forensics_tiou_reward": 0.19789832085371017,
      "step": 46
    },
    {
      "completion_length": 22.5625,
      "epoch": 0.1443932411674347,
      "grad_norm": 9.063969612121582,
      "kl": 0.30078125,
      "learning_rate": 9.646153846153846e-07,
      "loss": 0.2498,
      "reward": 0.6579698920249939,
      "reward_std": 0.09791078418493271,
      "reward_std_all": 0.1191321350634098,
      "rewards-max/forensics_sinkhorn_reward": 0.9649037420749664,
      "rewards-max/forensics_strict_edge_reward": 0.3408282548189163,
      "rewards-max/forensics_tiou_reward": 0.930481344461441,
      "rewards/forensics_sinkhorn_reward": 0.5668729543685913,
      "rewards/forensics_strict_edge_reward": 0.017699286341667175,
      "rewards/forensics_tiou_reward": 0.25835292786359787,
      "step": 47
    },
    {
      "completion_length": 21.796875,
      "epoch": 0.14746543778801843,
      "grad_norm": 7.141284942626953,
      "kl": 0.23095703125,
      "learning_rate": 9.638461538461537e-07,
      "loss": 0.4097,
      "reward": 0.6957532465457916,
      "reward_std": 0.057882703840732574,
      "reward_std_all": 0.08100996166467667,
      "rewards-max/forensics_sinkhorn_reward": 0.8973855078220367,
      "rewards-max/forensics_strict_edge_reward": 0.13013990595936775,
      "rewards-max/forensics_tiou_reward": 0.7957461476325989,
      "rewards/forensics_sinkhorn_reward": 0.5875504910945892,
      "rewards/forensics_strict_edge_reward": 0.007338703493587673,
      "rewards/forensics_tiou_reward": 0.22535163164138794,
      "step": 48
    },
    {
      "completion_length": 21.703125,
      "epoch": 0.15053763440860216,
      "grad_norm": 6.202637195587158,
      "kl": 0.2236328125,
      "learning_rate": 9.63076923076923e-07,
      "loss": 0.1535,
      "reward": 0.7117677927017212,
      "reward_std": 0.07436753436923027,
      "reward_std_all": 0.08464494906365871,
      "rewards-max/forensics_sinkhorn_reward": 0.8356577455997467,
      "rewards-max/forensics_strict_edge_reward": 0.007580853649415076,
      "rewards-max/forensics_tiou_reward": 0.6535336673259735,
      "rewards/forensics_sinkhorn_reward": 0.5792003870010376,
      "rewards/forensics_strict_edge_reward": 0.0003159284678986296,
      "rewards/forensics_tiou_reward": 0.17782237380743027,
      "step": 49
    },
    {
      "completion_length": 25.515625,
      "epoch": 0.15360983102918588,
      "grad_norm": 5.624034881591797,
      "kl": 0.2265625,
      "learning_rate": 9.623076923076922e-07,
      "loss": 0.3841,
      "reward": 0.6048153638839722,
      "reward_std": 0.1148572601377964,
      "reward_std_all": 0.14909683912992477,
      "rewards-max/forensics_sinkhorn_reward": 0.8602370917797089,
      "rewards-max/forensics_strict_edge_reward": 0.07428172510117292,
      "rewards-max/forensics_tiou_reward": 0.7821664214134216,
      "rewards/forensics_sinkhorn_reward": 0.4476906359195709,
      "rewards/forensics_strict_edge_reward": 0.0031545239908155054,
      "rewards/forensics_tiou_reward": 0.18212398886680603,
      "step": 50
    },
    {
      "completion_length": 25.28125,
      "epoch": 0.15668202764976957,
      "grad_norm": 6.133326053619385,
      "kl": 0.24853515625,
      "learning_rate": 9.615384615384615e-07,
      "loss": 0.4039,
      "reward": 0.6510983407497406,
      "reward_std": 0.08656837418675423,
      "reward_std_all": 0.11920152232050896,
      "rewards-max/forensics_sinkhorn_reward": 0.9077745378017426,
      "rewards-max/forensics_strict_edge_reward": 0.07590900175273418,
      "rewards-max/forensics_tiou_reward": 0.799401193857193,
      "rewards/forensics_sinkhorn_reward": 0.5187965929508209,
      "rewards/forensics_strict_edge_reward": 0.003172659664414823,
      "rewards/forensics_tiou_reward": 0.22111834585666656,
      "step": 51
    },
    {
      "completion_length": 25.359375,
      "epoch": 0.1597542242703533,
      "grad_norm": 5.236649036407471,
      "kl": 0.2236328125,
      "learning_rate": 9.607692307692308e-07,
      "loss": 0.2983,
      "reward": 0.706808865070343,
      "reward_std": 0.06691741570830345,
      "reward_std_all": 0.0817158930003643,
      "rewards-max/forensics_sinkhorn_reward": 0.9762270152568817,
      "rewards-max/forensics_strict_edge_reward": 0.3271576166152954,
      "rewards-max/forensics_tiou_reward": 0.9616460502147675,
      "rewards/forensics_sinkhorn_reward": 0.6047322452068329,
      "rewards/forensics_strict_edge_reward": 0.023497309535741806,
      "rewards/forensics_tiou_reward": 0.3365948349237442,
      "step": 52
    },
    {
      "completion_length": 25.84375,
      "epoch": 0.16282642089093702,
      "grad_norm": 8.187204360961914,
      "kl": 0.251953125,
      "learning_rate": 9.6e-07,
      "loss": 0.6931,
      "reward": 0.6830079853534698,
      "reward_std": 0.06791668757796288,
      "reward_std_all": 0.09339850023388863,
      "rewards-max/forensics_sinkhorn_reward": 0.9298369288444519,
      "rewards-max/forensics_strict_edge_reward": 0.49838246405124664,
      "rewards-max/forensics_tiou_reward": 0.855391800403595,
      "rewards/forensics_sinkhorn_reward": 0.5299429297447205,
      "rewards/forensics_strict_edge_reward": 0.02644503442570567,
      "rewards/forensics_tiou_reward": 0.2610948234796524,
      "step": 53
    },
    {
      "completion_length": 22.515625,
      "epoch": 0.16589861751152074,
      "grad_norm": 5.936975479125977,
      "kl": 0.22314453125,
      "learning_rate": 9.592307692307692e-07,
      "loss": 0.4822,
      "reward": 0.6890252828598022,
      "reward_std": 0.04633209854364395,
      "reward_std_all": 0.0775459036231041,
      "rewards-max/forensics_sinkhorn_reward": 0.9465627670288086,
      "rewards-max/forensics_strict_edge_reward": 0.5438719317317009,
      "rewards-max/forensics_tiou_reward": 0.9019484221935272,
      "rewards/forensics_sinkhorn_reward": 0.5837785303592682,
      "rewards/forensics_strict_edge_reward": 0.03763326839543879,
      "rewards/forensics_tiou_reward": 0.23775899410247803,
      "step": 54
    },
    {
      "completion_length": 24.859375,
      "epoch": 0.16897081413210446,
      "grad_norm": 6.960536956787109,
      "kl": 0.23046875,
      "learning_rate": 9.584615384615383e-07,
      "loss": 0.4846,
      "reward": 0.6228988468647003,
      "reward_std": 0.10509834438562393,
      "reward_std_all": 0.13565684109926224,
      "rewards-max/forensics_sinkhorn_reward": 0.8117028176784515,
      "rewards-max/forensics_strict_edge_reward": 0.039703802464828186,
      "rewards-max/forensics_tiou_reward": 0.6117427945137024,
      "rewards/forensics_sinkhorn_reward": 0.47044235467910767,
      "rewards/forensics_strict_edge_reward": 0.0018846286662039802,
      "rewards/forensics_tiou_reward": 0.16882973164319992,
      "step": 55
    },
    {
      "completion_length": 26.765625,
      "epoch": 0.17204301075268819,
      "grad_norm": 7.571044445037842,
      "kl": 0.26220703125,
      "learning_rate": 9.576923076923077e-07,
      "loss": 0.6766,
      "reward": 0.6270418465137482,
      "reward_std": 0.11115968599915504,
      "reward_std_all": 0.14182214438915253,
      "rewards-max/forensics_sinkhorn_reward": 0.9227036833763123,
      "rewards-max/forensics_strict_edge_reward": 0.004702921200077981,
      "rewards-max/forensics_tiou_reward": 0.7992222607135773,
      "rewards/forensics_sinkhorn_reward": 0.4877731502056122,
      "rewards/forensics_strict_edge_reward": 0.00041154338941851165,
      "rewards/forensics_tiou_reward": 0.24467384070158005,
      "step": 56
    },
    {
      "completion_length": 26.421875,
      "epoch": 0.17511520737327188,
      "grad_norm": 5.860063552856445,
      "kl": 0.259765625,
      "learning_rate": 9.569230769230768e-07,
      "loss": 0.3856,
      "reward": 0.6741425395011902,
      "reward_std": 0.03749843779951334,
      "reward_std_all": 0.08129160851240158,
      "rewards-max/forensics_sinkhorn_reward": 0.8823851644992828,
      "rewards-max/forensics_strict_edge_reward": 0.15612926706671715,
      "rewards-max/forensics_tiou_reward": 0.792861819267273,
      "rewards/forensics_sinkhorn_reward": 0.5155566036701202,
      "rewards/forensics_strict_edge_reward": 0.007169221993535757,
      "rewards/forensics_tiou_reward": 0.2744649648666382,
      "step": 57
    },
    {
      "completion_length": 23.0,
      "epoch": 0.1781874039938556,
      "grad_norm": 6.378067493438721,
      "kl": 0.18212890625,
      "learning_rate": 9.56153846153846e-07,
      "loss": 0.443,
      "reward": 0.6775027811527252,
      "reward_std": 0.05126640200614929,
      "reward_std_all": 0.08421312645077705,
      "rewards-max/forensics_sinkhorn_reward": 0.9843709468841553,
      "rewards-max/forensics_strict_edge_reward": 0.31976478546857834,
      "rewards-max/forensics_tiou_reward": 0.9851152300834656,
      "rewards/forensics_sinkhorn_reward": 0.5506415963172913,
      "rewards/forensics_strict_edge_reward": 0.015592400915920734,
      "rewards/forensics_tiou_reward": 0.25679294764995575,
      "step": 58
    },
    {
      "completion_length": 26.3125,
      "epoch": 0.18125960061443933,
      "grad_norm": 6.740096569061279,
      "kl": 0.2470703125,
      "learning_rate": 9.553846153846154e-07,
      "loss": 0.5775,
      "reward": 0.5546090602874756,
      "reward_std": 0.09741637110710144,
      "reward_std_all": 0.16165301203727722,
      "rewards-max/forensics_sinkhorn_reward": 0.9818667471408844,
      "rewards-max/forensics_strict_edge_reward": 0.18279672786593437,
      "rewards-max/forensics_tiou_reward": 0.9791266620159149,
      "rewards/forensics_sinkhorn_reward": 0.401819109916687,
      "rewards/forensics_strict_edge_reward": 0.00895152217708528,
      "rewards/forensics_tiou_reward": 0.17467762529850006,
      "step": 59
    },
    {
      "completion_length": 25.96875,
      "epoch": 0.18433179723502305,
      "grad_norm": 6.8356194496154785,
      "kl": 0.27978515625,
      "learning_rate": 9.546153846153845e-07,
      "loss": 0.0738,
      "reward": 0.6651937663555145,
      "reward_std": 0.12139313668012619,
      "reward_std_all": 0.13280784711241722,
      "rewards-max/forensics_sinkhorn_reward": 0.9127927124500275,
      "rewards-max/forensics_strict_edge_reward": 0.28016313537955284,
      "rewards-max/forensics_tiou_reward": 0.8092687427997589,
      "rewards/forensics_sinkhorn_reward": 0.5103559494018555,
      "rewards/forensics_strict_edge_reward": 0.01483380002900958,
      "rewards/forensics_tiou_reward": 0.2436404824256897,
      "step": 60
    },
    {
      "completion_length": 27.203125,
      "epoch": 0.18740399385560677,
      "grad_norm": 5.939598083496094,
      "kl": 0.2412109375,
      "learning_rate": 9.538461538461538e-07,
      "loss": 0.2334,
      "reward": 0.611267477273941,
      "reward_std": 0.12846869602799416,
      "reward_std_all": 0.15738511085510254,
      "rewards-max/forensics_sinkhorn_reward": 0.8384303748607635,
      "rewards-max/forensics_strict_edge_reward": 0.07341134804300964,
      "rewards-max/forensics_tiou_reward": 0.7080270051956177,
      "rewards/forensics_sinkhorn_reward": 0.4807131439447403,
      "rewards/forensics_strict_edge_reward": 0.006062568772904342,
      "rewards/forensics_tiou_reward": 0.2285255640745163,
      "step": 61
    },
    {
      "completion_length": 28.859375,
      "epoch": 0.19047619047619047,
      "grad_norm": 5.9984636306762695,
      "kl": 0.26171875,
      "learning_rate": 9.53076923076923e-07,
      "loss": 0.9298,
      "reward": 0.6052025854587555,
      "reward_std": 0.08302640914916992,
      "reward_std_all": 0.1349259838461876,
      "rewards-max/forensics_sinkhorn_reward": 0.8927110135555267,
      "rewards-max/forensics_strict_edge_reward": 0.014206655090674758,
      "rewards-max/forensics_tiou_reward": 0.7619015574455261,
      "rewards/forensics_sinkhorn_reward": 0.45096229016780853,
      "rewards/forensics_strict_edge_reward": 0.0008116197859635577,
      "rewards/forensics_tiou_reward": 0.17003869265317917,
      "step": 62
    },
    {
      "completion_length": 24.75,
      "epoch": 0.1935483870967742,
      "grad_norm": 5.960659980773926,
      "kl": 0.21240234375,
      "learning_rate": 9.523076923076923e-07,
      "loss": 0.1749,
      "reward": 0.6735385060310364,
      "reward_std": 0.11549662798643112,
      "reward_std_all": 0.12360000610351562,
      "rewards-max/forensics_sinkhorn_reward": 0.9238346517086029,
      "rewards-max/forensics_strict_edge_reward": 0.26249659806489944,
      "rewards-max/forensics_tiou_reward": 0.8452518582344055,
      "rewards/forensics_sinkhorn_reward": 0.5388524532318115,
      "rewards/forensics_strict_edge_reward": 0.011391129810363054,
      "rewards/forensics_tiou_reward": 0.24077250063419342,
      "step": 63
    },
    {
      "completion_length": 24.765625,
      "epoch": 0.1966205837173579,
      "grad_norm": 7.932613849639893,
      "kl": 0.224609375,
      "learning_rate": 9.515384615384615e-07,
      "loss": 0.9883,
      "reward": 0.6362729668617249,
      "reward_std": 0.11237429082393646,
      "reward_std_all": 0.1337043046951294,
      "rewards-max/forensics_sinkhorn_reward": 0.9584080278873444,
      "rewards-max/forensics_strict_edge_reward": 0.09803902776911855,
      "rewards-max/forensics_tiou_reward": 0.9077941477298737,
      "rewards/forensics_sinkhorn_reward": 0.4921126067638397,
      "rewards/forensics_strict_edge_reward": 0.006710962668876164,
      "rewards/forensics_tiou_reward": 0.25987982004880905,
      "step": 64
    },
    {
      "completion_length": 24.484375,
      "epoch": 0.19969278033794163,
      "grad_norm": 6.923433303833008,
      "kl": 0.19775390625,
      "learning_rate": 9.507692307692308e-07,
      "loss": 0.5094,
      "reward": 0.7027601599693298,
      "reward_std": 0.05488782748579979,
      "reward_std_all": 0.07684159278869629,
      "rewards-max/forensics_sinkhorn_reward": 0.9273669123649597,
      "rewards-max/forensics_strict_edge_reward": 0.08132950216531754,
      "rewards-max/forensics_tiou_reward": 0.8377790153026581,
      "rewards/forensics_sinkhorn_reward": 0.58687424659729,
      "rewards/forensics_strict_edge_reward": 0.003630526945926249,
      "rewards/forensics_tiou_reward": 0.2919265329837799,
      "step": 65
    },
    {
      "completion_length": 23.6875,
      "epoch": 0.20276497695852536,
      "grad_norm": 6.646726608276367,
      "kl": 0.25537109375,
      "learning_rate": 9.499999999999999e-07,
      "loss": 0.8043,
      "reward": 0.6196697354316711,
      "reward_std": 0.07984206080436707,
      "reward_std_all": 0.12560364603996277,
      "rewards-max/forensics_sinkhorn_reward": 0.8964976966381073,
      "rewards-max/forensics_strict_edge_reward": 0.09551665280014277,
      "rewards-max/forensics_tiou_reward": 0.8128584325313568,
      "rewards/forensics_sinkhorn_reward": 0.5066942274570465,
      "rewards/forensics_strict_edge_reward": 0.004004684800747782,
      "rewards/forensics_tiou_reward": 0.1679498478770256,
      "step": 66
    },
    {
      "completion_length": 24.703125,
      "epoch": 0.20583717357910905,
      "grad_norm": 6.543240070343018,
      "kl": 0.28515625,
      "learning_rate": 9.492307692307691e-07,
      "loss": 0.1551,
      "reward": 0.6370021402835846,
      "reward_std": 0.1328437626361847,
      "reward_std_all": 0.14500318467617035,
      "rewards-max/forensics_sinkhorn_reward": 0.8809637129306793,
      "rewards-max/forensics_strict_edge_reward": 0.06995366746559739,
      "rewards-max/forensics_tiou_reward": 0.735927015542984,
      "rewards/forensics_sinkhorn_reward": 0.4875611662864685,
      "rewards/forensics_strict_edge_reward": 0.004600713960826397,
      "rewards/forensics_tiou_reward": 0.23517527431249619,
      "step": 67
    },
    {
      "completion_length": 26.125,
      "epoch": 0.20890937019969277,
      "grad_norm": 6.581725120544434,
      "kl": 0.3037109375,
      "learning_rate": 9.484615384615384e-07,
      "loss": 0.5783,
      "reward": 0.6268743276596069,
      "reward_std": 0.10345739126205444,
      "reward_std_all": 0.13361292332410812,
      "rewards-max/forensics_sinkhorn_reward": 0.9542747437953949,
      "rewards-max/forensics_strict_edge_reward": 0.16016917303204536,
      "rewards-max/forensics_tiou_reward": 0.9013654589653015,
      "rewards/forensics_sinkhorn_reward": 0.5075769871473312,
      "rewards/forensics_strict_edge_reward": 0.008639580104500055,
      "rewards/forensics_tiou_reward": 0.2433336302638054,
      "step": 68
    },
    {
      "completion_length": 24.453125,
      "epoch": 0.2119815668202765,
      "grad_norm": 6.668725490570068,
      "kl": 0.2607421875,
      "learning_rate": 9.476923076923077e-07,
      "loss": 0.8093,
      "reward": 0.641578733921051,
      "reward_std": 0.09178872406482697,
      "reward_std_all": 0.12189074605703354,
      "rewards-max/forensics_sinkhorn_reward": 0.9252424836158752,
      "rewards-max/forensics_strict_edge_reward": 0.24330435530282557,
      "rewards-max/forensics_tiou_reward": 0.843779593706131,
      "rewards/forensics_sinkhorn_reward": 0.5030757635831833,
      "rewards/forensics_strict_edge_reward": 0.021494151034858078,
      "rewards/forensics_tiou_reward": 0.23356077075004578,
      "step": 69
    },
    {
      "completion_length": 25.578125,
      "epoch": 0.21505376344086022,
      "grad_norm": 7.827467918395996,
      "kl": 0.25927734375,
      "learning_rate": 9.46923076923077e-07,
      "loss": 0.3902,
      "reward": 0.6535875797271729,
      "reward_std": 0.0769102405756712,
      "reward_std_all": 0.11039720103144646,
      "rewards-max/forensics_sinkhorn_reward": 0.9004859924316406,
      "rewards-max/forensics_strict_edge_reward": 0.3424911629408598,
      "rewards-max/forensics_tiou_reward": 0.7765955924987793,
      "rewards/forensics_sinkhorn_reward": 0.5210138559341431,
      "rewards/forensics_strict_edge_reward": 0.01485834177583456,
      "rewards/forensics_tiou_reward": 0.24163435399532318,
      "step": 70
    },
    {
      "completion_length": 24.859375,
      "epoch": 0.21812596006144394,
      "grad_norm": 6.801212787628174,
      "kl": 0.412109375,
      "learning_rate": 9.461538461538461e-07,
      "loss": 0.6087,
      "reward": 0.6610190868377686,
      "reward_std": 0.08528131619095802,
      "reward_std_all": 0.10964996740221977,
      "rewards-max/forensics_sinkhorn_reward": 0.9604062139987946,
      "rewards-max/forensics_strict_edge_reward": 0.022901722230017185,
      "rewards-max/forensics_tiou_reward": 0.9151546061038971,
      "rewards/forensics_sinkhorn_reward": 0.5088961720466614,
      "rewards/forensics_strict_edge_reward": 0.0009821516287047416,
      "rewards/forensics_tiou_reward": 0.24668511748313904,
      "step": 71
    },
    {
      "completion_length": 21.734375,
      "epoch": 0.22119815668202766,
      "grad_norm": 6.934815406799316,
      "kl": 0.2939453125,
      "learning_rate": 9.453846153846153e-07,
      "loss": 0.1469,
      "reward": 0.6496914327144623,
      "reward_std": 0.07358513679355383,
      "reward_std_all": 0.11118406057357788,
      "rewards-max/forensics_sinkhorn_reward": 0.9138135015964508,
      "rewards-max/forensics_strict_edge_reward": 0.026983465999364853,
      "rewards-max/forensics_tiou_reward": 0.8223091661930084,
      "rewards/forensics_sinkhorn_reward": 0.5457803606987,
      "rewards/forensics_strict_edge_reward": 0.00113104292540811,
      "rewards/forensics_tiou_reward": 0.20746179670095444,
      "step": 72
    },
    {
      "completion_length": 21.953125,
      "epoch": 0.22427035330261136,
      "grad_norm": 6.214349746704102,
      "kl": 0.248046875,
      "learning_rate": 9.446153846153846e-07,
      "loss": 0.3054,
      "reward": 0.6811797022819519,
      "reward_std": 0.07487102970480919,
      "reward_std_all": 0.09466884657740593,
      "rewards-max/forensics_sinkhorn_reward": 0.9519692063331604,
      "rewards-max/forensics_strict_edge_reward": 0.12351062800735235,
      "rewards-max/forensics_tiou_reward": 0.8825201094150543,
      "rewards/forensics_sinkhorn_reward": 0.6033353209495544,
      "rewards/forensics_strict_edge_reward": 0.006288357079029083,
      "rewards/forensics_tiou_reward": 0.2790585309267044,
      "step": 73
    },
    {
      "completion_length": 24.3125,
      "epoch": 0.22734254992319508,
      "grad_norm": 7.139036655426025,
      "kl": 0.2412109375,
      "learning_rate": 9.438461538461538e-07,
      "loss": 0.2947,
      "reward": 0.6929177939891815,
      "reward_std": 0.07152633741497993,
      "reward_std_all": 0.08789272606372833,
      "rewards-max/forensics_sinkhorn_reward": 0.8465499877929688,
      "rewards-max/forensics_strict_edge_reward": 0.016763589694164693,
      "rewards-max/forensics_tiou_reward": 0.7526929080486298,
      "rewards/forensics_sinkhorn_reward": 0.5563032627105713,
      "rewards/forensics_strict_edge_reward": 0.0013489347620634362,
      "rewards/forensics_tiou_reward": 0.16865310072898865,
      "step": 74
    },
    {
      "completion_length": 21.765625,
      "epoch": 0.2304147465437788,
      "grad_norm": 8.53697395324707,
      "kl": 0.27734375,
      "learning_rate": 9.43076923076923e-07,
      "loss": 0.2635,
      "reward": 0.653946727514267,
      "reward_std": 0.07434065267443657,
      "reward_std_all": 0.10535956546664238,
      "rewards-max/forensics_sinkhorn_reward": 0.856048971414566,
      "rewards-max/forensics_strict_edge_reward": 0.0007034196307813545,
      "rewards-max/forensics_tiou_reward": 0.6172007024288177,
      "rewards/forensics_sinkhorn_reward": 0.5526880919933319,
      "rewards/forensics_strict_edge_reward": 2.931029118258266e-05,
      "rewards/forensics_tiou_reward": 0.22387902438640594,
      "step": 75
    },
    {
      "completion_length": 21.578125,
      "epoch": 0.23348694316436253,
      "grad_norm": 8.721624374389648,
      "kl": 0.353515625,
      "learning_rate": 9.423076923076923e-07,
      "loss": 0.4104,
      "reward": 0.6645663976669312,
      "reward_std": 0.09656339511275291,
      "reward_std_all": 0.11418482661247253,
      "rewards-max/forensics_sinkhorn_reward": 0.9352458417415619,
      "rewards-max/forensics_strict_edge_reward": 0.39614428783534095,
      "rewards-max/forensics_tiou_reward": 0.8067883253097534,
      "rewards/forensics_sinkhorn_reward": 0.5596814751625061,
      "rewards/forensics_strict_edge_reward": 0.017222215618858172,
      "rewards/forensics_tiou_reward": 0.23525745421648026,
      "step": 76
    },
    {
      "completion_length": 22.78125,
      "epoch": 0.23655913978494625,
      "grad_norm": 7.850525856018066,
      "kl": 0.3193359375,
      "learning_rate": 9.415384615384615e-07,
      "loss": -0.0014,
      "reward": 0.6248152554035187,
      "reward_std": 0.14021777361631393,
      "reward_std_all": 0.15291760861873627,
      "rewards-max/forensics_sinkhorn_reward": 0.8726848363876343,
      "rewards-max/forensics_strict_edge_reward": 0.03506167698651552,
      "rewards-max/forensics_tiou_reward": 0.7082406282424927,
      "rewards/forensics_sinkhorn_reward": 0.48001109063625336,
      "rewards/forensics_strict_edge_reward": 0.0017101969569921494,
      "rewards/forensics_tiou_reward": 0.15055106580257416,
      "step": 77
    },
    {
      "completion_length": 21.796875,
      "epoch": 0.23963133640552994,
      "grad_norm": 5.9468770027160645,
      "kl": 0.19873046875,
      "learning_rate": 9.407692307692308e-07,
      "loss": 0.3571,
      "reward": 0.7321111559867859,
      "reward_std": 0.06148539297282696,
      "reward_std_all": 0.07002006471157074,
      "rewards-max/forensics_sinkhorn_reward": 0.9563377797603607,
      "rewards-max/forensics_strict_edge_reward": 0.12088413536548615,
      "rewards-max/forensics_tiou_reward": 0.9033258855342865,
      "rewards/forensics_sinkhorn_reward": 0.6518141329288483,
      "rewards/forensics_strict_edge_reward": 0.008770657237619162,
      "rewards/forensics_tiou_reward": 0.36795641481876373,
      "step": 78
    },
    {
      "completion_length": 21.21875,
      "epoch": 0.24270353302611367,
      "grad_norm": 6.0112128257751465,
      "kl": 0.22412109375,
      "learning_rate": 9.399999999999999e-07,
      "loss": 0.3011,
      "reward": 0.7223767340183258,
      "reward_std": 0.04767436347901821,
      "reward_std_all": 0.06624428182840347,
      "rewards-max/forensics_sinkhorn_reward": 0.9578279554843903,
      "rewards-max/forensics_strict_edge_reward": 0.23647147417068481,
      "rewards-max/forensics_tiou_reward": 0.9039097726345062,
      "rewards/forensics_sinkhorn_reward": 0.6101439297199249,
      "rewards/forensics_strict_edge_reward": 0.012936137150973082,
      "rewards/forensics_tiou_reward": 0.2813671603798866,
      "step": 79
    },
    {
      "completion_length": 21.0625,
      "epoch": 0.2457757296466974,
      "grad_norm": 5.93679141998291,
      "kl": 0.322265625,
      "learning_rate": 9.392307692307691e-07,
      "loss": 0.1202,
      "reward": 0.7048754096031189,
      "reward_std": 0.08220724016427994,
      "reward_std_all": 0.0902990773320198,
      "rewards-max/forensics_sinkhorn_reward": 0.9794158637523651,
      "rewards-max/forensics_strict_edge_reward": 0.29562604427337646,
      "rewards-max/forensics_tiou_reward": 0.9742545187473297,
      "rewards/forensics_sinkhorn_reward": 0.641751617193222,
      "rewards/forensics_strict_edge_reward": 0.02477138303220272,
      "rewards/forensics_tiou_reward": 0.33506447076797485,
      "step": 80
    },
    {
      "completion_length": 22.03125,
      "epoch": 0.2488479262672811,
      "grad_norm": 11.032200813293457,
      "kl": 0.662109375,
      "learning_rate": 9.384615384615384e-07,
      "loss": 0.6336,
      "reward": 0.6553550958633423,
      "reward_std": 0.052410317584872246,
      "reward_std_all": 0.09584042057394981,
      "rewards-max/forensics_sinkhorn_reward": 0.829448401927948,
      "rewards-max/forensics_strict_edge_reward": 1.918750605511832e-06,
      "rewards-max/forensics_tiou_reward": 0.663599506020546,
      "rewards/forensics_sinkhorn_reward": 0.5311787724494934,
      "rewards/forensics_strict_edge_reward": 1.3459428238604687e-07,
      "rewards/forensics_tiou_reward": 0.21439751982688904,
      "step": 81
    },
    {
      "completion_length": 23.03125,
      "epoch": 0.2519201228878648,
      "grad_norm": 7.845782279968262,
      "kl": 0.26171875,
      "learning_rate": 9.376923076923076e-07,
      "loss": 0.5018,
      "reward": 0.6661651730537415,
      "reward_std": 0.07922758162021637,
      "reward_std_all": 0.10271195694804192,
      "rewards-max/forensics_sinkhorn_reward": 0.9836176037788391,
      "rewards-max/forensics_strict_edge_reward": 0.6548857688903809,
      "rewards-max/forensics_tiou_reward": 0.9827814698219299,
      "rewards/forensics_sinkhorn_reward": 0.5442734658718109,
      "rewards/forensics_strict_edge_reward": 0.027562405914068222,
      "rewards/forensics_tiou_reward": 0.2513616159558296,
      "step": 82
    },
    {
      "completion_length": 23.421875,
      "epoch": 0.25499231950844853,
      "grad_norm": 6.553048610687256,
      "kl": 0.37890625,
      "learning_rate": 9.36923076923077e-07,
      "loss": 0.4258,
      "reward": 0.6328037679195404,
      "reward_std": 0.10304749011993408,
      "reward_std_all": 0.13262351974844933,
      "rewards-max/forensics_sinkhorn_reward": 0.8279472589492798,
      "rewards-max/forensics_strict_edge_reward": 0.001755729434080422,
      "rewards-max/forensics_tiou_reward": 0.625878781080246,
      "rewards/forensics_sinkhorn_reward": 0.5079894363880157,
      "rewards/forensics_strict_edge_reward": 9.09781338123139e-05,
      "rewards/forensics_tiou_reward": 0.18926215171813965,
      "step": 83
    },
    {
      "completion_length": 20.21875,
      "epoch": 0.25806451612903225,
      "grad_norm": 7.892749309539795,
      "kl": 0.3134765625,
      "learning_rate": 9.361538461538461e-07,
      "loss": 0.4554,
      "reward": 0.6677021086215973,
      "reward_std": 0.08726832643151283,
      "reward_std_all": 0.10984897240996361,
      "rewards-max/forensics_sinkhorn_reward": 0.9441543221473694,
      "rewards-max/forensics_strict_edge_reward": 0.02178615738483733,
      "rewards-max/forensics_tiou_reward": 0.8696588277816772,
      "rewards/forensics_sinkhorn_reward": 0.5468272566795349,
      "rewards/forensics_strict_edge_reward": 0.0013822628175074285,
      "rewards/forensics_tiou_reward": 0.21041541546583176,
      "step": 84
    },
    {
      "completion_length": 23.375,
      "epoch": 0.261136712749616,
      "grad_norm": 8.466023445129395,
      "kl": 0.35302734375,
      "learning_rate": 9.353846153846153e-07,
      "loss": 0.4969,
      "reward": 0.6812966763973236,
      "reward_std": 0.043543255887925625,
      "reward_std_all": 0.08268647640943527,
      "rewards-max/forensics_sinkhorn_reward": 0.9252322912216187,
      "rewards-max/forensics_strict_edge_reward": 0.06811600946821272,
      "rewards-max/forensics_tiou_reward": 0.8599048256874084,
      "rewards/forensics_sinkhorn_reward": 0.5399543344974518,
      "rewards/forensics_strict_edge_reward": 0.0028382154650898883,
      "rewards/forensics_tiou_reward": 0.2033541277050972,
      "step": 85
    },
    {
      "completion_length": 22.9375,
      "epoch": 0.2642089093701997,
      "grad_norm": 6.417745113372803,
      "kl": 0.25634765625,
      "learning_rate": 9.346153846153846e-07,
      "loss": 0.3721,
      "reward": 0.7346004545688629,
      "reward_std": 0.0527922548353672,
      "reward_std_all": 0.06131123751401901,
      "rewards-max/forensics_sinkhorn_reward": 0.9657564759254456,
      "rewards-max/forensics_strict_edge_reward": 0.4636124148964882,
      "rewards-max/forensics_tiou_reward": 0.9458412826061249,
      "rewards/forensics_sinkhorn_reward": 0.6290736794471741,
      "rewards/forensics_strict_edge_reward": 0.034813910722732544,
      "rewards/forensics_tiou_reward": 0.353148877620697,
      "step": 86
    },
    {
      "completion_length": 23.421875,
      "epoch": 0.2672811059907834,
      "grad_norm": 7.548181056976318,
      "kl": 0.26171875,
      "learning_rate": 9.338461538461538e-07,
      "loss": 0.3616,
      "reward": 0.6691786646842957,
      "reward_std": 0.0665619745850563,
      "reward_std_all": 0.09704909473657608,
      "rewards-max/forensics_sinkhorn_reward": 0.9712095260620117,
      "rewards-max/forensics_strict_edge_reward": 0.4479029178619385,
      "rewards-max/forensics_tiou_reward": 0.9472731947898865,
      "rewards/forensics_sinkhorn_reward": 0.5517107248306274,
      "rewards/forensics_strict_edge_reward": 0.03686772100627422,
      "rewards/forensics_tiou_reward": 0.2739100977778435,
      "step": 87
    },
    {
      "completion_length": 22.109375,
      "epoch": 0.27035330261136714,
      "grad_norm": 7.527749061584473,
      "kl": 0.2783203125,
      "learning_rate": 9.33076923076923e-07,
      "loss": 0.3911,
      "reward": 0.6837450563907623,
      "reward_std": 0.05839063972234726,
      "reward_std_all": 0.08437519147992134,
      "rewards-max/forensics_sinkhorn_reward": 0.9799560904502869,
      "rewards-max/forensics_strict_edge_reward": 0.14492706209421158,
      "rewards-max/forensics_tiou_reward": 0.9735087454319,
      "rewards/forensics_sinkhorn_reward": 0.5688613951206207,
      "rewards/forensics_strict_edge_reward": 0.010727799497544765,
      "rewards/forensics_tiou_reward": 0.2610919177532196,
      "step": 88
    },
    {
      "completion_length": 22.421875,
      "epoch": 0.27342549923195086,
      "grad_norm": 9.357507705688477,
      "kl": 0.345703125,
      "learning_rate": 9.323076923076922e-07,
      "loss": 0.4132,
      "reward": 0.7071184515953064,
      "reward_std": 0.05857260152697563,
      "reward_std_all": 0.0767289437353611,
      "rewards-max/forensics_sinkhorn_reward": 0.888411283493042,
      "rewards-max/forensics_strict_edge_reward": 0.07814538100501522,
      "rewards-max/forensics_tiou_reward": 0.7390692830085754,
      "rewards/forensics_sinkhorn_reward": 0.5856000483036041,
      "rewards/forensics_strict_edge_reward": 0.004885817656031577,
      "rewards/forensics_tiou_reward": 0.25695740431547165,
      "step": 89
    },
    {
      "completion_length": 23.953125,
      "epoch": 0.2764976958525346,
      "grad_norm": 7.689690113067627,
      "kl": 0.2646484375,
      "learning_rate": 9.315384615384615e-07,
      "loss": 0.1093,
      "reward": 0.7089518308639526,
      "reward_std": 0.06363620981574059,
      "reward_std_all": 0.07785885035991669,
      "rewards-max/forensics_sinkhorn_reward": 0.9383449256420135,
      "rewards-max/forensics_strict_edge_reward": 0.335917916148901,
      "rewards-max/forensics_tiou_reward": 0.8832114338874817,
      "rewards/forensics_sinkhorn_reward": 0.6180711388587952,
      "rewards/forensics_strict_edge_reward": 0.03799408278428018,
      "rewards/forensics_tiou_reward": 0.3090725839138031,
      "step": 90
    },
    {
      "completion_length": 21.90625,
      "epoch": 0.27956989247311825,
      "grad_norm": 7.3871002197265625,
      "kl": 0.271484375,
      "learning_rate": 9.307692307692308e-07,
      "loss": 0.4215,
      "reward": 0.6574032008647919,
      "reward_std": 0.07496156916022301,
      "reward_std_all": 0.10690987855195999,
      "rewards-max/forensics_sinkhorn_reward": 0.977237343788147,
      "rewards-max/forensics_strict_edge_reward": 0.32482436299324036,
      "rewards-max/forensics_tiou_reward": 0.9653156101703644,
      "rewards/forensics_sinkhorn_reward": 0.5354529917240143,
      "rewards/forensics_strict_edge_reward": 0.017196659930050373,
      "rewards/forensics_tiou_reward": 0.24852197617292404,
      "step": 91
    },
    {
      "completion_length": 23.921875,
      "epoch": 0.282642089093702,
      "grad_norm": 6.8580803871154785,
      "kl": 0.345703125,
      "learning_rate": 9.3e-07,
      "loss": 0.3014,
      "reward": 0.6212573647499084,
      "reward_std": 0.10703380405902863,
      "reward_std_all": 0.14169516414403915,
      "rewards-max/forensics_sinkhorn_reward": 0.9522363245487213,
      "rewards-max/forensics_strict_edge_reward": 0.05680554499849677,
      "rewards-max/forensics_tiou_reward": 0.8846812546253204,
      "rewards/forensics_sinkhorn_reward": 0.5085788667201996,
      "rewards/forensics_strict_edge_reward": 0.0031244807178154588,
      "rewards/forensics_tiou_reward": 0.2570892199873924,
      "step": 92
    },
    {
      "completion_length": 27.453125,
      "epoch": 0.2857142857142857,
      "grad_norm": 9.173087120056152,
      "kl": 0.412109375,
      "learning_rate": 9.292307692307692e-07,
      "loss": 0.4079,
      "reward": 0.6022653877735138,
      "reward_std": 0.09673640131950378,
      "reward_std_all": 0.1440984085202217,
      "rewards-max/forensics_sinkhorn_reward": 0.8823326826095581,
      "rewards-max/forensics_strict_edge_reward": 0.028792994562536478,
      "rewards-max/forensics_tiou_reward": 0.7495701014995575,
      "rewards/forensics_sinkhorn_reward": 0.426814928650856,
      "rewards/forensics_strict_edge_reward": 0.001284269084862899,
      "rewards/forensics_tiou_reward": 0.19634037464857101,
      "step": 93
    },
    {
      "completion_length": 22.75,
      "epoch": 0.2887864823348694,
      "grad_norm": 9.150365829467773,
      "kl": 0.298828125,
      "learning_rate": 9.284615384615384e-07,
      "loss": 0.5442,
      "reward": 0.5926818251609802,
      "reward_std": 0.12235681712627411,
      "reward_std_all": 0.16452116519212723,
      "rewards-max/forensics_sinkhorn_reward": 0.9126644134521484,
      "rewards-max/forensics_strict_edge_reward": 0.2792669557966292,
      "rewards-max/forensics_tiou_reward": 0.8218323886394501,
      "rewards/forensics_sinkhorn_reward": 0.4853273779153824,
      "rewards/forensics_strict_edge_reward": 0.015613356314133853,
      "rewards/forensics_tiou_reward": 0.19646290689706802,
      "step": 94
    },
    {
      "completion_length": 27.0625,
      "epoch": 0.29185867895545314,
      "grad_norm": 7.805457592010498,
      "kl": 0.3388671875,
      "learning_rate": 9.276923076923076e-07,
      "loss": 0.5746,
      "reward": 0.6371923387050629,
      "reward_std": 0.09588049724698067,
      "reward_std_all": 0.1294984258711338,
      "rewards-max/forensics_sinkhorn_reward": 0.9734550416469574,
      "rewards-max/forensics_strict_edge_reward": 0.031641888825106435,
      "rewards-max/forensics_tiou_reward": 0.9422477781772614,
      "rewards/forensics_sinkhorn_reward": 0.5094963312149048,
      "rewards/forensics_strict_edge_reward": 0.0037399878638098016,
      "rewards/forensics_tiou_reward": 0.2995716333389282,
      "step": 95
    },
    {
      "completion_length": 22.9375,
      "epoch": 0.29493087557603687,
      "grad_norm": 7.147857666015625,
      "kl": 0.24560546875,
      "learning_rate": 9.26923076923077e-07,
      "loss": 0.296,
      "reward": 0.6823572516441345,
      "reward_std": 0.07321836706250906,
      "reward_std_all": 0.0930816289037466,
      "rewards-max/forensics_sinkhorn_reward": 0.9298993647098541,
      "rewards-max/forensics_strict_edge_reward": 0.10946938022971153,
      "rewards-max/forensics_tiou_reward": 0.8482304513454437,
      "rewards/forensics_sinkhorn_reward": 0.5603184103965759,
      "rewards/forensics_strict_edge_reward": 0.009150362107902765,
      "rewards/forensics_tiou_reward": 0.2547122463583946,
      "step": 96
    },
    {
      "completion_length": 24.28125,
      "epoch": 0.2980030721966206,
      "grad_norm": 7.460115909576416,
      "kl": 0.3466796875,
      "learning_rate": 9.261538461538461e-07,
      "loss": 0.5287,
      "reward": 0.647653728723526,
      "reward_std": 0.09330256655812263,
      "reward_std_all": 0.1222059316933155,
      "rewards-max/forensics_sinkhorn_reward": 0.9489806890487671,
      "rewards-max/forensics_strict_edge_reward": 0.17375990375876427,
      "rewards-max/forensics_tiou_reward": 0.8880233764648438,
      "rewards/forensics_sinkhorn_reward": 0.5385087132453918,
      "rewards/forensics_strict_edge_reward": 0.013692605018150061,
      "rewards/forensics_tiou_reward": 0.26171864569187164,
      "step": 97
    },
    {
      "completion_length": 22.65625,
      "epoch": 0.3010752688172043,
      "grad_norm": 7.04527473449707,
      "kl": 0.275390625,
      "learning_rate": 9.253846153846153e-07,
      "loss": 0.4238,
      "reward": 0.6953399479389191,
      "reward_std": 0.03814532142132521,
      "reward_std_all": 0.06832476891577244,
      "rewards-max/forensics_sinkhorn_reward": 0.8051332831382751,
      "rewards-max/forensics_strict_edge_reward": 0.1557092155862847,
      "rewards-max/forensics_tiou_reward": 0.6726722717285156,
      "rewards/forensics_sinkhorn_reward": 0.5491834729909897,
      "rewards/forensics_strict_edge_reward": 0.007441640688938378,
      "rewards/forensics_tiou_reward": 0.20238381624221802,
      "step": 98
    },
    {
      "completion_length": 23.109375,
      "epoch": 0.30414746543778803,
      "grad_norm": 8.765239715576172,
      "kl": 0.27880859375,
      "learning_rate": 9.246153846153846e-07,
      "loss": 0.1473,
      "reward": 0.6583681106567383,
      "reward_std": 0.08101974613964558,
      "reward_std_all": 0.11011885479092598,
      "rewards-max/forensics_sinkhorn_reward": 0.974000871181488,
      "rewards-max/forensics_strict_edge_reward": 0.1815263256430626,
      "rewards-max/forensics_tiou_reward": 0.9572990834712982,
      "rewards/forensics_sinkhorn_reward": 0.5285670459270477,
      "rewards/forensics_strict_edge_reward": 0.008964328793808818,
      "rewards/forensics_tiou_reward": 0.24295561015605927,
      "step": 99
    },
    {
      "completion_length": 22.65625,
      "epoch": 0.30721966205837176,
      "grad_norm": 7.341728210449219,
      "kl": 0.26123046875,
      "learning_rate": 9.238461538461538e-07,
      "loss": 0.3046,
      "reward": 0.689538449048996,
      "reward_std": 0.057242393493652344,
      "reward_std_all": 0.08385969698429108,
      "rewards-max/forensics_sinkhorn_reward": 0.9028859436511993,
      "rewards-max/forensics_strict_edge_reward": 0.017876665573567152,
      "rewards-max/forensics_tiou_reward": 0.7724985182285309,
      "rewards/forensics_sinkhorn_reward": 0.564387857913971,
      "rewards/forensics_strict_edge_reward": 0.0010577744687907398,
      "rewards/forensics_tiou_reward": 0.2516036927700043,
      "step": 100
    },
    {
      "completion_length": 22.09375,
      "epoch": 0.3102918586789555,
      "grad_norm": 7.245570659637451,
      "kl": 0.3720703125,
      "learning_rate": 9.230769230769231e-07,
      "loss": 0.0457,
      "reward": 0.6854436099529266,
      "reward_std": 0.08725164830684662,
      "reward_std_all": 0.1001085638999939,
      "rewards-max/forensics_sinkhorn_reward": 0.8615559339523315,
      "rewards-max/forensics_strict_edge_reward": 0.12076180730946362,
      "rewards-max/forensics_tiou_reward": 0.6981318891048431,
      "rewards/forensics_sinkhorn_reward": 0.5702385008335114,
      "rewards/forensics_strict_edge_reward": 0.0055656252152402885,
      "rewards/forensics_tiou_reward": 0.25737452507019043,
      "step": 101
    },
    {
      "completion_length": 23.78125,
      "epoch": 0.31336405529953915,
      "grad_norm": 7.860311508178711,
      "kl": 0.28125,
      "learning_rate": 9.223076923076922e-07,
      "loss": 0.6078,
      "reward": 0.6032379567623138,
      "reward_std": 0.146527461707592,
      "reward_std_all": 0.1664731465280056,
      "rewards-max/forensics_sinkhorn_reward": 0.9682471752166748,
      "rewards-max/forensics_strict_edge_reward": 0.40936681294544996,
      "rewards-max/forensics_tiou_reward": 0.9333525002002716,
      "rewards/forensics_sinkhorn_reward": 0.4838288426399231,
      "rewards/forensics_strict_edge_reward": 0.03469500908049383,
      "rewards/forensics_tiou_reward": 0.26966534554958344,
      "step": 102
    },
    {
      "completion_length": 24.015625,
      "epoch": 0.31643625192012287,
      "grad_norm": 8.209041595458984,
      "kl": 0.298828125,
      "learning_rate": 9.215384615384614e-07,
      "loss": 0.4449,
      "reward": 0.6350706219673157,
      "reward_std": 0.13700969517230988,
      "reward_std_all": 0.15242619812488556,
      "rewards-max/forensics_sinkhorn_reward": 0.911935567855835,
      "rewards-max/forensics_strict_edge_reward": 0.04459155444055796,
      "rewards-max/forensics_tiou_reward": 0.8166408240795135,
      "rewards/forensics_sinkhorn_reward": 0.5295437872409821,
      "rewards/forensics_strict_edge_reward": 0.0041381390183232725,
      "rewards/forensics_tiou_reward": 0.22873513400554657,
      "step": 103
    },
    {
      "completion_length": 24.6875,
      "epoch": 0.3195084485407066,
      "grad_norm": 8.642129898071289,
      "kl": 0.318359375,
      "learning_rate": 9.207692307692308e-07,
      "loss": 0.5064,
      "reward": 0.5875601768493652,
      "reward_std": 0.09581341966986656,
      "reward_std_all": 0.14243921637535095,
      "rewards-max/forensics_sinkhorn_reward": 0.8706892430782318,
      "rewards-max/forensics_strict_edge_reward": 0.11712515819817781,
      "rewards-max/forensics_tiou_reward": 0.7396859228610992,
      "rewards/forensics_sinkhorn_reward": 0.4529632478952408,
      "rewards/forensics_strict_edge_reward": 0.005003319529350847,
      "rewards/forensics_tiou_reward": 0.16683298349380493,
      "step": 104
    },
    {
      "completion_length": 23.953125,
      "epoch": 0.3225806451612903,
      "grad_norm": 6.573421955108643,
      "kl": 0.328125,
      "learning_rate": 9.2e-07,
      "loss": 0.1547,
      "reward": 0.6729561686515808,
      "reward_std": 0.1200333721935749,
      "reward_std_all": 0.12567725032567978,
      "rewards-max/forensics_sinkhorn_reward": 0.9635769724845886,
      "rewards-max/forensics_strict_edge_reward": 0.3427288681268692,
      "rewards-max/forensics_tiou_reward": 0.9324892163276672,
      "rewards/forensics_sinkhorn_reward": 0.5720445811748505,
      "rewards/forensics_strict_edge_reward": 0.02457645069807768,
      "rewards/forensics_tiou_reward": 0.2967911958694458,
      "step": 105
    },
    {
      "completion_length": 25.890625,
      "epoch": 0.32565284178187404,
      "grad_norm": 7.777248382568359,
      "kl": 0.24609375,
      "learning_rate": 9.192307692307692e-07,
      "loss": 0.208,
      "reward": 0.6634618639945984,
      "reward_std": 0.12019737437367439,
      "reward_std_all": 0.12970978394150734,
      "rewards-max/forensics_sinkhorn_reward": 0.8834024965763092,
      "rewards-max/forensics_strict_edge_reward": 0.03040526614176997,
      "rewards-max/forensics_tiou_reward": 0.7657809853553772,
      "rewards/forensics_sinkhorn_reward": 0.5078505128622055,
      "rewards/forensics_strict_edge_reward": 0.0023146100099058486,
      "rewards/forensics_tiou_reward": 0.23066333681344986,
      "step": 106
    },
    {
      "completion_length": 23.328125,
      "epoch": 0.32872503840245776,
      "grad_norm": 6.8429741859436035,
      "kl": 0.251953125,
      "learning_rate": 9.184615384615384e-07,
      "loss": 0.5054,
      "reward": 0.6601226627826691,
      "reward_std": 0.08115028589963913,
      "reward_std_all": 0.11030102521181107,
      "rewards-max/forensics_sinkhorn_reward": 0.963798850774765,
      "rewards-max/forensics_strict_edge_reward": 0.16439904272556305,
      "rewards-max/forensics_tiou_reward": 0.9163270592689514,
      "rewards/forensics_sinkhorn_reward": 0.5419783294200897,
      "rewards/forensics_strict_edge_reward": 0.012406505178660154,
      "rewards/forensics_tiou_reward": 0.30574384331703186,
      "step": 107
    },
    {
      "completion_length": 21.96875,
      "epoch": 0.3317972350230415,
      "grad_norm": 9.86298656463623,
      "kl": 0.2724609375,
      "learning_rate": 9.176923076923076e-07,
      "loss": 0.2148,
      "reward": 0.6956233084201813,
      "reward_std": 0.08374524116516113,
      "reward_std_all": 0.09884250536561012,
      "rewards-max/forensics_sinkhorn_reward": 0.9452185034751892,
      "rewards-max/forensics_strict_edge_reward": 0.05652732402086258,
      "rewards-max/forensics_tiou_reward": 0.8811068832874298,
      "rewards/forensics_sinkhorn_reward": 0.6075591146945953,
      "rewards/forensics_strict_edge_reward": 0.006393665913492441,
      "rewards/forensics_tiou_reward": 0.3436046838760376,
      "step": 108
    },
    {
      "completion_length": 25.015625,
      "epoch": 0.3348694316436252,
      "grad_norm": 7.098541259765625,
      "kl": 0.509765625,
      "learning_rate": 9.169230769230769e-07,
      "loss": 0.5215,
      "reward": 0.6169961392879486,
      "reward_std": 0.11930195614695549,
      "reward_std_all": 0.1461026668548584,
      "rewards-max/forensics_sinkhorn_reward": 0.9152398407459259,
      "rewards-max/forensics_strict_edge_reward": 0.04805204086005688,
      "rewards-max/forensics_tiou_reward": 0.8145231008529663,
      "rewards/forensics_sinkhorn_reward": 0.478699654340744,
      "rewards/forensics_strict_edge_reward": 0.0021445766906253994,
      "rewards/forensics_tiou_reward": 0.2054489105939865,
      "step": 109
    },
    {
      "completion_length": 22.8125,
      "epoch": 0.3379416282642089,
      "grad_norm": 7.912789344787598,
      "kl": 0.3056640625,
      "learning_rate": 9.161538461538462e-07,
      "loss": 0.4007,
      "reward": 0.7155248522758484,
      "reward_std": 0.06451568566262722,
      "reward_std_all": 0.07822934538125992,
      "rewards-max/forensics_sinkhorn_reward": 0.9445424377918243,
      "rewards-max/forensics_strict_edge_reward": 0.0713303706725128,
      "rewards-max/forensics_tiou_reward": 0.8827264904975891,
      "rewards/forensics_sinkhorn_reward": 0.6020147800445557,
      "rewards/forensics_strict_edge_reward": 0.005679107998730615,
      "rewards/forensics_tiou_reward": 0.26349764317274094,
      "step": 110
    },
    {
      "completion_length": 21.9375,
      "epoch": 0.34101382488479265,
      "grad_norm": 8.703043937683105,
      "kl": 0.2783203125,
      "learning_rate": 9.153846153846153e-07,
      "loss": 0.1523,
      "reward": 0.683584988117218,
      "reward_std": 0.09904959425330162,
      "reward_std_all": 0.10880675911903381,
      "rewards-max/forensics_sinkhorn_reward": 0.9785350859165192,
      "rewards-max/forensics_strict_edge_reward": 0.5038523152470589,
      "rewards-max/forensics_tiou_reward": 0.9672282636165619,
      "rewards/forensics_sinkhorn_reward": 0.592499315738678,
      "rewards/forensics_strict_edge_reward": 0.05487560760229826,
      "rewards/forensics_tiou_reward": 0.3300596550107002,
      "step": 111
    },
    {
      "completion_length": 22.609375,
      "epoch": 0.34408602150537637,
      "grad_norm": 8.911643981933594,
      "kl": 0.33203125,
      "learning_rate": 9.146153846153846e-07,
      "loss": 0.4364,
      "reward": 0.6819008588790894,
      "reward_std": 0.08432423695921898,
      "reward_std_all": 0.10503288358449936,
      "rewards-max/forensics_sinkhorn_reward": 0.9530323445796967,
      "rewards-max/forensics_strict_edge_reward": 0.2512385621666908,
      "rewards-max/forensics_tiou_reward": 0.8990578949451447,
      "rewards/forensics_sinkhorn_reward": 0.5957249402999878,
      "rewards/forensics_strict_edge_reward": 0.01672380557283759,
      "rewards/forensics_tiou_reward": 0.32985498011112213,
      "step": 112
    },
    {
      "completion_length": 24.03125,
      "epoch": 0.34715821812596004,
      "grad_norm": 7.4707207679748535,
      "kl": 0.310546875,
      "learning_rate": 9.138461538461538e-07,
      "loss": 0.3921,
      "reward": 0.676279753446579,
      "reward_std": 0.06262294203042984,
      "reward_std_all": 0.08924949914216995,
      "rewards-max/forensics_sinkhorn_reward": 0.9599973261356354,
      "rewards-max/forensics_strict_edge_reward": 0.3818525895476341,
      "rewards-max/forensics_tiou_reward": 0.9239414632320404,
      "rewards/forensics_sinkhorn_reward": 0.5451754629611969,
      "rewards/forensics_strict_edge_reward": 0.01649311906658113,
      "rewards/forensics_tiou_reward": 0.2918030098080635,
      "step": 113
    },
    {
      "completion_length": 24.28125,
      "epoch": 0.35023041474654376,
      "grad_norm": 6.69073486328125,
      "kl": 0.357421875,
      "learning_rate": 9.130769230769231e-07,
      "loss": 0.3042,
      "reward": 0.6344231367111206,
      "reward_std": 0.07478471472859383,
      "reward_std_all": 0.11564893648028374,
      "rewards-max/forensics_sinkhorn_reward": 0.9161455631256104,
      "rewards-max/forensics_strict_edge_reward": 0.20262090969481505,
      "rewards-max/forensics_tiou_reward": 0.8286730349063873,
      "rewards/forensics_sinkhorn_reward": 0.48495152592658997,
      "rewards/forensics_strict_edge_reward": 0.008560000100260368,
      "rewards/forensics_tiou_reward": 0.1880381815135479,
      "step": 114
    },
    {
      "completion_length": 24.328125,
      "epoch": 0.3533026113671275,
      "grad_norm": 7.114197731018066,
      "kl": 0.478515625,
      "learning_rate": 9.123076923076922e-07,
      "loss": 0.6364,
      "reward": 0.6596035361289978,
      "reward_std": 0.07987011596560478,
      "reward_std_all": 0.10854682326316833,
      "rewards-max/forensics_sinkhorn_reward": 0.8350997269153595,
      "rewards-max/forensics_strict_edge_reward": 0.12351400218904018,
      "rewards-max/forensics_tiou_reward": 0.7134167850017548,
      "rewards/forensics_sinkhorn_reward": 0.5134740471839905,
      "rewards/forensics_strict_edge_reward": 0.005353129585273564,
      "rewards/forensics_tiou_reward": 0.18699957430362701,
      "step": 115
    },
    {
      "completion_length": 26.21875,
      "epoch": 0.3563748079877112,
      "grad_norm": 5.747898101806641,
      "kl": 0.3232421875,
      "learning_rate": 9.115384615384614e-07,
      "loss": 0.3034,
      "reward": 0.6823802888393402,
      "reward_std": 0.10546384379267693,
      "reward_std_all": 0.1114983931183815,
      "rewards-max/forensics_sinkhorn_reward": 0.9757707417011261,
      "rewards-max/forensics_strict_edge_reward": 0.3043217329541221,
      "rewards-max/forensics_tiou_reward": 0.9644798934459686,
      "rewards/forensics_sinkhorn_reward": 0.5718118846416473,
      "rewards/forensics_strict_edge_reward": 0.012948773983225692,
      "rewards/forensics_tiou_reward": 0.29145294427871704,
      "step": 116
    },
    {
      "completion_length": 23.3125,
      "epoch": 0.35944700460829493,
      "grad_norm": 15.573790550231934,
      "kl": 0.955078125,
      "learning_rate": 9.107692307692308e-07,
      "loss": 0.4629,
      "reward": 0.6212317943572998,
      "reward_std": 0.1135740727186203,
      "reward_std_all": 0.1466195359826088,
      "rewards-max/forensics_sinkhorn_reward": 0.9428165555000305,
      "rewards-max/forensics_strict_edge_reward": 0.02529740147292614,
      "rewards-max/forensics_tiou_reward": 0.866753339767456,
      "rewards/forensics_sinkhorn_reward": 0.496863529086113,
      "rewards/forensics_strict_edge_reward": 0.002080805483274162,
      "rewards/forensics_tiou_reward": 0.19616077840328217,
      "step": 117
    },
    {
      "completion_length": 24.40625,
      "epoch": 0.36251920122887865,
      "grad_norm": 6.342844486236572,
      "kl": 0.375,
      "learning_rate": 9.1e-07,
      "loss": 0.3678,
      "reward": 0.6763631701469421,
      "reward_std": 0.04795115068554878,
      "reward_std_all": 0.08247961103916168,
      "rewards-max/forensics_sinkhorn_reward": 0.9589969515800476,
      "rewards-max/forensics_strict_edge_reward": 0.008074291050434113,
      "rewards-max/forensics_tiou_reward": 0.9052206575870514,
      "rewards/forensics_sinkhorn_reward": 0.5624092221260071,
      "rewards/forensics_strict_edge_reward": 0.0003828920307569206,
      "rewards/forensics_tiou_reward": 0.25577422231435776,
      "step": 118
    },
    {
      "completion_length": 23.25,
      "epoch": 0.3655913978494624,
      "grad_norm": 8.582140922546387,
      "kl": 0.3037109375,
      "learning_rate": 9.092307692307692e-07,
      "loss": 0.837,
      "reward": 0.6577684283256531,
      "reward_std": 0.0716112982481718,
      "reward_std_all": 0.1047898530960083,
      "rewards-max/forensics_sinkhorn_reward": 0.9507326185703278,
      "rewards-max/forensics_strict_edge_reward": 0.23010874539613724,
      "rewards-max/forensics_tiou_reward": 0.8959954082965851,
      "rewards/forensics_sinkhorn_reward": 0.5627455413341522,
      "rewards/forensics_strict_edge_reward": 0.012462008278816938,
      "rewards/forensics_tiou_reward": 0.3087894469499588,
      "step": 119
    },
    {
      "completion_length": 23.859375,
      "epoch": 0.3686635944700461,
      "grad_norm": 7.37802791595459,
      "kl": 0.3134765625,
      "learning_rate": 9.084615384615384e-07,
      "loss": 0.3768,
      "reward": 0.6267336308956146,
      "reward_std": 0.08447939157485962,
      "reward_std_all": 0.12258216366171837,
      "rewards-max/forensics_sinkhorn_reward": 0.9404972195625305,
      "rewards-max/forensics_strict_edge_reward": 0.08480102056637406,
      "rewards-max/forensics_tiou_reward": 0.868543803691864,
      "rewards/forensics_sinkhorn_reward": 0.5074075758457184,
      "rewards/forensics_strict_edge_reward": 0.004408006891026162,
      "rewards/forensics_tiou_reward": 0.2431580275297165,
      "step": 120
    },
    {
      "completion_length": 24.21875,
      "epoch": 0.3717357910906298,
      "grad_norm": 8.319513320922852,
      "kl": 0.3349609375,
      "learning_rate": 9.076923076923076e-07,
      "loss": 0.303,
      "reward": 0.6364075243473053,
      "reward_std": 0.0946289673447609,
      "reward_std_all": 0.12071672827005386,
      "rewards-max/forensics_sinkhorn_reward": 0.773699164390564,
      "rewards-max/forensics_strict_edge_reward": 0.014910894446074963,
      "rewards-max/forensics_tiou_reward": 0.6053286492824554,
      "rewards/forensics_sinkhorn_reward": 0.48028747737407684,
      "rewards/forensics_strict_edge_reward": 0.00135093426797539,
      "rewards/forensics_tiou_reward": 0.21591712534427643,
      "step": 121
    },
    {
      "completion_length": 24.171875,
      "epoch": 0.37480798771121354,
      "grad_norm": 6.864462852478027,
      "kl": 0.2939453125,
      "learning_rate": 9.069230769230769e-07,
      "loss": 0.19,
      "reward": 0.72353395819664,
      "reward_std": 0.0716211311519146,
      "reward_std_all": 0.07865647226572037,
      "rewards-max/forensics_sinkhorn_reward": 0.9441181421279907,
      "rewards-max/forensics_strict_edge_reward": 0.11190829426050186,
      "rewards-max/forensics_tiou_reward": 0.8950998187065125,
      "rewards/forensics_sinkhorn_reward": 0.6290019154548645,
      "rewards/forensics_strict_edge_reward": 0.011655391193926334,
      "rewards/forensics_tiou_reward": 0.3461327999830246,
      "step": 122
    },
    {
      "completion_length": 26.265625,
      "epoch": 0.3778801843317972,
      "grad_norm": 7.396755218505859,
      "kl": 0.318359375,
      "learning_rate": 9.061538461538461e-07,
      "loss": 0.278,
      "reward": 0.7000723779201508,
      "reward_std": 0.0656263679265976,
      "reward_std_all": 0.08398938179016113,
      "rewards-max/forensics_sinkhorn_reward": 0.9233592450618744,
      "rewards-max/forensics_strict_edge_reward": 0.1785433553159237,
      "rewards-max/forensics_tiou_reward": 0.8429342210292816,
      "rewards/forensics_sinkhorn_reward": 0.5903714597225189,
      "rewards/forensics_strict_edge_reward": 0.009084935707505792,
      "rewards/forensics_tiou_reward": 0.2930327132344246,
      "step": 123
    },
    {
      "completion_length": 22.484375,
      "epoch": 0.38095238095238093,
      "grad_norm": 10.205062866210938,
      "kl": 0.345703125,
      "learning_rate": 9.053846153846153e-07,
      "loss": 0.314,
      "reward": 0.7076875865459442,
      "reward_std": 0.10465195029973984,
      "reward_std_all": 0.10232741385698318,
      "rewards-max/forensics_sinkhorn_reward": 0.9699840545654297,
      "rewards-max/forensics_strict_edge_reward": 0.2055463194847107,
      "rewards-max/forensics_tiou_reward": 0.9403725564479828,
      "rewards/forensics_sinkhorn_reward": 0.6218309104442596,
      "rewards/forensics_strict_edge_reward": 0.015322148334234953,
      "rewards/forensics_tiou_reward": 0.3883993923664093,
      "step": 124
    },
    {
      "completion_length": 29.296875,
      "epoch": 0.38402457757296465,
      "grad_norm": 6.399243354797363,
      "kl": 0.2822265625,
      "learning_rate": 9.046153846153846e-07,
      "loss": 0.2179,
      "reward": 0.6977682411670685,
      "reward_std": 0.076755840331316,
      "reward_std_all": 0.08961419016122818,
      "rewards-max/forensics_sinkhorn_reward": 0.9488015472888947,
      "rewards-max/forensics_strict_edge_reward": 0.195326030254364,
      "rewards-max/forensics_tiou_reward": 0.8906608521938324,
      "rewards/forensics_sinkhorn_reward": 0.5399556159973145,
      "rewards/forensics_strict_edge_reward": 0.010185780003666878,
      "rewards/forensics_tiou_reward": 0.3381999731063843,
      "step": 125
    },
    {
      "completion_length": 22.546875,
      "epoch": 0.3870967741935484,
      "grad_norm": 6.60454797744751,
      "kl": 0.32421875,
      "learning_rate": 9.038461538461538e-07,
      "loss": 0.4955,
      "reward": 0.6372645199298859,
      "reward_std": 0.0764419287443161,
      "reward_std_all": 0.11423814296722412,
      "rewards-max/forensics_sinkhorn_reward": 0.891058087348938,
      "rewards-max/forensics_strict_edge_reward": 0.026643536602932727,
      "rewards-max/forensics_tiou_reward": 0.7627677023410797,
      "rewards/forensics_sinkhorn_reward": 0.5125419497489929,
      "rewards/forensics_strict_edge_reward": 0.002652056420174631,
      "rewards/forensics_tiou_reward": 0.18697787821292877,
      "step": 126
    },
    {
      "completion_length": 24.453125,
      "epoch": 0.3901689708141321,
      "grad_norm": 6.868506908416748,
      "kl": 0.2744140625,
      "learning_rate": 9.030769230769231e-07,
      "loss": 0.365,
      "reward": 0.6780570149421692,
      "reward_std": 0.07580157741904259,
      "reward_std_all": 0.09986630082130432,
      "rewards-max/forensics_sinkhorn_reward": 0.8563886284828186,
      "rewards-max/forensics_strict_edge_reward": 0.02247473336319672,
      "rewards-max/forensics_tiou_reward": 0.674530565738678,
      "rewards/forensics_sinkhorn_reward": 0.5437436699867249,
      "rewards/forensics_strict_edge_reward": 0.0019684778608279885,
      "rewards/forensics_tiou_reward": 0.26456600427627563,
      "step": 127
    },
    {
      "completion_length": 23.65625,
      "epoch": 0.3932411674347158,
      "grad_norm": 8.128719329833984,
      "kl": 0.3564453125,
      "learning_rate": 9.023076923076922e-07,
      "loss": 0.4685,
      "reward": 0.7219012379646301,
      "reward_std": 0.0955294817686081,
      "reward_std_all": 0.09185963124036789,
      "rewards-max/forensics_sinkhorn_reward": 0.9243842959403992,
      "rewards-max/forensics_strict_edge_reward": 0.14497714303433895,
      "rewards-max/forensics_tiou_reward": 0.8401767909526825,
      "rewards/forensics_sinkhorn_reward": 0.626205325126648,
      "rewards/forensics_strict_edge_reward": 0.01567086228169501,
      "rewards/forensics_tiou_reward": 0.35844922065734863,
      "step": 128
    },
    {
      "completion_length": 21.65625,
      "epoch": 0.39631336405529954,
      "grad_norm": 7.0510454177856445,
      "kl": 0.486328125,
      "learning_rate": 9.015384615384614e-07,
      "loss": 0.1434,
      "reward": 0.6675251126289368,
      "reward_std": 0.0849079042673111,
      "reward_std_all": 0.10630238801240921,
      "rewards-max/forensics_sinkhorn_reward": 0.8886216580867767,
      "rewards-max/forensics_strict_edge_reward": 0.4507608711719513,
      "rewards-max/forensics_tiou_reward": 0.7688140571117401,
      "rewards/forensics_sinkhorn_reward": 0.551700085401535,
      "rewards/forensics_strict_edge_reward": 0.01988648623228073,
      "rewards/forensics_tiou_reward": 0.2502002865076065,
      "step": 129
    },
    {
      "completion_length": 21.96875,
      "epoch": 0.39938556067588327,
      "grad_norm": 8.674551963806152,
      "kl": 0.3828125,
      "learning_rate": 9.007692307692307e-07,
      "loss": 0.15,
      "reward": 0.6381688714027405,
      "reward_std": 0.10138732194900513,
      "reward_std_all": 0.12999023497104645,
      "rewards-max/forensics_sinkhorn_reward": 0.9302818477153778,
      "rewards-max/forensics_strict_edge_reward": 0.24813437461853027,
      "rewards-max/forensics_tiou_reward": 0.8527910709381104,
      "rewards/forensics_sinkhorn_reward": 0.5146484076976776,
      "rewards/forensics_strict_edge_reward": 0.012647522613406181,
      "rewards/forensics_tiou_reward": 0.23131848871707916,
      "step": 130
    },
    {
      "completion_length": 23.6875,
      "epoch": 0.402457757296467,
      "grad_norm": 10.107806205749512,
      "kl": 0.4228515625,
      "learning_rate": 9e-07,
      "loss": 0.3137,
      "reward": 0.7060053050518036,
      "reward_std": 0.06611514277756214,
      "reward_std_all": 0.07831507548689842,
      "rewards-max/forensics_sinkhorn_reward": 0.9575806856155396,
      "rewards-max/forensics_strict_edge_reward": 0.4592139273881912,
      "rewards-max/forensics_tiou_reward": 0.925355851650238,
      "rewards/forensics_sinkhorn_reward": 0.5720586180686951,
      "rewards/forensics_strict_edge_reward": 0.019169614650309086,
      "rewards/forensics_tiou_reward": 0.2644352316856384,
      "step": 131
    },
    {
      "completion_length": 23.265625,
      "epoch": 0.4055299539170507,
      "grad_norm": 11.53878402709961,
      "kl": 0.654296875,
      "learning_rate": 8.992307692307693e-07,
      "loss": 0.2104,
      "reward": 0.6377729177474976,
      "reward_std": 0.16688823699951172,
      "reward_std_all": 0.165425106883049,
      "rewards-max/forensics_sinkhorn_reward": 0.947187751531601,
      "rewards-max/forensics_strict_edge_reward": 0.19119170680642128,
      "rewards-max/forensics_tiou_reward": 0.8937962353229523,
      "rewards/forensics_sinkhorn_reward": 0.5466712713241577,
      "rewards/forensics_strict_edge_reward": 0.022972004197072238,
      "rewards/forensics_tiou_reward": 0.2999253123998642,
      "step": 132
    },
    {
      "completion_length": 26.21875,
      "epoch": 0.40860215053763443,
      "grad_norm": 8.58533763885498,
      "kl": 0.3330078125,
      "learning_rate": 8.984615384615384e-07,
      "loss": 0.5842,
      "reward": 0.6155820488929749,
      "reward_std": 0.08552271500229836,
      "reward_std_all": 0.13436788320541382,
      "rewards-max/forensics_sinkhorn_reward": 0.9689424335956573,
      "rewards-max/forensics_strict_edge_reward": 0.4977126121520996,
      "rewards-max/forensics_tiou_reward": 0.944508284330368,
      "rewards/forensics_sinkhorn_reward": 0.5014693588018417,
      "rewards/forensics_strict_edge_reward": 0.022358364891260862,
      "rewards/forensics_tiou_reward": 0.22804804891347885,
      "step": 133
    },
    {
      "completion_length": 25.078125,
      "epoch": 0.4116743471582181,
      "grad_norm": 8.789997100830078,
      "kl": 0.294921875,
      "learning_rate": 8.976923076923076e-07,
      "loss": 0.7171,
      "reward": 0.6468852758407593,
      "reward_std": 0.09212018549442291,
      "reward_std_all": 0.12009403854608536,
      "rewards-max/forensics_sinkhorn_reward": 0.9229214489459991,
      "rewards-max/forensics_strict_edge_reward": 0.09814079920761287,
      "rewards-max/forensics_tiou_reward": 0.8224436640739441,
      "rewards/forensics_sinkhorn_reward": 0.5008554756641388,
      "rewards/forensics_strict_edge_reward": 0.005254055911791511,
      "rewards/forensics_tiou_reward": 0.22224904596805573,
      "step": 134
    },
    {
      "completion_length": 23.03125,
      "epoch": 0.4147465437788018,
      "grad_norm": 10.158905982971191,
      "kl": 0.326171875,
      "learning_rate": 8.969230769230769e-07,
      "loss": 0.6074,
      "reward": 0.6780283451080322,
      "reward_std": 0.09099799394607544,
      "reward_std_all": 0.10540280491113663,
      "rewards-max/forensics_sinkhorn_reward": 0.9684264063835144,
      "rewards-max/forensics_strict_edge_reward": 0.35866691172122955,
      "rewards-max/forensics_tiou_reward": 0.943135678768158,
      "rewards/forensics_sinkhorn_reward": 0.550849199295044,
      "rewards/forensics_strict_edge_reward": 0.026017394848167896,
      "rewards/forensics_tiou_reward": 0.2910715192556381,
      "step": 135
    },
    {
      "completion_length": 25.890625,
      "epoch": 0.41781874039938555,
      "grad_norm": 6.147733688354492,
      "kl": 0.4541015625,
      "learning_rate": 8.961538461538461e-07,
      "loss": 0.3031,
      "reward": 0.7075545787811279,
      "reward_std": 0.06650081276893616,
      "reward_std_all": 0.08323896303772926,
      "rewards-max/forensics_sinkhorn_reward": 0.9601745307445526,
      "rewards-max/forensics_strict_edge_reward": 0.020078774774447083,
      "rewards-max/forensics_tiou_reward": 0.9240604937076569,
      "rewards/forensics_sinkhorn_reward": 0.5855659544467926,
      "rewards/forensics_strict_edge_reward": 0.0024558854493079707,
      "rewards/forensics_tiou_reward": 0.36965008080005646,
      "step": 136
    },
    {
      "completion_length": 24.296875,
      "epoch": 0.42089093701996927,
      "grad_norm": 7.653773784637451,
      "kl": 0.390625,
      "learning_rate": 8.953846153846152e-07,
      "loss": 0.4089,
      "reward": 0.6743123531341553,
      "reward_std": 0.04920830950140953,
      "reward_std_all": 0.08703677728772163,
      "rewards-max/forensics_sinkhorn_reward": 0.8623957335948944,
      "rewards-max/forensics_strict_edge_reward": 0.14088464081084484,
      "rewards-max/forensics_tiou_reward": 0.6485702991485596,
      "rewards/forensics_sinkhorn_reward": 0.5475600063800812,
      "rewards/forensics_strict_edge_reward": 0.011549152565051957,
      "rewards/forensics_tiou_reward": 0.24607793986797333,
      "step": 137
    },
    {
      "completion_length": 25.078125,
      "epoch": 0.423963133640553,
      "grad_norm": 10.35831356048584,
      "kl": 0.3017578125,
      "learning_rate": 8.946153846153846e-07,
      "loss": 0.19,
      "reward": 0.6523562073707581,
      "reward_std": 0.11788557842373848,
      "reward_std_all": 0.13379783555865288,
      "rewards-max/forensics_sinkhorn_reward": 0.962790846824646,
      "rewards-max/forensics_strict_edge_reward": 0.11314093240071088,
      "rewards-max/forensics_tiou_reward": 0.9092036783695221,
      "rewards/forensics_sinkhorn_reward": 0.5122985243797302,
      "rewards/forensics_strict_edge_reward": 0.008725842424610164,
      "rewards/forensics_tiou_reward": 0.2627478986978531,
      "step": 138
    },
    {
      "completion_length": 24.203125,
      "epoch": 0.4270353302611367,
      "grad_norm": 7.248102188110352,
      "kl": 0.4091796875,
      "learning_rate": 8.938461538461538e-07,
      "loss": 0.0657,
      "reward": 0.7127441763877869,
      "reward_std": 0.05224981904029846,
      "reward_std_all": 0.07132037542760372,
      "rewards-max/forensics_sinkhorn_reward": 0.9845898747444153,
      "rewards-max/forensics_strict_edge_reward": 0.6934256851673126,
      "rewards-max/forensics_tiou_reward": 0.9869642555713654,
      "rewards/forensics_sinkhorn_reward": 0.612868994474411,
      "rewards/forensics_strict_edge_reward": 0.05061021074652672,
      "rewards/forensics_tiou_reward": 0.361317902803421,
      "step": 139
    },
    {
      "completion_length": 24.578125,
      "epoch": 0.43010752688172044,
      "grad_norm": 10.120370864868164,
      "kl": 0.29296875,
      "learning_rate": 8.930769230769231e-07,
      "loss": 0.4898,
      "reward": 0.7038536071777344,
      "reward_std": 0.08351645991206169,
      "reward_std_all": 0.08961831778287888,
      "rewards-max/forensics_sinkhorn_reward": 0.9606840014457703,
      "rewards-max/forensics_strict_edge_reward": 0.4920148327946663,
      "rewards-max/forensics_tiou_reward": 0.9308963716030121,
      "rewards/forensics_sinkhorn_reward": 0.6101877391338348,
      "rewards/forensics_strict_edge_reward": 0.025100132450461388,
      "rewards/forensics_tiou_reward": 0.3633806109428406,
      "step": 140
    },
    {
      "completion_length": 24.453125,
      "epoch": 0.43317972350230416,
      "grad_norm": 6.231029510498047,
      "kl": 0.435546875,
      "learning_rate": 8.923076923076923e-07,
      "loss": 0.1722,
      "reward": 0.6974201500415802,
      "reward_std": 0.058018190786242485,
      "reward_std_all": 0.07988247089087963,
      "rewards-max/forensics_sinkhorn_reward": 0.9790103137493134,
      "rewards-max/forensics_strict_edge_reward": 0.3306032679975033,
      "rewards-max/forensics_tiou_reward": 0.9658905267715454,
      "rewards/forensics_sinkhorn_reward": 0.5677536725997925,
      "rewards/forensics_strict_edge_reward": 0.019723741337656975,
      "rewards/forensics_tiou_reward": 0.3372327983379364,
      "step": 141
    },
    {
      "completion_length": 24.359375,
      "epoch": 0.4362519201228879,
      "grad_norm": 8.618459701538086,
      "kl": 0.36328125,
      "learning_rate": 8.915384615384614e-07,
      "loss": 0.3286,
      "reward": 0.7183058559894562,
      "reward_std": 0.07276074588298798,
      "reward_std_all": 0.07982758805155754,
      "rewards-max/forensics_sinkhorn_reward": 0.9780028164386749,
      "rewards-max/forensics_strict_edge_reward": 0.5129558444023132,
      "rewards-max/forensics_tiou_reward": 0.9675905704498291,
      "rewards/forensics_sinkhorn_reward": 0.6157523691654205,
      "rewards/forensics_strict_edge_reward": 0.027455979958176613,
      "rewards/forensics_tiou_reward": 0.32052598893642426,
      "step": 142
    },
    {
      "completion_length": 23.78125,
      "epoch": 0.4393241167434716,
      "grad_norm": 10.688720703125,
      "kl": 0.296875,
      "learning_rate": 8.907692307692307e-07,
      "loss": 0.4733,
      "reward": 0.7023454010486603,
      "reward_std": 0.04391520097851753,
      "reward_std_all": 0.06992100551724434,
      "rewards-max/forensics_sinkhorn_reward": 0.9227180778980255,
      "rewards-max/forensics_strict_edge_reward": 0.37015699967741966,
      "rewards-max/forensics_tiou_reward": 0.8492927551269531,
      "rewards/forensics_sinkhorn_reward": 0.5859768092632294,
      "rewards/forensics_strict_edge_reward": 0.04248179472051561,
      "rewards/forensics_tiou_reward": 0.30528561770915985,
      "step": 143
    },
    {
      "completion_length": 24.234375,
      "epoch": 0.4423963133640553,
      "grad_norm": 9.641036033630371,
      "kl": 0.41796875,
      "learning_rate": 8.9e-07,
      "loss": 0.3504,
      "reward": 0.6849057078361511,
      "reward_std": 0.0841999240219593,
      "reward_std_all": 0.09782916307449341,
      "rewards-max/forensics_sinkhorn_reward": 0.9675880968570709,
      "rewards-max/forensics_strict_edge_reward": 0.41594958305358887,
      "rewards-max/forensics_tiou_reward": 0.9381669461727142,
      "rewards/forensics_sinkhorn_reward": 0.5359977185726166,
      "rewards/forensics_strict_edge_reward": 0.023109788075089455,
      "rewards/forensics_tiou_reward": 0.2692837417125702,
      "step": 144
    },
    {
      "completion_length": 24.375,
      "epoch": 0.445468509984639,
      "grad_norm": 7.495700836181641,
      "kl": 0.326171875,
      "learning_rate": 8.892307692307693e-07,
      "loss": 0.1516,
      "reward": 0.7101984620094299,
      "reward_std": 0.0643758405931294,
      "reward_std_all": 0.08193213865160942,
      "rewards-max/forensics_sinkhorn_reward": 0.9778879880905151,
      "rewards-max/forensics_strict_edge_reward": 0.2017388865351677,
      "rewards-max/forensics_tiou_reward": 0.9648444056510925,
      "rewards/forensics_sinkhorn_reward": 0.6428460478782654,
      "rewards/forensics_strict_edge_reward": 0.015001815278083086,
      "rewards/forensics_tiou_reward": 0.40790437161922455,
      "step": 145
    },
    {
      "completion_length": 23.109375,
      "epoch": 0.4485407066052227,
      "grad_norm": 8.86291217803955,
      "kl": 0.322265625,
      "learning_rate": 8.884615384615384e-07,
      "loss": 0.4735,
      "reward": 0.6713484227657318,
      "reward_std": 0.08008669316768646,
      "reward_std_all": 0.10224473848938942,
      "rewards-max/forensics_sinkhorn_reward": 0.9696843028068542,
      "rewards-max/forensics_strict_edge_reward": 0.3457430303096771,
      "rewards-max/forensics_tiou_reward": 0.9464170634746552,
      "rewards/forensics_sinkhorn_reward": 0.5756288766860962,
      "rewards/forensics_strict_edge_reward": 0.02499157004058361,
      "rewards/forensics_tiou_reward": 0.3113323599100113,
      "step": 146
    },
    {
      "completion_length": 21.78125,
      "epoch": 0.45161290322580644,
      "grad_norm": 10.216455459594727,
      "kl": 0.38330078125,
      "learning_rate": 8.876923076923076e-07,
      "loss": 0.4934,
      "reward": 0.6917868256568909,
      "reward_std": 0.06107880175113678,
      "reward_std_all": 0.08561099879443645,
      "rewards-max/forensics_sinkhorn_reward": 0.9583728611469269,
      "rewards-max/forensics_strict_edge_reward": 0.25011650659143925,
      "rewards-max/forensics_tiou_reward": 0.9242581725120544,
      "rewards/forensics_sinkhorn_reward": 0.5968845784664154,
      "rewards/forensics_strict_edge_reward": 0.023754875059239566,
      "rewards/forensics_tiou_reward": 0.3120754212141037,
      "step": 147
    },
    {
      "completion_length": 22.25,
      "epoch": 0.45468509984639016,
      "grad_norm": 8.741990089416504,
      "kl": 0.27490234375,
      "learning_rate": 8.869230769230769e-07,
      "loss": 0.3381,
      "reward": 0.6950249373912811,
      "reward_std": 0.08913490734994411,
      "reward_std_all": 0.10129918903112411,
      "rewards-max/forensics_sinkhorn_reward": 0.9134423434734344,
      "rewards-max/forensics_strict_edge_reward": 0.2634928417392075,
      "rewards-max/forensics_tiou_reward": 0.8245970904827118,
      "rewards/forensics_sinkhorn_reward": 0.5744155049324036,
      "rewards/forensics_strict_edge_reward": 0.01229734264779836,
      "rewards/forensics_tiou_reward": 0.29391007125377655,
      "step": 148
    },
    {
      "completion_length": 24.375,
      "epoch": 0.4577572964669739,
      "grad_norm": 7.356131076812744,
      "kl": 0.41796875,
      "learning_rate": 8.861538461538461e-07,
      "loss": 0.4214,
      "reward": 0.72456294298172,
      "reward_std": 0.07076316326856613,
      "reward_std_all": 0.07637563161551952,
      "rewards-max/forensics_sinkhorn_reward": 0.8934732377529144,
      "rewards-max/forensics_strict_edge_reward": 0.5319918692111969,
      "rewards-max/forensics_tiou_reward": 0.8304270505905151,
      "rewards/forensics_sinkhorn_reward": 0.6036960482597351,
      "rewards/forensics_strict_edge_reward": 0.0390578331425786,
      "rewards/forensics_tiou_reward": 0.3701831102371216,
      "step": 149
    },
    {
      "completion_length": 23.265625,
      "epoch": 0.4608294930875576,
      "grad_norm": 8.87035083770752,
      "kl": 0.2763671875,
      "learning_rate": 8.853846153846154e-07,
      "loss": 0.1434,
      "reward": 0.7061339020729065,
      "reward_std": 0.06435155868530273,
      "reward_std_all": 0.08168813586235046,
      "rewards-max/forensics_sinkhorn_reward": 0.8663998544216156,
      "rewards-max/forensics_strict_edge_reward": 0.26541259613850343,
      "rewards-max/forensics_tiou_reward": 0.7053900063037872,
      "rewards/forensics_sinkhorn_reward": 0.593934565782547,
      "rewards/forensics_strict_edge_reward": 0.01675108699100747,
      "rewards/forensics_tiou_reward": 0.26705025881528854,
      "step": 150
    },
    {
      "completion_length": 21.5625,
      "epoch": 0.46390168970814133,
      "grad_norm": 8.500386238098145,
      "kl": 0.326171875,
      "learning_rate": 8.846153846153846e-07,
      "loss": 0.3504,
      "reward": 0.6674020886421204,
      "reward_std": 0.04778826795518398,
      "reward_std_all": 0.09084124118089676,
      "rewards-max/forensics_sinkhorn_reward": 0.8594225645065308,
      "rewards-max/forensics_strict_edge_reward": 0.4089423045516014,
      "rewards-max/forensics_tiou_reward": 0.7523640692234039,
      "rewards/forensics_sinkhorn_reward": 0.5602074265480042,
      "rewards/forensics_strict_edge_reward": 0.031123779248446226,
      "rewards/forensics_tiou_reward": 0.269271120429039,
      "step": 151
    },
    {
      "completion_length": 21.453125,
      "epoch": 0.46697388632872505,
      "grad_norm": 7.0468621253967285,
      "kl": 0.341796875,
      "learning_rate": 8.838461538461538e-07,
      "loss": 0.4376,
      "reward": 0.6933613419532776,
      "reward_std": 0.07251593843102455,
      "reward_std_all": 0.09056362509727478,
      "rewards-max/forensics_sinkhorn_reward": 0.9091338217258453,
      "rewards-max/forensics_strict_edge_reward": 0.19673356681596488,
      "rewards-max/forensics_tiou_reward": 0.779143214225769,
      "rewards/forensics_sinkhorn_reward": 0.6088178753852844,
      "rewards/forensics_strict_edge_reward": 0.021315802546723717,
      "rewards/forensics_tiou_reward": 0.3252509832382202,
      "step": 152
    },
    {
      "completion_length": 21.875,
      "epoch": 0.4700460829493088,
      "grad_norm": 8.60165786743164,
      "kl": 0.2265625,
      "learning_rate": 8.830769230769231e-07,
      "loss": 0.131,
      "reward": 0.7406733632087708,
      "reward_std": 0.037673261016607285,
      "reward_std_all": 0.049705143086612225,
      "rewards-max/forensics_sinkhorn_reward": 0.9610128700733185,
      "rewards-max/forensics_strict_edge_reward": 0.49020934104919434,
      "rewards-max/forensics_tiou_reward": 0.9300735592842102,
      "rewards/forensics_sinkhorn_reward": 0.6739597022533417,
      "rewards/forensics_strict_edge_reward": 0.029135676100850105,
      "rewards/forensics_tiou_reward": 0.4126415550708771,
      "step": 153
    },
    {
      "completion_length": 22.1875,
      "epoch": 0.4731182795698925,
      "grad_norm": 11.052698135375977,
      "kl": 0.7177734375,
      "learning_rate": 8.823076923076923e-07,
      "loss": 0.31,
      "reward": 0.6906349956989288,
      "reward_std": 0.0711616575717926,
      "reward_std_all": 0.08856440335512161,
      "rewards-max/forensics_sinkhorn_reward": 0.8832986056804657,
      "rewards-max/forensics_strict_edge_reward": 0.4047088027000427,
      "rewards-max/forensics_tiou_reward": 0.7930753231048584,
      "rewards/forensics_sinkhorn_reward": 0.5808746516704559,
      "rewards/forensics_strict_edge_reward": 0.021215814165771008,
      "rewards/forensics_tiou_reward": 0.2848556712269783,
      "step": 154
    },
    {
      "completion_length": 22.765625,
      "epoch": 0.47619047619047616,
      "grad_norm": 8.42642593383789,
      "kl": 0.361328125,
      "learning_rate": 8.815384615384614e-07,
      "loss": 0.3732,
      "reward": 0.7287660241127014,
      "reward_std": 0.04020201228559017,
      "reward_std_all": 0.056377969682216644,
      "rewards-max/forensics_sinkhorn_reward": 0.8814209401607513,
      "rewards-max/forensics_strict_edge_reward": 0.4116128236055374,
      "rewards-max/forensics_tiou_reward": 0.8102376461029053,
      "rewards/forensics_sinkhorn_reward": 0.6300651431083679,
      "rewards/forensics_strict_edge_reward": 0.036018874030560255,
      "rewards/forensics_tiou_reward": 0.30934426933526993,
      "step": 155
    },
    {
      "completion_length": 22.109375,
      "epoch": 0.4792626728110599,
      "grad_norm": 8.289921760559082,
      "kl": 0.384765625,
      "learning_rate": 8.807692307692307e-07,
      "loss": 0.5301,
      "reward": 0.6867094039916992,
      "reward_std": 0.06499532237648964,
      "reward_std_all": 0.08846499025821686,
      "rewards-max/forensics_sinkhorn_reward": 0.9086808860301971,
      "rewards-max/forensics_strict_edge_reward": 0.3752036914229393,
      "rewards-max/forensics_tiou_reward": 0.7914734780788422,
      "rewards/forensics_sinkhorn_reward": 0.5738388299942017,
      "rewards/forensics_strict_edge_reward": 0.016772192670032382,
      "rewards/forensics_tiou_reward": 0.2942611798644066,
      "step": 156
    },
    {
      "completion_length": 22.96875,
      "epoch": 0.4823348694316436,
      "grad_norm": 7.5519890785217285,
      "kl": 0.3349609375,
      "learning_rate": 8.799999999999999e-07,
      "loss": 0.2647,
      "reward": 0.7328740060329437,
      "reward_std": 0.07349110767245293,
      "reward_std_all": 0.07500413060188293,
      "rewards-max/forensics_sinkhorn_reward": 0.9723949134349823,
      "rewards-max/forensics_strict_edge_reward": 0.3396413326263428,
      "rewards-max/forensics_tiou_reward": 0.9511529803276062,
      "rewards/forensics_sinkhorn_reward": 0.6589792668819427,
      "rewards/forensics_strict_edge_reward": 0.029693241231143475,
      "rewards/forensics_tiou_reward": 0.4441232681274414,
      "step": 157
    },
    {
      "completion_length": 24.8125,
      "epoch": 0.48540706605222733,
      "grad_norm": 9.460902214050293,
      "kl": 0.349609375,
      "learning_rate": 8.792307692307693e-07,
      "loss": 0.045,
      "reward": 0.7133637368679047,
      "reward_std": 0.04713987372815609,
      "reward_std_all": 0.06756403669714928,
      "rewards-max/forensics_sinkhorn_reward": 0.9501349627971649,
      "rewards-max/forensics_strict_edge_reward": 0.2816109359264374,
      "rewards-max/forensics_tiou_reward": 0.8953342139720917,
      "rewards/forensics_sinkhorn_reward": 0.5871630609035492,
      "rewards/forensics_strict_edge_reward": 0.017979032825678587,
      "rewards/forensics_tiou_reward": 0.3639521598815918,
      "step": 158
    },
    {
      "completion_length": 24.15625,
      "epoch": 0.48847926267281105,
      "grad_norm": 8.016143798828125,
      "kl": 0.431640625,
      "learning_rate": 8.784615384615385e-07,
      "loss": 0.3405,
      "reward": 0.6732652187347412,
      "reward_std": 0.06728591397404671,
      "reward_std_all": 0.09625021368265152,
      "rewards-max/forensics_sinkhorn_reward": 0.9175986349582672,
      "rewards-max/forensics_strict_edge_reward": 0.11041954532265663,
      "rewards-max/forensics_tiou_reward": 0.8267011046409607,
      "rewards/forensics_sinkhorn_reward": 0.5322944521903992,
      "rewards/forensics_strict_edge_reward": 0.007295855728443712,
      "rewards/forensics_tiou_reward": 0.27508049458265305,
      "step": 159
    },
    {
      "completion_length": 23.46875,
      "epoch": 0.4915514592933948,
      "grad_norm": 12.609650611877441,
      "kl": 0.4228515625,
      "learning_rate": 8.776923076923076e-07,
      "loss": 0.3924,
      "reward": 0.7238990068435669,
      "reward_std": 0.06125400401651859,
      "reward_std_all": 0.07127600722014904,
      "rewards-max/forensics_sinkhorn_reward": 0.951969563961029,
      "rewards-max/forensics_strict_edge_reward": 0.38840945065021515,
      "rewards-max/forensics_tiou_reward": 0.912097305059433,
      "rewards/forensics_sinkhorn_reward": 0.5884752571582794,
      "rewards/forensics_strict_edge_reward": 0.02373388921841979,
      "rewards/forensics_tiou_reward": 0.3527938276529312,
      "step": 160
    },
    {
      "completion_length": 22.59375,
      "epoch": 0.4946236559139785,
      "grad_norm": 10.537129402160645,
      "kl": 0.322265625,
      "learning_rate": 8.769230769230769e-07,
      "loss": 0.4433,
      "reward": 0.6831831336021423,
      "reward_std": 0.08016844838857651,
      "reward_std_all": 0.09724629297852516,
      "rewards-max/forensics_sinkhorn_reward": 0.9854872524738312,
      "rewards-max/forensics_strict_edge_reward": 0.7421593070030212,
      "rewards-max/forensics_tiou_reward": 0.9906446635723114,
      "rewards/forensics_sinkhorn_reward": 0.5590400695800781,
      "rewards/forensics_strict_edge_reward": 0.03352787625044584,
      "rewards/forensics_tiou_reward": 0.33430516719818115,
      "step": 161
    },
    {
      "completion_length": 26.5,
      "epoch": 0.4976958525345622,
      "grad_norm": 12.940815925598145,
      "kl": 0.2978515625,
      "learning_rate": 8.761538461538461e-07,
      "loss": 0.3725,
      "reward": 0.7093576788902283,
      "reward_std": 0.0577614214271307,
      "reward_std_all": 0.07390498742461205,
      "rewards-max/forensics_sinkhorn_reward": 0.973799467086792,
      "rewards-max/forensics_strict_edge_reward": 0.5139672756195068,
      "rewards-max/forensics_tiou_reward": 0.9606402218341827,
      "rewards/forensics_sinkhorn_reward": 0.5598749816417694,
      "rewards/forensics_strict_edge_reward": 0.038950176909565926,
      "rewards/forensics_tiou_reward": 0.40842679142951965,
      "step": 162
    },
    {
      "completion_length": 21.53125,
      "epoch": 0.500768049155146,
      "grad_norm": 9.20637321472168,
      "kl": 0.4248046875,
      "learning_rate": 8.753846153846154e-07,
      "loss": 0.0609,
      "reward": 0.673048198223114,
      "reward_std": 0.06971045769751072,
      "reward_std_all": 0.0990586206316948,
      "rewards-max/forensics_sinkhorn_reward": 0.9382711350917816,
      "rewards-max/forensics_strict_edge_reward": 0.29109244234859943,
      "rewards-max/forensics_tiou_reward": 0.8771129548549652,
      "rewards/forensics_sinkhorn_reward": 0.577007383108139,
      "rewards/forensics_strict_edge_reward": 0.012637761072255671,
      "rewards/forensics_tiou_reward": 0.25452786684036255,
      "step": 163
    },
    {
      "completion_length": 26.765625,
      "epoch": 0.5038402457757296,
      "grad_norm": 10.636385917663574,
      "kl": 0.4208984375,
      "learning_rate": 8.746153846153845e-07,
      "loss": 0.4931,
      "reward": 0.6918764114379883,
      "reward_std": 0.08467650040984154,
      "reward_std_all": 0.09847413003444672,
      "rewards-max/forensics_sinkhorn_reward": 0.9590359330177307,
      "rewards-max/forensics_strict_edge_reward": 0.3948554955422878,
      "rewards-max/forensics_tiou_reward": 0.9263189435005188,
      "rewards/forensics_sinkhorn_reward": 0.5419657230377197,
      "rewards/forensics_strict_edge_reward": 0.01666127156931907,
      "rewards/forensics_tiou_reward": 0.28569725155830383,
      "step": 164
    },
    {
      "completion_length": 25.28125,
      "epoch": 0.5069124423963134,
      "grad_norm": 6.858279705047607,
      "kl": 0.375,
      "learning_rate": 8.738461538461538e-07,
      "loss": 0.3676,
      "reward": 0.6842033863067627,
      "reward_std": 0.08058230765163898,
      "reward_std_all": 0.09764283895492554,
      "rewards-max/forensics_sinkhorn_reward": 0.9861255884170532,
      "rewards-max/forensics_strict_edge_reward": 0.7265062034130096,
      "rewards-max/forensics_tiou_reward": 0.9923169612884521,
      "rewards/forensics_sinkhorn_reward": 0.5492876768112183,
      "rewards/forensics_strict_edge_reward": 0.07569453492760658,
      "rewards/forensics_tiou_reward": 0.40045931935310364,
      "step": 165
    },
    {
      "completion_length": 25.953125,
      "epoch": 0.5099846390168971,
      "grad_norm": 6.891895294189453,
      "kl": 0.34375,
      "learning_rate": 8.730769230769231e-07,
      "loss": 0.2885,
      "reward": 0.7080698609352112,
      "reward_std": 0.07903597131371498,
      "reward_std_all": 0.08980470895767212,
      "rewards-max/forensics_sinkhorn_reward": 0.955420583486557,
      "rewards-max/forensics_strict_edge_reward": 0.35890351980924606,
      "rewards-max/forensics_tiou_reward": 0.9089868068695068,
      "rewards/forensics_sinkhorn_reward": 0.5996374487876892,
      "rewards/forensics_strict_edge_reward": 0.04023379972204566,
      "rewards/forensics_tiou_reward": 0.3952977806329727,
      "step": 166
    },
    {
      "completion_length": 23.71875,
      "epoch": 0.5130568356374808,
      "grad_norm": 7.096706867218018,
      "kl": 0.6875,
      "learning_rate": 8.723076923076923e-07,
      "loss": 0.3868,
      "reward": 0.7158388197422028,
      "reward_std": 0.045646123588085175,
      "reward_std_all": 0.06456511467695236,
      "rewards-max/forensics_sinkhorn_reward": 0.9630930125713348,
      "rewards-max/forensics_strict_edge_reward": 0.5385349243879318,
      "rewards-max/forensics_tiou_reward": 0.935267984867096,
      "rewards/forensics_sinkhorn_reward": 0.6075871586799622,
      "rewards/forensics_strict_edge_reward": 0.053742521442472935,
      "rewards/forensics_tiou_reward": 0.41954636573791504,
      "step": 167
    },
    {
      "completion_length": 23.4375,
      "epoch": 0.5161290322580645,
      "grad_norm": 7.531307220458984,
      "kl": 0.46484375,
      "learning_rate": 8.715384615384614e-07,
      "loss": 0.2313,
      "reward": 0.6784879565238953,
      "reward_std": 0.04830264672636986,
      "reward_std_all": 0.08373765647411346,
      "rewards-max/forensics_sinkhorn_reward": 0.9811207354068756,
      "rewards-max/forensics_strict_edge_reward": 0.31730959936976433,
      "rewards-max/forensics_tiou_reward": 0.9770651459693909,
      "rewards/forensics_sinkhorn_reward": 0.5654587745666504,
      "rewards/forensics_strict_edge_reward": 0.033141680411063135,
      "rewards/forensics_tiou_reward": 0.3451838791370392,
      "step": 168
    },
    {
      "completion_length": 22.609375,
      "epoch": 0.5192012288786483,
      "grad_norm": 7.761549949645996,
      "kl": 0.345703125,
      "learning_rate": 8.707692307692307e-07,
      "loss": 0.0833,
      "reward": 0.7503311038017273,
      "reward_std": 0.05021502310410142,
      "reward_std_all": 0.055027782917022705,
      "rewards-max/forensics_sinkhorn_reward": 0.932536244392395,
      "rewards-max/forensics_strict_edge_reward": 0.265274815261364,
      "rewards-max/forensics_tiou_reward": 0.8688834607601166,
      "rewards/forensics_sinkhorn_reward": 0.642803430557251,
      "rewards/forensics_strict_edge_reward": 0.01541494601406157,
      "rewards/forensics_tiou_reward": 0.34454067051410675,
      "step": 169
    },
    {
      "completion_length": 23.703125,
      "epoch": 0.522273425499232,
      "grad_norm": 10.512971878051758,
      "kl": 0.373046875,
      "learning_rate": 8.699999999999999e-07,
      "loss": 0.1397,
      "reward": 0.7339924573898315,
      "reward_std": 0.0425959974527359,
      "reward_std_all": 0.05934363044798374,
      "rewards-max/forensics_sinkhorn_reward": 0.9850251972675323,
      "rewards-max/forensics_strict_edge_reward": 0.6839536428451538,
      "rewards-max/forensics_tiou_reward": 0.9889363944530487,
      "rewards/forensics_sinkhorn_reward": 0.6370934545993805,
      "rewards/forensics_strict_edge_reward": 0.05268525239080191,
      "rewards/forensics_tiou_reward": 0.3907013237476349,
      "step": 170
    },
    {
      "completion_length": 26.03125,
      "epoch": 0.5253456221198156,
      "grad_norm": 7.834076881408691,
      "kl": 0.5009765625,
      "learning_rate": 8.692307692307692e-07,
      "loss": 0.0352,
      "reward": 0.6872774362564087,
      "reward_std": 0.08753045462071896,
      "reward_std_all": 0.09825783595442772,
      "rewards-max/forensics_sinkhorn_reward": 0.8650694191455841,
      "rewards-max/forensics_strict_edge_reward": 0.004115016794173698,
      "rewards-max/forensics_tiou_reward": 0.6716097295284271,
      "rewards/forensics_sinkhorn_reward": 0.5742461085319519,
      "rewards/forensics_strict_edge_reward": 0.00039303764604792235,
      "rewards/forensics_tiou_reward": 0.22960666567087173,
      "step": 171
    },
    {
      "completion_length": 24.359375,
      "epoch": 0.5284178187403994,
      "grad_norm": 7.780462265014648,
      "kl": 0.33984375,
      "learning_rate": 8.684615384615385e-07,
      "loss": 0.3973,
      "reward": 0.7445123791694641,
      "reward_std": 0.07055432721972466,
      "reward_std_all": 0.06862817518413067,
      "rewards-max/forensics_sinkhorn_reward": 0.9506928622722626,
      "rewards-max/forensics_strict_edge_reward": 0.3421594202518463,
      "rewards-max/forensics_tiou_reward": 0.9083179533481598,
      "rewards/forensics_sinkhorn_reward": 0.6337657868862152,
      "rewards/forensics_strict_edge_reward": 0.0240458264015615,
      "rewards/forensics_tiou_reward": 0.3716711848974228,
      "step": 172
    },
    {
      "completion_length": 22.453125,
      "epoch": 0.5314900153609831,
      "grad_norm": 8.438396453857422,
      "kl": 0.69921875,
      "learning_rate": 8.676923076923076e-07,
      "loss": 0.4073,
      "reward": 0.7169879078865051,
      "reward_std": 0.1001577377319336,
      "reward_std_all": 0.09883615374565125,
      "rewards-max/forensics_sinkhorn_reward": 0.9837718605995178,
      "rewards-max/forensics_strict_edge_reward": 0.6643738597631454,
      "rewards-max/forensics_tiou_reward": 0.9827292263507843,
      "rewards/forensics_sinkhorn_reward": 0.6230911016464233,
      "rewards/forensics_strict_edge_reward": 0.05421325005590916,
      "rewards/forensics_tiou_reward": 0.39715367555618286,
      "step": 173
    },
    {
      "completion_length": 24.4375,
      "epoch": 0.5345622119815668,
      "grad_norm": 9.026039123535156,
      "kl": 0.318359375,
      "learning_rate": 8.669230769230769e-07,
      "loss": 0.3016,
      "reward": 0.7392838001251221,
      "reward_std": 0.04359443765133619,
      "reward_std_all": 0.05505991727113724,
      "rewards-max/forensics_sinkhorn_reward": 0.9289036095142365,
      "rewards-max/forensics_strict_edge_reward": 0.21364866942167282,
      "rewards-max/forensics_tiou_reward": 0.8501075804233551,
      "rewards/forensics_sinkhorn_reward": 0.6215589046478271,
      "rewards/forensics_strict_edge_reward": 0.025707476190291345,
      "rewards/forensics_tiou_reward": 0.3589998334646225,
      "step": 174
    },
    {
      "completion_length": 24.328125,
      "epoch": 0.5376344086021505,
      "grad_norm": 8.112221717834473,
      "kl": 0.5107421875,
      "learning_rate": 8.661538461538461e-07,
      "loss": 0.0967,
      "reward": 0.7048348486423492,
      "reward_std": 0.07160381600260735,
      "reward_std_all": 0.08409278839826584,
      "rewards-max/forensics_sinkhorn_reward": 0.9649006426334381,
      "rewards-max/forensics_strict_edge_reward": 0.28207121789455414,
      "rewards-max/forensics_tiou_reward": 0.9344085156917572,
      "rewards/forensics_sinkhorn_reward": 0.5771592557430267,
      "rewards/forensics_strict_edge_reward": 0.014434273354709148,
      "rewards/forensics_tiou_reward": 0.31712929904460907,
      "step": 175
    },
    {
      "completion_length": 25.703125,
      "epoch": 0.5407066052227343,
      "grad_norm": 6.444668292999268,
      "kl": 0.3828125,
      "learning_rate": 8.653846153846154e-07,
      "loss": 0.2366,
      "reward": 0.6911084949970245,
      "reward_std": 0.04371655359864235,
      "reward_std_all": 0.07606111094355583,
      "rewards-max/forensics_sinkhorn_reward": 0.9595734179019928,
      "rewards-max/forensics_strict_edge_reward": 0.3045084858313203,
      "rewards-max/forensics_tiou_reward": 0.9290362298488617,
      "rewards/forensics_sinkhorn_reward": 0.5767486393451691,
      "rewards/forensics_strict_edge_reward": 0.019337433157488704,
      "rewards/forensics_tiou_reward": 0.3422318994998932,
      "step": 176
    },
    {
      "completion_length": 23.15625,
      "epoch": 0.543778801843318,
      "grad_norm": 6.695522785186768,
      "kl": 0.2958984375,
      "learning_rate": 8.646153846153845e-07,
      "loss": 0.3428,
      "reward": 0.7175645232200623,
      "reward_std": 0.0685527678579092,
      "reward_std_all": 0.07962123304605484,
      "rewards-max/forensics_sinkhorn_reward": 0.9768767952919006,
      "rewards-max/forensics_strict_edge_reward": 0.4827862083911896,
      "rewards-max/forensics_tiou_reward": 0.9640598297119141,
      "rewards/forensics_sinkhorn_reward": 0.6464031934738159,
      "rewards/forensics_strict_edge_reward": 0.03120409045368433,
      "rewards/forensics_tiou_reward": 0.45833806693553925,
      "step": 177
    },
    {
      "completion_length": 22.421875,
      "epoch": 0.5468509984639017,
      "grad_norm": 7.994081497192383,
      "kl": 0.41796875,
      "learning_rate": 8.638461538461537e-07,
      "loss": 0.1817,
      "reward": 0.7137738168239594,
      "reward_std": 0.06923868507146835,
      "reward_std_all": 0.07725928351283073,
      "rewards-max/forensics_sinkhorn_reward": 0.9547812044620514,
      "rewards-max/forensics_strict_edge_reward": 0.21369408071041107,
      "rewards-max/forensics_tiou_reward": 0.9140904545783997,
      "rewards/forensics_sinkhorn_reward": 0.620446115732193,
      "rewards/forensics_strict_edge_reward": 0.01145265344530344,
      "rewards/forensics_tiou_reward": 0.29951339960098267,
      "step": 178
    },
    {
      "completion_length": 25.0,
      "epoch": 0.5499231950844854,
      "grad_norm": 8.088253021240234,
      "kl": 0.3818359375,
      "learning_rate": 8.630769230769231e-07,
      "loss": 0.5039,
      "reward": 0.7140334248542786,
      "reward_std": 0.08116514980792999,
      "reward_std_all": 0.08788200467824936,
      "rewards-max/forensics_sinkhorn_reward": 0.9565973579883575,
      "rewards-max/forensics_strict_edge_reward": 0.2397772977128625,
      "rewards-max/forensics_tiou_reward": 0.9076430797576904,
      "rewards/forensics_sinkhorn_reward": 0.6200776994228363,
      "rewards/forensics_strict_edge_reward": 0.031852858199272305,
      "rewards/forensics_tiou_reward": 0.3363545462489128,
      "step": 179
    },
    {
      "completion_length": 23.796875,
      "epoch": 0.5529953917050692,
      "grad_norm": 7.8705153465271,
      "kl": 0.2900390625,
      "learning_rate": 8.623076923076923e-07,
      "loss": 0.174,
      "reward": 0.7037628591060638,
      "reward_std": 0.07930002734065056,
      "reward_std_all": 0.09307575970888138,
      "rewards-max/forensics_sinkhorn_reward": 0.9783092439174652,
      "rewards-max/forensics_strict_edge_reward": 0.6005697548389435,
      "rewards-max/forensics_tiou_reward": 0.9726046323776245,
      "rewards/forensics_sinkhorn_reward": 0.6111794114112854,
      "rewards/forensics_strict_edge_reward": 0.03859691694378853,
      "rewards/forensics_tiou_reward": 0.34877797961235046,
      "step": 180
    },
    {
      "completion_length": 24.984375,
      "epoch": 0.5560675883256528,
      "grad_norm": 8.519804954528809,
      "kl": 0.3076171875,
      "learning_rate": 8.615384615384616e-07,
      "loss": -0.1967,
      "reward": 0.7350169122219086,
      "reward_std": 0.07275436073541641,
      "reward_std_all": 0.07204380817711353,
      "rewards-max/forensics_sinkhorn_reward": 0.9793530702590942,
      "rewards-max/forensics_strict_edge_reward": 0.5686750710010529,
      "rewards-max/forensics_tiou_reward": 0.9731620848178864,
      "rewards/forensics_sinkhorn_reward": 0.645029604434967,
      "rewards/forensics_strict_edge_reward": 0.03733584377914667,
      "rewards/forensics_tiou_reward": 0.4020150452852249,
      "step": 181
    },
    {
      "completion_length": 23.875,
      "epoch": 0.5591397849462365,
      "grad_norm": 9.98721694946289,
      "kl": 0.76953125,
      "learning_rate": 8.607692307692307e-07,
      "loss": 0.2839,
      "reward": 0.7129403054714203,
      "reward_std": 0.0792811531573534,
      "reward_std_all": 0.08616734854876995,
      "rewards-max/forensics_sinkhorn_reward": 0.9843117594718933,
      "rewards-max/forensics_strict_edge_reward": 0.711992472410202,
      "rewards-max/forensics_tiou_reward": 0.9853439927101135,
      "rewards/forensics_sinkhorn_reward": 0.6605785489082336,
      "rewards/forensics_strict_edge_reward": 0.07404628023505211,
      "rewards/forensics_tiou_reward": 0.4202304780483246,
      "step": 182
    },
    {
      "completion_length": 24.15625,
      "epoch": 0.5622119815668203,
      "grad_norm": 7.897840976715088,
      "kl": 0.375,
      "learning_rate": 8.599999999999999e-07,
      "loss": 0.2344,
      "reward": 0.6995872855186462,
      "reward_std": 0.08434377983212471,
      "reward_std_all": 0.09664230421185493,
      "rewards-max/forensics_sinkhorn_reward": 0.9721181988716125,
      "rewards-max/forensics_strict_edge_reward": 0.4834974706172943,
      "rewards-max/forensics_tiou_reward": 0.9570868611335754,
      "rewards/forensics_sinkhorn_reward": 0.5971918404102325,
      "rewards/forensics_strict_edge_reward": 0.03230508882552385,
      "rewards/forensics_tiou_reward": 0.3546532839536667,
      "step": 183
    },
    {
      "completion_length": 23.09375,
      "epoch": 0.565284178187404,
      "grad_norm": 6.110173225402832,
      "kl": 0.412109375,
      "learning_rate": 8.592307692307692e-07,
      "loss": 0.2485,
      "reward": 0.7109059691429138,
      "reward_std": 0.07318759523332119,
      "reward_std_all": 0.0837184265255928,
      "rewards-max/forensics_sinkhorn_reward": 0.9499955773353577,
      "rewards-max/forensics_strict_edge_reward": 0.30519454553723335,
      "rewards-max/forensics_tiou_reward": 0.9040571451187134,
      "rewards/forensics_sinkhorn_reward": 0.6235370934009552,
      "rewards/forensics_strict_edge_reward": 0.013681776821613312,
      "rewards/forensics_tiou_reward": 0.33681921660900116,
      "step": 184
    },
    {
      "completion_length": 22.4375,
      "epoch": 0.5683563748079877,
      "grad_norm": 10.982020378112793,
      "kl": 0.478515625,
      "learning_rate": 8.584615384615385e-07,
      "loss": 0.2234,
      "reward": 0.6859216690063477,
      "reward_std": 0.09664707258343697,
      "reward_std_all": 0.1031763069331646,
      "rewards-max/forensics_sinkhorn_reward": 0.9488217830657959,
      "rewards-max/forensics_strict_edge_reward": 0.4052717015147209,
      "rewards-max/forensics_tiou_reward": 0.8800025880336761,
      "rewards/forensics_sinkhorn_reward": 0.5776720345020294,
      "rewards/forensics_strict_edge_reward": 0.027043927228078246,
      "rewards/forensics_tiou_reward": 0.28146886080503464,
      "step": 185
    },
    {
      "completion_length": 22.84375,
      "epoch": 0.5714285714285714,
      "grad_norm": 8.639934539794922,
      "kl": 0.4072265625,
      "learning_rate": 8.576923076923076e-07,
      "loss": -0.0527,
      "reward": 0.7355904281139374,
      "reward_std": 0.05779901519417763,
      "reward_std_all": 0.06654025055468082,
      "rewards-max/forensics_sinkhorn_reward": 0.9695464968681335,
      "rewards-max/forensics_strict_edge_reward": 0.2612588554620743,
      "rewards-max/forensics_tiou_reward": 0.9423973262310028,
      "rewards/forensics_sinkhorn_reward": 0.646316409111023,
      "rewards/forensics_strict_edge_reward": 0.030770025216042995,
      "rewards/forensics_tiou_reward": 0.3976242244243622,
      "step": 186
    },
    {
      "completion_length": 26.21875,
      "epoch": 0.5745007680491552,
      "grad_norm": 8.417896270751953,
      "kl": 0.341796875,
      "learning_rate": 8.569230769230769e-07,
      "loss": 0.493,
      "reward": 0.6543833315372467,
      "reward_std": 0.12532109767198563,
      "reward_std_all": 0.13641640171408653,
      "rewards-max/forensics_sinkhorn_reward": 0.935571551322937,
      "rewards-max/forensics_strict_edge_reward": 0.1537964567542076,
      "rewards-max/forensics_tiou_reward": 0.8758652806282043,
      "rewards/forensics_sinkhorn_reward": 0.5093186795711517,
      "rewards/forensics_strict_edge_reward": 0.006599013460800052,
      "rewards/forensics_tiou_reward": 0.2755746468901634,
      "step": 187
    },
    {
      "completion_length": 21.96875,
      "epoch": 0.5775729646697388,
      "grad_norm": 7.680512428283691,
      "kl": 0.33984375,
      "learning_rate": 8.561538461538461e-07,
      "loss": 0.229,
      "reward": 0.6870976686477661,
      "reward_std": 0.08223536610603333,
      "reward_std_all": 0.0949506089091301,
      "rewards-max/forensics_sinkhorn_reward": 0.9657939970493317,
      "rewards-max/forensics_strict_edge_reward": 0.4068595767021179,
      "rewards-max/forensics_tiou_reward": 0.9361507296562195,
      "rewards/forensics_sinkhorn_reward": 0.6069599688053131,
      "rewards/forensics_strict_edge_reward": 0.05884275212883949,
      "rewards/forensics_tiou_reward": 0.35446797311306,
      "step": 188
    },
    {
      "completion_length": 22.03125,
      "epoch": 0.5806451612903226,
      "grad_norm": 9.090841293334961,
      "kl": 0.4658203125,
      "learning_rate": 8.553846153846154e-07,
      "loss": 0.0511,
      "reward": 0.7037010192871094,
      "reward_std": 0.06632944196462631,
      "reward_std_all": 0.07816953584551811,
      "rewards-max/forensics_sinkhorn_reward": 0.9640333950519562,
      "rewards-max/forensics_strict_edge_reward": 0.11750250356271863,
      "rewards-max/forensics_tiou_reward": 0.9253162741661072,
      "rewards/forensics_sinkhorn_reward": 0.646234542131424,
      "rewards/forensics_strict_edge_reward": 0.012854247441282496,
      "rewards/forensics_tiou_reward": 0.3780577927827835,
      "step": 189
    },
    {
      "completion_length": 24.984375,
      "epoch": 0.5837173579109063,
      "grad_norm": 8.528585433959961,
      "kl": 0.314453125,
      "learning_rate": 8.546153846153846e-07,
      "loss": 0.4096,
      "reward": 0.7124488353729248,
      "reward_std": 0.06594258733093739,
      "reward_std_all": 0.08065073005855083,
      "rewards-max/forensics_sinkhorn_reward": 0.9690888524055481,
      "rewards-max/forensics_strict_edge_reward": 0.5250524878501892,
      "rewards-max/forensics_tiou_reward": 0.9434195756912231,
      "rewards/forensics_sinkhorn_reward": 0.5949915051460266,
      "rewards/forensics_strict_edge_reward": 0.038720445707440376,
      "rewards/forensics_tiou_reward": 0.29662832617759705,
      "step": 190
    },
    {
      "completion_length": 23.53125,
      "epoch": 0.5867895545314901,
      "grad_norm": 6.631113052368164,
      "kl": 0.3046875,
      "learning_rate": 8.538461538461537e-07,
      "loss": 0.3525,
      "reward": 0.7201166152954102,
      "reward_std": 0.09355565533041954,
      "reward_std_all": 0.09177785739302635,
      "rewards-max/forensics_sinkhorn_reward": 0.9720273017883301,
      "rewards-max/forensics_strict_edge_reward": 0.5669038891792297,
      "rewards-max/forensics_tiou_reward": 0.9599409103393555,
      "rewards/forensics_sinkhorn_reward": 0.6154310405254364,
      "rewards/forensics_strict_edge_reward": 0.033452339470386505,
      "rewards/forensics_tiou_reward": 0.3689460754394531,
      "step": 191
    },
    {
      "completion_length": 23.65625,
      "epoch": 0.5898617511520737,
      "grad_norm": 7.3908514976501465,
      "kl": 0.48828125,
      "learning_rate": 8.530769230769231e-07,
      "loss": 0.5886,
      "reward": 0.721449226140976,
      "reward_std": 0.04299360513687134,
      "reward_std_all": 0.06421329826116562,
      "rewards-max/forensics_sinkhorn_reward": 0.9768735468387604,
      "rewards-max/forensics_strict_edge_reward": 0.5207925289869308,
      "rewards-max/forensics_tiou_reward": 0.9673677086830139,
      "rewards/forensics_sinkhorn_reward": 0.6384902000427246,
      "rewards/forensics_strict_edge_reward": 0.044035579077899456,
      "rewards/forensics_tiou_reward": 0.42097868025302887,
      "step": 192
    },
    {
      "completion_length": 26.328125,
      "epoch": 0.5929339477726574,
      "grad_norm": 6.980058670043945,
      "kl": 0.3544921875,
      "learning_rate": 8.523076923076923e-07,
      "loss": 0.3489,
      "reward": 0.7155221998691559,
      "reward_std": 0.05323285609483719,
      "reward_std_all": 0.06958790682256222,
      "rewards-max/forensics_sinkhorn_reward": 0.9474741816520691,
      "rewards-max/forensics_strict_edge_reward": 0.5761951208114624,
      "rewards-max/forensics_tiou_reward": 0.8958852291107178,
      "rewards/forensics_sinkhorn_reward": 0.5755581855773926,
      "rewards/forensics_strict_edge_reward": 0.03446176368743181,
      "rewards/forensics_tiou_reward": 0.3263547271490097,
      "step": 193
    },
    {
      "completion_length": 25.6875,
      "epoch": 0.5960061443932412,
      "grad_norm": 8.190871238708496,
      "kl": 0.3427734375,
      "learning_rate": 8.515384615384616e-07,
      "loss": -0.0342,
      "reward": 0.708268404006958,
      "reward_std": 0.05818825773894787,
      "reward_std_all": 0.07796911895275116,
      "rewards-max/forensics_sinkhorn_reward": 0.9569152593612671,
      "rewards-max/forensics_strict_edge_reward": 0.31561706960201263,
      "rewards-max/forensics_tiou_reward": 0.9146927297115326,
      "rewards/forensics_sinkhorn_reward": 0.5909741222858429,
      "rewards/forensics_strict_edge_reward": 0.03411435103043914,
      "rewards/forensics_tiou_reward": 0.3556179106235504,
      "step": 194
    },
    {
      "completion_length": 24.546875,
      "epoch": 0.5990783410138248,
      "grad_norm": 6.903352737426758,
      "kl": 0.3515625,
      "learning_rate": 8.507692307692307e-07,
      "loss": 0.0577,
      "reward": 0.7110490500926971,
      "reward_std": 0.08753374963998795,
      "reward_std_all": 0.09311685711145401,
      "rewards-max/forensics_sinkhorn_reward": 0.9575281739234924,
      "rewards-max/forensics_strict_edge_reward": 0.19811592623591423,
      "rewards-max/forensics_tiou_reward": 0.9127165079116821,
      "rewards/forensics_sinkhorn_reward": 0.6306317448616028,
      "rewards/forensics_strict_edge_reward": 0.008763329009525478,
      "rewards/forensics_tiou_reward": 0.4071061760187149,
      "step": 195
    },
    {
      "completion_length": 27.59375,
      "epoch": 0.6021505376344086,
      "grad_norm": 7.027512073516846,
      "kl": 0.4794921875,
      "learning_rate": 8.499999999999999e-07,
      "loss": 0.0164,
      "reward": 0.6494316756725311,
      "reward_std": 0.06675677560269833,
      "reward_std_all": 0.1060495376586914,
      "rewards-max/forensics_sinkhorn_reward": 0.8791178166866302,
      "rewards-max/forensics_strict_edge_reward": 0.12295542590436526,
      "rewards-max/forensics_tiou_reward": 0.740790992975235,
      "rewards/forensics_sinkhorn_reward": 0.5025681257247925,
      "rewards/forensics_strict_edge_reward": 0.011234891853746376,
      "rewards/forensics_tiou_reward": 0.26348260045051575,
      "step": 196
    },
    {
      "completion_length": 26.984375,
      "epoch": 0.6052227342549923,
      "grad_norm": 7.768043041229248,
      "kl": 0.3173828125,
      "learning_rate": 8.492307692307692e-07,
      "loss": 0.035,
      "reward": 0.6611106395721436,
      "reward_std": 0.08550039306282997,
      "reward_std_all": 0.11322693526744843,
      "rewards-max/forensics_sinkhorn_reward": 0.9825463891029358,
      "rewards-max/forensics_strict_edge_reward": 0.5734282732009888,
      "rewards-max/forensics_tiou_reward": 0.9804955124855042,
      "rewards/forensics_sinkhorn_reward": 0.5310185253620148,
      "rewards/forensics_strict_edge_reward": 0.034727152436971664,
      "rewards/forensics_tiou_reward": 0.3198009133338928,
      "step": 197
    },
    {
      "completion_length": 27.21875,
      "epoch": 0.6082949308755761,
      "grad_norm": 8.818231582641602,
      "kl": 0.4189453125,
      "learning_rate": 8.484615384615384e-07,
      "loss": 0.364,
      "reward": 0.6575512886047363,
      "reward_std": 0.06395049672573805,
      "reward_std_all": 0.10523782297968864,
      "rewards-max/forensics_sinkhorn_reward": 0.8418302536010742,
      "rewards-max/forensics_strict_edge_reward": 0.0989901339635253,
      "rewards-max/forensics_tiou_reward": 0.6509782075881958,
      "rewards/forensics_sinkhorn_reward": 0.5451295971870422,
      "rewards/forensics_strict_edge_reward": 0.004816401866264641,
      "rewards/forensics_tiou_reward": 0.2672139033675194,
      "step": 198
    },
    {
      "completion_length": 25.546875,
      "epoch": 0.6113671274961597,
      "grad_norm": 7.216836929321289,
      "kl": 0.361328125,
      "learning_rate": 8.476923076923077e-07,
      "loss": 0.6481,
      "reward": 0.6837886869907379,
      "reward_std": 0.09430751949548721,
      "reward_std_all": 0.10698067396879196,
      "rewards-max/forensics_sinkhorn_reward": 0.9391656219959259,
      "rewards-max/forensics_strict_edge_reward": 0.13080983608961105,
      "rewards-max/forensics_tiou_reward": 0.8739412426948547,
      "rewards/forensics_sinkhorn_reward": 0.5452632009983063,
      "rewards/forensics_strict_edge_reward": 0.011559352744370699,
      "rewards/forensics_tiou_reward": 0.3316873162984848,
      "step": 199
    },
    {
      "completion_length": 25.25,
      "epoch": 0.6144393241167435,
      "grad_norm": 7.7276129722595215,
      "kl": 0.4716796875,
      "learning_rate": 8.469230769230769e-07,
      "loss": 0.0716,
      "reward": 0.6932770609855652,
      "reward_std": 0.09774328395724297,
      "reward_std_all": 0.1025282945483923,
      "rewards-max/forensics_sinkhorn_reward": 0.956619530916214,
      "rewards-max/forensics_strict_edge_reward": 0.11384582379832864,
      "rewards-max/forensics_tiou_reward": 0.9195371568202972,
      "rewards/forensics_sinkhorn_reward": 0.5667236149311066,
      "rewards/forensics_strict_edge_reward": 0.010735757881775498,
      "rewards/forensics_tiou_reward": 0.38983054459095,
      "step": 200
    },
    {
      "completion_length": 25.6875,
      "epoch": 0.6175115207373272,
      "grad_norm": 10.908764839172363,
      "kl": 0.310546875,
      "learning_rate": 8.461538461538461e-07,
      "loss": 0.6534,
      "reward": 0.6140633225440979,
      "reward_std": 0.12854713946580887,
      "reward_std_all": 0.15411382168531418,
      "rewards-max/forensics_sinkhorn_reward": 0.9688935577869415,
      "rewards-max/forensics_strict_edge_reward": 0.37411409616470337,
      "rewards-max/forensics_tiou_reward": 0.9477225244045258,
      "rewards/forensics_sinkhorn_reward": 0.480867475271225,
      "rewards/forensics_strict_edge_reward": 0.022325380705296993,
      "rewards/forensics_tiou_reward": 0.29153452813625336,
      "step": 201
    },
    {
      "completion_length": 27.71875,
      "epoch": 0.620583717357911,
      "grad_norm": 7.968956470489502,
      "kl": 0.5224609375,
      "learning_rate": 8.453846153846154e-07,
      "loss": 0.4373,
      "reward": 0.6200647354125977,
      "reward_std": 0.13736089318990707,
      "reward_std_all": 0.15333925932645798,
      "rewards-max/forensics_sinkhorn_reward": 0.9428276717662811,
      "rewards-max/forensics_strict_edge_reward": 0.26304882764816284,
      "rewards-max/forensics_tiou_reward": 0.8530640602111816,
      "rewards/forensics_sinkhorn_reward": 0.46319788694381714,
      "rewards/forensics_strict_edge_reward": 0.015879815444350243,
      "rewards/forensics_tiou_reward": 0.2612884044647217,
      "step": 202
    },
    {
      "completion_length": 25.828125,
      "epoch": 0.6236559139784946,
      "grad_norm": 9.084380149841309,
      "kl": 0.4521484375,
      "learning_rate": 8.446153846153846e-07,
      "loss": 0.6271,
      "reward": 0.6964219510555267,
      "reward_std": 0.07400674559175968,
      "reward_std_all": 0.0918121486902237,
      "rewards-max/forensics_sinkhorn_reward": 0.9149397015571594,
      "rewards-max/forensics_strict_edge_reward": 0.2312263436615467,
      "rewards-max/forensics_tiou_reward": 0.8260609209537506,
      "rewards/forensics_sinkhorn_reward": 0.5798789858818054,
      "rewards/forensics_strict_edge_reward": 0.01665496636996977,
      "rewards/forensics_tiou_reward": 0.29544448107481003,
      "step": 203
    },
    {
      "completion_length": 24.359375,
      "epoch": 0.6267281105990783,
      "grad_norm": 7.614762306213379,
      "kl": 0.466796875,
      "learning_rate": 8.438461538461537e-07,
      "loss": 0.5568,
      "reward": 0.6461220979690552,
      "reward_std": 0.0975923016667366,
      "reward_std_all": 0.12569467723369598,
      "rewards-max/forensics_sinkhorn_reward": 0.9648136496543884,
      "rewards-max/forensics_strict_edge_reward": 0.4114516004920006,
      "rewards-max/forensics_tiou_reward": 0.9330912232398987,
      "rewards/forensics_sinkhorn_reward": 0.5217811465263367,
      "rewards/forensics_strict_edge_reward": 0.02605242235586047,
      "rewards/forensics_tiou_reward": 0.3091806322336197,
      "step": 204
    },
    {
      "completion_length": 24.96875,
      "epoch": 0.6298003072196621,
      "grad_norm": 8.939373970031738,
      "kl": 0.3603515625,
      "learning_rate": 8.43076923076923e-07,
      "loss": -0.065,
      "reward": 0.7157731056213379,
      "reward_std": 0.0613626092672348,
      "reward_std_all": 0.07652536407113075,
      "rewards-max/forensics_sinkhorn_reward": 0.9747723340988159,
      "rewards-max/forensics_strict_edge_reward": 0.508862778544426,
      "rewards-max/forensics_tiou_reward": 0.9610673189163208,
      "rewards/forensics_sinkhorn_reward": 0.6374042332172394,
      "rewards/forensics_strict_edge_reward": 0.05191217362880707,
      "rewards/forensics_tiou_reward": 0.402350515127182,
      "step": 205
    },
    {
      "completion_length": 25.671875,
      "epoch": 0.6328725038402457,
      "grad_norm": 9.182743072509766,
      "kl": 0.482421875,
      "learning_rate": 8.423076923076923e-07,
      "loss": 0.5933,
      "reward": 0.6287597715854645,
      "reward_std": 0.09782204777002335,
      "reward_std_all": 0.13196954503655434,
      "rewards-max/forensics_sinkhorn_reward": 0.9610952436923981,
      "rewards-max/forensics_strict_edge_reward": 0.3200121149420738,
      "rewards-max/forensics_tiou_reward": 0.9306846857070923,
      "rewards/forensics_sinkhorn_reward": 0.4805087000131607,
      "rewards/forensics_strict_edge_reward": 0.032643172424286604,
      "rewards/forensics_tiou_reward": 0.27276745438575745,
      "step": 206
    },
    {
      "completion_length": 27.21875,
      "epoch": 0.6359447004608295,
      "grad_norm": 7.736756801605225,
      "kl": 0.34765625,
      "learning_rate": 8.415384615384616e-07,
      "loss": 0.4578,
      "reward": 0.7498970925807953,
      "reward_std": 0.0471187736839056,
      "reward_std_all": 0.05189898423850536,
      "rewards-max/forensics_sinkhorn_reward": 0.9583328664302826,
      "rewards-max/forensics_strict_edge_reward": 0.24525797739624977,
      "rewards-max/forensics_tiou_reward": 0.9250832200050354,
      "rewards/forensics_sinkhorn_reward": 0.6296637058258057,
      "rewards/forensics_strict_edge_reward": 0.028995505068451166,
      "rewards/forensics_tiou_reward": 0.4071178734302521,
      "step": 207
    },
    {
      "completion_length": 24.515625,
      "epoch": 0.6390168970814132,
      "grad_norm": 6.063519477844238,
      "kl": 0.35546875,
      "learning_rate": 8.407692307692308e-07,
      "loss": 0.3575,
      "reward": 0.7125994563102722,
      "reward_std": 0.04015848971903324,
      "reward_std_all": 0.06643703952431679,
      "rewards-max/forensics_sinkhorn_reward": 0.9653811454772949,
      "rewards-max/forensics_strict_edge_reward": 0.3633866012096405,
      "rewards-max/forensics_tiou_reward": 0.9385569393634796,
      "rewards/forensics_sinkhorn_reward": 0.596338301897049,
      "rewards/forensics_strict_edge_reward": 0.04716229625046253,
      "rewards/forensics_tiou_reward": 0.3949459195137024,
      "step": 208
    },
    {
      "completion_length": 23.875,
      "epoch": 0.642089093701997,
      "grad_norm": 8.236322402954102,
      "kl": 0.4072265625,
      "learning_rate": 8.399999999999999e-07,
      "loss": 0.2055,
      "reward": 0.6665233373641968,
      "reward_std": 0.05948689021170139,
      "reward_std_all": 0.09920913726091385,
      "rewards-max/forensics_sinkhorn_reward": 0.9101299345493317,
      "rewards-max/forensics_strict_edge_reward": 0.05938864510972053,
      "rewards-max/forensics_tiou_reward": 0.8021713197231293,
      "rewards/forensics_sinkhorn_reward": 0.5463602393865585,
      "rewards/forensics_strict_edge_reward": 0.0046187190419004764,
      "rewards/forensics_tiou_reward": 0.24496114999055862,
      "step": 209
    },
    {
      "completion_length": 22.90625,
      "epoch": 0.6451612903225806,
      "grad_norm": 11.387603759765625,
      "kl": 0.359375,
      "learning_rate": 8.392307692307692e-07,
      "loss": 0.0755,
      "reward": 0.7032128274440765,
      "reward_std": 0.06841428950428963,
      "reward_std_all": 0.08583467453718185,
      "rewards-max/forensics_sinkhorn_reward": 0.8544276356697083,
      "rewards-max/forensics_strict_edge_reward": 0.3669360987842083,
      "rewards-max/forensics_tiou_reward": 0.7611782550811768,
      "rewards/forensics_sinkhorn_reward": 0.5778055489063263,
      "rewards/forensics_strict_edge_reward": 0.015967912040650845,
      "rewards/forensics_tiou_reward": 0.29295676201581955,
      "step": 210
    },
    {
      "completion_length": 23.75,
      "epoch": 0.6482334869431644,
      "grad_norm": 8.161431312561035,
      "kl": 0.453125,
      "learning_rate": 8.384615384615384e-07,
      "loss": 0.5863,
      "reward": 0.7024866342544556,
      "reward_std": 0.0708143599331379,
      "reward_std_all": 0.08841922134160995,
      "rewards-max/forensics_sinkhorn_reward": 0.969789445400238,
      "rewards-max/forensics_strict_edge_reward": 0.2087343130260706,
      "rewards-max/forensics_tiou_reward": 0.9424546957015991,
      "rewards/forensics_sinkhorn_reward": 0.5904141962528229,
      "rewards/forensics_strict_edge_reward": 0.014504401828162372,
      "rewards/forensics_tiou_reward": 0.3529997915029526,
      "step": 211
    },
    {
      "completion_length": 22.046875,
      "epoch": 0.6513056835637481,
      "grad_norm": 7.243438720703125,
      "kl": 0.328125,
      "learning_rate": 8.376923076923076e-07,
      "loss": 0.2772,
      "reward": 0.686052143573761,
      "reward_std": 0.04289484769105911,
      "reward_std_all": 0.07888818345963955,
      "rewards-max/forensics_sinkhorn_reward": 0.963022917509079,
      "rewards-max/forensics_strict_edge_reward": 0.1883164392784238,
      "rewards-max/forensics_tiou_reward": 0.9137840270996094,
      "rewards/forensics_sinkhorn_reward": 0.5933669209480286,
      "rewards/forensics_strict_edge_reward": 0.013547683076467365,
      "rewards/forensics_tiou_reward": 0.332695409655571,
      "step": 212
    },
    {
      "completion_length": 23.609375,
      "epoch": 0.6543778801843319,
      "grad_norm": 6.67525053024292,
      "kl": 0.3427734375,
      "learning_rate": 8.369230769230769e-07,
      "loss": 0.3975,
      "reward": 0.6920925676822662,
      "reward_std": 0.05329420045018196,
      "reward_std_all": 0.08134852349758148,
      "rewards-max/forensics_sinkhorn_reward": 0.9622220396995544,
      "rewards-max/forensics_strict_edge_reward": 0.18403177805885207,
      "rewards-max/forensics_tiou_reward": 0.9308774471282959,
      "rewards/forensics_sinkhorn_reward": 0.5510134547948837,
      "rewards/forensics_strict_edge_reward": 0.012600152521827113,
      "rewards/forensics_tiou_reward": 0.2960149496793747,
      "step": 213
    },
    {
      "completion_length": 23.28125,
      "epoch": 0.6574500768049155,
      "grad_norm": 8.989510536193848,
      "kl": 0.3154296875,
      "learning_rate": 8.361538461538461e-07,
      "loss": 0.2425,
      "reward": 0.708288848400116,
      "reward_std": 0.08218783512711525,
      "reward_std_all": 0.08788137882947922,
      "rewards-max/forensics_sinkhorn_reward": 0.9576031863689423,
      "rewards-max/forensics_strict_edge_reward": 0.34197375178337097,
      "rewards-max/forensics_tiou_reward": 0.924136221408844,
      "rewards/forensics_sinkhorn_reward": 0.6032068729400635,
      "rewards/forensics_strict_edge_reward": 0.03314104676246643,
      "rewards/forensics_tiou_reward": 0.32305265218019485,
      "step": 214
    },
    {
      "completion_length": 24.140625,
      "epoch": 0.6605222734254992,
      "grad_norm": 7.7050395011901855,
      "kl": 0.37890625,
      "learning_rate": 8.353846153846154e-07,
      "loss": 0.0715,
      "reward": 0.7062815725803375,
      "reward_std": 0.05687963590025902,
      "reward_std_all": 0.07726439833641052,
      "rewards-max/forensics_sinkhorn_reward": 0.9594324827194214,
      "rewards-max/forensics_strict_edge_reward": 0.3273772746324539,
      "rewards-max/forensics_tiou_reward": 0.9310666024684906,
      "rewards/forensics_sinkhorn_reward": 0.6393496990203857,
      "rewards/forensics_strict_edge_reward": 0.017567004077136517,
      "rewards/forensics_tiou_reward": 0.380636602640152,
      "step": 215
    },
    {
      "completion_length": 23.484375,
      "epoch": 0.663594470046083,
      "grad_norm": 7.953162670135498,
      "kl": 0.2822265625,
      "learning_rate": 8.346153846153846e-07,
      "loss": 0.3462,
      "reward": 0.6732445061206818,
      "reward_std": 0.08165746927261353,
      "reward_std_all": 0.1059129387140274,
      "rewards-max/forensics_sinkhorn_reward": 0.9674108326435089,
      "rewards-max/forensics_strict_edge_reward": 0.2154630869626999,
      "rewards-max/forensics_tiou_reward": 0.932295024394989,
      "rewards/forensics_sinkhorn_reward": 0.5765092372894287,
      "rewards/forensics_strict_edge_reward": 0.016559353098273277,
      "rewards/forensics_tiou_reward": 0.34794846177101135,
      "step": 216
    },
    {
      "completion_length": 22.578125,
      "epoch": 0.6666666666666666,
      "grad_norm": 8.0455904006958,
      "kl": 0.314453125,
      "learning_rate": 8.338461538461537e-07,
      "loss": 0.43,
      "reward": 0.7171709537506104,
      "reward_std": 0.03527240827679634,
      "reward_std_all": 0.06170428916811943,
      "rewards-max/forensics_sinkhorn_reward": 0.9242577850818634,
      "rewards-max/forensics_strict_edge_reward": 0.3424911629408598,
      "rewards-max/forensics_tiou_reward": 0.8329095244407654,
      "rewards/forensics_sinkhorn_reward": 0.6189569234848022,
      "rewards/forensics_strict_edge_reward": 0.03291188186267391,
      "rewards/forensics_tiou_reward": 0.33800166845321655,
      "step": 217
    },
    {
      "completion_length": 24.296875,
      "epoch": 0.6697388632872504,
      "grad_norm": 8.815828323364258,
      "kl": 0.4326171875,
      "learning_rate": 8.33076923076923e-07,
      "loss": -0.2873,
      "reward": 0.7396137118339539,
      "reward_std": 0.028176282881759107,
      "reward_std_all": 0.04863859200850129,
      "rewards-max/forensics_sinkhorn_reward": 0.9643817245960236,
      "rewards-max/forensics_strict_edge_reward": 0.2573440223932266,
      "rewards-max/forensics_tiou_reward": 0.9321936070919037,
      "rewards/forensics_sinkhorn_reward": 0.6264246106147766,
      "rewards/forensics_strict_edge_reward": 0.014456937089562416,
      "rewards/forensics_tiou_reward": 0.39453230798244476,
      "step": 218
    },
    {
      "completion_length": 24.46875,
      "epoch": 0.6728110599078341,
      "grad_norm": 6.514251232147217,
      "kl": 0.224609375,
      "learning_rate": 8.323076923076922e-07,
      "loss": 0.3475,
      "reward": 0.737669050693512,
      "reward_std": 0.0412812028080225,
      "reward_std_all": 0.05434403568506241,
      "rewards-max/forensics_sinkhorn_reward": 0.963090181350708,
      "rewards-max/forensics_strict_edge_reward": 0.1922721266746521,
      "rewards-max/forensics_tiou_reward": 0.9233271181583405,
      "rewards/forensics_sinkhorn_reward": 0.6152242720127106,
      "rewards/forensics_strict_edge_reward": 0.013553931843489408,
      "rewards/forensics_tiou_reward": 0.38166800141334534,
      "step": 219
    },
    {
      "completion_length": 26.703125,
      "epoch": 0.6758832565284179,
      "grad_norm": 6.8926310539245605,
      "kl": 0.302734375,
      "learning_rate": 8.315384615384616e-07,
      "loss": 0.6313,
      "reward": 0.6250999867916107,
      "reward_std": 0.12138698995113373,
      "reward_std_all": 0.14836639165878296,
      "rewards-max/forensics_sinkhorn_reward": 0.9474156200885773,
      "rewards-max/forensics_strict_edge_reward": 0.11453412473201752,
      "rewards-max/forensics_tiou_reward": 0.8935335874557495,
      "rewards/forensics_sinkhorn_reward": 0.5221900641918182,
      "rewards/forensics_strict_edge_reward": 0.008053992409259081,
      "rewards/forensics_tiou_reward": 0.2585442289710045,
      "step": 220
    },
    {
      "completion_length": 22.84375,
      "epoch": 0.6789554531490015,
      "grad_norm": 6.527693748474121,
      "kl": 0.27734375,
      "learning_rate": 8.307692307692308e-07,
      "loss": 0.2542,
      "reward": 0.6702221035957336,
      "reward_std": 0.08505402505397797,
      "reward_std_all": 0.10244356468319893,
      "rewards-max/forensics_sinkhorn_reward": 0.9834946393966675,
      "rewards-max/forensics_strict_edge_reward": 0.7958523333072662,
      "rewards-max/forensics_tiou_reward": 0.9862281978130341,
      "rewards/forensics_sinkhorn_reward": 0.5601299703121185,
      "rewards/forensics_strict_edge_reward": 0.05362598970532417,
      "rewards/forensics_tiou_reward": 0.3173784837126732,
      "step": 221
    },
    {
      "completion_length": 24.484375,
      "epoch": 0.6820276497695853,
      "grad_norm": 7.534584999084473,
      "kl": 0.3251953125,
      "learning_rate": 8.299999999999999e-07,
      "loss": 0.303,
      "reward": 0.697195291519165,
      "reward_std": 0.07678860053420067,
      "reward_std_all": 0.0933675542473793,
      "rewards-max/forensics_sinkhorn_reward": 0.9880179464817047,
      "rewards-max/forensics_strict_edge_reward": 0.7165313363075256,
      "rewards-max/forensics_tiou_reward": 0.9973492622375488,
      "rewards/forensics_sinkhorn_reward": 0.5955990552902222,
      "rewards/forensics_strict_edge_reward": 0.04543257039040327,
      "rewards/forensics_tiou_reward": 0.36668190360069275,
      "step": 222
    },
    {
      "completion_length": 22.328125,
      "epoch": 0.685099846390169,
      "grad_norm": 9.980000495910645,
      "kl": 0.322265625,
      "learning_rate": 8.292307692307692e-07,
      "loss": 0.018,
      "reward": 0.7316010892391205,
      "reward_std": 0.0601851511746645,
      "reward_std_all": 0.06567841023206711,
      "rewards-max/forensics_sinkhorn_reward": 0.9684528708457947,
      "rewards-max/forensics_strict_edge_reward": 0.3816703259944916,
      "rewards-max/forensics_tiou_reward": 0.9456493258476257,
      "rewards/forensics_sinkhorn_reward": 0.6208151578903198,
      "rewards/forensics_strict_edge_reward": 0.020710797514766455,
      "rewards/forensics_tiou_reward": 0.35206109285354614,
      "step": 223
    },
    {
      "completion_length": 22.984375,
      "epoch": 0.6881720430107527,
      "grad_norm": 8.061413764953613,
      "kl": 0.388671875,
      "learning_rate": 8.284615384615384e-07,
      "loss": 0.0286,
      "reward": 0.6842373907566071,
      "reward_std": 0.0803629606962204,
      "reward_std_all": 0.09665016457438469,
      "rewards-max/forensics_sinkhorn_reward": 0.9550053477287292,
      "rewards-max/forensics_strict_edge_reward": 0.15298296883702278,
      "rewards-max/forensics_tiou_reward": 0.9162484407424927,
      "rewards/forensics_sinkhorn_reward": 0.5718735754489899,
      "rewards/forensics_strict_edge_reward": 0.011157216504216194,
      "rewards/forensics_tiou_reward": 0.29594045877456665,
      "step": 224
    },
    {
      "completion_length": 23.5625,
      "epoch": 0.6912442396313364,
      "grad_norm": 10.059892654418945,
      "kl": 0.7109375,
      "learning_rate": 8.276923076923077e-07,
      "loss": 0.4343,
      "reward": 0.6874129176139832,
      "reward_std": 0.06729697436094284,
      "reward_std_all": 0.09237036481499672,
      "rewards-max/forensics_sinkhorn_reward": 0.9633693099021912,
      "rewards-max/forensics_strict_edge_reward": 0.2201457917690277,
      "rewards-max/forensics_tiou_reward": 0.9290217757225037,
      "rewards/forensics_sinkhorn_reward": 0.5940181314945221,
      "rewards/forensics_strict_edge_reward": 0.009323072386905551,
      "rewards/forensics_tiou_reward": 0.3270565867424011,
      "step": 225
    },
    {
      "completion_length": 23.359375,
      "epoch": 0.6943164362519201,
      "grad_norm": 8.510982513427734,
      "kl": 0.4248046875,
      "learning_rate": 8.269230769230768e-07,
      "loss": 0.3625,
      "reward": 0.6609015762805939,
      "reward_std": 0.08090079575777054,
      "reward_std_all": 0.10759418830275536,
      "rewards-max/forensics_sinkhorn_reward": 0.9272206127643585,
      "rewards-max/forensics_strict_edge_reward": 0.14129759604111314,
      "rewards-max/forensics_tiou_reward": 0.8815332651138306,
      "rewards/forensics_sinkhorn_reward": 0.4975399374961853,
      "rewards/forensics_strict_edge_reward": 0.011464237817563117,
      "rewards/forensics_tiou_reward": 0.24033711850643158,
      "step": 226
    },
    {
      "completion_length": 23.359375,
      "epoch": 0.6973886328725039,
      "grad_norm": 7.966771125793457,
      "kl": 0.2861328125,
      "learning_rate": 8.261538461538461e-07,
      "loss": 0.2163,
      "reward": 0.7126818299293518,
      "reward_std": 0.07381312735378742,
      "reward_std_all": 0.08425800502300262,
      "rewards-max/forensics_sinkhorn_reward": 0.9657067656517029,
      "rewards-max/forensics_strict_edge_reward": 0.6915990114212036,
      "rewards-max/forensics_tiou_reward": 0.9403234720230103,
      "rewards/forensics_sinkhorn_reward": 0.6091165840625763,
      "rewards/forensics_strict_edge_reward": 0.0514150895178318,
      "rewards/forensics_tiou_reward": 0.35975970327854156,
      "step": 227
    },
    {
      "completion_length": 22.453125,
      "epoch": 0.7004608294930875,
      "grad_norm": 12.27993106842041,
      "kl": 0.92578125,
      "learning_rate": 8.253846153846154e-07,
      "loss": -0.0056,
      "reward": 0.7110608220100403,
      "reward_std": 0.06644877791404724,
      "reward_std_all": 0.07931935042142868,
      "rewards-max/forensics_sinkhorn_reward": 0.9638498425483704,
      "rewards-max/forensics_strict_edge_reward": 0.3975932002067566,
      "rewards-max/forensics_tiou_reward": 0.9389183819293976,
      "rewards/forensics_sinkhorn_reward": 0.6260839700698853,
      "rewards/forensics_strict_edge_reward": 0.06076374463737011,
      "rewards/forensics_tiou_reward": 0.40544986724853516,
      "step": 228
    },
    {
      "completion_length": 21.546875,
      "epoch": 0.7035330261136713,
      "grad_norm": 8.035359382629395,
      "kl": 0.34375,
      "learning_rate": 8.246153846153846e-07,
      "loss": 0.1875,
      "reward": 0.7543755173683167,
      "reward_std": 0.04460816644132137,
      "reward_std_all": 0.048262473195791245,
      "rewards-max/forensics_sinkhorn_reward": 0.9456586539745331,
      "rewards-max/forensics_strict_edge_reward": 0.3348124772310257,
      "rewards-max/forensics_tiou_reward": 0.893714964389801,
      "rewards/forensics_sinkhorn_reward": 0.6822933554649353,
      "rewards/forensics_strict_edge_reward": 0.03260943293571472,
      "rewards/forensics_tiou_reward": 0.3957799822092056,
      "step": 229
    },
    {
      "completion_length": 21.15625,
      "epoch": 0.706605222734255,
      "grad_norm": 11.174281120300293,
      "kl": 0.3759765625,
      "learning_rate": 8.238461538461538e-07,
      "loss": 0.381,
      "reward": 0.657890111207962,
      "reward_std": 0.061003196984529495,
      "reward_std_all": 0.10077249258756638,
      "rewards-max/forensics_sinkhorn_reward": 0.9815672039985657,
      "rewards-max/forensics_strict_edge_reward": 0.14760751277208328,
      "rewards-max/forensics_tiou_reward": 0.9739746749401093,
      "rewards/forensics_sinkhorn_reward": 0.5808684825897217,
      "rewards/forensics_strict_edge_reward": 0.011324063409119844,
      "rewards/forensics_tiou_reward": 0.29530756175518036,
      "step": 230
    },
    {
      "completion_length": 22.96875,
      "epoch": 0.7096774193548387,
      "grad_norm": 11.259305000305176,
      "kl": 0.341796875,
      "learning_rate": 8.23076923076923e-07,
      "loss": -0.0813,
      "reward": 0.7124218344688416,
      "reward_std": 0.09232525154948235,
      "reward_std_all": 0.09252649173140526,
      "rewards-max/forensics_sinkhorn_reward": 0.9352788925170898,
      "rewards-max/forensics_strict_edge_reward": 0.39293370209634304,
      "rewards-max/forensics_tiou_reward": 0.8743256330490112,
      "rewards/forensics_sinkhorn_reward": 0.6143476963043213,
      "rewards/forensics_strict_edge_reward": 0.0339667130028829,
      "rewards/forensics_tiou_reward": 0.36856934428215027,
      "step": 231
    },
    {
      "completion_length": 21.671875,
      "epoch": 0.7127496159754224,
      "grad_norm": 8.780197143554688,
      "kl": 0.298828125,
      "learning_rate": 8.223076923076922e-07,
      "loss": 0.3571,
      "reward": 0.7154284715652466,
      "reward_std": 0.06925622001290321,
      "reward_std_all": 0.07916514202952385,
      "rewards-max/forensics_sinkhorn_reward": 0.9551731944084167,
      "rewards-max/forensics_strict_edge_reward": 0.2276683896780014,
      "rewards-max/forensics_tiou_reward": 0.9117296934127808,
      "rewards/forensics_sinkhorn_reward": 0.604409784078598,
      "rewards/forensics_strict_edge_reward": 0.0181797593832016,
      "rewards/forensics_tiou_reward": 0.3208170235157013,
      "step": 232
    },
    {
      "completion_length": 21.609375,
      "epoch": 0.7158218125960062,
      "grad_norm": 8.921835899353027,
      "kl": 0.29296875,
      "learning_rate": 8.215384615384616e-07,
      "loss": 0.3443,
      "reward": 0.6918611526489258,
      "reward_std": 0.10655640065670013,
      "reward_std_all": 0.11243942379951477,
      "rewards-max/forensics_sinkhorn_reward": 0.9426214694976807,
      "rewards-max/forensics_strict_edge_reward": 0.23797617852687836,
      "rewards-max/forensics_tiou_reward": 0.8808777630329132,
      "rewards/forensics_sinkhorn_reward": 0.6051466166973114,
      "rewards/forensics_strict_edge_reward": 0.01489336509257555,
      "rewards/forensics_tiou_reward": 0.3174060881137848,
      "step": 233
    },
    {
      "completion_length": 21.703125,
      "epoch": 0.7188940092165899,
      "grad_norm": 8.160965919494629,
      "kl": 0.47265625,
      "learning_rate": 8.207692307692308e-07,
      "loss": 0.4237,
      "reward": 0.6731264889240265,
      "reward_std": 0.06311210989952087,
      "reward_std_all": 0.09316546469926834,
      "rewards-max/forensics_sinkhorn_reward": 0.9296498000621796,
      "rewards-max/forensics_strict_edge_reward": 0.13320930674672127,
      "rewards-max/forensics_tiou_reward": 0.8614513576030731,
      "rewards/forensics_sinkhorn_reward": 0.5671036243438721,
      "rewards/forensics_strict_edge_reward": 0.01360535528510809,
      "rewards/forensics_tiou_reward": 0.2871899902820587,
      "step": 234
    },
    {
      "completion_length": 24.5625,
      "epoch": 0.7219662058371735,
      "grad_norm": 7.27317476272583,
      "kl": 0.2919921875,
      "learning_rate": 8.199999999999999e-07,
      "loss": -0.0474,
      "reward": 0.7445876896381378,
      "reward_std": 0.0718984603881836,
      "reward_std_all": 0.0712442398071289,
      "rewards-max/forensics_sinkhorn_reward": 0.9248270988464355,
      "rewards-max/forensics_strict_edge_reward": 0.23884401963451296,
      "rewards-max/forensics_tiou_reward": 0.832058310508728,
      "rewards/forensics_sinkhorn_reward": 0.6307679414749146,
      "rewards/forensics_strict_edge_reward": 0.03281253662072636,
      "rewards/forensics_tiou_reward": 0.33933496475219727,
      "step": 235
    },
    {
      "completion_length": 21.484375,
      "epoch": 0.7250384024577573,
      "grad_norm": 7.49907112121582,
      "kl": 0.3896484375,
      "learning_rate": 8.192307692307692e-07,
      "loss": 0.1658,
      "reward": 0.680495023727417,
      "reward_std": 0.12023656442761421,
      "reward_std_all": 0.12647103145718575,
      "rewards-max/forensics_sinkhorn_reward": 0.9206261932849884,
      "rewards-max/forensics_strict_edge_reward": 0.07720879465341568,
      "rewards-max/forensics_tiou_reward": 0.8161268830299377,
      "rewards/forensics_sinkhorn_reward": 0.5821930766105652,
      "rewards/forensics_strict_edge_reward": 0.004245585761964321,
      "rewards/forensics_tiou_reward": 0.29074352979660034,
      "step": 236
    },
    {
      "completion_length": 25.3125,
      "epoch": 0.728110599078341,
      "grad_norm": 6.756974697113037,
      "kl": 0.326171875,
      "learning_rate": 8.184615384615384e-07,
      "loss": 0.3598,
      "reward": 0.7051407694816589,
      "reward_std": 0.071234580129385,
      "reward_std_all": 0.08372451178729534,
      "rewards-max/forensics_sinkhorn_reward": 0.9073558747768402,
      "rewards-max/forensics_strict_edge_reward": 0.005938309128396213,
      "rewards-max/forensics_tiou_reward": 0.7952146828174591,
      "rewards/forensics_sinkhorn_reward": 0.5642452836036682,
      "rewards/forensics_strict_edge_reward": 0.0005178397404961288,
      "rewards/forensics_tiou_reward": 0.29012851417064667,
      "step": 237
    },
    {
      "completion_length": 22.484375,
      "epoch": 0.7311827956989247,
      "grad_norm": 7.316760540008545,
      "kl": 0.3603515625,
      "learning_rate": 8.176923076923077e-07,
      "loss": 0.5514,
      "reward": 0.7192253172397614,
      "reward_std": 0.06754104234278202,
      "reward_std_all": 0.07470078393816948,
      "rewards-max/forensics_sinkhorn_reward": 0.9525173604488373,
      "rewards-max/forensics_strict_edge_reward": 0.32719099521636963,
      "rewards-max/forensics_tiou_reward": 0.8954846858978271,
      "rewards/forensics_sinkhorn_reward": 0.6030823588371277,
      "rewards/forensics_strict_edge_reward": 0.02210918813943863,
      "rewards/forensics_tiou_reward": 0.3129819482564926,
      "step": 238
    },
    {
      "completion_length": 25.234375,
      "epoch": 0.7342549923195084,
      "grad_norm": 5.940828800201416,
      "kl": 0.3681640625,
      "learning_rate": 8.169230769230769e-07,
      "loss": 0.3478,
      "reward": 0.7448278069496155,
      "reward_std": 0.0619303397834301,
      "reward_std_all": 0.06164921820163727,
      "rewards-max/forensics_sinkhorn_reward": 0.9557332694530487,
      "rewards-max/forensics_strict_edge_reward": 0.42945683002471924,
      "rewards-max/forensics_tiou_reward": 0.9158900082111359,
      "rewards/forensics_sinkhorn_reward": 0.6367000937461853,
      "rewards/forensics_strict_edge_reward": 0.031224369071424007,
      "rewards/forensics_tiou_reward": 0.4544572979211807,
      "step": 239
    },
    {
      "completion_length": 23.96875,
      "epoch": 0.7373271889400922,
      "grad_norm": 7.728750705718994,
      "kl": 0.3544921875,
      "learning_rate": 8.16153846153846e-07,
      "loss": 0.3893,
      "reward": 0.6935971677303314,
      "reward_std": 0.1070343405008316,
      "reward_std_all": 0.10863891988992691,
      "rewards-max/forensics_sinkhorn_reward": 0.9805014729499817,
      "rewards-max/forensics_strict_edge_reward": 0.2838853932917118,
      "rewards-max/forensics_tiou_reward": 0.9781107306480408,
      "rewards/forensics_sinkhorn_reward": 0.6001266837120056,
      "rewards/forensics_strict_edge_reward": 0.014858916401863098,
      "rewards/forensics_tiou_reward": 0.3784617781639099,
      "step": 240
    },
    {
      "completion_length": 26.46875,
      "epoch": 0.7403993855606759,
      "grad_norm": 6.361184120178223,
      "kl": 0.3310546875,
      "learning_rate": 8.153846153846154e-07,
      "loss": 0.4768,
      "reward": 0.7318054437637329,
      "reward_std": 0.03615618031471968,
      "reward_std_all": 0.05411265231668949,
      "rewards-max/forensics_sinkhorn_reward": 0.9189267456531525,
      "rewards-max/forensics_strict_edge_reward": 0.10965754091739655,
      "rewards-max/forensics_tiou_reward": 0.8206874132156372,
      "rewards/forensics_sinkhorn_reward": 0.59328293800354,
      "rewards/forensics_strict_edge_reward": 0.008262506074970588,
      "rewards/forensics_tiou_reward": 0.34976665675640106,
      "step": 241
    },
    {
      "completion_length": 22.703125,
      "epoch": 0.7434715821812596,
      "grad_norm": 8.557918548583984,
      "kl": 0.345703125,
      "learning_rate": 8.146153846153846e-07,
      "loss": 0.3086,
      "reward": 0.6948961913585663,
      "reward_std": 0.0632150899618864,
      "reward_std_all": 0.08393166214227676,
      "rewards-max/forensics_sinkhorn_reward": 0.9509760141372681,
      "rewards-max/forensics_strict_edge_reward": 0.3576568216085434,
      "rewards-max/forensics_tiou_reward": 0.9117781221866608,
      "rewards/forensics_sinkhorn_reward": 0.5830361843109131,
      "rewards/forensics_strict_edge_reward": 0.019447398371994495,
      "rewards/forensics_tiou_reward": 0.33592621982097626,
      "step": 242
    },
    {
      "completion_length": 22.703125,
      "epoch": 0.7465437788018433,
      "grad_norm": 9.39486026763916,
      "kl": 0.330078125,
      "learning_rate": 8.138461538461538e-07,
      "loss": 0.1231,
      "reward": 0.7496029734611511,
      "reward_std": 0.06003832630813122,
      "reward_std_all": 0.059797147288918495,
      "rewards-max/forensics_sinkhorn_reward": 0.9579807221889496,
      "rewards-max/forensics_strict_edge_reward": 0.417997270822525,
      "rewards-max/forensics_tiou_reward": 0.924571692943573,
      "rewards/forensics_sinkhorn_reward": 0.6723122596740723,
      "rewards/forensics_strict_edge_reward": 0.02921285293996334,
      "rewards/forensics_tiou_reward": 0.4800693690776825,
      "step": 243
    },
    {
      "completion_length": 26.296875,
      "epoch": 0.7496159754224271,
      "grad_norm": 8.697653770446777,
      "kl": 0.3701171875,
      "learning_rate": 8.13076923076923e-07,
      "loss": 0.4428,
      "reward": 0.7433889508247375,
      "reward_std": 0.06856166571378708,
      "reward_std_all": 0.06763065606355667,
      "rewards-max/forensics_sinkhorn_reward": 0.9763828814029694,
      "rewards-max/forensics_strict_edge_reward": 0.5599738955497742,
      "rewards-max/forensics_tiou_reward": 0.9679414629936218,
      "rewards/forensics_sinkhorn_reward": 0.6147252023220062,
      "rewards/forensics_strict_edge_reward": 0.05374090559780598,
      "rewards/forensics_tiou_reward": 0.42036131024360657,
      "step": 244
    },
    {
      "completion_length": 22.640625,
      "epoch": 0.7526881720430108,
      "grad_norm": 6.710413455963135,
      "kl": 0.3447265625,
      "learning_rate": 8.123076923076922e-07,
      "loss": 0.282,
      "reward": 0.6853111386299133,
      "reward_std": 0.06190278008580208,
      "reward_std_all": 0.08869589492678642,
      "rewards-max/forensics_sinkhorn_reward": 0.9715130627155304,
      "rewards-max/forensics_strict_edge_reward": 0.34077148139476776,
      "rewards-max/forensics_tiou_reward": 0.9560723900794983,
      "rewards/forensics_sinkhorn_reward": 0.5817333459854126,
      "rewards/forensics_strict_edge_reward": 0.025304964743554592,
      "rewards/forensics_tiou_reward": 0.32167938351631165,
      "step": 245
    },
    {
      "completion_length": 23.328125,
      "epoch": 0.7557603686635944,
      "grad_norm": 9.143136024475098,
      "kl": 0.478515625,
      "learning_rate": 8.115384615384615e-07,
      "loss": 0.2164,
      "reward": 0.7196722328662872,
      "reward_std": 0.03904431685805321,
      "reward_std_all": 0.06029832921922207,
      "rewards-max/forensics_sinkhorn_reward": 0.9857493042945862,
      "rewards-max/forensics_strict_edge_reward": 0.606867641210556,
      "rewards-max/forensics_tiou_reward": 0.990781158208847,
      "rewards/forensics_sinkhorn_reward": 0.6225064992904663,
      "rewards/forensics_strict_edge_reward": 0.034372820518910885,
      "rewards/forensics_tiou_reward": 0.39030346274375916,
      "step": 246
    },
    {
      "completion_length": 23.4375,
      "epoch": 0.7588325652841782,
      "grad_norm": 9.511286735534668,
      "kl": 0.3779296875,
      "learning_rate": 8.107692307692308e-07,
      "loss": -0.0644,
      "reward": 0.7232284545898438,
      "reward_std": 0.05182734690606594,
      "reward_std_all": 0.06324328109622002,
      "rewards-max/forensics_sinkhorn_reward": 0.9710229337215424,
      "rewards-max/forensics_strict_edge_reward": 0.299286313354969,
      "rewards-max/forensics_tiou_reward": 0.949843168258667,
      "rewards/forensics_sinkhorn_reward": 0.618329793214798,
      "rewards/forensics_strict_edge_reward": 0.03495599050074816,
      "rewards/forensics_tiou_reward": 0.367155522108078,
      "step": 247
    },
    {
      "completion_length": 24.65625,
      "epoch": 0.7619047619047619,
      "grad_norm": 7.914295673370361,
      "kl": 0.4404296875,
      "learning_rate": 8.1e-07,
      "loss": 0.3034,
      "reward": 0.7399217188358307,
      "reward_std": 0.06263907998800278,
      "reward_std_all": 0.06589993461966515,
      "rewards-max/forensics_sinkhorn_reward": 0.9697797000408173,
      "rewards-max/forensics_strict_edge_reward": 0.513702392578125,
      "rewards-max/forensics_tiou_reward": 0.9522821605205536,
      "rewards/forensics_sinkhorn_reward": 0.6535695791244507,
      "rewards/forensics_strict_edge_reward": 0.0480746915563941,
      "rewards/forensics_tiou_reward": 0.41158586740493774,
      "step": 248
    },
    {
      "completion_length": 24.75,
      "epoch": 0.7649769585253456,
      "grad_norm": 11.01193618774414,
      "kl": 0.95703125,
      "learning_rate": 8.092307692307692e-07,
      "loss": 0.4809,
      "reward": 0.7044986188411713,
      "reward_std": 0.06056871451437473,
      "reward_std_all": 0.07835325226187706,
      "rewards-max/forensics_sinkhorn_reward": 0.965277224779129,
      "rewards-max/forensics_strict_edge_reward": 0.33831171528436244,
      "rewards-max/forensics_tiou_reward": 0.9440126121044159,
      "rewards/forensics_sinkhorn_reward": 0.5753662586212158,
      "rewards/forensics_strict_edge_reward": 0.0280754397099372,
      "rewards/forensics_tiou_reward": 0.33793212473392487,
      "step": 249
    },
    {
      "completion_length": 23.734375,
      "epoch": 0.7680491551459293,
      "grad_norm": 7.950465679168701,
      "kl": 0.3896484375,
      "learning_rate": 8.084615384615384e-07,
      "loss": -0.0295,
      "reward": 0.7385547459125519,
      "reward_std": 0.041319227777421474,
      "reward_std_all": 0.05194849893450737,
      "rewards-max/forensics_sinkhorn_reward": 0.9410113394260406,
      "rewards-max/forensics_strict_edge_reward": 0.409512179743615,
      "rewards-max/forensics_tiou_reward": 0.8560982346534729,
      "rewards/forensics_sinkhorn_reward": 0.6599619686603546,
      "rewards/forensics_strict_edge_reward": 0.04970188775223505,
      "rewards/forensics_tiou_reward": 0.3409934341907501,
      "step": 250
    },
    {
      "completion_length": 25.4375,
      "epoch": 0.7711213517665131,
      "grad_norm": 6.817476272583008,
      "kl": 0.400390625,
      "learning_rate": 8.076923076923077e-07,
      "loss": 0.1886,
      "reward": 0.6687543988227844,
      "reward_std": 0.08450718969106674,
      "reward_std_all": 0.1101539172232151,
      "rewards-max/forensics_sinkhorn_reward": 0.9440361559391022,
      "rewards-max/forensics_strict_edge_reward": 0.1733567714691162,
      "rewards-max/forensics_tiou_reward": 0.881979763507843,
      "rewards/forensics_sinkhorn_reward": 0.575139582157135,
      "rewards/forensics_strict_edge_reward": 0.010542726144194603,
      "rewards/forensics_tiou_reward": 0.33806586265563965,
      "step": 251
    },
    {
      "completion_length": 23.21875,
      "epoch": 0.7741935483870968,
      "grad_norm": 8.304311752319336,
      "kl": 0.396484375,
      "learning_rate": 8.069230769230769e-07,
      "loss": 0.2899,
      "reward": 0.6521530747413635,
      "reward_std": 0.10181523486971855,
      "reward_std_all": 0.12464074045419693,
      "rewards-max/forensics_sinkhorn_reward": 0.9577460587024689,
      "rewards-max/forensics_strict_edge_reward": 0.2865445762872696,
      "rewards-max/forensics_tiou_reward": 0.9210533499717712,
      "rewards/forensics_sinkhorn_reward": 0.5642094612121582,
      "rewards/forensics_strict_edge_reward": 0.01467727916315198,
      "rewards/forensics_tiou_reward": 0.2989942729473114,
      "step": 252
    },
    {
      "completion_length": 24.96875,
      "epoch": 0.7772657450076805,
      "grad_norm": 9.790471076965332,
      "kl": 0.3369140625,
      "learning_rate": 8.06153846153846e-07,
      "loss": 0.184,
      "reward": 0.7283967137336731,
      "reward_std": 0.05967786908149719,
      "reward_std_all": 0.06754737719893456,
      "rewards-max/forensics_sinkhorn_reward": 0.9337577819824219,
      "rewards-max/forensics_strict_edge_reward": 0.39255425706505775,
      "rewards-max/forensics_tiou_reward": 0.8573696315288544,
      "rewards/forensics_sinkhorn_reward": 0.6125377416610718,
      "rewards/forensics_strict_edge_reward": 0.02145132003352046,
      "rewards/forensics_tiou_reward": 0.39359961450099945,
      "step": 253
    },
    {
      "completion_length": 25.625,
      "epoch": 0.7803379416282642,
      "grad_norm": 8.813608169555664,
      "kl": 0.41015625,
      "learning_rate": 8.053846153846154e-07,
      "loss": 0.0299,
      "reward": 0.6722617447376251,
      "reward_std": 0.05718267895281315,
      "reward_std_all": 0.08883186802268028,
      "rewards-max/forensics_sinkhorn_reward": 0.9808870553970337,
      "rewards-max/forensics_strict_edge_reward": 0.5608736276626587,
      "rewards-max/forensics_tiou_reward": 0.9766592383384705,
      "rewards/forensics_sinkhorn_reward": 0.5379096865653992,
      "rewards/forensics_strict_edge_reward": 0.031817011535167694,
      "rewards/forensics_tiou_reward": 0.27529700100421906,
      "step": 254
    },
    {
      "completion_length": 24.390625,
      "epoch": 0.783410138248848,
      "grad_norm": 10.969520568847656,
      "kl": 0.38671875,
      "learning_rate": 8.046153846153846e-07,
      "loss": 0.172,
      "reward": 0.7389476001262665,
      "reward_std": 0.060856595635414124,
      "reward_std_all": 0.06332231312990189,
      "rewards-max/forensics_sinkhorn_reward": 0.9386171996593475,
      "rewards-max/forensics_strict_edge_reward": 0.43838582816533744,
      "rewards-max/forensics_tiou_reward": 0.8601987659931183,
      "rewards/forensics_sinkhorn_reward": 0.6191352009773254,
      "rewards/forensics_strict_edge_reward": 0.019078088575042784,
      "rewards/forensics_tiou_reward": 0.319888636469841,
      "step": 255
    },
    {
      "completion_length": 25.296875,
      "epoch": 0.7864823348694316,
      "grad_norm": 9.82104778289795,
      "kl": 0.35546875,
      "learning_rate": 8.038461538461538e-07,
      "loss": 0.2317,
      "reward": 0.7332598865032196,
      "reward_std": 0.07259577140212059,
      "reward_std_all": 0.07400189340114594,
      "rewards-max/forensics_sinkhorn_reward": 0.9057166874408722,
      "rewards-max/forensics_strict_edge_reward": 0.18729293905198574,
      "rewards-max/forensics_tiou_reward": 0.7920890748500824,
      "rewards/forensics_sinkhorn_reward": 0.6229850649833679,
      "rewards/forensics_strict_edge_reward": 0.012430275091901422,
      "rewards/forensics_tiou_reward": 0.32503150403499603,
      "step": 256
    },
    {
      "completion_length": 24.640625,
      "epoch": 0.7895545314900153,
      "grad_norm": 9.230565071105957,
      "kl": 0.4482421875,
      "learning_rate": 8.03076923076923e-07,
      "loss": 0.2455,
      "reward": 0.6714032590389252,
      "reward_std": 0.04251200798898935,
      "reward_std_all": 0.0850452184677124,
      "rewards-max/forensics_sinkhorn_reward": 0.9497411847114563,
      "rewards-max/forensics_strict_edge_reward": 0.467515729367733,
      "rewards-max/forensics_tiou_reward": 0.8965604901313782,
      "rewards/forensics_sinkhorn_reward": 0.5588342994451523,
      "rewards/forensics_strict_edge_reward": 0.02751230075955391,
      "rewards/forensics_tiou_reward": 0.37027789652347565,
      "step": 257
    },
    {
      "completion_length": 25.265625,
      "epoch": 0.7926267281105991,
      "grad_norm": 8.280991554260254,
      "kl": 0.287109375,
      "learning_rate": 8.023076923076922e-07,
      "loss": 0.353,
      "reward": 0.6624694764614105,
      "reward_std": 0.057279326021671295,
      "reward_std_all": 0.09367130696773529,
      "rewards-max/forensics_sinkhorn_reward": 0.9312179088592529,
      "rewards-max/forensics_strict_edge_reward": 0.2312060222029686,
      "rewards-max/forensics_tiou_reward": 0.8672734498977661,
      "rewards/forensics_sinkhorn_reward": 0.5267527103424072,
      "rewards/forensics_strict_edge_reward": 0.01026166696101427,
      "rewards/forensics_tiou_reward": 0.27973487973213196,
      "step": 258
    },
    {
      "completion_length": 28.859375,
      "epoch": 0.7956989247311828,
      "grad_norm": 9.579252243041992,
      "kl": 0.66015625,
      "learning_rate": 8.015384615384615e-07,
      "loss": 0.4729,
      "reward": 0.6796327829360962,
      "reward_std": 0.06178174167871475,
      "reward_std_all": 0.08986666798591614,
      "rewards-max/forensics_sinkhorn_reward": 0.9761010408401489,
      "rewards-max/forensics_strict_edge_reward": 0.4683289974927902,
      "rewards-max/forensics_tiou_reward": 0.9630827307701111,
      "rewards/forensics_sinkhorn_reward": 0.5735615789890289,
      "rewards/forensics_strict_edge_reward": 0.05225102975964546,
      "rewards/forensics_tiou_reward": 0.36918146908283234,
      "step": 259
    },
    {
      "completion_length": 24.375,
      "epoch": 0.7987711213517665,
      "grad_norm": 8.795024871826172,
      "kl": 0.35546875,
      "learning_rate": 8.007692307692308e-07,
      "loss": 0.3572,
      "reward": 0.7216277718544006,
      "reward_std": 0.033253773115575314,
      "reward_std_all": 0.05543544515967369,
      "rewards-max/forensics_sinkhorn_reward": 0.9253547191619873,
      "rewards-max/forensics_strict_edge_reward": 0.16070556640625,
      "rewards-max/forensics_tiou_reward": 0.8424284160137177,
      "rewards/forensics_sinkhorn_reward": 0.5891128480434418,
      "rewards/forensics_strict_edge_reward": 0.014676196267828345,
      "rewards/forensics_tiou_reward": 0.32889750599861145,
      "step": 260
    },
    {
      "completion_length": 24.4375,
      "epoch": 0.8018433179723502,
      "grad_norm": 8.286523818969727,
      "kl": 0.4921875,
      "learning_rate": 8e-07,
      "loss": 0.2047,
      "reward": 0.7320396304130554,
      "reward_std": 0.05099436920136213,
      "reward_std_all": 0.061250269412994385,
      "rewards-max/forensics_sinkhorn_reward": 0.9606360197067261,
      "rewards-max/forensics_strict_edge_reward": 0.4289688169956207,
      "rewards-max/forensics_tiou_reward": 0.9287247061729431,
      "rewards/forensics_sinkhorn_reward": 0.6060576140880585,
      "rewards/forensics_strict_edge_reward": 0.036759150214493275,
      "rewards/forensics_tiou_reward": 0.39400848746299744,
      "step": 261
    },
    {
      "completion_length": 21.46875,
      "epoch": 0.804915514592934,
      "grad_norm": 10.310356140136719,
      "kl": 0.5400390625,
      "learning_rate": 7.992307692307692e-07,
      "loss": 0.079,
      "reward": 0.7547568678855896,
      "reward_std": 0.04427293688058853,
      "reward_std_all": 0.04790765047073364,
      "rewards-max/forensics_sinkhorn_reward": 0.8956069052219391,
      "rewards-max/forensics_strict_edge_reward": 0.27208730578422546,
      "rewards-max/forensics_tiou_reward": 0.7803099751472473,
      "rewards/forensics_sinkhorn_reward": 0.6604096591472626,
      "rewards/forensics_strict_edge_reward": 0.02302060555666685,
      "rewards/forensics_tiou_reward": 0.39111684262752533,
      "step": 262
    },
    {
      "completion_length": 23.453125,
      "epoch": 0.8079877112135176,
      "grad_norm": 7.975942134857178,
      "kl": 0.345703125,
      "learning_rate": 7.984615384615384e-07,
      "loss": 0.147,
      "reward": 0.7202188968658447,
      "reward_std": 0.02960167173296213,
      "reward_std_all": 0.057332348078489304,
      "rewards-max/forensics_sinkhorn_reward": 0.8952053487300873,
      "rewards-max/forensics_strict_edge_reward": 0.16011690466439177,
      "rewards-max/forensics_tiou_reward": 0.7582042217254639,
      "rewards/forensics_sinkhorn_reward": 0.6017478406429291,
      "rewards/forensics_strict_edge_reward": 0.015532426314166514,
      "rewards/forensics_tiou_reward": 0.3297036439180374,
      "step": 263
    },
    {
      "completion_length": 26.34375,
      "epoch": 0.8110599078341014,
      "grad_norm": 7.059814929962158,
      "kl": 0.3798828125,
      "learning_rate": 7.976923076923077e-07,
      "loss": 0.323,
      "reward": 0.7182132601737976,
      "reward_std": 0.07916238903999329,
      "reward_std_all": 0.08406002447009087,
      "rewards-max/forensics_sinkhorn_reward": 0.9463907480239868,
      "rewards-max/forensics_strict_edge_reward": 0.42226019501686096,
      "rewards-max/forensics_tiou_reward": 0.8982571959495544,
      "rewards/forensics_sinkhorn_reward": 0.5915538668632507,
      "rewards/forensics_strict_edge_reward": 0.03178808465600014,
      "rewards/forensics_tiou_reward": 0.3790408819913864,
      "step": 264
    },
    {
      "completion_length": 25.125,
      "epoch": 0.8141321044546851,
      "grad_norm": 7.737633228302002,
      "kl": 0.46875,
      "learning_rate": 7.969230769230769e-07,
      "loss": 0.4999,
      "reward": 0.7018680274486542,
      "reward_std": 0.053376579657197,
      "reward_std_all": 0.07184142246842384,
      "rewards-max/forensics_sinkhorn_reward": 0.7783827483654022,
      "rewards-max/forensics_strict_edge_reward": 0.009482382148235047,
      "rewards-max/forensics_tiou_reward": 0.5788239240646362,
      "rewards/forensics_sinkhorn_reward": 0.4946082532405853,
      "rewards/forensics_strict_edge_reward": 0.0003976583129770006,
      "rewards/forensics_tiou_reward": 0.2411845177412033,
      "step": 265
    },
    {
      "completion_length": 24.796875,
      "epoch": 0.8172043010752689,
      "grad_norm": 9.120177268981934,
      "kl": 0.4921875,
      "learning_rate": 7.96153846153846e-07,
      "loss": 0.4915,
      "reward": 0.6738018691539764,
      "reward_std": 0.05087307933717966,
      "reward_std_all": 0.08732916414737701,
      "rewards-max/forensics_sinkhorn_reward": 0.9807698726654053,
      "rewards-max/forensics_strict_edge_reward": 0.75,
      "rewards-max/forensics_tiou_reward": 0.9765856862068176,
      "rewards/forensics_sinkhorn_reward": 0.5595385134220123,
      "rewards/forensics_strict_edge_reward": 0.06804482638835907,
      "rewards/forensics_tiou_reward": 0.3200513869524002,
      "step": 266
    },
    {
      "completion_length": 29.5625,
      "epoch": 0.8202764976958525,
      "grad_norm": 10.427560806274414,
      "kl": 0.349609375,
      "learning_rate": 7.953846153846154e-07,
      "loss": 0.5582,
      "reward": 0.6696697175502777,
      "reward_std": 0.07606150209903717,
      "reward_std_all": 0.0989280715584755,
      "rewards-max/forensics_sinkhorn_reward": 0.860556036233902,
      "rewards-max/forensics_strict_edge_reward": 0.25018105253229805,
      "rewards-max/forensics_tiou_reward": 0.7182471454143524,
      "rewards/forensics_sinkhorn_reward": 0.5002364367246628,
      "rewards/forensics_strict_edge_reward": 0.020097043507924184,
      "rewards/forensics_tiou_reward": 0.2519102320075035,
      "step": 267
    },
    {
      "completion_length": 24.21875,
      "epoch": 0.8233486943164362,
      "grad_norm": 7.501490592956543,
      "kl": 0.3583984375,
      "learning_rate": 7.946153846153846e-07,
      "loss": 0.1335,
      "reward": 0.6765473484992981,
      "reward_std": 0.06199915334582329,
      "reward_std_all": 0.0914275236427784,
      "rewards-max/forensics_sinkhorn_reward": 0.8569249212741852,
      "rewards-max/forensics_strict_edge_reward": 0.3093775361776352,
      "rewards-max/forensics_tiou_reward": 0.7245772480964661,
      "rewards/forensics_sinkhorn_reward": 0.5493283867835999,
      "rewards/forensics_strict_edge_reward": 0.0137975187972188,
      "rewards/forensics_tiou_reward": 0.2771760821342468,
      "step": 268
    },
    {
      "completion_length": 24.640625,
      "epoch": 0.82642089093702,
      "grad_norm": 12.934887886047363,
      "kl": 0.2548828125,
      "learning_rate": 7.938461538461538e-07,
      "loss": 0.0868,
      "reward": 0.7034363448619843,
      "reward_std": 0.09136184677481651,
      "reward_std_all": 0.09796323627233505,
      "rewards-max/forensics_sinkhorn_reward": 0.979043573141098,
      "rewards-max/forensics_strict_edge_reward": 0.2919655442237854,
      "rewards-max/forensics_tiou_reward": 0.9707396328449249,
      "rewards/forensics_sinkhorn_reward": 0.6170667409896851,
      "rewards/forensics_strict_edge_reward": 0.023290298879146576,
      "rewards/forensics_tiou_reward": 0.39783522486686707,
      "step": 269
    },
    {
      "completion_length": 27.5,
      "epoch": 0.8294930875576036,
      "grad_norm": 7.237906455993652,
      "kl": 0.275390625,
      "learning_rate": 7.930769230769231e-07,
      "loss": 0.5369,
      "reward": 0.7052353024482727,
      "reward_std": 0.06462963111698627,
      "reward_std_all": 0.08114335127174854,
      "rewards-max/forensics_sinkhorn_reward": 0.977886974811554,
      "rewards-max/forensics_strict_edge_reward": 0.32959966361522675,
      "rewards-max/forensics_tiou_reward": 0.9718769192695618,
      "rewards/forensics_sinkhorn_reward": 0.5646970868110657,
      "rewards/forensics_strict_edge_reward": 0.022494258359074593,
      "rewards/forensics_tiou_reward": 0.40659552812576294,
      "step": 270
    },
    {
      "completion_length": 26.28125,
      "epoch": 0.8325652841781874,
      "grad_norm": 7.054589748382568,
      "kl": 0.3193359375,
      "learning_rate": 7.923076923076922e-07,
      "loss": 0.1949,
      "reward": 0.7097110152244568,
      "reward_std": 0.0607534721493721,
      "reward_std_all": 0.07767027616500854,
      "rewards-max/forensics_sinkhorn_reward": 0.8574060797691345,
      "rewards-max/forensics_strict_edge_reward": 0.0612509156599117,
      "rewards-max/forensics_tiou_reward": 0.7047349214553833,
      "rewards/forensics_sinkhorn_reward": 0.5805114209651947,
      "rewards/forensics_strict_edge_reward": 0.002778939715199158,
      "rewards/forensics_tiou_reward": 0.29532133787870407,
      "step": 271
    },
    {
      "completion_length": 24.390625,
      "epoch": 0.8356374807987711,
      "grad_norm": 13.50279712677002,
      "kl": 0.4287109375,
      "learning_rate": 7.915384615384615e-07,
      "loss": -0.0054,
      "reward": 0.7829483151435852,
      "reward_std": 0.03352497145533562,
      "reward_std_all": 0.03208388667553663,
      "rewards-max/forensics_sinkhorn_reward": 0.9801077544689178,
      "rewards-max/forensics_strict_edge_reward": 0.6342070251703262,
      "rewards-max/forensics_tiou_reward": 0.9717110991477966,
      "rewards/forensics_sinkhorn_reward": 0.6891776621341705,
      "rewards/forensics_strict_edge_reward": 0.08974925242364407,
      "rewards/forensics_tiou_reward": 0.5153278708457947,
      "step": 272
    },
    {
      "completion_length": 24.71875,
      "epoch": 0.8387096774193549,
      "grad_norm": 8.535866737365723,
      "kl": 0.50048828125,
      "learning_rate": 7.907692307692307e-07,
      "loss": 0.1912,
      "reward": 0.6979793012142181,
      "reward_std": 0.1004759818315506,
      "reward_std_all": 0.10399926733225584,
      "rewards-max/forensics_sinkhorn_reward": 0.971331387758255,
      "rewards-max/forensics_strict_edge_reward": 0.6158150136470795,
      "rewards-max/forensics_tiou_reward": 0.9542983770370483,
      "rewards/forensics_sinkhorn_reward": 0.6143188774585724,
      "rewards/forensics_strict_edge_reward": 0.04842657409608364,
      "rewards/forensics_tiou_reward": 0.4362386018037796,
      "step": 273
    },
    {
      "completion_length": 24.265625,
      "epoch": 0.8417818740399385,
      "grad_norm": 7.052660942077637,
      "kl": 0.5439453125,
      "learning_rate": 7.9e-07,
      "loss": 0.2656,
      "reward": 0.6600248217582703,
      "reward_std": 0.06429086625576019,
      "reward_std_all": 0.10422755032777786,
      "rewards-max/forensics_sinkhorn_reward": 0.9654925465583801,
      "rewards-max/forensics_strict_edge_reward": 0.2674800604581833,
      "rewards-max/forensics_tiou_reward": 0.9395022094249725,
      "rewards/forensics_sinkhorn_reward": 0.5793818235397339,
      "rewards/forensics_strict_edge_reward": 0.030216801911592484,
      "rewards/forensics_tiou_reward": 0.3540710508823395,
      "step": 274
    },
    {
      "completion_length": 25.1875,
      "epoch": 0.8448540706605223,
      "grad_norm": 7.237254619598389,
      "kl": 0.4384765625,
      "learning_rate": 7.892307692307692e-07,
      "loss": 0.1934,
      "reward": 0.7137223184108734,
      "reward_std": 0.06899527460336685,
      "reward_std_all": 0.07856925949454308,
      "rewards-max/forensics_sinkhorn_reward": 0.9662214815616608,
      "rewards-max/forensics_strict_edge_reward": 0.32628850638866425,
      "rewards-max/forensics_tiou_reward": 0.9432494938373566,
      "rewards/forensics_sinkhorn_reward": 0.6351147890090942,
      "rewards/forensics_strict_edge_reward": 0.030260776169598103,
      "rewards/forensics_tiou_reward": 0.39678002893924713,
      "step": 275
    },
    {
      "completion_length": 23.84375,
      "epoch": 0.847926267281106,
      "grad_norm": 6.730179786682129,
      "kl": 0.3330078125,
      "learning_rate": 7.884615384615384e-07,
      "loss": 0.1968,
      "reward": 0.7475717067718506,
      "reward_std": 0.07183906063437462,
      "reward_std_all": 0.06770635396242142,
      "rewards-max/forensics_sinkhorn_reward": 0.9471864700317383,
      "rewards-max/forensics_strict_edge_reward": 0.15774135291576385,
      "rewards-max/forensics_tiou_reward": 0.9032393991947174,
      "rewards/forensics_sinkhorn_reward": 0.6816554665565491,
      "rewards/forensics_strict_edge_reward": 0.010844529140740633,
      "rewards/forensics_tiou_reward": 0.42178402841091156,
      "step": 276
    },
    {
      "completion_length": 25.546875,
      "epoch": 0.8509984639016898,
      "grad_norm": 6.681519985198975,
      "kl": 0.3828125,
      "learning_rate": 7.876923076923077e-07,
      "loss": 0.2916,
      "reward": 0.7409079074859619,
      "reward_std": 0.03780953772366047,
      "reward_std_all": 0.051742393523454666,
      "rewards-max/forensics_sinkhorn_reward": 0.9609402418136597,
      "rewards-max/forensics_strict_edge_reward": 0.4723781943321228,
      "rewards-max/forensics_tiou_reward": 0.9276837110519409,
      "rewards/forensics_sinkhorn_reward": 0.6432891488075256,
      "rewards/forensics_strict_edge_reward": 0.04593455698341131,
      "rewards/forensics_tiou_reward": 0.38698606193065643,
      "step": 277
    },
    {
      "completion_length": 27.125,
      "epoch": 0.8540706605222734,
      "grad_norm": 7.352272987365723,
      "kl": 0.2958984375,
      "learning_rate": 7.869230769230769e-07,
      "loss": 0.3617,
      "reward": 0.6865908205509186,
      "reward_std": 0.0692290123552084,
      "reward_std_all": 0.08765288069844246,
      "rewards-max/forensics_sinkhorn_reward": 0.949640691280365,
      "rewards-max/forensics_strict_edge_reward": 0.2290111407637596,
      "rewards-max/forensics_tiou_reward": 0.8871223032474518,
      "rewards/forensics_sinkhorn_reward": 0.5307718515396118,
      "rewards/forensics_strict_edge_reward": 0.01907912315800786,
      "rewards/forensics_tiou_reward": 0.3009094148874283,
      "step": 278
    },
    {
      "completion_length": 24.0625,
      "epoch": 0.8571428571428571,
      "grad_norm": 9.464210510253906,
      "kl": 0.4111328125,
      "learning_rate": 7.861538461538461e-07,
      "loss": 0.1495,
      "reward": 0.694323867559433,
      "reward_std": 0.0772484689950943,
      "reward_std_all": 0.09134603664278984,
      "rewards-max/forensics_sinkhorn_reward": 0.9688431620597839,
      "rewards-max/forensics_strict_edge_reward": 0.3355042338371277,
      "rewards-max/forensics_tiou_reward": 0.9461092352867126,
      "rewards/forensics_sinkhorn_reward": 0.5955770015716553,
      "rewards/forensics_strict_edge_reward": 0.02028611581772566,
      "rewards/forensics_tiou_reward": 0.35853584110736847,
      "step": 279
    },
    {
      "completion_length": 26.015625,
      "epoch": 0.8602150537634409,
      "grad_norm": 8.903818130493164,
      "kl": 0.3515625,
      "learning_rate": 7.853846153846153e-07,
      "loss": 0.1086,
      "reward": 0.7370942831039429,
      "reward_std": 0.06330000143498182,
      "reward_std_all": 0.06540916673839092,
      "rewards-max/forensics_sinkhorn_reward": 0.9701685905456543,
      "rewards-max/forensics_strict_edge_reward": 0.4613439440727234,
      "rewards-max/forensics_tiou_reward": 0.9501999318599701,
      "rewards/forensics_sinkhorn_reward": 0.6575473248958588,
      "rewards/forensics_strict_edge_reward": 0.041438521817326546,
      "rewards/forensics_tiou_reward": 0.4236191362142563,
      "step": 280
    },
    {
      "completion_length": 24.296875,
      "epoch": 0.8632872503840245,
      "grad_norm": 8.296707153320312,
      "kl": 0.3251953125,
      "learning_rate": 7.846153846153846e-07,
      "loss": -0.1517,
      "reward": 0.7471871972084045,
      "reward_std": 0.06557972356677055,
      "reward_std_all": 0.06532565318048,
      "rewards-max/forensics_sinkhorn_reward": 0.9806494414806366,
      "rewards-max/forensics_strict_edge_reward": 0.7989390194416046,
      "rewards-max/forensics_tiou_reward": 0.9800819456577301,
      "rewards/forensics_sinkhorn_reward": 0.6687635481357574,
      "rewards/forensics_strict_edge_reward": 0.08170269057154655,
      "rewards/forensics_tiou_reward": 0.44829125702381134,
      "step": 281
    },
    {
      "completion_length": 25.765625,
      "epoch": 0.8663594470046083,
      "grad_norm": 8.173111915588379,
      "kl": 0.3779296875,
      "learning_rate": 7.838461538461539e-07,
      "loss": 0.3383,
      "reward": 0.6934057474136353,
      "reward_std": 0.06724909320473671,
      "reward_std_all": 0.08489689230918884,
      "rewards-max/forensics_sinkhorn_reward": 0.9034817218780518,
      "rewards-max/forensics_strict_edge_reward": 0.3862220346927643,
      "rewards-max/forensics_tiou_reward": 0.8146329522132874,
      "rewards/forensics_sinkhorn_reward": 0.5860535204410553,
      "rewards/forensics_strict_edge_reward": 0.018279753625392914,
      "rewards/forensics_tiou_reward": 0.32550448179244995,
      "step": 282
    },
    {
      "completion_length": 25.25,
      "epoch": 0.869431643625192,
      "grad_norm": 6.765259265899658,
      "kl": 0.388671875,
      "learning_rate": 7.830769230769231e-07,
      "loss": 0.5756,
      "reward": 0.665877491235733,
      "reward_std": 0.1033351942896843,
      "reward_std_all": 0.11890434846282005,
      "rewards-max/forensics_sinkhorn_reward": 0.9721370935440063,
      "rewards-max/forensics_strict_edge_reward": 0.3058830425143242,
      "rewards-max/forensics_tiou_reward": 0.9548113942146301,
      "rewards/forensics_sinkhorn_reward": 0.5728846192359924,
      "rewards/forensics_strict_edge_reward": 0.031310442835092545,
      "rewards/forensics_tiou_reward": 0.3714163154363632,
      "step": 283
    },
    {
      "completion_length": 24.203125,
      "epoch": 0.8725038402457758,
      "grad_norm": 5.93157434463501,
      "kl": 0.3359375,
      "learning_rate": 7.823076923076922e-07,
      "loss": 0.2526,
      "reward": 0.729588508605957,
      "reward_std": 0.08022509887814522,
      "reward_std_all": 0.08282696828246117,
      "rewards-max/forensics_sinkhorn_reward": 0.9789608120918274,
      "rewards-max/forensics_strict_edge_reward": 0.5115250647068024,
      "rewards-max/forensics_tiou_reward": 0.9706712365150452,
      "rewards/forensics_sinkhorn_reward": 0.6528258323669434,
      "rewards/forensics_strict_edge_reward": 0.04902293160557747,
      "rewards/forensics_tiou_reward": 0.45656590163707733,
      "step": 284
    },
    {
      "completion_length": 25.671875,
      "epoch": 0.8755760368663594,
      "grad_norm": 7.794654369354248,
      "kl": 0.46484375,
      "learning_rate": 7.815384615384615e-07,
      "loss": 0.5533,
      "reward": 0.6929652690887451,
      "reward_std": 0.032871516421437263,
      "reward_std_all": 0.06711886450648308,
      "rewards-max/forensics_sinkhorn_reward": 0.9347859621047974,
      "rewards-max/forensics_strict_edge_reward": 0.04668188374489546,
      "rewards-max/forensics_tiou_reward": 0.8516444265842438,
      "rewards/forensics_sinkhorn_reward": 0.5736425817012787,
      "rewards/forensics_strict_edge_reward": 0.00233267352450639,
      "rewards/forensics_tiou_reward": 0.2798157185316086,
      "step": 285
    },
    {
      "completion_length": 25.796875,
      "epoch": 0.8786482334869432,
      "grad_norm": 5.690019607543945,
      "kl": 0.427734375,
      "learning_rate": 7.807692307692307e-07,
      "loss": 0.5035,
      "reward": 0.6724013686180115,
      "reward_std": 0.0629255659878254,
      "reward_std_all": 0.09158996492624283,
      "rewards-max/forensics_sinkhorn_reward": 0.9353583753108978,
      "rewards-max/forensics_strict_edge_reward": 0.3341168984770775,
      "rewards-max/forensics_tiou_reward": 0.8519214689731598,
      "rewards/forensics_sinkhorn_reward": 0.521239310503006,
      "rewards/forensics_strict_edge_reward": 0.015110721811652184,
      "rewards/forensics_tiou_reward": 0.2810675799846649,
      "step": 286
    },
    {
      "completion_length": 24.859375,
      "epoch": 0.8817204301075269,
      "grad_norm": 7.84441614151001,
      "kl": 0.4228515625,
      "learning_rate": 7.799999999999999e-07,
      "loss": -0.011,
      "reward": 0.6809009909629822,
      "reward_std": 0.11473438516259193,
      "reward_std_all": 0.11560338363051414,
      "rewards-max/forensics_sinkhorn_reward": 0.9711863100528717,
      "rewards-max/forensics_strict_edge_reward": 0.291122630238533,
      "rewards-max/forensics_tiou_reward": 0.9470271468162537,
      "rewards/forensics_sinkhorn_reward": 0.5771358907222748,
      "rewards/forensics_strict_edge_reward": 0.024492562282830477,
      "rewards/forensics_tiou_reward": 0.3786438852548599,
      "step": 287
    },
    {
      "completion_length": 28.78125,
      "epoch": 0.8847926267281107,
      "grad_norm": 8.042600631713867,
      "kl": 0.4345703125,
      "learning_rate": 7.792307692307693e-07,
      "loss": 0.4213,
      "reward": 0.6746273338794708,
      "reward_std": 0.07380573265254498,
      "reward_std_all": 0.10170046612620354,
      "rewards-max/forensics_sinkhorn_reward": 0.9656964540481567,
      "rewards-max/forensics_strict_edge_reward": 0.2994947135448456,
      "rewards-max/forensics_tiou_reward": 0.9410785734653473,
      "rewards/forensics_sinkhorn_reward": 0.5439267158508301,
      "rewards/forensics_strict_edge_reward": 0.01872907020151615,
      "rewards/forensics_tiou_reward": 0.31340697407722473,
      "step": 288
    },
    {
      "completion_length": 25.078125,
      "epoch": 0.8878648233486943,
      "grad_norm": 7.670064449310303,
      "kl": 0.365234375,
      "learning_rate": 7.784615384615384e-07,
      "loss": 0.448,
      "reward": 0.6913935840129852,
      "reward_std": 0.09293121844530106,
      "reward_std_all": 0.10173530131578445,
      "rewards-max/forensics_sinkhorn_reward": 0.9536641836166382,
      "rewards-max/forensics_strict_edge_reward": 0.35114636272192,
      "rewards-max/forensics_tiou_reward": 0.9152528643608093,
      "rewards/forensics_sinkhorn_reward": 0.5894969701766968,
      "rewards/forensics_strict_edge_reward": 0.03098798217251897,
      "rewards/forensics_tiou_reward": 0.34373530000448227,
      "step": 289
    },
    {
      "completion_length": 29.109375,
      "epoch": 0.890937019969278,
      "grad_norm": 7.9386162757873535,
      "kl": 0.396484375,
      "learning_rate": 7.776923076923077e-07,
      "loss": 0.2211,
      "reward": 0.689576119184494,
      "reward_std": 0.06479708664119244,
      "reward_std_all": 0.090312160551548,
      "rewards-max/forensics_sinkhorn_reward": 0.9526481032371521,
      "rewards-max/forensics_strict_edge_reward": 0.27545714378356934,
      "rewards-max/forensics_tiou_reward": 0.9151362180709839,
      "rewards/forensics_sinkhorn_reward": 0.612300455570221,
      "rewards/forensics_strict_edge_reward": 0.03491910011507571,
      "rewards/forensics_tiou_reward": 0.39181894063949585,
      "step": 290
    },
    {
      "completion_length": 25.46875,
      "epoch": 0.8940092165898618,
      "grad_norm": 10.22054672241211,
      "kl": 0.416015625,
      "learning_rate": 7.769230769230769e-07,
      "loss": 0.1225,
      "reward": 0.711938202381134,
      "reward_std": 0.07470633462071419,
      "reward_std_all": 0.08242437988519669,
      "rewards-max/forensics_sinkhorn_reward": 0.9769964814186096,
      "rewards-max/forensics_strict_edge_reward": 0.49268636107444763,
      "rewards-max/forensics_tiou_reward": 0.9625706076622009,
      "rewards/forensics_sinkhorn_reward": 0.6329280734062195,
      "rewards/forensics_strict_edge_reward": 0.057834260165691376,
      "rewards/forensics_tiou_reward": 0.42135007679462433,
      "step": 291
    },
    {
      "completion_length": 22.75,
      "epoch": 0.8970814132104454,
      "grad_norm": 6.5405144691467285,
      "kl": 0.3935546875,
      "learning_rate": 7.761538461538461e-07,
      "loss": 0.1112,
      "reward": 0.7319860756397247,
      "reward_std": 0.07030266150832176,
      "reward_std_all": 0.07358012720942497,
      "rewards-max/forensics_sinkhorn_reward": 0.9755669832229614,
      "rewards-max/forensics_strict_edge_reward": 0.7162975072860718,
      "rewards-max/forensics_tiou_reward": 0.9632150530815125,
      "rewards/forensics_sinkhorn_reward": 0.603665828704834,
      "rewards/forensics_strict_edge_reward": 0.06099522486329079,
      "rewards/forensics_tiou_reward": 0.39052078127861023,
      "step": 292
    },
    {
      "completion_length": 24.734375,
      "epoch": 0.9001536098310292,
      "grad_norm": 8.90070915222168,
      "kl": 0.34765625,
      "learning_rate": 7.753846153846153e-07,
      "loss": 0.403,
      "reward": 0.7591840028762817,
      "reward_std": 0.049046218395233154,
      "reward_std_all": 0.05000858195126057,
      "rewards-max/forensics_sinkhorn_reward": 0.9416612088680267,
      "rewards-max/forensics_strict_edge_reward": 0.22649170830845833,
      "rewards-max/forensics_tiou_reward": 0.8798099756240845,
      "rewards/forensics_sinkhorn_reward": 0.6597484052181244,
      "rewards/forensics_strict_edge_reward": 0.01618189737200737,
      "rewards/forensics_tiou_reward": 0.3830728977918625,
      "step": 293
    },
    {
      "completion_length": 29.140625,
      "epoch": 0.9032258064516129,
      "grad_norm": 8.800174713134766,
      "kl": 0.4140625,
      "learning_rate": 7.746153846153845e-07,
      "loss": 0.0173,
      "reward": 0.670829564332962,
      "reward_std": 0.08750513754785061,
      "reward_std_all": 0.11028655990958214,
      "rewards-max/forensics_sinkhorn_reward": 0.9263834953308105,
      "rewards-max/forensics_strict_edge_reward": 0.13275554263964295,
      "rewards-max/forensics_tiou_reward": 0.8432034850120544,
      "rewards/forensics_sinkhorn_reward": 0.5311858057975769,
      "rewards/forensics_strict_edge_reward": 0.0070707721170037985,
      "rewards/forensics_tiou_reward": 0.32871393859386444,
      "step": 294
    },
    {
      "completion_length": 23.796875,
      "epoch": 0.9062980030721967,
      "grad_norm": 7.338511943817139,
      "kl": 0.396484375,
      "learning_rate": 7.738461538461539e-07,
      "loss": 0.2812,
      "reward": 0.7496210336685181,
      "reward_std": 0.047037769109010696,
      "reward_std_all": 0.05137709528207779,
      "rewards-max/forensics_sinkhorn_reward": 0.9788776934146881,
      "rewards-max/forensics_strict_edge_reward": 0.5149057954549789,
      "rewards-max/forensics_tiou_reward": 0.9652353227138519,
      "rewards/forensics_sinkhorn_reward": 0.6455433964729309,
      "rewards/forensics_strict_edge_reward": 0.05066600255668163,
      "rewards/forensics_tiou_reward": 0.3914645314216614,
      "step": 295
    },
    {
      "completion_length": 25.359375,
      "epoch": 0.9093701996927803,
      "grad_norm": 10.398367881774902,
      "kl": 0.4658203125,
      "learning_rate": 7.730769230769231e-07,
      "loss": 0.1617,
      "reward": 0.7116318047046661,
      "reward_std": 0.08359481953084469,
      "reward_std_all": 0.09096366539597511,
      "rewards-max/forensics_sinkhorn_reward": 0.9764096736907959,
      "rewards-max/forensics_strict_edge_reward": 0.4991503059864044,
      "rewards-max/forensics_tiou_reward": 0.9645285904407501,
      "rewards/forensics_sinkhorn_reward": 0.6305560767650604,
      "rewards/forensics_strict_edge_reward": 0.07417769357562065,
      "rewards/forensics_tiou_reward": 0.425239160656929,
      "step": 296
    },
    {
      "completion_length": 26.53125,
      "epoch": 0.9124423963133641,
      "grad_norm": 10.681870460510254,
      "kl": 0.3984375,
      "learning_rate": 7.723076923076923e-07,
      "loss": 0.2625,
      "reward": 0.6889337003231049,
      "reward_std": 0.12055891007184982,
      "reward_std_all": 0.11722313985228539,
      "rewards-max/forensics_sinkhorn_reward": 0.9557367563247681,
      "rewards-max/forensics_strict_edge_reward": 0.32453522831201553,
      "rewards-max/forensics_tiou_reward": 0.9210862219333649,
      "rewards/forensics_sinkhorn_reward": 0.5576828867197037,
      "rewards/forensics_strict_edge_reward": 0.021817723289132118,
      "rewards/forensics_tiou_reward": 0.3669288009405136,
      "step": 297
    },
    {
      "completion_length": 26.3125,
      "epoch": 0.9155145929339478,
      "grad_norm": 10.526775360107422,
      "kl": 0.3486328125,
      "learning_rate": 7.715384615384615e-07,
      "loss": 0.0757,
      "reward": 0.7417366802692413,
      "reward_std": 0.018387835938483477,
      "reward_std_all": 0.04088976141065359,
      "rewards-max/forensics_sinkhorn_reward": 0.9814761579036713,
      "rewards-max/forensics_strict_edge_reward": 0.4923376441001892,
      "rewards-max/forensics_tiou_reward": 0.9780594706535339,
      "rewards/forensics_sinkhorn_reward": 0.6327582895755768,
      "rewards/forensics_strict_edge_reward": 0.0567687526345253,
      "rewards/forensics_tiou_reward": 0.4398665726184845,
      "step": 298
    },
    {
      "completion_length": 25.453125,
      "epoch": 0.9185867895545314,
      "grad_norm": 6.069691181182861,
      "kl": 0.365234375,
      "learning_rate": 7.707692307692307e-07,
      "loss": 0.0591,
      "reward": 0.6738491654396057,
      "reward_std": 0.04023285396397114,
      "reward_std_all": 0.08141455799341202,
      "rewards-max/forensics_sinkhorn_reward": 0.9619916677474976,
      "rewards-max/forensics_strict_edge_reward": 0.5898665860295296,
      "rewards-max/forensics_tiou_reward": 0.9079471528530121,
      "rewards/forensics_sinkhorn_reward": 0.5479779839515686,
      "rewards/forensics_strict_edge_reward": 0.025192948058247566,
      "rewards/forensics_tiou_reward": 0.37342076003551483,
      "step": 299
    },
    {
      "completion_length": 23.234375,
      "epoch": 0.9216589861751152,
      "grad_norm": 14.078354835510254,
      "kl": 0.4580078125,
      "learning_rate": 7.699999999999999e-07,
      "loss": 0.321,
      "reward": 0.7024874091148376,
      "reward_std": 0.05871153436601162,
      "reward_std_all": 0.08084731921553612,
      "rewards-max/forensics_sinkhorn_reward": 0.9843969345092773,
      "rewards-max/forensics_strict_edge_reward": 0.6747872233390808,
      "rewards-max/forensics_tiou_reward": 0.989159345626831,
      "rewards/forensics_sinkhorn_reward": 0.6271602511405945,
      "rewards/forensics_strict_edge_reward": 0.06620882451534271,
      "rewards/forensics_tiou_reward": 0.34765782952308655,
      "step": 300
    },
    {
      "completion_length": 22.375,
      "epoch": 0.9247311827956989,
      "grad_norm": 7.225747585296631,
      "kl": 0.3330078125,
      "learning_rate": 7.692307692307693e-07,
      "loss": -0.1135,
      "reward": 0.7729068398475647,
      "reward_std": 0.03051698813214898,
      "reward_std_all": 0.03252718597650528,
      "rewards-max/forensics_sinkhorn_reward": 0.9788226187229156,
      "rewards-max/forensics_strict_edge_reward": 0.6477618515491486,
      "rewards-max/forensics_tiou_reward": 0.9625750184059143,
      "rewards/forensics_sinkhorn_reward": 0.6932465434074402,
      "rewards/forensics_strict_edge_reward": 0.0750733520835638,
      "rewards/forensics_tiou_reward": 0.4543085843324661,
      "step": 301
    },
    {
      "completion_length": 21.125,
      "epoch": 0.9278033794162827,
      "grad_norm": 9.533697128295898,
      "kl": 0.33203125,
      "learning_rate": 7.684615384615384e-07,
      "loss": -0.0442,
      "reward": 0.7268124520778656,
      "reward_std": 0.06035310262814164,
      "reward_std_all": 0.06788078788667917,
      "rewards-max/forensics_sinkhorn_reward": 0.9787189364433289,
      "rewards-max/forensics_strict_edge_reward": 0.5543488264083862,
      "rewards-max/forensics_tiou_reward": 0.9675154983997345,
      "rewards/forensics_sinkhorn_reward": 0.679309070110321,
      "rewards/forensics_strict_edge_reward": 0.05986354872584343,
      "rewards/forensics_tiou_reward": 0.4182501882314682,
      "step": 302
    },
    {
      "completion_length": 22.78125,
      "epoch": 0.9308755760368663,
      "grad_norm": 8.04935359954834,
      "kl": 0.455078125,
      "learning_rate": 7.676923076923077e-07,
      "loss": 0.1472,
      "reward": 0.7502857446670532,
      "reward_std": 0.0610730554908514,
      "reward_std_all": 0.06151713244616985,
      "rewards-max/forensics_sinkhorn_reward": 0.9609134197235107,
      "rewards-max/forensics_strict_edge_reward": 0.39454537630081177,
      "rewards-max/forensics_tiou_reward": 0.9222703576087952,
      "rewards/forensics_sinkhorn_reward": 0.6749624609947205,
      "rewards/forensics_strict_edge_reward": 0.027100663632154465,
      "rewards/forensics_tiou_reward": 0.39490607380867004,
      "step": 303
    },
    {
      "completion_length": 22.015625,
      "epoch": 0.9339477726574501,
      "grad_norm": 7.867414951324463,
      "kl": 0.45703125,
      "learning_rate": 7.669230769230769e-07,
      "loss": 0.1812,
      "reward": 0.7346336245536804,
      "reward_std": 0.051108554005622864,
      "reward_std_all": 0.058777645230293274,
      "rewards-max/forensics_sinkhorn_reward": 0.9447959661483765,
      "rewards-max/forensics_strict_edge_reward": 0.386706106364727,
      "rewards-max/forensics_tiou_reward": 0.8715566098690033,
      "rewards/forensics_sinkhorn_reward": 0.6764531135559082,
      "rewards/forensics_strict_edge_reward": 0.034471027087420225,
      "rewards/forensics_tiou_reward": 0.451199546456337,
      "step": 304
    },
    {
      "completion_length": 22.71875,
      "epoch": 0.9370199692780338,
      "grad_norm": 6.484719753265381,
      "kl": 0.4296875,
      "learning_rate": 7.661538461538461e-07,
      "loss": -0.122,
      "reward": 0.7523020207881927,
      "reward_std": 0.0752050057053566,
      "reward_std_all": 0.06916430965065956,
      "rewards-max/forensics_sinkhorn_reward": 0.9664249122142792,
      "rewards-max/forensics_strict_edge_reward": 0.36706169694662094,
      "rewards-max/forensics_tiou_reward": 0.9378497302532196,
      "rewards/forensics_sinkhorn_reward": 0.6522350609302521,
      "rewards/forensics_strict_edge_reward": 0.04957539169117808,
      "rewards/forensics_tiou_reward": 0.43075619637966156,
      "step": 305
    },
    {
      "completion_length": 23.171875,
      "epoch": 0.9400921658986175,
      "grad_norm": 8.881013870239258,
      "kl": 0.3896484375,
      "learning_rate": 7.653846153846153e-07,
      "loss": 0.2384,
      "reward": 0.7081878185272217,
      "reward_std": 0.07383800111711025,
      "reward_std_all": 0.08534533157944679,
      "rewards-max/forensics_sinkhorn_reward": 0.9742185473442078,
      "rewards-max/forensics_strict_edge_reward": 0.5454942062497139,
      "rewards-max/forensics_tiou_reward": 0.9557178318500519,
      "rewards/forensics_sinkhorn_reward": 0.6029965281486511,
      "rewards/forensics_strict_edge_reward": 0.025902779307216406,
      "rewards/forensics_tiou_reward": 0.34646646678447723,
      "step": 306
    },
    {
      "completion_length": 22.84375,
      "epoch": 0.9431643625192012,
      "grad_norm": 9.853114128112793,
      "kl": 0.3291015625,
      "learning_rate": 7.646153846153845e-07,
      "loss": 0.2147,
      "reward": 0.760531485080719,
      "reward_std": 0.05005796626210213,
      "reward_std_all": 0.049665676429867744,
      "rewards-max/forensics_sinkhorn_reward": 0.9208266139030457,
      "rewards-max/forensics_strict_edge_reward": 0.6448107361793518,
      "rewards-max/forensics_tiou_reward": 0.8453322350978851,
      "rewards/forensics_sinkhorn_reward": 0.6525402069091797,
      "rewards/forensics_strict_edge_reward": 0.06797603331506252,
      "rewards/forensics_tiou_reward": 0.37919478118419647,
      "step": 307
    },
    {
      "completion_length": 26.5,
      "epoch": 0.946236559139785,
      "grad_norm": 8.559029579162598,
      "kl": 0.5166015625,
      "learning_rate": 7.638461538461539e-07,
      "loss": -0.1489,
      "reward": 0.6454769968986511,
      "reward_std": 0.09158681705594063,
      "reward_std_all": 0.12468856573104858,
      "rewards-max/forensics_sinkhorn_reward": 0.9727913737297058,
      "rewards-max/forensics_strict_edge_reward": 0.4490189552307129,
      "rewards-max/forensics_tiou_reward": 0.952208399772644,
      "rewards/forensics_sinkhorn_reward": 0.5788426399230957,
      "rewards/forensics_strict_edge_reward": 0.029527362436056137,
      "rewards/forensics_tiou_reward": 0.35551169514656067,
      "step": 308
    },
    {
      "completion_length": 23.8125,
      "epoch": 0.9493087557603687,
      "grad_norm": 7.845417022705078,
      "kl": 0.40625,
      "learning_rate": 7.630769230769231e-07,
      "loss": 0.3597,
      "reward": 0.6957470774650574,
      "reward_std": 0.07166729122400284,
      "reward_std_all": 0.08812358602881432,
      "rewards-max/forensics_sinkhorn_reward": 0.943899005651474,
      "rewards-max/forensics_strict_edge_reward": 0.27603506995365024,
      "rewards-max/forensics_tiou_reward": 0.883102297782898,
      "rewards/forensics_sinkhorn_reward": 0.6040243208408356,
      "rewards/forensics_strict_edge_reward": 0.03424474810890388,
      "rewards/forensics_tiou_reward": 0.3250056803226471,
      "step": 309
    },
    {
      "completion_length": 25.125,
      "epoch": 0.9523809523809523,
      "grad_norm": 9.423465728759766,
      "kl": 0.3251953125,
      "learning_rate": 7.623076923076923e-07,
      "loss": 0.1103,
      "reward": 0.7620459198951721,
      "reward_std": 0.03606227971613407,
      "reward_std_all": 0.041537001729011536,
      "rewards-max/forensics_sinkhorn_reward": 0.9830825924873352,
      "rewards-max/forensics_strict_edge_reward": 0.5134405195713043,
      "rewards-max/forensics_tiou_reward": 0.9850400984287262,
      "rewards/forensics_sinkhorn_reward": 0.6725606918334961,
      "rewards/forensics_strict_edge_reward": 0.03384675830602646,
      "rewards/forensics_tiou_reward": 0.4197605550289154,
      "step": 310
    },
    {
      "completion_length": 26.375,
      "epoch": 0.9554531490015361,
      "grad_norm": 7.428229331970215,
      "kl": 0.31640625,
      "learning_rate": 7.615384615384615e-07,
      "loss": 0.1248,
      "reward": 0.6647837460041046,
      "reward_std": 0.05838080868124962,
      "reward_std_all": 0.09170203655958176,
      "rewards-max/forensics_sinkhorn_reward": 0.9508767426013947,
      "rewards-max/forensics_strict_edge_reward": 0.4288911297917366,
      "rewards-max/forensics_tiou_reward": 0.9121536314487457,
      "rewards/forensics_sinkhorn_reward": 0.5305395126342773,
      "rewards/forensics_strict_edge_reward": 0.018323025200515985,
      "rewards/forensics_tiou_reward": 0.3011608272790909,
      "step": 311
    },
    {
      "completion_length": 26.15625,
      "epoch": 0.9585253456221198,
      "grad_norm": 7.00751256942749,
      "kl": 0.3271484375,
      "learning_rate": 7.607692307692307e-07,
      "loss": 0.2682,
      "reward": 0.7150354385375977,
      "reward_std": 0.06987903267145157,
      "reward_std_all": 0.08198207244277,
      "rewards-max/forensics_sinkhorn_reward": 0.9617578089237213,
      "rewards-max/forensics_strict_edge_reward": 0.22766375541687012,
      "rewards-max/forensics_tiou_reward": 0.9286436140537262,
      "rewards/forensics_sinkhorn_reward": 0.602028101682663,
      "rewards/forensics_strict_edge_reward": 0.01474428130313754,
      "rewards/forensics_tiou_reward": 0.3791487067937851,
      "step": 312
    },
    {
      "completion_length": 24.78125,
      "epoch": 0.9615975422427036,
      "grad_norm": 11.220947265625,
      "kl": 0.455078125,
      "learning_rate": 7.599999999999999e-07,
      "loss": 0.0694,
      "reward": 0.673189640045166,
      "reward_std": 0.05606594309210777,
      "reward_std_all": 0.08851369470357895,
      "rewards-max/forensics_sinkhorn_reward": 0.9377117455005646,
      "rewards-max/forensics_strict_edge_reward": 0.0881829485297203,
      "rewards-max/forensics_tiou_reward": 0.859235405921936,
      "rewards/forensics_sinkhorn_reward": 0.5705821216106415,
      "rewards/forensics_strict_edge_reward": 0.010686890687793493,
      "rewards/forensics_tiou_reward": 0.3433423936367035,
      "step": 313
    },
    {
      "completion_length": 21.609375,
      "epoch": 0.9646697388632872,
      "grad_norm": 10.19406795501709,
      "kl": 0.3955078125,
      "learning_rate": 7.592307692307692e-07,
      "loss": 0.1413,
      "reward": 0.7133647799491882,
      "reward_std": 0.07674005255103111,
      "reward_std_all": 0.08348517492413521,
      "rewards-max/forensics_sinkhorn_reward": 0.9538969099521637,
      "rewards-max/forensics_strict_edge_reward": 0.41057318868115544,
      "rewards-max/forensics_tiou_reward": 0.910595029592514,
      "rewards/forensics_sinkhorn_reward": 0.6169018745422363,
      "rewards/forensics_strict_edge_reward": 0.036462856936850585,
      "rewards/forensics_tiou_reward": 0.3387851417064667,
      "step": 314
    },
    {
      "completion_length": 28.65625,
      "epoch": 0.967741935483871,
      "grad_norm": 6.937694549560547,
      "kl": 0.361328125,
      "learning_rate": 7.584615384615384e-07,
      "loss": 0.4204,
      "reward": 0.6554026007652283,
      "reward_std": 0.06496597826480865,
      "reward_std_all": 0.10156746581196785,
      "rewards-max/forensics_sinkhorn_reward": 0.9536848366260529,
      "rewards-max/forensics_strict_edge_reward": 0.20125575363636017,
      "rewards-max/forensics_tiou_reward": 0.916606068611145,
      "rewards/forensics_sinkhorn_reward": 0.5256815552711487,
      "rewards/forensics_strict_edge_reward": 0.01626070123165846,
      "rewards/forensics_tiou_reward": 0.3420868068933487,
      "step": 315
    },
    {
      "completion_length": 26.828125,
      "epoch": 0.9708141321044547,
      "grad_norm": 10.130130767822266,
      "kl": 0.3408203125,
      "learning_rate": 7.576923076923077e-07,
      "loss": 0.2298,
      "reward": 0.7121904492378235,
      "reward_std": 0.07995254918932915,
      "reward_std_all": 0.086051095277071,
      "rewards-max/forensics_sinkhorn_reward": 0.974128246307373,
      "rewards-max/forensics_strict_edge_reward": 0.19371240958571434,
      "rewards-max/forensics_tiou_reward": 0.9623467922210693,
      "rewards/forensics_sinkhorn_reward": 0.5381325930356979,
      "rewards/forensics_strict_edge_reward": 0.018041594768874347,
      "rewards/forensics_tiou_reward": 0.36459778249263763,
      "step": 316
    },
    {
      "completion_length": 22.609375,
      "epoch": 0.9738863287250384,
      "grad_norm": 7.567885398864746,
      "kl": 0.3662109375,
      "learning_rate": 7.569230769230769e-07,
      "loss": 0.0936,
      "reward": 0.7485354244709015,
      "reward_std": 0.044162189587950706,
      "reward_std_all": 0.053612176328897476,
      "rewards-max/forensics_sinkhorn_reward": 0.9607181251049042,
      "rewards-max/forensics_strict_edge_reward": 0.5734860002994537,
      "rewards-max/forensics_tiou_reward": 0.9238609373569489,
      "rewards/forensics_sinkhorn_reward": 0.6625774502754211,
      "rewards/forensics_strict_edge_reward": 0.03611167334020138,
      "rewards/forensics_tiou_reward": 0.4031110107898712,
      "step": 317
    },
    {
      "completion_length": 24.9375,
      "epoch": 0.9769585253456221,
      "grad_norm": 9.294951438903809,
      "kl": 0.4052734375,
      "learning_rate": 7.561538461538461e-07,
      "loss": -0.0861,
      "reward": 0.7093897461891174,
      "reward_std": 0.04084879998117685,
      "reward_std_all": 0.06814747303724289,
      "rewards-max/forensics_sinkhorn_reward": 0.9551569521427155,
      "rewards-max/forensics_strict_edge_reward": 0.3192719966173172,
      "rewards-max/forensics_tiou_reward": 0.9126816391944885,
      "rewards/forensics_sinkhorn_reward": 0.6060220301151276,
      "rewards/forensics_strict_edge_reward": 0.023656069301068783,
      "rewards/forensics_tiou_reward": 0.34284211695194244,
      "step": 318
    },
    {
      "completion_length": 25.5,
      "epoch": 0.9800307219662059,
      "grad_norm": 9.246077537536621,
      "kl": 0.2861328125,
      "learning_rate": 7.553846153846154e-07,
      "loss": 0.0823,
      "reward": 0.6835789680480957,
      "reward_std": 0.09701603278517723,
      "reward_std_all": 0.10740550607442856,
      "rewards-max/forensics_sinkhorn_reward": 0.9768449068069458,
      "rewards-max/forensics_strict_edge_reward": 0.5543488264083862,
      "rewards-max/forensics_tiou_reward": 0.9686307311058044,
      "rewards/forensics_sinkhorn_reward": 0.5551808476448059,
      "rewards/forensics_strict_edge_reward": 0.03980223275721073,
      "rewards/forensics_tiou_reward": 0.3693729192018509,
      "step": 319
    },
    {
      "completion_length": 24.234375,
      "epoch": 0.9831029185867896,
      "grad_norm": 10.16265869140625,
      "kl": 0.4296875,
      "learning_rate": 7.546153846153845e-07,
      "loss": 0.2107,
      "reward": 0.7431134581565857,
      "reward_std": 0.04090100899338722,
      "reward_std_all": 0.05419265665113926,
      "rewards-max/forensics_sinkhorn_reward": 0.9788696765899658,
      "rewards-max/forensics_strict_edge_reward": 0.38056008517742157,
      "rewards-max/forensics_tiou_reward": 0.9682457745075226,
      "rewards/forensics_sinkhorn_reward": 0.7050062119960785,
      "rewards/forensics_strict_edge_reward": 0.05698336474597454,
      "rewards/forensics_tiou_reward": 0.5002851337194443,
      "step": 320
    },
    {
      "completion_length": 24.9375,
      "epoch": 0.9861751152073732,
      "grad_norm": 7.675189018249512,
      "kl": 0.275390625,
      "learning_rate": 7.538461538461538e-07,
      "loss": 0.3567,
      "reward": 0.6982126832008362,
      "reward_std": 0.04847119376063347,
      "reward_std_all": 0.07590058445930481,
      "rewards-max/forensics_sinkhorn_reward": 0.916392982006073,
      "rewards-max/forensics_strict_edge_reward": 0.15503494255244732,
      "rewards-max/forensics_tiou_reward": 0.820159912109375,
      "rewards/forensics_sinkhorn_reward": 0.5815980732440948,
      "rewards/forensics_strict_edge_reward": 0.007518122554756701,
      "rewards/forensics_tiou_reward": 0.32125823199748993,
      "step": 321
    },
    {
      "completion_length": 26.359375,
      "epoch": 0.989247311827957,
      "grad_norm": 7.28254508972168,
      "kl": 0.34765625,
      "learning_rate": 7.530769230769231e-07,
      "loss": 0.331,
      "reward": 0.6977369785308838,
      "reward_std": 0.07054933160543442,
      "reward_std_all": 0.08807684853672981,
      "rewards-max/forensics_sinkhorn_reward": 0.9491972029209137,
      "rewards-max/forensics_strict_edge_reward": 0.19055698812007904,
      "rewards-max/forensics_tiou_reward": 0.9001670777797699,
      "rewards/forensics_sinkhorn_reward": 0.5602310299873352,
      "rewards/forensics_strict_edge_reward": 0.013344904407858849,
      "rewards/forensics_tiou_reward": 0.3776431381702423,
      "step": 322
    },
    {
      "completion_length": 25.53125,
      "epoch": 0.9923195084485407,
      "grad_norm": 6.97634220123291,
      "kl": 0.533203125,
      "learning_rate": 7.523076923076923e-07,
      "loss": 0.4385,
      "reward": 0.6904036104679108,
      "reward_std": 0.08454121463000774,
      "reward_std_all": 0.09727947413921356,
      "rewards-max/forensics_sinkhorn_reward": 0.9672290980815887,
      "rewards-max/forensics_strict_edge_reward": 0.252693772315979,
      "rewards-max/forensics_tiou_reward": 0.936723917722702,
      "rewards/forensics_sinkhorn_reward": 0.5610057413578033,
      "rewards/forensics_strict_edge_reward": 0.017477679066359997,
      "rewards/forensics_tiou_reward": 0.3263914883136749,
      "step": 323
    },
    {
      "completion_length": 25.546875,
      "epoch": 0.9953917050691244,
      "grad_norm": 8.240628242492676,
      "kl": 0.728515625,
      "learning_rate": 7.515384615384615e-07,
      "loss": 0.3362,
      "reward": 0.7318499982357025,
      "reward_std": 0.052237652242183685,
      "reward_std_all": 0.06288736686110497,
      "rewards-max/forensics_sinkhorn_reward": 0.9753551185131073,
      "rewards-max/forensics_strict_edge_reward": 0.3787429630756378,
      "rewards-max/forensics_tiou_reward": 0.9607414603233337,
      "rewards/forensics_sinkhorn_reward": 0.6208849251270294,
      "rewards/forensics_strict_edge_reward": 0.04402424953877926,
      "rewards/forensics_tiou_reward": 0.48123280704021454,
      "step": 324
    },
    {
      "completion_length": 26.046875,
      "epoch": 0.9984639016897081,
      "grad_norm": 10.829663276672363,
      "kl": 0.7470703125,
      "learning_rate": 7.507692307692307e-07,
      "loss": 0.5563,
      "reward": 0.6935741305351257,
      "reward_std": 0.09955982863903046,
      "reward_std_all": 0.10431636497378349,
      "rewards-max/forensics_sinkhorn_reward": 0.9485808610916138,
      "rewards-max/forensics_strict_edge_reward": 0.2097843885421753,
      "rewards-max/forensics_tiou_reward": 0.8951280415058136,
      "rewards/forensics_sinkhorn_reward": 0.5967921912670135,
      "rewards/forensics_strict_edge_reward": 0.019335933960974216,
      "rewards/forensics_tiou_reward": 0.38831816613674164,
      "step": 325
    },
    {
      "completion_length": 18.80000114440918,
      "epoch": 1.0,
      "grad_norm": 6.299590110778809,
      "kl": 0.259765625,
      "learning_rate": 7.5e-07,
      "loss": 0.1568,
      "reward": 0.7038647532463074,
      "reward_std": 0.07714565098285675,
      "reward_std_all": 0.08943378925323486,
      "rewards-max/forensics_sinkhorn_reward": 0.9849702715873718,
      "rewards-max/forensics_strict_edge_reward": 0.3678794503211975,
      "rewards-max/forensics_tiou_reward": 0.9881454110145569,
      "rewards/forensics_sinkhorn_reward": 0.6239210963249207,
      "rewards/forensics_strict_edge_reward": 0.07357589155435562,
      "rewards/forensics_tiou_reward": 0.26159772276878357,
      "step": 326
    },
    {
      "completion_length": 27.953125,
      "epoch": 1.0030721966205838,
      "grad_norm": 6.167362213134766,
      "kl": 0.4501953125,
      "learning_rate": 7.492307692307692e-07,
      "loss": 0.2451,
      "reward": 0.7148918509483337,
      "reward_std": 0.05338750034570694,
      "reward_std_all": 0.0690564438700676,
      "rewards-max/forensics_sinkhorn_reward": 0.937593400478363,
      "rewards-max/forensics_strict_edge_reward": 0.25321530317887664,
      "rewards-max/forensics_tiou_reward": 0.8783177137374878,
      "rewards/forensics_sinkhorn_reward": 0.6020452380180359,
      "rewards/forensics_strict_edge_reward": 0.019697177573107183,
      "rewards/forensics_tiou_reward": 0.36313392221927643,
      "step": 327
    },
    {
      "completion_length": 24.15625,
      "epoch": 1.0061443932411673,
      "grad_norm": 8.754875183105469,
      "kl": 0.361328125,
      "learning_rate": 7.484615384615384e-07,
      "loss": 0.3886,
      "reward": 0.7364874482154846,
      "reward_std": 0.07685594633221626,
      "reward_std_all": 0.07517009228467941,
      "rewards-max/forensics_sinkhorn_reward": 0.9798663556575775,
      "rewards-max/forensics_strict_edge_reward": 0.47400766611099243,
      "rewards-max/forensics_tiou_reward": 0.9735434055328369,
      "rewards/forensics_sinkhorn_reward": 0.6389974653720856,
      "rewards/forensics_strict_edge_reward": 0.04606620967388153,
      "rewards/forensics_tiou_reward": 0.4391465187072754,
      "step": 328
    },
    {
      "completion_length": 27.546875,
      "epoch": 1.0092165898617511,
      "grad_norm": 7.154808521270752,
      "kl": 0.3974609375,
      "learning_rate": 7.476923076923077e-07,
      "loss": 0.4505,
      "reward": 0.6617894172668457,
      "reward_std": 0.08392316848039627,
      "reward_std_all": 0.1080937460064888,
      "rewards-max/forensics_sinkhorn_reward": 0.9701743721961975,
      "rewards-max/forensics_strict_edge_reward": 0.37661072611808777,
      "rewards-max/forensics_tiou_reward": 0.9474197924137115,
      "rewards/forensics_sinkhorn_reward": 0.5352914929389954,
      "rewards/forensics_strict_edge_reward": 0.021242516115307808,
      "rewards/forensics_tiou_reward": 0.36107200384140015,
      "step": 329
    },
    {
      "completion_length": 27.90625,
      "epoch": 1.012288786482335,
      "grad_norm": 8.014079093933105,
      "kl": 0.3759765625,
      "learning_rate": 7.469230769230769e-07,
      "loss": 0.3942,
      "reward": 0.693494588136673,
      "reward_std": 0.0815393254160881,
      "reward_std_all": 0.0961967185139656,
      "rewards-max/forensics_sinkhorn_reward": 0.9726941585540771,
      "rewards-max/forensics_strict_edge_reward": 0.4636124148964882,
      "rewards-max/forensics_tiou_reward": 0.9520912766456604,
      "rewards/forensics_sinkhorn_reward": 0.552621141076088,
      "rewards/forensics_strict_edge_reward": 0.033423683140426874,
      "rewards/forensics_tiou_reward": 0.40317749977111816,
      "step": 330
    },
    {
      "completion_length": 29.3125,
      "epoch": 1.0153609831029187,
      "grad_norm": 7.08407735824585,
      "kl": 0.7724609375,
      "learning_rate": 7.461538461538461e-07,
      "loss": 0.5486,
      "reward": 0.6888374984264374,
      "reward_std": 0.1018579937517643,
      "reward_std_all": 0.11048709601163864,
      "rewards-max/forensics_sinkhorn_reward": 0.9786687195301056,
      "rewards-max/forensics_strict_edge_reward": 0.6032288670539856,
      "rewards-max/forensics_tiou_reward": 0.9720938503742218,
      "rewards/forensics_sinkhorn_reward": 0.565423846244812,
      "rewards/forensics_strict_edge_reward": 0.06882678717374802,
      "rewards/forensics_tiou_reward": 0.4011589586734772,
      "step": 331
    },
    {
      "completion_length": 28.03125,
      "epoch": 1.0184331797235022,
      "grad_norm": 6.226096153259277,
      "kl": 0.2861328125,
      "learning_rate": 7.453846153846154e-07,
      "loss": 0.3321,
      "reward": 0.6762525737285614,
      "reward_std": 0.07328472658991814,
      "reward_std_all": 0.09468119032680988,
      "rewards-max/forensics_sinkhorn_reward": 0.9125233292579651,
      "rewards-max/forensics_strict_edge_reward": 0.08022034727036953,
      "rewards-max/forensics_tiou_reward": 0.8124056458473206,
      "rewards/forensics_sinkhorn_reward": 0.5104714632034302,
      "rewards/forensics_strict_edge_reward": 0.004379368736408651,
      "rewards/forensics_tiou_reward": 0.3470168858766556,
      "step": 332
    },
    {
      "completion_length": 24.671875,
      "epoch": 1.021505376344086,
      "grad_norm": 6.618248462677002,
      "kl": 0.4140625,
      "learning_rate": 7.446153846153845e-07,
      "loss": 0.3259,
      "reward": 0.719462662935257,
      "reward_std": 0.0829564556479454,
      "reward_std_all": 0.08558216691017151,
      "rewards-max/forensics_sinkhorn_reward": 0.9794808328151703,
      "rewards-max/forensics_strict_edge_reward": 0.6442374289035797,
      "rewards-max/forensics_tiou_reward": 0.9726704359054565,
      "rewards/forensics_sinkhorn_reward": 0.6253509223461151,
      "rewards/forensics_strict_edge_reward": 0.0572514533996582,
      "rewards/forensics_tiou_reward": 0.4123949855566025,
      "step": 333
    },
    {
      "completion_length": 28.15625,
      "epoch": 1.0245775729646698,
      "grad_norm": 8.93328857421875,
      "kl": 0.541015625,
      "learning_rate": 7.438461538461538e-07,
      "loss": 0.3328,
      "reward": 0.6965509057044983,
      "reward_std": 0.07531542330980301,
      "reward_std_all": 0.09065248817205429,
      "rewards-max/forensics_sinkhorn_reward": 0.947857141494751,
      "rewards-max/forensics_strict_edge_reward": 0.4018307626247406,
      "rewards-max/forensics_tiou_reward": 0.904437243938446,
      "rewards/forensics_sinkhorn_reward": 0.5786987543106079,
      "rewards/forensics_strict_edge_reward": 0.05226378608494997,
      "rewards/forensics_tiou_reward": 0.37702950835227966,
      "step": 334
    },
    {
      "completion_length": 23.390625,
      "epoch": 1.0276497695852536,
      "grad_norm": 8.032510757446289,
      "kl": 0.5498046875,
      "learning_rate": 7.43076923076923e-07,
      "loss": 0.012,
      "reward": 0.7069809138774872,
      "reward_std": 0.09023300930857658,
      "reward_std_all": 0.094998300075531,
      "rewards-max/forensics_sinkhorn_reward": 0.9760256111621857,
      "rewards-max/forensics_strict_edge_reward": 0.4322323799133301,
      "rewards-max/forensics_tiou_reward": 0.9620792865753174,
      "rewards/forensics_sinkhorn_reward": 0.6235249638557434,
      "rewards/forensics_strict_edge_reward": 0.043165381997823715,
      "rewards/forensics_tiou_reward": 0.4194822311401367,
      "step": 335
    },
    {
      "completion_length": 28.015625,
      "epoch": 1.0307219662058371,
      "grad_norm": 9.856558799743652,
      "kl": 0.3876953125,
      "learning_rate": 7.423076923076923e-07,
      "loss": 0.6139,
      "reward": 0.7006586492061615,
      "reward_std": 0.0551003348082304,
      "reward_std_all": 0.0746452547609806,
      "rewards-max/forensics_sinkhorn_reward": 0.9583381414413452,
      "rewards-max/forensics_strict_edge_reward": 0.18192462623119354,
      "rewards-max/forensics_tiou_reward": 0.9272813498973846,
      "rewards/forensics_sinkhorn_reward": 0.5215939879417419,
      "rewards/forensics_strict_edge_reward": 0.012613029684871435,
      "rewards/forensics_tiou_reward": 0.34290216863155365,
      "step": 336
    },
    {
      "completion_length": 22.015625,
      "epoch": 1.033794162826421,
      "grad_norm": 8.526281356811523,
      "kl": 0.51953125,
      "learning_rate": 7.415384615384616e-07,
      "loss": 0.0159,
      "reward": 0.7579930424690247,
      "reward_std": 0.02139008278027177,
      "reward_std_all": 0.036065077409148216,
      "rewards-max/forensics_sinkhorn_reward": 0.9600741565227509,
      "rewards-max/forensics_strict_edge_reward": 0.4002290964126587,
      "rewards-max/forensics_tiou_reward": 0.9273351430892944,
      "rewards/forensics_sinkhorn_reward": 0.6631386280059814,
      "rewards/forensics_strict_edge_reward": 0.055324602872133255,
      "rewards/forensics_tiou_reward": 0.36510422825813293,
      "step": 337
    },
    {
      "completion_length": 25.1875,
      "epoch": 1.0368663594470047,
      "grad_norm": 9.950227737426758,
      "kl": 0.3798828125,
      "learning_rate": 7.407692307692307e-07,
      "loss": 0.4637,
      "reward": 0.7246573865413666,
      "reward_std": 0.07140827551484108,
      "reward_std_all": 0.07558783888816833,
      "rewards-max/forensics_sinkhorn_reward": 0.9695887267589569,
      "rewards-max/forensics_strict_edge_reward": 0.12081857398152351,
      "rewards-max/forensics_tiou_reward": 0.9498373866081238,
      "rewards/forensics_sinkhorn_reward": 0.6054162681102753,
      "rewards/forensics_strict_edge_reward": 0.01185500156134367,
      "rewards/forensics_tiou_reward": 0.3389647305011749,
      "step": 338
    },
    {
      "completion_length": 25.75,
      "epoch": 1.0399385560675882,
      "grad_norm": 9.937435150146484,
      "kl": 0.4052734375,
      "learning_rate": 7.4e-07,
      "loss": 0.3411,
      "reward": 0.7117573618888855,
      "reward_std": 0.06401179730892181,
      "reward_std_all": 0.07723189517855644,
      "rewards-max/forensics_sinkhorn_reward": 0.9744923114776611,
      "rewards-max/forensics_strict_edge_reward": 0.3340276926755905,
      "rewards-max/forensics_tiou_reward": 0.9479511678218842,
      "rewards/forensics_sinkhorn_reward": 0.5633820593357086,
      "rewards/forensics_strict_edge_reward": 0.03372384328395128,
      "rewards/forensics_tiou_reward": 0.37436775863170624,
      "step": 339
    },
    {
      "completion_length": 27.25,
      "epoch": 1.043010752688172,
      "grad_norm": 5.598576545715332,
      "kl": 0.3291015625,
      "learning_rate": 7.392307692307692e-07,
      "loss": 0.219,
      "reward": 0.7453837096691132,
      "reward_std": 0.05077718384563923,
      "reward_std_all": 0.05714017525315285,
      "rewards-max/forensics_sinkhorn_reward": 0.9821487069129944,
      "rewards-max/forensics_strict_edge_reward": 0.5674866065382957,
      "rewards-max/forensics_tiou_reward": 0.9862693846225739,
      "rewards/forensics_sinkhorn_reward": 0.6600542366504669,
      "rewards/forensics_strict_edge_reward": 0.03695438802242279,
      "rewards/forensics_tiou_reward": 0.45814330875873566,
      "step": 340
    },
    {
      "completion_length": 22.96875,
      "epoch": 1.0460829493087558,
      "grad_norm": 9.176945686340332,
      "kl": 0.365234375,
      "learning_rate": 7.384615384615384e-07,
      "loss": 0.4629,
      "reward": 0.7191495001316071,
      "reward_std": 0.04100160673260689,
      "reward_std_all": 0.061121050268411636,
      "rewards-max/forensics_sinkhorn_reward": 0.9567378163337708,
      "rewards-max/forensics_strict_edge_reward": 0.2156689465045929,
      "rewards-max/forensics_tiou_reward": 0.9166657328605652,
      "rewards/forensics_sinkhorn_reward": 0.5655657947063446,
      "rewards/forensics_strict_edge_reward": 0.011132110375910997,
      "rewards/forensics_tiou_reward": 0.3434039205312729,
      "step": 341
    },
    {
      "completion_length": 25.515625,
      "epoch": 1.0491551459293396,
      "grad_norm": 7.4527668952941895,
      "kl": 0.439453125,
      "learning_rate": 7.376923076923076e-07,
      "loss": 0.2597,
      "reward": 0.6869219541549683,
      "reward_std": 0.09947646036744118,
      "reward_std_all": 0.11340627446770668,
      "rewards-max/forensics_sinkhorn_reward": 0.9836703836917877,
      "rewards-max/forensics_strict_edge_reward": 0.646886333823204,
      "rewards-max/forensics_tiou_reward": 0.9860365092754364,
      "rewards/forensics_sinkhorn_reward": 0.5972523987293243,
      "rewards/forensics_strict_edge_reward": 0.04411652870476246,
      "rewards/forensics_tiou_reward": 0.3859596997499466,
      "step": 342
    },
    {
      "completion_length": 25.90625,
      "epoch": 1.0522273425499231,
      "grad_norm": 11.500772476196289,
      "kl": 0.4130859375,
      "learning_rate": 7.369230769230769e-07,
      "loss": 0.3921,
      "reward": 0.6785349547863007,
      "reward_std": 0.07817816361784935,
      "reward_std_all": 0.10117920115590096,
      "rewards-max/forensics_sinkhorn_reward": 0.9825147390365601,
      "rewards-max/forensics_strict_edge_reward": 0.6049591153860092,
      "rewards-max/forensics_tiou_reward": 0.9838264286518097,
      "rewards/forensics_sinkhorn_reward": 0.5914056599140167,
      "rewards/forensics_strict_edge_reward": 0.06680218316614628,
      "rewards/forensics_tiou_reward": 0.4065902531147003,
      "step": 343
    },
    {
      "completion_length": 25.625,
      "epoch": 1.055299539170507,
      "grad_norm": 8.28559398651123,
      "kl": 0.33203125,
      "learning_rate": 7.361538461538461e-07,
      "loss": 0.3501,
      "reward": 0.7233308255672455,
      "reward_std": 0.046745709143579006,
      "reward_std_all": 0.06326028890907764,
      "rewards-max/forensics_sinkhorn_reward": 0.9798959195613861,
      "rewards-max/forensics_strict_edge_reward": 0.5681687146425247,
      "rewards-max/forensics_tiou_reward": 0.9734334647655487,
      "rewards/forensics_sinkhorn_reward": 0.5805175304412842,
      "rewards/forensics_strict_edge_reward": 0.042252964340150356,
      "rewards/forensics_tiou_reward": 0.4314386397600174,
      "step": 344
    },
    {
      "completion_length": 22.5,
      "epoch": 1.0583717357910907,
      "grad_norm": 9.560919761657715,
      "kl": 0.4326171875,
      "learning_rate": 7.353846153846154e-07,
      "loss": -0.1003,
      "reward": 0.7638283669948578,
      "reward_std": 0.04375390708446503,
      "reward_std_all": 0.04523095302283764,
      "rewards-max/forensics_sinkhorn_reward": 0.9688973724842072,
      "rewards-max/forensics_strict_edge_reward": 0.26407431811094284,
      "rewards-max/forensics_tiou_reward": 0.9460159242153168,
      "rewards/forensics_sinkhorn_reward": 0.7083572149276733,
      "rewards/forensics_strict_edge_reward": 0.026536171324551105,
      "rewards/forensics_tiou_reward": 0.5050846636295319,
      "step": 345
    },
    {
      "completion_length": 25.171875,
      "epoch": 1.0614439324116745,
      "grad_norm": 13.512022972106934,
      "kl": 0.3828125,
      "learning_rate": 7.346153846153846e-07,
      "loss": 0.308,
      "reward": 0.7456508278846741,
      "reward_std": 0.04662775993347168,
      "reward_std_all": 0.05464194342494011,
      "rewards-max/forensics_sinkhorn_reward": 0.9546339809894562,
      "rewards-max/forensics_strict_edge_reward": 0.3316391110420227,
      "rewards-max/forensics_tiou_reward": 0.8904066979885101,
      "rewards/forensics_sinkhorn_reward": 0.6377090513706207,
      "rewards/forensics_strict_edge_reward": 0.026085272431373596,
      "rewards/forensics_tiou_reward": 0.3956010639667511,
      "step": 346
    },
    {
      "completion_length": 23.03125,
      "epoch": 1.064516129032258,
      "grad_norm": 7.724719047546387,
      "kl": 0.32421875,
      "learning_rate": 7.338461538461538e-07,
      "loss": 0.0117,
      "reward": 0.7814080715179443,
      "reward_std": 0.022885517217218876,
      "reward_std_all": 0.025972958654165268,
      "rewards-max/forensics_sinkhorn_reward": 0.9709329009056091,
      "rewards-max/forensics_strict_edge_reward": 0.277015283703804,
      "rewards-max/forensics_tiou_reward": 0.943852424621582,
      "rewards/forensics_sinkhorn_reward": 0.7189554870128632,
      "rewards/forensics_strict_edge_reward": 0.02817194862291217,
      "rewards/forensics_tiou_reward": 0.4496214985847473,
      "step": 347
    },
    {
      "completion_length": 23.1875,
      "epoch": 1.0675883256528418,
      "grad_norm": 8.693140983581543,
      "kl": 0.41796875,
      "learning_rate": 7.33076923076923e-07,
      "loss": 0.0875,
      "reward": 0.7268176674842834,
      "reward_std": 0.04158054105937481,
      "reward_std_all": 0.06246541999280453,
      "rewards-max/forensics_sinkhorn_reward": 0.985531747341156,
      "rewards-max/forensics_strict_edge_reward": 0.835382878780365,
      "rewards-max/forensics_tiou_reward": 0.9905379116535187,
      "rewards/forensics_sinkhorn_reward": 0.6846655011177063,
      "rewards/forensics_strict_edge_reward": 0.09610839188098907,
      "rewards/forensics_tiou_reward": 0.4526590406894684,
      "step": 348
    },
    {
      "completion_length": 23.75,
      "epoch": 1.0706605222734256,
      "grad_norm": 7.252242565155029,
      "kl": 0.4609375,
      "learning_rate": 7.323076923076923e-07,
      "loss": 0.2582,
      "reward": 0.7189093232154846,
      "reward_std": 0.009702297858893871,
      "reward_std_all": 0.048692477867007256,
      "rewards-max/forensics_sinkhorn_reward": 0.9763169288635254,
      "rewards-max/forensics_strict_edge_reward": 0.51341712474823,
      "rewards-max/forensics_tiou_reward": 0.9690779745578766,
      "rewards/forensics_sinkhorn_reward": 0.6336264908313751,
      "rewards/forensics_strict_edge_reward": 0.04717734642326832,
      "rewards/forensics_tiou_reward": 0.43055446445941925,
      "step": 349
    },
    {
      "completion_length": 24.8125,
      "epoch": 1.0737327188940091,
      "grad_norm": 9.403846740722656,
      "kl": 0.298828125,
      "learning_rate": 7.315384615384616e-07,
      "loss": -0.1076,
      "reward": 0.7470448315143585,
      "reward_std": 0.02351155411452055,
      "reward_std_all": 0.04112078994512558,
      "rewards-max/forensics_sinkhorn_reward": 0.977959156036377,
      "rewards-max/forensics_strict_edge_reward": 0.6149742305278778,
      "rewards-max/forensics_tiou_reward": 0.9709273278713226,
      "rewards/forensics_sinkhorn_reward": 0.7287253439426422,
      "rewards/forensics_strict_edge_reward": 0.07510843314230442,
      "rewards/forensics_tiou_reward": 0.5087071061134338,
      "step": 350
    },
    {
      "completion_length": 26.375,
      "epoch": 1.076804915514593,
      "grad_norm": 8.275656700134277,
      "kl": 0.404296875,
      "learning_rate": 7.307692307692307e-07,
      "loss": 0.0582,
      "reward": 0.7374346852302551,
      "reward_std": 0.02536086644977331,
      "reward_std_all": 0.04802560620009899,
      "rewards-max/forensics_sinkhorn_reward": 0.975814938545227,
      "rewards-max/forensics_strict_edge_reward": 0.5050012171268463,
      "rewards-max/forensics_tiou_reward": 0.9644027352333069,
      "rewards/forensics_sinkhorn_reward": 0.645668089389801,
      "rewards/forensics_strict_edge_reward": 0.04120209626853466,
      "rewards/forensics_tiou_reward": 0.44021327793598175,
      "step": 351
    },
    {
      "completion_length": 24.59375,
      "epoch": 1.0798771121351767,
      "grad_norm": 10.320450782775879,
      "kl": 0.3466796875,
      "learning_rate": 7.3e-07,
      "loss": 0.3163,
      "reward": 0.7014794945716858,
      "reward_std": 0.07136653549969196,
      "reward_std_all": 0.08842561393976212,
      "rewards-max/forensics_sinkhorn_reward": 0.956143319606781,
      "rewards-max/forensics_strict_edge_reward": 0.3313475623726845,
      "rewards-max/forensics_tiou_reward": 0.9227888286113739,
      "rewards/forensics_sinkhorn_reward": 0.6144866347312927,
      "rewards/forensics_strict_edge_reward": 0.032935792580246925,
      "rewards/forensics_tiou_reward": 0.41528794169425964,
      "step": 352
    },
    {
      "completion_length": 22.859375,
      "epoch": 1.0829493087557605,
      "grad_norm": 9.1630277633667,
      "kl": 0.505859375,
      "learning_rate": 7.292307692307692e-07,
      "loss": 0.1015,
      "reward": 0.7551121115684509,
      "reward_std": 0.03319784812629223,
      "reward_std_all": 0.04056677874177694,
      "rewards-max/forensics_sinkhorn_reward": 0.9821502268314362,
      "rewards-max/forensics_strict_edge_reward": 0.6516932249069214,
      "rewards-max/forensics_tiou_reward": 0.9812420010566711,
      "rewards/forensics_sinkhorn_reward": 0.6442587673664093,
      "rewards/forensics_strict_edge_reward": 0.052888380363583565,
      "rewards/forensics_tiou_reward": 0.3916681855916977,
      "step": 353
    },
    {
      "completion_length": 23.421875,
      "epoch": 1.086021505376344,
      "grad_norm": 12.10097599029541,
      "kl": 0.529296875,
      "learning_rate": 7.284615384615384e-07,
      "loss": -0.0405,
      "reward": 0.7157991826534271,
      "reward_std": 0.07574532553553581,
      "reward_std_all": 0.07961887866258621,
      "rewards-max/forensics_sinkhorn_reward": 0.931818276643753,
      "rewards-max/forensics_strict_edge_reward": 0.20482546836137772,
      "rewards-max/forensics_tiou_reward": 0.8516659140586853,
      "rewards/forensics_sinkhorn_reward": 0.5701562762260437,
      "rewards/forensics_strict_edge_reward": 0.016741957049816847,
      "rewards/forensics_tiou_reward": 0.31733438372612,
      "step": 354
    },
    {
      "completion_length": 23.234375,
      "epoch": 1.0890937019969278,
      "grad_norm": 9.650479316711426,
      "kl": 0.4130859375,
      "learning_rate": 7.276923076923076e-07,
      "loss": 0.2006,
      "reward": 0.6939060091972351,
      "reward_std": 0.07490918040275574,
      "reward_std_all": 0.09471633285284042,
      "rewards-max/forensics_sinkhorn_reward": 0.9809505343437195,
      "rewards-max/forensics_strict_edge_reward": 0.4826948940753937,
      "rewards-max/forensics_tiou_reward": 0.9776048064231873,
      "rewards/forensics_sinkhorn_reward": 0.6100506782531738,
      "rewards/forensics_strict_edge_reward": 0.05893502198159695,
      "rewards/forensics_tiou_reward": 0.37648941576480865,
      "step": 355
    },
    {
      "completion_length": 25.828125,
      "epoch": 1.0921658986175116,
      "grad_norm": 10.85537338256836,
      "kl": 0.384765625,
      "learning_rate": 7.269230769230769e-07,
      "loss": 0.3035,
      "reward": 0.7235418260097504,
      "reward_std": 0.05101047642529011,
      "reward_std_all": 0.06578710116446018,
      "rewards-max/forensics_sinkhorn_reward": 0.9670742452144623,
      "rewards-max/forensics_strict_edge_reward": 0.39300739392638206,
      "rewards-max/forensics_tiou_reward": 0.9373022317886353,
      "rewards/forensics_sinkhorn_reward": 0.6055653095245361,
      "rewards/forensics_strict_edge_reward": 0.020238644443452358,
      "rewards/forensics_tiou_reward": 0.3585125803947449,
      "step": 356
    },
    {
      "completion_length": 26.859375,
      "epoch": 1.0952380952380953,
      "grad_norm": 10.130902290344238,
      "kl": 0.47265625,
      "learning_rate": 7.261538461538461e-07,
      "loss": 0.1874,
      "reward": 0.7650751173496246,
      "reward_std": 0.03005181159824133,
      "reward_std_all": 0.03501879423856735,
      "rewards-max/forensics_sinkhorn_reward": 0.9483939409255981,
      "rewards-max/forensics_strict_edge_reward": 0.24411102384328842,
      "rewards-max/forensics_tiou_reward": 0.901274710893631,
      "rewards/forensics_sinkhorn_reward": 0.6726334393024445,
      "rewards/forensics_strict_edge_reward": 0.02224409836344421,
      "rewards/forensics_tiou_reward": 0.4679280072450638,
      "step": 357
    },
    {
      "completion_length": 26.453125,
      "epoch": 1.098310291858679,
      "grad_norm": 7.036717414855957,
      "kl": 0.3779296875,
      "learning_rate": 7.253846153846154e-07,
      "loss": -0.4143,
      "reward": 0.7319696843624115,
      "reward_std": 0.05829307809472084,
      "reward_std_all": 0.06377600505948067,
      "rewards-max/forensics_sinkhorn_reward": 0.9660634100437164,
      "rewards-max/forensics_strict_edge_reward": 0.6082725524902344,
      "rewards-max/forensics_tiou_reward": 0.9504069089889526,
      "rewards/forensics_sinkhorn_reward": 0.6268683671951294,
      "rewards/forensics_strict_edge_reward": 0.04378329124301672,
      "rewards/forensics_tiou_reward": 0.4424930214881897,
      "step": 358
    },
    {
      "completion_length": 25.5,
      "epoch": 1.1013824884792627,
      "grad_norm": 6.856736660003662,
      "kl": 0.3369140625,
      "learning_rate": 7.246153846153846e-07,
      "loss": -0.0307,
      "reward": 0.7428246140480042,
      "reward_std": 0.035801539197564125,
      "reward_std_all": 0.04998430050909519,
      "rewards-max/forensics_sinkhorn_reward": 0.9400519132614136,
      "rewards-max/forensics_strict_edge_reward": 0.2605968713760376,
      "rewards-max/forensics_tiou_reward": 0.8830547034740448,
      "rewards/forensics_sinkhorn_reward": 0.6193317770957947,
      "rewards/forensics_strict_edge_reward": 0.018991936929523945,
      "rewards/forensics_tiou_reward": 0.3926986753940582,
      "step": 359
    },
    {
      "completion_length": 22.796875,
      "epoch": 1.1044546850998465,
      "grad_norm": 8.842378616333008,
      "kl": 0.4267578125,
      "learning_rate": 7.238461538461538e-07,
      "loss": 0.3592,
      "reward": 0.7677593231201172,
      "reward_std": 0.04572707787156105,
      "reward_std_all": 0.044132012873888016,
      "rewards-max/forensics_sinkhorn_reward": 0.9796815514564514,
      "rewards-max/forensics_strict_edge_reward": 0.6095658242702484,
      "rewards-max/forensics_tiou_reward": 0.9747484624385834,
      "rewards/forensics_sinkhorn_reward": 0.6914392411708832,
      "rewards/forensics_strict_edge_reward": 0.047164563089609146,
      "rewards/forensics_tiou_reward": 0.4880591630935669,
      "step": 360
    },
    {
      "completion_length": 23.40625,
      "epoch": 1.10752688172043,
      "grad_norm": 7.259132385253906,
      "kl": 0.3701171875,
      "learning_rate": 7.23076923076923e-07,
      "loss": 0.2942,
      "reward": 0.749425858259201,
      "reward_std": 0.05459332466125488,
      "reward_std_all": 0.05712978541851044,
      "rewards-max/forensics_sinkhorn_reward": 0.9701041877269745,
      "rewards-max/forensics_strict_edge_reward": 0.38722455501556396,
      "rewards-max/forensics_tiou_reward": 0.9516313970088959,
      "rewards/forensics_sinkhorn_reward": 0.6527205109596252,
      "rewards/forensics_strict_edge_reward": 0.034493859857320786,
      "rewards/forensics_tiou_reward": 0.4595689922571182,
      "step": 361
    },
    {
      "completion_length": 27.65625,
      "epoch": 1.1105990783410138,
      "grad_norm": 11.235627174377441,
      "kl": 0.37890625,
      "learning_rate": 7.223076923076922e-07,
      "loss": 0.4489,
      "reward": 0.704245924949646,
      "reward_std": 0.07480945624411106,
      "reward_std_all": 0.08451301231980324,
      "rewards-max/forensics_sinkhorn_reward": 0.9573482573032379,
      "rewards-max/forensics_strict_edge_reward": 0.38750791549682617,
      "rewards-max/forensics_tiou_reward": 0.922169953584671,
      "rewards/forensics_sinkhorn_reward": 0.5884615778923035,
      "rewards/forensics_strict_edge_reward": 0.040094624273478985,
      "rewards/forensics_tiou_reward": 0.4014656990766525,
      "step": 362
    },
    {
      "completion_length": 27.1875,
      "epoch": 1.1136712749615976,
      "grad_norm": 6.431179046630859,
      "kl": 0.392578125,
      "learning_rate": 7.215384615384616e-07,
      "loss": 0.31,
      "reward": 0.7550376057624817,
      "reward_std": 0.0564802847802639,
      "reward_std_all": 0.055289603769779205,
      "rewards-max/forensics_sinkhorn_reward": 0.9550719261169434,
      "rewards-max/forensics_strict_edge_reward": 0.37587641179561615,
      "rewards-max/forensics_tiou_reward": 0.920470654964447,
      "rewards/forensics_sinkhorn_reward": 0.7010980844497681,
      "rewards/forensics_strict_edge_reward": 0.04683338291943073,
      "rewards/forensics_tiou_reward": 0.47839823365211487,
      "step": 363
    },
    {
      "completion_length": 26.21875,
      "epoch": 1.1167434715821813,
      "grad_norm": 7.556069850921631,
      "kl": 0.626953125,
      "learning_rate": 7.207692307692307e-07,
      "loss": 0.1406,
      "reward": 0.7224821150302887,
      "reward_std": 0.07700434699654579,
      "reward_std_all": 0.08068297058343887,
      "rewards-max/forensics_sinkhorn_reward": 0.9697921276092529,
      "rewards-max/forensics_strict_edge_reward": 0.44084134697914124,
      "rewards-max/forensics_tiou_reward": 0.9500086009502411,
      "rewards/forensics_sinkhorn_reward": 0.6230516731739044,
      "rewards/forensics_strict_edge_reward": 0.03220092132687569,
      "rewards/forensics_tiou_reward": 0.42053452134132385,
      "step": 364
    },
    {
      "completion_length": 22.140625,
      "epoch": 1.119815668202765,
      "grad_norm": 8.216044425964355,
      "kl": 0.4736328125,
      "learning_rate": 7.2e-07,
      "loss": 0.0637,
      "reward": 0.7561615705490112,
      "reward_std": 0.025889011099934578,
      "reward_std_all": 0.03884538821876049,
      "rewards-max/forensics_sinkhorn_reward": 0.9629258811473846,
      "rewards-max/forensics_strict_edge_reward": 0.3938451409339905,
      "rewards-max/forensics_tiou_reward": 0.9378785192966461,
      "rewards/forensics_sinkhorn_reward": 0.7081233859062195,
      "rewards/forensics_strict_edge_reward": 0.06848330982029438,
      "rewards/forensics_tiou_reward": 0.4349363297224045,
      "step": 365
    },
    {
      "completion_length": 22.46875,
      "epoch": 1.1228878648233487,
      "grad_norm": 11.43066692352295,
      "kl": 0.4228515625,
      "learning_rate": 7.192307692307692e-07,
      "loss": -0.0448,
      "reward": 0.7419189214706421,
      "reward_std": 0.033800575882196426,
      "reward_std_all": 0.046390374191105366,
      "rewards-max/forensics_sinkhorn_reward": 0.9653492271900177,
      "rewards-max/forensics_strict_edge_reward": 0.40081964433193207,
      "rewards-max/forensics_tiou_reward": 0.9354577660560608,
      "rewards/forensics_sinkhorn_reward": 0.6804967820644379,
      "rewards/forensics_strict_edge_reward": 0.028377532958984375,
      "rewards/forensics_tiou_reward": 0.4418223202228546,
      "step": 366
    },
    {
      "completion_length": 28.96875,
      "epoch": 1.1259600614439325,
      "grad_norm": 7.639456748962402,
      "kl": 0.443359375,
      "learning_rate": 7.184615384615384e-07,
      "loss": 0.2842,
      "reward": 0.6772345900535583,
      "reward_std": 0.06163043435662985,
      "reward_std_all": 0.08990811929106712,
      "rewards-max/forensics_sinkhorn_reward": 0.9668143093585968,
      "rewards-max/forensics_strict_edge_reward": 0.3846660181879997,
      "rewards-max/forensics_tiou_reward": 0.9416136741638184,
      "rewards/forensics_sinkhorn_reward": 0.5384190082550049,
      "rewards/forensics_strict_edge_reward": 0.02971021458506584,
      "rewards/forensics_tiou_reward": 0.3684868812561035,
      "step": 367
    },
    {
      "completion_length": 26.890625,
      "epoch": 1.129032258064516,
      "grad_norm": 10.721099853515625,
      "kl": 0.3779296875,
      "learning_rate": 7.176923076923077e-07,
      "loss": 0.2847,
      "reward": 0.7441534698009491,
      "reward_std": 0.05080906022340059,
      "reward_std_all": 0.0579607542604208,
      "rewards-max/forensics_sinkhorn_reward": 0.9649240076541901,
      "rewards-max/forensics_strict_edge_reward": 0.5493997037410736,
      "rewards-max/forensics_tiou_reward": 0.9431999027729034,
      "rewards/forensics_sinkhorn_reward": 0.629025399684906,
      "rewards/forensics_strict_edge_reward": 0.03365749679505825,
      "rewards/forensics_tiou_reward": 0.36731094121932983,
      "step": 368
    },
    {
      "completion_length": 25.890625,
      "epoch": 1.1321044546850998,
      "grad_norm": 7.502996444702148,
      "kl": 0.46484375,
      "learning_rate": 7.169230769230768e-07,
      "loss": 0.6648,
      "reward": 0.69427689909935,
      "reward_std": 0.04737398959696293,
      "reward_std_all": 0.07397659122943878,
      "rewards-max/forensics_sinkhorn_reward": 0.9785566031932831,
      "rewards-max/forensics_strict_edge_reward": 0.04468680638819933,
      "rewards-max/forensics_tiou_reward": 0.963026762008667,
      "rewards/forensics_sinkhorn_reward": 0.5076618194580078,
      "rewards/forensics_strict_edge_reward": 0.0032727973302826285,
      "rewards/forensics_tiou_reward": 0.344005823135376,
      "step": 369
    },
    {
      "completion_length": 23.671875,
      "epoch": 1.1351766513056836,
      "grad_norm": 12.063202857971191,
      "kl": 0.4306640625,
      "learning_rate": 7.161538461538461e-07,
      "loss": 0.1429,
      "reward": 0.7522335350513458,
      "reward_std": 0.04263170063495636,
      "reward_std_all": 0.04848555475473404,
      "rewards-max/forensics_sinkhorn_reward": 0.9588699340820312,
      "rewards-max/forensics_strict_edge_reward": 0.41313453018665314,
      "rewards-max/forensics_tiou_reward": 0.9227710962295532,
      "rewards/forensics_sinkhorn_reward": 0.6421920955181122,
      "rewards/forensics_strict_edge_reward": 0.03584033763036132,
      "rewards/forensics_tiou_reward": 0.4238757938146591,
      "step": 370
    },
    {
      "completion_length": 22.796875,
      "epoch": 1.1382488479262673,
      "grad_norm": 8.096929550170898,
      "kl": 0.41796875,
      "learning_rate": 7.153846153846154e-07,
      "loss": 0.0669,
      "reward": 0.7328611016273499,
      "reward_std": 0.033108160831034184,
      "reward_std_all": 0.05568409524857998,
      "rewards-max/forensics_sinkhorn_reward": 0.9619988203048706,
      "rewards-max/forensics_strict_edge_reward": 0.44371652603149414,
      "rewards-max/forensics_tiou_reward": 0.9235460162162781,
      "rewards/forensics_sinkhorn_reward": 0.648949921131134,
      "rewards/forensics_strict_edge_reward": 0.046413519885391,
      "rewards/forensics_tiou_reward": 0.43484827876091003,
      "step": 371
    },
    {
      "completion_length": 27.140625,
      "epoch": 1.141321044546851,
      "grad_norm": 10.379178047180176,
      "kl": 0.361328125,
      "learning_rate": 7.146153846153846e-07,
      "loss": 0.5452,
      "reward": 0.6953640878200531,
      "reward_std": 0.10225540399551392,
      "reward_std_all": 0.10751906037330627,
      "rewards-max/forensics_sinkhorn_reward": 0.9763643443584442,
      "rewards-max/forensics_strict_edge_reward": 0.435063898563385,
      "rewards-max/forensics_tiou_reward": 0.9641026556491852,
      "rewards/forensics_sinkhorn_reward": 0.5972251892089844,
      "rewards/forensics_strict_edge_reward": 0.03573988098651171,
      "rewards/forensics_tiou_reward": 0.40975816547870636,
      "step": 372
    },
    {
      "completion_length": 25.953125,
      "epoch": 1.1443932411674347,
      "grad_norm": 8.235209465026855,
      "kl": 0.51171875,
      "learning_rate": 7.138461538461538e-07,
      "loss": 0.3072,
      "reward": 0.7353129982948303,
      "reward_std": 0.05593625828623772,
      "reward_std_all": 0.06358929350972176,
      "rewards-max/forensics_sinkhorn_reward": 0.9711468517780304,
      "rewards-max/forensics_strict_edge_reward": 0.49232783913612366,
      "rewards-max/forensics_tiou_reward": 0.96223184466362,
      "rewards/forensics_sinkhorn_reward": 0.6063202321529388,
      "rewards/forensics_strict_edge_reward": 0.04886683076620102,
      "rewards/forensics_tiou_reward": 0.4535215198993683,
      "step": 373
    },
    {
      "completion_length": 24.46875,
      "epoch": 1.1474654377880185,
      "grad_norm": 9.26113510131836,
      "kl": 0.5849609375,
      "learning_rate": 7.13076923076923e-07,
      "loss": 0.1881,
      "reward": 0.7248422801494598,
      "reward_std": 0.029108926653862,
      "reward_std_all": 0.05361328274011612,
      "rewards-max/forensics_sinkhorn_reward": 0.9784978926181793,
      "rewards-max/forensics_strict_edge_reward": 0.6824378371238708,
      "rewards-max/forensics_tiou_reward": 0.9747187197208405,
      "rewards/forensics_sinkhorn_reward": 0.6279168725013733,
      "rewards/forensics_strict_edge_reward": 0.06210637465119362,
      "rewards/forensics_tiou_reward": 0.3946201503276825,
      "step": 374
    },
    {
      "completion_length": 27.078125,
      "epoch": 1.1505376344086022,
      "grad_norm": 10.290824890136719,
      "kl": 0.3427734375,
      "learning_rate": 7.123076923076922e-07,
      "loss": 0.0887,
      "reward": 0.7255149781703949,
      "reward_std": 0.03922307211905718,
      "reward_std_all": 0.05801551416516304,
      "rewards-max/forensics_sinkhorn_reward": 0.9534409046173096,
      "rewards-max/forensics_strict_edge_reward": 0.3223366439342499,
      "rewards-max/forensics_tiou_reward": 0.9063718020915985,
      "rewards/forensics_sinkhorn_reward": 0.6456587314605713,
      "rewards/forensics_strict_edge_reward": 0.04243287397548556,
      "rewards/forensics_tiou_reward": 0.4472058117389679,
      "step": 375
    },
    {
      "completion_length": 27.015625,
      "epoch": 1.1536098310291858,
      "grad_norm": 8.851548194885254,
      "kl": 0.4111328125,
      "learning_rate": 7.115384615384616e-07,
      "loss": 0.3327,
      "reward": 0.6716160476207733,
      "reward_std": 0.07972955703735352,
      "reward_std_all": 0.1022631898522377,
      "rewards-max/forensics_sinkhorn_reward": 0.9739818871021271,
      "rewards-max/forensics_strict_edge_reward": 0.5404154062271118,
      "rewards-max/forensics_tiou_reward": 0.9629254937171936,
      "rewards/forensics_sinkhorn_reward": 0.5645211935043335,
      "rewards/forensics_strict_edge_reward": 0.05227397941052914,
      "rewards/forensics_tiou_reward": 0.42532485723495483,
      "step": 376
    },
    {
      "completion_length": 25.921875,
      "epoch": 1.1566820276497696,
      "grad_norm": 7.41989278793335,
      "kl": 0.7705078125,
      "learning_rate": 7.107692307692308e-07,
      "loss": 0.2954,
      "reward": 0.7513752579689026,
      "reward_std": 0.041669223457574844,
      "reward_std_all": 0.04881153628230095,
      "rewards-max/forensics_sinkhorn_reward": 0.9725787043571472,
      "rewards-max/forensics_strict_edge_reward": 0.5413877367973328,
      "rewards-max/forensics_tiou_reward": 0.9561901688575745,
      "rewards/forensics_sinkhorn_reward": 0.6345208883285522,
      "rewards/forensics_strict_edge_reward": 0.03676585014909506,
      "rewards/forensics_tiou_reward": 0.41368281841278076,
      "step": 377
    },
    {
      "completion_length": 23.59375,
      "epoch": 1.1597542242703534,
      "grad_norm": 16.127742767333984,
      "kl": 0.82421875,
      "learning_rate": 7.1e-07,
      "loss": 0.1235,
      "reward": 0.7669001221656799,
      "reward_std": 0.034057799726724625,
      "reward_std_all": 0.04048597440123558,
      "rewards-max/forensics_sinkhorn_reward": 0.9650785624980927,
      "rewards-max/forensics_strict_edge_reward": 0.4092717915773392,
      "rewards-max/forensics_tiou_reward": 0.9416736662387848,
      "rewards/forensics_sinkhorn_reward": 0.713068425655365,
      "rewards/forensics_strict_edge_reward": 0.0443891417235136,
      "rewards/forensics_tiou_reward": 0.5080675184726715,
      "step": 378
    },
    {
      "completion_length": 22.828125,
      "epoch": 1.1628264208909371,
      "grad_norm": 11.494624137878418,
      "kl": 0.806640625,
      "learning_rate": 7.092307692307692e-07,
      "loss": -0.0831,
      "reward": 0.7359397113323212,
      "reward_std": 0.0579571183770895,
      "reward_std_all": 0.06209065206348896,
      "rewards-max/forensics_sinkhorn_reward": 0.9363455772399902,
      "rewards-max/forensics_strict_edge_reward": 0.42371489852666855,
      "rewards-max/forensics_tiou_reward": 0.8800764381885529,
      "rewards/forensics_sinkhorn_reward": 0.615137368440628,
      "rewards/forensics_strict_edge_reward": 0.03202784061431885,
      "rewards/forensics_tiou_reward": 0.39685890078544617,
      "step": 379
    },
    {
      "completion_length": 20.9375,
      "epoch": 1.1658986175115207,
      "grad_norm": 7.639415740966797,
      "kl": 0.53125,
      "learning_rate": 7.084615384615384e-07,
      "loss": -0.0063,
      "reward": 0.724777340888977,
      "reward_std": 0.0713751595467329,
      "reward_std_all": 0.07491854205727577,
      "rewards-max/forensics_sinkhorn_reward": 0.9782154858112335,
      "rewards-max/forensics_strict_edge_reward": 0.44491301476955414,
      "rewards-max/forensics_tiou_reward": 0.9700110554695129,
      "rewards/forensics_sinkhorn_reward": 0.6844986379146576,
      "rewards/forensics_strict_edge_reward": 0.0358501560986042,
      "rewards/forensics_tiou_reward": 0.4480496942996979,
      "step": 380
    },
    {
      "completion_length": 23.546875,
      "epoch": 1.1689708141321045,
      "grad_norm": 8.685205459594727,
      "kl": 0.3515625,
      "learning_rate": 7.076923076923077e-07,
      "loss": 0.2084,
      "reward": 0.7237232029438019,
      "reward_std": 0.043766969814896584,
      "reward_std_all": 0.062264321371912956,
      "rewards-max/forensics_sinkhorn_reward": 0.97038134932518,
      "rewards-max/forensics_strict_edge_reward": 0.49555233120918274,
      "rewards-max/forensics_tiou_reward": 0.9567206501960754,
      "rewards/forensics_sinkhorn_reward": 0.6591762006282806,
      "rewards/forensics_strict_edge_reward": 0.034669702872633934,
      "rewards/forensics_tiou_reward": 0.4188540130853653,
      "step": 381
    },
    {
      "completion_length": 22.703125,
      "epoch": 1.1720430107526882,
      "grad_norm": 10.520283699035645,
      "kl": 0.4541015625,
      "learning_rate": 7.069230769230768e-07,
      "loss": 0.149,
      "reward": 0.7063935399055481,
      "reward_std": 0.06066183000802994,
      "reward_std_all": 0.07945004105567932,
      "rewards-max/forensics_sinkhorn_reward": 0.9706220030784607,
      "rewards-max/forensics_strict_edge_reward": 0.4538624435663223,
      "rewards-max/forensics_tiou_reward": 0.9515982866287231,
      "rewards/forensics_sinkhorn_reward": 0.6065117418766022,
      "rewards/forensics_strict_edge_reward": 0.021993019618093967,
      "rewards/forensics_tiou_reward": 0.3999011516571045,
      "step": 382
    },
    {
      "completion_length": 24.890625,
      "epoch": 1.1751152073732718,
      "grad_norm": 8.170886039733887,
      "kl": 0.3740234375,
      "learning_rate": 7.061538461538462e-07,
      "loss": 0.2569,
      "reward": 0.7659257352352142,
      "reward_std": 0.03102712519466877,
      "reward_std_all": 0.036911845207214355,
      "rewards-max/forensics_sinkhorn_reward": 0.9744243025779724,
      "rewards-max/forensics_strict_edge_reward": 0.6824378371238708,
      "rewards-max/forensics_tiou_reward": 0.966713547706604,
      "rewards/forensics_sinkhorn_reward": 0.6695897281169891,
      "rewards/forensics_strict_edge_reward": 0.06565830484032631,
      "rewards/forensics_tiou_reward": 0.4929352253675461,
      "step": 383
    },
    {
      "completion_length": 27.078125,
      "epoch": 1.1781874039938556,
      "grad_norm": 8.186261177062988,
      "kl": 0.3388671875,
      "learning_rate": 7.053846153846154e-07,
      "loss": 0.305,
      "reward": 0.7100338935852051,
      "reward_std": 0.05556943453848362,
      "reward_std_all": 0.07219532504677773,
      "rewards-max/forensics_sinkhorn_reward": 0.9805449843406677,
      "rewards-max/forensics_strict_edge_reward": 0.46424447000026703,
      "rewards-max/forensics_tiou_reward": 0.9782147705554962,
      "rewards/forensics_sinkhorn_reward": 0.6002281904220581,
      "rewards/forensics_strict_edge_reward": 0.030002796091139317,
      "rewards/forensics_tiou_reward": 0.4243834912776947,
      "step": 384
    },
    {
      "completion_length": 21.796875,
      "epoch": 1.1812596006144394,
      "grad_norm": 7.50511360168457,
      "kl": 0.3212890625,
      "learning_rate": 7.046153846153846e-07,
      "loss": 0.0315,
      "reward": 0.7332686185836792,
      "reward_std": 0.051023880019783974,
      "reward_std_all": 0.06393202394247055,
      "rewards-max/forensics_sinkhorn_reward": 0.9686526656150818,
      "rewards-max/forensics_strict_edge_reward": 0.5072369128465652,
      "rewards-max/forensics_tiou_reward": 0.9495969116687775,
      "rewards/forensics_sinkhorn_reward": 0.6730931401252747,
      "rewards/forensics_strict_edge_reward": 0.05354314483702183,
      "rewards/forensics_tiou_reward": 0.43975530564785004,
      "step": 385
    },
    {
      "completion_length": 21.6875,
      "epoch": 1.1843317972350231,
      "grad_norm": 9.416152954101562,
      "kl": 0.5166015625,
      "learning_rate": 7.038461538461539e-07,
      "loss": -0.0413,
      "reward": 0.7014324069023132,
      "reward_std": 0.048774316906929016,
      "reward_std_all": 0.07511243969202042,
      "rewards-max/forensics_sinkhorn_reward": 0.9721862077713013,
      "rewards-max/forensics_strict_edge_reward": 0.6499901115894318,
      "rewards-max/forensics_tiou_reward": 0.9571697413921356,
      "rewards/forensics_sinkhorn_reward": 0.6373586058616638,
      "rewards/forensics_strict_edge_reward": 0.04688326083123684,
      "rewards/forensics_tiou_reward": 0.39166614413261414,
      "step": 386
    },
    {
      "completion_length": 22.828125,
      "epoch": 1.1874039938556067,
      "grad_norm": 13.899160385131836,
      "kl": 0.40234375,
      "learning_rate": 7.03076923076923e-07,
      "loss": 0.4232,
      "reward": 0.7251668870449066,
      "reward_std": 0.07644667103886604,
      "reward_std_all": 0.079881701618433,
      "rewards-max/forensics_sinkhorn_reward": 0.9741111099720001,
      "rewards-max/forensics_strict_edge_reward": 0.5550410598516464,
      "rewards-max/forensics_tiou_reward": 0.9574251472949982,
      "rewards/forensics_sinkhorn_reward": 0.6529212594032288,
      "rewards/forensics_strict_edge_reward": 0.03734648786485195,
      "rewards/forensics_tiou_reward": 0.3855222463607788,
      "step": 387
    },
    {
      "completion_length": 22.25,
      "epoch": 1.1904761904761905,
      "grad_norm": 7.194213390350342,
      "kl": 0.3349609375,
      "learning_rate": 7.023076923076922e-07,
      "loss": 0.2162,
      "reward": 0.7802532315254211,
      "reward_std": 0.039605654776096344,
      "reward_std_all": 0.037140337750315666,
      "rewards-max/forensics_sinkhorn_reward": 0.9693228900432587,
      "rewards-max/forensics_strict_edge_reward": 0.5900419056415558,
      "rewards-max/forensics_tiou_reward": 0.9452542960643768,
      "rewards/forensics_sinkhorn_reward": 0.7314608693122864,
      "rewards/forensics_strict_edge_reward": 0.057005349546670914,
      "rewards/forensics_tiou_reward": 0.49492427706718445,
      "step": 388
    },
    {
      "completion_length": 24.796875,
      "epoch": 1.1935483870967742,
      "grad_norm": 9.769319534301758,
      "kl": 0.568359375,
      "learning_rate": 7.015384615384615e-07,
      "loss": 0.425,
      "reward": 0.7028170526027679,
      "reward_std": 0.057155415415763855,
      "reward_std_all": 0.07945867627859116,
      "rewards-max/forensics_sinkhorn_reward": 0.966462254524231,
      "rewards-max/forensics_strict_edge_reward": 0.16973301768302917,
      "rewards-max/forensics_tiou_reward": 0.9436717927455902,
      "rewards/forensics_sinkhorn_reward": 0.6091645956039429,
      "rewards/forensics_strict_edge_reward": 0.015559525229036808,
      "rewards/forensics_tiou_reward": 0.3361925035715103,
      "step": 389
    },
    {
      "completion_length": 24.28125,
      "epoch": 1.1966205837173578,
      "grad_norm": 5.891523361206055,
      "kl": 0.4072265625,
      "learning_rate": 7.007692307692308e-07,
      "loss": 0.3936,
      "reward": 0.7432433366775513,
      "reward_std": 0.07125844061374664,
      "reward_std_all": 0.06870242208242416,
      "rewards-max/forensics_sinkhorn_reward": 0.9740346074104309,
      "rewards-max/forensics_strict_edge_reward": 0.3245256543159485,
      "rewards-max/forensics_tiou_reward": 0.951018214225769,
      "rewards/forensics_sinkhorn_reward": 0.6612151265144348,
      "rewards/forensics_strict_edge_reward": 0.03327763080596924,
      "rewards/forensics_tiou_reward": 0.41418541967868805,
      "step": 390
    },
    {
      "completion_length": 24.328125,
      "epoch": 1.1996927803379416,
      "grad_norm": 8.15308666229248,
      "kl": 0.4326171875,
      "learning_rate": 7e-07,
      "loss": -0.0081,
      "reward": 0.7405611276626587,
      "reward_std": 0.06278225407004356,
      "reward_std_all": 0.06848520413041115,
      "rewards-max/forensics_sinkhorn_reward": 0.9738243818283081,
      "rewards-max/forensics_strict_edge_reward": 0.2973281219601631,
      "rewards-max/forensics_tiou_reward": 0.9562790095806122,
      "rewards/forensics_sinkhorn_reward": 0.6758401989936829,
      "rewards/forensics_strict_edge_reward": 0.03503248794004321,
      "rewards/forensics_tiou_reward": 0.46714216470718384,
      "step": 391
    },
    {
      "completion_length": 26.828125,
      "epoch": 1.2027649769585254,
      "grad_norm": 7.92012882232666,
      "kl": 0.4150390625,
      "learning_rate": 6.992307692307692e-07,
      "loss": 0.4219,
      "reward": 0.730207622051239,
      "reward_std": 0.043943123891949654,
      "reward_std_all": 0.060212260112166405,
      "rewards-max/forensics_sinkhorn_reward": 0.9730331897735596,
      "rewards-max/forensics_strict_edge_reward": 0.48575712740421295,
      "rewards-max/forensics_tiou_reward": 0.9570416510105133,
      "rewards/forensics_sinkhorn_reward": 0.6474034786224365,
      "rewards/forensics_strict_edge_reward": 0.04664638917893171,
      "rewards/forensics_tiou_reward": 0.4380984902381897,
      "step": 392
    },
    {
      "completion_length": 26.5,
      "epoch": 1.2058371735791091,
      "grad_norm": 7.177093029022217,
      "kl": 0.498046875,
      "learning_rate": 6.984615384615384e-07,
      "loss": 0.2542,
      "reward": 0.7013468742370605,
      "reward_std": 0.0697909714654088,
      "reward_std_all": 0.08431877568364143,
      "rewards-max/forensics_sinkhorn_reward": 0.9795741140842438,
      "rewards-max/forensics_strict_edge_reward": 0.4555841088294983,
      "rewards-max/forensics_tiou_reward": 0.9724728465080261,
      "rewards/forensics_sinkhorn_reward": 0.5968109667301178,
      "rewards/forensics_strict_edge_reward": 0.021481262519955635,
      "rewards/forensics_tiou_reward": 0.3461812734603882,
      "step": 393
    },
    {
      "completion_length": 21.34375,
      "epoch": 1.2089093701996927,
      "grad_norm": 8.370214462280273,
      "kl": 0.36328125,
      "learning_rate": 6.976923076923077e-07,
      "loss": -0.047,
      "reward": 0.7600159347057343,
      "reward_std": 0.029689996503293514,
      "reward_std_all": 0.03828869201242924,
      "rewards-max/forensics_sinkhorn_reward": 0.9720287919044495,
      "rewards-max/forensics_strict_edge_reward": 0.5404154062271118,
      "rewards-max/forensics_tiou_reward": 0.9594940841197968,
      "rewards/forensics_sinkhorn_reward": 0.7011977434158325,
      "rewards/forensics_strict_edge_reward": 0.06785587780177593,
      "rewards/forensics_tiou_reward": 0.4983924478292465,
      "step": 394
    },
    {
      "completion_length": 23.125,
      "epoch": 1.2119815668202765,
      "grad_norm": 9.000720977783203,
      "kl": 0.53125,
      "learning_rate": 6.969230769230769e-07,
      "loss": 0.0885,
      "reward": 0.7388715147972107,
      "reward_std": 0.04455965664237738,
      "reward_std_all": 0.056276097893714905,
      "rewards-max/forensics_sinkhorn_reward": 0.9520083665847778,
      "rewards-max/forensics_strict_edge_reward": 0.6536494493484497,
      "rewards-max/forensics_tiou_reward": 0.9092406034469604,
      "rewards/forensics_sinkhorn_reward": 0.6573532223701477,
      "rewards/forensics_strict_edge_reward": 0.04891707934439182,
      "rewards/forensics_tiou_reward": 0.4015294909477234,
      "step": 395
    },
    {
      "completion_length": 24.359375,
      "epoch": 1.2150537634408602,
      "grad_norm": 10.39194107055664,
      "kl": 0.439453125,
      "learning_rate": 6.961538461538461e-07,
      "loss": 0.286,
      "reward": 0.691210150718689,
      "reward_std": 0.047195034101605415,
      "reward_std_all": 0.0772707611322403,
      "rewards-max/forensics_sinkhorn_reward": 0.9606907963752747,
      "rewards-max/forensics_strict_edge_reward": 0.44084134697914124,
      "rewards-max/forensics_tiou_reward": 0.9322436451911926,
      "rewards/forensics_sinkhorn_reward": 0.590470165014267,
      "rewards/forensics_strict_edge_reward": 0.04828348010778427,
      "rewards/forensics_tiou_reward": 0.3459104597568512,
      "step": 396
    },
    {
      "completion_length": 26.296875,
      "epoch": 1.218125960061444,
      "grad_norm": 8.457493782043457,
      "kl": 0.53125,
      "learning_rate": 6.953846153846154e-07,
      "loss": 0.2554,
      "reward": 0.718974769115448,
      "reward_std": 0.023498819675296545,
      "reward_std_all": 0.05176583491265774,
      "rewards-max/forensics_sinkhorn_reward": 0.9791274070739746,
      "rewards-max/forensics_strict_edge_reward": 0.6326714754104614,
      "rewards-max/forensics_tiou_reward": 0.9737348258495331,
      "rewards/forensics_sinkhorn_reward": 0.589484304189682,
      "rewards/forensics_strict_edge_reward": 0.039482494816184044,
      "rewards/forensics_tiou_reward": 0.3641757220029831,
      "step": 397
    },
    {
      "completion_length": 22.328125,
      "epoch": 1.2211981566820276,
      "grad_norm": 10.92202091217041,
      "kl": 0.44140625,
      "learning_rate": 6.946153846153846e-07,
      "loss": 0.0863,
      "reward": 0.7576190829277039,
      "reward_std": 0.043318730778992176,
      "reward_std_all": 0.04801577143371105,
      "rewards-max/forensics_sinkhorn_reward": 0.9612368047237396,
      "rewards-max/forensics_strict_edge_reward": 0.5550410598516464,
      "rewards-max/forensics_tiou_reward": 0.9366772770881653,
      "rewards/forensics_sinkhorn_reward": 0.7072578072547913,
      "rewards/forensics_strict_edge_reward": 0.07276388257741928,
      "rewards/forensics_tiou_reward": 0.5097542107105255,
      "step": 398
    },
    {
      "completion_length": 22.609375,
      "epoch": 1.2242703533026114,
      "grad_norm": 9.799542427062988,
      "kl": 0.4658203125,
      "learning_rate": 6.938461538461539e-07,
      "loss": -0.1705,
      "reward": 0.7749230265617371,
      "reward_std": 0.0434187687933445,
      "reward_std_all": 0.04092000424861908,
      "rewards-max/forensics_sinkhorn_reward": 0.9804295599460602,
      "rewards-max/forensics_strict_edge_reward": 0.6236754059791565,
      "rewards-max/forensics_tiou_reward": 0.9803090989589691,
      "rewards/forensics_sinkhorn_reward": 0.7207445204257965,
      "rewards/forensics_strict_edge_reward": 0.06471369042992592,
      "rewards/forensics_tiou_reward": 0.5288257002830505,
      "step": 399
    },
    {
      "completion_length": 29.75,
      "epoch": 1.2273425499231951,
      "grad_norm": 6.3636651039123535,
      "kl": 0.37109375,
      "learning_rate": 6.93076923076923e-07,
      "loss": -0.0855,
      "reward": 0.7439825534820557,
      "reward_std": 0.05757230531889945,
      "reward_std_all": 0.05844243010506034,
      "rewards-max/forensics_sinkhorn_reward": 0.9585978984832764,
      "rewards-max/forensics_strict_edge_reward": 0.5433326959609985,
      "rewards-max/forensics_tiou_reward": 0.9333770573139191,
      "rewards/forensics_sinkhorn_reward": 0.665877103805542,
      "rewards/forensics_strict_edge_reward": 0.0775154996663332,
      "rewards/forensics_tiou_reward": 0.5428109765052795,
      "step": 400
    },
    {
      "completion_length": 24.5,
      "epoch": 1.230414746543779,
      "grad_norm": 7.778439521789551,
      "kl": 0.4736328125,
      "learning_rate": 6.923076923076922e-07,
      "loss": 0.1647,
      "reward": 0.7268684208393097,
      "reward_std": 0.03478834219276905,
      "reward_std_all": 0.056907206773757935,
      "rewards-max/forensics_sinkhorn_reward": 0.97856804728508,
      "rewards-max/forensics_strict_edge_reward": 0.43684302270412445,
      "rewards-max/forensics_tiou_reward": 0.972334086894989,
      "rewards/forensics_sinkhorn_reward": 0.6185241639614105,
      "rewards/forensics_strict_edge_reward": 0.044479917734861374,
      "rewards/forensics_tiou_reward": 0.43221326172351837,
      "step": 401
    },
    {
      "completion_length": 25.515625,
      "epoch": 1.2334869431643625,
      "grad_norm": 7.564115047454834,
      "kl": 0.32421875,
      "learning_rate": 6.915384615384615e-07,
      "loss": 0.0205,
      "reward": 0.7342462539672852,
      "reward_std": 0.0811147429049015,
      "reward_std_all": 0.07775668520480394,
      "rewards-max/forensics_sinkhorn_reward": 0.9427378475666046,
      "rewards-max/forensics_strict_edge_reward": 0.3340276926755905,
      "rewards-max/forensics_tiou_reward": 0.8924407958984375,
      "rewards/forensics_sinkhorn_reward": 0.6344722807407379,
      "rewards/forensics_strict_edge_reward": 0.041339475428685546,
      "rewards/forensics_tiou_reward": 0.37542977929115295,
      "step": 402
    },
    {
      "completion_length": 23.71875,
      "epoch": 1.2365591397849462,
      "grad_norm": 8.845738410949707,
      "kl": 0.5498046875,
      "learning_rate": 6.907692307692307e-07,
      "loss": -0.0154,
      "reward": 0.7748509347438812,
      "reward_std": 0.021926602348685265,
      "reward_std_all": 0.027554095722734928,
      "rewards-max/forensics_sinkhorn_reward": 0.9616283178329468,
      "rewards-max/forensics_strict_edge_reward": 0.258957639336586,
      "rewards-max/forensics_tiou_reward": 0.9380612671375275,
      "rewards/forensics_sinkhorn_reward": 0.6785073578357697,
      "rewards/forensics_strict_edge_reward": 0.02703909995034337,
      "rewards/forensics_tiou_reward": 0.41923634707927704,
      "step": 403
    },
    {
      "completion_length": 25.609375,
      "epoch": 1.23963133640553,
      "grad_norm": 7.782578468322754,
      "kl": 0.3984375,
      "learning_rate": 6.9e-07,
      "loss": 0.0727,
      "reward": 0.7595896422863007,
      "reward_std": 0.02902725338935852,
      "reward_std_all": 0.037465726025402546,
      "rewards-max/forensics_sinkhorn_reward": 0.9744422137737274,
      "rewards-max/forensics_strict_edge_reward": 0.411786213517189,
      "rewards-max/forensics_tiou_reward": 0.9600937366485596,
      "rewards/forensics_sinkhorn_reward": 0.6540349721908569,
      "rewards/forensics_strict_edge_reward": 0.03504706360399723,
      "rewards/forensics_tiou_reward": 0.4395977854728699,
      "step": 404
    },
    {
      "completion_length": 24.9375,
      "epoch": 1.2427035330261136,
      "grad_norm": 8.258418083190918,
      "kl": 0.388671875,
      "learning_rate": 6.892307692307692e-07,
      "loss": 0.0436,
      "reward": 0.72025927901268,
      "reward_std": 0.055045200511813164,
      "reward_std_all": 0.07150095328688622,
      "rewards-max/forensics_sinkhorn_reward": 0.9716452062129974,
      "rewards-max/forensics_strict_edge_reward": 0.5230375230312347,
      "rewards-max/forensics_tiou_reward": 0.9546147286891937,
      "rewards/forensics_sinkhorn_reward": 0.6764965355396271,
      "rewards/forensics_strict_edge_reward": 0.031014326959848404,
      "rewards/forensics_tiou_reward": 0.46519410610198975,
      "step": 405
    },
    {
      "completion_length": 22.671875,
      "epoch": 1.2457757296466974,
      "grad_norm": 6.899761199951172,
      "kl": 0.546875,
      "learning_rate": 6.884615384615384e-07,
      "loss": 0.0876,
      "reward": 0.7460108399391174,
      "reward_std": 0.04731729440391064,
      "reward_std_all": 0.05347694642841816,
      "rewards-max/forensics_sinkhorn_reward": 0.9661462903022766,
      "rewards-max/forensics_strict_edge_reward": 0.41527149081230164,
      "rewards-max/forensics_tiou_reward": 0.9257379174232483,
      "rewards/forensics_sinkhorn_reward": 0.7143835425376892,
      "rewards/forensics_strict_edge_reward": 0.04021025262773037,
      "rewards/forensics_tiou_reward": 0.49090924859046936,
      "step": 406
    },
    {
      "completion_length": 23.421875,
      "epoch": 1.2488479262672811,
      "grad_norm": 7.409153461456299,
      "kl": 0.3076171875,
      "learning_rate": 6.876923076923077e-07,
      "loss": 0.118,
      "reward": 0.7236951291561127,
      "reward_std": 0.04179948940873146,
      "reward_std_all": 0.06125678867101669,
      "rewards-max/forensics_sinkhorn_reward": 0.9734441339969635,
      "rewards-max/forensics_strict_edge_reward": 0.4736122563481331,
      "rewards-max/forensics_tiou_reward": 0.9557122588157654,
      "rewards/forensics_sinkhorn_reward": 0.650139182806015,
      "rewards/forensics_strict_edge_reward": 0.02831873018294573,
      "rewards/forensics_tiou_reward": 0.3782506287097931,
      "step": 407
    },
    {
      "completion_length": 25.515625,
      "epoch": 1.251920122887865,
      "grad_norm": 7.224094867706299,
      "kl": 0.4140625,
      "learning_rate": 6.869230769230769e-07,
      "loss": 0.5383,
      "reward": 0.7086615562438965,
      "reward_std": 0.0660974346101284,
      "reward_std_all": 0.08094129711389542,
      "rewards-max/forensics_sinkhorn_reward": 0.9603358507156372,
      "rewards-max/forensics_strict_edge_reward": 0.5132495760917664,
      "rewards-max/forensics_tiou_reward": 0.9176158308982849,
      "rewards/forensics_sinkhorn_reward": 0.5746178925037384,
      "rewards/forensics_strict_edge_reward": 0.039289627224206924,
      "rewards/forensics_tiou_reward": 0.3589554876089096,
      "step": 408
    },
    {
      "completion_length": 24.59375,
      "epoch": 1.2549923195084485,
      "grad_norm": 9.189226150512695,
      "kl": 0.43359375,
      "learning_rate": 6.861538461538461e-07,
      "loss": 0.5396,
      "reward": 0.7016341984272003,
      "reward_std": 0.08060689643025398,
      "reward_std_all": 0.08876851946115494,
      "rewards-max/forensics_sinkhorn_reward": 0.9492961466312408,
      "rewards-max/forensics_strict_edge_reward": 0.40749260783195496,
      "rewards-max/forensics_tiou_reward": 0.901716560125351,
      "rewards/forensics_sinkhorn_reward": 0.600368320941925,
      "rewards/forensics_strict_edge_reward": 0.02816020743921399,
      "rewards/forensics_tiou_reward": 0.3959873914718628,
      "step": 409
    },
    {
      "completion_length": 23.828125,
      "epoch": 1.2580645161290323,
      "grad_norm": 12.040114402770996,
      "kl": 0.60107421875,
      "learning_rate": 6.853846153846153e-07,
      "loss": -0.1039,
      "reward": 0.7444164752960205,
      "reward_std": 0.01781076076440513,
      "reward_std_all": 0.04137174040079117,
      "rewards-max/forensics_sinkhorn_reward": 0.9719636738300323,
      "rewards-max/forensics_strict_edge_reward": 0.45597073435783386,
      "rewards-max/forensics_tiou_reward": 0.9532131254673004,
      "rewards/forensics_sinkhorn_reward": 0.6736637949943542,
      "rewards/forensics_strict_edge_reward": 0.05921022407710552,
      "rewards/forensics_tiou_reward": 0.43585920333862305,
      "step": 410
    },
    {
      "completion_length": 27.15625,
      "epoch": 1.261136712749616,
      "grad_norm": 9.200875282287598,
      "kl": 0.361328125,
      "learning_rate": 6.846153846153846e-07,
      "loss": 0.6726,
      "reward": 0.7238431870937347,
      "reward_std": 0.07240470871329308,
      "reward_std_all": 0.07717863470315933,
      "rewards-max/forensics_sinkhorn_reward": 0.9771661162376404,
      "rewards-max/forensics_strict_edge_reward": 0.4231441020965576,
      "rewards-max/forensics_tiou_reward": 0.967311829328537,
      "rewards/forensics_sinkhorn_reward": 0.6169688403606415,
      "rewards/forensics_strict_edge_reward": 0.03554191812872887,
      "rewards/forensics_tiou_reward": 0.43185843527317047,
      "step": 411
    },
    {
      "completion_length": 26.890625,
      "epoch": 1.2642089093701996,
      "grad_norm": 8.501766204833984,
      "kl": 0.404296875,
      "learning_rate": 6.838461538461539e-07,
      "loss": 0.2609,
      "reward": 0.7283974885940552,
      "reward_std": 0.029983963817358017,
      "reward_std_all": 0.052085068076848984,
      "rewards-max/forensics_sinkhorn_reward": 0.9638689756393433,
      "rewards-max/forensics_strict_edge_reward": 0.1977136880159378,
      "rewards-max/forensics_tiou_reward": 0.9316089451313019,
      "rewards/forensics_sinkhorn_reward": 0.6206145882606506,
      "rewards/forensics_strict_edge_reward": 0.014140257146209478,
      "rewards/forensics_tiou_reward": 0.4330779016017914,
      "step": 412
    },
    {
      "completion_length": 24.796875,
      "epoch": 1.2672811059907834,
      "grad_norm": 9.803497314453125,
      "kl": 0.32421875,
      "learning_rate": 6.83076923076923e-07,
      "loss": 0.0606,
      "reward": 0.7107594907283783,
      "reward_std": 0.03795023821294308,
      "reward_std_all": 0.06311127170920372,
      "rewards-max/forensics_sinkhorn_reward": 0.9747052490711212,
      "rewards-max/forensics_strict_edge_reward": 0.47400766611099243,
      "rewards-max/forensics_tiou_reward": 0.9570774734020233,
      "rewards/forensics_sinkhorn_reward": 0.6229256987571716,
      "rewards/forensics_strict_edge_reward": 0.05041766166687012,
      "rewards/forensics_tiou_reward": 0.4336320161819458,
      "step": 413
    },
    {
      "completion_length": 22.796875,
      "epoch": 1.2703533026113671,
      "grad_norm": 8.998908042907715,
      "kl": 0.4677734375,
      "learning_rate": 6.823076923076922e-07,
      "loss": 0.3971,
      "reward": 0.6578530669212341,
      "reward_std": 0.09562119469046593,
      "reward_std_all": 0.11703650653362274,
      "rewards-max/forensics_sinkhorn_reward": 0.9700630903244019,
      "rewards-max/forensics_strict_edge_reward": 0.45270346850156784,
      "rewards-max/forensics_tiou_reward": 0.9498468041419983,
      "rewards/forensics_sinkhorn_reward": 0.5710904002189636,
      "rewards/forensics_strict_edge_reward": 0.04123315401375294,
      "rewards/forensics_tiou_reward": 0.34343473613262177,
      "step": 414
    },
    {
      "completion_length": 30.640625,
      "epoch": 1.273425499231951,
      "grad_norm": 5.928400993347168,
      "kl": 0.4267578125,
      "learning_rate": 6.815384615384615e-07,
      "loss": 0.1933,
      "reward": 0.7410317361354828,
      "reward_std": 0.04127793200314045,
      "reward_std_all": 0.0541912205517292,
      "rewards-max/forensics_sinkhorn_reward": 0.9822025001049042,
      "rewards-max/forensics_strict_edge_reward": 0.3939525932073593,
      "rewards-max/forensics_tiou_reward": 0.9809406995773315,
      "rewards/forensics_sinkhorn_reward": 0.6150864958763123,
      "rewards/forensics_strict_edge_reward": 0.04452706128358841,
      "rewards/forensics_tiou_reward": 0.48091956973075867,
      "step": 415
    },
    {
      "completion_length": 25.3125,
      "epoch": 1.2764976958525347,
      "grad_norm": 9.739285469055176,
      "kl": 0.447265625,
      "learning_rate": 6.807692307692307e-07,
      "loss": 0.5088,
      "reward": 0.7231851518154144,
      "reward_std": 0.06343104690313339,
      "reward_std_all": 0.07561991363763809,
      "rewards-max/forensics_sinkhorn_reward": 0.9487358331680298,
      "rewards-max/forensics_strict_edge_reward": 0.25596073269844055,
      "rewards-max/forensics_tiou_reward": 0.9012022614479065,
      "rewards/forensics_sinkhorn_reward": 0.5736642181873322,
      "rewards/forensics_strict_edge_reward": 0.03232367243617773,
      "rewards/forensics_tiou_reward": 0.396337628364563,
      "step": 416
    },
    {
      "completion_length": 26.5625,
      "epoch": 1.2795698924731183,
      "grad_norm": 8.520977020263672,
      "kl": 0.333984375,
      "learning_rate": 6.800000000000001e-07,
      "loss": 0.1474,
      "reward": 0.6697049736976624,
      "reward_std": 0.04461165866814554,
      "reward_std_all": 0.08618681505322456,
      "rewards-max/forensics_sinkhorn_reward": 0.9740351736545563,
      "rewards-max/forensics_strict_edge_reward": 0.5006151795387268,
      "rewards-max/forensics_tiou_reward": 0.9614725410938263,
      "rewards/forensics_sinkhorn_reward": 0.5879684686660767,
      "rewards/forensics_strict_edge_reward": 0.03837294317781925,
      "rewards/forensics_tiou_reward": 0.35448959469795227,
      "step": 417
    },
    {
      "completion_length": 24.84375,
      "epoch": 1.282642089093702,
      "grad_norm": 7.8792338371276855,
      "kl": 0.3984375,
      "learning_rate": 6.792307692307692e-07,
      "loss": 0.4371,
      "reward": 0.7425399422645569,
      "reward_std": 0.06558779627084732,
      "reward_std_all": 0.06695323809981346,
      "rewards-max/forensics_sinkhorn_reward": 0.929701030254364,
      "rewards-max/forensics_strict_edge_reward": 0.21054257079958916,
      "rewards-max/forensics_tiou_reward": 0.8488752841949463,
      "rewards/forensics_sinkhorn_reward": 0.6054241061210632,
      "rewards/forensics_strict_edge_reward": 0.011961011216044426,
      "rewards/forensics_tiou_reward": 0.4379001259803772,
      "step": 418
    },
    {
      "completion_length": 21.53125,
      "epoch": 1.2857142857142856,
      "grad_norm": 10.121471405029297,
      "kl": 0.49609375,
      "learning_rate": 6.784615384615384e-07,
      "loss": 0.3721,
      "reward": 0.7731655538082123,
      "reward_std": 0.03791062720119953,
      "reward_std_all": 0.037458499893546104,
      "rewards-max/forensics_sinkhorn_reward": 0.9631558954715729,
      "rewards-max/forensics_strict_edge_reward": 0.477401964366436,
      "rewards-max/forensics_tiou_reward": 0.9402468800544739,
      "rewards/forensics_sinkhorn_reward": 0.6693347990512848,
      "rewards/forensics_strict_edge_reward": 0.02653572428971529,
      "rewards/forensics_tiou_reward": 0.4179767668247223,
      "step": 419
    },
    {
      "completion_length": 27.515625,
      "epoch": 1.2887864823348694,
      "grad_norm": 9.146854400634766,
      "kl": 0.712890625,
      "learning_rate": 6.776923076923077e-07,
      "loss": 0.6445,
      "reward": 0.6529355049133301,
      "reward_std": 0.10065440461039543,
      "reward_std_all": 0.12378286570310593,
      "rewards-max/forensics_sinkhorn_reward": 0.9733733534812927,
      "rewards-max/forensics_strict_edge_reward": 0.2946265935897827,
      "rewards-max/forensics_tiou_reward": 0.9555570185184479,
      "rewards/forensics_sinkhorn_reward": 0.5470793843269348,
      "rewards/forensics_strict_edge_reward": 0.013597304932773113,
      "rewards/forensics_tiou_reward": 0.3160362243652344,
      "step": 420
    },
    {
      "completion_length": 27.59375,
      "epoch": 1.2918586789554531,
      "grad_norm": 8.32073974609375,
      "kl": 0.5009765625,
      "learning_rate": 6.769230769230769e-07,
      "loss": 0.4081,
      "reward": 0.7160086333751678,
      "reward_std": 0.08082883059978485,
      "reward_std_all": 0.08517679199576378,
      "rewards-max/forensics_sinkhorn_reward": 0.9614133536815643,
      "rewards-max/forensics_strict_edge_reward": 0.3855053186416626,
      "rewards-max/forensics_tiou_reward": 0.9272721409797668,
      "rewards/forensics_sinkhorn_reward": 0.628368616104126,
      "rewards/forensics_strict_edge_reward": 0.04250462353229523,
      "rewards/forensics_tiou_reward": 0.43124212324619293,
      "step": 421
    },
    {
      "completion_length": 24.375,
      "epoch": 1.294930875576037,
      "grad_norm": 8.636025428771973,
      "kl": 0.3955078125,
      "learning_rate": 6.761538461538461e-07,
      "loss": 0.2846,
      "reward": 0.7031205296516418,
      "reward_std": 0.06290603429079056,
      "reward_std_all": 0.08309195935726166,
      "rewards-max/forensics_sinkhorn_reward": 0.9688480198383331,
      "rewards-max/forensics_strict_edge_reward": 0.402368500828743,
      "rewards-max/forensics_tiou_reward": 0.9479462206363678,
      "rewards/forensics_sinkhorn_reward": 0.6067444980144501,
      "rewards/forensics_strict_edge_reward": 0.060527621768414974,
      "rewards/forensics_tiou_reward": 0.4032721370458603,
      "step": 422
    },
    {
      "completion_length": 23.859375,
      "epoch": 1.2980030721966207,
      "grad_norm": 9.247530937194824,
      "kl": 0.4658203125,
      "learning_rate": 6.753846153846153e-07,
      "loss": 0.384,
      "reward": 0.7530005574226379,
      "reward_std": 0.053572794422507286,
      "reward_std_all": 0.0542199332267046,
      "rewards-max/forensics_sinkhorn_reward": 0.9781591296195984,
      "rewards-max/forensics_strict_edge_reward": 0.538444846868515,
      "rewards-max/forensics_tiou_reward": 0.9712475836277008,
      "rewards/forensics_sinkhorn_reward": 0.681351512670517,
      "rewards/forensics_strict_edge_reward": 0.04387149214744568,
      "rewards/forensics_tiou_reward": 0.47150078415870667,
      "step": 423
    },
    {
      "completion_length": 23.546875,
      "epoch": 1.3010752688172043,
      "grad_norm": 8.686728477478027,
      "kl": 0.517578125,
      "learning_rate": 6.746153846153845e-07,
      "loss": 0.0853,
      "reward": 0.7450543940067291,
      "reward_std": 0.05201881565153599,
      "reward_std_all": 0.05665905401110649,
      "rewards-max/forensics_sinkhorn_reward": 0.9801489412784576,
      "rewards-max/forensics_strict_edge_reward": 0.709127277135849,
      "rewards-max/forensics_tiou_reward": 0.97701096534729,
      "rewards/forensics_sinkhorn_reward": 0.6596719920635223,
      "rewards/forensics_strict_edge_reward": 0.057818181812763214,
      "rewards/forensics_tiou_reward": 0.49338655173778534,
      "step": 424
    },
    {
      "completion_length": 23.203125,
      "epoch": 1.304147465437788,
      "grad_norm": 9.55072021484375,
      "kl": 0.419921875,
      "learning_rate": 6.738461538461539e-07,
      "loss": 0.0195,
      "reward": 0.7677633762359619,
      "reward_std": 0.05012483801692724,
      "reward_std_all": 0.04780455119907856,
      "rewards-max/forensics_sinkhorn_reward": 0.9436380863189697,
      "rewards-max/forensics_strict_edge_reward": 0.4556138813495636,
      "rewards-max/forensics_tiou_reward": 0.8931132555007935,
      "rewards/forensics_sinkhorn_reward": 0.6842723488807678,
      "rewards/forensics_strict_edge_reward": 0.04659818671643734,
      "rewards/forensics_tiou_reward": 0.44797028601169586,
      "step": 425
    },
    {
      "completion_length": 23.34375,
      "epoch": 1.3072196620583718,
      "grad_norm": 10.897377967834473,
      "kl": 0.443359375,
      "learning_rate": 6.730769230769231e-07,
      "loss": 0.0139,
      "reward": 0.74300417304039,
      "reward_std": 0.043399980291724205,
      "reward_std_all": 0.051864391192793846,
      "rewards-max/forensics_sinkhorn_reward": 0.9665304124355316,
      "rewards-max/forensics_strict_edge_reward": 0.4257437437772751,
      "rewards-max/forensics_tiou_reward": 0.9377773702144623,
      "rewards/forensics_sinkhorn_reward": 0.6828148663043976,
      "rewards/forensics_strict_edge_reward": 0.0552561953663826,
      "rewards/forensics_tiou_reward": 0.4415595829486847,
      "step": 426
    },
    {
      "completion_length": 24.859375,
      "epoch": 1.3102918586789554,
      "grad_norm": 8.923003196716309,
      "kl": 0.412109375,
      "learning_rate": 6.723076923076922e-07,
      "loss": -0.0074,
      "reward": 0.7727555632591248,
      "reward_std": 0.022346148267388344,
      "reward_std_all": 0.030463355593383312,
      "rewards-max/forensics_sinkhorn_reward": 0.9769544899463654,
      "rewards-max/forensics_strict_edge_reward": 0.3965047597885132,
      "rewards-max/forensics_tiou_reward": 0.9612801373004913,
      "rewards/forensics_sinkhorn_reward": 0.707245945930481,
      "rewards/forensics_strict_edge_reward": 0.04790941160172224,
      "rewards/forensics_tiou_reward": 0.5231082737445831,
      "step": 427
    },
    {
      "completion_length": 23.5625,
      "epoch": 1.3133640552995391,
      "grad_norm": 9.403181076049805,
      "kl": 0.666015625,
      "learning_rate": 6.715384615384615e-07,
      "loss": 0.1397,
      "reward": 0.7409678399562836,
      "reward_std": 0.05257626622915268,
      "reward_std_all": 0.05889175459742546,
      "rewards-max/forensics_sinkhorn_reward": 0.9790477156639099,
      "rewards-max/forensics_strict_edge_reward": 0.6076286435127258,
      "rewards-max/forensics_tiou_reward": 0.9726262092590332,
      "rewards/forensics_sinkhorn_reward": 0.674843043088913,
      "rewards/forensics_strict_edge_reward": 0.05104915425181389,
      "rewards/forensics_tiou_reward": 0.4509253203868866,
      "step": 428
    },
    {
      "completion_length": 24.09375,
      "epoch": 1.316436251920123,
      "grad_norm": 8.982510566711426,
      "kl": 0.6044921875,
      "learning_rate": 6.707692307692307e-07,
      "loss": -0.0188,
      "reward": 0.7690284848213196,
      "reward_std": 0.04227147996425629,
      "reward_std_all": 0.041498471051454544,
      "rewards-max/forensics_sinkhorn_reward": 0.964955598115921,
      "rewards-max/forensics_strict_edge_reward": 0.5424859970808029,
      "rewards-max/forensics_tiou_reward": 0.9429454505443573,
      "rewards/forensics_sinkhorn_reward": 0.7039133906364441,
      "rewards/forensics_strict_edge_reward": 0.0741224680095911,
      "rewards/forensics_tiou_reward": 0.4914439916610718,
      "step": 429
    },
    {
      "completion_length": 28.125,
      "epoch": 1.3195084485407067,
      "grad_norm": 12.692078590393066,
      "kl": 0.5888671875,
      "learning_rate": 6.7e-07,
      "loss": 0.1599,
      "reward": 0.7669806182384491,
      "reward_std": 0.019411688670516014,
      "reward_std_all": 0.030826987698674202,
      "rewards-max/forensics_sinkhorn_reward": 0.9737865030765533,
      "rewards-max/forensics_strict_edge_reward": 0.255331352353096,
      "rewards-max/forensics_tiou_reward": 0.966130793094635,
      "rewards/forensics_sinkhorn_reward": 0.6603758335113525,
      "rewards/forensics_strict_edge_reward": 0.01973313931375742,
      "rewards/forensics_tiou_reward": 0.41684870421886444,
      "step": 430
    },
    {
      "completion_length": 25.375,
      "epoch": 1.3225806451612903,
      "grad_norm": 9.502702713012695,
      "kl": 0.70703125,
      "learning_rate": 6.692307692307692e-07,
      "loss": 0.2015,
      "reward": 0.7148742973804474,
      "reward_std": 0.05519145913422108,
      "reward_std_all": 0.0721953809261322,
      "rewards-max/forensics_sinkhorn_reward": 0.9735608398914337,
      "rewards-max/forensics_strict_edge_reward": 0.5333996266126633,
      "rewards-max/forensics_tiou_reward": 0.9666212201118469,
      "rewards/forensics_sinkhorn_reward": 0.610621452331543,
      "rewards/forensics_strict_edge_reward": 0.060477860271930695,
      "rewards/forensics_tiou_reward": 0.40151576697826385,
      "step": 431
    },
    {
      "completion_length": 24.515625,
      "epoch": 1.325652841781874,
      "grad_norm": 13.312777519226074,
      "kl": 0.400390625,
      "learning_rate": 6.684615384615384e-07,
      "loss": 0.1257,
      "reward": 0.7512377798557281,
      "reward_std": 0.04509218968451023,
      "reward_std_all": 0.05057078041136265,
      "rewards-max/forensics_sinkhorn_reward": 0.9677848517894745,
      "rewards-max/forensics_strict_edge_reward": 0.27642149291932583,
      "rewards-max/forensics_tiou_reward": 0.9458533227443695,
      "rewards/forensics_sinkhorn_reward": 0.6731201112270355,
      "rewards/forensics_strict_edge_reward": 0.03338289260864258,
      "rewards/forensics_tiou_reward": 0.4549817591905594,
      "step": 432
    },
    {
      "completion_length": 24.265625,
      "epoch": 1.3287250384024578,
      "grad_norm": 8.76033878326416,
      "kl": 0.4013671875,
      "learning_rate": 6.676923076923077e-07,
      "loss": 0.176,
      "reward": 0.7477844953536987,
      "reward_std": 0.03966693487018347,
      "reward_std_all": 0.0494155203923583,
      "rewards-max/forensics_sinkhorn_reward": 0.9656383991241455,
      "rewards-max/forensics_strict_edge_reward": 0.5246936976909637,
      "rewards-max/forensics_tiou_reward": 0.9350980818271637,
      "rewards/forensics_sinkhorn_reward": 0.6700291633605957,
      "rewards/forensics_strict_edge_reward": 0.06828159838914871,
      "rewards/forensics_tiou_reward": 0.5363175123929977,
      "step": 433
    },
    {
      "completion_length": 29.5,
      "epoch": 1.3317972350230414,
      "grad_norm": 7.301154136657715,
      "kl": 0.5712890625,
      "learning_rate": 6.669230769230769e-07,
      "loss": 0.316,
      "reward": 0.7151168882846832,
      "reward_std": 0.07087455317378044,
      "reward_std_all": 0.0847487561404705,
      "rewards-max/forensics_sinkhorn_reward": 0.9669630825519562,
      "rewards-max/forensics_strict_edge_reward": 0.38121218234300613,
      "rewards-max/forensics_tiou_reward": 0.9450710117816925,
      "rewards/forensics_sinkhorn_reward": 0.5815569013357162,
      "rewards/forensics_strict_edge_reward": 0.026614185189828277,
      "rewards/forensics_tiou_reward": 0.415753036737442,
      "step": 434
    },
    {
      "completion_length": 24.96875,
      "epoch": 1.3348694316436251,
      "grad_norm": 7.904009819030762,
      "kl": 0.4111328125,
      "learning_rate": 6.661538461538462e-07,
      "loss": 0.2325,
      "reward": 0.7676292657852173,
      "reward_std": 0.02589552104473114,
      "reward_std_all": 0.03240233100950718,
      "rewards-max/forensics_sinkhorn_reward": 0.9647584557533264,
      "rewards-max/forensics_strict_edge_reward": 0.39300739392638206,
      "rewards-max/forensics_tiou_reward": 0.9317120611667633,
      "rewards/forensics_sinkhorn_reward": 0.6622025072574615,
      "rewards/forensics_strict_edge_reward": 0.039588995510712266,
      "rewards/forensics_tiou_reward": 0.3939902037382126,
      "step": 435
    },
    {
      "completion_length": 24.53125,
      "epoch": 1.337941628264209,
      "grad_norm": 13.770967483520508,
      "kl": 0.8544921875,
      "learning_rate": 6.653846153846153e-07,
      "loss": -0.1442,
      "reward": 0.7621445953845978,
      "reward_std": 0.04073503613471985,
      "reward_std_all": 0.04448621720075607,
      "rewards-max/forensics_sinkhorn_reward": 0.9797956943511963,
      "rewards-max/forensics_strict_edge_reward": 0.6690258681774139,
      "rewards-max/forensics_tiou_reward": 0.9752444922924042,
      "rewards/forensics_sinkhorn_reward": 0.6948783695697784,
      "rewards/forensics_strict_edge_reward": 0.08038875460624695,
      "rewards/forensics_tiou_reward": 0.4842734783887863,
      "step": 436
    },
    {
      "completion_length": 26.578125,
      "epoch": 1.3410138248847927,
      "grad_norm": 8.202507972717285,
      "kl": 0.4169921875,
      "learning_rate": 6.646153846153845e-07,
      "loss": -0.1819,
      "reward": 0.7468417286872864,
      "reward_std": 0.04719328787177801,
      "reward_std_all": 0.056088026612997055,
      "rewards-max/forensics_sinkhorn_reward": 0.9739333093166351,
      "rewards-max/forensics_strict_edge_reward": 0.5669038891792297,
      "rewards-max/forensics_tiou_reward": 0.9631167650222778,
      "rewards/forensics_sinkhorn_reward": 0.6835623383522034,
      "rewards/forensics_strict_edge_reward": 0.07922461442649364,
      "rewards/forensics_tiou_reward": 0.5197018533945084,
      "step": 437
    },
    {
      "completion_length": 23.390625,
      "epoch": 1.3440860215053765,
      "grad_norm": 7.984729766845703,
      "kl": 0.4970703125,
      "learning_rate": 6.638461538461539e-07,
      "loss": 0.2051,
      "reward": 0.6940881609916687,
      "reward_std": 0.06313697621226311,
      "reward_std_all": 0.0857381783425808,
      "rewards-max/forensics_sinkhorn_reward": 0.9601354598999023,
      "rewards-max/forensics_strict_edge_reward": 0.08595200430136174,
      "rewards-max/forensics_tiou_reward": 0.9283446073532104,
      "rewards/forensics_sinkhorn_reward": 0.5742176473140717,
      "rewards/forensics_strict_edge_reward": 0.008811664416498388,
      "rewards/forensics_tiou_reward": 0.32438839972019196,
      "step": 438
    },
    {
      "completion_length": 23.703125,
      "epoch": 1.34715821812596,
      "grad_norm": 9.899158477783203,
      "kl": 0.494140625,
      "learning_rate": 6.630769230769231e-07,
      "loss": -0.0653,
      "reward": 0.6925685107707977,
      "reward_std": 0.058623479679226875,
      "reward_std_all": 0.0815078541636467,
      "rewards-max/forensics_sinkhorn_reward": 0.9385079145431519,
      "rewards-max/forensics_strict_edge_reward": 0.30848536640405655,
      "rewards-max/forensics_tiou_reward": 0.8713442981243134,
      "rewards/forensics_sinkhorn_reward": 0.5809817016124725,
      "rewards/forensics_strict_edge_reward": 0.025120371486991644,
      "rewards/forensics_tiou_reward": 0.34230002760887146,
      "step": 439
    },
    {
      "completion_length": 26.125,
      "epoch": 1.3502304147465438,
      "grad_norm": 8.416667938232422,
      "kl": 0.341796875,
      "learning_rate": 6.623076923076923e-07,
      "loss": 0.028,
      "reward": 0.7651340067386627,
      "reward_std": 0.028886941727250814,
      "reward_std_all": 0.035428912844508886,
      "rewards-max/forensics_sinkhorn_reward": 0.9640239477157593,
      "rewards-max/forensics_strict_edge_reward": 0.45856474339962006,
      "rewards-max/forensics_tiou_reward": 0.931614488363266,
      "rewards/forensics_sinkhorn_reward": 0.6908419132232666,
      "rewards/forensics_strict_edge_reward": 0.04572121985256672,
      "rewards/forensics_tiou_reward": 0.5106676518917084,
      "step": 440
    },
    {
      "completion_length": 26.765625,
      "epoch": 1.3533026113671274,
      "grad_norm": 19.527957916259766,
      "kl": 0.533203125,
      "learning_rate": 6.615384615384615e-07,
      "loss": 0.1632,
      "reward": 0.6623249053955078,
      "reward_std": 0.0453422125428915,
      "reward_std_all": 0.0892326571047306,
      "rewards-max/forensics_sinkhorn_reward": 0.9791128933429718,
      "rewards-max/forensics_strict_edge_reward": 0.6516547650098801,
      "rewards-max/forensics_tiou_reward": 0.9667350351810455,
      "rewards/forensics_sinkhorn_reward": 0.5403304100036621,
      "rewards/forensics_strict_edge_reward": 0.047206612303853035,
      "rewards/forensics_tiou_reward": 0.361442431807518,
      "step": 441
    },
    {
      "completion_length": 22.1875,
      "epoch": 1.3563748079877112,
      "grad_norm": 10.98839282989502,
      "kl": 0.5751953125,
      "learning_rate": 6.607692307692307e-07,
      "loss": 0.1901,
      "reward": 0.718356728553772,
      "reward_std": 0.07089937664568424,
      "reward_std_all": 0.07917642965912819,
      "rewards-max/forensics_sinkhorn_reward": 0.9685247838497162,
      "rewards-max/forensics_strict_edge_reward": 0.6170401871204376,
      "rewards-max/forensics_tiou_reward": 0.9469417035579681,
      "rewards/forensics_sinkhorn_reward": 0.6570611298084259,
      "rewards/forensics_strict_edge_reward": 0.03635330963879824,
      "rewards/forensics_tiou_reward": 0.4092804044485092,
      "step": 442
    },
    {
      "completion_length": 27.578125,
      "epoch": 1.359447004608295,
      "grad_norm": 8.545907020568848,
      "kl": 0.330078125,
      "learning_rate": 6.6e-07,
      "loss": 0.4905,
      "reward": 0.6841777563095093,
      "reward_std": 0.04569455422461033,
      "reward_std_all": 0.08222130686044693,
      "rewards-max/forensics_sinkhorn_reward": 0.9716392159461975,
      "rewards-max/forensics_strict_edge_reward": 0.561550498008728,
      "rewards-max/forensics_tiou_reward": 0.9581185281276703,
      "rewards/forensics_sinkhorn_reward": 0.5712748765945435,
      "rewards/forensics_strict_edge_reward": 0.03994634002447128,
      "rewards/forensics_tiou_reward": 0.39590854942798615,
      "step": 443
    },
    {
      "completion_length": 22.734375,
      "epoch": 1.3625192012288787,
      "grad_norm": 11.393245697021484,
      "kl": 0.3740234375,
      "learning_rate": 6.592307692307691e-07,
      "loss": 0.0582,
      "reward": 0.7567350268363953,
      "reward_std": 0.018473239615559578,
      "reward_std_all": 0.03611946664750576,
      "rewards-max/forensics_sinkhorn_reward": 0.9803791344165802,
      "rewards-max/forensics_strict_edge_reward": 0.756708562374115,
      "rewards-max/forensics_tiou_reward": 0.9795479476451874,
      "rewards/forensics_sinkhorn_reward": 0.7123481929302216,
      "rewards/forensics_strict_edge_reward": 0.07706145569682121,
      "rewards/forensics_tiou_reward": 0.5170685052871704,
      "step": 444
    },
    {
      "completion_length": 23.59375,
      "epoch": 1.3655913978494625,
      "grad_norm": 10.363692283630371,
      "kl": 0.4658203125,
      "learning_rate": 6.584615384615384e-07,
      "loss": 0.2251,
      "reward": 0.7163142263889313,
      "reward_std": 0.045508552342653275,
      "reward_std_all": 0.06808033213019371,
      "rewards-max/forensics_sinkhorn_reward": 0.985969215631485,
      "rewards-max/forensics_strict_edge_reward": 0.8351600170135498,
      "rewards-max/forensics_tiou_reward": 0.9910872876644135,
      "rewards/forensics_sinkhorn_reward": 0.6569797098636627,
      "rewards/forensics_strict_edge_reward": 0.08602601289749146,
      "rewards/forensics_tiou_reward": 0.5006237775087357,
      "step": 445
    },
    {
      "completion_length": 23.453125,
      "epoch": 1.368663594470046,
      "grad_norm": 8.391596794128418,
      "kl": 0.4140625,
      "learning_rate": 6.576923076923077e-07,
      "loss": 0.1034,
      "reward": 0.7724479734897614,
      "reward_std": 0.04668933153152466,
      "reward_std_all": 0.04330449178814888,
      "rewards-max/forensics_sinkhorn_reward": 0.9854118824005127,
      "rewards-max/forensics_strict_edge_reward": 0.7676310539245605,
      "rewards-max/forensics_tiou_reward": 0.9907310605049133,
      "rewards/forensics_sinkhorn_reward": 0.7296321392059326,
      "rewards/forensics_strict_edge_reward": 0.09500230103731155,
      "rewards/forensics_tiou_reward": 0.5985677242279053,
      "step": 446
    },
    {
      "completion_length": 25.296875,
      "epoch": 1.3717357910906298,
      "grad_norm": 8.449241638183594,
      "kl": 0.34375,
      "learning_rate": 6.569230769230769e-07,
      "loss": 0.1111,
      "reward": 0.7197552025318146,
      "reward_std": 0.0548529289662838,
      "reward_std_all": 0.0681647602468729,
      "rewards-max/forensics_sinkhorn_reward": 0.9805055856704712,
      "rewards-max/forensics_strict_edge_reward": 0.47859422117471695,
      "rewards-max/forensics_tiou_reward": 0.9755824506282806,
      "rewards/forensics_sinkhorn_reward": 0.6250716149806976,
      "rewards/forensics_strict_edge_reward": 0.05729990592226386,
      "rewards/forensics_tiou_reward": 0.4609462767839432,
      "step": 447
    },
    {
      "completion_length": 22.875,
      "epoch": 1.3748079877112136,
      "grad_norm": 10.206066131591797,
      "kl": 0.408203125,
      "learning_rate": 6.561538461538462e-07,
      "loss": 0.2382,
      "reward": 0.6956617534160614,
      "reward_std": 0.08798831515014172,
      "reward_std_all": 0.09886889159679413,
      "rewards-max/forensics_sinkhorn_reward": 0.9183154404163361,
      "rewards-max/forensics_strict_edge_reward": 0.26828642189502716,
      "rewards-max/forensics_tiou_reward": 0.8193849921226501,
      "rewards/forensics_sinkhorn_reward": 0.6057492792606354,
      "rewards/forensics_strict_edge_reward": 0.02715510129928589,
      "rewards/forensics_tiou_reward": 0.3564010262489319,
      "step": 448
    },
    {
      "completion_length": 24.03125,
      "epoch": 1.3778801843317972,
      "grad_norm": 15.484441757202148,
      "kl": 0.3642578125,
      "learning_rate": 6.553846153846153e-07,
      "loss": 0.1239,
      "reward": 0.6916158199310303,
      "reward_std": 0.06772053055465221,
      "reward_std_all": 0.08953410014510155,
      "rewards-max/forensics_sinkhorn_reward": 0.9852667450904846,
      "rewards-max/forensics_strict_edge_reward": 0.6341890394687653,
      "rewards-max/forensics_tiou_reward": 0.9902299046516418,
      "rewards/forensics_sinkhorn_reward": 0.6015522480010986,
      "rewards/forensics_strict_edge_reward": 0.043425651267170906,
      "rewards/forensics_tiou_reward": 0.4490467756986618,
      "step": 449
    },
    {
      "completion_length": 29.546875,
      "epoch": 1.380952380952381,
      "grad_norm": 11.097712516784668,
      "kl": 0.37890625,
      "learning_rate": 6.546153846153845e-07,
      "loss": 0.2247,
      "reward": 0.7036961913108826,
      "reward_std": 0.05111510120332241,
      "reward_std_all": 0.07280591130256653,
      "rewards-max/forensics_sinkhorn_reward": 0.9622532725334167,
      "rewards-max/forensics_strict_edge_reward": 0.409380704164505,
      "rewards-max/forensics_tiou_reward": 0.9358187317848206,
      "rewards/forensics_sinkhorn_reward": 0.5378946661949158,
      "rewards/forensics_strict_edge_reward": 0.02480613999068737,
      "rewards/forensics_tiou_reward": 0.3446474075317383,
      "step": 450
    },
    {
      "completion_length": 24.65625,
      "epoch": 1.3840245775729647,
      "grad_norm": 9.933623313903809,
      "kl": 0.3876953125,
      "learning_rate": 6.538461538461538e-07,
      "loss": 0.2511,
      "reward": 0.7196792662143707,
      "reward_std": 0.10814552381634712,
      "reward_std_all": 0.10227175056934357,
      "rewards-max/forensics_sinkhorn_reward": 0.9703494608402252,
      "rewards-max/forensics_strict_edge_reward": 0.28488798439502716,
      "rewards-max/forensics_tiou_reward": 0.9416711032390594,
      "rewards/forensics_sinkhorn_reward": 0.624748021364212,
      "rewards/forensics_strict_edge_reward": 0.023530295118689537,
      "rewards/forensics_tiou_reward": 0.441670298576355,
      "step": 451
    },
    {
      "completion_length": 23.515625,
      "epoch": 1.3870967741935485,
      "grad_norm": 7.7684102058410645,
      "kl": 0.474609375,
      "learning_rate": 6.530769230769231e-07,
      "loss": 0.4353,
      "reward": 0.6751275658607483,
      "reward_std": 0.09016253799200058,
      "reward_std_all": 0.10932360216975212,
      "rewards-max/forensics_sinkhorn_reward": 0.9597819447517395,
      "rewards-max/forensics_strict_edge_reward": 0.34215399622917175,
      "rewards-max/forensics_tiou_reward": 0.9241982102394104,
      "rewards/forensics_sinkhorn_reward": 0.5937884747982025,
      "rewards/forensics_strict_edge_reward": 0.03826399799436331,
      "rewards/forensics_tiou_reward": 0.41479212045669556,
      "step": 452
    },
    {
      "completion_length": 22.078125,
      "epoch": 1.390168970814132,
      "grad_norm": 16.507097244262695,
      "kl": 1.3310546875,
      "learning_rate": 6.523076923076923e-07,
      "loss": 0.0191,
      "reward": 0.7780415117740631,
      "reward_std": 0.020564310252666473,
      "reward_std_all": 0.026674830354750156,
      "rewards-max/forensics_sinkhorn_reward": 0.976600855588913,
      "rewards-max/forensics_strict_edge_reward": 0.4712325632572174,
      "rewards-max/forensics_tiou_reward": 0.9675925374031067,
      "rewards/forensics_sinkhorn_reward": 0.6993202567100525,
      "rewards/forensics_strict_edge_reward": 0.0592840239405632,
      "rewards/forensics_tiou_reward": 0.47635504603385925,
      "step": 453
    },
    {
      "completion_length": 23.3125,
      "epoch": 1.3932411674347158,
      "grad_norm": 9.769418716430664,
      "kl": 0.43359375,
      "learning_rate": 6.515384615384615e-07,
      "loss": 0.2137,
      "reward": 0.7706507444381714,
      "reward_std": 0.03406541422009468,
      "reward_std_all": 0.03626155108213425,
      "rewards-max/forensics_sinkhorn_reward": 0.9662946462631226,
      "rewards-max/forensics_strict_edge_reward": 0.4633841961622238,
      "rewards-max/forensics_tiou_reward": 0.9465183913707733,
      "rewards/forensics_sinkhorn_reward": 0.7095046639442444,
      "rewards/forensics_strict_edge_reward": 0.06560653075575829,
      "rewards/forensics_tiou_reward": 0.5217322111129761,
      "step": 454
    },
    {
      "completion_length": 21.78125,
      "epoch": 1.3963133640552996,
      "grad_norm": 10.358469009399414,
      "kl": 0.4853515625,
      "learning_rate": 6.507692307692307e-07,
      "loss": 0.1197,
      "reward": 0.7095420658588409,
      "reward_std": 0.03371686954051256,
      "reward_std_all": 0.06201455742120743,
      "rewards-max/forensics_sinkhorn_reward": 0.9779273569583893,
      "rewards-max/forensics_strict_edge_reward": 0.5896764099597931,
      "rewards-max/forensics_tiou_reward": 0.9701747000217438,
      "rewards/forensics_sinkhorn_reward": 0.6305658221244812,
      "rewards/forensics_strict_edge_reward": 0.07268255855888128,
      "rewards/forensics_tiou_reward": 0.41591405868530273,
      "step": 455
    },
    {
      "completion_length": 22.8125,
      "epoch": 1.3993855606758832,
      "grad_norm": 9.169208526611328,
      "kl": 0.453125,
      "learning_rate": 6.5e-07,
      "loss": 0.1458,
      "reward": 0.7212508916854858,
      "reward_std": 0.061902567744255066,
      "reward_std_all": 0.06999385263770819,
      "rewards-max/forensics_sinkhorn_reward": 0.9780868589878082,
      "rewards-max/forensics_strict_edge_reward": 0.42236119508743286,
      "rewards-max/forensics_tiou_reward": 0.9723710417747498,
      "rewards/forensics_sinkhorn_reward": 0.6657077670097351,
      "rewards/forensics_strict_edge_reward": 0.0497548459097743,
      "rewards/forensics_tiou_reward": 0.4726417660713196,
      "step": 456
    },
    {
      "completion_length": 21.875,
      "epoch": 1.402457757296467,
      "grad_norm": 9.697223663330078,
      "kl": 0.4921875,
      "learning_rate": 6.492307692307692e-07,
      "loss": 0.3858,
      "reward": 0.6812630593776703,
      "reward_std": 0.07608775421977043,
      "reward_std_all": 0.09948383271694183,
      "rewards-max/forensics_sinkhorn_reward": 0.9751814305782318,
      "rewards-max/forensics_strict_edge_reward": 0.6274375021457672,
      "rewards-max/forensics_tiou_reward": 0.9652600586414337,
      "rewards/forensics_sinkhorn_reward": 0.5749825239181519,
      "rewards/forensics_strict_edge_reward": 0.05612847954034805,
      "rewards/forensics_tiou_reward": 0.3660246282815933,
      "step": 457
    },
    {
      "completion_length": 23.796875,
      "epoch": 1.4055299539170507,
      "grad_norm": 8.581483840942383,
      "kl": 0.73828125,
      "learning_rate": 6.484615384615383e-07,
      "loss": 0.1389,
      "reward": 0.7405033111572266,
      "reward_std": 0.044291749596595764,
      "reward_std_all": 0.055235665291547775,
      "rewards-max/forensics_sinkhorn_reward": 0.9555651545524597,
      "rewards-max/forensics_strict_edge_reward": 0.3830736055970192,
      "rewards-max/forensics_tiou_reward": 0.9185663759708405,
      "rewards/forensics_sinkhorn_reward": 0.6348216831684113,
      "rewards/forensics_strict_edge_reward": 0.04356513498350978,
      "rewards/forensics_tiou_reward": 0.3945069909095764,
      "step": 458
    },
    {
      "completion_length": 22.140625,
      "epoch": 1.4086021505376345,
      "grad_norm": 10.958930969238281,
      "kl": 0.3759765625,
      "learning_rate": 6.476923076923077e-07,
      "loss": -0.0364,
      "reward": 0.7666519284248352,
      "reward_std": 0.031465657986700535,
      "reward_std_all": 0.03540065046399832,
      "rewards-max/forensics_sinkhorn_reward": 0.9847302436828613,
      "rewards-max/forensics_strict_edge_reward": 0.6168450862169266,
      "rewards-max/forensics_tiou_reward": 0.987016350030899,
      "rewards/forensics_sinkhorn_reward": 0.7409057021141052,
      "rewards/forensics_strict_edge_reward": 0.07395638898015022,
      "rewards/forensics_tiou_reward": 0.5355564653873444,
      "step": 459
    },
    {
      "completion_length": 21.6875,
      "epoch": 1.411674347158218,
      "grad_norm": 9.118985176086426,
      "kl": 0.36328125,
      "learning_rate": 6.469230769230769e-07,
      "loss": 0.0515,
      "reward": 0.7122199833393097,
      "reward_std": 0.05498011037707329,
      "reward_std_all": 0.07665249332785606,
      "rewards-max/forensics_sinkhorn_reward": 0.932732880115509,
      "rewards-max/forensics_strict_edge_reward": 0.24438381753861904,
      "rewards-max/forensics_tiou_reward": 0.8683727085590363,
      "rewards/forensics_sinkhorn_reward": 0.6312851011753082,
      "rewards/forensics_strict_edge_reward": 0.014296527428086847,
      "rewards/forensics_tiou_reward": 0.3681087791919708,
      "step": 460
    },
    {
      "completion_length": 21.328125,
      "epoch": 1.4147465437788018,
      "grad_norm": 7.846642971038818,
      "kl": 0.5078125,
      "learning_rate": 6.461538461538462e-07,
      "loss": 0.4452,
      "reward": 0.7241347134113312,
      "reward_std": 0.10144324414432049,
      "reward_std_all": 0.09574658796191216,
      "rewards-max/forensics_sinkhorn_reward": 0.9800398349761963,
      "rewards-max/forensics_strict_edge_reward": 0.4361082762479782,
      "rewards-max/forensics_tiou_reward": 0.9750886857509613,
      "rewards/forensics_sinkhorn_reward": 0.690365195274353,
      "rewards/forensics_strict_edge_reward": 0.047658505849540234,
      "rewards/forensics_tiou_reward": 0.4460630714893341,
      "step": 461
    },
    {
      "completion_length": 23.1875,
      "epoch": 1.4178187403993856,
      "grad_norm": 7.165112495422363,
      "kl": 0.4208984375,
      "learning_rate": 6.453846153846153e-07,
      "loss": 0.1587,
      "reward": 0.7454788088798523,
      "reward_std": 0.06200086884200573,
      "reward_std_all": 0.06511920318007469,
      "rewards-max/forensics_sinkhorn_reward": 0.9797086417675018,
      "rewards-max/forensics_strict_edge_reward": 0.638425350189209,
      "rewards-max/forensics_tiou_reward": 0.973941296339035,
      "rewards/forensics_sinkhorn_reward": 0.6932376027107239,
      "rewards/forensics_strict_edge_reward": 0.055472357198596,
      "rewards/forensics_tiou_reward": 0.49239102005958557,
      "step": 462
    },
    {
      "completion_length": 24.09375,
      "epoch": 1.4208909370199692,
      "grad_norm": 6.494134902954102,
      "kl": 0.42578125,
      "learning_rate": 6.446153846153845e-07,
      "loss": -0.0051,
      "reward": 0.7432644963264465,
      "reward_std": 0.04268603213131428,
      "reward_std_all": 0.052913935855031013,
      "rewards-max/forensics_sinkhorn_reward": 0.9417691826820374,
      "rewards-max/forensics_strict_edge_reward": 0.3609570264816284,
      "rewards-max/forensics_tiou_reward": 0.8927499353885651,
      "rewards/forensics_sinkhorn_reward": 0.6849569380283356,
      "rewards/forensics_strict_edge_reward": 0.03134461119771004,
      "rewards/forensics_tiou_reward": 0.4224333018064499,
      "step": 463
    },
    {
      "completion_length": 23.921875,
      "epoch": 1.423963133640553,
      "grad_norm": 10.49038028717041,
      "kl": 0.400390625,
      "learning_rate": 6.438461538461538e-07,
      "loss": 0.1961,
      "reward": 0.6924852728843689,
      "reward_std": 0.07670367509126663,
      "reward_std_all": 0.09617716819047928,
      "rewards-max/forensics_sinkhorn_reward": 0.9614168107509613,
      "rewards-max/forensics_strict_edge_reward": 0.43609243631362915,
      "rewards-max/forensics_tiou_reward": 0.929293304681778,
      "rewards/forensics_sinkhorn_reward": 0.6198104619979858,
      "rewards/forensics_strict_edge_reward": 0.031787317246198654,
      "rewards/forensics_tiou_reward": 0.3781522661447525,
      "step": 464
    },
    {
      "completion_length": 23.1875,
      "epoch": 1.4270353302611367,
      "grad_norm": 6.184193134307861,
      "kl": 0.62890625,
      "learning_rate": 6.43076923076923e-07,
      "loss": 0.0476,
      "reward": 0.7535503804683685,
      "reward_std": 0.024445160757750273,
      "reward_std_all": 0.03928213380277157,
      "rewards-max/forensics_sinkhorn_reward": 0.9528708755970001,
      "rewards-max/forensics_strict_edge_reward": 0.1327925645673531,
      "rewards-max/forensics_tiou_reward": 0.9068792760372162,
      "rewards/forensics_sinkhorn_reward": 0.689842015504837,
      "rewards/forensics_strict_edge_reward": 0.018337037047785998,
      "rewards/forensics_tiou_reward": 0.47149212658405304,
      "step": 465
    },
    {
      "completion_length": 25.1875,
      "epoch": 1.4301075268817205,
      "grad_norm": 6.557056427001953,
      "kl": 0.4833984375,
      "learning_rate": 6.423076923076924e-07,
      "loss": 0.4003,
      "reward": 0.7098952233791351,
      "reward_std": 0.05389171838760376,
      "reward_std_all": 0.07460621371865273,
      "rewards-max/forensics_sinkhorn_reward": 0.9477372467517853,
      "rewards-max/forensics_strict_edge_reward": 0.2766973301768303,
      "rewards-max/forensics_tiou_reward": 0.8948107063770294,
      "rewards/forensics_sinkhorn_reward": 0.5684399902820587,
      "rewards/forensics_strict_edge_reward": 0.0406628642231226,
      "rewards/forensics_tiou_reward": 0.38783058524131775,
      "step": 466
    },
    {
      "completion_length": 23.484375,
      "epoch": 1.4331797235023043,
      "grad_norm": 11.726506233215332,
      "kl": 0.4365234375,
      "learning_rate": 6.415384615384615e-07,
      "loss": 0.1325,
      "reward": 0.7573072016239166,
      "reward_std": 0.04291601572185755,
      "reward_std_all": 0.04576718062162399,
      "rewards-max/forensics_sinkhorn_reward": 0.976455569267273,
      "rewards-max/forensics_strict_edge_reward": 0.42036955058574677,
      "rewards-max/forensics_tiou_reward": 0.9624517858028412,
      "rewards/forensics_sinkhorn_reward": 0.681984931230545,
      "rewards/forensics_strict_edge_reward": 0.03703606501221657,
      "rewards/forensics_tiou_reward": 0.38792990148067474,
      "step": 467
    },
    {
      "completion_length": 22.953125,
      "epoch": 1.4362519201228878,
      "grad_norm": 8.848037719726562,
      "kl": 0.376953125,
      "learning_rate": 6.407692307692307e-07,
      "loss": -0.1812,
      "reward": 0.6912875175476074,
      "reward_std": 0.05209015868604183,
      "reward_std_all": 0.07887226901948452,
      "rewards-max/forensics_sinkhorn_reward": 0.9560765624046326,
      "rewards-max/forensics_strict_edge_reward": 0.3375745862722397,
      "rewards-max/forensics_tiou_reward": 0.8933897912502289,
      "rewards/forensics_sinkhorn_reward": 0.5891177654266357,
      "rewards/forensics_strict_edge_reward": 0.02175889629870653,
      "rewards/forensics_tiou_reward": 0.3345988988876343,
      "step": 468
    },
    {
      "completion_length": 24.640625,
      "epoch": 1.4393241167434716,
      "grad_norm": 9.081385612487793,
      "kl": 0.4794921875,
      "learning_rate": 6.4e-07,
      "loss": 0.2346,
      "reward": 0.7140857577323914,
      "reward_std": 0.09385916590690613,
      "reward_std_all": 0.09407950192689896,
      "rewards-max/forensics_sinkhorn_reward": 0.9630861580371857,
      "rewards-max/forensics_strict_edge_reward": 0.8333333432674408,
      "rewards-max/forensics_tiou_reward": 0.9379942417144775,
      "rewards/forensics_sinkhorn_reward": 0.642863541841507,
      "rewards/forensics_strict_edge_reward": 0.06110377609729767,
      "rewards/forensics_tiou_reward": 0.41732098162174225,
      "step": 469
    },
    {
      "completion_length": 22.359375,
      "epoch": 1.4423963133640554,
      "grad_norm": 6.585651397705078,
      "kl": 0.556640625,
      "learning_rate": 6.392307692307692e-07,
      "loss": -0.0432,
      "reward": 0.7705866694450378,
      "reward_std": 0.05170060694217682,
      "reward_std_all": 0.04750663600862026,
      "rewards-max/forensics_sinkhorn_reward": 0.9740768373012543,
      "rewards-max/forensics_strict_edge_reward": 0.6876014769077301,
      "rewards-max/forensics_tiou_reward": 0.9638490974903107,
      "rewards/forensics_sinkhorn_reward": 0.7308458685874939,
      "rewards/forensics_strict_edge_reward": 0.10936786606907845,
      "rewards/forensics_tiou_reward": 0.5434620678424835,
      "step": 470
    },
    {
      "completion_length": 24.609375,
      "epoch": 1.445468509984639,
      "grad_norm": 10.956815719604492,
      "kl": 0.4384765625,
      "learning_rate": 6.384615384615383e-07,
      "loss": 0.1089,
      "reward": 0.7431385219097137,
      "reward_std": 0.025728169828653336,
      "reward_std_all": 0.04707011952996254,
      "rewards-max/forensics_sinkhorn_reward": 0.9713191390037537,
      "rewards-max/forensics_strict_edge_reward": 0.6286808848381042,
      "rewards-max/forensics_tiou_reward": 0.954111784696579,
      "rewards/forensics_sinkhorn_reward": 0.6518160700798035,
      "rewards/forensics_strict_edge_reward": 0.08031724579632282,
      "rewards/forensics_tiou_reward": 0.4857761263847351,
      "step": 471
    },
    {
      "completion_length": 25.078125,
      "epoch": 1.4485407066052227,
      "grad_norm": 10.743945121765137,
      "kl": 0.4833984375,
      "learning_rate": 6.376923076923077e-07,
      "loss": 0.2701,
      "reward": 0.7674471735954285,
      "reward_std": 0.04998531565070152,
      "reward_std_all": 0.04738743230700493,
      "rewards-max/forensics_sinkhorn_reward": 0.9609414637088776,
      "rewards-max/forensics_strict_edge_reward": 0.3893286883831024,
      "rewards-max/forensics_tiou_reward": 0.9289872944355011,
      "rewards/forensics_sinkhorn_reward": 0.6783013343811035,
      "rewards/forensics_strict_edge_reward": 0.025947081856429577,
      "rewards/forensics_tiou_reward": 0.46227771043777466,
      "step": 472
    },
    {
      "completion_length": 25.0625,
      "epoch": 1.4516129032258065,
      "grad_norm": 8.405562400817871,
      "kl": 0.3447265625,
      "learning_rate": 6.369230769230769e-07,
      "loss": -0.0196,
      "reward": 0.7542021572589874,
      "reward_std": 0.038593060337007046,
      "reward_std_all": 0.04345940425992012,
      "rewards-max/forensics_sinkhorn_reward": 0.9805354177951813,
      "rewards-max/forensics_strict_edge_reward": 0.5321442186832428,
      "rewards-max/forensics_tiou_reward": 0.9777294099330902,
      "rewards/forensics_sinkhorn_reward": 0.6790537238121033,
      "rewards/forensics_strict_edge_reward": 0.050396887585520744,
      "rewards/forensics_tiou_reward": 0.45508721470832825,
      "step": 473
    },
    {
      "completion_length": 24.234375,
      "epoch": 1.4546850998463903,
      "grad_norm": 10.027205467224121,
      "kl": 0.3212890625,
      "learning_rate": 6.361538461538462e-07,
      "loss": -0.0573,
      "reward": 0.7298604846000671,
      "reward_std": 0.04384115431457758,
      "reward_std_all": 0.06009341776371002,
      "rewards-max/forensics_sinkhorn_reward": 0.9772579669952393,
      "rewards-max/forensics_strict_edge_reward": 0.5753126740455627,
      "rewards-max/forensics_tiou_reward": 0.9662855267524719,
      "rewards/forensics_sinkhorn_reward": 0.6663847267627716,
      "rewards/forensics_strict_edge_reward": 0.061320871114730835,
      "rewards/forensics_tiou_reward": 0.4881960451602936,
      "step": 474
    },
    {
      "completion_length": 25.390625,
      "epoch": 1.4577572964669738,
      "grad_norm": 8.305709838867188,
      "kl": 0.4755859375,
      "learning_rate": 6.353846153846154e-07,
      "loss": 0.0643,
      "reward": 0.7501600384712219,
      "reward_std": 0.04981953650712967,
      "reward_std_all": 0.052983684465289116,
      "rewards-max/forensics_sinkhorn_reward": 0.9849956035614014,
      "rewards-max/forensics_strict_edge_reward": 0.5919729918241501,
      "rewards-max/forensics_tiou_reward": 0.9891049265861511,
      "rewards/forensics_sinkhorn_reward": 0.6415739357471466,
      "rewards/forensics_strict_edge_reward": 0.06106788106262684,
      "rewards/forensics_tiou_reward": 0.4410376697778702,
      "step": 475
    },
    {
      "completion_length": 23.09375,
      "epoch": 1.4608294930875576,
      "grad_norm": 7.614587306976318,
      "kl": 0.505859375,
      "learning_rate": 6.346153846153845e-07,
      "loss": -0.0388,
      "reward": 0.7536744177341461,
      "reward_std": 0.06734922807663679,
      "reward_std_all": 0.06307394988834858,
      "rewards-max/forensics_sinkhorn_reward": 0.9881173968315125,
      "rewards-max/forensics_strict_edge_reward": 0.8079957664012909,
      "rewards-max/forensics_tiou_reward": 0.9967509806156158,
      "rewards/forensics_sinkhorn_reward": 0.6743534505367279,
      "rewards/forensics_strict_edge_reward": 0.12222453951835632,
      "rewards/forensics_tiou_reward": 0.49165409803390503,
      "step": 476
    },
    {
      "completion_length": 24.203125,
      "epoch": 1.4639016897081414,
      "grad_norm": 7.6903557777404785,
      "kl": 0.4443359375,
      "learning_rate": 6.338461538461538e-07,
      "loss": 0.0891,
      "reward": 0.7555505335330963,
      "reward_std": 0.052161650732159615,
      "reward_std_all": 0.052376653999090195,
      "rewards-max/forensics_sinkhorn_reward": 0.9852459728717804,
      "rewards-max/forensics_strict_edge_reward": 0.6822756230831146,
      "rewards-max/forensics_tiou_reward": 0.9894453585147858,
      "rewards/forensics_sinkhorn_reward": 0.6948771178722382,
      "rewards/forensics_strict_edge_reward": 0.05937955342233181,
      "rewards/forensics_tiou_reward": 0.5062588304281235,
      "step": 477
    },
    {
      "completion_length": 25.0,
      "epoch": 1.466973886328725,
      "grad_norm": 9.16733455657959,
      "kl": 0.44140625,
      "learning_rate": 6.33076923076923e-07,
      "loss": 0.1452,
      "reward": 0.7539741396903992,
      "reward_std": 0.03978918446227908,
      "reward_std_all": 0.046482279896736145,
      "rewards-max/forensics_sinkhorn_reward": 0.9235842227935791,
      "rewards-max/forensics_strict_edge_reward": 0.42826876416802406,
      "rewards-max/forensics_tiou_reward": 0.8538927733898163,
      "rewards/forensics_sinkhorn_reward": 0.635143518447876,
      "rewards/forensics_strict_edge_reward": 0.04157118918374181,
      "rewards/forensics_tiou_reward": 0.43918725848197937,
      "step": 478
    },
    {
      "completion_length": 23.03125,
      "epoch": 1.4700460829493087,
      "grad_norm": 8.722310066223145,
      "kl": 0.4931640625,
      "learning_rate": 6.323076923076924e-07,
      "loss": 0.2389,
      "reward": 0.7667661011219025,
      "reward_std": 0.037194281816482544,
      "reward_std_all": 0.03985998407006264,
      "rewards-max/forensics_sinkhorn_reward": 0.9674946367740631,
      "rewards-max/forensics_strict_edge_reward": 0.48620253801345825,
      "rewards-max/forensics_tiou_reward": 0.9417827725410461,
      "rewards/forensics_sinkhorn_reward": 0.6971712112426758,
      "rewards/forensics_strict_edge_reward": 0.060680849477648735,
      "rewards/forensics_tiou_reward": 0.41873860359191895,
      "step": 479
    },
    {
      "completion_length": 21.90625,
      "epoch": 1.4731182795698925,
      "grad_norm": 12.088628768920898,
      "kl": 0.4208984375,
      "learning_rate": 6.315384615384615e-07,
      "loss": 0.2787,
      "reward": 0.725390613079071,
      "reward_std": 0.07724517397582531,
      "reward_std_all": 0.07902391999959946,
      "rewards-max/forensics_sinkhorn_reward": 0.9757013618946075,
      "rewards-max/forensics_strict_edge_reward": 0.5205644369125366,
      "rewards-max/forensics_tiou_reward": 0.9643739461898804,
      "rewards/forensics_sinkhorn_reward": 0.6234913468360901,
      "rewards/forensics_strict_edge_reward": 0.06876332312822342,
      "rewards/forensics_tiou_reward": 0.38418300449848175,
      "step": 480
    },
    {
      "completion_length": 26.859375,
      "epoch": 1.4761904761904763,
      "grad_norm": 7.053993225097656,
      "kl": 0.3515625,
      "learning_rate": 6.307692307692307e-07,
      "loss": -0.043,
      "reward": 0.7587490975856781,
      "reward_std": 0.038096860982477665,
      "reward_std_all": 0.043162195943295956,
      "rewards-max/forensics_sinkhorn_reward": 0.9730388522148132,
      "rewards-max/forensics_strict_edge_reward": 0.6082656681537628,
      "rewards-max/forensics_tiou_reward": 0.9575416743755341,
      "rewards/forensics_sinkhorn_reward": 0.6693129539489746,
      "rewards/forensics_strict_edge_reward": 0.06669963896274567,
      "rewards/forensics_tiou_reward": 0.5022462606430054,
      "step": 481
    },
    {
      "completion_length": 23.1875,
      "epoch": 1.4792626728110598,
      "grad_norm": 8.122594833374023,
      "kl": 0.34423828125,
      "learning_rate": 6.3e-07,
      "loss": 0.1452,
      "reward": 0.7103897929191589,
      "reward_std": 0.0491778589785099,
      "reward_std_all": 0.06860064901411533,
      "rewards-max/forensics_sinkhorn_reward": 0.9697766602039337,
      "rewards-max/forensics_strict_edge_reward": 0.4623936712741852,
      "rewards-max/forensics_tiou_reward": 0.9541921317577362,
      "rewards/forensics_sinkhorn_reward": 0.6001084744930267,
      "rewards/forensics_strict_edge_reward": 0.024495521560311317,
      "rewards/forensics_tiou_reward": 0.3532903790473938,
      "step": 482
    },
    {
      "completion_length": 25.0625,
      "epoch": 1.4823348694316436,
      "grad_norm": 7.585169315338135,
      "kl": 0.494140625,
      "learning_rate": 6.292307692307692e-07,
      "loss": 0.2798,
      "reward": 0.6988574862480164,
      "reward_std": 0.06384317390620708,
      "reward_std_all": 0.08273591846227646,
      "rewards-max/forensics_sinkhorn_reward": 0.9778887629508972,
      "rewards-max/forensics_strict_edge_reward": 0.48720505833625793,
      "rewards-max/forensics_tiou_reward": 0.9687720537185669,
      "rewards/forensics_sinkhorn_reward": 0.6095490157604218,
      "rewards/forensics_strict_edge_reward": 0.048793964087963104,
      "rewards/forensics_tiou_reward": 0.40271949768066406,
      "step": 483
    },
    {
      "completion_length": 23.59375,
      "epoch": 1.4854070660522274,
      "grad_norm": 8.957603454589844,
      "kl": 0.4228515625,
      "learning_rate": 6.284615384615385e-07,
      "loss": 0.2422,
      "reward": 0.7547312080860138,
      "reward_std": 0.03561784140765667,
      "reward_std_all": 0.04541466571390629,
      "rewards-max/forensics_sinkhorn_reward": 0.9826482534408569,
      "rewards-max/forensics_strict_edge_reward": 0.7958523333072662,
      "rewards-max/forensics_tiou_reward": 0.9832581877708435,
      "rewards/forensics_sinkhorn_reward": 0.659687876701355,
      "rewards/forensics_strict_edge_reward": 0.06272128596901894,
      "rewards/forensics_tiou_reward": 0.4156845211982727,
      "step": 484
    },
    {
      "completion_length": 27.0625,
      "epoch": 1.488479262672811,
      "grad_norm": 7.086518287658691,
      "kl": 0.4462890625,
      "learning_rate": 6.276923076923076e-07,
      "loss": 0.1901,
      "reward": 0.7490722835063934,
      "reward_std": 0.05459355376660824,
      "reward_std_all": 0.055986298248171806,
      "rewards-max/forensics_sinkhorn_reward": 0.9267778098583221,
      "rewards-max/forensics_strict_edge_reward": 0.09420423395931721,
      "rewards-max/forensics_tiou_reward": 0.84548020362854,
      "rewards/forensics_sinkhorn_reward": 0.6264604330062866,
      "rewards/forensics_strict_edge_reward": 0.009183074813336134,
      "rewards/forensics_tiou_reward": 0.43022583425045013,
      "step": 485
    },
    {
      "completion_length": 27.5625,
      "epoch": 1.4915514592933947,
      "grad_norm": 10.255091667175293,
      "kl": 0.412109375,
      "learning_rate": 6.269230769230769e-07,
      "loss": -0.0349,
      "reward": 0.6973784863948822,
      "reward_std": 0.05275457538664341,
      "reward_std_all": 0.07727847620844841,
      "rewards-max/forensics_sinkhorn_reward": 0.9627488553524017,
      "rewards-max/forensics_strict_edge_reward": 0.3853638991713524,
      "rewards-max/forensics_tiou_reward": 0.942511796951294,
      "rewards/forensics_sinkhorn_reward": 0.5753296613693237,
      "rewards/forensics_strict_edge_reward": 0.03913795854896307,
      "rewards/forensics_tiou_reward": 0.3840419799089432,
      "step": 486
    },
    {
      "completion_length": 25.890625,
      "epoch": 1.4946236559139785,
      "grad_norm": 10.506112098693848,
      "kl": 0.544921875,
      "learning_rate": 6.261538461538462e-07,
      "loss": 0.1609,
      "reward": 0.7188074588775635,
      "reward_std": 0.034794194623827934,
      "reward_std_all": 0.059629637748003006,
      "rewards-max/forensics_sinkhorn_reward": 0.9754881262779236,
      "rewards-max/forensics_strict_edge_reward": 0.3726078122854233,
      "rewards-max/forensics_tiou_reward": 0.9655772745609283,
      "rewards/forensics_sinkhorn_reward": 0.6748263835906982,
      "rewards/forensics_strict_edge_reward": 0.050506966188549995,
      "rewards/forensics_tiou_reward": 0.4538416862487793,
      "step": 487
    },
    {
      "completion_length": 24.53125,
      "epoch": 1.4976958525345623,
      "grad_norm": 8.445721626281738,
      "kl": 0.306640625,
      "learning_rate": 6.253846153846154e-07,
      "loss": 0.0772,
      "reward": 0.7624388337135315,
      "reward_std": 0.047959149815142155,
      "reward_std_all": 0.0494390744715929,
      "rewards-max/forensics_sinkhorn_reward": 0.9569584429264069,
      "rewards-max/forensics_strict_edge_reward": 0.3816106468439102,
      "rewards-max/forensics_tiou_reward": 0.9137140214443207,
      "rewards/forensics_sinkhorn_reward": 0.6998084783554077,
      "rewards/forensics_strict_edge_reward": 0.03849701210856438,
      "rewards/forensics_tiou_reward": 0.4924115389585495,
      "step": 488
    },
    {
      "completion_length": 29.3125,
      "epoch": 1.500768049155146,
      "grad_norm": 6.7419657707214355,
      "kl": 0.38671875,
      "learning_rate": 6.246153846153845e-07,
      "loss": 0.5078,
      "reward": 0.7082773745059967,
      "reward_std": 0.06685769557952881,
      "reward_std_all": 0.08188755437731743,
      "rewards-max/forensics_sinkhorn_reward": 0.9801830649375916,
      "rewards-max/forensics_strict_edge_reward": 0.46825139969587326,
      "rewards-max/forensics_tiou_reward": 0.9771648943424225,
      "rewards/forensics_sinkhorn_reward": 0.5754684209823608,
      "rewards/forensics_strict_edge_reward": 0.03685175767168403,
      "rewards/forensics_tiou_reward": 0.4067540615797043,
      "step": 489
    },
    {
      "completion_length": 25.453125,
      "epoch": 1.5038402457757296,
      "grad_norm": 8.039287567138672,
      "kl": 0.400390625,
      "learning_rate": 6.238461538461538e-07,
      "loss": 0.1341,
      "reward": 0.739010751247406,
      "reward_std": 0.01938769780099392,
      "reward_std_all": 0.04134721867740154,
      "rewards-max/forensics_sinkhorn_reward": 0.9847636520862579,
      "rewards-max/forensics_strict_edge_reward": 0.569886103272438,
      "rewards-max/forensics_tiou_reward": 0.9890545308589935,
      "rewards/forensics_sinkhorn_reward": 0.6529848277568817,
      "rewards/forensics_strict_edge_reward": 0.05353086069226265,
      "rewards/forensics_tiou_reward": 0.494667649269104,
      "step": 490
    },
    {
      "completion_length": 22.640625,
      "epoch": 1.5069124423963134,
      "grad_norm": 8.735465049743652,
      "kl": 0.48828125,
      "learning_rate": 6.23076923076923e-07,
      "loss": -0.119,
      "reward": 0.7327516973018646,
      "reward_std": 0.048233821988105774,
      "reward_std_all": 0.05683094263076782,
      "rewards-max/forensics_sinkhorn_reward": 0.9825131297111511,
      "rewards-max/forensics_strict_edge_reward": 0.7165313363075256,
      "rewards-max/forensics_tiou_reward": 0.9841318130493164,
      "rewards/forensics_sinkhorn_reward": 0.642939567565918,
      "rewards/forensics_strict_edge_reward": 0.09058857336640358,
      "rewards/forensics_tiou_reward": 0.48621784150600433,
      "step": 491
    },
    {
      "completion_length": 23.890625,
      "epoch": 1.509984639016897,
      "grad_norm": 8.625006675720215,
      "kl": 0.5703125,
      "learning_rate": 6.223076923076924e-07,
      "loss": -0.0252,
      "reward": 0.7635016143321991,
      "reward_std": 0.009558877209201455,
      "reward_std_all": 0.027932350523769855,
      "rewards-max/forensics_sinkhorn_reward": 0.9811680018901825,
      "rewards-max/forensics_strict_edge_reward": 0.6556849330663681,
      "rewards-max/forensics_tiou_reward": 0.9767940044403076,
      "rewards/forensics_sinkhorn_reward": 0.6951484084129333,
      "rewards/forensics_strict_edge_reward": 0.10301159881055355,
      "rewards/forensics_tiou_reward": 0.5032356679439545,
      "step": 492
    },
    {
      "completion_length": 26.6875,
      "epoch": 1.5130568356374807,
      "grad_norm": 9.172259330749512,
      "kl": 0.505859375,
      "learning_rate": 6.215384615384615e-07,
      "loss": 0.3537,
      "reward": 0.7259408533573151,
      "reward_std": 0.059664286673069,
      "reward_std_all": 0.07164985314011574,
      "rewards-max/forensics_sinkhorn_reward": 0.9788969457149506,
      "rewards-max/forensics_strict_edge_reward": 0.47703270614147186,
      "rewards-max/forensics_tiou_reward": 0.9685169458389282,
      "rewards/forensics_sinkhorn_reward": 0.6688809096813202,
      "rewards/forensics_strict_edge_reward": 0.05437064729630947,
      "rewards/forensics_tiou_reward": 0.4960724711418152,
      "step": 493
    },
    {
      "completion_length": 26.828125,
      "epoch": 1.5161290322580645,
      "grad_norm": 8.172687530517578,
      "kl": 0.4189453125,
      "learning_rate": 6.207692307692307e-07,
      "loss": 0.3143,
      "reward": 0.7416037023067474,
      "reward_std": 0.035865627229213715,
      "reward_std_all": 0.04895767197012901,
      "rewards-max/forensics_sinkhorn_reward": 0.9714607894420624,
      "rewards-max/forensics_strict_edge_reward": 0.5051241368055344,
      "rewards-max/forensics_tiou_reward": 0.9426276087760925,
      "rewards/forensics_sinkhorn_reward": 0.6409632563591003,
      "rewards/forensics_strict_edge_reward": 0.05111084412783384,
      "rewards/forensics_tiou_reward": 0.4970713257789612,
      "step": 494
    },
    {
      "completion_length": 24.140625,
      "epoch": 1.5192012288786483,
      "grad_norm": 9.42849063873291,
      "kl": 1.177734375,
      "learning_rate": 6.2e-07,
      "loss": 0.1329,
      "reward": 0.7678490281105042,
      "reward_std": 0.046185326762497425,
      "reward_std_all": 0.04461380746215582,
      "rewards-max/forensics_sinkhorn_reward": 0.9767951667308807,
      "rewards-max/forensics_strict_edge_reward": 0.40391434729099274,
      "rewards-max/forensics_tiou_reward": 0.9630483984947205,
      "rewards/forensics_sinkhorn_reward": 0.7246692478656769,
      "rewards/forensics_strict_edge_reward": 0.05605308525264263,
      "rewards/forensics_tiou_reward": 0.5148997157812119,
      "step": 495
    },
    {
      "completion_length": 22.859375,
      "epoch": 1.522273425499232,
      "grad_norm": 9.106968879699707,
      "kl": 0.4970703125,
      "learning_rate": 6.192307692307692e-07,
      "loss": 0.1051,
      "reward": 0.7525162696838379,
      "reward_std": 0.05323404632508755,
      "reward_std_all": 0.053748197853565216,
      "rewards-max/forensics_sinkhorn_reward": 0.9723758101463318,
      "rewards-max/forensics_strict_edge_reward": 0.3332277685403824,
      "rewards-max/forensics_tiou_reward": 0.9539276659488678,
      "rewards/forensics_sinkhorn_reward": 0.7213783264160156,
      "rewards/forensics_strict_edge_reward": 0.046701787039637566,
      "rewards/forensics_tiou_reward": 0.4958636313676834,
      "step": 496
    },
    {
      "completion_length": 26.484375,
      "epoch": 1.5253456221198156,
      "grad_norm": 11.452098846435547,
      "kl": 0.603515625,
      "learning_rate": 6.184615384615385e-07,
      "loss": 0.4067,
      "reward": 0.724409282207489,
      "reward_std": 0.03739912249147892,
      "reward_std_all": 0.055690374225378036,
      "rewards-max/forensics_sinkhorn_reward": 0.9297823309898376,
      "rewards-max/forensics_strict_edge_reward": 0.6266564428806305,
      "rewards-max/forensics_tiou_reward": 0.8633140325546265,
      "rewards/forensics_sinkhorn_reward": 0.592170923948288,
      "rewards/forensics_strict_edge_reward": 0.03916511684656143,
      "rewards/forensics_tiou_reward": 0.33217132091522217,
      "step": 497
    },
    {
      "completion_length": 24.09375,
      "epoch": 1.5284178187403994,
      "grad_norm": 8.902318000793457,
      "kl": 0.3671875,
      "learning_rate": 6.176923076923076e-07,
      "loss": 0.0471,
      "reward": 0.7585231065750122,
      "reward_std": 0.026802494190633297,
      "reward_std_all": 0.03721419721841812,
      "rewards-max/forensics_sinkhorn_reward": 0.9682784378528595,
      "rewards-max/forensics_strict_edge_reward": 0.5506324470043182,
      "rewards-max/forensics_tiou_reward": 0.9467223286628723,
      "rewards/forensics_sinkhorn_reward": 0.692671924829483,
      "rewards/forensics_strict_edge_reward": 0.04974024556577206,
      "rewards/forensics_tiou_reward": 0.4555783122777939,
      "step": 498
    },
    {
      "completion_length": 25.390625,
      "epoch": 1.531490015360983,
      "grad_norm": 9.412637710571289,
      "kl": 0.552734375,
      "learning_rate": 6.169230769230768e-07,
      "loss": -0.1109,
      "reward": 0.7519569396972656,
      "reward_std": 0.06563579849898815,
      "reward_std_all": 0.06264323927462101,
      "rewards-max/forensics_sinkhorn_reward": 0.9341621696949005,
      "rewards-max/forensics_strict_edge_reward": 0.19946293719112873,
      "rewards-max/forensics_tiou_reward": 0.8745712637901306,
      "rewards/forensics_sinkhorn_reward": 0.6353689134120941,
      "rewards/forensics_strict_edge_reward": 0.021516295499168336,
      "rewards/forensics_tiou_reward": 0.41806378960609436,
      "step": 499
    },
    {
      "completion_length": 26.9375,
      "epoch": 1.5345622119815667,
      "grad_norm": 13.39993953704834,
      "kl": 0.6640625,
      "learning_rate": 6.161538461538462e-07,
      "loss": 0.5913,
      "reward": 0.6674134433269501,
      "reward_std": 0.053049162961542606,
      "reward_std_all": 0.09111480042338371,
      "rewards-max/forensics_sinkhorn_reward": 0.9470375776290894,
      "rewards-max/forensics_strict_edge_reward": 0.08508010301738977,
      "rewards-max/forensics_tiou_reward": 0.9007907509803772,
      "rewards/forensics_sinkhorn_reward": 0.5413097739219666,
      "rewards/forensics_strict_edge_reward": 0.003634421489550732,
      "rewards/forensics_tiou_reward": 0.27688494324684143,
      "step": 500
    },
    {
      "completion_length": 23.796875,
      "epoch": 1.5376344086021505,
      "grad_norm": 10.102437973022461,
      "kl": 0.392578125,
      "learning_rate": 6.153846153846154e-07,
      "loss": 0.2553,
      "reward": 0.6839948892593384,
      "reward_std": 0.05814180430024862,
      "reward_std_all": 0.08861139789223671,
      "rewards-max/forensics_sinkhorn_reward": 0.9766928851604462,
      "rewards-max/forensics_strict_edge_reward": 0.3885071277618408,
      "rewards-max/forensics_tiou_reward": 0.965539425611496,
      "rewards/forensics_sinkhorn_reward": 0.6062208861112595,
      "rewards/forensics_strict_edge_reward": 0.035418190993368626,
      "rewards/forensics_tiou_reward": 0.4063337594270706,
      "step": 501
    },
    {
      "completion_length": 26.390625,
      "epoch": 1.5407066052227343,
      "grad_norm": 8.727957725524902,
      "kl": 0.4365234375,
      "learning_rate": 6.146153846153845e-07,
      "loss": 0.4616,
      "reward": 0.7004228234291077,
      "reward_std": 0.06307137385010719,
      "reward_std_all": 0.08423915132880211,
      "rewards-max/forensics_sinkhorn_reward": 0.9600973427295685,
      "rewards-max/forensics_strict_edge_reward": 0.36510249972343445,
      "rewards-max/forensics_tiou_reward": 0.928239643573761,
      "rewards/forensics_sinkhorn_reward": 0.6080293357372284,
      "rewards/forensics_strict_edge_reward": 0.030529027804732323,
      "rewards/forensics_tiou_reward": 0.4205305725336075,
      "step": 502
    },
    {
      "completion_length": 27.890625,
      "epoch": 1.543778801843318,
      "grad_norm": 9.007235527038574,
      "kl": 0.4814453125,
      "learning_rate": 6.138461538461538e-07,
      "loss": -0.0874,
      "reward": 0.7740837931632996,
      "reward_std": 0.029966607224196196,
      "reward_std_all": 0.03180336393415928,
      "rewards-max/forensics_sinkhorn_reward": 0.9851424396038055,
      "rewards-max/forensics_strict_edge_reward": 0.5555870831012726,
      "rewards-max/forensics_tiou_reward": 0.9869257509708405,
      "rewards/forensics_sinkhorn_reward": 0.7411418855190277,
      "rewards/forensics_strict_edge_reward": 0.05759931355714798,
      "rewards/forensics_tiou_reward": 0.5488272905349731,
      "step": 503
    },
    {
      "completion_length": 25.9375,
      "epoch": 1.5468509984639018,
      "grad_norm": 10.400407791137695,
      "kl": 0.546875,
      "learning_rate": 6.13076923076923e-07,
      "loss": 0.1202,
      "reward": 0.7677581012248993,
      "reward_std": 0.03908269293606281,
      "reward_std_all": 0.04216916114091873,
      "rewards-max/forensics_sinkhorn_reward": 0.9289906024932861,
      "rewards-max/forensics_strict_edge_reward": 0.2155831065028906,
      "rewards-max/forensics_tiou_reward": 0.8627417683601379,
      "rewards/forensics_sinkhorn_reward": 0.6823006868362427,
      "rewards/forensics_strict_edge_reward": 0.016310061677359045,
      "rewards/forensics_tiou_reward": 0.4498632550239563,
      "step": 504
    },
    {
      "completion_length": 22.984375,
      "epoch": 1.5499231950844854,
      "grad_norm": 8.688766479492188,
      "kl": 0.5244140625,
      "learning_rate": 6.123076923076923e-07,
      "loss": 0.0811,
      "reward": 0.7616424858570099,
      "reward_std": 0.055547814816236496,
      "reward_std_all": 0.05286861211061478,
      "rewards-max/forensics_sinkhorn_reward": 0.9811468422412872,
      "rewards-max/forensics_strict_edge_reward": 0.7958523333072662,
      "rewards-max/forensics_tiou_reward": 0.9791205823421478,
      "rewards/forensics_sinkhorn_reward": 0.6967310607433319,
      "rewards/forensics_strict_edge_reward": 0.06817616149783134,
      "rewards/forensics_tiou_reward": 0.4636716991662979,
      "step": 505
    },
    {
      "completion_length": 24.25,
      "epoch": 1.5529953917050692,
      "grad_norm": 7.377137184143066,
      "kl": 0.3876953125,
      "learning_rate": 6.115384615384616e-07,
      "loss": 0.127,
      "reward": 0.7622172832489014,
      "reward_std": 0.013461593072861433,
      "reward_std_all": 0.028285948559641838,
      "rewards-max/forensics_sinkhorn_reward": 0.9758709073066711,
      "rewards-max/forensics_strict_edge_reward": 0.45321224629879,
      "rewards-max/forensics_tiou_reward": 0.962649405002594,
      "rewards/forensics_sinkhorn_reward": 0.6352973580360413,
      "rewards/forensics_strict_edge_reward": 0.025934182107448578,
      "rewards/forensics_tiou_reward": 0.45865464210510254,
      "step": 506
    },
    {
      "completion_length": 26.296875,
      "epoch": 1.5560675883256527,
      "grad_norm": 10.500572204589844,
      "kl": 0.4384765625,
      "learning_rate": 6.107692307692307e-07,
      "loss": 0.1936,
      "reward": 0.7131396234035492,
      "reward_std": 0.08418079279363155,
      "reward_std_all": 0.08805539086461067,
      "rewards-max/forensics_sinkhorn_reward": 0.961548388004303,
      "rewards-max/forensics_strict_edge_reward": 0.084713414311409,
      "rewards-max/forensics_tiou_reward": 0.9281338155269623,
      "rewards/forensics_sinkhorn_reward": 0.6264060139656067,
      "rewards/forensics_strict_edge_reward": 0.004891482647508383,
      "rewards/forensics_tiou_reward": 0.44557712972164154,
      "step": 507
    },
    {
      "completion_length": 28.359375,
      "epoch": 1.5591397849462365,
      "grad_norm": 9.78591537475586,
      "kl": 0.626953125,
      "learning_rate": 6.1e-07,
      "loss": 0.0937,
      "reward": 0.7659182548522949,
      "reward_std": 0.03406408615410328,
      "reward_std_all": 0.03766162507236004,
      "rewards-max/forensics_sinkhorn_reward": 0.943540632724762,
      "rewards-max/forensics_strict_edge_reward": 0.3550787717103958,
      "rewards-max/forensics_tiou_reward": 0.8999111652374268,
      "rewards/forensics_sinkhorn_reward": 0.6944225430488586,
      "rewards/forensics_strict_edge_reward": 0.04071158356964588,
      "rewards/forensics_tiou_reward": 0.47878946363925934,
      "step": 508
    },
    {
      "completion_length": 24.390625,
      "epoch": 1.5622119815668203,
      "grad_norm": 9.556239128112793,
      "kl": 0.53515625,
      "learning_rate": 6.092307692307692e-07,
      "loss": 0.2352,
      "reward": 0.740987628698349,
      "reward_std": 0.0470479279756546,
      "reward_std_all": 0.056448476389050484,
      "rewards-max/forensics_sinkhorn_reward": 0.9838499426841736,
      "rewards-max/forensics_strict_edge_reward": 0.5192448049783707,
      "rewards-max/forensics_tiou_reward": 0.9848227500915527,
      "rewards/forensics_sinkhorn_reward": 0.6753365993499756,
      "rewards/forensics_strict_edge_reward": 0.04849177785217762,
      "rewards/forensics_tiou_reward": 0.509553849697113,
      "step": 509
    },
    {
      "completion_length": 22.953125,
      "epoch": 1.565284178187404,
      "grad_norm": 13.546948432922363,
      "kl": 0.384765625,
      "learning_rate": 6.084615384615385e-07,
      "loss": 0.0903,
      "reward": 0.7334685027599335,
      "reward_std": 0.04193994775414467,
      "reward_std_all": 0.05821720138192177,
      "rewards-max/forensics_sinkhorn_reward": 0.9622537195682526,
      "rewards-max/forensics_strict_edge_reward": 0.41129170358181,
      "rewards-max/forensics_tiou_reward": 0.9359394311904907,
      "rewards/forensics_sinkhorn_reward": 0.6343879699707031,
      "rewards/forensics_strict_edge_reward": 0.02740293089300394,
      "rewards/forensics_tiou_reward": 0.37002983689308167,
      "step": 510
    },
    {
      "completion_length": 23.9375,
      "epoch": 1.5683563748079878,
      "grad_norm": 9.072713851928711,
      "kl": 0.439453125,
      "learning_rate": 6.076923076923076e-07,
      "loss": 0.1184,
      "reward": 0.7797021269798279,
      "reward_std": 0.03510421235114336,
      "reward_std_all": 0.033769406378269196,
      "rewards-max/forensics_sinkhorn_reward": 0.9838775992393494,
      "rewards-max/forensics_strict_edge_reward": 0.4905853122472763,
      "rewards-max/forensics_tiou_reward": 0.9843424558639526,
      "rewards/forensics_sinkhorn_reward": 0.7044509649276733,
      "rewards/forensics_strict_edge_reward": 0.06854842603206635,
      "rewards/forensics_tiou_reward": 0.5022091716527939,
      "step": 511
    },
    {
      "completion_length": 21.625,
      "epoch": 1.5714285714285714,
      "grad_norm": 11.03469181060791,
      "kl": 0.591796875,
      "learning_rate": 6.069230769230768e-07,
      "loss": -0.0788,
      "reward": 0.7447215616703033,
      "reward_std": 0.06158973649144173,
      "reward_std_all": 0.06715194880962372,
      "rewards-max/forensics_sinkhorn_reward": 0.9736148416996002,
      "rewards-max/forensics_strict_edge_reward": 0.3803163915872574,
      "rewards-max/forensics_tiou_reward": 0.9597395062446594,
      "rewards/forensics_sinkhorn_reward": 0.7239767014980316,
      "rewards/forensics_strict_edge_reward": 0.03759920783340931,
      "rewards/forensics_tiou_reward": 0.5186914801597595,
      "step": 512
    },
    {
      "completion_length": 25.078125,
      "epoch": 1.5745007680491552,
      "grad_norm": 8.718282699584961,
      "kl": 0.5048828125,
      "learning_rate": 6.061538461538462e-07,
      "loss": 0.1287,
      "reward": 0.7865814566612244,
      "reward_std": 0.0228535826317966,
      "reward_std_all": 0.024076296016573906,
      "rewards-max/forensics_sinkhorn_reward": 0.9289024174213409,
      "rewards-max/forensics_strict_edge_reward": 0.07982965372502804,
      "rewards-max/forensics_tiou_reward": 0.8445302248001099,
      "rewards/forensics_sinkhorn_reward": 0.6569907367229462,
      "rewards/forensics_strict_edge_reward": 0.0074720680131576955,
      "rewards/forensics_tiou_reward": 0.42333173751831055,
      "step": 513
    },
    {
      "completion_length": 23.859375,
      "epoch": 1.5775729646697387,
      "grad_norm": 11.142707824707031,
      "kl": 0.564453125,
      "learning_rate": 6.053846153846154e-07,
      "loss": 0.2259,
      "reward": 0.7246575951576233,
      "reward_std": 0.06686029024422169,
      "reward_std_all": 0.07446232251822948,
      "rewards-max/forensics_sinkhorn_reward": 0.9811519682407379,
      "rewards-max/forensics_strict_edge_reward": 0.6718102097511292,
      "rewards-max/forensics_tiou_reward": 0.9795336723327637,
      "rewards/forensics_sinkhorn_reward": 0.6478096842765808,
      "rewards/forensics_strict_edge_reward": 0.04588477313518524,
      "rewards/forensics_tiou_reward": 0.43348756432533264,
      "step": 514
    },
    {
      "completion_length": 26.5625,
      "epoch": 1.5806451612903225,
      "grad_norm": 8.585702896118164,
      "kl": 0.5078125,
      "learning_rate": 6.046153846153847e-07,
      "loss": 0.1916,
      "reward": 0.7034004032611847,
      "reward_std": 0.10325799882411957,
      "reward_std_all": 0.10471731424331665,
      "rewards-max/forensics_sinkhorn_reward": 0.981769323348999,
      "rewards-max/forensics_strict_edge_reward": 0.6052050739526749,
      "rewards-max/forensics_tiou_reward": 0.9797244071960449,
      "rewards/forensics_sinkhorn_reward": 0.6237743496894836,
      "rewards/forensics_strict_edge_reward": 0.05190142244100571,
      "rewards/forensics_tiou_reward": 0.4273584187030792,
      "step": 515
    },
    {
      "completion_length": 24.34375,
      "epoch": 1.5837173579109063,
      "grad_norm": 10.080856323242188,
      "kl": 0.486328125,
      "learning_rate": 6.038461538461538e-07,
      "loss": -0.012,
      "reward": 0.7557598054409027,
      "reward_std": 0.060854725539684296,
      "reward_std_all": 0.05763611942529678,
      "rewards-max/forensics_sinkhorn_reward": 0.9770813584327698,
      "rewards-max/forensics_strict_edge_reward": 0.824172168970108,
      "rewards-max/forensics_tiou_reward": 0.9689697027206421,
      "rewards/forensics_sinkhorn_reward": 0.7221381366252899,
      "rewards/forensics_strict_edge_reward": 0.1449608877301216,
      "rewards/forensics_tiou_reward": 0.585309773683548,
      "step": 516
    },
    {
      "completion_length": 21.546875,
      "epoch": 1.58678955453149,
      "grad_norm": 7.183210372924805,
      "kl": 0.51171875,
      "learning_rate": 6.03076923076923e-07,
      "loss": 0.3292,
      "reward": 0.7430444359779358,
      "reward_std": 0.0754714086651802,
      "reward_std_all": 0.07354752346873283,
      "rewards-max/forensics_sinkhorn_reward": 0.952062577009201,
      "rewards-max/forensics_strict_edge_reward": 0.13686756044626236,
      "rewards-max/forensics_tiou_reward": 0.8945659101009369,
      "rewards/forensics_sinkhorn_reward": 0.6509764790534973,
      "rewards/forensics_strict_edge_reward": 0.009350916370749474,
      "rewards/forensics_tiou_reward": 0.41344229876995087,
      "step": 517
    },
    {
      "completion_length": 24.515625,
      "epoch": 1.5898617511520738,
      "grad_norm": 10.106578826904297,
      "kl": 0.580078125,
      "learning_rate": 6.023076923076923e-07,
      "loss": 0.3204,
      "reward": 0.6871753036975861,
      "reward_std": 0.06917204428464174,
      "reward_std_all": 0.09043360128998756,
      "rewards-max/forensics_sinkhorn_reward": 0.9840236008167267,
      "rewards-max/forensics_strict_edge_reward": 0.5919846817851067,
      "rewards-max/forensics_tiou_reward": 0.9828638434410095,
      "rewards/forensics_sinkhorn_reward": 0.5892327427864075,
      "rewards/forensics_strict_edge_reward": 0.07297483459115028,
      "rewards/forensics_tiou_reward": 0.4088207185268402,
      "step": 518
    },
    {
      "completion_length": 22.03125,
      "epoch": 1.5929339477726574,
      "grad_norm": 10.618659973144531,
      "kl": 0.48046875,
      "learning_rate": 6.015384615384615e-07,
      "loss": 0.2376,
      "reward": 0.7332921624183655,
      "reward_std": 0.06914109457284212,
      "reward_std_all": 0.07461390178650618,
      "rewards-max/forensics_sinkhorn_reward": 0.9625665843486786,
      "rewards-max/forensics_strict_edge_reward": 0.3730776160955429,
      "rewards-max/forensics_tiou_reward": 0.9282371401786804,
      "rewards/forensics_sinkhorn_reward": 0.6637627184391022,
      "rewards/forensics_strict_edge_reward": 0.040838029235601425,
      "rewards/forensics_tiou_reward": 0.477274551987648,
      "step": 519
    },
    {
      "completion_length": 25.6875,
      "epoch": 1.5960061443932412,
      "grad_norm": 9.115177154541016,
      "kl": 0.4306640625,
      "learning_rate": 6.007692307692307e-07,
      "loss": 0.225,
      "reward": 0.6650347113609314,
      "reward_std": 0.10610242933034897,
      "reward_std_all": 0.11906605586409569,
      "rewards-max/forensics_sinkhorn_reward": 0.9687577188014984,
      "rewards-max/forensics_strict_edge_reward": 0.47277915477752686,
      "rewards-max/forensics_tiou_reward": 0.9502709805965424,
      "rewards/forensics_sinkhorn_reward": 0.5700543820858002,
      "rewards/forensics_strict_edge_reward": 0.043937692418694496,
      "rewards/forensics_tiou_reward": 0.35816289484500885,
      "step": 520
    },
    {
      "completion_length": 24.921875,
      "epoch": 1.5990783410138247,
      "grad_norm": 9.39517879486084,
      "kl": 0.396484375,
      "learning_rate": 6e-07,
      "loss": 0.172,
      "reward": 0.6972700953483582,
      "reward_std": 0.049636343494057655,
      "reward_std_all": 0.07776741497218609,
      "rewards-max/forensics_sinkhorn_reward": 0.9073953628540039,
      "rewards-max/forensics_strict_edge_reward": 0.26066684257239103,
      "rewards-max/forensics_tiou_reward": 0.8087548017501831,
      "rewards/forensics_sinkhorn_reward": 0.6202430427074432,
      "rewards/forensics_strict_edge_reward": 0.036075483410968445,
      "rewards/forensics_tiou_reward": 0.31661464273929596,
      "step": 521
    },
    {
      "completion_length": 22.546875,
      "epoch": 1.6021505376344085,
      "grad_norm": 11.927628517150879,
      "kl": 0.40234375,
      "learning_rate": 5.992307692307692e-07,
      "loss": -0.0261,
      "reward": 0.736598551273346,
      "reward_std": 0.054375067353248596,
      "reward_std_all": 0.05997077748179436,
      "rewards-max/forensics_sinkhorn_reward": 0.9777185022830963,
      "rewards-max/forensics_strict_edge_reward": 0.6215263903141022,
      "rewards-max/forensics_tiou_reward": 0.9669695496559143,
      "rewards/forensics_sinkhorn_reward": 0.6192023754119873,
      "rewards/forensics_strict_edge_reward": 0.05681795906275511,
      "rewards/forensics_tiou_reward": 0.3968418091535568,
      "step": 522
    },
    {
      "completion_length": 25.078125,
      "epoch": 1.6052227342549923,
      "grad_norm": 9.295808792114258,
      "kl": 0.4267578125,
      "learning_rate": 5.984615384615385e-07,
      "loss": -0.0173,
      "reward": 0.71058589220047,
      "reward_std": 0.03950000798795372,
      "reward_std_all": 0.06867030449211597,
      "rewards-max/forensics_sinkhorn_reward": 0.9666501879692078,
      "rewards-max/forensics_strict_edge_reward": 0.4158726632595062,
      "rewards-max/forensics_tiou_reward": 0.940361499786377,
      "rewards/forensics_sinkhorn_reward": 0.6553550958633423,
      "rewards/forensics_strict_edge_reward": 0.035995678044855595,
      "rewards/forensics_tiou_reward": 0.45719730854034424,
      "step": 523
    },
    {
      "completion_length": 21.984375,
      "epoch": 1.608294930875576,
      "grad_norm": 11.9675931930542,
      "kl": 0.3994140625,
      "learning_rate": 5.976923076923077e-07,
      "loss": 0.1834,
      "reward": 0.7644036412239075,
      "reward_std": 0.05488167889416218,
      "reward_std_all": 0.05155613087117672,
      "rewards-max/forensics_sinkhorn_reward": 0.9766814708709717,
      "rewards-max/forensics_strict_edge_reward": 0.3678794503211975,
      "rewards-max/forensics_tiou_reward": 0.9686234891414642,
      "rewards/forensics_sinkhorn_reward": 0.7330040335655212,
      "rewards/forensics_strict_edge_reward": 0.059555198065936565,
      "rewards/forensics_tiou_reward": 0.5066210627555847,
      "step": 524
    },
    {
      "completion_length": 22.40625,
      "epoch": 1.6113671274961598,
      "grad_norm": 9.774992942810059,
      "kl": 0.41796875,
      "learning_rate": 5.969230769230768e-07,
      "loss": 0.3615,
      "reward": 0.6626931428909302,
      "reward_std": 0.09541890770196915,
      "reward_std_all": 0.11224716156721115,
      "rewards-max/forensics_sinkhorn_reward": 0.9824004173278809,
      "rewards-max/forensics_strict_edge_reward": 0.7311047911643982,
      "rewards-max/forensics_tiou_reward": 0.9823416471481323,
      "rewards/forensics_sinkhorn_reward": 0.5627035796642303,
      "rewards/forensics_strict_edge_reward": 0.04454144276678562,
      "rewards/forensics_tiou_reward": 0.3281673789024353,
      "step": 525
    },
    {
      "completion_length": 27.6875,
      "epoch": 1.6144393241167436,
      "grad_norm": 11.458837509155273,
      "kl": 0.4482421875,
      "learning_rate": 5.961538461538461e-07,
      "loss": 0.118,
      "reward": 0.7845608294010162,
      "reward_std": 0.025300596840679646,
      "reward_std_all": 0.025008386932313442,
      "rewards-max/forensics_sinkhorn_reward": 0.976762056350708,
      "rewards-max/forensics_strict_edge_reward": 0.3850628435611725,
      "rewards-max/forensics_tiou_reward": 0.9679021239280701,
      "rewards/forensics_sinkhorn_reward": 0.7600843906402588,
      "rewards/forensics_strict_edge_reward": 0.07528371177613735,
      "rewards/forensics_tiou_reward": 0.5607739090919495,
      "step": 526
    },
    {
      "completion_length": 23.390625,
      "epoch": 1.6175115207373272,
      "grad_norm": 6.762235641479492,
      "kl": 0.365234375,
      "learning_rate": 5.953846153846154e-07,
      "loss": 0.0198,
      "reward": 0.7733538448810577,
      "reward_std": 0.039975838735699654,
      "reward_std_all": 0.03884996101260185,
      "rewards-max/forensics_sinkhorn_reward": 0.9123793840408325,
      "rewards-max/forensics_strict_edge_reward": 0.6915990114212036,
      "rewards-max/forensics_tiou_reward": 0.839792937040329,
      "rewards/forensics_sinkhorn_reward": 0.6995562314987183,
      "rewards/forensics_strict_edge_reward": 0.09432783164083958,
      "rewards/forensics_tiou_reward": 0.5307851880788803,
      "step": 527
    },
    {
      "completion_length": 21.734375,
      "epoch": 1.620583717357911,
      "grad_norm": 7.895699501037598,
      "kl": 0.490234375,
      "learning_rate": 5.946153846153847e-07,
      "loss": 0.0893,
      "reward": 0.6976286768913269,
      "reward_std": 0.05882357992231846,
      "reward_std_all": 0.0816822275519371,
      "rewards-max/forensics_sinkhorn_reward": 0.979179322719574,
      "rewards-max/forensics_strict_edge_reward": 0.742465078830719,
      "rewards-max/forensics_tiou_reward": 0.9762685894966125,
      "rewards/forensics_sinkhorn_reward": 0.6142599880695343,
      "rewards/forensics_strict_edge_reward": 0.08244302496314049,
      "rewards/forensics_tiou_reward": 0.4166431725025177,
      "step": 528
    },
    {
      "completion_length": 24.4375,
      "epoch": 1.6236559139784945,
      "grad_norm": 10.182711601257324,
      "kl": 0.4404296875,
      "learning_rate": 5.938461538461538e-07,
      "loss": 0.1386,
      "reward": 0.7475475668907166,
      "reward_std": 0.057729871943593025,
      "reward_std_all": 0.05939491093158722,
      "rewards-max/forensics_sinkhorn_reward": 0.9771408140659332,
      "rewards-max/forensics_strict_edge_reward": 0.6282242834568024,
      "rewards-max/forensics_tiou_reward": 0.9690362811088562,
      "rewards/forensics_sinkhorn_reward": 0.6617728769779205,
      "rewards/forensics_strict_edge_reward": 0.06699417717754841,
      "rewards/forensics_tiou_reward": 0.4856681227684021,
      "step": 529
    },
    {
      "completion_length": 22.65625,
      "epoch": 1.6267281105990783,
      "grad_norm": 7.588809967041016,
      "kl": 0.49609375,
      "learning_rate": 5.93076923076923e-07,
      "loss": -0.01,
      "reward": 0.7785310447216034,
      "reward_std": 0.03178897826001048,
      "reward_std_all": 0.03353715850971639,
      "rewards-max/forensics_sinkhorn_reward": 0.978787362575531,
      "rewards-max/forensics_strict_edge_reward": 0.6706924736499786,
      "rewards-max/forensics_tiou_reward": 0.9738681614398956,
      "rewards/forensics_sinkhorn_reward": 0.697808712720871,
      "rewards/forensics_strict_edge_reward": 0.06452317908406258,
      "rewards/forensics_tiou_reward": 0.5094455778598785,
      "step": 530
    },
    {
      "completion_length": 25.28125,
      "epoch": 1.629800307219662,
      "grad_norm": 8.16008472442627,
      "kl": 0.3935546875,
      "learning_rate": 5.923076923076923e-07,
      "loss": 0.1074,
      "reward": 0.73261359333992,
      "reward_std": 0.050173183903098106,
      "reward_std_all": 0.05918198823928833,
      "rewards-max/forensics_sinkhorn_reward": 0.9819170236587524,
      "rewards-max/forensics_strict_edge_reward": 0.6535452902317047,
      "rewards-max/forensics_tiou_reward": 0.980717808008194,
      "rewards/forensics_sinkhorn_reward": 0.6417645812034607,
      "rewards/forensics_strict_edge_reward": 0.040201250463724136,
      "rewards/forensics_tiou_reward": 0.40419721603393555,
      "step": 531
    },
    {
      "completion_length": 27.53125,
      "epoch": 1.6328725038402458,
      "grad_norm": 9.93983268737793,
      "kl": 0.578125,
      "learning_rate": 5.915384615384615e-07,
      "loss": 0.3717,
      "reward": 0.7114652991294861,
      "reward_std": 0.08106765896081924,
      "reward_std_all": 0.08541569858789444,
      "rewards-max/forensics_sinkhorn_reward": 0.9858860075473785,
      "rewards-max/forensics_strict_edge_reward": 0.7933230996131897,
      "rewards-max/forensics_tiou_reward": 0.9899345636367798,
      "rewards/forensics_sinkhorn_reward": 0.5988418161869049,
      "rewards/forensics_strict_edge_reward": 0.05142940208315849,
      "rewards/forensics_tiou_reward": 0.40266838669776917,
      "step": 532
    },
    {
      "completion_length": 25.6875,
      "epoch": 1.6359447004608296,
      "grad_norm": 7.6877288818359375,
      "kl": 0.4423828125,
      "learning_rate": 5.907692307692307e-07,
      "loss": 0.2947,
      "reward": 0.7590085864067078,
      "reward_std": 0.04028348345309496,
      "reward_std_all": 0.04494927637279034,
      "rewards-max/forensics_sinkhorn_reward": 0.9824303686618805,
      "rewards-max/forensics_strict_edge_reward": 0.6597102135419846,
      "rewards-max/forensics_tiou_reward": 0.983662486076355,
      "rewards/forensics_sinkhorn_reward": 0.6859167218208313,
      "rewards/forensics_strict_edge_reward": 0.05156360752880573,
      "rewards/forensics_tiou_reward": 0.47623634338378906,
      "step": 533
    },
    {
      "completion_length": 21.03125,
      "epoch": 1.6390168970814132,
      "grad_norm": 9.147974967956543,
      "kl": 0.4423828125,
      "learning_rate": 5.9e-07,
      "loss": 0.1457,
      "reward": 0.7191123068332672,
      "reward_std": 0.04569824039936066,
      "reward_std_all": 0.060770658776164055,
      "rewards-max/forensics_sinkhorn_reward": 0.9670785367488861,
      "rewards-max/forensics_strict_edge_reward": 0.26193115580827,
      "rewards-max/forensics_tiou_reward": 0.9401440024375916,
      "rewards/forensics_sinkhorn_reward": 0.6169931590557098,
      "rewards/forensics_strict_edge_reward": 0.021749238949269056,
      "rewards/forensics_tiou_reward": 0.3240158408880234,
      "step": 534
    },
    {
      "completion_length": 24.09375,
      "epoch": 1.642089093701997,
      "grad_norm": 10.271404266357422,
      "kl": 0.380859375,
      "learning_rate": 5.892307692307692e-07,
      "loss": 0.2226,
      "reward": 0.7127095460891724,
      "reward_std": 0.04707490932196379,
      "reward_std_all": 0.06866844370961189,
      "rewards-max/forensics_sinkhorn_reward": 0.976148933172226,
      "rewards-max/forensics_strict_edge_reward": 0.2919003367424011,
      "rewards-max/forensics_tiou_reward": 0.9632066786289215,
      "rewards/forensics_sinkhorn_reward": 0.5914875864982605,
      "rewards/forensics_strict_edge_reward": 0.026346100494265556,
      "rewards/forensics_tiou_reward": 0.3854388892650604,
      "step": 535
    },
    {
      "completion_length": 25.46875,
      "epoch": 1.6451612903225805,
      "grad_norm": 7.752186298370361,
      "kl": 0.46875,
      "learning_rate": 5.884615384615385e-07,
      "loss": 0.2897,
      "reward": 0.7056735157966614,
      "reward_std": 0.0842539630830288,
      "reward_std_all": 0.09571394696831703,
      "rewards-max/forensics_sinkhorn_reward": 0.9573953151702881,
      "rewards-max/forensics_strict_edge_reward": 0.3399255871772766,
      "rewards-max/forensics_tiou_reward": 0.9268338978290558,
      "rewards/forensics_sinkhorn_reward": 0.6541038155555725,
      "rewards/forensics_strict_edge_reward": 0.04264266416430473,
      "rewards/forensics_tiou_reward": 0.46715840697288513,
      "step": 536
    },
    {
      "completion_length": 22.90625,
      "epoch": 1.6482334869431643,
      "grad_norm": 9.056018829345703,
      "kl": 0.7578125,
      "learning_rate": 5.876923076923077e-07,
      "loss": 0.1339,
      "reward": 0.7617098987102509,
      "reward_std": 0.040481146425008774,
      "reward_std_all": 0.0418685432523489,
      "rewards-max/forensics_sinkhorn_reward": 0.9623379707336426,
      "rewards-max/forensics_strict_edge_reward": 0.4909685105085373,
      "rewards-max/forensics_tiou_reward": 0.9274797439575195,
      "rewards/forensics_sinkhorn_reward": 0.6853280365467072,
      "rewards/forensics_strict_edge_reward": 0.06734594982117414,
      "rewards/forensics_tiou_reward": 0.4960471987724304,
      "step": 537
    },
    {
      "completion_length": 24.421875,
      "epoch": 1.651305683563748,
      "grad_norm": 8.03991985321045,
      "kl": 0.4091796875,
      "learning_rate": 5.869230769230768e-07,
      "loss": 0.0309,
      "reward": 0.7643337547779083,
      "reward_std": 0.0434592105448246,
      "reward_std_all": 0.04416347201913595,
      "rewards-max/forensics_sinkhorn_reward": 0.9739974141120911,
      "rewards-max/forensics_strict_edge_reward": 0.4899691939353943,
      "rewards-max/forensics_tiou_reward": 0.9676641523838043,
      "rewards/forensics_sinkhorn_reward": 0.7247742712497711,
      "rewards/forensics_strict_edge_reward": 0.0434699859470129,
      "rewards/forensics_tiou_reward": 0.5215798616409302,
      "step": 538
    },
    {
      "completion_length": 26.5625,
      "epoch": 1.6543778801843319,
      "grad_norm": 9.63774299621582,
      "kl": 0.48046875,
      "learning_rate": 5.861538461538461e-07,
      "loss": -0.0703,
      "reward": 0.7058911025524139,
      "reward_std": 0.05272211320698261,
      "reward_std_all": 0.07565205171704292,
      "rewards-max/forensics_sinkhorn_reward": 0.9657326936721802,
      "rewards-max/forensics_strict_edge_reward": 0.4616380035877228,
      "rewards-max/forensics_tiou_reward": 0.942882239818573,
      "rewards/forensics_sinkhorn_reward": 0.6416614055633545,
      "rewards/forensics_strict_edge_reward": 0.04874487966299057,
      "rewards/forensics_tiou_reward": 0.43358585238456726,
      "step": 539
    },
    {
      "completion_length": 29.1875,
      "epoch": 1.6574500768049156,
      "grad_norm": 12.585140228271484,
      "kl": 0.919921875,
      "learning_rate": 5.853846153846154e-07,
      "loss": 0.3394,
      "reward": 0.7388626635074615,
      "reward_std": 0.04871348291635513,
      "reward_std_all": 0.05771918594837189,
      "rewards-max/forensics_sinkhorn_reward": 0.9737769067287445,
      "rewards-max/forensics_strict_edge_reward": 0.4142165631055832,
      "rewards-max/forensics_tiou_reward": 0.9597022831439972,
      "rewards/forensics_sinkhorn_reward": 0.6257129609584808,
      "rewards/forensics_strict_edge_reward": 0.04833949729800224,
      "rewards/forensics_tiou_reward": 0.474589005112648,
      "step": 540
    },
    {
      "completion_length": 23.515625,
      "epoch": 1.6605222734254992,
      "grad_norm": 8.001949310302734,
      "kl": 0.4990234375,
      "learning_rate": 5.846153846153847e-07,
      "loss": 0.0168,
      "reward": 0.7725372314453125,
      "reward_std": 0.02505988162010908,
      "reward_std_all": 0.030333097092807293,
      "rewards-max/forensics_sinkhorn_reward": 0.9870869517326355,
      "rewards-max/forensics_strict_edge_reward": 0.8106843829154968,
      "rewards-max/forensics_tiou_reward": 0.9945466220378876,
      "rewards/forensics_sinkhorn_reward": 0.7039337456226349,
      "rewards/forensics_strict_edge_reward": 0.08412478491663933,
      "rewards/forensics_tiou_reward": 0.5178523063659668,
      "step": 541
    },
    {
      "completion_length": 23.953125,
      "epoch": 1.663594470046083,
      "grad_norm": 9.359672546386719,
      "kl": 0.609375,
      "learning_rate": 5.838461538461538e-07,
      "loss": 0.1805,
      "reward": 0.7793063521385193,
      "reward_std": 0.02423115074634552,
      "reward_std_all": 0.027488925494253635,
      "rewards-max/forensics_sinkhorn_reward": 0.9805929958820343,
      "rewards-max/forensics_strict_edge_reward": 0.6593321859836578,
      "rewards-max/forensics_tiou_reward": 0.9799731075763702,
      "rewards/forensics_sinkhorn_reward": 0.7004181742668152,
      "rewards/forensics_strict_edge_reward": 0.05975761637091637,
      "rewards/forensics_tiou_reward": 0.5118690431118011,
      "step": 542
    },
    {
      "completion_length": 24.109375,
      "epoch": 1.6666666666666665,
      "grad_norm": 9.165403366088867,
      "kl": 0.4111328125,
      "learning_rate": 5.83076923076923e-07,
      "loss": 0.2097,
      "reward": 0.7428118586540222,
      "reward_std": 0.04932361841201782,
      "reward_std_all": 0.05739833414554596,
      "rewards-max/forensics_sinkhorn_reward": 0.9780130386352539,
      "rewards-max/forensics_strict_edge_reward": 0.5905451774597168,
      "rewards-max/forensics_tiou_reward": 0.9726217687129974,
      "rewards/forensics_sinkhorn_reward": 0.6183100640773773,
      "rewards/forensics_strict_edge_reward": 0.07004008255898952,
      "rewards/forensics_tiou_reward": 0.4230218827724457,
      "step": 543
    },
    {
      "completion_length": 28.140625,
      "epoch": 1.6697388632872503,
      "grad_norm": 7.6951165199279785,
      "kl": 0.6357421875,
      "learning_rate": 5.823076923076923e-07,
      "loss": 0.1973,
      "reward": 0.7324918508529663,
      "reward_std": 0.05340409278869629,
      "reward_std_all": 0.06117321364581585,
      "rewards-max/forensics_sinkhorn_reward": 0.9644751846790314,
      "rewards-max/forensics_strict_edge_reward": 0.43737562000751495,
      "rewards-max/forensics_tiou_reward": 0.9366917014122009,
      "rewards/forensics_sinkhorn_reward": 0.6388637125492096,
      "rewards/forensics_strict_edge_reward": 0.05266566015779972,
      "rewards/forensics_tiou_reward": 0.44487637281417847,
      "step": 544
    },
    {
      "completion_length": 23.765625,
      "epoch": 1.672811059907834,
      "grad_norm": 10.04746150970459,
      "kl": 0.3515625,
      "learning_rate": 5.815384615384615e-07,
      "loss": 0.0442,
      "reward": 0.7587890625,
      "reward_std": 0.055221322225406766,
      "reward_std_all": 0.05440614698454738,
      "rewards-max/forensics_sinkhorn_reward": 0.9132061302661896,
      "rewards-max/forensics_strict_edge_reward": 0.265866382047534,
      "rewards-max/forensics_tiou_reward": 0.8227893114089966,
      "rewards/forensics_sinkhorn_reward": 0.6731162071228027,
      "rewards/forensics_strict_edge_reward": 0.039995403261855245,
      "rewards/forensics_tiou_reward": 0.5016593933105469,
      "step": 545
    },
    {
      "completion_length": 24.71875,
      "epoch": 1.6758832565284179,
      "grad_norm": 9.757089614868164,
      "kl": 0.4140625,
      "learning_rate": 5.807692307692307e-07,
      "loss": 0.1753,
      "reward": 0.7099589705467224,
      "reward_std": 0.041170903481543064,
      "reward_std_all": 0.0668822368606925,
      "rewards-max/forensics_sinkhorn_reward": 0.9738285839557648,
      "rewards-max/forensics_strict_edge_reward": 0.5579978227615356,
      "rewards-max/forensics_tiou_reward": 0.962613433599472,
      "rewards/forensics_sinkhorn_reward": 0.6272741705179214,
      "rewards/forensics_strict_edge_reward": 0.04505594726651907,
      "rewards/forensics_tiou_reward": 0.4521457701921463,
      "step": 546
    },
    {
      "completion_length": 25.59375,
      "epoch": 1.6789554531490016,
      "grad_norm": 9.486364364624023,
      "kl": 0.662109375,
      "learning_rate": 5.8e-07,
      "loss": 0.2384,
      "reward": 0.7723639011383057,
      "reward_std": 0.034495290368795395,
      "reward_std_all": 0.0349587369710207,
      "rewards-max/forensics_sinkhorn_reward": 0.9784802794456482,
      "rewards-max/forensics_strict_edge_reward": 0.711992472410202,
      "rewards-max/forensics_tiou_reward": 0.9739551246166229,
      "rewards/forensics_sinkhorn_reward": 0.6601786911487579,
      "rewards/forensics_strict_edge_reward": 0.06361385993659496,
      "rewards/forensics_tiou_reward": 0.5116853564977646,
      "step": 547
    },
    {
      "completion_length": 23.890625,
      "epoch": 1.6820276497695854,
      "grad_norm": 8.67254638671875,
      "kl": 0.65625,
      "learning_rate": 5.792307692307692e-07,
      "loss": 0.2431,
      "reward": 0.7331402003765106,
      "reward_std": 0.07655412703752518,
      "reward_std_all": 0.07871909812092781,
      "rewards-max/forensics_sinkhorn_reward": 0.9429672062397003,
      "rewards-max/forensics_strict_edge_reward": 0.39734840393066406,
      "rewards-max/forensics_tiou_reward": 0.8877113461494446,
      "rewards/forensics_sinkhorn_reward": 0.5972388982772827,
      "rewards/forensics_strict_edge_reward": 0.031755562871694565,
      "rewards/forensics_tiou_reward": 0.40686580538749695,
      "step": 548
    },
    {
      "completion_length": 25.03125,
      "epoch": 1.685099846390169,
      "grad_norm": 6.0762176513671875,
      "kl": 0.625,
      "learning_rate": 5.784615384615385e-07,
      "loss": 0.0396,
      "reward": 0.7516117990016937,
      "reward_std": 0.036525944247841835,
      "reward_std_all": 0.04571564123034477,
      "rewards-max/forensics_sinkhorn_reward": 0.9715235829353333,
      "rewards-max/forensics_strict_edge_reward": 0.35313303023576736,
      "rewards-max/forensics_tiou_reward": 0.9442373514175415,
      "rewards/forensics_sinkhorn_reward": 0.699618399143219,
      "rewards/forensics_strict_edge_reward": 0.034491296857595444,
      "rewards/forensics_tiou_reward": 0.4948142319917679,
      "step": 549
    },
    {
      "completion_length": 24.875,
      "epoch": 1.6881720430107527,
      "grad_norm": 9.541680335998535,
      "kl": 0.3583984375,
      "learning_rate": 5.776923076923077e-07,
      "loss": -0.0545,
      "reward": 0.7803973257541656,
      "reward_std": 0.020805551670491695,
      "reward_std_all": 0.025599461048841476,
      "rewards-max/forensics_sinkhorn_reward": 0.9628866910934448,
      "rewards-max/forensics_strict_edge_reward": 0.350901335477829,
      "rewards-max/forensics_tiou_reward": 0.9360342919826508,
      "rewards/forensics_sinkhorn_reward": 0.7187386453151703,
      "rewards/forensics_strict_edge_reward": 0.0263533778488636,
      "rewards/forensics_tiou_reward": 0.4626421183347702,
      "step": 550
    },
    {
      "completion_length": 22.90625,
      "epoch": 1.6912442396313363,
      "grad_norm": 10.647790908813477,
      "kl": 0.4990234375,
      "learning_rate": 5.769230769230768e-07,
      "loss": 0.1729,
      "reward": 0.7597472369670868,
      "reward_std": 0.05599492788314819,
      "reward_std_all": 0.0548255629837513,
      "rewards-max/forensics_sinkhorn_reward": 0.9766599237918854,
      "rewards-max/forensics_strict_edge_reward": 0.6149742305278778,
      "rewards-max/forensics_tiou_reward": 0.9658802449703217,
      "rewards/forensics_sinkhorn_reward": 0.7228062748908997,
      "rewards/forensics_strict_edge_reward": 0.07667197659611702,
      "rewards/forensics_tiou_reward": 0.5385032892227173,
      "step": 551
    },
    {
      "completion_length": 24.984375,
      "epoch": 1.69431643625192,
      "grad_norm": 10.660898208618164,
      "kl": 0.3896484375,
      "learning_rate": 5.761538461538461e-07,
      "loss": -0.0087,
      "reward": 0.7887901961803436,
      "reward_std": 0.014186218846589327,
      "reward_std_all": 0.017759107053279877,
      "rewards-max/forensics_sinkhorn_reward": 0.9741193354129791,
      "rewards-max/forensics_strict_edge_reward": 0.35759691894054413,
      "rewards-max/forensics_tiou_reward": 0.9638043940067291,
      "rewards/forensics_sinkhorn_reward": 0.726997047662735,
      "rewards/forensics_strict_edge_reward": 0.03506939113140106,
      "rewards/forensics_tiou_reward": 0.5539747476577759,
      "step": 552
    },
    {
      "completion_length": 23.890625,
      "epoch": 1.6973886328725039,
      "grad_norm": 8.896968841552734,
      "kl": 0.4208984375,
      "learning_rate": 5.753846153846153e-07,
      "loss": 0.1535,
      "reward": 0.7332258224487305,
      "reward_std": 0.06637955456972122,
      "reward_std_all": 0.07058033160865307,
      "rewards-max/forensics_sinkhorn_reward": 0.9409318566322327,
      "rewards-max/forensics_strict_edge_reward": 0.18750645220279694,
      "rewards-max/forensics_tiou_reward": 0.8822008669376373,
      "rewards/forensics_sinkhorn_reward": 0.6392422020435333,
      "rewards/forensics_strict_edge_reward": 0.022425557486712933,
      "rewards/forensics_tiou_reward": 0.40558284521102905,
      "step": 553
    },
    {
      "completion_length": 25.625,
      "epoch": 1.7004608294930876,
      "grad_norm": 8.265668869018555,
      "kl": 0.4443359375,
      "learning_rate": 5.746153846153847e-07,
      "loss": 0.0228,
      "reward": 0.7892541587352753,
      "reward_std": 0.02010409254580736,
      "reward_std_all": 0.02086698543280363,
      "rewards-max/forensics_sinkhorn_reward": 0.9492296874523163,
      "rewards-max/forensics_strict_edge_reward": 0.4746953248977661,
      "rewards-max/forensics_tiou_reward": 0.9075746238231659,
      "rewards/forensics_sinkhorn_reward": 0.7519005537033081,
      "rewards/forensics_strict_edge_reward": 0.05068868026137352,
      "rewards/forensics_tiou_reward": 0.49282221496105194,
      "step": 554
    },
    {
      "completion_length": 25.25,
      "epoch": 1.7035330261136714,
      "grad_norm": 7.930532455444336,
      "kl": 0.5625,
      "learning_rate": 5.738461538461539e-07,
      "loss": 0.213,
      "reward": 0.7249281406402588,
      "reward_std": 0.033709848299622536,
      "reward_std_all": 0.05395781248807907,
      "rewards-max/forensics_sinkhorn_reward": 0.9479553401470184,
      "rewards-max/forensics_strict_edge_reward": 0.3480508103966713,
      "rewards-max/forensics_tiou_reward": 0.8932521343231201,
      "rewards/forensics_sinkhorn_reward": 0.6469079256057739,
      "rewards/forensics_strict_edge_reward": 0.021213097730651498,
      "rewards/forensics_tiou_reward": 0.4243694245815277,
      "step": 555
    },
    {
      "completion_length": 25.546875,
      "epoch": 1.706605222734255,
      "grad_norm": 11.324052810668945,
      "kl": 0.439453125,
      "learning_rate": 5.73076923076923e-07,
      "loss": 0.2355,
      "reward": 0.7096608579158783,
      "reward_std": 0.06263907812535763,
      "reward_std_all": 0.07950440794229507,
      "rewards-max/forensics_sinkhorn_reward": 0.97834911942482,
      "rewards-max/forensics_strict_edge_reward": 0.09770023263990879,
      "rewards-max/forensics_tiou_reward": 0.973990261554718,
      "rewards/forensics_sinkhorn_reward": 0.5839092433452606,
      "rewards/forensics_strict_edge_reward": 0.00941101647913456,
      "rewards/forensics_tiou_reward": 0.3861891031265259,
      "step": 556
    },
    {
      "completion_length": 24.40625,
      "epoch": 1.7096774193548387,
      "grad_norm": 122.59892272949219,
      "kl": 4.0,
      "learning_rate": 5.723076923076923e-07,
      "loss": 0.1972,
      "reward": 0.7289917767047882,
      "reward_std": 0.027309422381222248,
      "reward_std_all": 0.05050148069858551,
      "rewards-max/forensics_sinkhorn_reward": 0.9715653657913208,
      "rewards-max/forensics_strict_edge_reward": 0.5650669932365417,
      "rewards-max/forensics_tiou_reward": 0.9553554952144623,
      "rewards/forensics_sinkhorn_reward": 0.6502136886119843,
      "rewards/forensics_strict_edge_reward": 0.05102528817951679,
      "rewards/forensics_tiou_reward": 0.4157259911298752,
      "step": 557
    },
    {
      "completion_length": 22.625,
      "epoch": 1.7127496159754223,
      "grad_norm": 10.121587753295898,
      "kl": 1.1640625,
      "learning_rate": 5.715384615384615e-07,
      "loss": 0.1402,
      "reward": 0.7398632764816284,
      "reward_std": 0.053222037851810455,
      "reward_std_all": 0.06005527637898922,
      "rewards-max/forensics_sinkhorn_reward": 0.9761276841163635,
      "rewards-max/forensics_strict_edge_reward": 0.6732408702373505,
      "rewards-max/forensics_tiou_reward": 0.9688803851604462,
      "rewards/forensics_sinkhorn_reward": 0.693794846534729,
      "rewards/forensics_strict_edge_reward": 0.0980965867638588,
      "rewards/forensics_tiou_reward": 0.4934665560722351,
      "step": 558
    },
    {
      "completion_length": 26.34375,
      "epoch": 1.715821812596006,
      "grad_norm": 9.513270378112793,
      "kl": 0.474609375,
      "learning_rate": 5.707692307692307e-07,
      "loss": -0.0837,
      "reward": 0.7526687979698181,
      "reward_std": 0.04757130518555641,
      "reward_std_all": 0.05041223205626011,
      "rewards-max/forensics_sinkhorn_reward": 0.9815301597118378,
      "rewards-max/forensics_strict_edge_reward": 0.5321668684482574,
      "rewards-max/forensics_tiou_reward": 0.9771071672439575,
      "rewards/forensics_sinkhorn_reward": 0.6468468904495239,
      "rewards/forensics_strict_edge_reward": 0.0637735053896904,
      "rewards/forensics_tiou_reward": 0.5058344602584839,
      "step": 559
    },
    {
      "completion_length": 27.1875,
      "epoch": 1.7188940092165899,
      "grad_norm": 15.317214965820312,
      "kl": 0.56640625,
      "learning_rate": 5.699999999999999e-07,
      "loss": 0.3844,
      "reward": 0.6969679594039917,
      "reward_std": 0.03297539800405502,
      "reward_std_all": 0.06717530637979507,
      "rewards-max/forensics_sinkhorn_reward": 0.9376865327358246,
      "rewards-max/forensics_strict_edge_reward": 0.4099828153848648,
      "rewards-max/forensics_tiou_reward": 0.8770986497402191,
      "rewards/forensics_sinkhorn_reward": 0.5718016624450684,
      "rewards/forensics_strict_edge_reward": 0.041709586046636105,
      "rewards/forensics_tiou_reward": 0.3328656256198883,
      "step": 560
    },
    {
      "completion_length": 23.921875,
      "epoch": 1.7219662058371736,
      "grad_norm": 8.618310928344727,
      "kl": 0.5244140625,
      "learning_rate": 5.692307692307692e-07,
      "loss": 0.3079,
      "reward": 0.7838656604290009,
      "reward_std": 0.021421998739242554,
      "reward_std_all": 0.02365281991660595,
      "rewards-max/forensics_sinkhorn_reward": 0.9784161448478699,
      "rewards-max/forensics_strict_edge_reward": 0.5062990635633469,
      "rewards-max/forensics_tiou_reward": 0.9726257920265198,
      "rewards/forensics_sinkhorn_reward": 0.6667899787425995,
      "rewards/forensics_strict_edge_reward": 0.0498965959995985,
      "rewards/forensics_tiou_reward": 0.4999515116214752,
      "step": 561
    },
    {
      "completion_length": 26.40625,
      "epoch": 1.7250384024577574,
      "grad_norm": 7.165635585784912,
      "kl": 0.4619140625,
      "learning_rate": 5.684615384615385e-07,
      "loss": 0.0779,
      "reward": 0.7610887885093689,
      "reward_std": 0.041797252371907234,
      "reward_std_all": 0.044077866710722446,
      "rewards-max/forensics_sinkhorn_reward": 0.9692765176296234,
      "rewards-max/forensics_strict_edge_reward": 0.332689493894577,
      "rewards-max/forensics_tiou_reward": 0.9500758051872253,
      "rewards/forensics_sinkhorn_reward": 0.6786957681179047,
      "rewards/forensics_strict_edge_reward": 0.02238440327346325,
      "rewards/forensics_tiou_reward": 0.4637027382850647,
      "step": 562
    },
    {
      "completion_length": 25.34375,
      "epoch": 1.728110599078341,
      "grad_norm": 8.167126655578613,
      "kl": 0.47265625,
      "learning_rate": 5.676923076923077e-07,
      "loss": 0.214,
      "reward": 0.7573545277118683,
      "reward_std": 0.04351396672427654,
      "reward_std_all": 0.04683132842183113,
      "rewards-max/forensics_sinkhorn_reward": 0.9379667639732361,
      "rewards-max/forensics_strict_edge_reward": 0.36069802939891815,
      "rewards-max/forensics_tiou_reward": 0.8739612996578217,
      "rewards/forensics_sinkhorn_reward": 0.6585976481437683,
      "rewards/forensics_strict_edge_reward": 0.02229217439889908,
      "rewards/forensics_tiou_reward": 0.41804203391075134,
      "step": 563
    },
    {
      "completion_length": 26.796875,
      "epoch": 1.7311827956989247,
      "grad_norm": 11.389451026916504,
      "kl": 0.408203125,
      "learning_rate": 5.669230769230769e-07,
      "loss": 0.2221,
      "reward": 0.7407947480678558,
      "reward_std": 0.03648890182375908,
      "reward_std_all": 0.0507020577788353,
      "rewards-max/forensics_sinkhorn_reward": 0.9534457921981812,
      "rewards-max/forensics_strict_edge_reward": 0.23711124621331692,
      "rewards-max/forensics_tiou_reward": 0.9119083285331726,
      "rewards/forensics_sinkhorn_reward": 0.6026850640773773,
      "rewards/forensics_strict_edge_reward": 0.03484715649392456,
      "rewards/forensics_tiou_reward": 0.4308522045612335,
      "step": 564
    },
    {
      "completion_length": 24.78125,
      "epoch": 1.7342549923195083,
      "grad_norm": 8.89052677154541,
      "kl": 0.4150390625,
      "learning_rate": 5.661538461538461e-07,
      "loss": 0.2234,
      "reward": 0.7489824891090393,
      "reward_std": 0.02932136319577694,
      "reward_std_all": 0.043486964888870716,
      "rewards-max/forensics_sinkhorn_reward": 0.9813655614852905,
      "rewards-max/forensics_strict_edge_reward": 0.7082725167274475,
      "rewards-max/forensics_tiou_reward": 0.9790439903736115,
      "rewards/forensics_sinkhorn_reward": 0.6674066185951233,
      "rewards/forensics_strict_edge_reward": 0.06685162708163261,
      "rewards/forensics_tiou_reward": 0.4737294167280197,
      "step": 565
    },
    {
      "completion_length": 24.515625,
      "epoch": 1.737327188940092,
      "grad_norm": 7.4597344398498535,
      "kl": 0.4990234375,
      "learning_rate": 5.653846153846153e-07,
      "loss": 0.2647,
      "reward": 0.714821845293045,
      "reward_std": 0.05609079636633396,
      "reward_std_all": 0.07265500165522099,
      "rewards-max/forensics_sinkhorn_reward": 0.9751328527927399,
      "rewards-max/forensics_strict_edge_reward": 0.2650461904704571,
      "rewards-max/forensics_tiou_reward": 0.9620298743247986,
      "rewards/forensics_sinkhorn_reward": 0.6498958468437195,
      "rewards/forensics_strict_edge_reward": 0.020758034894242883,
      "rewards/forensics_tiou_reward": 0.41428127884864807,
      "step": 566
    },
    {
      "completion_length": 22.40625,
      "epoch": 1.7403993855606759,
      "grad_norm": 8.696730613708496,
      "kl": 0.400390625,
      "learning_rate": 5.646153846153847e-07,
      "loss": 0.0538,
      "reward": 0.7484473884105682,
      "reward_std": 0.048766206949949265,
      "reward_std_all": 0.05143783800303936,
      "rewards-max/forensics_sinkhorn_reward": 0.9760004878044128,
      "rewards-max/forensics_strict_edge_reward": 0.5422053933143616,
      "rewards-max/forensics_tiou_reward": 0.9676810204982758,
      "rewards/forensics_sinkhorn_reward": 0.6997427344322205,
      "rewards/forensics_strict_edge_reward": 0.04875701479613781,
      "rewards/forensics_tiou_reward": 0.4788644015789032,
      "step": 567
    },
    {
      "completion_length": 24.125,
      "epoch": 1.7434715821812596,
      "grad_norm": 10.74374008178711,
      "kl": 0.4033203125,
      "learning_rate": 5.638461538461539e-07,
      "loss": 0.3311,
      "reward": 0.7612054347991943,
      "reward_std": 0.053426796570420265,
      "reward_std_all": 0.053454525768756866,
      "rewards-max/forensics_sinkhorn_reward": 0.9716354608535767,
      "rewards-max/forensics_strict_edge_reward": 0.13687143847346306,
      "rewards-max/forensics_tiou_reward": 0.9518543481826782,
      "rewards/forensics_sinkhorn_reward": 0.6701849400997162,
      "rewards/forensics_strict_edge_reward": 0.0211234581656754,
      "rewards/forensics_tiou_reward": 0.4541633874177933,
      "step": 568
    },
    {
      "completion_length": 28.546875,
      "epoch": 1.7465437788018434,
      "grad_norm": 9.299304962158203,
      "kl": 0.4521484375,
      "learning_rate": 5.63076923076923e-07,
      "loss": 0.2764,
      "reward": 0.694433867931366,
      "reward_std": 0.07102589681744576,
      "reward_std_all": 0.08987167850136757,
      "rewards-max/forensics_sinkhorn_reward": 0.9723978936672211,
      "rewards-max/forensics_strict_edge_reward": 0.28738873451948166,
      "rewards-max/forensics_tiou_reward": 0.9534787833690643,
      "rewards/forensics_sinkhorn_reward": 0.618224561214447,
      "rewards/forensics_strict_edge_reward": 0.017368401400744915,
      "rewards/forensics_tiou_reward": 0.41867929697036743,
      "step": 569
    },
    {
      "completion_length": 24.25,
      "epoch": 1.7496159754224272,
      "grad_norm": 9.841641426086426,
      "kl": 0.4306640625,
      "learning_rate": 5.623076923076923e-07,
      "loss": 0.1257,
      "reward": 0.7582314610481262,
      "reward_std": 0.03538531716912985,
      "reward_std_all": 0.04154381342232227,
      "rewards-max/forensics_sinkhorn_reward": 0.9823110997676849,
      "rewards-max/forensics_strict_edge_reward": 0.57895427942276,
      "rewards-max/forensics_tiou_reward": 0.9824756383895874,
      "rewards/forensics_sinkhorn_reward": 0.7118222415447235,
      "rewards/forensics_strict_edge_reward": 0.06453319080173969,
      "rewards/forensics_tiou_reward": 0.48403897881507874,
      "step": 570
    },
    {
      "completion_length": 24.671875,
      "epoch": 1.7526881720430108,
      "grad_norm": 8.545550346374512,
      "kl": 0.4228515625,
      "learning_rate": 5.615384615384615e-07,
      "loss": 0.2197,
      "reward": 0.7767648696899414,
      "reward_std": 0.009008655324578285,
      "reward_std_all": 0.019219462294131517,
      "rewards-max/forensics_sinkhorn_reward": 0.9791446328163147,
      "rewards-max/forensics_strict_edge_reward": 0.639672726392746,
      "rewards-max/forensics_tiou_reward": 0.9745802581310272,
      "rewards/forensics_sinkhorn_reward": 0.6707507967948914,
      "rewards/forensics_strict_edge_reward": 0.03234555013477802,
      "rewards/forensics_tiou_reward": 0.45302820205688477,
      "step": 571
    },
    {
      "completion_length": 27.390625,
      "epoch": 1.7557603686635943,
      "grad_norm": 7.168601989746094,
      "kl": 0.4619140625,
      "learning_rate": 5.607692307692308e-07,
      "loss": 0.3357,
      "reward": 0.658898651599884,
      "reward_std": 0.06776893883943558,
      "reward_std_all": 0.10405494272708893,
      "rewards-max/forensics_sinkhorn_reward": 0.9714383184909821,
      "rewards-max/forensics_strict_edge_reward": 0.4042638689279556,
      "rewards-max/forensics_tiou_reward": 0.9553734958171844,
      "rewards/forensics_sinkhorn_reward": 0.5286359488964081,
      "rewards/forensics_strict_edge_reward": 0.04340531677007675,
      "rewards/forensics_tiou_reward": 0.3803142160177231,
      "step": 572
    },
    {
      "completion_length": 27.078125,
      "epoch": 1.758832565284178,
      "grad_norm": 7.548270225524902,
      "kl": 0.720703125,
      "learning_rate": 5.6e-07,
      "loss": 0.3346,
      "reward": 0.7178652584552765,
      "reward_std": 0.051609646528959274,
      "reward_std_all": 0.0694572925567627,
      "rewards-max/forensics_sinkhorn_reward": 0.9748704135417938,
      "rewards-max/forensics_strict_edge_reward": 0.51341712474823,
      "rewards-max/forensics_tiou_reward": 0.964354932308197,
      "rewards/forensics_sinkhorn_reward": 0.6440240740776062,
      "rewards/forensics_strict_edge_reward": 0.045622263103723526,
      "rewards/forensics_tiou_reward": 0.4470492750406265,
      "step": 573
    },
    {
      "completion_length": 26.109375,
      "epoch": 1.7619047619047619,
      "grad_norm": 8.947154998779297,
      "kl": 0.49609375,
      "learning_rate": 5.592307692307691e-07,
      "loss": 0.0522,
      "reward": 0.7330451011657715,
      "reward_std": 0.03781027439981699,
      "reward_std_all": 0.05249655619263649,
      "rewards-max/forensics_sinkhorn_reward": 0.977922260761261,
      "rewards-max/forensics_strict_edge_reward": 0.5205644369125366,
      "rewards-max/forensics_tiou_reward": 0.9700529277324677,
      "rewards/forensics_sinkhorn_reward": 0.6496995389461517,
      "rewards/forensics_strict_edge_reward": 0.05716056935489178,
      "rewards/forensics_tiou_reward": 0.4705338776111603,
      "step": 574
    },
    {
      "completion_length": 26.65625,
      "epoch": 1.7649769585253456,
      "grad_norm": 7.4464430809021,
      "kl": 0.509765625,
      "learning_rate": 5.584615384615385e-07,
      "loss": 0.1905,
      "reward": 0.7734163105487823,
      "reward_std": 0.026160457637161016,
      "reward_std_all": 0.03197260852903128,
      "rewards-max/forensics_sinkhorn_reward": 0.9220143258571625,
      "rewards-max/forensics_strict_edge_reward": 0.23538962751626968,
      "rewards-max/forensics_tiou_reward": 0.8343152701854706,
      "rewards/forensics_sinkhorn_reward": 0.6618694365024567,
      "rewards/forensics_strict_edge_reward": 0.021075021475553513,
      "rewards/forensics_tiou_reward": 0.45312874019145966,
      "step": 575
    },
    {
      "completion_length": 25.21875,
      "epoch": 1.7680491551459294,
      "grad_norm": 10.264065742492676,
      "kl": 0.525390625,
      "learning_rate": 5.576923076923077e-07,
      "loss": 0.3007,
      "reward": 0.7108580470085144,
      "reward_std": 0.07462415471673012,
      "reward_std_all": 0.08264517784118652,
      "rewards-max/forensics_sinkhorn_reward": 0.9672077596187592,
      "rewards-max/forensics_strict_edge_reward": 0.32832057029008865,
      "rewards-max/forensics_tiou_reward": 0.937745064496994,
      "rewards/forensics_sinkhorn_reward": 0.6150048971176147,
      "rewards/forensics_strict_edge_reward": 0.023842911701649427,
      "rewards/forensics_tiou_reward": 0.4305282533168793,
      "step": 576
    },
    {
      "completion_length": 26.078125,
      "epoch": 1.7711213517665132,
      "grad_norm": 12.641167640686035,
      "kl": 0.388671875,
      "learning_rate": 5.569230769230769e-07,
      "loss": 0.3683,
      "reward": 0.6578471660614014,
      "reward_std": 0.06925978511571884,
      "reward_std_all": 0.10408858209848404,
      "rewards-max/forensics_sinkhorn_reward": 0.98061403632164,
      "rewards-max/forensics_strict_edge_reward": 0.5770299434661865,
      "rewards-max/forensics_tiou_reward": 0.9789745211601257,
      "rewards/forensics_sinkhorn_reward": 0.5621135830879211,
      "rewards/forensics_strict_edge_reward": 0.05737586319446564,
      "rewards/forensics_tiou_reward": 0.3748790770769119,
      "step": 577
    },
    {
      "completion_length": 25.9375,
      "epoch": 1.7741935483870968,
      "grad_norm": 9.278329849243164,
      "kl": 0.4345703125,
      "learning_rate": 5.561538461538461e-07,
      "loss": 0.1772,
      "reward": 0.7426457405090332,
      "reward_std": 0.032275582663714886,
      "reward_std_all": 0.04607158061116934,
      "rewards-max/forensics_sinkhorn_reward": 0.9687567949295044,
      "rewards-max/forensics_strict_edge_reward": 0.44491301476955414,
      "rewards-max/forensics_tiou_reward": 0.9484756588935852,
      "rewards/forensics_sinkhorn_reward": 0.6570367813110352,
      "rewards/forensics_strict_edge_reward": 0.04630611278116703,
      "rewards/forensics_tiou_reward": 0.4355829209089279,
      "step": 578
    },
    {
      "completion_length": 24.53125,
      "epoch": 1.7772657450076805,
      "grad_norm": 9.360533714294434,
      "kl": 0.6064453125,
      "learning_rate": 5.553846153846153e-07,
      "loss": 0.3199,
      "reward": 0.7152424454689026,
      "reward_std": 0.05632364749908447,
      "reward_std_all": 0.06952846981585026,
      "rewards-max/forensics_sinkhorn_reward": 0.9647625684738159,
      "rewards-max/forensics_strict_edge_reward": 0.36884360760450363,
      "rewards-max/forensics_tiou_reward": 0.9301083981990814,
      "rewards/forensics_sinkhorn_reward": 0.5917242765426636,
      "rewards/forensics_strict_edge_reward": 0.036752935498952866,
      "rewards/forensics_tiou_reward": 0.3844972252845764,
      "step": 579
    },
    {
      "completion_length": 27.796875,
      "epoch": 1.780337941628264,
      "grad_norm": 6.249568462371826,
      "kl": 0.513671875,
      "learning_rate": 5.546153846153846e-07,
      "loss": 0.1786,
      "reward": 0.7095991671085358,
      "reward_std": 0.06786846742033958,
      "reward_std_all": 0.07916662469506264,
      "rewards-max/forensics_sinkhorn_reward": 0.9855956435203552,
      "rewards-max/forensics_strict_edge_reward": 0.6892678737640381,
      "rewards-max/forensics_tiou_reward": 0.9902778267860413,
      "rewards/forensics_sinkhorn_reward": 0.5739737749099731,
      "rewards/forensics_strict_edge_reward": 0.0990525558590889,
      "rewards/forensics_tiou_reward": 0.42145003378391266,
      "step": 580
    },
    {
      "completion_length": 22.4375,
      "epoch": 1.7834101382488479,
      "grad_norm": 12.369274139404297,
      "kl": 1.0703125,
      "learning_rate": 5.538461538461539e-07,
      "loss": 0.3572,
      "reward": 0.760563313961029,
      "reward_std": 0.039452371187508106,
      "reward_std_all": 0.04224035330116749,
      "rewards-max/forensics_sinkhorn_reward": 0.9847496449947357,
      "rewards-max/forensics_strict_edge_reward": 0.6071805953979492,
      "rewards-max/forensics_tiou_reward": 0.9872004985809326,
      "rewards/forensics_sinkhorn_reward": 0.684243381023407,
      "rewards/forensics_strict_edge_reward": 0.06932093761861324,
      "rewards/forensics_tiou_reward": 0.5112937688827515,
      "step": 581
    },
    {
      "completion_length": 25.875,
      "epoch": 1.7864823348694316,
      "grad_norm": 9.307243347167969,
      "kl": 0.5908203125,
      "learning_rate": 5.53076923076923e-07,
      "loss": 0.1076,
      "reward": 0.745855063199997,
      "reward_std": 0.059245532378554344,
      "reward_std_all": 0.062073685228824615,
      "rewards-max/forensics_sinkhorn_reward": 0.9525032341480255,
      "rewards-max/forensics_strict_edge_reward": 0.27902691066265106,
      "rewards-max/forensics_tiou_reward": 0.9046005010604858,
      "rewards/forensics_sinkhorn_reward": 0.6889981627464294,
      "rewards/forensics_strict_edge_reward": 0.031208294443786144,
      "rewards/forensics_tiou_reward": 0.4618871957063675,
      "step": 582
    },
    {
      "completion_length": 24.359375,
      "epoch": 1.7895545314900154,
      "grad_norm": 11.638238906860352,
      "kl": 0.64453125,
      "learning_rate": 5.523076923076923e-07,
      "loss": 0.0504,
      "reward": 0.7518846094608307,
      "reward_std": 0.049488335847854614,
      "reward_std_all": 0.05370000749826431,
      "rewards-max/forensics_sinkhorn_reward": 0.9463345408439636,
      "rewards-max/forensics_strict_edge_reward": 0.23662097565829754,
      "rewards-max/forensics_tiou_reward": 0.8857398629188538,
      "rewards/forensics_sinkhorn_reward": 0.6450121998786926,
      "rewards/forensics_strict_edge_reward": 0.010352708748541772,
      "rewards/forensics_tiou_reward": 0.3911304473876953,
      "step": 583
    },
    {
      "completion_length": 27.703125,
      "epoch": 1.7926267281105992,
      "grad_norm": 7.026988506317139,
      "kl": 0.453125,
      "learning_rate": 5.515384615384615e-07,
      "loss": 0.261,
      "reward": 0.7441991567611694,
      "reward_std": 0.06459135003387928,
      "reward_std_all": 0.06775850057601929,
      "rewards-max/forensics_sinkhorn_reward": 0.9795875549316406,
      "rewards-max/forensics_strict_edge_reward": 0.4533289223909378,
      "rewards-max/forensics_tiou_reward": 0.9732473492622375,
      "rewards/forensics_sinkhorn_reward": 0.6819094121456146,
      "rewards/forensics_strict_edge_reward": 0.04842933639883995,
      "rewards/forensics_tiou_reward": 0.5377589762210846,
      "step": 584
    },
    {
      "completion_length": 24.140625,
      "epoch": 1.7956989247311828,
      "grad_norm": 7.201074600219727,
      "kl": 0.4072265625,
      "learning_rate": 5.507692307692308e-07,
      "loss": 0.1436,
      "reward": 0.7215280830860138,
      "reward_std": 0.03750233165919781,
      "reward_std_all": 0.06094178184866905,
      "rewards-max/forensics_sinkhorn_reward": 0.9683228135108948,
      "rewards-max/forensics_strict_edge_reward": 0.285599522292614,
      "rewards-max/forensics_tiou_reward": 0.94507697224617,
      "rewards/forensics_sinkhorn_reward": 0.6326992511749268,
      "rewards/forensics_strict_edge_reward": 0.022019917611032724,
      "rewards/forensics_tiou_reward": 0.4024640619754791,
      "step": 585
    },
    {
      "completion_length": 25.25,
      "epoch": 1.7987711213517665,
      "grad_norm": 9.77525520324707,
      "kl": 0.525390625,
      "learning_rate": 5.5e-07,
      "loss": 0.0993,
      "reward": 0.7036147117614746,
      "reward_std": 0.1012013740837574,
      "reward_std_all": 0.09978343173861504,
      "rewards-max/forensics_sinkhorn_reward": 0.973596841096878,
      "rewards-max/forensics_strict_edge_reward": 0.5147627294063568,
      "rewards-max/forensics_tiou_reward": 0.9526476860046387,
      "rewards/forensics_sinkhorn_reward": 0.6170173287391663,
      "rewards/forensics_strict_edge_reward": 0.03712902031838894,
      "rewards/forensics_tiou_reward": 0.42915569245815277,
      "step": 586
    },
    {
      "completion_length": 22.90625,
      "epoch": 1.80184331797235,
      "grad_norm": 10.812640190124512,
      "kl": 0.4755859375,
      "learning_rate": 5.492307692307691e-07,
      "loss": -0.1875,
      "reward": 0.7526851296424866,
      "reward_std": 0.026690112426877022,
      "reward_std_all": 0.04114456661045551,
      "rewards-max/forensics_sinkhorn_reward": 0.9642491638660431,
      "rewards-max/forensics_strict_edge_reward": 0.5062626451253891,
      "rewards-max/forensics_tiou_reward": 0.935581237077713,
      "rewards/forensics_sinkhorn_reward": 0.6810154914855957,
      "rewards/forensics_strict_edge_reward": 0.04865542519837618,
      "rewards/forensics_tiou_reward": 0.47692443430423737,
      "step": 587
    },
    {
      "completion_length": 28.796875,
      "epoch": 1.8049155145929339,
      "grad_norm": 8.149312973022461,
      "kl": 0.33203125,
      "learning_rate": 5.484615384615385e-07,
      "loss": 0.3306,
      "reward": 0.7605228424072266,
      "reward_std": 0.05183058977127075,
      "reward_std_all": 0.050744397565722466,
      "rewards-max/forensics_sinkhorn_reward": 0.9687153995037079,
      "rewards-max/forensics_strict_edge_reward": 0.5854147225618362,
      "rewards-max/forensics_tiou_reward": 0.9529967308044434,
      "rewards/forensics_sinkhorn_reward": 0.6538789570331573,
      "rewards/forensics_strict_edge_reward": 0.033443390391767025,
      "rewards/forensics_tiou_reward": 0.49760353565216064,
      "step": 588
    },
    {
      "completion_length": 26.140625,
      "epoch": 1.8079877112135176,
      "grad_norm": 13.18924617767334,
      "kl": 0.431640625,
      "learning_rate": 5.476923076923077e-07,
      "loss": 0.2173,
      "reward": 0.7218535244464874,
      "reward_std": 0.034845998510718346,
      "reward_std_all": 0.060765933245420456,
      "rewards-max/forensics_sinkhorn_reward": 0.9674052894115448,
      "rewards-max/forensics_strict_edge_reward": 0.3001856058835983,
      "rewards-max/forensics_tiou_reward": 0.945216029882431,
      "rewards/forensics_sinkhorn_reward": 0.6404089033603668,
      "rewards/forensics_strict_edge_reward": 0.03561961464583874,
      "rewards/forensics_tiou_reward": 0.4363618493080139,
      "step": 589
    },
    {
      "completion_length": 26.75,
      "epoch": 1.8110599078341014,
      "grad_norm": 8.313339233398438,
      "kl": 0.42578125,
      "learning_rate": 5.469230769230769e-07,
      "loss": 0.0721,
      "reward": 0.7568722665309906,
      "reward_std": 0.027132540941238403,
      "reward_std_all": 0.03684343956410885,
      "rewards-max/forensics_sinkhorn_reward": 0.9715560376644135,
      "rewards-max/forensics_strict_edge_reward": 0.39376434683799744,
      "rewards-max/forensics_tiou_reward": 0.955272376537323,
      "rewards/forensics_sinkhorn_reward": 0.6622994542121887,
      "rewards/forensics_strict_edge_reward": 0.027680289931595325,
      "rewards/forensics_tiou_reward": 0.4988171011209488,
      "step": 590
    },
    {
      "completion_length": 25.890625,
      "epoch": 1.8141321044546852,
      "grad_norm": 8.786826133728027,
      "kl": 0.4560546875,
      "learning_rate": 5.461538461538461e-07,
      "loss": -0.0536,
      "reward": 0.7482689917087555,
      "reward_std": 0.04663458839058876,
      "reward_std_all": 0.051839714869856834,
      "rewards-max/forensics_sinkhorn_reward": 0.956380695104599,
      "rewards-max/forensics_strict_edge_reward": 0.3015466257929802,
      "rewards-max/forensics_tiou_reward": 0.9179542064666748,
      "rewards/forensics_sinkhorn_reward": 0.6560167074203491,
      "rewards/forensics_strict_edge_reward": 0.028699652291834354,
      "rewards/forensics_tiou_reward": 0.49806544184684753,
      "step": 591
    },
    {
      "completion_length": 27.34375,
      "epoch": 1.817204301075269,
      "grad_norm": 12.948328018188477,
      "kl": 0.4755859375,
      "learning_rate": 5.453846153846153e-07,
      "loss": 0.1205,
      "reward": 0.7531337141990662,
      "reward_std": 0.06212018430233002,
      "reward_std_all": 0.06137232482433319,
      "rewards-max/forensics_sinkhorn_reward": 0.9847646355628967,
      "rewards-max/forensics_strict_edge_reward": 0.6365986764431,
      "rewards-max/forensics_tiou_reward": 0.9876275062561035,
      "rewards/forensics_sinkhorn_reward": 0.6252189576625824,
      "rewards/forensics_strict_edge_reward": 0.0607891920953989,
      "rewards/forensics_tiou_reward": 0.491731196641922,
      "step": 592
    },
    {
      "completion_length": 27.875,
      "epoch": 1.8202764976958525,
      "grad_norm": 8.916441917419434,
      "kl": 0.5078125,
      "learning_rate": 5.446153846153846e-07,
      "loss": -0.035,
      "reward": 0.7660504281520844,
      "reward_std": 0.044451446272432804,
      "reward_std_all": 0.04405864514410496,
      "rewards-max/forensics_sinkhorn_reward": 0.9642497897148132,
      "rewards-max/forensics_strict_edge_reward": 0.5030358955264091,
      "rewards-max/forensics_tiou_reward": 0.9349642693996429,
      "rewards/forensics_sinkhorn_reward": 0.6447291076183319,
      "rewards/forensics_strict_edge_reward": 0.05822915397584438,
      "rewards/forensics_tiou_reward": 0.49739648401737213,
      "step": 593
    },
    {
      "completion_length": 24.109375,
      "epoch": 1.823348694316436,
      "grad_norm": 9.37510871887207,
      "kl": 0.51953125,
      "learning_rate": 5.438461538461538e-07,
      "loss": 0.0131,
      "reward": 0.7463989853858948,
      "reward_std": 0.034657303243875504,
      "reward_std_all": 0.04597601667046547,
      "rewards-max/forensics_sinkhorn_reward": 0.9766533970832825,
      "rewards-max/forensics_strict_edge_reward": 0.48061278462409973,
      "rewards-max/forensics_tiou_reward": 0.9652641415596008,
      "rewards/forensics_sinkhorn_reward": 0.7005332112312317,
      "rewards/forensics_strict_edge_reward": 0.08598291501402855,
      "rewards/forensics_tiou_reward": 0.5193963944911957,
      "step": 594
    },
    {
      "completion_length": 25.84375,
      "epoch": 1.8264208909370199,
      "grad_norm": 9.511271476745605,
      "kl": 0.34375,
      "learning_rate": 5.430769230769231e-07,
      "loss": 0.0864,
      "reward": 0.7623125910758972,
      "reward_std": 0.040303305722773075,
      "reward_std_all": 0.04269589111208916,
      "rewards-max/forensics_sinkhorn_reward": 0.9590875208377838,
      "rewards-max/forensics_strict_edge_reward": 0.3005751669406891,
      "rewards-max/forensics_tiou_reward": 0.917805403470993,
      "rewards/forensics_sinkhorn_reward": 0.6815930306911469,
      "rewards/forensics_strict_edge_reward": 0.03380454517900944,
      "rewards/forensics_tiou_reward": 0.4907533824443817,
      "step": 595
    },
    {
      "completion_length": 22.453125,
      "epoch": 1.8294930875576036,
      "grad_norm": 13.135984420776367,
      "kl": 0.494140625,
      "learning_rate": 5.423076923076923e-07,
      "loss": 0.1255,
      "reward": 0.6962616145610809,
      "reward_std": 0.08965543657541275,
      "reward_std_all": 0.0999438427388668,
      "rewards-max/forensics_sinkhorn_reward": 0.9774373471736908,
      "rewards-max/forensics_strict_edge_reward": 0.756708562374115,
      "rewards-max/forensics_tiou_reward": 0.9742472469806671,
      "rewards/forensics_sinkhorn_reward": 0.6143309772014618,
      "rewards/forensics_strict_edge_reward": 0.07651368528604507,
      "rewards/forensics_tiou_reward": 0.42365722358226776,
      "step": 596
    },
    {
      "completion_length": 22.921875,
      "epoch": 1.8325652841781874,
      "grad_norm": 9.20547103881836,
      "kl": 0.40234375,
      "learning_rate": 5.415384615384615e-07,
      "loss": 0.2471,
      "reward": 0.7194787859916687,
      "reward_std": 0.10745706781744957,
      "reward_std_all": 0.10040181875228882,
      "rewards-max/forensics_sinkhorn_reward": 0.9855837821960449,
      "rewards-max/forensics_strict_edge_reward": 0.765928328037262,
      "rewards-max/forensics_tiou_reward": 0.988962709903717,
      "rewards/forensics_sinkhorn_reward": 0.6702369749546051,
      "rewards/forensics_strict_edge_reward": 0.07882682234048843,
      "rewards/forensics_tiou_reward": 0.5051085352897644,
      "step": 597
    },
    {
      "completion_length": 22.59375,
      "epoch": 1.8356374807987712,
      "grad_norm": 9.64263916015625,
      "kl": 0.453125,
      "learning_rate": 5.407692307692308e-07,
      "loss": 0.2695,
      "reward": 0.7348185777664185,
      "reward_std": 0.028542807325720787,
      "reward_std_all": 0.04967646859586239,
      "rewards-max/forensics_sinkhorn_reward": 0.9620010256767273,
      "rewards-max/forensics_strict_edge_reward": 0.3298750054091215,
      "rewards-max/forensics_tiou_reward": 0.9248538613319397,
      "rewards/forensics_sinkhorn_reward": 0.6722593605518341,
      "rewards/forensics_strict_edge_reward": 0.029992697527632117,
      "rewards/forensics_tiou_reward": 0.4964190721511841,
      "step": 598
    },
    {
      "completion_length": 25.578125,
      "epoch": 1.838709677419355,
      "grad_norm": 9.483880996704102,
      "kl": 0.869140625,
      "learning_rate": 5.4e-07,
      "loss": 0.0559,
      "reward": 0.6891281008720398,
      "reward_std": 0.0809406004846096,
      "reward_std_all": 0.09591235965490341,
      "rewards-max/forensics_sinkhorn_reward": 0.9871760308742523,
      "rewards-max/forensics_strict_edge_reward": 0.6858793199062347,
      "rewards-max/forensics_tiou_reward": 0.9943775236606598,
      "rewards/forensics_sinkhorn_reward": 0.5669283270835876,
      "rewards/forensics_strict_edge_reward": 0.045351725071668625,
      "rewards/forensics_tiou_reward": 0.35485953092575073,
      "step": 599
    },
    {
      "completion_length": 24.625,
      "epoch": 1.8417818740399385,
      "grad_norm": 10.189544677734375,
      "kl": 0.4482421875,
      "learning_rate": 5.392307692307691e-07,
      "loss": 0.0215,
      "reward": 0.7692737281322479,
      "reward_std": 0.042092589661478996,
      "reward_std_all": 0.04240104928612709,
      "rewards-max/forensics_sinkhorn_reward": 0.9787796139717102,
      "rewards-max/forensics_strict_edge_reward": 0.49048639088869095,
      "rewards-max/forensics_tiou_reward": 0.9739380776882172,
      "rewards/forensics_sinkhorn_reward": 0.7253825068473816,
      "rewards/forensics_strict_edge_reward": 0.05458778887987137,
      "rewards/forensics_tiou_reward": 0.5523480176925659,
      "step": 600
    },
    {
      "completion_length": 22.40625,
      "epoch": 1.8448540706605223,
      "grad_norm": 10.878210067749023,
      "kl": 0.4140625,
      "learning_rate": 5.384615384615384e-07,
      "loss": 0.0947,
      "reward": 0.7565539181232452,
      "reward_std": 0.026245756074786186,
      "reward_std_all": 0.038566749542951584,
      "rewards-max/forensics_sinkhorn_reward": 0.9652184844017029,
      "rewards-max/forensics_strict_edge_reward": 0.5832466036081314,
      "rewards-max/forensics_tiou_reward": 0.9304872751235962,
      "rewards/forensics_sinkhorn_reward": 0.6887872219085693,
      "rewards/forensics_strict_edge_reward": 0.07082131505012512,
      "rewards/forensics_tiou_reward": 0.46794408559799194,
      "step": 601
    },
    {
      "completion_length": 22.9375,
      "epoch": 1.8479262672811059,
      "grad_norm": 7.9968695640563965,
      "kl": 0.4814453125,
      "learning_rate": 5.376923076923077e-07,
      "loss": 0.2677,
      "reward": 0.7569900453090668,
      "reward_std": 0.03903952240943909,
      "reward_std_all": 0.04392857663333416,
      "rewards-max/forensics_sinkhorn_reward": 0.9789445102214813,
      "rewards-max/forensics_strict_edge_reward": 0.6736744344234467,
      "rewards-max/forensics_tiou_reward": 0.9751909375190735,
      "rewards/forensics_sinkhorn_reward": 0.7020451426506042,
      "rewards/forensics_strict_edge_reward": 0.0473616486415267,
      "rewards/forensics_tiou_reward": 0.45062026381492615,
      "step": 602
    },
    {
      "completion_length": 24.25,
      "epoch": 1.8509984639016897,
      "grad_norm": 7.0674848556518555,
      "kl": 0.466796875,
      "learning_rate": 5.369230769230769e-07,
      "loss": 0.0855,
      "reward": 0.7733884453773499,
      "reward_std": 0.027898642234504223,
      "reward_std_all": 0.032554104924201965,
      "rewards-max/forensics_sinkhorn_reward": 0.9789683818817139,
      "rewards-max/forensics_strict_edge_reward": 0.6419725120067596,
      "rewards-max/forensics_tiou_reward": 0.9740356802940369,
      "rewards/forensics_sinkhorn_reward": 0.6787348687648773,
      "rewards/forensics_strict_edge_reward": 0.09934039413928986,
      "rewards/forensics_tiou_reward": 0.48007863759994507,
      "step": 603
    },
    {
      "completion_length": 23.453125,
      "epoch": 1.8540706605222734,
      "grad_norm": 10.017725944519043,
      "kl": 0.328125,
      "learning_rate": 5.361538461538462e-07,
      "loss": 0.0112,
      "reward": 0.7631419599056244,
      "reward_std": 0.01563355326652527,
      "reward_std_all": 0.03163339011371136,
      "rewards-max/forensics_sinkhorn_reward": 0.948340654373169,
      "rewards-max/forensics_strict_edge_reward": 0.3362295627593994,
      "rewards-max/forensics_tiou_reward": 0.8966287672519684,
      "rewards/forensics_sinkhorn_reward": 0.7342836558818817,
      "rewards/forensics_strict_edge_reward": 0.034465583972632885,
      "rewards/forensics_tiou_reward": 0.518687054514885,
      "step": 604
    },
    {
      "completion_length": 22.5625,
      "epoch": 1.8571428571428572,
      "grad_norm": 9.106054306030273,
      "kl": 0.52734375,
      "learning_rate": 5.353846153846153e-07,
      "loss": 0.0799,
      "reward": 0.7831277549266815,
      "reward_std": 0.01358474139124155,
      "reward_std_all": 0.019471215084195137,
      "rewards-max/forensics_sinkhorn_reward": 0.9652921557426453,
      "rewards-max/forensics_strict_edge_reward": 0.411786213517189,
      "rewards-max/forensics_tiou_reward": 0.9377665519714355,
      "rewards/forensics_sinkhorn_reward": 0.7327864170074463,
      "rewards/forensics_strict_edge_reward": 0.05005046911537647,
      "rewards/forensics_tiou_reward": 0.5633810758590698,
      "step": 605
    },
    {
      "completion_length": 25.046875,
      "epoch": 1.860215053763441,
      "grad_norm": 8.772161483764648,
      "kl": 0.5556640625,
      "learning_rate": 5.346153846153846e-07,
      "loss": 0.2766,
      "reward": 0.6515107750892639,
      "reward_std": 0.06357370968908072,
      "reward_std_all": 0.10470832511782646,
      "rewards-max/forensics_sinkhorn_reward": 0.9647443890571594,
      "rewards-max/forensics_strict_edge_reward": 0.26573626697063446,
      "rewards-max/forensics_tiou_reward": 0.9334564208984375,
      "rewards/forensics_sinkhorn_reward": 0.5771284699440002,
      "rewards/forensics_strict_edge_reward": 0.022141636814922094,
      "rewards/forensics_tiou_reward": 0.3599821925163269,
      "step": 606
    },
    {
      "completion_length": 24.765625,
      "epoch": 1.8632872503840245,
      "grad_norm": 6.580840587615967,
      "kl": 0.4443359375,
      "learning_rate": 5.338461538461538e-07,
      "loss": 0.2317,
      "reward": 0.75871741771698,
      "reward_std": 0.04780454281717539,
      "reward_std_all": 0.049407076090574265,
      "rewards-max/forensics_sinkhorn_reward": 0.9753386378288269,
      "rewards-max/forensics_strict_edge_reward": 0.23922576755285263,
      "rewards-max/forensics_tiou_reward": 0.9588203132152557,
      "rewards/forensics_sinkhorn_reward": 0.7206855118274689,
      "rewards/forensics_strict_edge_reward": 0.03491044603288174,
      "rewards/forensics_tiou_reward": 0.5259638726711273,
      "step": 607
    },
    {
      "completion_length": 26.75,
      "epoch": 1.8663594470046083,
      "grad_norm": 10.930176734924316,
      "kl": 0.53125,
      "learning_rate": 5.33076923076923e-07,
      "loss": 0.0284,
      "reward": 0.7480847835540771,
      "reward_std": 0.025611200369894505,
      "reward_std_all": 0.03963985852897167,
      "rewards-max/forensics_sinkhorn_reward": 0.9814234375953674,
      "rewards-max/forensics_strict_edge_reward": 0.7237150073051453,
      "rewards-max/forensics_tiou_reward": 0.9805614352226257,
      "rewards/forensics_sinkhorn_reward": 0.6852594614028931,
      "rewards/forensics_strict_edge_reward": 0.055494705215096474,
      "rewards/forensics_tiou_reward": 0.5158688724040985,
      "step": 608
    },
    {
      "completion_length": 27.546875,
      "epoch": 1.8694316436251919,
      "grad_norm": 8.746185302734375,
      "kl": 0.375,
      "learning_rate": 5.323076923076923e-07,
      "loss": -0.0574,
      "reward": 0.7549104988574982,
      "reward_std": 0.04456157237291336,
      "reward_std_all": 0.04756338708102703,
      "rewards-max/forensics_sinkhorn_reward": 0.9761228859424591,
      "rewards-max/forensics_strict_edge_reward": 0.3886224180459976,
      "rewards-max/forensics_tiou_reward": 0.9697935581207275,
      "rewards/forensics_sinkhorn_reward": 0.6514598727226257,
      "rewards/forensics_strict_edge_reward": 0.035949286073446274,
      "rewards/forensics_tiou_reward": 0.49143871665000916,
      "step": 609
    },
    {
      "completion_length": 25.609375,
      "epoch": 1.8725038402457757,
      "grad_norm": 8.30842113494873,
      "kl": 0.4111328125,
      "learning_rate": 5.315384615384615e-07,
      "loss": 0.1218,
      "reward": 0.7479100823402405,
      "reward_std": 0.04443752043880522,
      "reward_std_all": 0.05109075969085097,
      "rewards-max/forensics_sinkhorn_reward": 0.9845863878726959,
      "rewards-max/forensics_strict_edge_reward": 0.7467312514781952,
      "rewards-max/forensics_tiou_reward": 0.9881643354892731,
      "rewards/forensics_sinkhorn_reward": 0.6707657277584076,
      "rewards/forensics_strict_edge_reward": 0.06344163045287132,
      "rewards/forensics_tiou_reward": 0.4910443127155304,
      "step": 610
    },
    {
      "completion_length": 20.46875,
      "epoch": 1.8755760368663594,
      "grad_norm": 21.443389892578125,
      "kl": 1.0654296875,
      "learning_rate": 5.307692307692308e-07,
      "loss": -0.0117,
      "reward": 0.7569704055786133,
      "reward_std": 0.056542620062828064,
      "reward_std_all": 0.0552595891058445,
      "rewards-max/forensics_sinkhorn_reward": 0.9778523743152618,
      "rewards-max/forensics_strict_edge_reward": 0.7345792949199677,
      "rewards-max/forensics_tiou_reward": 0.969251811504364,
      "rewards/forensics_sinkhorn_reward": 0.7466896176338196,
      "rewards/forensics_strict_edge_reward": 0.09826207906007767,
      "rewards/forensics_tiou_reward": 0.5173511505126953,
      "step": 611
    },
    {
      "completion_length": 22.453125,
      "epoch": 1.8786482334869432,
      "grad_norm": 11.032743453979492,
      "kl": 0.494140625,
      "learning_rate": 5.3e-07,
      "loss": -0.0533,
      "reward": 0.7142145931720734,
      "reward_std": 0.0436019841581583,
      "reward_std_all": 0.06443290412425995,
      "rewards-max/forensics_sinkhorn_reward": 0.9379273355007172,
      "rewards-max/forensics_strict_edge_reward": 0.2088332585990429,
      "rewards-max/forensics_tiou_reward": 0.8755508959293365,
      "rewards/forensics_sinkhorn_reward": 0.6135198771953583,
      "rewards/forensics_strict_edge_reward": 0.015109932282939553,
      "rewards/forensics_tiou_reward": 0.4286445826292038,
      "step": 612
    },
    {
      "completion_length": 24.28125,
      "epoch": 1.881720430107527,
      "grad_norm": 9.099703788757324,
      "kl": 0.5380859375,
      "learning_rate": 5.292307692307692e-07,
      "loss": 0.1796,
      "reward": 0.7695170342922211,
      "reward_std": 0.03582999063655734,
      "reward_std_all": 0.039299788884818554,
      "rewards-max/forensics_sinkhorn_reward": 0.9706440269947052,
      "rewards-max/forensics_strict_edge_reward": 0.6376232653856277,
      "rewards-max/forensics_tiou_reward": 0.9461857676506042,
      "rewards/forensics_sinkhorn_reward": 0.6951867043972015,
      "rewards/forensics_strict_edge_reward": 0.05185465328395367,
      "rewards/forensics_tiou_reward": 0.4973922669887543,
      "step": 613
    },
    {
      "completion_length": 24.328125,
      "epoch": 1.8847926267281108,
      "grad_norm": 11.760027885437012,
      "kl": 0.697265625,
      "learning_rate": 5.284615384615384e-07,
      "loss": 0.207,
      "reward": 0.7498385310173035,
      "reward_std": 0.04675361420959234,
      "reward_std_all": 0.05163033585995436,
      "rewards-max/forensics_sinkhorn_reward": 0.980927973985672,
      "rewards-max/forensics_strict_edge_reward": 0.38209572434425354,
      "rewards-max/forensics_tiou_reward": 0.9775249660015106,
      "rewards/forensics_sinkhorn_reward": 0.6742313802242279,
      "rewards/forensics_strict_edge_reward": 0.03758633974939585,
      "rewards/forensics_tiou_reward": 0.45616260170936584,
      "step": 614
    },
    {
      "completion_length": 26.0,
      "epoch": 1.8878648233486943,
      "grad_norm": 8.065606117248535,
      "kl": 0.583984375,
      "learning_rate": 5.276923076923076e-07,
      "loss": -0.1155,
      "reward": 0.753225177526474,
      "reward_std": 0.05519013851881027,
      "reward_std_all": 0.054801713675260544,
      "rewards-max/forensics_sinkhorn_reward": 0.9661325812339783,
      "rewards-max/forensics_strict_edge_reward": 0.4745020121335983,
      "rewards-max/forensics_tiou_reward": 0.9419885873794556,
      "rewards/forensics_sinkhorn_reward": 0.6347824335098267,
      "rewards/forensics_strict_edge_reward": 0.036011939868330956,
      "rewards/forensics_tiou_reward": 0.43180735409259796,
      "step": 615
    },
    {
      "completion_length": 26.046875,
      "epoch": 1.8909370199692779,
      "grad_norm": 7.4120306968688965,
      "kl": 0.4130859375,
      "learning_rate": 5.269230769230769e-07,
      "loss": 0.1326,
      "reward": 0.7594576776027679,
      "reward_std": 0.03303883224725723,
      "reward_std_all": 0.040945421904325485,
      "rewards-max/forensics_sinkhorn_reward": 0.9662649035453796,
      "rewards-max/forensics_strict_edge_reward": 0.24597223848104477,
      "rewards-max/forensics_tiou_reward": 0.947005569934845,
      "rewards/forensics_sinkhorn_reward": 0.6838105916976929,
      "rewards/forensics_strict_edge_reward": 0.026325412094593048,
      "rewards/forensics_tiou_reward": 0.5385235846042633,
      "step": 616
    },
    {
      "completion_length": 22.953125,
      "epoch": 1.8940092165898617,
      "grad_norm": 12.292673110961914,
      "kl": 0.583984375,
      "learning_rate": 5.261538461538462e-07,
      "loss": -0.1484,
      "reward": 0.7668816149234772,
      "reward_std": 0.05373137444257736,
      "reward_std_all": 0.05040904879570007,
      "rewards-max/forensics_sinkhorn_reward": 0.9566015899181366,
      "rewards-max/forensics_strict_edge_reward": 0.32561273872852325,
      "rewards-max/forensics_tiou_reward": 0.9188840985298157,
      "rewards/forensics_sinkhorn_reward": 0.7435448467731476,
      "rewards/forensics_strict_edge_reward": 0.037835896015167236,
      "rewards/forensics_tiou_reward": 0.5210738182067871,
      "step": 617
    },
    {
      "completion_length": 24.640625,
      "epoch": 1.8970814132104454,
      "grad_norm": 9.807506561279297,
      "kl": 0.3994140625,
      "learning_rate": 5.253846153846153e-07,
      "loss": 0.0155,
      "reward": 0.7607437968254089,
      "reward_std": 0.012224585749208927,
      "reward_std_all": 0.027835799381136894,
      "rewards-max/forensics_sinkhorn_reward": 0.9750677943229675,
      "rewards-max/forensics_strict_edge_reward": 0.39538006484508514,
      "rewards-max/forensics_tiou_reward": 0.964117169380188,
      "rewards/forensics_sinkhorn_reward": 0.7075563669204712,
      "rewards/forensics_strict_edge_reward": 0.04267165903002024,
      "rewards/forensics_tiou_reward": 0.4828393906354904,
      "step": 618
    },
    {
      "completion_length": 24.421875,
      "epoch": 1.9001536098310292,
      "grad_norm": 8.610296249389648,
      "kl": 0.4052734375,
      "learning_rate": 5.246153846153846e-07,
      "loss": 0.1043,
      "reward": 0.7156772017478943,
      "reward_std": 0.06081448495388031,
      "reward_std_all": 0.07397903874516487,
      "rewards-max/forensics_sinkhorn_reward": 0.9748809635639191,
      "rewards-max/forensics_strict_edge_reward": 0.6138583719730377,
      "rewards-max/forensics_tiou_reward": 0.9632463157176971,
      "rewards/forensics_sinkhorn_reward": 0.6569679975509644,
      "rewards/forensics_strict_edge_reward": 0.07930329814553261,
      "rewards/forensics_tiou_reward": 0.5212744921445847,
      "step": 619
    },
    {
      "completion_length": 23.546875,
      "epoch": 1.903225806451613,
      "grad_norm": 8.334692001342773,
      "kl": 0.39453125,
      "learning_rate": 5.238461538461538e-07,
      "loss": -0.0716,
      "reward": 0.7693097591400146,
      "reward_std": 0.03776501654647291,
      "reward_std_all": 0.038443869445472956,
      "rewards-max/forensics_sinkhorn_reward": 0.9871883988380432,
      "rewards-max/forensics_strict_edge_reward": 0.9232408702373505,
      "rewards-max/forensics_tiou_reward": 0.9952094852924347,
      "rewards/forensics_sinkhorn_reward": 0.6712642908096313,
      "rewards/forensics_strict_edge_reward": 0.073422035202384,
      "rewards/forensics_tiou_reward": 0.48922109603881836,
      "step": 620
    },
    {
      "completion_length": 25.203125,
      "epoch": 1.9062980030721968,
      "grad_norm": 8.06514835357666,
      "kl": 0.400390625,
      "learning_rate": 5.23076923076923e-07,
      "loss": -0.0914,
      "reward": 0.776239812374115,
      "reward_std": 0.02161787450313568,
      "reward_std_all": 0.029436267912387848,
      "rewards-max/forensics_sinkhorn_reward": 0.9838946163654327,
      "rewards-max/forensics_strict_edge_reward": 0.6398434042930603,
      "rewards-max/forensics_tiou_reward": 0.9853783547878265,
      "rewards/forensics_sinkhorn_reward": 0.7357923984527588,
      "rewards/forensics_strict_edge_reward": 0.09025942161679268,
      "rewards/forensics_tiou_reward": 0.5765398144721985,
      "step": 621
    },
    {
      "completion_length": 22.25,
      "epoch": 1.9093701996927803,
      "grad_norm": 8.757158279418945,
      "kl": 0.765625,
      "learning_rate": 5.223076923076924e-07,
      "loss": 0.0777,
      "reward": 0.75141242146492,
      "reward_std": 0.04279971309006214,
      "reward_std_all": 0.04677187465131283,
      "rewards-max/forensics_sinkhorn_reward": 0.9785903096199036,
      "rewards-max/forensics_strict_edge_reward": 0.498907133936882,
      "rewards-max/forensics_tiou_reward": 0.9727277457714081,
      "rewards/forensics_sinkhorn_reward": 0.6947367489337921,
      "rewards/forensics_strict_edge_reward": 0.03463442251086235,
      "rewards/forensics_tiou_reward": 0.4355352371931076,
      "step": 622
    },
    {
      "completion_length": 25.625,
      "epoch": 1.912442396313364,
      "grad_norm": 9.382254600524902,
      "kl": 0.703125,
      "learning_rate": 5.215384615384615e-07,
      "loss": -0.1115,
      "reward": 0.7509975433349609,
      "reward_std": 0.037185488268733025,
      "reward_std_all": 0.04807364195585251,
      "rewards-max/forensics_sinkhorn_reward": 0.9845291078090668,
      "rewards-max/forensics_strict_edge_reward": 0.6105617582798004,
      "rewards-max/forensics_tiou_reward": 0.9887564182281494,
      "rewards/forensics_sinkhorn_reward": 0.7172706723213196,
      "rewards/forensics_strict_edge_reward": 0.060546642169356346,
      "rewards/forensics_tiou_reward": 0.5593342185020447,
      "step": 623
    },
    {
      "completion_length": 26.890625,
      "epoch": 1.9155145929339477,
      "grad_norm": 10.780383110046387,
      "kl": 0.697265625,
      "learning_rate": 5.207692307692308e-07,
      "loss": 0.0881,
      "reward": 0.7436663508415222,
      "reward_std": 0.04811233468353748,
      "reward_std_all": 0.05462963506579399,
      "rewards-max/forensics_sinkhorn_reward": 0.9824548661708832,
      "rewards-max/forensics_strict_edge_reward": 0.5486607998609543,
      "rewards-max/forensics_tiou_reward": 0.9802001714706421,
      "rewards/forensics_sinkhorn_reward": 0.6372676491737366,
      "rewards/forensics_strict_edge_reward": 0.046935731545090675,
      "rewards/forensics_tiou_reward": 0.4313664883375168,
      "step": 624
    },
    {
      "completion_length": 28.375,
      "epoch": 1.9185867895545314,
      "grad_norm": 8.9025297164917,
      "kl": 0.5908203125,
      "learning_rate": 5.2e-07,
      "loss": 0.0395,
      "reward": 0.7065707743167877,
      "reward_std": 0.03475692309439182,
      "reward_std_all": 0.06648040376603603,
      "rewards-max/forensics_sinkhorn_reward": 0.9675474464893341,
      "rewards-max/forensics_strict_edge_reward": 0.53313247859478,
      "rewards-max/forensics_tiou_reward": 0.9469412863254547,
      "rewards/forensics_sinkhorn_reward": 0.6470930874347687,
      "rewards/forensics_strict_edge_reward": 0.06455943919718266,
      "rewards/forensics_tiou_reward": 0.4494034945964813,
      "step": 625
    },
    {
      "completion_length": 24.671875,
      "epoch": 1.9216589861751152,
      "grad_norm": 7.189444541931152,
      "kl": 0.5625,
      "learning_rate": 5.192307692307692e-07,
      "loss": 0.1463,
      "reward": 0.7631583213806152,
      "reward_std": 0.013367797248065472,
      "reward_std_all": 0.028130538761615753,
      "rewards-max/forensics_sinkhorn_reward": 0.9626717865467072,
      "rewards-max/forensics_strict_edge_reward": 0.44223348796367645,
      "rewards-max/forensics_tiou_reward": 0.9351585805416107,
      "rewards/forensics_sinkhorn_reward": 0.6978369057178497,
      "rewards/forensics_strict_edge_reward": 0.04225330054759979,
      "rewards/forensics_tiou_reward": 0.4797481596469879,
      "step": 626
    },
    {
      "completion_length": 25.578125,
      "epoch": 1.924731182795699,
      "grad_norm": 8.260209083557129,
      "kl": 0.4169921875,
      "learning_rate": 5.184615384615384e-07,
      "loss": 0.0591,
      "reward": 0.7503425478935242,
      "reward_std": 0.03369938116520643,
      "reward_std_all": 0.04325081780552864,
      "rewards-max/forensics_sinkhorn_reward": 0.9810862243175507,
      "rewards-max/forensics_strict_edge_reward": 0.6526533365249634,
      "rewards-max/forensics_tiou_reward": 0.980586051940918,
      "rewards/forensics_sinkhorn_reward": 0.6918725967407227,
      "rewards/forensics_strict_edge_reward": 0.04707571491599083,
      "rewards/forensics_tiou_reward": 0.5563943982124329,
      "step": 627
    },
    {
      "completion_length": 25.078125,
      "epoch": 1.9278033794162828,
      "grad_norm": 9.74130916595459,
      "kl": 0.576171875,
      "learning_rate": 5.176923076923076e-07,
      "loss": 0.2033,
      "reward": 0.7720814049243927,
      "reward_std": 0.04530736431479454,
      "reward_std_all": 0.042749198153615,
      "rewards-max/forensics_sinkhorn_reward": 0.9760995805263519,
      "rewards-max/forensics_strict_edge_reward": 0.39962221682071686,
      "rewards-max/forensics_tiou_reward": 0.9661596417427063,
      "rewards/forensics_sinkhorn_reward": 0.6954417824745178,
      "rewards/forensics_strict_edge_reward": 0.050171807408332825,
      "rewards/forensics_tiou_reward": 0.4930192083120346,
      "step": 628
    },
    {
      "completion_length": 23.703125,
      "epoch": 1.9308755760368663,
      "grad_norm": 10.295766830444336,
      "kl": 0.4921875,
      "learning_rate": 5.16923076923077e-07,
      "loss": 0.0779,
      "reward": 0.728902280330658,
      "reward_std": 0.06583728268742561,
      "reward_std_all": 0.06907589919865131,
      "rewards-max/forensics_sinkhorn_reward": 0.9691755473613739,
      "rewards-max/forensics_strict_edge_reward": 0.609315887093544,
      "rewards-max/forensics_tiou_reward": 0.954018771648407,
      "rewards/forensics_sinkhorn_reward": 0.6447120308876038,
      "rewards/forensics_strict_edge_reward": 0.04036960005760193,
      "rewards/forensics_tiou_reward": 0.4535859525203705,
      "step": 629
    },
    {
      "completion_length": 26.9375,
      "epoch": 1.93394777265745,
      "grad_norm": 7.69195032119751,
      "kl": 0.525390625,
      "learning_rate": 5.161538461538462e-07,
      "loss": 0.0355,
      "reward": 0.7346175312995911,
      "reward_std": 0.04675816558301449,
      "reward_std_all": 0.058490386232733727,
      "rewards-max/forensics_sinkhorn_reward": 0.9751054346561432,
      "rewards-max/forensics_strict_edge_reward": 0.3886224180459976,
      "rewards-max/forensics_tiou_reward": 0.9631458520889282,
      "rewards/forensics_sinkhorn_reward": 0.623367041349411,
      "rewards/forensics_strict_edge_reward": 0.040845511481165886,
      "rewards/forensics_tiou_reward": 0.4509088397026062,
      "step": 630
    },
    {
      "completion_length": 26.953125,
      "epoch": 1.9370199692780337,
      "grad_norm": 9.494501113891602,
      "kl": 0.5859375,
      "learning_rate": 5.153846153846153e-07,
      "loss": 0.0375,
      "reward": 0.7124539017677307,
      "reward_std": 0.07924556359648705,
      "reward_std_all": 0.08369983732700348,
      "rewards-max/forensics_sinkhorn_reward": 0.97956982254982,
      "rewards-max/forensics_strict_edge_reward": 0.6062345206737518,
      "rewards-max/forensics_tiou_reward": 0.9687568545341492,
      "rewards/forensics_sinkhorn_reward": 0.6512994170188904,
      "rewards/forensics_strict_edge_reward": 0.07196204736828804,
      "rewards/forensics_tiou_reward": 0.5004493296146393,
      "step": 631
    },
    {
      "completion_length": 24.109375,
      "epoch": 1.9400921658986174,
      "grad_norm": 8.73715877532959,
      "kl": 0.4248046875,
      "learning_rate": 5.146153846153846e-07,
      "loss": 0.0107,
      "reward": 0.7089962363243103,
      "reward_std": 0.06469153612852097,
      "reward_std_all": 0.08174499869346619,
      "rewards-max/forensics_sinkhorn_reward": 0.9410479068756104,
      "rewards-max/forensics_strict_edge_reward": 0.3230769634246826,
      "rewards-max/forensics_tiou_reward": 0.8866673409938812,
      "rewards/forensics_sinkhorn_reward": 0.6355780065059662,
      "rewards/forensics_strict_edge_reward": 0.030117479618638754,
      "rewards/forensics_tiou_reward": 0.41566887497901917,
      "step": 632
    },
    {
      "completion_length": 24.609375,
      "epoch": 1.9431643625192012,
      "grad_norm": 9.770853996276855,
      "kl": 0.4736328125,
      "learning_rate": 5.138461538461538e-07,
      "loss": -0.2991,
      "reward": 0.7099737524986267,
      "reward_std": 0.12005923315882683,
      "reward_std_all": 0.11217719316482544,
      "rewards-max/forensics_sinkhorn_reward": 0.9477909505367279,
      "rewards-max/forensics_strict_edge_reward": 0.2302108332514763,
      "rewards-max/forensics_tiou_reward": 0.8949525952339172,
      "rewards/forensics_sinkhorn_reward": 0.589911937713623,
      "rewards/forensics_strict_edge_reward": 0.03294603060930967,
      "rewards/forensics_tiou_reward": 0.40621377527713776,
      "step": 633
    },
    {
      "completion_length": 24.765625,
      "epoch": 1.946236559139785,
      "grad_norm": 8.244340896606445,
      "kl": 0.4794921875,
      "learning_rate": 5.13076923076923e-07,
      "loss": -0.0495,
      "reward": 0.7579388320446014,
      "reward_std": 0.048108287155628204,
      "reward_std_all": 0.048304734751582146,
      "rewards-max/forensics_sinkhorn_reward": 0.9659408330917358,
      "rewards-max/forensics_strict_edge_reward": 0.568371519446373,
      "rewards-max/forensics_tiou_reward": 0.9442170262336731,
      "rewards/forensics_sinkhorn_reward": 0.683285266160965,
      "rewards/forensics_strict_edge_reward": 0.08107257168740034,
      "rewards/forensics_tiou_reward": 0.5093249976634979,
      "step": 634
    },
    {
      "completion_length": 27.15625,
      "epoch": 1.9493087557603688,
      "grad_norm": 7.797660827636719,
      "kl": 0.6083984375,
      "learning_rate": 5.123076923076923e-07,
      "loss": -0.2788,
      "reward": 0.7295879125595093,
      "reward_std": 0.04137203097343445,
      "reward_std_all": 0.05914044938981533,
      "rewards-max/forensics_sinkhorn_reward": 0.9680424332618713,
      "rewards-max/forensics_strict_edge_reward": 0.5721234083175659,
      "rewards-max/forensics_tiou_reward": 0.9475648701190948,
      "rewards/forensics_sinkhorn_reward": 0.6455729901790619,
      "rewards/forensics_strict_edge_reward": 0.05386615730822086,
      "rewards/forensics_tiou_reward": 0.3999814838171005,
      "step": 635
    },
    {
      "completion_length": 23.8125,
      "epoch": 1.9523809523809523,
      "grad_norm": 8.097285270690918,
      "kl": 0.369140625,
      "learning_rate": 5.115384615384615e-07,
      "loss": 0.4272,
      "reward": 0.7390068173408508,
      "reward_std": 0.049689756706357,
      "reward_std_all": 0.05808538198471069,
      "rewards-max/forensics_sinkhorn_reward": 0.9669918417930603,
      "rewards-max/forensics_strict_edge_reward": 0.15060257533332333,
      "rewards-max/forensics_tiou_reward": 0.9393068552017212,
      "rewards/forensics_sinkhorn_reward": 0.6225714981555939,
      "rewards/forensics_strict_edge_reward": 0.009178612450341461,
      "rewards/forensics_tiou_reward": 0.4477717727422714,
      "step": 636
    },
    {
      "completion_length": 27.40625,
      "epoch": 1.955453149001536,
      "grad_norm": 8.732812881469727,
      "kl": 0.486328125,
      "learning_rate": 5.107692307692308e-07,
      "loss": 0.3616,
      "reward": 0.7275669574737549,
      "reward_std": 0.04314592853188515,
      "reward_std_all": 0.0628848560154438,
      "rewards-max/forensics_sinkhorn_reward": 0.9339307546615601,
      "rewards-max/forensics_strict_edge_reward": 0.2760728895664215,
      "rewards-max/forensics_tiou_reward": 0.8740580976009369,
      "rewards/forensics_sinkhorn_reward": 0.6227250099182129,
      "rewards/forensics_strict_edge_reward": 0.038330236449837685,
      "rewards/forensics_tiou_reward": 0.4277259558439255,
      "step": 637
    },
    {
      "completion_length": 28.75,
      "epoch": 1.9585253456221197,
      "grad_norm": 7.583689212799072,
      "kl": 0.4765625,
      "learning_rate": 5.1e-07,
      "loss": 0.0285,
      "reward": 0.7654030919075012,
      "reward_std": 0.05869393236935139,
      "reward_std_all": 0.05406299978494644,
      "rewards-max/forensics_sinkhorn_reward": 0.9675961136817932,
      "rewards-max/forensics_strict_edge_reward": 0.4981888085603714,
      "rewards-max/forensics_tiou_reward": 0.9411172568798065,
      "rewards/forensics_sinkhorn_reward": 0.7136395573616028,
      "rewards/forensics_strict_edge_reward": 0.057014184072613716,
      "rewards/forensics_tiou_reward": 0.5454835295677185,
      "step": 638
    },
    {
      "completion_length": 24.640625,
      "epoch": 1.9615975422427034,
      "grad_norm": 7.830003261566162,
      "kl": 0.3134765625,
      "learning_rate": 5.092307692307692e-07,
      "loss": -0.1324,
      "reward": 0.7561525702476501,
      "reward_std": 0.04991652723401785,
      "reward_std_all": 0.05136351566761732,
      "rewards-max/forensics_sinkhorn_reward": 0.9786379337310791,
      "rewards-max/forensics_strict_edge_reward": 0.6798866391181946,
      "rewards-max/forensics_tiou_reward": 0.9758245050907135,
      "rewards/forensics_sinkhorn_reward": 0.7092811465263367,
      "rewards/forensics_strict_edge_reward": 0.09391134232282639,
      "rewards/forensics_tiou_reward": 0.4965611398220062,
      "step": 639
    },
    {
      "completion_length": 22.125,
      "epoch": 1.9646697388632872,
      "grad_norm": 9.46375560760498,
      "kl": 0.857421875,
      "learning_rate": 5.084615384615384e-07,
      "loss": 0.2737,
      "reward": 0.7264959514141083,
      "reward_std": 0.07573520392179489,
      "reward_std_all": 0.08051169663667679,
      "rewards-max/forensics_sinkhorn_reward": 0.977269321680069,
      "rewards-max/forensics_strict_edge_reward": 0.6076286435127258,
      "rewards-max/forensics_tiou_reward": 0.969152420759201,
      "rewards/forensics_sinkhorn_reward": 0.6778141260147095,
      "rewards/forensics_strict_edge_reward": 0.054521530866622925,
      "rewards/forensics_tiou_reward": 0.4363018274307251,
      "step": 640
    },
    {
      "completion_length": 25.046875,
      "epoch": 1.967741935483871,
      "grad_norm": 7.51240348815918,
      "kl": 0.400390625,
      "learning_rate": 5.076923076923076e-07,
      "loss": -0.0681,
      "reward": 0.7650386691093445,
      "reward_std": 0.03662066999822855,
      "reward_std_all": 0.03873406816273928,
      "rewards-max/forensics_sinkhorn_reward": 0.9663010239601135,
      "rewards-max/forensics_strict_edge_reward": 0.4503692537546158,
      "rewards-max/forensics_tiou_reward": 0.9442845582962036,
      "rewards/forensics_sinkhorn_reward": 0.7018408179283142,
      "rewards/forensics_strict_edge_reward": 0.06822915188968182,
      "rewards/forensics_tiou_reward": 0.5363203585147858,
      "step": 641
    },
    {
      "completion_length": 23.5625,
      "epoch": 1.9708141321044548,
      "grad_norm": 7.983953475952148,
      "kl": 0.439453125,
      "learning_rate": 5.06923076923077e-07,
      "loss": 0.113,
      "reward": 0.7665106952190399,
      "reward_std": 0.05874912440776825,
      "reward_std_all": 0.05376586318016052,
      "rewards-max/forensics_sinkhorn_reward": 0.9882718324661255,
      "rewards-max/forensics_strict_edge_reward": 0.952418714761734,
      "rewards-max/forensics_tiou_reward": 0.9978364408016205,
      "rewards/forensics_sinkhorn_reward": 0.7469938099384308,
      "rewards/forensics_strict_edge_reward": 0.1256699040532112,
      "rewards/forensics_tiou_reward": 0.5754321217536926,
      "step": 642
    },
    {
      "completion_length": 23.890625,
      "epoch": 1.9738863287250386,
      "grad_norm": 9.543411254882812,
      "kl": 0.5712890625,
      "learning_rate": 5.061538461538462e-07,
      "loss": 0.2998,
      "reward": 0.736214816570282,
      "reward_std": 0.0504688061773777,
      "reward_std_all": 0.05957382544875145,
      "rewards-max/forensics_sinkhorn_reward": 0.9661588072776794,
      "rewards-max/forensics_strict_edge_reward": 0.2774585038423538,
      "rewards-max/forensics_tiou_reward": 0.9378353357315063,
      "rewards/forensics_sinkhorn_reward": 0.598191499710083,
      "rewards/forensics_strict_edge_reward": 0.02384125255048275,
      "rewards/forensics_tiou_reward": 0.4095078110694885,
      "step": 643
    },
    {
      "completion_length": 23.734375,
      "epoch": 1.976958525345622,
      "grad_norm": 10.251206398010254,
      "kl": 0.5947265625,
      "learning_rate": 5.053846153846154e-07,
      "loss": 0.2549,
      "reward": 0.7337405681610107,
      "reward_std": 0.06229928694665432,
      "reward_std_all": 0.06949535384774208,
      "rewards-max/forensics_sinkhorn_reward": 0.9759613573551178,
      "rewards-max/forensics_strict_edge_reward": 0.44549836218357086,
      "rewards-max/forensics_tiou_reward": 0.9603748023509979,
      "rewards/forensics_sinkhorn_reward": 0.6930748224258423,
      "rewards/forensics_strict_edge_reward": 0.043057993054389954,
      "rewards/forensics_tiou_reward": 0.4801108092069626,
      "step": 644
    },
    {
      "completion_length": 25.1875,
      "epoch": 1.9800307219662059,
      "grad_norm": 7.864706993103027,
      "kl": 0.3818359375,
      "learning_rate": 5.046153846153846e-07,
      "loss": 0.3303,
      "reward": 0.7352389991283417,
      "reward_std": 0.059518806636333466,
      "reward_std_all": 0.06470109522342682,
      "rewards-max/forensics_sinkhorn_reward": 0.9780513346195221,
      "rewards-max/forensics_strict_edge_reward": 0.6005697548389435,
      "rewards-max/forensics_tiou_reward": 0.9719972610473633,
      "rewards/forensics_sinkhorn_reward": 0.6552978157997131,
      "rewards/forensics_strict_edge_reward": 0.08245376124978065,
      "rewards/forensics_tiou_reward": 0.4998088777065277,
      "step": 645
    },
    {
      "completion_length": 25.65625,
      "epoch": 1.9831029185867894,
      "grad_norm": 9.324623107910156,
      "kl": 0.3916015625,
      "learning_rate": 5.038461538461538e-07,
      "loss": 0.1763,
      "reward": 0.7505972981452942,
      "reward_std": 0.03957538679242134,
      "reward_std_all": 0.04864780604839325,
      "rewards-max/forensics_sinkhorn_reward": 0.9518121182918549,
      "rewards-max/forensics_strict_edge_reward": 0.17688647098839283,
      "rewards-max/forensics_tiou_reward": 0.9025538563728333,
      "rewards/forensics_sinkhorn_reward": 0.6643968224525452,
      "rewards/forensics_strict_edge_reward": 0.021735436283051968,
      "rewards/forensics_tiou_reward": 0.46804460883140564,
      "step": 646
    },
    {
      "completion_length": 23.703125,
      "epoch": 1.9861751152073732,
      "grad_norm": 11.464422225952148,
      "kl": 0.4921875,
      "learning_rate": 5.03076923076923e-07,
      "loss": 0.1256,
      "reward": 0.7870436310768127,
      "reward_std": 0.023613831494003534,
      "reward_std_all": 0.024309254251420498,
      "rewards-max/forensics_sinkhorn_reward": 0.9639352560043335,
      "rewards-max/forensics_strict_edge_reward": 0.3480503261089325,
      "rewards-max/forensics_tiou_reward": 0.9439428746700287,
      "rewards/forensics_sinkhorn_reward": 0.7041946053504944,
      "rewards/forensics_strict_edge_reward": 0.021106778644025326,
      "rewards/forensics_tiou_reward": 0.4339616447687149,
      "step": 647
    },
    {
      "completion_length": 28.203125,
      "epoch": 1.989247311827957,
      "grad_norm": 10.057032585144043,
      "kl": 0.455078125,
      "learning_rate": 5.023076923076923e-07,
      "loss": 0.177,
      "reward": 0.7065036296844482,
      "reward_std": 0.04765598475933075,
      "reward_std_all": 0.07257692143321037,
      "rewards-max/forensics_sinkhorn_reward": 0.9710382521152496,
      "rewards-max/forensics_strict_edge_reward": 0.4131776839494705,
      "rewards-max/forensics_tiou_reward": 0.9517691433429718,
      "rewards/forensics_sinkhorn_reward": 0.630883127450943,
      "rewards/forensics_strict_edge_reward": 0.04825916886329651,
      "rewards/forensics_tiou_reward": 0.48590612411499023,
      "step": 648
    },
    {
      "completion_length": 25.296875,
      "epoch": 1.9923195084485408,
      "grad_norm": 8.54576587677002,
      "kl": 0.4658203125,
      "learning_rate": 5.015384615384614e-07,
      "loss": -0.1119,
      "reward": 0.7564697861671448,
      "reward_std": 0.025493478402495384,
      "reward_std_all": 0.03618000168353319,
      "rewards-max/forensics_sinkhorn_reward": 0.9738735854625702,
      "rewards-max/forensics_strict_edge_reward": 0.3784440755844116,
      "rewards-max/forensics_tiou_reward": 0.9624233543872833,
      "rewards/forensics_sinkhorn_reward": 0.6840881109237671,
      "rewards/forensics_strict_edge_reward": 0.042034195736050606,
      "rewards/forensics_tiou_reward": 0.527394562959671,
      "step": 649
    },
    {
      "completion_length": 23.1875,
      "epoch": 1.9953917050691246,
      "grad_norm": 8.484838485717773,
      "kl": 0.4765625,
      "learning_rate": 5.007692307692308e-07,
      "loss": 0.1391,
      "reward": 0.7310166358947754,
      "reward_std": 0.06823333725333214,
      "reward_std_all": 0.06952713802456856,
      "rewards-max/forensics_sinkhorn_reward": 0.9804619550704956,
      "rewards-max/forensics_strict_edge_reward": 0.6302271485328674,
      "rewards-max/forensics_tiou_reward": 0.97981858253479,
      "rewards/forensics_sinkhorn_reward": 0.6469120681285858,
      "rewards/forensics_strict_edge_reward": 0.06088072806596756,
      "rewards/forensics_tiou_reward": 0.47083593904972076,
      "step": 650
    },
    {
      "completion_length": 25.28125,
      "epoch": 1.998463901689708,
      "grad_norm": 9.220866203308105,
      "kl": 0.4052734375,
      "learning_rate": 5e-07,
      "loss": 0.287,
      "reward": 0.7473039031028748,
      "reward_std": 0.05954922363162041,
      "reward_std_all": 0.05916214548051357,
      "rewards-max/forensics_sinkhorn_reward": 0.9691986441612244,
      "rewards-max/forensics_strict_edge_reward": 0.46446214616298676,
      "rewards-max/forensics_tiou_reward": 0.9309081733226776,
      "rewards/forensics_sinkhorn_reward": 0.6431684494018555,
      "rewards/forensics_strict_edge_reward": 0.02845107391476631,
      "rewards/forensics_tiou_reward": 0.510439395904541,
      "step": 651
    },
    {
      "completion_length": 24.399999618530273,
      "epoch": 2.0,
      "grad_norm": 5.972357273101807,
      "kl": 0.80859375,
      "learning_rate": 4.992307692307692e-07,
      "loss": 0.119,
      "reward": 0.5435653924942017,
      "reward_std": 0.0440434031188488,
      "reward_std_all": 0.14407508075237274,
      "rewards-max/forensics_sinkhorn_reward": 0.6851330995559692,
      "rewards-max/forensics_strict_edge_reward": 0.004701781086623669,
      "rewards-max/forensics_tiou_reward": 0.43586108088493347,
      "rewards/forensics_sinkhorn_reward": 0.3893527388572693,
      "rewards/forensics_strict_edge_reward": 0.0009419447160325944,
      "rewards/forensics_tiou_reward": 0.16591016948223114,
      "step": 652
    },
    {
      "completion_length": 22.0,
      "epoch": 2.003072196620584,
      "grad_norm": 8.584124565124512,
      "kl": 0.501953125,
      "learning_rate": 4.984615384615384e-07,
      "loss": -0.005,
      "reward": 0.7556856870651245,
      "reward_std": 0.017423197394236922,
      "reward_std_all": 0.03475397266447544,
      "rewards-max/forensics_sinkhorn_reward": 0.9678347110748291,
      "rewards-max/forensics_strict_edge_reward": 0.6966203600168228,
      "rewards-max/forensics_tiou_reward": 0.9520581364631653,
      "rewards/forensics_sinkhorn_reward": 0.6366041600704193,
      "rewards/forensics_strict_edge_reward": 0.0696968138217926,
      "rewards/forensics_tiou_reward": 0.46536001563072205,
      "step": 653
    },
    {
      "completion_length": 25.03125,
      "epoch": 2.0061443932411676,
      "grad_norm": 8.937191009521484,
      "kl": 0.453125,
      "learning_rate": 4.976923076923077e-07,
      "loss": -0.1018,
      "reward": 0.7599141597747803,
      "reward_std": 0.03132901433855295,
      "reward_std_all": 0.03864874690771103,
      "rewards-max/forensics_sinkhorn_reward": 0.9877842366695404,
      "rewards-max/forensics_strict_edge_reward": 0.6886506080627441,
      "rewards-max/forensics_tiou_reward": 0.9954663217067719,
      "rewards/forensics_sinkhorn_reward": 0.7069060802459717,
      "rewards/forensics_strict_edge_reward": 0.1026785597205162,
      "rewards/forensics_tiou_reward": 0.5111735165119171,
      "step": 654
    },
    {
      "completion_length": 28.234375,
      "epoch": 2.0092165898617513,
      "grad_norm": 7.304177284240723,
      "kl": 0.607421875,
      "learning_rate": 4.969230769230769e-07,
      "loss": 0.0925,
      "reward": 0.7834768295288086,
      "reward_std": 0.0317467269487679,
      "reward_std_all": 0.030602372251451015,
      "rewards-max/forensics_sinkhorn_reward": 0.9404640793800354,
      "rewards-max/forensics_strict_edge_reward": 0.33302006497979164,
      "rewards-max/forensics_tiou_reward": 0.8853645622730255,
      "rewards/forensics_sinkhorn_reward": 0.7191675305366516,
      "rewards/forensics_strict_edge_reward": 0.04526465805247426,
      "rewards/forensics_tiou_reward": 0.5156634449958801,
      "step": 655
    },
    {
      "completion_length": 24.46875,
      "epoch": 2.0122887864823347,
      "grad_norm": 8.426223754882812,
      "kl": 0.650390625,
      "learning_rate": 4.961538461538462e-07,
      "loss": 0.2481,
      "reward": 0.7479296624660492,
      "reward_std": 0.07218513265252113,
      "reward_std_all": 0.06843555718660355,
      "rewards-max/forensics_sinkhorn_reward": 0.9794796407222748,
      "rewards-max/forensics_strict_edge_reward": 0.824172168970108,
      "rewards-max/forensics_tiou_reward": 0.9736082255840302,
      "rewards/forensics_sinkhorn_reward": 0.6706903874874115,
      "rewards/forensics_strict_edge_reward": 0.07786879315972328,
      "rewards/forensics_tiou_reward": 0.5152424275875092,
      "step": 656
    },
    {
      "completion_length": 25.9375,
      "epoch": 2.0153609831029184,
      "grad_norm": 8.294068336486816,
      "kl": 0.576171875,
      "learning_rate": 4.953846153846154e-07,
      "loss": 0.1339,
      "reward": 0.7268795073032379,
      "reward_std": 0.06265752203762531,
      "reward_std_all": 0.07313108816742897,
      "rewards-max/forensics_sinkhorn_reward": 0.9676123857498169,
      "rewards-max/forensics_strict_edge_reward": 0.45854122936725616,
      "rewards-max/forensics_tiou_reward": 0.9416330754756927,
      "rewards/forensics_sinkhorn_reward": 0.6652329564094543,
      "rewards/forensics_strict_edge_reward": 0.04690377600491047,
      "rewards/forensics_tiou_reward": 0.4977833926677704,
      "step": 657
    },
    {
      "completion_length": 25.453125,
      "epoch": 2.0184331797235022,
      "grad_norm": 12.68608283996582,
      "kl": 0.541015625,
      "learning_rate": 4.946153846153846e-07,
      "loss": -0.0731,
      "reward": 0.769486129283905,
      "reward_std": 0.036884333938360214,
      "reward_std_all": 0.03749613091349602,
      "rewards-max/forensics_sinkhorn_reward": 0.9798749685287476,
      "rewards-max/forensics_strict_edge_reward": 0.6515887677669525,
      "rewards-max/forensics_tiou_reward": 0.9765490889549255,
      "rewards/forensics_sinkhorn_reward": 0.6530567705631256,
      "rewards/forensics_strict_edge_reward": 0.05814998224377632,
      "rewards/forensics_tiou_reward": 0.48604026436805725,
      "step": 658
    },
    {
      "completion_length": 22.546875,
      "epoch": 2.021505376344086,
      "grad_norm": 10.570765495300293,
      "kl": 0.6796875,
      "learning_rate": 4.938461538461538e-07,
      "loss": 0.1734,
      "reward": 0.7256127297878265,
      "reward_std": 0.058001033030450344,
      "reward_std_all": 0.0675609540194273,
      "rewards-max/forensics_sinkhorn_reward": 0.9746812582015991,
      "rewards-max/forensics_strict_edge_reward": 0.4251216650009155,
      "rewards-max/forensics_tiou_reward": 0.9611612260341644,
      "rewards/forensics_sinkhorn_reward": 0.6647452116012573,
      "rewards/forensics_strict_edge_reward": 0.043571203015744686,
      "rewards/forensics_tiou_reward": 0.4534592628479004,
      "step": 659
    },
    {
      "completion_length": 23.890625,
      "epoch": 2.02457757296467,
      "grad_norm": 8.345492362976074,
      "kl": 0.724609375,
      "learning_rate": 4.93076923076923e-07,
      "loss": -0.059,
      "reward": 0.7477774322032928,
      "reward_std": 0.02490113116800785,
      "reward_std_all": 0.04274108214303851,
      "rewards-max/forensics_sinkhorn_reward": 0.9840119481086731,
      "rewards-max/forensics_strict_edge_reward": 0.4974842667579651,
      "rewards-max/forensics_tiou_reward": 0.9850876033306122,
      "rewards/forensics_sinkhorn_reward": 0.7084186971187592,
      "rewards/forensics_strict_edge_reward": 0.06690939515829086,
      "rewards/forensics_tiou_reward": 0.5095901042222977,
      "step": 660
    },
    {
      "completion_length": 26.71875,
      "epoch": 2.0276497695852536,
      "grad_norm": 8.322662353515625,
      "kl": 0.703125,
      "learning_rate": 4.923076923076923e-07,
      "loss": -0.0136,
      "reward": 0.7355703115463257,
      "reward_std": 0.04231216013431549,
      "reward_std_all": 0.05686655454337597,
      "rewards-max/forensics_sinkhorn_reward": 0.9827508330345154,
      "rewards-max/forensics_strict_edge_reward": 0.5587892681360245,
      "rewards-max/forensics_tiou_reward": 0.9834524393081665,
      "rewards/forensics_sinkhorn_reward": 0.6823016703128815,
      "rewards/forensics_strict_edge_reward": 0.04825214855372906,
      "rewards/forensics_tiou_reward": 0.45407959818840027,
      "step": 661
    },
    {
      "completion_length": 23.796875,
      "epoch": 2.0307219662058373,
      "grad_norm": 8.49409294128418,
      "kl": 0.4013671875,
      "learning_rate": 4.915384615384615e-07,
      "loss": -0.0047,
      "reward": 0.7908967137336731,
      "reward_std": 0.020450366660952568,
      "reward_std_all": 0.020636992529034615,
      "rewards-max/forensics_sinkhorn_reward": 0.9822184145450592,
      "rewards-max/forensics_strict_edge_reward": 0.756708562374115,
      "rewards-max/forensics_tiou_reward": 0.98367640376091,
      "rewards/forensics_sinkhorn_reward": 0.7512143850326538,
      "rewards/forensics_strict_edge_reward": 0.08308817259967327,
      "rewards/forensics_tiou_reward": 0.5559586882591248,
      "step": 662
    },
    {
      "completion_length": 24.6875,
      "epoch": 2.033794162826421,
      "grad_norm": 9.415334701538086,
      "kl": 0.509765625,
      "learning_rate": 4.907692307692308e-07,
      "loss": 0.086,
      "reward": 0.7718183696269989,
      "reward_std": 0.058040738105773926,
      "reward_std_all": 0.052575403824448586,
      "rewards-max/forensics_sinkhorn_reward": 0.9629652798175812,
      "rewards-max/forensics_strict_edge_reward": 0.21591167524456978,
      "rewards-max/forensics_tiou_reward": 0.9221157133579254,
      "rewards/forensics_sinkhorn_reward": 0.7398472428321838,
      "rewards/forensics_strict_edge_reward": 0.03470308147370815,
      "rewards/forensics_tiou_reward": 0.5622615665197372,
      "step": 663
    },
    {
      "completion_length": 26.765625,
      "epoch": 2.0368663594470044,
      "grad_norm": 10.488137245178223,
      "kl": 0.75390625,
      "learning_rate": 4.9e-07,
      "loss": 0.2958,
      "reward": 0.7210399806499481,
      "reward_std": 0.06338011287152767,
      "reward_std_all": 0.07369189709424973,
      "rewards-max/forensics_sinkhorn_reward": 0.9780005216598511,
      "rewards-max/forensics_strict_edge_reward": 0.7246644794940948,
      "rewards-max/forensics_tiou_reward": 0.9726545810699463,
      "rewards/forensics_sinkhorn_reward": 0.6359301507472992,
      "rewards/forensics_strict_edge_reward": 0.08395780995488167,
      "rewards/forensics_tiou_reward": 0.45918750762939453,
      "step": 664
    },
    {
      "completion_length": 26.515625,
      "epoch": 2.0399385560675882,
      "grad_norm": 9.105920791625977,
      "kl": 0.34375,
      "learning_rate": 4.892307692307692e-07,
      "loss": -0.2136,
      "reward": 0.8024219870567322,
      "reward_std": 0.005606764752883464,
      "reward_std_all": 0.007205056957900524,
      "rewards-max/forensics_sinkhorn_reward": 0.9800857305526733,
      "rewards-max/forensics_strict_edge_reward": 0.8169413805007935,
      "rewards-max/forensics_tiou_reward": 0.9810543060302734,
      "rewards/forensics_sinkhorn_reward": 0.7787728905677795,
      "rewards/forensics_strict_edge_reward": 0.09491118788719177,
      "rewards/forensics_tiou_reward": 0.5999360978603363,
      "step": 665
    },
    {
      "completion_length": 22.65625,
      "epoch": 2.043010752688172,
      "grad_norm": 10.616456031799316,
      "kl": 0.455078125,
      "learning_rate": 4.884615384615384e-07,
      "loss": -0.0747,
      "reward": 0.7834183573722839,
      "reward_std": 0.035886434838175774,
      "reward_std_all": 0.03324375208467245,
      "rewards-max/forensics_sinkhorn_reward": 0.9847981929779053,
      "rewards-max/forensics_strict_edge_reward": 0.8418737053871155,
      "rewards-max/forensics_tiou_reward": 0.9898942112922668,
      "rewards/forensics_sinkhorn_reward": 0.7558088600635529,
      "rewards/forensics_strict_edge_reward": 0.07558072730898857,
      "rewards/forensics_tiou_reward": 0.5428484082221985,
      "step": 666
    },
    {
      "completion_length": 24.375,
      "epoch": 2.046082949308756,
      "grad_norm": 12.556605339050293,
      "kl": 0.48046875,
      "learning_rate": 4.876923076923077e-07,
      "loss": -0.0886,
      "reward": 0.7189721763134003,
      "reward_std": 0.06227685324847698,
      "reward_std_all": 0.07823273912072182,
      "rewards-max/forensics_sinkhorn_reward": 0.9867448806762695,
      "rewards-max/forensics_strict_edge_reward": 0.8582656681537628,
      "rewards-max/forensics_tiou_reward": 0.9930993318557739,
      "rewards/forensics_sinkhorn_reward": 0.7070910334587097,
      "rewards/forensics_strict_edge_reward": 0.07247588038444519,
      "rewards/forensics_tiou_reward": 0.5356591045856476,
      "step": 667
    },
    {
      "completion_length": 22.890625,
      "epoch": 2.0491551459293396,
      "grad_norm": 10.401562690734863,
      "kl": 0.6259765625,
      "learning_rate": 4.869230769230769e-07,
      "loss": -0.4213,
      "reward": 0.7344475388526917,
      "reward_std": 0.01279621000867337,
      "reward_std_all": 0.04449213296175003,
      "rewards-max/forensics_sinkhorn_reward": 0.9803128242492676,
      "rewards-max/forensics_strict_edge_reward": 0.6132667809724808,
      "rewards-max/forensics_tiou_reward": 0.9757797718048096,
      "rewards/forensics_sinkhorn_reward": 0.7313399314880371,
      "rewards/forensics_strict_edge_reward": 0.10663034580647945,
      "rewards/forensics_tiou_reward": 0.5962493121623993,
      "step": 668
    },
    {
      "completion_length": 23.375,
      "epoch": 2.0522273425499233,
      "grad_norm": 11.751771926879883,
      "kl": 0.37109375,
      "learning_rate": 4.861538461538461e-07,
      "loss": -0.1624,
      "reward": 0.7588198482990265,
      "reward_std": 0.059430841356515884,
      "reward_std_all": 0.0558307021856308,
      "rewards-max/forensics_sinkhorn_reward": 0.9806148111820221,
      "rewards-max/forensics_strict_edge_reward": 0.835382878780365,
      "rewards-max/forensics_tiou_reward": 0.9808234870433807,
      "rewards/forensics_sinkhorn_reward": 0.7094148397445679,
      "rewards/forensics_strict_edge_reward": 0.0883314199745655,
      "rewards/forensics_tiou_reward": 0.5354293882846832,
      "step": 669
    },
    {
      "completion_length": 23.546875,
      "epoch": 2.055299539170507,
      "grad_norm": 9.763998985290527,
      "kl": 0.58984375,
      "learning_rate": 4.853846153846154e-07,
      "loss": -0.0797,
      "reward": 0.743597686290741,
      "reward_std": 0.034660580568015575,
      "reward_std_all": 0.047943372279405594,
      "rewards-max/forensics_sinkhorn_reward": 0.9886049628257751,
      "rewards-max/forensics_strict_edge_reward": 0.9836080372333527,
      "rewards-max/forensics_tiou_reward": 0.9989905059337616,
      "rewards/forensics_sinkhorn_reward": 0.7015693783760071,
      "rewards/forensics_strict_edge_reward": 0.14760678261518478,
      "rewards/forensics_tiou_reward": 0.5153864026069641,
      "step": 670
    },
    {
      "completion_length": 21.5625,
      "epoch": 2.0583717357910905,
      "grad_norm": 9.184829711914062,
      "kl": 0.5361328125,
      "learning_rate": 4.846153846153846e-07,
      "loss": 0.1923,
      "reward": 0.7107875943183899,
      "reward_std": 0.030409510247409344,
      "reward_std_all": 0.0595558937638998,
      "rewards-max/forensics_sinkhorn_reward": 0.9812517762184143,
      "rewards-max/forensics_strict_edge_reward": 0.5050012171268463,
      "rewards-max/forensics_tiou_reward": 0.978785902261734,
      "rewards/forensics_sinkhorn_reward": 0.6643700301647186,
      "rewards/forensics_strict_edge_reward": 0.06289427727460861,
      "rewards/forensics_tiou_reward": 0.44562971591949463,
      "step": 671
    },
    {
      "completion_length": 27.6875,
      "epoch": 2.0614439324116742,
      "grad_norm": 7.964746475219727,
      "kl": 0.51171875,
      "learning_rate": 4.838461538461539e-07,
      "loss": -0.0194,
      "reward": 0.7403744459152222,
      "reward_std": 0.03978107147850096,
      "reward_std_all": 0.04979968210682273,
      "rewards-max/forensics_sinkhorn_reward": 0.9849500954151154,
      "rewards-max/forensics_strict_edge_reward": 0.8418737053871155,
      "rewards-max/forensics_tiou_reward": 0.9893614947795868,
      "rewards/forensics_sinkhorn_reward": 0.656443178653717,
      "rewards/forensics_strict_edge_reward": 0.07409025728702545,
      "rewards/forensics_tiou_reward": 0.5070490092039108,
      "step": 672
    },
    {
      "completion_length": 27.3125,
      "epoch": 2.064516129032258,
      "grad_norm": 8.725215911865234,
      "kl": 0.59765625,
      "learning_rate": 4.83076923076923e-07,
      "loss": -0.0,
      "reward": 0.7627489566802979,
      "reward_std": 0.031425971537828445,
      "reward_std_all": 0.03712864965200424,
      "rewards-max/forensics_sinkhorn_reward": 0.9807568490505219,
      "rewards-max/forensics_strict_edge_reward": 0.7047859728336334,
      "rewards-max/forensics_tiou_reward": 0.9775735437870026,
      "rewards/forensics_sinkhorn_reward": 0.7235098481178284,
      "rewards/forensics_strict_edge_reward": 0.0769902840256691,
      "rewards/forensics_tiou_reward": 0.5435462892055511,
      "step": 673
    },
    {
      "completion_length": 22.828125,
      "epoch": 2.067588325652842,
      "grad_norm": 8.210840225219727,
      "kl": 0.42578125,
      "learning_rate": 4.823076923076923e-07,
      "loss": 0.1386,
      "reward": 0.7153899967670441,
      "reward_std": 0.026884766295552254,
      "reward_std_all": 0.0576600506901741,
      "rewards-max/forensics_sinkhorn_reward": 0.9854908883571625,
      "rewards-max/forensics_strict_edge_reward": 0.8079957664012909,
      "rewards-max/forensics_tiou_reward": 0.9907732605934143,
      "rewards/forensics_sinkhorn_reward": 0.6590668261051178,
      "rewards/forensics_strict_edge_reward": 0.15059997141361237,
      "rewards/forensics_tiou_reward": 0.5215553492307663,
      "step": 674
    },
    {
      "completion_length": 24.59375,
      "epoch": 2.0706605222734256,
      "grad_norm": 9.407227516174316,
      "kl": 0.63671875,
      "learning_rate": 4.815384615384615e-07,
      "loss": 0.1146,
      "reward": 0.7599422931671143,
      "reward_std": 0.05545857176184654,
      "reward_std_all": 0.053197259083390236,
      "rewards-max/forensics_sinkhorn_reward": 0.9656216502189636,
      "rewards-max/forensics_strict_edge_reward": 0.514603697694838,
      "rewards-max/forensics_tiou_reward": 0.9431284666061401,
      "rewards/forensics_sinkhorn_reward": 0.7080641388893127,
      "rewards/forensics_strict_edge_reward": 0.07334065542090684,
      "rewards/forensics_tiou_reward": 0.5000176280736923,
      "step": 675
    },
    {
      "completion_length": 23.546875,
      "epoch": 2.0737327188940093,
      "grad_norm": 10.95350170135498,
      "kl": 0.490234375,
      "learning_rate": 4.807692307692307e-07,
      "loss": -0.0084,
      "reward": 0.7614875435829163,
      "reward_std": 0.010172897251322865,
      "reward_std_all": 0.027084945468232036,
      "rewards-max/forensics_sinkhorn_reward": 0.9770113825798035,
      "rewards-max/forensics_strict_edge_reward": 0.5144400894641876,
      "rewards-max/forensics_tiou_reward": 0.9645033180713654,
      "rewards/forensics_sinkhorn_reward": 0.6840532422065735,
      "rewards/forensics_strict_edge_reward": 0.06467522121965885,
      "rewards/forensics_tiou_reward": 0.5199654400348663,
      "step": 676
    },
    {
      "completion_length": 23.515625,
      "epoch": 2.076804915514593,
      "grad_norm": 9.063121795654297,
      "kl": 0.56640625,
      "learning_rate": 4.8e-07,
      "loss": 0.1615,
      "reward": 0.7758253812789917,
      "reward_std": 0.03754499927163124,
      "reward_std_all": 0.03703240118920803,
      "rewards-max/forensics_sinkhorn_reward": 0.9557233452796936,
      "rewards-max/forensics_strict_edge_reward": 0.3680838346481323,
      "rewards-max/forensics_tiou_reward": 0.9206410348415375,
      "rewards/forensics_sinkhorn_reward": 0.6938876807689667,
      "rewards/forensics_strict_edge_reward": 0.03325390163809061,
      "rewards/forensics_tiou_reward": 0.4528942406177521,
      "step": 677
    },
    {
      "completion_length": 24.625,
      "epoch": 2.0798771121351765,
      "grad_norm": 10.195016860961914,
      "kl": 0.4599609375,
      "learning_rate": 4.792307692307692e-07,
      "loss": -0.0557,
      "reward": 0.789449542760849,
      "reward_std": 0.023352724267169833,
      "reward_std_all": 0.023027152521535754,
      "rewards-max/forensics_sinkhorn_reward": 0.9789941012859344,
      "rewards-max/forensics_strict_edge_reward": 0.568440854549408,
      "rewards-max/forensics_tiou_reward": 0.9725576043128967,
      "rewards/forensics_sinkhorn_reward": 0.7689905762672424,
      "rewards/forensics_strict_edge_reward": 0.07118046935647726,
      "rewards/forensics_tiou_reward": 0.5745604038238525,
      "step": 678
    },
    {
      "completion_length": 22.515625,
      "epoch": 2.0829493087557602,
      "grad_norm": 14.062173843383789,
      "kl": 0.68359375,
      "learning_rate": 4.784615384615384e-07,
      "loss": -0.105,
      "reward": 0.7082260847091675,
      "reward_std": 0.07987850159406662,
      "reward_std_all": 0.09238788485527039,
      "rewards-max/forensics_sinkhorn_reward": 0.9779860973358154,
      "rewards-max/forensics_strict_edge_reward": 0.7226715534925461,
      "rewards-max/forensics_tiou_reward": 0.973359078168869,
      "rewards/forensics_sinkhorn_reward": 0.6175424158573151,
      "rewards/forensics_strict_edge_reward": 0.03843730129301548,
      "rewards/forensics_tiou_reward": 0.34991031885147095,
      "step": 679
    },
    {
      "completion_length": 28.796875,
      "epoch": 2.086021505376344,
      "grad_norm": 14.239874839782715,
      "kl": 0.4794921875,
      "learning_rate": 4.776923076923077e-07,
      "loss": -0.0673,
      "reward": 0.7040892243385315,
      "reward_std": 0.03759151138365269,
      "reward_std_all": 0.06628837063908577,
      "rewards-max/forensics_sinkhorn_reward": 0.975633829832077,
      "rewards-max/forensics_strict_edge_reward": 0.4972463697195053,
      "rewards-max/forensics_tiou_reward": 0.9629756510257721,
      "rewards/forensics_sinkhorn_reward": 0.6349876523017883,
      "rewards/forensics_strict_edge_reward": 0.05265680607408285,
      "rewards/forensics_tiou_reward": 0.47914573550224304,
      "step": 680
    },
    {
      "completion_length": 26.390625,
      "epoch": 2.089093701996928,
      "grad_norm": 9.669878005981445,
      "kl": 0.578125,
      "learning_rate": 4.769230769230769e-07,
      "loss": -0.0141,
      "reward": 0.7909595370292664,
      "reward_std": 0.013534557539969683,
      "reward_std_all": 0.01613216334953904,
      "rewards-max/forensics_sinkhorn_reward": 0.9774435758590698,
      "rewards-max/forensics_strict_edge_reward": 0.5368025302886963,
      "rewards-max/forensics_tiou_reward": 0.9658215045928955,
      "rewards/forensics_sinkhorn_reward": 0.7468937039375305,
      "rewards/forensics_strict_edge_reward": 0.06484897993505001,
      "rewards/forensics_tiou_reward": 0.5416140258312225,
      "step": 681
    },
    {
      "completion_length": 23.03125,
      "epoch": 2.0921658986175116,
      "grad_norm": 12.28387451171875,
      "kl": 0.4296875,
      "learning_rate": 4.7615384615384613e-07,
      "loss": -0.0749,
      "reward": 0.7596421241760254,
      "reward_std": 0.027318977750837803,
      "reward_std_all": 0.03952999925240874,
      "rewards-max/forensics_sinkhorn_reward": 0.9658332467079163,
      "rewards-max/forensics_strict_edge_reward": 0.2855285704135895,
      "rewards-max/forensics_tiou_reward": 0.9414447247982025,
      "rewards/forensics_sinkhorn_reward": 0.7003965675830841,
      "rewards/forensics_strict_edge_reward": 0.025498253293335438,
      "rewards/forensics_tiou_reward": 0.45791125297546387,
      "step": 682
    },
    {
      "completion_length": 23.328125,
      "epoch": 2.0952380952380953,
      "grad_norm": 8.214191436767578,
      "kl": 0.4873046875,
      "learning_rate": 4.753846153846154e-07,
      "loss": -0.0267,
      "reward": 0.7750751376152039,
      "reward_std": 0.03647145535796881,
      "reward_std_all": 0.03619304858148098,
      "rewards-max/forensics_sinkhorn_reward": 0.9758054614067078,
      "rewards-max/forensics_strict_edge_reward": 0.5168293714523315,
      "rewards-max/forensics_tiou_reward": 0.9641169607639313,
      "rewards/forensics_sinkhorn_reward": 0.7232114970684052,
      "rewards/forensics_strict_edge_reward": 0.04955442529171705,
      "rewards/forensics_tiou_reward": 0.5173899829387665,
      "step": 683
    },
    {
      "completion_length": 24.359375,
      "epoch": 2.098310291858679,
      "grad_norm": 11.518294334411621,
      "kl": 0.4052734375,
      "learning_rate": 4.7461538461538456e-07,
      "loss": -0.1152,
      "reward": 0.7594351768493652,
      "reward_std": 0.011158821173012257,
      "reward_std_all": 0.028800971806049347,
      "rewards-max/forensics_sinkhorn_reward": 0.9748018682003021,
      "rewards-max/forensics_strict_edge_reward": 0.5052894204854965,
      "rewards-max/forensics_tiou_reward": 0.9614351093769073,
      "rewards/forensics_sinkhorn_reward": 0.6922366619110107,
      "rewards/forensics_strict_edge_reward": 0.07180510088801384,
      "rewards/forensics_tiou_reward": 0.5168545842170715,
      "step": 684
    },
    {
      "completion_length": 23.015625,
      "epoch": 2.1013824884792625,
      "grad_norm": 11.523262023925781,
      "kl": 0.6953125,
      "learning_rate": 4.738461538461538e-07,
      "loss": -0.1051,
      "reward": 0.7224356830120087,
      "reward_std": 0.02126745507121086,
      "reward_std_all": 0.050203295424580574,
      "rewards-max/forensics_sinkhorn_reward": 0.984619140625,
      "rewards-max/forensics_strict_edge_reward": 0.6420994997024536,
      "rewards-max/forensics_tiou_reward": 0.9885313212871552,
      "rewards/forensics_sinkhorn_reward": 0.6798825562000275,
      "rewards/forensics_strict_edge_reward": 0.0849902331829071,
      "rewards/forensics_tiou_reward": 0.4709038883447647,
      "step": 685
    },
    {
      "completion_length": 24.875,
      "epoch": 2.1044546850998462,
      "grad_norm": 11.473949432373047,
      "kl": 0.505859375,
      "learning_rate": 4.7307692307692304e-07,
      "loss": 0.3336,
      "reward": 0.7697411179542542,
      "reward_std": 0.04645860567688942,
      "reward_std_all": 0.0447880532592535,
      "rewards-max/forensics_sinkhorn_reward": 0.9770529866218567,
      "rewards-max/forensics_strict_edge_reward": 0.3434191644191742,
      "rewards-max/forensics_tiou_reward": 0.9675582349300385,
      "rewards/forensics_sinkhorn_reward": 0.7128873467445374,
      "rewards/forensics_strict_edge_reward": 0.03483485709875822,
      "rewards/forensics_tiou_reward": 0.5256893932819366,
      "step": 686
    },
    {
      "completion_length": 23.8125,
      "epoch": 2.10752688172043,
      "grad_norm": 9.75922679901123,
      "kl": 0.376953125,
      "learning_rate": 4.723076923076923e-07,
      "loss": -0.1482,
      "reward": 0.7614808678627014,
      "reward_std": 0.02382338047027588,
      "reward_std_all": 0.03436243534088135,
      "rewards-max/forensics_sinkhorn_reward": 0.9592724442481995,
      "rewards-max/forensics_strict_edge_reward": 0.30970802903175354,
      "rewards-max/forensics_tiou_reward": 0.9248886108398438,
      "rewards/forensics_sinkhorn_reward": 0.7274251282215118,
      "rewards/forensics_strict_edge_reward": 0.028741289861500263,
      "rewards/forensics_tiou_reward": 0.5477981865406036,
      "step": 687
    },
    {
      "completion_length": 27.015625,
      "epoch": 2.110599078341014,
      "grad_norm": 8.74160099029541,
      "kl": 0.4287109375,
      "learning_rate": 4.715384615384615e-07,
      "loss": 0.1806,
      "reward": 0.7309214472770691,
      "reward_std": 0.048972807824611664,
      "reward_std_all": 0.06202109716832638,
      "rewards-max/forensics_sinkhorn_reward": 0.9709332585334778,
      "rewards-max/forensics_strict_edge_reward": 0.31460704654455185,
      "rewards-max/forensics_tiou_reward": 0.954556941986084,
      "rewards/forensics_sinkhorn_reward": 0.6520479321479797,
      "rewards/forensics_strict_edge_reward": 0.04084168467670679,
      "rewards/forensics_tiou_reward": 0.46238307654857635,
      "step": 688
    },
    {
      "completion_length": 24.140625,
      "epoch": 2.1136712749615976,
      "grad_norm": 11.95111083984375,
      "kl": 0.4501953125,
      "learning_rate": 4.7076923076923073e-07,
      "loss": -0.3309,
      "reward": 0.7064228057861328,
      "reward_std": 0.044167812913656235,
      "reward_std_all": 0.07223577424883842,
      "rewards-max/forensics_sinkhorn_reward": 0.9819552600383759,
      "rewards-max/forensics_strict_edge_reward": 0.7474130392074585,
      "rewards-max/forensics_tiou_reward": 0.9814596772193909,
      "rewards/forensics_sinkhorn_reward": 0.7024372816085815,
      "rewards/forensics_strict_edge_reward": 0.05811312794685364,
      "rewards/forensics_tiou_reward": 0.5235875397920609,
      "step": 689
    },
    {
      "completion_length": 24.328125,
      "epoch": 2.1167434715821813,
      "grad_norm": 10.365182876586914,
      "kl": 0.3642578125,
      "learning_rate": 4.6999999999999995e-07,
      "loss": -0.262,
      "reward": 0.7961583733558655,
      "reward_std": 0.008850174257531762,
      "reward_std_all": 0.010994309326633811,
      "rewards-max/forensics_sinkhorn_reward": 0.9791545271873474,
      "rewards-max/forensics_strict_edge_reward": 0.6456629633903503,
      "rewards-max/forensics_tiou_reward": 0.9721168875694275,
      "rewards/forensics_sinkhorn_reward": 0.7764291167259216,
      "rewards/forensics_strict_edge_reward": 0.07511194981634617,
      "rewards/forensics_tiou_reward": 0.5688748955726624,
      "step": 690
    },
    {
      "completion_length": 22.046875,
      "epoch": 2.119815668202765,
      "grad_norm": 13.09277057647705,
      "kl": 0.6474609375,
      "learning_rate": 4.692307692307692e-07,
      "loss": -0.0495,
      "reward": 0.7929612696170807,
      "reward_std": 0.016952892299741507,
      "reward_std_all": 0.017195909284055233,
      "rewards-max/forensics_sinkhorn_reward": 0.9731916189193726,
      "rewards-max/forensics_strict_edge_reward": 0.5954558104276657,
      "rewards-max/forensics_tiou_reward": 0.960831344127655,
      "rewards/forensics_sinkhorn_reward": 0.7824113070964813,
      "rewards/forensics_strict_edge_reward": 0.09467091597616673,
      "rewards/forensics_tiou_reward": 0.5868170857429504,
      "step": 691
    },
    {
      "completion_length": 25.921875,
      "epoch": 2.122887864823349,
      "grad_norm": 8.190454483032227,
      "kl": 0.4951171875,
      "learning_rate": 4.684615384615385e-07,
      "loss": 0.0224,
      "reward": 0.7207907438278198,
      "reward_std": 0.017867788905277848,
      "reward_std_all": 0.05275881476700306,
      "rewards-max/forensics_sinkhorn_reward": 0.980479896068573,
      "rewards-max/forensics_strict_edge_reward": 0.538355216383934,
      "rewards-max/forensics_tiou_reward": 0.9749569296836853,
      "rewards/forensics_sinkhorn_reward": 0.6546481847763062,
      "rewards/forensics_strict_edge_reward": 0.04964614100754261,
      "rewards/forensics_tiou_reward": 0.453134223818779,
      "step": 692
    },
    {
      "completion_length": 22.109375,
      "epoch": 2.1259600614439322,
      "grad_norm": 9.636356353759766,
      "kl": 0.494140625,
      "learning_rate": 4.6769230769230764e-07,
      "loss": 0.1896,
      "reward": 0.7508397400379181,
      "reward_std": 0.0741011444479227,
      "reward_std_all": 0.06931454688310623,
      "rewards-max/forensics_sinkhorn_reward": 0.9801667928695679,
      "rewards-max/forensics_strict_edge_reward": 0.6149742305278778,
      "rewards-max/forensics_tiou_reward": 0.9775809049606323,
      "rewards/forensics_sinkhorn_reward": 0.7104683816432953,
      "rewards/forensics_strict_edge_reward": 0.06120016612112522,
      "rewards/forensics_tiou_reward": 0.5121873617172241,
      "step": 693
    },
    {
      "completion_length": 26.140625,
      "epoch": 2.129032258064516,
      "grad_norm": 10.987188339233398,
      "kl": 0.4521484375,
      "learning_rate": 4.669230769230769e-07,
      "loss": 0.216,
      "reward": 0.7612160444259644,
      "reward_std": 0.04499215632677078,
      "reward_std_all": 0.047077251598238945,
      "rewards-max/forensics_sinkhorn_reward": 0.8800490498542786,
      "rewards-max/forensics_strict_edge_reward": 0.4109206646680832,
      "rewards-max/forensics_tiou_reward": 0.8364180624485016,
      "rewards/forensics_sinkhorn_reward": 0.6181278228759766,
      "rewards/forensics_strict_edge_reward": 0.03426147811114788,
      "rewards/forensics_tiou_reward": 0.44480475783348083,
      "step": 694
    },
    {
      "completion_length": 23.484375,
      "epoch": 2.1321044546851,
      "grad_norm": 8.43676471710205,
      "kl": 0.470703125,
      "learning_rate": 4.661538461538461e-07,
      "loss": 0.0055,
      "reward": 0.7671032845973969,
      "reward_std": 0.04030764475464821,
      "reward_std_all": 0.0405991580337286,
      "rewards-max/forensics_sinkhorn_reward": 0.9739782512187958,
      "rewards-max/forensics_strict_edge_reward": 0.6799494326114655,
      "rewards-max/forensics_tiou_reward": 0.9622370004653931,
      "rewards/forensics_sinkhorn_reward": 0.7173576951026917,
      "rewards/forensics_strict_edge_reward": 0.08498015441000462,
      "rewards/forensics_tiou_reward": 0.5239707231521606,
      "step": 695
    },
    {
      "completion_length": 25.015625,
      "epoch": 2.1351766513056836,
      "grad_norm": 9.508655548095703,
      "kl": 0.5068359375,
      "learning_rate": 4.653846153846154e-07,
      "loss": 0.0915,
      "reward": 0.7751108407974243,
      "reward_std": 0.02248068992048502,
      "reward_std_all": 0.02729765512049198,
      "rewards-max/forensics_sinkhorn_reward": 0.970711886882782,
      "rewards-max/forensics_strict_edge_reward": 0.4287079870700836,
      "rewards-max/forensics_tiou_reward": 0.9509982764720917,
      "rewards/forensics_sinkhorn_reward": 0.6819362044334412,
      "rewards/forensics_strict_edge_reward": 0.05556747317314148,
      "rewards/forensics_tiou_reward": 0.4720753878355026,
      "step": 696
    },
    {
      "completion_length": 24.875,
      "epoch": 2.1382488479262673,
      "grad_norm": 10.144627571105957,
      "kl": 0.62890625,
      "learning_rate": 4.646153846153846e-07,
      "loss": -0.2052,
      "reward": 0.7829472124576569,
      "reward_std": 0.035294645465910435,
      "reward_std_all": 0.03304215334355831,
      "rewards-max/forensics_sinkhorn_reward": 0.9846201837062836,
      "rewards-max/forensics_strict_edge_reward": 0.7047859728336334,
      "rewards-max/forensics_tiou_reward": 0.98830446600914,
      "rewards/forensics_sinkhorn_reward": 0.7739684879779816,
      "rewards/forensics_strict_edge_reward": 0.087168850004673,
      "rewards/forensics_tiou_reward": 0.5733936131000519,
      "step": 697
    },
    {
      "completion_length": 22.390625,
      "epoch": 2.141321044546851,
      "grad_norm": 10.43372631072998,
      "kl": 0.63671875,
      "learning_rate": 4.638461538461538e-07,
      "loss": 0.1088,
      "reward": 0.7682049572467804,
      "reward_std": 0.05668642371892929,
      "reward_std_all": 0.05216522887349129,
      "rewards-max/forensics_sinkhorn_reward": 0.9584217071533203,
      "rewards-max/forensics_strict_edge_reward": 0.6732408702373505,
      "rewards-max/forensics_tiou_reward": 0.9264744818210602,
      "rewards/forensics_sinkhorn_reward": 0.6889311075210571,
      "rewards/forensics_strict_edge_reward": 0.04077967070043087,
      "rewards/forensics_tiou_reward": 0.5066099464893341,
      "step": 698
    },
    {
      "completion_length": 25.65625,
      "epoch": 2.144393241167435,
      "grad_norm": 8.530420303344727,
      "kl": 0.5283203125,
      "learning_rate": 4.6307692307692303e-07,
      "loss": 0.0131,
      "reward": 0.777293473482132,
      "reward_std": 0.040202898904681206,
      "reward_std_all": 0.03820767626166344,
      "rewards-max/forensics_sinkhorn_reward": 0.9798260927200317,
      "rewards-max/forensics_strict_edge_reward": 0.5984183251857758,
      "rewards-max/forensics_tiou_reward": 0.9749988317489624,
      "rewards/forensics_sinkhorn_reward": 0.7625626623630524,
      "rewards/forensics_strict_edge_reward": 0.0772629976272583,
      "rewards/forensics_tiou_reward": 0.5877911448478699,
      "step": 699
    },
    {
      "completion_length": 23.859375,
      "epoch": 2.1474654377880182,
      "grad_norm": 14.880219459533691,
      "kl": 0.4228515625,
      "learning_rate": 4.623076923076923e-07,
      "loss": 0.0916,
      "reward": 0.7180322110652924,
      "reward_std": 0.04265177808701992,
      "reward_std_all": 0.06291325017809868,
      "rewards-max/forensics_sinkhorn_reward": 0.9756615459918976,
      "rewards-max/forensics_strict_edge_reward": 0.5458240509033203,
      "rewards-max/forensics_tiou_reward": 0.9659587442874908,
      "rewards/forensics_sinkhorn_reward": 0.6379597187042236,
      "rewards/forensics_strict_edge_reward": 0.0670523326843977,
      "rewards/forensics_tiou_reward": 0.4444683790206909,
      "step": 700
    },
    {
      "completion_length": 21.875,
      "epoch": 2.150537634408602,
      "grad_norm": 9.340474128723145,
      "kl": 0.4453125,
      "learning_rate": 4.6153846153846156e-07,
      "loss": 0.0807,
      "reward": 0.7860113978385925,
      "reward_std": 0.02510610595345497,
      "reward_std_all": 0.024716351181268692,
      "rewards-max/forensics_sinkhorn_reward": 0.9861066043376923,
      "rewards-max/forensics_strict_edge_reward": 0.882964164018631,
      "rewards-max/forensics_tiou_reward": 0.9923409223556519,
      "rewards/forensics_sinkhorn_reward": 0.7603214979171753,
      "rewards/forensics_strict_edge_reward": 0.12381044030189514,
      "rewards/forensics_tiou_reward": 0.5560940802097321,
      "step": 701
    },
    {
      "completion_length": 28.390625,
      "epoch": 2.153609831029186,
      "grad_norm": 9.572286605834961,
      "kl": 0.4033203125,
      "learning_rate": 4.607692307692307e-07,
      "loss": 0.1893,
      "reward": 0.7259372472763062,
      "reward_std": 0.06775877252221107,
      "reward_std_all": 0.0742133017629385,
      "rewards-max/forensics_sinkhorn_reward": 0.9807775914669037,
      "rewards-max/forensics_strict_edge_reward": 0.5071596875786781,
      "rewards-max/forensics_tiou_reward": 0.9789898991584778,
      "rewards/forensics_sinkhorn_reward": 0.662551999092102,
      "rewards/forensics_strict_edge_reward": 0.0679379552602768,
      "rewards/forensics_tiou_reward": 0.5414798706769943,
      "step": 702
    },
    {
      "completion_length": 28.15625,
      "epoch": 2.1566820276497696,
      "grad_norm": 9.496773719787598,
      "kl": 0.458984375,
      "learning_rate": 4.6e-07,
      "loss": -0.1259,
      "reward": 0.7540884613990784,
      "reward_std": 0.04362742695957422,
      "reward_std_all": 0.05052410252392292,
      "rewards-max/forensics_sinkhorn_reward": 0.9704682528972626,
      "rewards-max/forensics_strict_edge_reward": 0.3159602954983711,
      "rewards-max/forensics_tiou_reward": 0.95252725481987,
      "rewards/forensics_sinkhorn_reward": 0.6709045469760895,
      "rewards/forensics_strict_edge_reward": 0.030676299706101418,
      "rewards/forensics_tiou_reward": 0.46033646166324615,
      "step": 703
    },
    {
      "completion_length": 25.703125,
      "epoch": 2.1597542242703534,
      "grad_norm": 8.924802780151367,
      "kl": 0.458984375,
      "learning_rate": 4.592307692307692e-07,
      "loss": 0.3079,
      "reward": 0.7448606789112091,
      "reward_std": 0.052837446331977844,
      "reward_std_all": 0.056920696049928665,
      "rewards-max/forensics_sinkhorn_reward": 0.988029956817627,
      "rewards-max/forensics_strict_edge_reward": 0.7933230996131897,
      "rewards-max/forensics_tiou_reward": 0.9969565868377686,
      "rewards/forensics_sinkhorn_reward": 0.6529901921749115,
      "rewards/forensics_strict_edge_reward": 0.06129937432706356,
      "rewards/forensics_tiou_reward": 0.4989579766988754,
      "step": 704
    },
    {
      "completion_length": 23.84375,
      "epoch": 2.162826420890937,
      "grad_norm": 7.797768592834473,
      "kl": 0.4453125,
      "learning_rate": 4.5846153846153847e-07,
      "loss": 0.1979,
      "reward": 0.761246532201767,
      "reward_std": 0.06691469065845013,
      "reward_std_all": 0.06078128516674042,
      "rewards-max/forensics_sinkhorn_reward": 0.9805806279182434,
      "rewards-max/forensics_strict_edge_reward": 0.5308436453342438,
      "rewards-max/forensics_tiou_reward": 0.9761548042297363,
      "rewards/forensics_sinkhorn_reward": 0.7219172716140747,
      "rewards/forensics_strict_edge_reward": 0.09046000987291336,
      "rewards/forensics_tiou_reward": 0.5781587362289429,
      "step": 705
    },
    {
      "completion_length": 25.09375,
      "epoch": 2.165898617511521,
      "grad_norm": 8.167438507080078,
      "kl": 0.5224609375,
      "learning_rate": 4.5769230769230763e-07,
      "loss": 0.3004,
      "reward": 0.7195405960083008,
      "reward_std": 0.09021671861410141,
      "reward_std_all": 0.09094734489917755,
      "rewards-max/forensics_sinkhorn_reward": 0.9846095740795135,
      "rewards-max/forensics_strict_edge_reward": 0.6934256851673126,
      "rewards-max/forensics_tiou_reward": 0.987191915512085,
      "rewards/forensics_sinkhorn_reward": 0.6617989838123322,
      "rewards/forensics_strict_edge_reward": 0.07769710198044777,
      "rewards/forensics_tiou_reward": 0.4865823984146118,
      "step": 706
    },
    {
      "completion_length": 23.859375,
      "epoch": 2.1689708141321047,
      "grad_norm": 8.650129318237305,
      "kl": 0.3564453125,
      "learning_rate": 4.569230769230769e-07,
      "loss": -0.1203,
      "reward": 0.7768633961677551,
      "reward_std": 0.03350404277443886,
      "reward_std_all": 0.03336010314524174,
      "rewards-max/forensics_sinkhorn_reward": 0.9823057949542999,
      "rewards-max/forensics_strict_edge_reward": 0.6408714950084686,
      "rewards-max/forensics_tiou_reward": 0.9780605137348175,
      "rewards/forensics_sinkhorn_reward": 0.7512198090553284,
      "rewards/forensics_strict_edge_reward": 0.0865611582994461,
      "rewards/forensics_tiou_reward": 0.611767053604126,
      "step": 707
    },
    {
      "completion_length": 28.4375,
      "epoch": 2.172043010752688,
      "grad_norm": 8.292681694030762,
      "kl": 0.4580078125,
      "learning_rate": 4.561538461538461e-07,
      "loss": 0.1102,
      "reward": 0.7699525654315948,
      "reward_std": 0.03231258690357208,
      "reward_std_all": 0.036008622497320175,
      "rewards-max/forensics_sinkhorn_reward": 0.9746558964252472,
      "rewards-max/forensics_strict_edge_reward": 0.3045279011130333,
      "rewards-max/forensics_tiou_reward": 0.9603826105594635,
      "rewards/forensics_sinkhorn_reward": 0.6614250540733337,
      "rewards/forensics_strict_edge_reward": 0.039951346814632416,
      "rewards/forensics_tiou_reward": 0.540543407201767,
      "step": 708
    },
    {
      "completion_length": 26.515625,
      "epoch": 2.175115207373272,
      "grad_norm": 11.169523239135742,
      "kl": 0.486328125,
      "learning_rate": 4.553846153846154e-07,
      "loss": 0.086,
      "reward": 0.7686682641506195,
      "reward_std": 0.03644043020904064,
      "reward_std_all": 0.03870330564677715,
      "rewards-max/forensics_sinkhorn_reward": 0.9275712072849274,
      "rewards-max/forensics_strict_edge_reward": 0.23528613150119781,
      "rewards-max/forensics_tiou_reward": 0.8503296375274658,
      "rewards/forensics_sinkhorn_reward": 0.7107438743114471,
      "rewards/forensics_strict_edge_reward": 0.03068972541950643,
      "rewards/forensics_tiou_reward": 0.539794310927391,
      "step": 709
    },
    {
      "completion_length": 25.3125,
      "epoch": 2.1781874039938556,
      "grad_norm": 15.361620903015137,
      "kl": 0.3759765625,
      "learning_rate": 4.546153846153846e-07,
      "loss": -0.0399,
      "reward": 0.7487713098526001,
      "reward_std": 0.03229503985494375,
      "reward_std_all": 0.044694872573018074,
      "rewards-max/forensics_sinkhorn_reward": 0.9825847744941711,
      "rewards-max/forensics_strict_edge_reward": 0.709127277135849,
      "rewards-max/forensics_tiou_reward": 0.9831078350543976,
      "rewards/forensics_sinkhorn_reward": 0.671698272228241,
      "rewards/forensics_strict_edge_reward": 0.08730534464120865,
      "rewards/forensics_tiou_reward": 0.5164456367492676,
      "step": 710
    },
    {
      "completion_length": 25.34375,
      "epoch": 2.1812596006144394,
      "grad_norm": 8.735158920288086,
      "kl": 0.4267578125,
      "learning_rate": 4.538461538461538e-07,
      "loss": 0.0329,
      "reward": 0.7601158022880554,
      "reward_std": 0.04775994084775448,
      "reward_std_all": 0.04831390269100666,
      "rewards-max/forensics_sinkhorn_reward": 0.9743960797786713,
      "rewards-max/forensics_strict_edge_reward": 0.4833206683397293,
      "rewards-max/forensics_tiou_reward": 0.9629499316215515,
      "rewards/forensics_sinkhorn_reward": 0.7265999913215637,
      "rewards/forensics_strict_edge_reward": 0.10409770160913467,
      "rewards/forensics_tiou_reward": 0.589234322309494,
      "step": 711
    },
    {
      "completion_length": 24.890625,
      "epoch": 2.184331797235023,
      "grad_norm": 12.359153747558594,
      "kl": 0.705078125,
      "learning_rate": 4.5307692307692307e-07,
      "loss": 0.0625,
      "reward": 0.7259060442447662,
      "reward_std": 0.052428700029850006,
      "reward_std_all": 0.06992959603667259,
      "rewards-max/forensics_sinkhorn_reward": 0.9719498753547668,
      "rewards-max/forensics_strict_edge_reward": 0.6168365329504013,
      "rewards-max/forensics_tiou_reward": 0.9591512680053711,
      "rewards/forensics_sinkhorn_reward": 0.6886754631996155,
      "rewards/forensics_strict_edge_reward": 0.07103481702506542,
      "rewards/forensics_tiou_reward": 0.482571005821228,
      "step": 712
    },
    {
      "completion_length": 20.625,
      "epoch": 2.187403993855607,
      "grad_norm": 11.386198997497559,
      "kl": 0.546875,
      "learning_rate": 4.523076923076923e-07,
      "loss": 0.0168,
      "reward": 0.7576905786991119,
      "reward_std": 0.05635101534426212,
      "reward_std_all": 0.05410696566104889,
      "rewards-max/forensics_sinkhorn_reward": 0.9821040630340576,
      "rewards-max/forensics_strict_edge_reward": 0.8187680542469025,
      "rewards-max/forensics_tiou_reward": 0.9834510087966919,
      "rewards/forensics_sinkhorn_reward": 0.7347280085086823,
      "rewards/forensics_strict_edge_reward": 0.11513106152415276,
      "rewards/forensics_tiou_reward": 0.5056071132421494,
      "step": 713
    },
    {
      "completion_length": 26.015625,
      "epoch": 2.1904761904761907,
      "grad_norm": 9.32951545715332,
      "kl": 0.556640625,
      "learning_rate": 4.5153846153846155e-07,
      "loss": -0.0198,
      "reward": 0.7760668396949768,
      "reward_std": 0.040649767383001745,
      "reward_std_all": 0.03851220873184502,
      "rewards-max/forensics_sinkhorn_reward": 0.9839105308055878,
      "rewards-max/forensics_strict_edge_reward": 0.7533732652664185,
      "rewards-max/forensics_tiou_reward": 0.9848350882530212,
      "rewards/forensics_sinkhorn_reward": 0.7195973694324493,
      "rewards/forensics_strict_edge_reward": 0.10987839102745056,
      "rewards/forensics_tiou_reward": 0.5858187675476074,
      "step": 714
    },
    {
      "completion_length": 23.984375,
      "epoch": 2.193548387096774,
      "grad_norm": 11.642987251281738,
      "kl": 0.5068359375,
      "learning_rate": 4.507692307692307e-07,
      "loss": -0.2425,
      "reward": 0.7605946958065033,
      "reward_std": 0.05931730195879936,
      "reward_std_all": 0.05559181421995163,
      "rewards-max/forensics_sinkhorn_reward": 0.9695707559585571,
      "rewards-max/forensics_strict_edge_reward": 0.5680216699838638,
      "rewards-max/forensics_tiou_reward": 0.9526011347770691,
      "rewards/forensics_sinkhorn_reward": 0.6927456855773926,
      "rewards/forensics_strict_edge_reward": 0.12573165073990822,
      "rewards/forensics_tiou_reward": 0.5124794542789459,
      "step": 715
    },
    {
      "completion_length": 26.90625,
      "epoch": 2.196620583717358,
      "grad_norm": 10.423253059387207,
      "kl": 0.8271484375,
      "learning_rate": 4.5e-07,
      "loss": 0.1251,
      "reward": 0.776750385761261,
      "reward_std": 0.04004458570852876,
      "reward_std_all": 0.037906312849372625,
      "rewards-max/forensics_sinkhorn_reward": 0.9660995006561279,
      "rewards-max/forensics_strict_edge_reward": 0.29590949416160583,
      "rewards-max/forensics_tiou_reward": 0.9261957108974457,
      "rewards/forensics_sinkhorn_reward": 0.658606767654419,
      "rewards/forensics_strict_edge_reward": 0.0459944186732173,
      "rewards/forensics_tiou_reward": 0.5158311575651169,
      "step": 716
    },
    {
      "completion_length": 28.421875,
      "epoch": 2.1996927803379416,
      "grad_norm": 12.982059478759766,
      "kl": 1.0029296875,
      "learning_rate": 4.492307692307692e-07,
      "loss": -0.0332,
      "reward": 0.7759573757648468,
      "reward_std": 0.028866860084235668,
      "reward_std_all": 0.03049489203840494,
      "rewards-max/forensics_sinkhorn_reward": 0.9874328076839447,
      "rewards-max/forensics_strict_edge_reward": 0.6944838762283325,
      "rewards-max/forensics_tiou_reward": 0.9963929653167725,
      "rewards/forensics_sinkhorn_reward": 0.7192754745483398,
      "rewards/forensics_strict_edge_reward": 0.09240962564945221,
      "rewards/forensics_tiou_reward": 0.596410721540451,
      "step": 717
    },
    {
      "completion_length": 24.015625,
      "epoch": 2.2027649769585254,
      "grad_norm": 11.601127624511719,
      "kl": 0.3994140625,
      "learning_rate": 4.4846153846153846e-07,
      "loss": 0.0383,
      "reward": 0.7367661595344543,
      "reward_std": 0.028297673910856247,
      "reward_std_all": 0.0501142218708992,
      "rewards-max/forensics_sinkhorn_reward": 0.9677209258079529,
      "rewards-max/forensics_strict_edge_reward": 0.38874343037605286,
      "rewards-max/forensics_tiou_reward": 0.9457152187824249,
      "rewards/forensics_sinkhorn_reward": 0.6942636668682098,
      "rewards/forensics_strict_edge_reward": 0.04187672212719917,
      "rewards/forensics_tiou_reward": 0.4771801829338074,
      "step": 718
    },
    {
      "completion_length": 22.859375,
      "epoch": 2.205837173579109,
      "grad_norm": 10.657294273376465,
      "kl": 0.728515625,
      "learning_rate": 4.476923076923076e-07,
      "loss": 0.1159,
      "reward": 0.7951069474220276,
      "reward_std": 0.01434423215687275,
      "reward_std_all": 0.014949996024370193,
      "rewards-max/forensics_sinkhorn_reward": 0.9785924553871155,
      "rewards-max/forensics_strict_edge_reward": 0.5322106778621674,
      "rewards-max/forensics_tiou_reward": 0.9718333780765533,
      "rewards/forensics_sinkhorn_reward": 0.766506165266037,
      "rewards/forensics_strict_edge_reward": 0.07952200248837471,
      "rewards/forensics_tiou_reward": 0.5738574862480164,
      "step": 719
    },
    {
      "completion_length": 24.4375,
      "epoch": 2.208909370199693,
      "grad_norm": 7.909279823303223,
      "kl": 0.548828125,
      "learning_rate": 4.469230769230769e-07,
      "loss": 0.0213,
      "reward": 0.7931223213672638,
      "reward_std": 0.022536814212799072,
      "reward_std_all": 0.02160835452377796,
      "rewards-max/forensics_sinkhorn_reward": 0.9860227108001709,
      "rewards-max/forensics_strict_edge_reward": 0.5755647718906403,
      "rewards-max/forensics_tiou_reward": 0.9914596080780029,
      "rewards/forensics_sinkhorn_reward": 0.8047694861888885,
      "rewards/forensics_strict_edge_reward": 0.05121711269021034,
      "rewards/forensics_tiou_reward": 0.643237441778183,
      "step": 720
    },
    {
      "completion_length": 23.828125,
      "epoch": 2.2119815668202767,
      "grad_norm": 11.239764213562012,
      "kl": 0.396484375,
      "learning_rate": 4.4615384615384615e-07,
      "loss": -0.066,
      "reward": 0.7884197533130646,
      "reward_std": 0.02697950741276145,
      "reward_std_all": 0.025440256111323833,
      "rewards-max/forensics_sinkhorn_reward": 0.9889904856681824,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.7883268892765045,
      "rewards/forensics_strict_edge_reward": 0.1747564822435379,
      "rewards/forensics_tiou_reward": 0.6621358096599579,
      "step": 721
    },
    {
      "completion_length": 25.8125,
      "epoch": 2.21505376344086,
      "grad_norm": 9.030620574951172,
      "kl": 0.5078125,
      "learning_rate": 4.4538461538461537e-07,
      "loss": 0.4453,
      "reward": 0.7629390954971313,
      "reward_std": 0.03529119258746505,
      "reward_std_all": 0.04106130264699459,
      "rewards-max/forensics_sinkhorn_reward": 0.9837594330310822,
      "rewards-max/forensics_strict_edge_reward": 0.6789796948432922,
      "rewards-max/forensics_tiou_reward": 0.9865724444389343,
      "rewards/forensics_sinkhorn_reward": 0.6843457818031311,
      "rewards/forensics_strict_edge_reward": 0.08765525743365288,
      "rewards/forensics_tiou_reward": 0.5705370604991913,
      "step": 722
    },
    {
      "completion_length": 23.875,
      "epoch": 2.218125960061444,
      "grad_norm": 12.488584518432617,
      "kl": 0.578125,
      "learning_rate": 4.4461538461538463e-07,
      "loss": 0.4163,
      "reward": 0.7457166910171509,
      "reward_std": 0.0484086312353611,
      "reward_std_all": 0.0569739006459713,
      "rewards-max/forensics_sinkhorn_reward": 0.9810453653335571,
      "rewards-max/forensics_strict_edge_reward": 0.5731385350227356,
      "rewards-max/forensics_tiou_reward": 0.9769717156887054,
      "rewards/forensics_sinkhorn_reward": 0.6723307073116302,
      "rewards/forensics_strict_edge_reward": 0.06293688341975212,
      "rewards/forensics_tiou_reward": 0.4205082952976227,
      "step": 723
    },
    {
      "completion_length": 24.71875,
      "epoch": 2.2211981566820276,
      "grad_norm": 11.483543395996094,
      "kl": 0.748046875,
      "learning_rate": 4.438461538461538e-07,
      "loss": 0.1653,
      "reward": 0.7847300171852112,
      "reward_std": 0.028812464326620102,
      "reward_std_all": 0.02836239244788885,
      "rewards-max/forensics_sinkhorn_reward": 0.9597552120685577,
      "rewards-max/forensics_strict_edge_reward": 0.4589299261569977,
      "rewards-max/forensics_tiou_reward": 0.9292121827602386,
      "rewards/forensics_sinkhorn_reward": 0.6994362473487854,
      "rewards/forensics_strict_edge_reward": 0.05393584072589874,
      "rewards/forensics_tiou_reward": 0.5597751438617706,
      "step": 724
    },
    {
      "completion_length": 27.890625,
      "epoch": 2.2242703533026114,
      "grad_norm": 8.249343872070312,
      "kl": 0.38671875,
      "learning_rate": 4.4307692307692306e-07,
      "loss": -0.1476,
      "reward": 0.7421686053276062,
      "reward_std": 0.025723755359649658,
      "reward_std_all": 0.04363525100052357,
      "rewards-max/forensics_sinkhorn_reward": 0.9823082983493805,
      "rewards-max/forensics_strict_edge_reward": 0.6035806834697723,
      "rewards-max/forensics_tiou_reward": 0.9827800989151001,
      "rewards/forensics_sinkhorn_reward": 0.6918683052062988,
      "rewards/forensics_strict_edge_reward": 0.08164768293499947,
      "rewards/forensics_tiou_reward": 0.5521611869335175,
      "step": 725
    },
    {
      "completion_length": 23.5625,
      "epoch": 2.227342549923195,
      "grad_norm": 10.941254615783691,
      "kl": 0.8583984375,
      "learning_rate": 4.423076923076923e-07,
      "loss": 0.0168,
      "reward": 0.7883410751819611,
      "reward_std": 0.016999084036797285,
      "reward_std_all": 0.019155094400048256,
      "rewards-max/forensics_sinkhorn_reward": 0.9852037131786346,
      "rewards-max/forensics_strict_edge_reward": 0.7542104423046112,
      "rewards-max/forensics_tiou_reward": 0.9894362688064575,
      "rewards/forensics_sinkhorn_reward": 0.777229905128479,
      "rewards/forensics_strict_edge_reward": 0.08271397650241852,
      "rewards/forensics_tiou_reward": 0.5982456803321838,
      "step": 726
    },
    {
      "completion_length": 25.5625,
      "epoch": 2.230414746543779,
      "grad_norm": 16.68073272705078,
      "kl": 0.4482421875,
      "learning_rate": 4.4153846153846154e-07,
      "loss": 0.3695,
      "reward": 0.7380048930644989,
      "reward_std": 0.053804339841008186,
      "reward_std_all": 0.062001269310712814,
      "rewards-max/forensics_sinkhorn_reward": 0.9842748641967773,
      "rewards-max/forensics_strict_edge_reward": 0.625903993844986,
      "rewards-max/forensics_tiou_reward": 0.9830868244171143,
      "rewards/forensics_sinkhorn_reward": 0.6726308465003967,
      "rewards/forensics_strict_edge_reward": 0.08782286010682583,
      "rewards/forensics_tiou_reward": 0.5243059992790222,
      "step": 727
    },
    {
      "completion_length": 23.265625,
      "epoch": 2.2334869431643627,
      "grad_norm": 11.100578308105469,
      "kl": 0.458984375,
      "learning_rate": 4.407692307692307e-07,
      "loss": 0.0138,
      "reward": 0.7395652532577515,
      "reward_std": 0.05278707295656204,
      "reward_std_all": 0.058411287143826485,
      "rewards-max/forensics_sinkhorn_reward": 0.9822669923305511,
      "rewards-max/forensics_strict_edge_reward": 0.6352442353963852,
      "rewards-max/forensics_tiou_reward": 0.976190596818924,
      "rewards/forensics_sinkhorn_reward": 0.6633268296718597,
      "rewards/forensics_strict_edge_reward": 0.045784697867929935,
      "rewards/forensics_tiou_reward": 0.42811135947704315,
      "step": 728
    },
    {
      "completion_length": 25.484375,
      "epoch": 2.236559139784946,
      "grad_norm": 11.054034233093262,
      "kl": 0.666015625,
      "learning_rate": 4.3999999999999997e-07,
      "loss": 0.0253,
      "reward": 0.7671032845973969,
      "reward_std": 0.03420219849795103,
      "reward_std_all": 0.03627857565879822,
      "rewards-max/forensics_sinkhorn_reward": 0.9706676602363586,
      "rewards-max/forensics_strict_edge_reward": 0.6575653851032257,
      "rewards-max/forensics_tiou_reward": 0.9535880088806152,
      "rewards/forensics_sinkhorn_reward": 0.6389342248439789,
      "rewards/forensics_strict_edge_reward": 0.08668273314833641,
      "rewards/forensics_tiou_reward": 0.5008189976215363,
      "step": 729
    },
    {
      "completion_length": 27.890625,
      "epoch": 2.23963133640553,
      "grad_norm": 8.130170822143555,
      "kl": 0.5,
      "learning_rate": 4.3923076923076924e-07,
      "loss": -0.0894,
      "reward": 0.7901804447174072,
      "reward_std": 0.02558869682252407,
      "reward_std_all": 0.024919189512729645,
      "rewards-max/forensics_sinkhorn_reward": 0.9764063358306885,
      "rewards-max/forensics_strict_edge_reward": 0.4591679126024246,
      "rewards-max/forensics_tiou_reward": 0.9676614105701447,
      "rewards/forensics_sinkhorn_reward": 0.7465067207813263,
      "rewards/forensics_strict_edge_reward": 0.06247681751847267,
      "rewards/forensics_tiou_reward": 0.5161884129047394,
      "step": 730
    },
    {
      "completion_length": 23.1875,
      "epoch": 2.2427035330261136,
      "grad_norm": 11.247013092041016,
      "kl": 0.4716796875,
      "learning_rate": 4.3846153846153845e-07,
      "loss": -0.2306,
      "reward": 0.7994395196437836,
      "reward_std": 0.007850904483348131,
      "reward_std_all": 0.009559475351125002,
      "rewards-max/forensics_sinkhorn_reward": 0.9774641692638397,
      "rewards-max/forensics_strict_edge_reward": 0.5776711404323578,
      "rewards-max/forensics_tiou_reward": 0.9722424447536469,
      "rewards/forensics_sinkhorn_reward": 0.7698494493961334,
      "rewards/forensics_strict_edge_reward": 0.09180417284369469,
      "rewards/forensics_tiou_reward": 0.5569977164268494,
      "step": 731
    },
    {
      "completion_length": 25.1875,
      "epoch": 2.2457757296466974,
      "grad_norm": 9.708617210388184,
      "kl": 0.474609375,
      "learning_rate": 4.376923076923077e-07,
      "loss": 0.0012,
      "reward": 0.7433944642543793,
      "reward_std": 0.04083417955553159,
      "reward_std_all": 0.0493240513605997,
      "rewards-max/forensics_sinkhorn_reward": 0.9750188291072845,
      "rewards-max/forensics_strict_edge_reward": 0.6083320677280426,
      "rewards-max/forensics_tiou_reward": 0.9610786437988281,
      "rewards/forensics_sinkhorn_reward": 0.7216299176216125,
      "rewards/forensics_strict_edge_reward": 0.05940689146518707,
      "rewards/forensics_tiou_reward": 0.5346309393644333,
      "step": 732
    },
    {
      "completion_length": 26.203125,
      "epoch": 2.248847926267281,
      "grad_norm": 10.59656047821045,
      "kl": 0.5107421875,
      "learning_rate": 4.369230769230769e-07,
      "loss": 0.0202,
      "reward": 0.7567912340164185,
      "reward_std": 0.02198181487619877,
      "reward_std_all": 0.035741278901696205,
      "rewards-max/forensics_sinkhorn_reward": 0.9803148508071899,
      "rewards-max/forensics_strict_edge_reward": 0.482917919754982,
      "rewards-max/forensics_tiou_reward": 0.9750872254371643,
      "rewards/forensics_sinkhorn_reward": 0.684937983751297,
      "rewards/forensics_strict_edge_reward": 0.04990745149552822,
      "rewards/forensics_tiou_reward": 0.47344332933425903,
      "step": 733
    },
    {
      "completion_length": 26.578125,
      "epoch": 2.251920122887865,
      "grad_norm": 9.928791999816895,
      "kl": 0.51953125,
      "learning_rate": 4.3615384615384614e-07,
      "loss": 0.1248,
      "reward": 0.7587080299854279,
      "reward_std": 0.0373055343516171,
      "reward_std_all": 0.044087911024689674,
      "rewards-max/forensics_sinkhorn_reward": 0.9793419241905212,
      "rewards-max/forensics_strict_edge_reward": 0.49768924713134766,
      "rewards-max/forensics_tiou_reward": 0.9734764695167542,
      "rewards/forensics_sinkhorn_reward": 0.6488706171512604,
      "rewards/forensics_strict_edge_reward": 0.08836853876709938,
      "rewards/forensics_tiou_reward": 0.5063779652118683,
      "step": 734
    },
    {
      "completion_length": 26.171875,
      "epoch": 2.2549923195084487,
      "grad_norm": 9.51589584350586,
      "kl": 0.64453125,
      "learning_rate": 4.3538461538461536e-07,
      "loss": 0.1861,
      "reward": 0.7452628016471863,
      "reward_std": 0.048853855580091476,
      "reward_std_all": 0.05322292447090149,
      "rewards-max/forensics_sinkhorn_reward": 0.9824167490005493,
      "rewards-max/forensics_strict_edge_reward": 0.4493289589881897,
      "rewards-max/forensics_tiou_reward": 0.984302282333374,
      "rewards/forensics_sinkhorn_reward": 0.6403367519378662,
      "rewards/forensics_strict_edge_reward": 0.0503233028575778,
      "rewards/forensics_tiou_reward": 0.46241725981235504,
      "step": 735
    },
    {
      "completion_length": 29.546875,
      "epoch": 2.258064516129032,
      "grad_norm": 8.754242897033691,
      "kl": 0.712890625,
      "learning_rate": 4.346153846153846e-07,
      "loss": -0.0732,
      "reward": 0.7799818217754364,
      "reward_std": 0.02922158781439066,
      "reward_std_all": 0.029397962614893913,
      "rewards-max/forensics_sinkhorn_reward": 0.9676986634731293,
      "rewards-max/forensics_strict_edge_reward": 0.5629483610391617,
      "rewards-max/forensics_tiou_reward": 0.9388181567192078,
      "rewards/forensics_sinkhorn_reward": 0.7608563303947449,
      "rewards/forensics_strict_edge_reward": 0.055765691213309765,
      "rewards/forensics_tiou_reward": 0.6079974472522736,
      "step": 736
    },
    {
      "completion_length": 26.46875,
      "epoch": 2.261136712749616,
      "grad_norm": 7.689254283905029,
      "kl": 0.5595703125,
      "learning_rate": 4.338461538461538e-07,
      "loss": -0.0267,
      "reward": 0.7664250433444977,
      "reward_std": 0.022462680004537106,
      "reward_std_all": 0.03372129797935486,
      "rewards-max/forensics_sinkhorn_reward": 0.9797099232673645,
      "rewards-max/forensics_strict_edge_reward": 0.5690367668867111,
      "rewards-max/forensics_tiou_reward": 0.9736348986625671,
      "rewards/forensics_sinkhorn_reward": 0.7427870035171509,
      "rewards/forensics_strict_edge_reward": 0.07846855372190475,
      "rewards/forensics_tiou_reward": 0.6072455644607544,
      "step": 737
    },
    {
      "completion_length": 23.75,
      "epoch": 2.2642089093701996,
      "grad_norm": 10.384353637695312,
      "kl": 0.658203125,
      "learning_rate": 4.3307692307692305e-07,
      "loss": 0.2035,
      "reward": 0.7689789235591888,
      "reward_std": 0.041913808323442936,
      "reward_std_all": 0.04129341244697571,
      "rewards-max/forensics_sinkhorn_reward": 0.9793766438961029,
      "rewards-max/forensics_strict_edge_reward": 0.8135445415973663,
      "rewards-max/forensics_tiou_reward": 0.9777216911315918,
      "rewards/forensics_sinkhorn_reward": 0.720708429813385,
      "rewards/forensics_strict_edge_reward": 0.08154206350445747,
      "rewards/forensics_tiou_reward": 0.4940827190876007,
      "step": 738
    },
    {
      "completion_length": 23.59375,
      "epoch": 2.2672811059907834,
      "grad_norm": 11.485599517822266,
      "kl": 0.677734375,
      "learning_rate": 4.3230769230769227e-07,
      "loss": 0.1421,
      "reward": 0.7613239586353302,
      "reward_std": 0.03717127814888954,
      "reward_std_all": 0.0404552835971117,
      "rewards-max/forensics_sinkhorn_reward": 0.9849717915058136,
      "rewards-max/forensics_strict_edge_reward": 0.8108834326267242,
      "rewards-max/forensics_tiou_reward": 0.9905534982681274,
      "rewards/forensics_sinkhorn_reward": 0.6564265191555023,
      "rewards/forensics_strict_edge_reward": 0.11350843124091625,
      "rewards/forensics_tiou_reward": 0.4609441012144089,
      "step": 739
    },
    {
      "completion_length": 26.234375,
      "epoch": 2.270353302611367,
      "grad_norm": 8.190691947937012,
      "kl": 0.435546875,
      "learning_rate": 4.3153846153846153e-07,
      "loss": 0.0752,
      "reward": 0.7576558887958527,
      "reward_std": 0.04655821807682514,
      "reward_std_all": 0.04922317899763584,
      "rewards-max/forensics_sinkhorn_reward": 0.9670135676860809,
      "rewards-max/forensics_strict_edge_reward": 0.5333003252744675,
      "rewards-max/forensics_tiou_reward": 0.9467412233352661,
      "rewards/forensics_sinkhorn_reward": 0.6373725533485413,
      "rewards/forensics_strict_edge_reward": 0.039598140865564346,
      "rewards/forensics_tiou_reward": 0.3896615207195282,
      "step": 740
    },
    {
      "completion_length": 26.25,
      "epoch": 2.273425499231951,
      "grad_norm": 10.95861530303955,
      "kl": 0.646484375,
      "learning_rate": 4.307692307692308e-07,
      "loss": 0.1029,
      "reward": 0.7890398800373077,
      "reward_std": 0.007832844508811831,
      "reward_std_all": 0.01404433511197567,
      "rewards-max/forensics_sinkhorn_reward": 0.9717181622982025,
      "rewards-max/forensics_strict_edge_reward": 0.4626377075910568,
      "rewards-max/forensics_tiou_reward": 0.9532730579376221,
      "rewards/forensics_sinkhorn_reward": 0.6724022030830383,
      "rewards/forensics_strict_edge_reward": 0.04651689529418945,
      "rewards/forensics_tiou_reward": 0.5166154503822327,
      "step": 741
    },
    {
      "completion_length": 23.1875,
      "epoch": 2.2764976958525347,
      "grad_norm": 9.461380958557129,
      "kl": 0.607421875,
      "learning_rate": 4.2999999999999996e-07,
      "loss": 0.2577,
      "reward": 0.7130581140518188,
      "reward_std": 0.10003451630473137,
      "reward_std_all": 0.09662522375583649,
      "rewards-max/forensics_sinkhorn_reward": 0.9762349724769592,
      "rewards-max/forensics_strict_edge_reward": 0.24938895553350449,
      "rewards-max/forensics_tiou_reward": 0.968731552362442,
      "rewards/forensics_sinkhorn_reward": 0.6903720200061798,
      "rewards/forensics_strict_edge_reward": 0.03811698779463768,
      "rewards/forensics_tiou_reward": 0.5002108514308929,
      "step": 742
    },
    {
      "completion_length": 26.015625,
      "epoch": 2.279569892473118,
      "grad_norm": 10.083898544311523,
      "kl": 0.580078125,
      "learning_rate": 4.292307692307692e-07,
      "loss": -0.2391,
      "reward": 0.7907600402832031,
      "reward_std": 0.02154557593166828,
      "reward_std_all": 0.02114716451615095,
      "rewards-max/forensics_sinkhorn_reward": 0.9855741858482361,
      "rewards-max/forensics_strict_edge_reward": 0.7309097647666931,
      "rewards-max/forensics_tiou_reward": 0.9886184334754944,
      "rewards/forensics_sinkhorn_reward": 0.7713192701339722,
      "rewards/forensics_strict_edge_reward": 0.10395770892500877,
      "rewards/forensics_tiou_reward": 0.6037168800830841,
      "step": 743
    },
    {
      "completion_length": 25.21875,
      "epoch": 2.282642089093702,
      "grad_norm": 10.307121276855469,
      "kl": 0.5224609375,
      "learning_rate": 4.2846153846153844e-07,
      "loss": 0.3803,
      "reward": 0.7016094326972961,
      "reward_std": 0.051508307456970215,
      "reward_std_all": 0.07370436191558838,
      "rewards-max/forensics_sinkhorn_reward": 0.9697422683238983,
      "rewards-max/forensics_strict_edge_reward": 0.5371811389923096,
      "rewards-max/forensics_tiou_reward": 0.9521774053573608,
      "rewards/forensics_sinkhorn_reward": 0.6165651977062225,
      "rewards/forensics_strict_edge_reward": 0.06568105332553387,
      "rewards/forensics_tiou_reward": 0.43379856646060944,
      "step": 744
    },
    {
      "completion_length": 25.453125,
      "epoch": 2.2857142857142856,
      "grad_norm": 10.982922554016113,
      "kl": 0.4248046875,
      "learning_rate": 4.276923076923077e-07,
      "loss": 0.2735,
      "reward": 0.7466784417629242,
      "reward_std": 0.02272575395181775,
      "reward_std_all": 0.040226236917078495,
      "rewards-max/forensics_sinkhorn_reward": 0.966562807559967,
      "rewards-max/forensics_strict_edge_reward": 0.40120235085487366,
      "rewards-max/forensics_tiou_reward": 0.9468941688537598,
      "rewards/forensics_sinkhorn_reward": 0.630784809589386,
      "rewards/forensics_strict_edge_reward": 0.034177946858108044,
      "rewards/forensics_tiou_reward": 0.45420652627944946,
      "step": 745
    },
    {
      "completion_length": 24.0625,
      "epoch": 2.2887864823348694,
      "grad_norm": 8.859371185302734,
      "kl": 0.6318359375,
      "learning_rate": 4.2692307692307687e-07,
      "loss": -0.0766,
      "reward": 0.7655918598175049,
      "reward_std": 0.0261928616091609,
      "reward_std_all": 0.0332572003826499,
      "rewards-max/forensics_sinkhorn_reward": 0.9856826066970825,
      "rewards-max/forensics_strict_edge_reward": 0.7026884853839874,
      "rewards-max/forensics_tiou_reward": 0.9894932806491852,
      "rewards/forensics_sinkhorn_reward": 0.7630048096179962,
      "rewards/forensics_strict_edge_reward": 0.10737545788288116,
      "rewards/forensics_tiou_reward": 0.6625529825687408,
      "step": 746
    },
    {
      "completion_length": 23.21875,
      "epoch": 2.291858678955453,
      "grad_norm": 10.641454696655273,
      "kl": 0.5390625,
      "learning_rate": 4.2615384615384613e-07,
      "loss": 0.2695,
      "reward": 0.7674623429775238,
      "reward_std": 0.04306042566895485,
      "reward_std_all": 0.043320540338754654,
      "rewards-max/forensics_sinkhorn_reward": 0.9856838881969452,
      "rewards-max/forensics_strict_edge_reward": 0.6637322008609772,
      "rewards-max/forensics_tiou_reward": 0.9906345903873444,
      "rewards/forensics_sinkhorn_reward": 0.7022225260734558,
      "rewards/forensics_strict_edge_reward": 0.05881350859999657,
      "rewards/forensics_tiou_reward": 0.45164071023464203,
      "step": 747
    },
    {
      "completion_length": 23.234375,
      "epoch": 2.294930875576037,
      "grad_norm": 8.128152847290039,
      "kl": 0.6201171875,
      "learning_rate": 4.2538461538461535e-07,
      "loss": 0.1296,
      "reward": 0.7597671747207642,
      "reward_std": 0.05711403861641884,
      "reward_std_all": 0.05439658835530281,
      "rewards-max/forensics_sinkhorn_reward": 0.9703193604946136,
      "rewards-max/forensics_strict_edge_reward": 0.8224053680896759,
      "rewards-max/forensics_tiou_reward": 0.9499211609363556,
      "rewards/forensics_sinkhorn_reward": 0.668902575969696,
      "rewards/forensics_strict_edge_reward": 0.10567304119467735,
      "rewards/forensics_tiou_reward": 0.4793878048658371,
      "step": 748
    },
    {
      "completion_length": 22.546875,
      "epoch": 2.2980030721966207,
      "grad_norm": 13.924028396606445,
      "kl": 0.4296875,
      "learning_rate": 4.246153846153846e-07,
      "loss": -0.0058,
      "reward": 0.7906827032566071,
      "reward_std": 0.01481121452525258,
      "reward_std_all": 0.016648522578179836,
      "rewards-max/forensics_sinkhorn_reward": 0.967603474855423,
      "rewards-max/forensics_strict_edge_reward": 0.2955048084259033,
      "rewards-max/forensics_tiou_reward": 0.9395286738872528,
      "rewards/forensics_sinkhorn_reward": 0.7857683598995209,
      "rewards/forensics_strict_edge_reward": 0.02821490168571472,
      "rewards/forensics_tiou_reward": 0.571516364812851,
      "step": 749
    },
    {
      "completion_length": 26.4375,
      "epoch": 2.3010752688172045,
      "grad_norm": 12.265645980834961,
      "kl": 0.470703125,
      "learning_rate": 4.2384615384615383e-07,
      "loss": -0.1682,
      "reward": 0.7432157695293427,
      "reward_std": 0.013852004893124104,
      "reward_std_all": 0.03784917201846838,
      "rewards-max/forensics_sinkhorn_reward": 0.9821996092796326,
      "rewards-max/forensics_strict_edge_reward": 0.7727466821670532,
      "rewards-max/forensics_tiou_reward": 0.983046680688858,
      "rewards/forensics_sinkhorn_reward": 0.7179175019264221,
      "rewards/forensics_strict_edge_reward": 0.10266559198498726,
      "rewards/forensics_tiou_reward": 0.5427004992961884,
      "step": 750
    },
    {
      "completion_length": 25.796875,
      "epoch": 2.3041474654377883,
      "grad_norm": 9.14106273651123,
      "kl": 0.5849609375,
      "learning_rate": 4.2307692307692304e-07,
      "loss": -0.0012,
      "reward": 0.7193397879600525,
      "reward_std": 0.06222692131996155,
      "reward_std_all": 0.07657638564705849,
      "rewards-max/forensics_sinkhorn_reward": 0.9716533720493317,
      "rewards-max/forensics_strict_edge_reward": 0.5703445225954056,
      "rewards-max/forensics_tiou_reward": 0.9564368426799774,
      "rewards/forensics_sinkhorn_reward": 0.6379004716873169,
      "rewards/forensics_strict_edge_reward": 0.033110180869698524,
      "rewards/forensics_tiou_reward": 0.4386066198348999,
      "step": 751
    },
    {
      "completion_length": 24.828125,
      "epoch": 2.3072196620583716,
      "grad_norm": 9.228615760803223,
      "kl": 0.5087890625,
      "learning_rate": 4.223076923076923e-07,
      "loss": 0.2679,
      "reward": 0.7466163039207458,
      "reward_std": 0.034113314002752304,
      "reward_std_all": 0.04724365100264549,
      "rewards-max/forensics_sinkhorn_reward": 0.9811403453350067,
      "rewards-max/forensics_strict_edge_reward": 0.502220444381237,
      "rewards-max/forensics_tiou_reward": 0.9784161150455475,
      "rewards/forensics_sinkhorn_reward": 0.6870910823345184,
      "rewards/forensics_strict_edge_reward": 0.0317134503275156,
      "rewards/forensics_tiou_reward": 0.5101870894432068,
      "step": 752
    },
    {
      "completion_length": 24.890625,
      "epoch": 2.3102918586789554,
      "grad_norm": 9.017988204956055,
      "kl": 0.3935546875,
      "learning_rate": 4.215384615384615e-07,
      "loss": 0.0083,
      "reward": 0.7808476388454437,
      "reward_std": 0.02053201664239168,
      "reward_std_all": 0.02418587077409029,
      "rewards-max/forensics_sinkhorn_reward": 0.9582830667495728,
      "rewards-max/forensics_strict_edge_reward": 0.5971094369888306,
      "rewards-max/forensics_tiou_reward": 0.9226938486099243,
      "rewards/forensics_sinkhorn_reward": 0.7264038622379303,
      "rewards/forensics_strict_edge_reward": 0.05210054200142622,
      "rewards/forensics_tiou_reward": 0.5175300389528275,
      "step": 753
    },
    {
      "completion_length": 26.390625,
      "epoch": 2.313364055299539,
      "grad_norm": 7.465929985046387,
      "kl": 0.6875,
      "learning_rate": 4.207692307692308e-07,
      "loss": -0.0385,
      "reward": 0.7952890992164612,
      "reward_std": 0.017895383294671774,
      "reward_std_all": 0.017426904290914536,
      "rewards-max/forensics_sinkhorn_reward": 0.9733932912349701,
      "rewards-max/forensics_strict_edge_reward": 0.6279962807893753,
      "rewards-max/forensics_tiou_reward": 0.9613611400127411,
      "rewards/forensics_sinkhorn_reward": 0.7865515649318695,
      "rewards/forensics_strict_edge_reward": 0.05856937915086746,
      "rewards/forensics_tiou_reward": 0.5791018605232239,
      "step": 754
    },
    {
      "completion_length": 21.796875,
      "epoch": 2.316436251920123,
      "grad_norm": 9.61178207397461,
      "kl": 0.466796875,
      "learning_rate": 4.1999999999999995e-07,
      "loss": -0.2374,
      "reward": 0.7865866422653198,
      "reward_std": 0.02218713564798236,
      "reward_std_all": 0.022883080877363682,
      "rewards-max/forensics_sinkhorn_reward": 0.9840056598186493,
      "rewards-max/forensics_strict_edge_reward": 0.6930722296237946,
      "rewards-max/forensics_tiou_reward": 0.9851466417312622,
      "rewards/forensics_sinkhorn_reward": 0.7695984542369843,
      "rewards/forensics_strict_edge_reward": 0.09207502380013466,
      "rewards/forensics_tiou_reward": 0.5531576722860336,
      "step": 755
    },
    {
      "completion_length": 24.015625,
      "epoch": 2.3195084485407067,
      "grad_norm": 9.0160493850708,
      "kl": 0.4345703125,
      "learning_rate": 4.192307692307692e-07,
      "loss": -0.0333,
      "reward": 0.7916843593120575,
      "reward_std": 0.015852161217480898,
      "reward_std_all": 0.01674602832645178,
      "rewards-max/forensics_sinkhorn_reward": 0.9760592579841614,
      "rewards-max/forensics_strict_edge_reward": 0.6676279902458191,
      "rewards-max/forensics_tiou_reward": 0.9667518138885498,
      "rewards/forensics_sinkhorn_reward": 0.787697821855545,
      "rewards/forensics_strict_edge_reward": 0.08473297953605652,
      "rewards/forensics_tiou_reward": 0.62572380900383,
      "step": 756
    },
    {
      "completion_length": 23.015625,
      "epoch": 2.3225806451612905,
      "grad_norm": 7.958619117736816,
      "kl": 0.421875,
      "learning_rate": 4.1846153846153843e-07,
      "loss": 0.079,
      "reward": 0.7521614134311676,
      "reward_std": 0.05379139259457588,
      "reward_std_all": 0.05400528386235237,
      "rewards-max/forensics_sinkhorn_reward": 0.9800349175930023,
      "rewards-max/forensics_strict_edge_reward": 0.8079957664012909,
      "rewards-max/forensics_tiou_reward": 0.9791860580444336,
      "rewards/forensics_sinkhorn_reward": 0.6841175258159637,
      "rewards/forensics_strict_edge_reward": 0.07651828974485397,
      "rewards/forensics_tiou_reward": 0.45838378369808197,
      "step": 757
    },
    {
      "completion_length": 21.8125,
      "epoch": 2.3256528417818743,
      "grad_norm": 14.354581832885742,
      "kl": 0.44140625,
      "learning_rate": 4.176923076923077e-07,
      "loss": -0.033,
      "reward": 0.7783250510692596,
      "reward_std": 0.02511359052732587,
      "reward_std_all": 0.028729810379445553,
      "rewards-max/forensics_sinkhorn_reward": 0.9880844354629517,
      "rewards-max/forensics_strict_edge_reward": 0.9375866651535034,
      "rewards-max/forensics_tiou_reward": 0.9972178637981415,
      "rewards/forensics_sinkhorn_reward": 0.7566193640232086,
      "rewards/forensics_strict_edge_reward": 0.1457476057112217,
      "rewards/forensics_tiou_reward": 0.5329979509115219,
      "step": 758
    },
    {
      "completion_length": 24.0,
      "epoch": 2.3287250384024576,
      "grad_norm": 9.718174934387207,
      "kl": 0.3486328125,
      "learning_rate": 4.1692307692307686e-07,
      "loss": 0.1307,
      "reward": 0.7687184512615204,
      "reward_std": 0.038605086505413055,
      "reward_std_all": 0.039576622657477856,
      "rewards-max/forensics_sinkhorn_reward": 0.9792854189872742,
      "rewards-max/forensics_strict_edge_reward": 0.5738120675086975,
      "rewards-max/forensics_tiou_reward": 0.9738231599330902,
      "rewards/forensics_sinkhorn_reward": 0.7217866480350494,
      "rewards/forensics_strict_edge_reward": 0.05086746998131275,
      "rewards/forensics_tiou_reward": 0.5102702975273132,
      "step": 759
    },
    {
      "completion_length": 24.875,
      "epoch": 2.3317972350230414,
      "grad_norm": 9.04605484008789,
      "kl": 0.568359375,
      "learning_rate": 4.161538461538461e-07,
      "loss": 0.0202,
      "reward": 0.7142127454280853,
      "reward_std": 0.07303262688219547,
      "reward_std_all": 0.08164595067501068,
      "rewards-max/forensics_sinkhorn_reward": 0.9809072017669678,
      "rewards-max/forensics_strict_edge_reward": 0.5862489938735962,
      "rewards-max/forensics_tiou_reward": 0.9794842302799225,
      "rewards/forensics_sinkhorn_reward": 0.6432949006557465,
      "rewards/forensics_strict_edge_reward": 0.05517387576401234,
      "rewards/forensics_tiou_reward": 0.44599829614162445,
      "step": 760
    },
    {
      "completion_length": 24.4375,
      "epoch": 2.334869431643625,
      "grad_norm": 9.406132698059082,
      "kl": 0.4033203125,
      "learning_rate": 4.153846153846154e-07,
      "loss": -0.1182,
      "reward": 0.7935228049755096,
      "reward_std": 0.020445337053388357,
      "reward_std_all": 0.019800817128270864,
      "rewards-max/forensics_sinkhorn_reward": 0.9868049621582031,
      "rewards-max/forensics_strict_edge_reward": 0.6096635609865189,
      "rewards-max/forensics_tiou_reward": 0.9917881190776825,
      "rewards/forensics_sinkhorn_reward": 0.7715112566947937,
      "rewards/forensics_strict_edge_reward": 0.06611206382513046,
      "rewards/forensics_tiou_reward": 0.6055835783481598,
      "step": 761
    },
    {
      "completion_length": 28.0,
      "epoch": 2.337941628264209,
      "grad_norm": 9.52772331237793,
      "kl": 0.521484375,
      "learning_rate": 4.146153846153846e-07,
      "loss": -0.0852,
      "reward": 0.7917119562625885,
      "reward_std": 0.01738073187880218,
      "reward_std_all": 0.01793626882135868,
      "rewards-max/forensics_sinkhorn_reward": 0.9767674803733826,
      "rewards-max/forensics_strict_edge_reward": 0.5159863531589508,
      "rewards-max/forensics_tiou_reward": 0.9677605926990509,
      "rewards/forensics_sinkhorn_reward": 0.7334726750850677,
      "rewards/forensics_strict_edge_reward": 0.06924204528331757,
      "rewards/forensics_tiou_reward": 0.6238678991794586,
      "step": 762
    },
    {
      "completion_length": 25.703125,
      "epoch": 2.3410138248847927,
      "grad_norm": 11.124678611755371,
      "kl": 0.794921875,
      "learning_rate": 4.1384615384615387e-07,
      "loss": -0.337,
      "reward": 0.7904204726219177,
      "reward_std": 0.02889412734657526,
      "reward_std_all": 0.02632133848965168,
      "rewards-max/forensics_sinkhorn_reward": 0.9854390621185303,
      "rewards-max/forensics_strict_edge_reward": 0.6934256851673126,
      "rewards-max/forensics_tiou_reward": 0.9908468425273895,
      "rewards/forensics_sinkhorn_reward": 0.7948287725448608,
      "rewards/forensics_strict_edge_reward": 0.17304648458957672,
      "rewards/forensics_tiou_reward": 0.63744255900383,
      "step": 763
    },
    {
      "completion_length": 22.75,
      "epoch": 2.3440860215053765,
      "grad_norm": 10.70434284210205,
      "kl": 1.4765625,
      "learning_rate": 4.1307692307692303e-07,
      "loss": 0.1194,
      "reward": 0.7790245413780212,
      "reward_std": 0.0456222128123045,
      "reward_std_all": 0.041154950857162476,
      "rewards-max/forensics_sinkhorn_reward": 0.9812035262584686,
      "rewards-max/forensics_strict_edge_reward": 0.5288843363523483,
      "rewards-max/forensics_tiou_reward": 0.9768109619617462,
      "rewards/forensics_sinkhorn_reward": 0.7799404263496399,
      "rewards/forensics_strict_edge_reward": 0.09601227939128876,
      "rewards/forensics_tiou_reward": 0.6013153195381165,
      "step": 764
    },
    {
      "completion_length": 22.1875,
      "epoch": 2.3471582181259603,
      "grad_norm": 15.101365089416504,
      "kl": 0.4599609375,
      "learning_rate": 4.123076923076923e-07,
      "loss": 0.0096,
      "reward": 0.7670651972293854,
      "reward_std": 0.030763667076826096,
      "reward_std_all": 0.036818817257881165,
      "rewards-max/forensics_sinkhorn_reward": 0.9813423454761505,
      "rewards-max/forensics_strict_edge_reward": 0.6149742305278778,
      "rewards-max/forensics_tiou_reward": 0.9778095483779907,
      "rewards/forensics_sinkhorn_reward": 0.7394543886184692,
      "rewards/forensics_strict_edge_reward": 0.09669229947030544,
      "rewards/forensics_tiou_reward": 0.5689843595027924,
      "step": 765
    },
    {
      "completion_length": 25.5625,
      "epoch": 2.3502304147465436,
      "grad_norm": 9.925936698913574,
      "kl": 0.3671875,
      "learning_rate": 4.115384615384615e-07,
      "loss": -0.2977,
      "reward": 0.794734537601471,
      "reward_std": 0.010469310567714274,
      "reward_std_all": 0.012410720344632864,
      "rewards-max/forensics_sinkhorn_reward": 0.9613696038722992,
      "rewards-max/forensics_strict_edge_reward": 0.5860248506069183,
      "rewards-max/forensics_tiou_reward": 0.9330649077892303,
      "rewards/forensics_sinkhorn_reward": 0.7519755363464355,
      "rewards/forensics_strict_edge_reward": 0.0722082182765007,
      "rewards/forensics_tiou_reward": 0.5718221962451935,
      "step": 766
    },
    {
      "completion_length": 22.546875,
      "epoch": 2.3533026113671274,
      "grad_norm": 16.466938018798828,
      "kl": 0.478515625,
      "learning_rate": 4.107692307692308e-07,
      "loss": 0.241,
      "reward": 0.7496432363986969,
      "reward_std": 0.02671959064900875,
      "reward_std_all": 0.04154863953590393,
      "rewards-max/forensics_sinkhorn_reward": 0.9812467098236084,
      "rewards-max/forensics_strict_edge_reward": 0.882964164018631,
      "rewards-max/forensics_tiou_reward": 0.981674075126648,
      "rewards/forensics_sinkhorn_reward": 0.6536675095558167,
      "rewards/forensics_strict_edge_reward": 0.09053783491253853,
      "rewards/forensics_tiou_reward": 0.49347957968711853,
      "step": 767
    },
    {
      "completion_length": 24.28125,
      "epoch": 2.356374807987711,
      "grad_norm": 10.056743621826172,
      "kl": 0.6337890625,
      "learning_rate": 4.0999999999999994e-07,
      "loss": 0.2822,
      "reward": 0.7897500395774841,
      "reward_std": 0.024358680471777916,
      "reward_std_all": 0.023619321174919605,
      "rewards-max/forensics_sinkhorn_reward": 0.9790621399879456,
      "rewards-max/forensics_strict_edge_reward": 0.5112937390804291,
      "rewards-max/forensics_tiou_reward": 0.9723609685897827,
      "rewards/forensics_sinkhorn_reward": 0.6939843893051147,
      "rewards/forensics_strict_edge_reward": 0.046551245264708996,
      "rewards/forensics_tiou_reward": 0.5401802957057953,
      "step": 768
    },
    {
      "completion_length": 25.703125,
      "epoch": 2.359447004608295,
      "grad_norm": 9.367069244384766,
      "kl": 0.4521484375,
      "learning_rate": 4.092307692307692e-07,
      "loss": -0.0718,
      "reward": 0.7766650021076202,
      "reward_std": 0.03409591969102621,
      "reward_std_all": 0.03330150991678238,
      "rewards-max/forensics_sinkhorn_reward": 0.9793814122676849,
      "rewards-max/forensics_strict_edge_reward": 0.6609791368246078,
      "rewards-max/forensics_tiou_reward": 0.9753893911838531,
      "rewards/forensics_sinkhorn_reward": 0.7278980314731598,
      "rewards/forensics_strict_edge_reward": 0.11192400008440018,
      "rewards/forensics_tiou_reward": 0.5823907554149628,
      "step": 769
    },
    {
      "completion_length": 27.90625,
      "epoch": 2.3625192012288787,
      "grad_norm": 12.571938514709473,
      "kl": 0.84765625,
      "learning_rate": 4.0846153846153847e-07,
      "loss": 0.1319,
      "reward": 0.6817530393600464,
      "reward_std": 0.05769771710038185,
      "reward_std_all": 0.08704234659671783,
      "rewards-max/forensics_sinkhorn_reward": 0.977714329957962,
      "rewards-max/forensics_strict_edge_reward": 0.4613604247570038,
      "rewards-max/forensics_tiou_reward": 0.9679357707500458,
      "rewards/forensics_sinkhorn_reward": 0.6241425275802612,
      "rewards/forensics_strict_edge_reward": 0.04759644716978073,
      "rewards/forensics_tiou_reward": 0.4836461991071701,
      "step": 770
    },
    {
      "completion_length": 27.125,
      "epoch": 2.3655913978494625,
      "grad_norm": 9.542746543884277,
      "kl": 0.4560546875,
      "learning_rate": 4.076923076923077e-07,
      "loss": 0.3254,
      "reward": 0.7406974136829376,
      "reward_std": 0.05301376432180405,
      "reward_std_all": 0.06082248315215111,
      "rewards-max/forensics_sinkhorn_reward": 0.9683262407779694,
      "rewards-max/forensics_strict_edge_reward": 0.5879503488540649,
      "rewards-max/forensics_tiou_reward": 0.952931821346283,
      "rewards/forensics_sinkhorn_reward": 0.6467870473861694,
      "rewards/forensics_strict_edge_reward": 0.0824664905667305,
      "rewards/forensics_tiou_reward": 0.4699876010417938,
      "step": 771
    },
    {
      "completion_length": 24.90625,
      "epoch": 2.3686635944700463,
      "grad_norm": 12.631492614746094,
      "kl": 0.431640625,
      "learning_rate": 4.069230769230769e-07,
      "loss": 0.2099,
      "reward": 0.7487880885601044,
      "reward_std": 0.023754728958010674,
      "reward_std_all": 0.04101403243839741,
      "rewards-max/forensics_sinkhorn_reward": 0.9672158658504486,
      "rewards-max/forensics_strict_edge_reward": 0.49555233120918274,
      "rewards-max/forensics_tiou_reward": 0.9497284591197968,
      "rewards/forensics_sinkhorn_reward": 0.6663925647735596,
      "rewards/forensics_strict_edge_reward": 0.0803149864077568,
      "rewards/forensics_tiou_reward": 0.5043189525604248,
      "step": 772
    },
    {
      "completion_length": 23.59375,
      "epoch": 2.3717357910906296,
      "grad_norm": 11.931177139282227,
      "kl": 0.794921875,
      "learning_rate": 4.061538461538461e-07,
      "loss": 0.0995,
      "reward": 0.7433716952800751,
      "reward_std": 0.06479539349675179,
      "reward_std_all": 0.06674631126224995,
      "rewards-max/forensics_sinkhorn_reward": 0.9770286977291107,
      "rewards-max/forensics_strict_edge_reward": 0.7933230996131897,
      "rewards-max/forensics_tiou_reward": 0.9726606011390686,
      "rewards/forensics_sinkhorn_reward": 0.7243398725986481,
      "rewards/forensics_strict_edge_reward": 0.08757542446255684,
      "rewards/forensics_tiou_reward": 0.529400646686554,
      "step": 773
    },
    {
      "completion_length": 25.5625,
      "epoch": 2.3748079877112134,
      "grad_norm": 11.471327781677246,
      "kl": 0.4130859375,
      "learning_rate": 4.053846153846154e-07,
      "loss": 0.0501,
      "reward": 0.7855928540229797,
      "reward_std": 0.018537352792918682,
      "reward_std_all": 0.020805840380489826,
      "rewards-max/forensics_sinkhorn_reward": 0.9550004601478577,
      "rewards-max/forensics_strict_edge_reward": 0.24896636605262756,
      "rewards-max/forensics_tiou_reward": 0.9168738722801208,
      "rewards/forensics_sinkhorn_reward": 0.7098945081233978,
      "rewards/forensics_strict_edge_reward": 0.027011987753212452,
      "rewards/forensics_tiou_reward": 0.5352224111557007,
      "step": 774
    },
    {
      "completion_length": 26.390625,
      "epoch": 2.377880184331797,
      "grad_norm": 10.511483192443848,
      "kl": 0.740234375,
      "learning_rate": 4.046153846153846e-07,
      "loss": 0.2478,
      "reward": 0.7704610824584961,
      "reward_std": 0.0297176381573081,
      "reward_std_all": 0.03282663319259882,
      "rewards-max/forensics_sinkhorn_reward": 0.9801853001117706,
      "rewards-max/forensics_strict_edge_reward": 0.398559108376503,
      "rewards-max/forensics_tiou_reward": 0.9717946350574493,
      "rewards/forensics_sinkhorn_reward": 0.6785888075828552,
      "rewards/forensics_strict_edge_reward": 0.042679375037550926,
      "rewards/forensics_tiou_reward": 0.4938378185033798,
      "step": 775
    },
    {
      "completion_length": 23.8125,
      "epoch": 2.380952380952381,
      "grad_norm": 9.428051948547363,
      "kl": 0.732421875,
      "learning_rate": 4.0384615384615386e-07,
      "loss": 0.0768,
      "reward": 0.756767213344574,
      "reward_std": 0.048312172293663025,
      "reward_std_all": 0.04954245314002037,
      "rewards-max/forensics_sinkhorn_reward": 0.9763608574867249,
      "rewards-max/forensics_strict_edge_reward": 0.5895819067955017,
      "rewards-max/forensics_tiou_reward": 0.9668032824993134,
      "rewards/forensics_sinkhorn_reward": 0.6736385524272919,
      "rewards/forensics_strict_edge_reward": 0.07452215440571308,
      "rewards/forensics_tiou_reward": 0.46930406987667084,
      "step": 776
    },
    {
      "completion_length": 24.46875,
      "epoch": 2.3840245775729647,
      "grad_norm": 7.6169538497924805,
      "kl": 0.3896484375,
      "learning_rate": 4.03076923076923e-07,
      "loss": 0.0793,
      "reward": 0.7849392294883728,
      "reward_std": 0.03673025034368038,
      "reward_std_all": 0.033603242598474026,
      "rewards-max/forensics_sinkhorn_reward": 0.9794751703739166,
      "rewards-max/forensics_strict_edge_reward": 0.5703445225954056,
      "rewards-max/forensics_tiou_reward": 0.9726933836936951,
      "rewards/forensics_sinkhorn_reward": 0.7715774774551392,
      "rewards/forensics_strict_edge_reward": 0.05639331974089146,
      "rewards/forensics_tiou_reward": 0.6524031162261963,
      "step": 777
    },
    {
      "completion_length": 26.515625,
      "epoch": 2.3870967741935485,
      "grad_norm": 11.069575309753418,
      "kl": 0.4404296875,
      "learning_rate": 4.023076923076923e-07,
      "loss": -0.1936,
      "reward": 0.7532163560390472,
      "reward_std": 0.04127915855497122,
      "reward_std_all": 0.05074683390557766,
      "rewards-max/forensics_sinkhorn_reward": 0.9805654585361481,
      "rewards-max/forensics_strict_edge_reward": 0.5809490978717804,
      "rewards-max/forensics_tiou_reward": 0.9777320921421051,
      "rewards/forensics_sinkhorn_reward": 0.7645584940910339,
      "rewards/forensics_strict_edge_reward": 0.08010946214199066,
      "rewards/forensics_tiou_reward": 0.6013730764389038,
      "step": 778
    },
    {
      "completion_length": 30.40625,
      "epoch": 2.3901689708141323,
      "grad_norm": 7.354000091552734,
      "kl": 0.388671875,
      "learning_rate": 4.015384615384615e-07,
      "loss": 0.0037,
      "reward": 0.7855532765388489,
      "reward_std": 0.02462196908891201,
      "reward_std_all": 0.025059999898076057,
      "rewards-max/forensics_sinkhorn_reward": 0.9731674194335938,
      "rewards-max/forensics_strict_edge_reward": 0.3981529176235199,
      "rewards-max/forensics_tiou_reward": 0.9637992978096008,
      "rewards/forensics_sinkhorn_reward": 0.7298097014427185,
      "rewards/forensics_strict_edge_reward": 0.07341228425502777,
      "rewards/forensics_tiou_reward": 0.5719238519668579,
      "step": 779
    },
    {
      "completion_length": 25.984375,
      "epoch": 2.3932411674347156,
      "grad_norm": 7.038487434387207,
      "kl": 0.560546875,
      "learning_rate": 4.0076923076923077e-07,
      "loss": 0.1739,
      "reward": 0.7667277157306671,
      "reward_std": 0.044123077765107155,
      "reward_std_all": 0.04587460681796074,
      "rewards-max/forensics_sinkhorn_reward": 0.9804566502571106,
      "rewards-max/forensics_strict_edge_reward": 0.28781314939260483,
      "rewards-max/forensics_tiou_reward": 0.9766160547733307,
      "rewards/forensics_sinkhorn_reward": 0.7253953218460083,
      "rewards/forensics_strict_edge_reward": 0.051129854284226894,
      "rewards/forensics_tiou_reward": 0.5421042144298553,
      "step": 780
    },
    {
      "completion_length": 25.59375,
      "epoch": 2.3963133640552994,
      "grad_norm": 8.702478408813477,
      "kl": 0.47265625,
      "learning_rate": 4e-07,
      "loss": 0.2656,
      "reward": 0.7686610519886017,
      "reward_std": 0.050733789801597595,
      "reward_std_all": 0.04841371811926365,
      "rewards-max/forensics_sinkhorn_reward": 0.9792609512805939,
      "rewards-max/forensics_strict_edge_reward": 0.47295713424682617,
      "rewards-max/forensics_tiou_reward": 0.9696680009365082,
      "rewards/forensics_sinkhorn_reward": 0.6999225914478302,
      "rewards/forensics_strict_edge_reward": 0.08989225700497627,
      "rewards/forensics_tiou_reward": 0.556341290473938,
      "step": 781
    },
    {
      "completion_length": 24.828125,
      "epoch": 2.399385560675883,
      "grad_norm": 10.334897994995117,
      "kl": 0.4833984375,
      "learning_rate": 3.992307692307692e-07,
      "loss": 0.0476,
      "reward": 0.7556944787502289,
      "reward_std": 0.02034828346222639,
      "reward_std_all": 0.03306836821138859,
      "rewards-max/forensics_sinkhorn_reward": 0.9708013832569122,
      "rewards-max/forensics_strict_edge_reward": 0.3554850406944752,
      "rewards-max/forensics_tiou_reward": 0.9435211718082428,
      "rewards/forensics_sinkhorn_reward": 0.6953985095024109,
      "rewards/forensics_strict_edge_reward": 0.04425761243328452,
      "rewards/forensics_tiou_reward": 0.47458943724632263,
      "step": 782
    },
    {
      "completion_length": 27.265625,
      "epoch": 2.402457757296467,
      "grad_norm": 10.36472225189209,
      "kl": 0.501953125,
      "learning_rate": 3.9846153846153846e-07,
      "loss": 0.0243,
      "reward": 0.7450304329395294,
      "reward_std": 0.05145992524921894,
      "reward_std_all": 0.05854209139943123,
      "rewards-max/forensics_sinkhorn_reward": 0.9767437875270844,
      "rewards-max/forensics_strict_edge_reward": 0.39520856738090515,
      "rewards-max/forensics_tiou_reward": 0.9661878645420074,
      "rewards/forensics_sinkhorn_reward": 0.6677435636520386,
      "rewards/forensics_strict_edge_reward": 0.051399193704128265,
      "rewards/forensics_tiou_reward": 0.5372661352157593,
      "step": 783
    },
    {
      "completion_length": 26.625,
      "epoch": 2.4055299539170507,
      "grad_norm": 11.536273002624512,
      "kl": 0.4462890625,
      "learning_rate": 3.976923076923077e-07,
      "loss": -0.2105,
      "reward": 0.7839712500572205,
      "reward_std": 0.015023495070636272,
      "reward_std_all": 0.021206960082054138,
      "rewards-max/forensics_sinkhorn_reward": 0.9846190214157104,
      "rewards-max/forensics_strict_edge_reward": 0.6105617582798004,
      "rewards-max/forensics_tiou_reward": 0.987868994474411,
      "rewards/forensics_sinkhorn_reward": 0.7611639201641083,
      "rewards/forensics_strict_edge_reward": 0.08733916655182838,
      "rewards/forensics_tiou_reward": 0.6239194869995117,
      "step": 784
    },
    {
      "completion_length": 24.71875,
      "epoch": 2.4086021505376345,
      "grad_norm": 10.805769920349121,
      "kl": 0.5390625,
      "learning_rate": 3.969230769230769e-07,
      "loss": -0.1012,
      "reward": 0.7786628603935242,
      "reward_std": 0.03092356724664569,
      "reward_std_all": 0.03130614827387035,
      "rewards-max/forensics_sinkhorn_reward": 0.9780990183353424,
      "rewards-max/forensics_strict_edge_reward": 0.5497699081897736,
      "rewards-max/forensics_tiou_reward": 0.9716692566871643,
      "rewards/forensics_sinkhorn_reward": 0.7045019567012787,
      "rewards/forensics_strict_edge_reward": 0.0640603806823492,
      "rewards/forensics_tiou_reward": 0.5111907720565796,
      "step": 785
    },
    {
      "completion_length": 23.140625,
      "epoch": 2.4116743471582183,
      "grad_norm": 10.282508850097656,
      "kl": 0.548828125,
      "learning_rate": 3.961538461538461e-07,
      "loss": 0.0352,
      "reward": 0.7765130698680878,
      "reward_std": 0.021736904978752136,
      "reward_std_all": 0.02674274891614914,
      "rewards-max/forensics_sinkhorn_reward": 0.9787427484989166,
      "rewards-max/forensics_strict_edge_reward": 0.6718102097511292,
      "rewards-max/forensics_tiou_reward": 0.9720540344715118,
      "rewards/forensics_sinkhorn_reward": 0.7105867564678192,
      "rewards/forensics_strict_edge_reward": 0.07507460564374924,
      "rewards/forensics_tiou_reward": 0.5240313112735748,
      "step": 786
    },
    {
      "completion_length": 21.171875,
      "epoch": 2.4147465437788016,
      "grad_norm": 9.206660270690918,
      "kl": 0.869140625,
      "learning_rate": 3.9538461538461537e-07,
      "loss": 0.0584,
      "reward": 0.7810788452625275,
      "reward_std": 0.03209549933671951,
      "reward_std_all": 0.031419022008776665,
      "rewards-max/forensics_sinkhorn_reward": 0.9820781052112579,
      "rewards-max/forensics_strict_edge_reward": 0.7026884853839874,
      "rewards-max/forensics_tiou_reward": 0.9790282845497131,
      "rewards/forensics_sinkhorn_reward": 0.763883650302887,
      "rewards/forensics_strict_edge_reward": 0.10636167973279953,
      "rewards/forensics_tiou_reward": 0.578168660402298,
      "step": 787
    },
    {
      "completion_length": 24.71875,
      "epoch": 2.4178187403993854,
      "grad_norm": 8.439703941345215,
      "kl": 0.5966796875,
      "learning_rate": 3.946153846153846e-07,
      "loss": -0.1038,
      "reward": 0.7882620990276337,
      "reward_std": 0.02935889922082424,
      "reward_std_all": 0.027100905776023865,
      "rewards-max/forensics_sinkhorn_reward": 0.9863358736038208,
      "rewards-max/forensics_strict_edge_reward": 0.7710188329219818,
      "rewards-max/forensics_tiou_reward": 0.9921844303607941,
      "rewards/forensics_sinkhorn_reward": 0.7660156190395355,
      "rewards/forensics_strict_edge_reward": 0.13247622549533844,
      "rewards/forensics_tiou_reward": 0.6448090970516205,
      "step": 788
    },
    {
      "completion_length": 22.28125,
      "epoch": 2.420890937019969,
      "grad_norm": 12.208122253417969,
      "kl": 0.423828125,
      "learning_rate": 3.9384615384615385e-07,
      "loss": -0.2784,
      "reward": 0.7521876394748688,
      "reward_std": 0.02257020794786513,
      "reward_std_all": 0.037329668179154396,
      "rewards-max/forensics_sinkhorn_reward": 0.9833586513996124,
      "rewards-max/forensics_strict_edge_reward": 0.8351600170135498,
      "rewards-max/forensics_tiou_reward": 0.9859802424907684,
      "rewards/forensics_sinkhorn_reward": 0.7121477127075195,
      "rewards/forensics_strict_edge_reward": 0.0484829843044281,
      "rewards/forensics_tiou_reward": 0.5103061646223068,
      "step": 789
    },
    {
      "completion_length": 23.828125,
      "epoch": 2.423963133640553,
      "grad_norm": 9.852266311645508,
      "kl": 0.4189453125,
      "learning_rate": 3.9307692307692306e-07,
      "loss": -0.1108,
      "reward": 0.7796495258808136,
      "reward_std": 0.01816902169957757,
      "reward_std_all": 0.024126150412485003,
      "rewards-max/forensics_sinkhorn_reward": 0.9767841398715973,
      "rewards-max/forensics_strict_edge_reward": 0.5221183001995087,
      "rewards-max/forensics_tiou_reward": 0.9679281115531921,
      "rewards/forensics_sinkhorn_reward": 0.7968379557132721,
      "rewards/forensics_strict_edge_reward": 0.08330662921071053,
      "rewards/forensics_tiou_reward": 0.5860921740531921,
      "step": 790
    },
    {
      "completion_length": 24.890625,
      "epoch": 2.4270353302611367,
      "grad_norm": 7.374654769897461,
      "kl": 0.568359375,
      "learning_rate": 3.923076923076923e-07,
      "loss": 0.2219,
      "reward": 0.733448714017868,
      "reward_std": 0.07094144634902477,
      "reward_std_all": 0.07348862569779158,
      "rewards-max/forensics_sinkhorn_reward": 0.9851996004581451,
      "rewards-max/forensics_strict_edge_reward": 0.6979960203170776,
      "rewards-max/forensics_tiou_reward": 0.9909516274929047,
      "rewards/forensics_sinkhorn_reward": 0.7122669816017151,
      "rewards/forensics_strict_edge_reward": 0.09548987820744514,
      "rewards/forensics_tiou_reward": 0.5292864441871643,
      "step": 791
    },
    {
      "completion_length": 25.390625,
      "epoch": 2.4301075268817205,
      "grad_norm": 9.805647850036621,
      "kl": 0.564453125,
      "learning_rate": 3.9153846153846154e-07,
      "loss": -0.1472,
      "reward": 0.7839109301567078,
      "reward_std": 0.019019631203263998,
      "reward_std_all": 0.021724273916333914,
      "rewards-max/forensics_sinkhorn_reward": 0.9819464981555939,
      "rewards-max/forensics_strict_edge_reward": 0.38481538742780685,
      "rewards-max/forensics_tiou_reward": 0.9820312857627869,
      "rewards/forensics_sinkhorn_reward": 0.7417087554931641,
      "rewards/forensics_strict_edge_reward": 0.06913349591195583,
      "rewards/forensics_tiou_reward": 0.5785110890865326,
      "step": 792
    },
    {
      "completion_length": 23.0,
      "epoch": 2.4331797235023043,
      "grad_norm": 10.54482364654541,
      "kl": 0.4697265625,
      "learning_rate": 3.9076923076923076e-07,
      "loss": -0.2771,
      "reward": 0.7827675938606262,
      "reward_std": 0.022923351265490055,
      "reward_std_all": 0.024267064407467842,
      "rewards-max/forensics_sinkhorn_reward": 0.9880258738994598,
      "rewards-max/forensics_strict_edge_reward": 0.7933230996131897,
      "rewards-max/forensics_tiou_reward": 0.9973886907100677,
      "rewards/forensics_sinkhorn_reward": 0.7318210005760193,
      "rewards/forensics_strict_edge_reward": 0.1123577430844307,
      "rewards/forensics_tiou_reward": 0.58330237865448,
      "step": 793
    },
    {
      "completion_length": 25.59375,
      "epoch": 2.436251920122888,
      "grad_norm": 8.475801467895508,
      "kl": 0.765625,
      "learning_rate": 3.8999999999999997e-07,
      "loss": -0.0748,
      "reward": 0.7630971074104309,
      "reward_std": 0.04507813788950443,
      "reward_std_all": 0.045523613691329956,
      "rewards-max/forensics_sinkhorn_reward": 0.9807379245758057,
      "rewards-max/forensics_strict_edge_reward": 0.49555233120918274,
      "rewards-max/forensics_tiou_reward": 0.9791945219039917,
      "rewards/forensics_sinkhorn_reward": 0.7604233622550964,
      "rewards/forensics_strict_edge_reward": 0.06174040213227272,
      "rewards/forensics_tiou_reward": 0.5626015365123749,
      "step": 794
    },
    {
      "completion_length": 29.28125,
      "epoch": 2.439324116743472,
      "grad_norm": 8.809706687927246,
      "kl": 0.4208984375,
      "learning_rate": 3.892307692307692e-07,
      "loss": -0.2968,
      "reward": 0.7552712559700012,
      "reward_std": 0.02175265667028725,
      "reward_std_all": 0.03680377209093422,
      "rewards-max/forensics_sinkhorn_reward": 0.9782158136367798,
      "rewards-max/forensics_strict_edge_reward": 0.41505207121372223,
      "rewards-max/forensics_tiou_reward": 0.975719153881073,
      "rewards/forensics_sinkhorn_reward": 0.7197922170162201,
      "rewards/forensics_strict_edge_reward": 0.04494474641978741,
      "rewards/forensics_tiou_reward": 0.5180491209030151,
      "step": 795
    },
    {
      "completion_length": 24.328125,
      "epoch": 2.442396313364055,
      "grad_norm": 11.109801292419434,
      "kl": 0.666015625,
      "learning_rate": 3.8846153846153845e-07,
      "loss": -0.0094,
      "reward": 0.7256869077682495,
      "reward_std": 0.06799695827066898,
      "reward_std_all": 0.07168583944439888,
      "rewards-max/forensics_sinkhorn_reward": 0.9858556985855103,
      "rewards-max/forensics_strict_edge_reward": 0.6231168210506439,
      "rewards-max/forensics_tiou_reward": 0.9901590645313263,
      "rewards/forensics_sinkhorn_reward": 0.749809592962265,
      "rewards/forensics_strict_edge_reward": 0.09649625420570374,
      "rewards/forensics_tiou_reward": 0.6160161197185516,
      "step": 796
    },
    {
      "completion_length": 22.5,
      "epoch": 2.445468509984639,
      "grad_norm": 15.69754409790039,
      "kl": 0.5068359375,
      "learning_rate": 3.8769230769230767e-07,
      "loss": 0.1399,
      "reward": 0.7681024968624115,
      "reward_std": 0.048268452286720276,
      "reward_std_all": 0.04548710398375988,
      "rewards-max/forensics_sinkhorn_reward": 0.9817879498004913,
      "rewards-max/forensics_strict_edge_reward": 0.731900691986084,
      "rewards-max/forensics_tiou_reward": 0.9805569052696228,
      "rewards/forensics_sinkhorn_reward": 0.7020170092582703,
      "rewards/forensics_strict_edge_reward": 0.08381645381450653,
      "rewards/forensics_tiou_reward": 0.6111981570720673,
      "step": 797
    },
    {
      "completion_length": 23.84375,
      "epoch": 2.4485407066052227,
      "grad_norm": 9.317622184753418,
      "kl": 0.662109375,
      "learning_rate": 3.8692307692307693e-07,
      "loss": -0.0198,
      "reward": 0.7759822010993958,
      "reward_std": 0.03770533483475447,
      "reward_std_all": 0.03664213977754116,
      "rewards-max/forensics_sinkhorn_reward": 0.9771629273891449,
      "rewards-max/forensics_strict_edge_reward": 0.6446520984172821,
      "rewards-max/forensics_tiou_reward": 0.9670630395412445,
      "rewards/forensics_sinkhorn_reward": 0.7610690593719482,
      "rewards/forensics_strict_edge_reward": 0.057646628469228745,
      "rewards/forensics_tiou_reward": 0.5546845495700836,
      "step": 798
    },
    {
      "completion_length": 25.53125,
      "epoch": 2.4516129032258065,
      "grad_norm": 9.705069541931152,
      "kl": 0.529296875,
      "learning_rate": 3.8615384615384615e-07,
      "loss": -0.2342,
      "reward": 0.7808151245117188,
      "reward_std": 0.023835788015276194,
      "reward_std_all": 0.026139866560697556,
      "rewards-max/forensics_sinkhorn_reward": 0.9881771504878998,
      "rewards-max/forensics_strict_edge_reward": 0.7401526570320129,
      "rewards-max/forensics_tiou_reward": 0.9972514808177948,
      "rewards/forensics_sinkhorn_reward": 0.7457562685012817,
      "rewards/forensics_strict_edge_reward": 0.09055418148636818,
      "rewards/forensics_tiou_reward": 0.5849889516830444,
      "step": 799
    },
    {
      "completion_length": 23.96875,
      "epoch": 2.4546850998463903,
      "grad_norm": 9.422497749328613,
      "kl": 0.4658203125,
      "learning_rate": 3.8538461538461536e-07,
      "loss": -0.0583,
      "reward": 0.7804489135742188,
      "reward_std": 0.025254830718040466,
      "reward_std_all": 0.02650711126625538,
      "rewards-max/forensics_sinkhorn_reward": 0.9739952087402344,
      "rewards-max/forensics_strict_edge_reward": 0.5072369128465652,
      "rewards-max/forensics_tiou_reward": 0.9598346650600433,
      "rewards/forensics_sinkhorn_reward": 0.7562038898468018,
      "rewards/forensics_strict_edge_reward": 0.05314525216817856,
      "rewards/forensics_tiou_reward": 0.5510856211185455,
      "step": 800
    },
    {
      "completion_length": 23.8125,
      "epoch": 2.457757296466974,
      "grad_norm": 9.206683158874512,
      "kl": 0.564453125,
      "learning_rate": 3.8461538461538463e-07,
      "loss": -0.0066,
      "reward": 0.7799613177776337,
      "reward_std": 0.025494597852230072,
      "reward_std_all": 0.02674275077879429,
      "rewards-max/forensics_sinkhorn_reward": 0.9853737056255341,
      "rewards-max/forensics_strict_edge_reward": 0.7212935090065002,
      "rewards-max/forensics_tiou_reward": 0.9905551970005035,
      "rewards/forensics_sinkhorn_reward": 0.7642802596092224,
      "rewards/forensics_strict_edge_reward": 0.10472648963332176,
      "rewards/forensics_tiou_reward": 0.5754736810922623,
      "step": 801
    },
    {
      "completion_length": 23.453125,
      "epoch": 2.460829493087558,
      "grad_norm": 10.339627265930176,
      "kl": 0.515625,
      "learning_rate": 3.8384615384615384e-07,
      "loss": 0.0554,
      "reward": 0.7169402241706848,
      "reward_std": 0.02628823136910796,
      "reward_std_all": 0.056859200820326805,
      "rewards-max/forensics_sinkhorn_reward": 0.9755926132202148,
      "rewards-max/forensics_strict_edge_reward": 0.6032288670539856,
      "rewards-max/forensics_tiou_reward": 0.9684375822544098,
      "rewards/forensics_sinkhorn_reward": 0.6577324271202087,
      "rewards/forensics_strict_edge_reward": 0.05658333748579025,
      "rewards/forensics_tiou_reward": 0.46762295067310333,
      "step": 802
    },
    {
      "completion_length": 26.78125,
      "epoch": 2.463901689708141,
      "grad_norm": 9.550678253173828,
      "kl": 0.4501953125,
      "learning_rate": 3.8307692307692305e-07,
      "loss": 0.0516,
      "reward": 0.743900328874588,
      "reward_std": 0.040286652743816376,
      "reward_std_all": 0.0521986298263073,
      "rewards-max/forensics_sinkhorn_reward": 0.9817537367343903,
      "rewards-max/forensics_strict_edge_reward": 0.49256695806980133,
      "rewards-max/forensics_tiou_reward": 0.981867104768753,
      "rewards/forensics_sinkhorn_reward": 0.6999776065349579,
      "rewards/forensics_strict_edge_reward": 0.05590617097914219,
      "rewards/forensics_tiou_reward": 0.5309987366199493,
      "step": 803
    },
    {
      "completion_length": 24.390625,
      "epoch": 2.466973886328725,
      "grad_norm": 13.701578140258789,
      "kl": 0.560546875,
      "learning_rate": 3.8230769230769227e-07,
      "loss": 0.0682,
      "reward": 0.7053759694099426,
      "reward_std": 0.018355919048190117,
      "reward_std_all": 0.060347530990839005,
      "rewards-max/forensics_sinkhorn_reward": 0.9856491982936859,
      "rewards-max/forensics_strict_edge_reward": 0.6061122119426727,
      "rewards-max/forensics_tiou_reward": 0.9907944202423096,
      "rewards/forensics_sinkhorn_reward": 0.6335303485393524,
      "rewards/forensics_strict_edge_reward": 0.049567198380827904,
      "rewards/forensics_tiou_reward": 0.4651476889848709,
      "step": 804
    },
    {
      "completion_length": 24.15625,
      "epoch": 2.4700460829493087,
      "grad_norm": 10.247763633728027,
      "kl": 0.455078125,
      "learning_rate": 3.8153846153846153e-07,
      "loss": 0.1143,
      "reward": 0.7839117646217346,
      "reward_std": 0.02830328419804573,
      "reward_std_all": 0.028008513152599335,
      "rewards-max/forensics_sinkhorn_reward": 0.9866279363632202,
      "rewards-max/forensics_strict_edge_reward": 0.6419725120067596,
      "rewards-max/forensics_tiou_reward": 0.9937855899333954,
      "rewards/forensics_sinkhorn_reward": 0.7169810235500336,
      "rewards/forensics_strict_edge_reward": 0.06480048783123493,
      "rewards/forensics_tiou_reward": 0.5239230096340179,
      "step": 805
    },
    {
      "completion_length": 22.890625,
      "epoch": 2.4731182795698925,
      "grad_norm": 11.383173942565918,
      "kl": 0.4716796875,
      "learning_rate": 3.8076923076923075e-07,
      "loss": 0.0406,
      "reward": 0.7873570621013641,
      "reward_std": 0.013273493386805058,
      "reward_std_all": 0.018522498197853565,
      "rewards-max/forensics_sinkhorn_reward": 0.9677472114562988,
      "rewards-max/forensics_strict_edge_reward": 0.4599933922290802,
      "rewards-max/forensics_tiou_reward": 0.9469680786132812,
      "rewards/forensics_sinkhorn_reward": 0.712051510810852,
      "rewards/forensics_strict_edge_reward": 0.0527775464579463,
      "rewards/forensics_tiou_reward": 0.5015612840652466,
      "step": 806
    },
    {
      "completion_length": 27.046875,
      "epoch": 2.4761904761904763,
      "grad_norm": 12.630731582641602,
      "kl": 0.671875,
      "learning_rate": 3.7999999999999996e-07,
      "loss": 0.1739,
      "reward": 0.7542282342910767,
      "reward_std": 0.03239421918988228,
      "reward_std_all": 0.04187491722404957,
      "rewards-max/forensics_sinkhorn_reward": 0.9497998952865601,
      "rewards-max/forensics_strict_edge_reward": 0.16161373257637024,
      "rewards-max/forensics_tiou_reward": 0.9076993465423584,
      "rewards/forensics_sinkhorn_reward": 0.6244585812091827,
      "rewards/forensics_strict_edge_reward": 0.015563775319606066,
      "rewards/forensics_tiou_reward": 0.40338435769081116,
      "step": 807
    },
    {
      "completion_length": 25.015625,
      "epoch": 2.47926267281106,
      "grad_norm": 12.90700912475586,
      "kl": 0.6015625,
      "learning_rate": 3.792307692307692e-07,
      "loss": 0.1314,
      "reward": 0.7841853201389313,
      "reward_std": 0.03154017589986324,
      "reward_std_all": 0.03046078234910965,
      "rewards-max/forensics_sinkhorn_reward": 0.973375529050827,
      "rewards-max/forensics_strict_edge_reward": 0.5082854479551315,
      "rewards-max/forensics_tiou_reward": 0.9593062996864319,
      "rewards/forensics_sinkhorn_reward": 0.7045159339904785,
      "rewards/forensics_strict_edge_reward": 0.05983203463256359,
      "rewards/forensics_tiou_reward": 0.541456788778305,
      "step": 808
    },
    {
      "completion_length": 25.859375,
      "epoch": 2.482334869431644,
      "grad_norm": 9.27469253540039,
      "kl": 0.5595703125,
      "learning_rate": 3.7846153846153844e-07,
      "loss": -0.0416,
      "reward": 0.7698245644569397,
      "reward_std": 0.028023707680404186,
      "reward_std_all": 0.03402324952185154,
      "rewards-max/forensics_sinkhorn_reward": 0.9845440089702606,
      "rewards-max/forensics_strict_edge_reward": 0.5905451774597168,
      "rewards-max/forensics_tiou_reward": 0.9891113638877869,
      "rewards/forensics_sinkhorn_reward": 0.7164622247219086,
      "rewards/forensics_strict_edge_reward": 0.0932079553604126,
      "rewards/forensics_tiou_reward": 0.568995475769043,
      "step": 809
    },
    {
      "completion_length": 27.328125,
      "epoch": 2.485407066052227,
      "grad_norm": 12.614872932434082,
      "kl": 0.5244140625,
      "learning_rate": 3.776923076923077e-07,
      "loss": -0.1766,
      "reward": 0.7804047763347626,
      "reward_std": 0.02308226330205798,
      "reward_std_all": 0.02679105568677187,
      "rewards-max/forensics_sinkhorn_reward": 0.9762562811374664,
      "rewards-max/forensics_strict_edge_reward": 0.6421390473842621,
      "rewards-max/forensics_tiou_reward": 0.9687552750110626,
      "rewards/forensics_sinkhorn_reward": 0.7480338215827942,
      "rewards/forensics_strict_edge_reward": 0.07682057842612267,
      "rewards/forensics_tiou_reward": 0.543892115354538,
      "step": 810
    },
    {
      "completion_length": 25.953125,
      "epoch": 2.488479262672811,
      "grad_norm": 11.074884414672852,
      "kl": 0.576171875,
      "learning_rate": 3.769230769230769e-07,
      "loss": -0.2522,
      "reward": 0.7926889955997467,
      "reward_std": 0.008885209914296865,
      "reward_std_all": 0.012584447860717773,
      "rewards-max/forensics_sinkhorn_reward": 0.973188042640686,
      "rewards-max/forensics_strict_edge_reward": 0.34002022445201874,
      "rewards-max/forensics_tiou_reward": 0.9540142118930817,
      "rewards/forensics_sinkhorn_reward": 0.7255344688892365,
      "rewards/forensics_strict_edge_reward": 0.04638863913714886,
      "rewards/forensics_tiou_reward": 0.5603610277175903,
      "step": 811
    },
    {
      "completion_length": 24.875,
      "epoch": 2.4915514592933947,
      "grad_norm": 9.64462661743164,
      "kl": 0.4921875,
      "learning_rate": 3.7615384615384614e-07,
      "loss": 0.2273,
      "reward": 0.7074811458587646,
      "reward_std": 0.050729963928461075,
      "reward_std_all": 0.07402431219816208,
      "rewards-max/forensics_sinkhorn_reward": 0.9673584699630737,
      "rewards-max/forensics_strict_edge_reward": 0.4592139273881912,
      "rewards-max/forensics_tiou_reward": 0.944782942533493,
      "rewards/forensics_sinkhorn_reward": 0.5920615196228027,
      "rewards/forensics_strict_edge_reward": 0.06068394100293517,
      "rewards/forensics_tiou_reward": 0.42286448180675507,
      "step": 812
    },
    {
      "completion_length": 28.015625,
      "epoch": 2.4946236559139785,
      "grad_norm": 10.372604370117188,
      "kl": 0.416015625,
      "learning_rate": 3.7538461538461535e-07,
      "loss": -0.0016,
      "reward": 0.7915409505367279,
      "reward_std": 0.022218135185539722,
      "reward_std_all": 0.021447401493787766,
      "rewards-max/forensics_sinkhorn_reward": 0.9866439700126648,
      "rewards-max/forensics_strict_edge_reward": 0.8959447741508484,
      "rewards-max/forensics_tiou_reward": 0.9928799867630005,
      "rewards/forensics_sinkhorn_reward": 0.8090168535709381,
      "rewards/forensics_strict_edge_reward": 0.13072886317968369,
      "rewards/forensics_tiou_reward": 0.6717319488525391,
      "step": 813
    },
    {
      "completion_length": 29.09375,
      "epoch": 2.4976958525345623,
      "grad_norm": 9.265728950500488,
      "kl": 0.69140625,
      "learning_rate": 3.746153846153846e-07,
      "loss": -0.1062,
      "reward": 0.7988206148147583,
      "reward_std": 0.007142889546230435,
      "reward_std_all": 0.009467415511608124,
      "rewards-max/forensics_sinkhorn_reward": 0.9818325042724609,
      "rewards-max/forensics_strict_edge_reward": 0.5532653331756592,
      "rewards-max/forensics_tiou_reward": 0.9800354838371277,
      "rewards/forensics_sinkhorn_reward": 0.7603744566440582,
      "rewards/forensics_strict_edge_reward": 0.05378951504826546,
      "rewards/forensics_tiou_reward": 0.6005455553531647,
      "step": 814
    },
    {
      "completion_length": 22.859375,
      "epoch": 2.500768049155146,
      "grad_norm": 12.307969093322754,
      "kl": 0.630859375,
      "learning_rate": 3.7384615384615383e-07,
      "loss": 0.0971,
      "reward": 0.7527711689472198,
      "reward_std": 0.04617526661604643,
      "reward_std_all": 0.05178176797926426,
      "rewards-max/forensics_sinkhorn_reward": 0.9726395308971405,
      "rewards-max/forensics_strict_edge_reward": 0.5952507928013802,
      "rewards-max/forensics_tiou_reward": 0.9601186513900757,
      "rewards/forensics_sinkhorn_reward": 0.6803472340106964,
      "rewards/forensics_strict_edge_reward": 0.07912898622453213,
      "rewards/forensics_tiou_reward": 0.43869927525520325,
      "step": 815
    },
    {
      "completion_length": 23.390625,
      "epoch": 2.50384024577573,
      "grad_norm": 13.982381820678711,
      "kl": 0.578125,
      "learning_rate": 3.7307692307692304e-07,
      "loss": 0.0366,
      "reward": 0.7767420411109924,
      "reward_std": 0.03979360684752464,
      "reward_std_all": 0.03819585219025612,
      "rewards-max/forensics_sinkhorn_reward": 0.9823606610298157,
      "rewards-max/forensics_strict_edge_reward": 0.7406637370586395,
      "rewards-max/forensics_tiou_reward": 0.9833325743675232,
      "rewards/forensics_sinkhorn_reward": 0.7662275433540344,
      "rewards/forensics_strict_edge_reward": 0.2032587006688118,
      "rewards/forensics_tiou_reward": 0.5984484851360321,
      "step": 816
    },
    {
      "completion_length": 24.859375,
      "epoch": 2.506912442396313,
      "grad_norm": 9.071664810180664,
      "kl": 0.6328125,
      "learning_rate": 3.7230769230769226e-07,
      "loss": 0.2297,
      "reward": 0.7471844553947449,
      "reward_std": 0.03273650445044041,
      "reward_std_all": 0.044241705909371376,
      "rewards-max/forensics_sinkhorn_reward": 0.9771628975868225,
      "rewards-max/forensics_strict_edge_reward": 0.5932522118091583,
      "rewards-max/forensics_tiou_reward": 0.9719240367412567,
      "rewards/forensics_sinkhorn_reward": 0.6587211787700653,
      "rewards/forensics_strict_edge_reward": 0.0804042425006628,
      "rewards/forensics_tiou_reward": 0.5036302208900452,
      "step": 817
    },
    {
      "completion_length": 26.015625,
      "epoch": 2.509984639016897,
      "grad_norm": 8.559953689575195,
      "kl": 0.4375,
      "learning_rate": 3.715384615384615e-07,
      "loss": 0.0343,
      "reward": 0.7450062334537506,
      "reward_std": 0.040543499402701855,
      "reward_std_all": 0.05064588598906994,
      "rewards-max/forensics_sinkhorn_reward": 0.9717219471931458,
      "rewards-max/forensics_strict_edge_reward": 0.34965111315250397,
      "rewards-max/forensics_tiou_reward": 0.9513580799102783,
      "rewards/forensics_sinkhorn_reward": 0.6803912818431854,
      "rewards/forensics_strict_edge_reward": 0.02254130132496357,
      "rewards/forensics_tiou_reward": 0.5046545565128326,
      "step": 818
    },
    {
      "completion_length": 27.375,
      "epoch": 2.5130568356374807,
      "grad_norm": 7.962421417236328,
      "kl": 0.583984375,
      "learning_rate": 3.707692307692308e-07,
      "loss": 0.0004,
      "reward": 0.7651661038398743,
      "reward_std": 0.035355940693989396,
      "reward_std_all": 0.040340120322071016,
      "rewards-max/forensics_sinkhorn_reward": 0.9785706996917725,
      "rewards-max/forensics_strict_edge_reward": 0.5580610483884811,
      "rewards-max/forensics_tiou_reward": 0.9732078611850739,
      "rewards/forensics_sinkhorn_reward": 0.739619106054306,
      "rewards/forensics_strict_edge_reward": 0.061635732650756836,
      "rewards/forensics_tiou_reward": 0.609061986207962,
      "step": 819
    },
    {
      "completion_length": 25.734375,
      "epoch": 2.5161290322580645,
      "grad_norm": 8.398163795471191,
      "kl": 0.58203125,
      "learning_rate": 3.7e-07,
      "loss": 0.1604,
      "reward": 0.7678946554660797,
      "reward_std": 0.04339907690882683,
      "reward_std_all": 0.04246230982244015,
      "rewards-max/forensics_sinkhorn_reward": 0.980284184217453,
      "rewards-max/forensics_strict_edge_reward": 0.6824378371238708,
      "rewards-max/forensics_tiou_reward": 0.979476124048233,
      "rewards/forensics_sinkhorn_reward": 0.7105869054794312,
      "rewards/forensics_strict_edge_reward": 0.06133623421192169,
      "rewards/forensics_tiou_reward": 0.5333559513092041,
      "step": 820
    },
    {
      "completion_length": 29.8125,
      "epoch": 2.5192012288786483,
      "grad_norm": 7.750259876251221,
      "kl": 0.4375,
      "learning_rate": 3.692307692307692e-07,
      "loss": 0.3074,
      "reward": 0.7503183484077454,
      "reward_std": 0.0624307282269001,
      "reward_std_all": 0.06131839379668236,
      "rewards-max/forensics_sinkhorn_reward": 0.9825065433979034,
      "rewards-max/forensics_strict_edge_reward": 0.5685442388057709,
      "rewards-max/forensics_tiou_reward": 0.9865938425064087,
      "rewards/forensics_sinkhorn_reward": 0.663768857717514,
      "rewards/forensics_strict_edge_reward": 0.11188926175236702,
      "rewards/forensics_tiou_reward": 0.5668710470199585,
      "step": 821
    },
    {
      "completion_length": 25.546875,
      "epoch": 2.522273425499232,
      "grad_norm": 9.609511375427246,
      "kl": 0.5244140625,
      "learning_rate": 3.6846153846153843e-07,
      "loss": -0.0188,
      "reward": 0.7623614966869354,
      "reward_std": 0.030983940698206425,
      "reward_std_all": 0.039945222437381744,
      "rewards-max/forensics_sinkhorn_reward": 0.9795764684677124,
      "rewards-max/forensics_strict_edge_reward": 0.6865042746067047,
      "rewards-max/forensics_tiou_reward": 0.9773224294185638,
      "rewards/forensics_sinkhorn_reward": 0.7470776438713074,
      "rewards/forensics_strict_edge_reward": 0.09396098367869854,
      "rewards/forensics_tiou_reward": 0.5568988919258118,
      "step": 822
    },
    {
      "completion_length": 25.9375,
      "epoch": 2.525345622119816,
      "grad_norm": 10.17660140991211,
      "kl": 0.4345703125,
      "learning_rate": 3.676923076923077e-07,
      "loss": -0.0388,
      "reward": 0.7895260453224182,
      "reward_std": 0.028587867505848408,
      "reward_std_all": 0.026832828298211098,
      "rewards-max/forensics_sinkhorn_reward": 0.9742801487445831,
      "rewards-max/forensics_strict_edge_reward": 0.6420994997024536,
      "rewards-max/forensics_tiou_reward": 0.9663333296775818,
      "rewards/forensics_sinkhorn_reward": 0.7498356401920319,
      "rewards/forensics_strict_edge_reward": 0.10714834183454514,
      "rewards/forensics_tiou_reward": 0.5786210596561432,
      "step": 823
    },
    {
      "completion_length": 26.78125,
      "epoch": 2.528417818740399,
      "grad_norm": 8.576749801635742,
      "kl": 0.5390625,
      "learning_rate": 3.669230769230769e-07,
      "loss": 0.0238,
      "reward": 0.7503414154052734,
      "reward_std": 0.046257125213742256,
      "reward_std_all": 0.05004885978996754,
      "rewards-max/forensics_sinkhorn_reward": 0.9748852550983429,
      "rewards-max/forensics_strict_edge_reward": 0.7457418143749237,
      "rewards-max/forensics_tiou_reward": 0.9676859676837921,
      "rewards/forensics_sinkhorn_reward": 0.6893176436424255,
      "rewards/forensics_strict_edge_reward": 0.10240504890680313,
      "rewards/forensics_tiou_reward": 0.5355744063854218,
      "step": 824
    },
    {
      "completion_length": 25.59375,
      "epoch": 2.531490015360983,
      "grad_norm": 19.435970306396484,
      "kl": 1.38671875,
      "learning_rate": 3.661538461538461e-07,
      "loss": 0.1642,
      "reward": 0.7603723704814911,
      "reward_std": 0.0282282792031765,
      "reward_std_all": 0.03603881597518921,
      "rewards-max/forensics_sinkhorn_reward": 0.9734451770782471,
      "rewards-max/forensics_strict_edge_reward": 0.4812778830528259,
      "rewards-max/forensics_tiou_reward": 0.9597258567810059,
      "rewards/forensics_sinkhorn_reward": 0.7228546738624573,
      "rewards/forensics_strict_edge_reward": 0.0415749866515398,
      "rewards/forensics_tiou_reward": 0.5348649322986603,
      "step": 825
    },
    {
      "completion_length": 24.84375,
      "epoch": 2.5345622119815667,
      "grad_norm": 9.193868637084961,
      "kl": 0.5361328125,
      "learning_rate": 3.6538461538461534e-07,
      "loss": -0.1231,
      "reward": 0.7404166460037231,
      "reward_std": 0.018919500056654215,
      "reward_std_all": 0.04118264280259609,
      "rewards-max/forensics_sinkhorn_reward": 0.9807112812995911,
      "rewards-max/forensics_strict_edge_reward": 0.6358188390731812,
      "rewards-max/forensics_tiou_reward": 0.9797324240207672,
      "rewards/forensics_sinkhorn_reward": 0.6938670873641968,
      "rewards/forensics_strict_edge_reward": 0.09634054824709892,
      "rewards/forensics_tiou_reward": 0.5108021348714828,
      "step": 826
    },
    {
      "completion_length": 23.359375,
      "epoch": 2.5376344086021505,
      "grad_norm": 14.815820693969727,
      "kl": 1.1962890625,
      "learning_rate": 3.646153846153846e-07,
      "loss": 0.0744,
      "reward": 0.7592766284942627,
      "reward_std": 0.04877249989658594,
      "reward_std_all": 0.048902819864451885,
      "rewards-max/forensics_sinkhorn_reward": 0.9627987742424011,
      "rewards-max/forensics_strict_edge_reward": 0.5244988799095154,
      "rewards-max/forensics_tiou_reward": 0.932572215795517,
      "rewards/forensics_sinkhorn_reward": 0.713601291179657,
      "rewards/forensics_strict_edge_reward": 0.09283479116857052,
      "rewards/forensics_tiou_reward": 0.5301229655742645,
      "step": 827
    },
    {
      "completion_length": 23.75,
      "epoch": 2.5407066052227343,
      "grad_norm": 14.464690208435059,
      "kl": 0.541015625,
      "learning_rate": 3.638461538461538e-07,
      "loss": 0.3086,
      "reward": 0.7074914574623108,
      "reward_std": 0.05353287793695927,
      "reward_std_all": 0.07277827709913254,
      "rewards-max/forensics_sinkhorn_reward": 0.9819866120815277,
      "rewards-max/forensics_strict_edge_reward": 0.7336080372333527,
      "rewards-max/forensics_tiou_reward": 0.9823478758335114,
      "rewards/forensics_sinkhorn_reward": 0.6351424753665924,
      "rewards/forensics_strict_edge_reward": 0.06512279063463211,
      "rewards/forensics_tiou_reward": 0.4223337322473526,
      "step": 828
    },
    {
      "completion_length": 23.5625,
      "epoch": 2.543778801843318,
      "grad_norm": 9.21496295928955,
      "kl": 0.3994140625,
      "learning_rate": 3.6307692307692303e-07,
      "loss": 0.0161,
      "reward": 0.7434105277061462,
      "reward_std": 0.06652611866593361,
      "reward_std_all": 0.06640737876296043,
      "rewards-max/forensics_sinkhorn_reward": 0.9816738665103912,
      "rewards-max/forensics_strict_edge_reward": 0.5853640735149384,
      "rewards-max/forensics_tiou_reward": 0.9789004325866699,
      "rewards/forensics_sinkhorn_reward": 0.725479930639267,
      "rewards/forensics_strict_edge_reward": 0.11453580111265182,
      "rewards/forensics_tiou_reward": 0.5611849427223206,
      "step": 829
    },
    {
      "completion_length": 26.09375,
      "epoch": 2.546850998463902,
      "grad_norm": 9.179615020751953,
      "kl": 0.525390625,
      "learning_rate": 3.623076923076923e-07,
      "loss": 0.0138,
      "reward": 0.7717209458351135,
      "reward_std": 0.02413521276321262,
      "reward_std_all": 0.030076215974986553,
      "rewards-max/forensics_sinkhorn_reward": 0.9808962643146515,
      "rewards-max/forensics_strict_edge_reward": 0.6112343668937683,
      "rewards-max/forensics_tiou_reward": 0.9820695519447327,
      "rewards/forensics_sinkhorn_reward": 0.6696154475212097,
      "rewards/forensics_strict_edge_reward": 0.07096663024276495,
      "rewards/forensics_tiou_reward": 0.5052915662527084,
      "step": 830
    },
    {
      "completion_length": 27.234375,
      "epoch": 2.549923195084485,
      "grad_norm": 10.076265335083008,
      "kl": 0.6982421875,
      "learning_rate": 3.615384615384615e-07,
      "loss": 0.0186,
      "reward": 0.7250734567642212,
      "reward_std": 0.05420373869128525,
      "reward_std_all": 0.06355382455512881,
      "rewards-max/forensics_sinkhorn_reward": 0.9776284396648407,
      "rewards-max/forensics_strict_edge_reward": 0.580880731344223,
      "rewards-max/forensics_tiou_reward": 0.9723628461360931,
      "rewards/forensics_sinkhorn_reward": 0.661047101020813,
      "rewards/forensics_strict_edge_reward": 0.0656477864831686,
      "rewards/forensics_tiou_reward": 0.47798506915569305,
      "step": 831
    },
    {
      "completion_length": 25.1875,
      "epoch": 2.5529953917050694,
      "grad_norm": 9.297563552856445,
      "kl": 0.619140625,
      "learning_rate": 3.607692307692308e-07,
      "loss": -0.0055,
      "reward": 0.8027363121509552,
      "reward_std": 0.005298048024997115,
      "reward_std_all": 0.006442180136218667,
      "rewards-max/forensics_sinkhorn_reward": 0.9827278852462769,
      "rewards-max/forensics_strict_edge_reward": 0.4213927835226059,
      "rewards-max/forensics_tiou_reward": 0.9850271940231323,
      "rewards/forensics_sinkhorn_reward": 0.7622725665569305,
      "rewards/forensics_strict_edge_reward": 0.0629151202738285,
      "rewards/forensics_tiou_reward": 0.6061173677444458,
      "step": 832
    },
    {
      "completion_length": 24.515625,
      "epoch": 2.5560675883256527,
      "grad_norm": 9.273186683654785,
      "kl": 0.361328125,
      "learning_rate": 3.6e-07,
      "loss": 0.0125,
      "reward": 0.7605176568031311,
      "reward_std": 0.015159075614064932,
      "reward_std_all": 0.03073298931121826,
      "rewards-max/forensics_sinkhorn_reward": 0.9878800511360168,
      "rewards-max/forensics_strict_edge_reward": 0.5758698284626007,
      "rewards-max/forensics_tiou_reward": 0.9959734380245209,
      "rewards/forensics_sinkhorn_reward": 0.7146493792533875,
      "rewards/forensics_strict_edge_reward": 0.08344958536326885,
      "rewards/forensics_tiou_reward": 0.5372522175312042,
      "step": 833
    },
    {
      "completion_length": 22.453125,
      "epoch": 2.5591397849462365,
      "grad_norm": 10.236160278320312,
      "kl": 0.373046875,
      "learning_rate": 3.592307692307692e-07,
      "loss": -0.1243,
      "reward": 0.7632718086242676,
      "reward_std": 0.04055050574243069,
      "reward_std_all": 0.04221925139427185,
      "rewards-max/forensics_sinkhorn_reward": 0.9881396591663361,
      "rewards-max/forensics_strict_edge_reward": 0.756708562374115,
      "rewards-max/forensics_tiou_reward": 0.9969333708286285,
      "rewards/forensics_sinkhorn_reward": 0.72925665974617,
      "rewards/forensics_strict_edge_reward": 0.11162614822387695,
      "rewards/forensics_tiou_reward": 0.5476300716400146,
      "step": 834
    },
    {
      "completion_length": 25.296875,
      "epoch": 2.5622119815668203,
      "grad_norm": 8.743009567260742,
      "kl": 0.4892578125,
      "learning_rate": 3.584615384615384e-07,
      "loss": 0.318,
      "reward": 0.7482857406139374,
      "reward_std": 0.03410269878804684,
      "reward_std_all": 0.047369509004056454,
      "rewards-max/forensics_sinkhorn_reward": 0.9644593894481659,
      "rewards-max/forensics_strict_edge_reward": 0.6487045586109161,
      "rewards-max/forensics_tiou_reward": 0.9450119435787201,
      "rewards/forensics_sinkhorn_reward": 0.6621375679969788,
      "rewards/forensics_strict_edge_reward": 0.03762350603938103,
      "rewards/forensics_tiou_reward": 0.4474058449268341,
      "step": 835
    },
    {
      "completion_length": 25.015625,
      "epoch": 2.565284178187404,
      "grad_norm": 9.066826820373535,
      "kl": 0.4443359375,
      "learning_rate": 3.576923076923077e-07,
      "loss": 0.2359,
      "reward": 0.7289322912693024,
      "reward_std": 0.04721424728631973,
      "reward_std_all": 0.06024380400776863,
      "rewards-max/forensics_sinkhorn_reward": 0.9759465456008911,
      "rewards-max/forensics_strict_edge_reward": 0.5579978227615356,
      "rewards-max/forensics_tiou_reward": 0.964979350566864,
      "rewards/forensics_sinkhorn_reward": 0.642768919467926,
      "rewards/forensics_strict_edge_reward": 0.05242317542433739,
      "rewards/forensics_tiou_reward": 0.42657411098480225,
      "step": 836
    },
    {
      "completion_length": 22.078125,
      "epoch": 2.568356374807988,
      "grad_norm": 8.723791122436523,
      "kl": 0.576171875,
      "learning_rate": 3.569230769230769e-07,
      "loss": 0.1931,
      "reward": 0.7669084370136261,
      "reward_std": 0.042855761013925076,
      "reward_std_all": 0.04253888688981533,
      "rewards-max/forensics_sinkhorn_reward": 0.9638553559780121,
      "rewards-max/forensics_strict_edge_reward": 0.42958900332450867,
      "rewards-max/forensics_tiou_reward": 0.9402340054512024,
      "rewards/forensics_sinkhorn_reward": 0.7164221405982971,
      "rewards/forensics_strict_edge_reward": 0.04827316105365753,
      "rewards/forensics_tiou_reward": 0.5427830517292023,
      "step": 837
    },
    {
      "completion_length": 23.84375,
      "epoch": 2.571428571428571,
      "grad_norm": 10.32559871673584,
      "kl": 0.533203125,
      "learning_rate": 3.561538461538461e-07,
      "loss": 0.0314,
      "reward": 0.7908149063587189,
      "reward_std": 0.011568314861506224,
      "reward_std_all": 0.015520932152867317,
      "rewards-max/forensics_sinkhorn_reward": 0.9809672832489014,
      "rewards-max/forensics_strict_edge_reward": 0.6065583229064941,
      "rewards-max/forensics_tiou_reward": 0.9787634015083313,
      "rewards/forensics_sinkhorn_reward": 0.7926440834999084,
      "rewards/forensics_strict_edge_reward": 0.08275985345244408,
      "rewards/forensics_tiou_reward": 0.5905833542346954,
      "step": 838
    },
    {
      "completion_length": 31.0625,
      "epoch": 2.5745007680491554,
      "grad_norm": 8.070738792419434,
      "kl": 0.560546875,
      "learning_rate": 3.553846153846154e-07,
      "loss": 0.339,
      "reward": 0.6994648575782776,
      "reward_std": 0.08648830279707909,
      "reward_std_all": 0.09350274503231049,
      "rewards-max/forensics_sinkhorn_reward": 0.9696229994297028,
      "rewards-max/forensics_strict_edge_reward": 0.21245912462472916,
      "rewards-max/forensics_tiou_reward": 0.9533098638057709,
      "rewards/forensics_sinkhorn_reward": 0.5790383219718933,
      "rewards/forensics_strict_edge_reward": 0.01776686543598771,
      "rewards/forensics_tiou_reward": 0.3882576525211334,
      "step": 839
    },
    {
      "completion_length": 22.78125,
      "epoch": 2.5775729646697387,
      "grad_norm": 10.114624977111816,
      "kl": 0.484375,
      "learning_rate": 3.546153846153846e-07,
      "loss": -0.0582,
      "reward": 0.7685348093509674,
      "reward_std": 0.011984213953837752,
      "reward_std_all": 0.02635856345295906,
      "rewards-max/forensics_sinkhorn_reward": 0.9775539040565491,
      "rewards-max/forensics_strict_edge_reward": 0.684698224067688,
      "rewards-max/forensics_tiou_reward": 0.9712312519550323,
      "rewards/forensics_sinkhorn_reward": 0.7285445034503937,
      "rewards/forensics_strict_edge_reward": 0.08079851046204567,
      "rewards/forensics_tiou_reward": 0.5454814285039902,
      "step": 840
    },
    {
      "completion_length": 24.34375,
      "epoch": 2.5806451612903225,
      "grad_norm": 10.26390266418457,
      "kl": 0.4677734375,
      "learning_rate": 3.5384615384615386e-07,
      "loss": 0.0326,
      "reward": 0.7831893563270569,
      "reward_std": 0.028200209140777588,
      "reward_std_all": 0.02794572338461876,
      "rewards-max/forensics_sinkhorn_reward": 0.9686367213726044,
      "rewards-max/forensics_strict_edge_reward": 0.3942713588476181,
      "rewards-max/forensics_tiou_reward": 0.9446439445018768,
      "rewards/forensics_sinkhorn_reward": 0.7526263892650604,
      "rewards/forensics_strict_edge_reward": 0.06031453888863325,
      "rewards/forensics_tiou_reward": 0.5266083478927612,
      "step": 841
    },
    {
      "completion_length": 20.96875,
      "epoch": 2.5837173579109063,
      "grad_norm": 8.037723541259766,
      "kl": 0.490234375,
      "learning_rate": 3.530769230769231e-07,
      "loss": -0.0756,
      "reward": 0.7729936838150024,
      "reward_std": 0.033295087050646544,
      "reward_std_all": 0.03393417410552502,
      "rewards-max/forensics_sinkhorn_reward": 0.9819857776165009,
      "rewards-max/forensics_strict_edge_reward": 0.7558856904506683,
      "rewards-max/forensics_tiou_reward": 0.9835017323493958,
      "rewards/forensics_sinkhorn_reward": 0.7444454729557037,
      "rewards/forensics_strict_edge_reward": 0.08844536915421486,
      "rewards/forensics_tiou_reward": 0.5123624354600906,
      "step": 842
    },
    {
      "completion_length": 26.96875,
      "epoch": 2.58678955453149,
      "grad_norm": 11.389286041259766,
      "kl": 0.72265625,
      "learning_rate": 3.523076923076923e-07,
      "loss": 0.0134,
      "reward": 0.7869434952735901,
      "reward_std": 0.02871153224259615,
      "reward_std_all": 0.02738203853368759,
      "rewards-max/forensics_sinkhorn_reward": 0.95992511510849,
      "rewards-max/forensics_strict_edge_reward": 0.5373985022306442,
      "rewards-max/forensics_tiou_reward": 0.9346423447132111,
      "rewards/forensics_sinkhorn_reward": 0.7128380239009857,
      "rewards/forensics_strict_edge_reward": 0.06566719524562359,
      "rewards/forensics_tiou_reward": 0.5163847655057907,
      "step": 843
    },
    {
      "completion_length": 25.0,
      "epoch": 2.589861751152074,
      "grad_norm": 14.475425720214844,
      "kl": 0.583984375,
      "learning_rate": 3.515384615384615e-07,
      "loss": -0.0491,
      "reward": 0.7837367355823517,
      "reward_std": 0.038381281308829784,
      "reward_std_all": 0.035164509899914265,
      "rewards-max/forensics_sinkhorn_reward": 0.9806428849697113,
      "rewards-max/forensics_strict_edge_reward": 0.5210062116384506,
      "rewards-max/forensics_tiou_reward": 0.9758707880973816,
      "rewards/forensics_sinkhorn_reward": 0.7746685743331909,
      "rewards/forensics_strict_edge_reward": 0.09301231987774372,
      "rewards/forensics_tiou_reward": 0.5365476906299591,
      "step": 844
    },
    {
      "completion_length": 24.8125,
      "epoch": 2.592933947772657,
      "grad_norm": 11.830148696899414,
      "kl": 0.4755859375,
      "learning_rate": 3.5076923076923077e-07,
      "loss": 0.0827,
      "reward": 0.7055903077125549,
      "reward_std": 0.04380744695663452,
      "reward_std_all": 0.06637781113386154,
      "rewards-max/forensics_sinkhorn_reward": 0.9861929416656494,
      "rewards-max/forensics_strict_edge_reward": 0.8959447741508484,
      "rewards-max/forensics_tiou_reward": 0.9927589595317841,
      "rewards/forensics_sinkhorn_reward": 0.6497633457183838,
      "rewards/forensics_strict_edge_reward": 0.10024473071098328,
      "rewards/forensics_tiou_reward": 0.4144233912229538,
      "step": 845
    },
    {
      "completion_length": 27.390625,
      "epoch": 2.5960061443932414,
      "grad_norm": 8.144842147827148,
      "kl": 1.08203125,
      "learning_rate": 3.5e-07,
      "loss": 0.1273,
      "reward": 0.7904582917690277,
      "reward_std": 0.026944475248456,
      "reward_std_all": 0.02529030479490757,
      "rewards-max/forensics_sinkhorn_reward": 0.983148992061615,
      "rewards-max/forensics_strict_edge_reward": 0.8032653331756592,
      "rewards-max/forensics_tiou_reward": 0.9848375916481018,
      "rewards/forensics_sinkhorn_reward": 0.753411203622818,
      "rewards/forensics_strict_edge_reward": 0.10297110676765442,
      "rewards/forensics_tiou_reward": 0.5928606390953064,
      "step": 846
    },
    {
      "completion_length": 26.234375,
      "epoch": 2.5990783410138247,
      "grad_norm": 9.626211166381836,
      "kl": 0.5400390625,
      "learning_rate": 3.492307692307692e-07,
      "loss": 0.1452,
      "reward": 0.794494241476059,
      "reward_std": 0.017470243386924267,
      "reward_std_all": 0.017313295975327492,
      "rewards-max/forensics_sinkhorn_reward": 0.9699920117855072,
      "rewards-max/forensics_strict_edge_reward": 0.4029012620449066,
      "rewards-max/forensics_tiou_reward": 0.9530980885028839,
      "rewards/forensics_sinkhorn_reward": 0.7567697167396545,
      "rewards/forensics_strict_edge_reward": 0.053500786423683167,
      "rewards/forensics_tiou_reward": 0.614484041929245,
      "step": 847
    },
    {
      "completion_length": 27.125,
      "epoch": 2.6021505376344085,
      "grad_norm": 9.975096702575684,
      "kl": 0.4375,
      "learning_rate": 3.4846153846153846e-07,
      "loss": 0.1083,
      "reward": 0.766288697719574,
      "reward_std": 0.022074616514146328,
      "reward_std_all": 0.03211375419050455,
      "rewards-max/forensics_sinkhorn_reward": 0.9727839827537537,
      "rewards-max/forensics_strict_edge_reward": 0.48137304186820984,
      "rewards-max/forensics_tiou_reward": 0.9549819231033325,
      "rewards/forensics_sinkhorn_reward": 0.6751887202262878,
      "rewards/forensics_strict_edge_reward": 0.05177725479006767,
      "rewards/forensics_tiou_reward": 0.48073312640190125,
      "step": 848
    },
    {
      "completion_length": 24.421875,
      "epoch": 2.6052227342549923,
      "grad_norm": 9.651800155639648,
      "kl": 0.505859375,
      "learning_rate": 3.476923076923077e-07,
      "loss": 0.0921,
      "reward": 0.7902928590774536,
      "reward_std": 0.0219752611592412,
      "reward_std_all": 0.0216464726254344,
      "rewards-max/forensics_sinkhorn_reward": 0.9792308211326599,
      "rewards-max/forensics_strict_edge_reward": 0.38006649911403656,
      "rewards-max/forensics_tiou_reward": 0.9759885370731354,
      "rewards/forensics_sinkhorn_reward": 0.7343031764030457,
      "rewards/forensics_strict_edge_reward": 0.05826609022915363,
      "rewards/forensics_tiou_reward": 0.6275595128536224,
      "step": 849
    },
    {
      "completion_length": 22.015625,
      "epoch": 2.608294930875576,
      "grad_norm": 15.834524154663086,
      "kl": 1.0625,
      "learning_rate": 3.4692307692307694e-07,
      "loss": 0.2058,
      "reward": 0.7844524383544922,
      "reward_std": 0.029979856219142675,
      "reward_std_all": 0.028681505471467972,
      "rewards-max/forensics_sinkhorn_reward": 0.9846440851688385,
      "rewards-max/forensics_strict_edge_reward": 0.5829301476478577,
      "rewards-max/forensics_tiou_reward": 0.9834010601043701,
      "rewards/forensics_sinkhorn_reward": 0.7168264985084534,
      "rewards/forensics_strict_edge_reward": 0.09436957910656929,
      "rewards/forensics_tiou_reward": 0.5997882187366486,
      "step": 850
    },
    {
      "completion_length": 23.921875,
      "epoch": 2.61136712749616,
      "grad_norm": 9.576458930969238,
      "kl": 0.4423828125,
      "learning_rate": 3.461538461538461e-07,
      "loss": -0.0148,
      "reward": 0.7537919580936432,
      "reward_std": 0.04188608191907406,
      "reward_std_all": 0.05014077574014664,
      "rewards-max/forensics_sinkhorn_reward": 0.9856651127338409,
      "rewards-max/forensics_strict_edge_reward": 0.6942952275276184,
      "rewards-max/forensics_tiou_reward": 0.9912989735603333,
      "rewards/forensics_sinkhorn_reward": 0.6988898813724518,
      "rewards/forensics_strict_edge_reward": 0.08248025737702847,
      "rewards/forensics_tiou_reward": 0.5574366748332977,
      "step": 851
    },
    {
      "completion_length": 29.40625,
      "epoch": 2.6144393241167436,
      "grad_norm": 8.677658081054688,
      "kl": 0.392578125,
      "learning_rate": 3.4538461538461537e-07,
      "loss": 0.2137,
      "reward": 0.7626381516456604,
      "reward_std": 0.026404659263789654,
      "reward_std_all": 0.03573635593056679,
      "rewards-max/forensics_sinkhorn_reward": 0.9825809001922607,
      "rewards-max/forensics_strict_edge_reward": 0.4917614609003067,
      "rewards-max/forensics_tiou_reward": 0.9831755459308624,
      "rewards/forensics_sinkhorn_reward": 0.6551982164382935,
      "rewards/forensics_strict_edge_reward": 0.09151551499962807,
      "rewards/forensics_tiou_reward": 0.5291748642921448,
      "step": 852
    },
    {
      "completion_length": 25.671875,
      "epoch": 2.6175115207373274,
      "grad_norm": 9.660345077514648,
      "kl": 0.4619140625,
      "learning_rate": 3.446153846153846e-07,
      "loss": 0.2065,
      "reward": 0.758335679769516,
      "reward_std": 0.03157241549342871,
      "reward_std_all": 0.040141511708498,
      "rewards-max/forensics_sinkhorn_reward": 0.9799339473247528,
      "rewards-max/forensics_strict_edge_reward": 0.37195122241973877,
      "rewards-max/forensics_tiou_reward": 0.975162535905838,
      "rewards/forensics_sinkhorn_reward": 0.6448052525520325,
      "rewards/forensics_strict_edge_reward": 0.04022164177149534,
      "rewards/forensics_tiou_reward": 0.49070650339126587,
      "step": 853
    },
    {
      "completion_length": 26.390625,
      "epoch": 2.6205837173579107,
      "grad_norm": 10.337597846984863,
      "kl": 0.62890625,
      "learning_rate": 3.4384615384615385e-07,
      "loss": -0.0809,
      "reward": 0.7756595611572266,
      "reward_std": 0.014131552772596478,
      "reward_std_all": 0.02409596648067236,
      "rewards-max/forensics_sinkhorn_reward": 0.9880473911762238,
      "rewards-max/forensics_strict_edge_reward": 0.9677534997463226,
      "rewards-max/forensics_tiou_reward": 0.9976431131362915,
      "rewards/forensics_sinkhorn_reward": 0.7728185057640076,
      "rewards/forensics_strict_edge_reward": 0.11364811658859253,
      "rewards/forensics_tiou_reward": 0.578251302242279,
      "step": 854
    },
    {
      "completion_length": 23.0625,
      "epoch": 2.6236559139784945,
      "grad_norm": 10.858332633972168,
      "kl": 0.4560546875,
      "learning_rate": 3.4307692307692307e-07,
      "loss": -0.1775,
      "reward": 0.776188313961029,
      "reward_std": 0.0152078066021204,
      "reward_std_all": 0.022548608481884003,
      "rewards-max/forensics_sinkhorn_reward": 0.9814395010471344,
      "rewards-max/forensics_strict_edge_reward": 0.6014226377010345,
      "rewards-max/forensics_tiou_reward": 0.9812106788158417,
      "rewards/forensics_sinkhorn_reward": 0.7503533065319061,
      "rewards/forensics_strict_edge_reward": 0.06547221913933754,
      "rewards/forensics_tiou_reward": 0.5627352297306061,
      "step": 855
    },
    {
      "completion_length": 26.328125,
      "epoch": 2.6267281105990783,
      "grad_norm": 9.647109031677246,
      "kl": 0.521484375,
      "learning_rate": 3.423076923076923e-07,
      "loss": 0.0462,
      "reward": 0.7973045706748962,
      "reward_std": 0.011717105517163873,
      "reward_std_all": 0.013067645020782948,
      "rewards-max/forensics_sinkhorn_reward": 0.972034364938736,
      "rewards-max/forensics_strict_edge_reward": 0.6188668608665466,
      "rewards-max/forensics_tiou_reward": 0.9574106633663177,
      "rewards/forensics_sinkhorn_reward": 0.719071239233017,
      "rewards/forensics_strict_edge_reward": 0.047792645171284676,
      "rewards/forensics_tiou_reward": 0.495551735162735,
      "step": 856
    },
    {
      "completion_length": 25.59375,
      "epoch": 2.629800307219662,
      "grad_norm": 9.63338565826416,
      "kl": 0.70703125,
      "learning_rate": 3.415384615384615e-07,
      "loss": 0.4185,
      "reward": 0.7677405774593353,
      "reward_std": 0.04809276573359966,
      "reward_std_all": 0.04622989147901535,
      "rewards-max/forensics_sinkhorn_reward": 0.9846272170543671,
      "rewards-max/forensics_strict_edge_reward": 0.8418737053871155,
      "rewards-max/forensics_tiou_reward": 0.9893312156200409,
      "rewards/forensics_sinkhorn_reward": 0.6540359258651733,
      "rewards/forensics_strict_edge_reward": 0.08432142995297909,
      "rewards/forensics_tiou_reward": 0.5578608512878418,
      "step": 857
    },
    {
      "completion_length": 26.0,
      "epoch": 2.632872503840246,
      "grad_norm": 10.19042682647705,
      "kl": 0.427734375,
      "learning_rate": 3.4076923076923076e-07,
      "loss": -0.0105,
      "reward": 0.7435713708400726,
      "reward_std": 0.031478650867938995,
      "reward_std_all": 0.045878393575549126,
      "rewards-max/forensics_sinkhorn_reward": 0.9720806777477264,
      "rewards-max/forensics_strict_edge_reward": 0.6274375021457672,
      "rewards-max/forensics_tiou_reward": 0.9580526053905487,
      "rewards/forensics_sinkhorn_reward": 0.6755455732345581,
      "rewards/forensics_strict_edge_reward": 0.08996732905507088,
      "rewards/forensics_tiou_reward": 0.4724280536174774,
      "step": 858
    },
    {
      "completion_length": 24.84375,
      "epoch": 2.6359447004608296,
      "grad_norm": 11.260661125183105,
      "kl": 0.544921875,
      "learning_rate": 3.4000000000000003e-07,
      "loss": 0.0518,
      "reward": 0.744766891002655,
      "reward_std": 0.022701053880155087,
      "reward_std_all": 0.04365735501050949,
      "rewards-max/forensics_sinkhorn_reward": 0.9801835119724274,
      "rewards-max/forensics_strict_edge_reward": 0.48720505833625793,
      "rewards-max/forensics_tiou_reward": 0.9770364165306091,
      "rewards/forensics_sinkhorn_reward": 0.7345510721206665,
      "rewards/forensics_strict_edge_reward": 0.07733399420976639,
      "rewards/forensics_tiou_reward": 0.566619411110878,
      "step": 859
    },
    {
      "completion_length": 28.546875,
      "epoch": 2.6390168970814134,
      "grad_norm": 9.069779396057129,
      "kl": 0.591796875,
      "learning_rate": 3.392307692307692e-07,
      "loss": 0.1527,
      "reward": 0.7297816276550293,
      "reward_std": 0.046327786054462194,
      "reward_std_all": 0.06436919700354338,
      "rewards-max/forensics_sinkhorn_reward": 0.9779162406921387,
      "rewards-max/forensics_strict_edge_reward": 0.6666710078716278,
      "rewards-max/forensics_tiou_reward": 0.9726923704147339,
      "rewards/forensics_sinkhorn_reward": 0.6560535728931427,
      "rewards/forensics_strict_edge_reward": 0.07261200249195099,
      "rewards/forensics_tiou_reward": 0.5278148651123047,
      "step": 860
    },
    {
      "completion_length": 23.734375,
      "epoch": 2.6420890937019967,
      "grad_norm": 9.955589294433594,
      "kl": 0.5556640625,
      "learning_rate": 3.3846153846153845e-07,
      "loss": 0.0982,
      "reward": 0.7825644910335541,
      "reward_std": 0.037240657955408096,
      "reward_std_all": 0.034428153187036514,
      "rewards-max/forensics_sinkhorn_reward": 0.974354088306427,
      "rewards-max/forensics_strict_edge_reward": 0.45128685235977173,
      "rewards-max/forensics_tiou_reward": 0.9625149071216583,
      "rewards/forensics_sinkhorn_reward": 0.7620202600955963,
      "rewards/forensics_strict_edge_reward": 0.08132137916982174,
      "rewards/forensics_tiou_reward": 0.584664523601532,
      "step": 861
    },
    {
      "completion_length": 25.234375,
      "epoch": 2.6451612903225805,
      "grad_norm": 12.77173900604248,
      "kl": 0.4521484375,
      "learning_rate": 3.3769230769230767e-07,
      "loss": 0.0028,
      "reward": 0.718127965927124,
      "reward_std": 0.045707935467362404,
      "reward_std_all": 0.0633893683552742,
      "rewards-max/forensics_sinkhorn_reward": 0.9781230986118317,
      "rewards-max/forensics_strict_edge_reward": 0.7165639698505402,
      "rewards-max/forensics_tiou_reward": 0.9735158085823059,
      "rewards/forensics_sinkhorn_reward": 0.6442696452140808,
      "rewards/forensics_strict_edge_reward": 0.07458221167325974,
      "rewards/forensics_tiou_reward": 0.4856446087360382,
      "step": 862
    },
    {
      "completion_length": 24.859375,
      "epoch": 2.6482334869431643,
      "grad_norm": 10.763263702392578,
      "kl": 0.4501953125,
      "learning_rate": 3.3692307692307693e-07,
      "loss": 0.0069,
      "reward": 0.7721766531467438,
      "reward_std": 0.024954643100500107,
      "reward_std_all": 0.030966809950768948,
      "rewards-max/forensics_sinkhorn_reward": 0.9851487874984741,
      "rewards-max/forensics_strict_edge_reward": 0.5470435917377472,
      "rewards-max/forensics_tiou_reward": 0.9885019659996033,
      "rewards/forensics_sinkhorn_reward": 0.7102630436420441,
      "rewards/forensics_strict_edge_reward": 0.09106917679309845,
      "rewards/forensics_tiou_reward": 0.6100459396839142,
      "step": 863
    },
    {
      "completion_length": 26.171875,
      "epoch": 2.651305683563748,
      "grad_norm": 10.668020248413086,
      "kl": 0.734375,
      "learning_rate": 3.361538461538461e-07,
      "loss": 0.1673,
      "reward": 0.7905551791191101,
      "reward_std": 0.026013354770839214,
      "reward_std_all": 0.02453605178743601,
      "rewards-max/forensics_sinkhorn_reward": 0.9842628836631775,
      "rewards-max/forensics_strict_edge_reward": 0.5829301476478577,
      "rewards-max/forensics_tiou_reward": 0.9863501191139221,
      "rewards/forensics_sinkhorn_reward": 0.7272416949272156,
      "rewards/forensics_strict_edge_reward": 0.10134666040539742,
      "rewards/forensics_tiou_reward": 0.5953335165977478,
      "step": 864
    },
    {
      "completion_length": 23.5625,
      "epoch": 2.654377880184332,
      "grad_norm": 10.357695579528809,
      "kl": 0.54296875,
      "learning_rate": 3.3538461538461536e-07,
      "loss": -0.0748,
      "reward": 0.7847015261650085,
      "reward_std": 0.029025171883404255,
      "reward_std_all": 0.02804678212851286,
      "rewards-max/forensics_sinkhorn_reward": 0.9819933772087097,
      "rewards-max/forensics_strict_edge_reward": 0.5972513854503632,
      "rewards-max/forensics_tiou_reward": 0.9800315201282501,
      "rewards/forensics_sinkhorn_reward": 0.7464413046836853,
      "rewards/forensics_strict_edge_reward": 0.07027145847678185,
      "rewards/forensics_tiou_reward": 0.5569757521152496,
      "step": 865
    },
    {
      "completion_length": 25.703125,
      "epoch": 2.6574500768049156,
      "grad_norm": 8.21828842163086,
      "kl": 0.5859375,
      "learning_rate": 3.346153846153846e-07,
      "loss": 0.0009,
      "reward": 0.781063973903656,
      "reward_std": 0.04339101351797581,
      "reward_std_all": 0.03952496126294136,
      "rewards-max/forensics_sinkhorn_reward": 0.9809049367904663,
      "rewards-max/forensics_strict_edge_reward": 0.5813090205192566,
      "rewards-max/forensics_tiou_reward": 0.977433830499649,
      "rewards/forensics_sinkhorn_reward": 0.7655418813228607,
      "rewards/forensics_strict_edge_reward": 0.08555423840880394,
      "rewards/forensics_tiou_reward": 0.6087403297424316,
      "step": 866
    },
    {
      "completion_length": 23.796875,
      "epoch": 2.6605222734254994,
      "grad_norm": 10.173506736755371,
      "kl": 0.66015625,
      "learning_rate": 3.3384615384615384e-07,
      "loss": -0.0505,
      "reward": 0.7432775795459747,
      "reward_std": 0.02656538935843855,
      "reward_std_all": 0.04477735701948404,
      "rewards-max/forensics_sinkhorn_reward": 0.9681352376937866,
      "rewards-max/forensics_strict_edge_reward": 0.7565742135047913,
      "rewards-max/forensics_tiou_reward": 0.9521730840206146,
      "rewards/forensics_sinkhorn_reward": 0.6460874080657959,
      "rewards/forensics_strict_edge_reward": 0.09711124747991562,
      "rewards/forensics_tiou_reward": 0.4386938363313675,
      "step": 867
    },
    {
      "completion_length": 26.984375,
      "epoch": 2.6635944700460827,
      "grad_norm": 8.3209867477417,
      "kl": 0.5361328125,
      "learning_rate": 3.330769230769231e-07,
      "loss": -0.0799,
      "reward": 0.7851834893226624,
      "reward_std": 0.029169853776693344,
      "reward_std_all": 0.027868769131600857,
      "rewards-max/forensics_sinkhorn_reward": 0.9860181212425232,
      "rewards-max/forensics_strict_edge_reward": 0.459838330745697,
      "rewards-max/forensics_tiou_reward": 0.9901362955570221,
      "rewards/forensics_sinkhorn_reward": 0.7306198179721832,
      "rewards/forensics_strict_edge_reward": 0.07034002244472504,
      "rewards/forensics_tiou_reward": 0.5419031977653503,
      "step": 868
    },
    {
      "completion_length": 25.015625,
      "epoch": 2.6666666666666665,
      "grad_norm": 13.057233810424805,
      "kl": 0.4912109375,
      "learning_rate": 3.3230769230769227e-07,
      "loss": -0.0653,
      "reward": 0.7710899710655212,
      "reward_std": 0.026639029383659363,
      "reward_std_all": 0.03182474710047245,
      "rewards-max/forensics_sinkhorn_reward": 0.9882086217403412,
      "rewards-max/forensics_strict_edge_reward": 0.824172168970108,
      "rewards-max/forensics_tiou_reward": 0.997759222984314,
      "rewards/forensics_sinkhorn_reward": 0.7498583793640137,
      "rewards/forensics_strict_edge_reward": 0.09879285097122192,
      "rewards/forensics_tiou_reward": 0.5607201755046844,
      "step": 869
    },
    {
      "completion_length": 23.25,
      "epoch": 2.6697388632872503,
      "grad_norm": 8.584007263183594,
      "kl": 0.537109375,
      "learning_rate": 3.3153846153846154e-07,
      "loss": -0.0589,
      "reward": 0.7911846339702606,
      "reward_std": 0.011122627882286906,
      "reward_std_all": 0.01495773484930396,
      "rewards-max/forensics_sinkhorn_reward": 0.9848944842815399,
      "rewards-max/forensics_strict_edge_reward": 0.7244620621204376,
      "rewards-max/forensics_tiou_reward": 0.9882415533065796,
      "rewards/forensics_sinkhorn_reward": 0.7521028220653534,
      "rewards/forensics_strict_edge_reward": 0.09387410432100296,
      "rewards/forensics_tiou_reward": 0.5999959707260132,
      "step": 870
    },
    {
      "completion_length": 27.265625,
      "epoch": 2.672811059907834,
      "grad_norm": 11.917552947998047,
      "kl": 0.6982421875,
      "learning_rate": 3.3076923076923075e-07,
      "loss": -0.1045,
      "reward": 0.7283750176429749,
      "reward_std": 0.03919653408229351,
      "reward_std_all": 0.059947993606328964,
      "rewards-max/forensics_sinkhorn_reward": 0.9836196303367615,
      "rewards-max/forensics_strict_edge_reward": 0.598973348736763,
      "rewards-max/forensics_tiou_reward": 0.9863943755626678,
      "rewards/forensics_sinkhorn_reward": 0.6550782322883606,
      "rewards/forensics_strict_edge_reward": 0.09014034643769264,
      "rewards/forensics_tiou_reward": 0.4866252839565277,
      "step": 871
    },
    {
      "completion_length": 22.09375,
      "epoch": 2.675883256528418,
      "grad_norm": 9.310062408447266,
      "kl": 0.4755859375,
      "learning_rate": 3.3e-07,
      "loss": -0.0478,
      "reward": 0.7275511920452118,
      "reward_std": 0.06932362727820873,
      "reward_std_all": 0.07351235672831535,
      "rewards-max/forensics_sinkhorn_reward": 0.9758844077587128,
      "rewards-max/forensics_strict_edge_reward": 0.6505971103906631,
      "rewards-max/forensics_tiou_reward": 0.9628866910934448,
      "rewards/forensics_sinkhorn_reward": 0.6830758452415466,
      "rewards/forensics_strict_edge_reward": 0.08602645806968212,
      "rewards/forensics_tiou_reward": 0.5290116667747498,
      "step": 872
    },
    {
      "completion_length": 23.4375,
      "epoch": 2.6789554531490016,
      "grad_norm": 9.377123832702637,
      "kl": 0.3857421875,
      "learning_rate": 3.292307692307692e-07,
      "loss": -0.0534,
      "reward": 0.7932745814323425,
      "reward_std": 0.02105956058949232,
      "reward_std_all": 0.02024768479168415,
      "rewards-max/forensics_sinkhorn_reward": 0.9839857816696167,
      "rewards-max/forensics_strict_edge_reward": 0.7026884853839874,
      "rewards-max/forensics_tiou_reward": 0.9859989285469055,
      "rewards/forensics_sinkhorn_reward": 0.7618825435638428,
      "rewards/forensics_strict_edge_reward": 0.08754871413111687,
      "rewards/forensics_tiou_reward": 0.6489973962306976,
      "step": 873
    },
    {
      "completion_length": 25.546875,
      "epoch": 2.6820276497695854,
      "grad_norm": 10.932950973510742,
      "kl": 0.4765625,
      "learning_rate": 3.2846153846153844e-07,
      "loss": 0.1226,
      "reward": 0.739291250705719,
      "reward_std": 0.03570437431335449,
      "reward_std_all": 0.0484761968255043,
      "rewards-max/forensics_sinkhorn_reward": 0.9784140884876251,
      "rewards-max/forensics_strict_edge_reward": 0.5844750255346298,
      "rewards-max/forensics_tiou_reward": 0.9721134603023529,
      "rewards/forensics_sinkhorn_reward": 0.6775039732456207,
      "rewards/forensics_strict_edge_reward": 0.055526284500956535,
      "rewards/forensics_tiou_reward": 0.4929490387439728,
      "step": 874
    },
    {
      "completion_length": 25.21875,
      "epoch": 2.6850998463901687,
      "grad_norm": 10.181096076965332,
      "kl": 0.625,
      "learning_rate": 3.2769230769230766e-07,
      "loss": 0.1053,
      "reward": 0.78457972407341,
      "reward_std": 0.02595531684346497,
      "reward_std_all": 0.026391152990981936,
      "rewards-max/forensics_sinkhorn_reward": 0.9611918330192566,
      "rewards-max/forensics_strict_edge_reward": 0.6735590398311615,
      "rewards-max/forensics_tiou_reward": 0.9334003627300262,
      "rewards/forensics_sinkhorn_reward": 0.7316974401473999,
      "rewards/forensics_strict_edge_reward": 0.05281962454319,
      "rewards/forensics_tiou_reward": 0.5123312920331955,
      "step": 875
    },
    {
      "completion_length": 26.109375,
      "epoch": 2.688172043010753,
      "grad_norm": 24.131786346435547,
      "kl": 0.689453125,
      "learning_rate": 3.269230769230769e-07,
      "loss": 0.0876,
      "reward": 0.752017617225647,
      "reward_std": 0.052765464410185814,
      "reward_std_all": 0.05768749676644802,
      "rewards-max/forensics_sinkhorn_reward": 0.982894092798233,
      "rewards-max/forensics_strict_edge_reward": 0.4851513057947159,
      "rewards-max/forensics_tiou_reward": 0.9781172275543213,
      "rewards/forensics_sinkhorn_reward": 0.702238142490387,
      "rewards/forensics_strict_edge_reward": 0.046191995963454247,
      "rewards/forensics_tiou_reward": 0.5100109279155731,
      "step": 876
    },
    {
      "completion_length": 22.875,
      "epoch": 2.6912442396313363,
      "grad_norm": 9.9088716506958,
      "kl": 0.58203125,
      "learning_rate": 3.2615384615384614e-07,
      "loss": -0.0912,
      "reward": 0.7847720980644226,
      "reward_std": 0.010064569767564535,
      "reward_std_all": 0.01669110730290413,
      "rewards-max/forensics_sinkhorn_reward": 0.9865045845508575,
      "rewards-max/forensics_strict_edge_reward": 0.8608275353908539,
      "rewards-max/forensics_tiou_reward": 0.9932883679866791,
      "rewards/forensics_sinkhorn_reward": 0.7618214190006256,
      "rewards/forensics_strict_edge_reward": 0.1250014863908291,
      "rewards/forensics_tiou_reward": 0.5996325016021729,
      "step": 877
    },
    {
      "completion_length": 22.328125,
      "epoch": 2.69431643625192,
      "grad_norm": 10.432679176330566,
      "kl": 0.552734375,
      "learning_rate": 3.2538461538461535e-07,
      "loss": -0.1462,
      "reward": 0.7721695601940155,
      "reward_std": 0.043058031937107444,
      "reward_std_all": 0.041318153496831656,
      "rewards-max/forensics_sinkhorn_reward": 0.9852616190910339,
      "rewards-max/forensics_strict_edge_reward": 0.670350581407547,
      "rewards-max/forensics_tiou_reward": 0.9897796213626862,
      "rewards/forensics_sinkhorn_reward": 0.7694630026817322,
      "rewards/forensics_strict_edge_reward": 0.13862962275743484,
      "rewards/forensics_tiou_reward": 0.6098763942718506,
      "step": 878
    },
    {
      "completion_length": 23.609375,
      "epoch": 2.697388632872504,
      "grad_norm": 10.045900344848633,
      "kl": 0.59765625,
      "learning_rate": 3.246153846153846e-07,
      "loss": 0.136,
      "reward": 0.7776282131671906,
      "reward_std": 0.02496999641880393,
      "reward_std_all": 0.029432824812829494,
      "rewards-max/forensics_sinkhorn_reward": 0.9798094034194946,
      "rewards-max/forensics_strict_edge_reward": 0.580880731344223,
      "rewards-max/forensics_tiou_reward": 0.973273754119873,
      "rewards/forensics_sinkhorn_reward": 0.7248512506484985,
      "rewards/forensics_strict_edge_reward": 0.06120758503675461,
      "rewards/forensics_tiou_reward": 0.5666192173957825,
      "step": 879
    },
    {
      "completion_length": 27.484375,
      "epoch": 2.7004608294930876,
      "grad_norm": 10.898778915405273,
      "kl": 0.537109375,
      "learning_rate": 3.2384615384615383e-07,
      "loss": 0.0638,
      "reward": 0.7762983441352844,
      "reward_std": 0.03301512636244297,
      "reward_std_all": 0.033880929462611675,
      "rewards-max/forensics_sinkhorn_reward": 0.9825751483440399,
      "rewards-max/forensics_strict_edge_reward": 0.6082755923271179,
      "rewards-max/forensics_tiou_reward": 0.9831344485282898,
      "rewards/forensics_sinkhorn_reward": 0.7438390851020813,
      "rewards/forensics_strict_edge_reward": 0.0911477729678154,
      "rewards/forensics_tiou_reward": 0.595063328742981,
      "step": 880
    },
    {
      "completion_length": 23.359375,
      "epoch": 2.7035330261136714,
      "grad_norm": 8.370628356933594,
      "kl": 0.529296875,
      "learning_rate": 3.230769230769231e-07,
      "loss": 0.2145,
      "reward": 0.7531152963638306,
      "reward_std": 0.05310897156596184,
      "reward_std_all": 0.05466720834374428,
      "rewards-max/forensics_sinkhorn_reward": 0.9677965044975281,
      "rewards-max/forensics_strict_edge_reward": 0.5895819067955017,
      "rewards-max/forensics_tiou_reward": 0.9506665766239166,
      "rewards/forensics_sinkhorn_reward": 0.70295649766922,
      "rewards/forensics_strict_edge_reward": 0.0803120769560337,
      "rewards/forensics_tiou_reward": 0.4833041578531265,
      "step": 881
    },
    {
      "completion_length": 25.34375,
      "epoch": 2.7066052227342547,
      "grad_norm": 13.157440185546875,
      "kl": 1.2275390625,
      "learning_rate": 3.2230769230769226e-07,
      "loss": 0.1298,
      "reward": 0.751970112323761,
      "reward_std": 0.0265790531411767,
      "reward_std_all": 0.040161299519240856,
      "rewards-max/forensics_sinkhorn_reward": 0.9849522709846497,
      "rewards-max/forensics_strict_edge_reward": 0.8959447741508484,
      "rewards-max/forensics_tiou_reward": 0.9903245270252228,
      "rewards/forensics_sinkhorn_reward": 0.7378002107143402,
      "rewards/forensics_strict_edge_reward": 0.09132902696728706,
      "rewards/forensics_tiou_reward": 0.5433843731880188,
      "step": 882
    },
    {
      "completion_length": 25.921875,
      "epoch": 2.709677419354839,
      "grad_norm": 8.9498872756958,
      "kl": 0.6025390625,
      "learning_rate": 3.215384615384615e-07,
      "loss": 0.0386,
      "reward": 0.775014728307724,
      "reward_std": 0.037511939532123506,
      "reward_std_all": 0.03667119541205466,
      "rewards-max/forensics_sinkhorn_reward": 0.9857535660266876,
      "rewards-max/forensics_strict_edge_reward": 0.8617841005325317,
      "rewards-max/forensics_tiou_reward": 0.9940324425697327,
      "rewards/forensics_sinkhorn_reward": 0.7602419257164001,
      "rewards/forensics_strict_edge_reward": 0.12293742224574089,
      "rewards/forensics_tiou_reward": 0.623824805021286,
      "step": 883
    },
    {
      "completion_length": 28.515625,
      "epoch": 2.7127496159754223,
      "grad_norm": 8.540913581848145,
      "kl": 0.646484375,
      "learning_rate": 3.2076923076923074e-07,
      "loss": 0.171,
      "reward": 0.7443719506263733,
      "reward_std": 0.03652871586382389,
      "reward_std_all": 0.050838084891438484,
      "rewards-max/forensics_sinkhorn_reward": 0.9673136472702026,
      "rewards-max/forensics_strict_edge_reward": 0.6029106974601746,
      "rewards-max/forensics_tiou_reward": 0.9451044499874115,
      "rewards/forensics_sinkhorn_reward": 0.6975981295108795,
      "rewards/forensics_strict_edge_reward": 0.05966814886778593,
      "rewards/forensics_tiou_reward": 0.5353531688451767,
      "step": 884
    },
    {
      "completion_length": 26.890625,
      "epoch": 2.715821812596006,
      "grad_norm": 9.437710762023926,
      "kl": 0.685546875,
      "learning_rate": 3.2e-07,
      "loss": 0.1337,
      "reward": 0.7398887276649475,
      "reward_std": 0.06903484091162682,
      "reward_std_all": 0.06780504062771797,
      "rewards-max/forensics_sinkhorn_reward": 0.9778977632522583,
      "rewards-max/forensics_strict_edge_reward": 0.6637322008609772,
      "rewards-max/forensics_tiou_reward": 0.96809321641922,
      "rewards/forensics_sinkhorn_reward": 0.66834956407547,
      "rewards/forensics_strict_edge_reward": 0.07527562975883484,
      "rewards/forensics_tiou_reward": 0.5186923742294312,
      "step": 885
    },
    {
      "completion_length": 25.109375,
      "epoch": 2.71889400921659,
      "grad_norm": 8.112653732299805,
      "kl": 0.4091796875,
      "learning_rate": 3.1923076923076917e-07,
      "loss": 0.0412,
      "reward": 0.7613032460212708,
      "reward_std": 0.03266558004543185,
      "reward_std_all": 0.03904082719236612,
      "rewards-max/forensics_sinkhorn_reward": 0.9805528521537781,
      "rewards-max/forensics_strict_edge_reward": 0.5274569988250732,
      "rewards-max/forensics_tiou_reward": 0.9766847491264343,
      "rewards/forensics_sinkhorn_reward": 0.7305839657783508,
      "rewards/forensics_strict_edge_reward": 0.09577112458646297,
      "rewards/forensics_tiou_reward": 0.5597183704376221,
      "step": 886
    },
    {
      "completion_length": 23.21875,
      "epoch": 2.7219662058371736,
      "grad_norm": 13.432151794433594,
      "kl": 0.4619140625,
      "learning_rate": 3.1846153846153843e-07,
      "loss": -0.1395,
      "reward": 0.7708941698074341,
      "reward_std": 0.016343056224286556,
      "reward_std_all": 0.02593611367046833,
      "rewards-max/forensics_sinkhorn_reward": 0.9601215124130249,
      "rewards-max/forensics_strict_edge_reward": 0.33779143542051315,
      "rewards-max/forensics_tiou_reward": 0.9285487234592438,
      "rewards/forensics_sinkhorn_reward": 0.7205386161804199,
      "rewards/forensics_strict_edge_reward": 0.049938717391341925,
      "rewards/forensics_tiou_reward": 0.5039217174053192,
      "step": 887
    },
    {
      "completion_length": 21.78125,
      "epoch": 2.7250384024577574,
      "grad_norm": 9.33926773071289,
      "kl": 0.4853515625,
      "learning_rate": 3.176923076923077e-07,
      "loss": -0.0987,
      "reward": 0.7666634321212769,
      "reward_std": 0.017388093750923872,
      "reward_std_all": 0.03156346268951893,
      "rewards-max/forensics_sinkhorn_reward": 0.9753691554069519,
      "rewards-max/forensics_strict_edge_reward": 0.6839397251605988,
      "rewards-max/forensics_tiou_reward": 0.9661369025707245,
      "rewards/forensics_sinkhorn_reward": 0.7805997431278229,
      "rewards/forensics_strict_edge_reward": 0.07563776150345802,
      "rewards/forensics_tiou_reward": 0.6151678562164307,
      "step": 888
    },
    {
      "completion_length": 24.890625,
      "epoch": 2.7281105990783407,
      "grad_norm": 12.329784393310547,
      "kl": 0.708984375,
      "learning_rate": 3.169230769230769e-07,
      "loss": -0.0427,
      "reward": 0.7398850619792938,
      "reward_std": 0.056796152144670486,
      "reward_std_all": 0.06168123707175255,
      "rewards-max/forensics_sinkhorn_reward": 0.9662491679191589,
      "rewards-max/forensics_strict_edge_reward": 0.7310135364532471,
      "rewards-max/forensics_tiou_reward": 0.9445401430130005,
      "rewards/forensics_sinkhorn_reward": 0.6875413954257965,
      "rewards/forensics_strict_edge_reward": 0.10493568331003189,
      "rewards/forensics_tiou_reward": 0.48900267481803894,
      "step": 889
    },
    {
      "completion_length": 24.90625,
      "epoch": 2.731182795698925,
      "grad_norm": 11.835711479187012,
      "kl": 0.326171875,
      "learning_rate": 3.161538461538462e-07,
      "loss": -0.012,
      "reward": 0.7636541724205017,
      "reward_std": 0.023699039593338966,
      "reward_std_all": 0.032141102477908134,
      "rewards-max/forensics_sinkhorn_reward": 0.9751457870006561,
      "rewards-max/forensics_strict_edge_reward": 0.43032702803611755,
      "rewards-max/forensics_tiou_reward": 0.966536819934845,
      "rewards/forensics_sinkhorn_reward": 0.7071916460990906,
      "rewards/forensics_strict_edge_reward": 0.06056860089302063,
      "rewards/forensics_tiou_reward": 0.515119656920433,
      "step": 890
    },
    {
      "completion_length": 25.65625,
      "epoch": 2.7342549923195083,
      "grad_norm": 12.753255844116211,
      "kl": 0.9111328125,
      "learning_rate": 3.1538461538461534e-07,
      "loss": -0.0408,
      "reward": 0.745188981294632,
      "reward_std": 0.04060332290828228,
      "reward_std_all": 0.050485437735915184,
      "rewards-max/forensics_sinkhorn_reward": 0.9878764152526855,
      "rewards-max/forensics_strict_edge_reward": 0.9201722145080566,
      "rewards-max/forensics_tiou_reward": 0.9969423413276672,
      "rewards/forensics_sinkhorn_reward": 0.7385973632335663,
      "rewards/forensics_strict_edge_reward": 0.12306907027959824,
      "rewards/forensics_tiou_reward": 0.5573552548885345,
      "step": 891
    },
    {
      "completion_length": 24.140625,
      "epoch": 2.737327188940092,
      "grad_norm": 11.685639381408691,
      "kl": 0.6396484375,
      "learning_rate": 3.146153846153846e-07,
      "loss": -0.0685,
      "reward": 0.7645426690578461,
      "reward_std": 0.03475529979914427,
      "reward_std_all": 0.039524239487946033,
      "rewards-max/forensics_sinkhorn_reward": 0.9854018688201904,
      "rewards-max/forensics_strict_edge_reward": 0.7542104423046112,
      "rewards-max/forensics_tiou_reward": 0.9896702766418457,
      "rewards/forensics_sinkhorn_reward": 0.7473359107971191,
      "rewards/forensics_strict_edge_reward": 0.10470669716596603,
      "rewards/forensics_tiou_reward": 0.5751511454582214,
      "step": 892
    },
    {
      "completion_length": 21.75,
      "epoch": 2.740399385560676,
      "grad_norm": 11.267704963684082,
      "kl": 0.607421875,
      "learning_rate": 3.138461538461538e-07,
      "loss": 0.3628,
      "reward": 0.742933839559555,
      "reward_std": 0.04760919604450464,
      "reward_std_all": 0.05546143651008606,
      "rewards-max/forensics_sinkhorn_reward": 0.9776101410388947,
      "rewards-max/forensics_strict_edge_reward": 0.6839397251605988,
      "rewards-max/forensics_tiou_reward": 0.9730555713176727,
      "rewards/forensics_sinkhorn_reward": 0.6578704118728638,
      "rewards/forensics_strict_edge_reward": 0.047348495572805405,
      "rewards/forensics_tiou_reward": 0.44395747780799866,
      "step": 893
    },
    {
      "completion_length": 25.9375,
      "epoch": 2.7434715821812596,
      "grad_norm": 10.763030052185059,
      "kl": 0.46875,
      "learning_rate": 3.130769230769231e-07,
      "loss": -0.1685,
      "reward": 0.7659470438957214,
      "reward_std": 0.029889723286032677,
      "reward_std_all": 0.03834337554872036,
      "rewards-max/forensics_sinkhorn_reward": 0.9846719205379486,
      "rewards-max/forensics_strict_edge_reward": 0.6032288670539856,
      "rewards-max/forensics_tiou_reward": 0.9892145097255707,
      "rewards/forensics_sinkhorn_reward": 0.7398848235607147,
      "rewards/forensics_strict_edge_reward": 0.08888478577136993,
      "rewards/forensics_tiou_reward": 0.5825648009777069,
      "step": 894
    },
    {
      "completion_length": 26.65625,
      "epoch": 2.7465437788018434,
      "grad_norm": 8.361323356628418,
      "kl": 0.806640625,
      "learning_rate": 3.1230769230769225e-07,
      "loss": 0.1537,
      "reward": 0.7722070515155792,
      "reward_std": 0.03890063054859638,
      "reward_std_all": 0.03818936087191105,
      "rewards-max/forensics_sinkhorn_reward": 0.983148992061615,
      "rewards-max/forensics_strict_edge_reward": 0.6092631667852402,
      "rewards-max/forensics_tiou_reward": 0.9847958385944366,
      "rewards/forensics_sinkhorn_reward": 0.7041885554790497,
      "rewards/forensics_strict_edge_reward": 0.07551166787743568,
      "rewards/forensics_tiou_reward": 0.5387943387031555,
      "step": 895
    },
    {
      "completion_length": 24.90625,
      "epoch": 2.749615975422427,
      "grad_norm": 9.29175090789795,
      "kl": 0.5830078125,
      "learning_rate": 3.115384615384615e-07,
      "loss": -0.0478,
      "reward": 0.7981041371822357,
      "reward_std": 0.014364702627062798,
      "reward_std_all": 0.014286681544035673,
      "rewards-max/forensics_sinkhorn_reward": 0.9702348709106445,
      "rewards-max/forensics_strict_edge_reward": 0.5496069490909576,
      "rewards-max/forensics_tiou_reward": 0.9576009809970856,
      "rewards/forensics_sinkhorn_reward": 0.7664465308189392,
      "rewards/forensics_strict_edge_reward": 0.08196217566728592,
      "rewards/forensics_tiou_reward": 0.6323586702346802,
      "step": 896
    },
    {
      "completion_length": 25.421875,
      "epoch": 2.752688172043011,
      "grad_norm": 9.992111206054688,
      "kl": 0.4580078125,
      "learning_rate": 3.1076923076923073e-07,
      "loss": 0.159,
      "reward": 0.760577917098999,
      "reward_std": 0.05970475822687149,
      "reward_std_all": 0.0571402944624424,
      "rewards-max/forensics_sinkhorn_reward": 0.9874861836433411,
      "rewards-max/forensics_strict_edge_reward": 0.7745758593082428,
      "rewards-max/forensics_tiou_reward": 0.9962947368621826,
      "rewards/forensics_sinkhorn_reward": 0.7412918210029602,
      "rewards/forensics_strict_edge_reward": 0.14540718123316765,
      "rewards/forensics_tiou_reward": 0.6267196536064148,
      "step": 897
    },
    {
      "completion_length": 23.46875,
      "epoch": 2.7557603686635943,
      "grad_norm": 11.270904541015625,
      "kl": 0.69921875,
      "learning_rate": 3.1e-07,
      "loss": 0.0295,
      "reward": 0.7539873123168945,
      "reward_std": 0.06543938629329205,
      "reward_std_all": 0.06346364505589008,
      "rewards-max/forensics_sinkhorn_reward": 0.986861914396286,
      "rewards-max/forensics_strict_edge_reward": 0.8221496343612671,
      "rewards-max/forensics_tiou_reward": 0.9954239726066589,
      "rewards/forensics_sinkhorn_reward": 0.7481542527675629,
      "rewards/forensics_strict_edge_reward": 0.10696272924542427,
      "rewards/forensics_tiou_reward": 0.5663795471191406,
      "step": 898
    },
    {
      "completion_length": 24.546875,
      "epoch": 2.758832565284178,
      "grad_norm": 39.56271743774414,
      "kl": 0.525390625,
      "learning_rate": 3.0923076923076926e-07,
      "loss": 0.0285,
      "reward": 0.7467203140258789,
      "reward_std": 0.05567758530378342,
      "reward_std_all": 0.058982957154512405,
      "rewards-max/forensics_sinkhorn_reward": 0.9850981831550598,
      "rewards-max/forensics_strict_edge_reward": 0.6934256851673126,
      "rewards-max/forensics_tiou_reward": 0.9887931942939758,
      "rewards/forensics_sinkhorn_reward": 0.7184978425502777,
      "rewards/forensics_strict_edge_reward": 0.08909297734498978,
      "rewards/forensics_tiou_reward": 0.5421924144029617,
      "step": 899
    },
    {
      "completion_length": 25.21875,
      "epoch": 2.761904761904762,
      "grad_norm": 10.427136421203613,
      "kl": 0.51953125,
      "learning_rate": 3.084615384615384e-07,
      "loss": 0.0878,
      "reward": 0.750744491815567,
      "reward_std": 0.078192800283432,
      "reward_std_all": 0.07173370569944382,
      "rewards-max/forensics_sinkhorn_reward": 0.9851876497268677,
      "rewards-max/forensics_strict_edge_reward": 0.7073164582252502,
      "rewards-max/forensics_tiou_reward": 0.9904463589191437,
      "rewards/forensics_sinkhorn_reward": 0.7010544240474701,
      "rewards/forensics_strict_edge_reward": 0.07790184393525124,
      "rewards/forensics_tiou_reward": 0.5569956004619598,
      "step": 900
    },
    {
      "completion_length": 25.796875,
      "epoch": 2.7649769585253456,
      "grad_norm": 12.558570861816406,
      "kl": 0.40625,
      "learning_rate": 3.076923076923077e-07,
      "loss": -0.0412,
      "reward": 0.7361905872821808,
      "reward_std": 0.02591159287840128,
      "reward_std_all": 0.048070117831230164,
      "rewards-max/forensics_sinkhorn_reward": 0.9850387871265411,
      "rewards-max/forensics_strict_edge_reward": 0.6515887677669525,
      "rewards-max/forensics_tiou_reward": 0.9886859655380249,
      "rewards/forensics_sinkhorn_reward": 0.6887798607349396,
      "rewards/forensics_strict_edge_reward": 0.09040708094835281,
      "rewards/forensics_tiou_reward": 0.5310027897357941,
      "step": 901
    },
    {
      "completion_length": 25.53125,
      "epoch": 2.7680491551459294,
      "grad_norm": 11.268178939819336,
      "kl": 0.5615234375,
      "learning_rate": 3.069230769230769e-07,
      "loss": 0.1265,
      "reward": 0.7526275515556335,
      "reward_std": 0.0483546257019043,
      "reward_std_all": 0.05129438079893589,
      "rewards-max/forensics_sinkhorn_reward": 0.9793622493743896,
      "rewards-max/forensics_strict_edge_reward": 0.7055691182613373,
      "rewards-max/forensics_tiou_reward": 0.9734194874763489,
      "rewards/forensics_sinkhorn_reward": 0.69993856549263,
      "rewards/forensics_strict_edge_reward": 0.08445252850651741,
      "rewards/forensics_tiou_reward": 0.5324806571006775,
      "step": 902
    },
    {
      "completion_length": 26.65625,
      "epoch": 2.771121351766513,
      "grad_norm": 8.930623054504395,
      "kl": 0.783203125,
      "learning_rate": 3.0615384615384617e-07,
      "loss": 0.1665,
      "reward": 0.7723359763622284,
      "reward_std": 0.03648820240050554,
      "reward_std_all": 0.03761539235711098,
      "rewards-max/forensics_sinkhorn_reward": 0.9861944317817688,
      "rewards-max/forensics_strict_edge_reward": 0.8326062560081482,
      "rewards-max/forensics_tiou_reward": 0.9917571246623993,
      "rewards/forensics_sinkhorn_reward": 0.7311864495277405,
      "rewards/forensics_strict_edge_reward": 0.09749547392129898,
      "rewards/forensics_tiou_reward": 0.5423036515712738,
      "step": 903
    },
    {
      "completion_length": 26.1875,
      "epoch": 2.774193548387097,
      "grad_norm": 8.323593139648438,
      "kl": 0.595703125,
      "learning_rate": 3.0538461538461533e-07,
      "loss": 0.0743,
      "reward": 0.7646610140800476,
      "reward_std": 0.05366552993655205,
      "reward_std_all": 0.05050742533057928,
      "rewards-max/forensics_sinkhorn_reward": 0.977061927318573,
      "rewards-max/forensics_strict_edge_reward": 0.5918685793876648,
      "rewards-max/forensics_tiou_reward": 0.9703916013240814,
      "rewards/forensics_sinkhorn_reward": 0.7197253406047821,
      "rewards/forensics_strict_edge_reward": 0.09576206468045712,
      "rewards/forensics_tiou_reward": 0.5335916578769684,
      "step": 904
    },
    {
      "completion_length": 25.796875,
      "epoch": 2.7772657450076803,
      "grad_norm": 9.534226417541504,
      "kl": 0.4130859375,
      "learning_rate": 3.046153846153846e-07,
      "loss": 0.1944,
      "reward": 0.7765412330627441,
      "reward_std": 0.030514865182340145,
      "reward_std_all": 0.0322694955393672,
      "rewards-max/forensics_sinkhorn_reward": 0.9792514741420746,
      "rewards-max/forensics_strict_edge_reward": 0.6183506101369858,
      "rewards-max/forensics_tiou_reward": 0.9716558456420898,
      "rewards/forensics_sinkhorn_reward": 0.6846189200878143,
      "rewards/forensics_strict_edge_reward": 0.05173273291438818,
      "rewards/forensics_tiou_reward": 0.5160589516162872,
      "step": 905
    },
    {
      "completion_length": 24.8125,
      "epoch": 2.780337941628264,
      "grad_norm": 10.067448616027832,
      "kl": 0.640625,
      "learning_rate": 3.038461538461538e-07,
      "loss": -0.0932,
      "reward": 0.7979380190372467,
      "reward_std": 0.014441067352890968,
      "reward_std_all": 0.014421505853533745,
      "rewards-max/forensics_sinkhorn_reward": 0.9823325276374817,
      "rewards-max/forensics_strict_edge_reward": 0.7727466821670532,
      "rewards-max/forensics_tiou_reward": 0.9829649031162262,
      "rewards/forensics_sinkhorn_reward": 0.8002220988273621,
      "rewards/forensics_strict_edge_reward": 0.08921327441930771,
      "rewards/forensics_tiou_reward": 0.6335941255092621,
      "step": 906
    },
    {
      "completion_length": 23.484375,
      "epoch": 2.783410138248848,
      "grad_norm": 11.019368171691895,
      "kl": 0.4892578125,
      "learning_rate": 3.030769230769231e-07,
      "loss": 0.1107,
      "reward": 0.7583251595497131,
      "reward_std": 0.027327903546392918,
      "reward_std_all": 0.03954040817916393,
      "rewards-max/forensics_sinkhorn_reward": 0.9838652908802032,
      "rewards-max/forensics_strict_edge_reward": 0.7294844686985016,
      "rewards-max/forensics_tiou_reward": 0.9869596660137177,
      "rewards/forensics_sinkhorn_reward": 0.6999389827251434,
      "rewards/forensics_strict_edge_reward": 0.10477781295776367,
      "rewards/forensics_tiou_reward": 0.531567707657814,
      "step": 907
    },
    {
      "completion_length": 24.296875,
      "epoch": 2.7864823348694316,
      "grad_norm": 15.199822425842285,
      "kl": 0.568359375,
      "learning_rate": 3.0230769230769235e-07,
      "loss": -0.2306,
      "reward": 0.7659741044044495,
      "reward_std": 0.02891514776274562,
      "reward_std_all": 0.03386581316590309,
      "rewards-max/forensics_sinkhorn_reward": 0.9849233031272888,
      "rewards-max/forensics_strict_edge_reward": 0.4424546957015991,
      "rewards-max/forensics_tiou_reward": 0.9863275587558746,
      "rewards/forensics_sinkhorn_reward": 0.7490664124488831,
      "rewards/forensics_strict_edge_reward": 0.07890484482049942,
      "rewards/forensics_tiou_reward": 0.5815132260322571,
      "step": 908
    },
    {
      "completion_length": 26.453125,
      "epoch": 2.7895545314900154,
      "grad_norm": 14.463375091552734,
      "kl": 0.443359375,
      "learning_rate": 3.015384615384615e-07,
      "loss": -0.0143,
      "reward": 0.7623918354511261,
      "reward_std": 0.013567181304097176,
      "reward_std_all": 0.028419493231922388,
      "rewards-max/forensics_sinkhorn_reward": 0.9825213849544525,
      "rewards-max/forensics_strict_edge_reward": 0.5824257880449295,
      "rewards-max/forensics_tiou_reward": 0.9861132502555847,
      "rewards/forensics_sinkhorn_reward": 0.7520244419574738,
      "rewards/forensics_strict_edge_reward": 0.06149819307029247,
      "rewards/forensics_tiou_reward": 0.5715409517288208,
      "step": 909
    },
    {
      "completion_length": 25.96875,
      "epoch": 2.792626728110599,
      "grad_norm": 9.172273635864258,
      "kl": 0.671875,
      "learning_rate": 3.0076923076923077e-07,
      "loss": -0.0447,
      "reward": 0.7365492284297943,
      "reward_std": 0.03205355326645076,
      "reward_std_all": 0.05011086701415479,
      "rewards-max/forensics_sinkhorn_reward": 0.9838671088218689,
      "rewards-max/forensics_strict_edge_reward": 0.595930814743042,
      "rewards-max/forensics_tiou_reward": 0.9820037186145782,
      "rewards/forensics_sinkhorn_reward": 0.7269633412361145,
      "rewards/forensics_strict_edge_reward": 0.07570827193558216,
      "rewards/forensics_tiou_reward": 0.5805322378873825,
      "step": 910
    },
    {
      "completion_length": 25.359375,
      "epoch": 2.795698924731183,
      "grad_norm": 11.492026329040527,
      "kl": 0.484375,
      "learning_rate": 3e-07,
      "loss": 0.018,
      "reward": 0.7222779095172882,
      "reward_std": 0.04417121037840843,
      "reward_std_all": 0.06586599722504616,
      "rewards-max/forensics_sinkhorn_reward": 0.9862079620361328,
      "rewards-max/forensics_strict_edge_reward": 0.5653136372566223,
      "rewards-max/forensics_tiou_reward": 0.992818295955658,
      "rewards/forensics_sinkhorn_reward": 0.7056935429573059,
      "rewards/forensics_strict_edge_reward": 0.049496978521347046,
      "rewards/forensics_tiou_reward": 0.5287244021892548,
      "step": 911
    },
    {
      "completion_length": 25.484375,
      "epoch": 2.7987711213517663,
      "grad_norm": 20.202306747436523,
      "kl": 0.6279296875,
      "learning_rate": 2.9923076923076925e-07,
      "loss": -0.273,
      "reward": 0.7734220623970032,
      "reward_std": 0.048460302874445915,
      "reward_std_all": 0.044632965698838234,
      "rewards-max/forensics_sinkhorn_reward": 0.9708192646503448,
      "rewards-max/forensics_strict_edge_reward": 0.5876468569040298,
      "rewards-max/forensics_tiou_reward": 0.954740583896637,
      "rewards/forensics_sinkhorn_reward": 0.752714216709137,
      "rewards/forensics_strict_edge_reward": 0.07677809707820415,
      "rewards/forensics_tiou_reward": 0.5603033304214478,
      "step": 912
    },
    {
      "completion_length": 28.515625,
      "epoch": 2.80184331797235,
      "grad_norm": 9.849528312683105,
      "kl": 0.630859375,
      "learning_rate": 2.984615384615384e-07,
      "loss": -0.0643,
      "reward": 0.7782610356807709,
      "reward_std": 0.012435679323971272,
      "reward_std_all": 0.022520122583955526,
      "rewards-max/forensics_sinkhorn_reward": 0.9767630100250244,
      "rewards-max/forensics_strict_edge_reward": 0.4555388540029526,
      "rewards-max/forensics_tiou_reward": 0.9714826345443726,
      "rewards/forensics_sinkhorn_reward": 0.7480870485305786,
      "rewards/forensics_strict_edge_reward": 0.09179031103849411,
      "rewards/forensics_tiou_reward": 0.5729217380285263,
      "step": 913
    },
    {
      "completion_length": 23.578125,
      "epoch": 2.804915514592934,
      "grad_norm": 11.317831039428711,
      "kl": 0.7578125,
      "learning_rate": 2.976923076923077e-07,
      "loss": 0.0451,
      "reward": 0.7712900042533875,
      "reward_std": 0.03939133509993553,
      "reward_std_all": 0.0389340091496706,
      "rewards-max/forensics_sinkhorn_reward": 0.9771162569522858,
      "rewards-max/forensics_strict_edge_reward": 0.6747751235961914,
      "rewards-max/forensics_tiou_reward": 0.9667057394981384,
      "rewards/forensics_sinkhorn_reward": 0.7237690985202789,
      "rewards/forensics_strict_edge_reward": 0.060638852417469025,
      "rewards/forensics_tiou_reward": 0.530892014503479,
      "step": 914
    },
    {
      "completion_length": 23.921875,
      "epoch": 2.8079877112135176,
      "grad_norm": 11.841866493225098,
      "kl": 0.470703125,
      "learning_rate": 2.969230769230769e-07,
      "loss": -0.0554,
      "reward": 0.7853429317474365,
      "reward_std": 0.02635961677879095,
      "reward_std_all": 0.026025408413261175,
      "rewards-max/forensics_sinkhorn_reward": 0.9783117175102234,
      "rewards-max/forensics_strict_edge_reward": 0.6174952685832977,
      "rewards-max/forensics_tiou_reward": 0.9718858897686005,
      "rewards/forensics_sinkhorn_reward": 0.7304064631462097,
      "rewards/forensics_strict_edge_reward": 0.054512400180101395,
      "rewards/forensics_tiou_reward": 0.5506678819656372,
      "step": 915
    },
    {
      "completion_length": 23.109375,
      "epoch": 2.8110599078341014,
      "grad_norm": 9.380657196044922,
      "kl": 0.7890625,
      "learning_rate": 2.9615384615384616e-07,
      "loss": -0.11,
      "reward": 0.7356233596801758,
      "reward_std": 0.02025989769026637,
      "reward_std_all": 0.04348804987967014,
      "rewards-max/forensics_sinkhorn_reward": 0.96953085064888,
      "rewards-max/forensics_strict_edge_reward": 0.6724361926317215,
      "rewards-max/forensics_tiou_reward": 0.948944628238678,
      "rewards/forensics_sinkhorn_reward": 0.6924930810928345,
      "rewards/forensics_strict_edge_reward": 0.08879491500556469,
      "rewards/forensics_tiou_reward": 0.5212976187467575,
      "step": 916
    },
    {
      "completion_length": 28.03125,
      "epoch": 2.814132104454685,
      "grad_norm": 9.046551704406738,
      "kl": 0.52734375,
      "learning_rate": 2.953846153846154e-07,
      "loss": -0.0756,
      "reward": 0.7636097371578217,
      "reward_std": 0.04768247529864311,
      "reward_std_all": 0.046998798847198486,
      "rewards-max/forensics_sinkhorn_reward": 0.9818823337554932,
      "rewards-max/forensics_strict_edge_reward": 0.4776906371116638,
      "rewards-max/forensics_tiou_reward": 0.9789772629737854,
      "rewards/forensics_sinkhorn_reward": 0.7331433892250061,
      "rewards/forensics_strict_edge_reward": 0.06683937460184097,
      "rewards/forensics_tiou_reward": 0.5584449768066406,
      "step": 917
    },
    {
      "completion_length": 22.3125,
      "epoch": 2.817204301075269,
      "grad_norm": 20.161836624145508,
      "kl": 1.197265625,
      "learning_rate": 2.946153846153846e-07,
      "loss": -0.0679,
      "reward": 0.7881590723991394,
      "reward_std": 0.010304499417543411,
      "reward_std_all": 0.01675821654498577,
      "rewards-max/forensics_sinkhorn_reward": 0.9707561135292053,
      "rewards-max/forensics_strict_edge_reward": 0.5957051217556,
      "rewards-max/forensics_tiou_reward": 0.9554891586303711,
      "rewards/forensics_sinkhorn_reward": 0.7624086737632751,
      "rewards/forensics_strict_edge_reward": 0.08892233297228813,
      "rewards/forensics_tiou_reward": 0.5847118198871613,
      "step": 918
    },
    {
      "completion_length": 24.4375,
      "epoch": 2.8202764976958523,
      "grad_norm": 13.119771003723145,
      "kl": 0.521484375,
      "learning_rate": 2.9384615384615385e-07,
      "loss": 0.0205,
      "reward": 0.7787257730960846,
      "reward_std": 0.02876970451325178,
      "reward_std_all": 0.029313436709344387,
      "rewards-max/forensics_sinkhorn_reward": 0.9805365800857544,
      "rewards-max/forensics_strict_edge_reward": 0.3858771324157715,
      "rewards-max/forensics_tiou_reward": 0.9784537553787231,
      "rewards/forensics_sinkhorn_reward": 0.7358832061290741,
      "rewards/forensics_strict_edge_reward": 0.03919639065861702,
      "rewards/forensics_tiou_reward": 0.5610276460647583,
      "step": 919
    },
    {
      "completion_length": 25.984375,
      "epoch": 2.823348694316436,
      "grad_norm": 11.534014701843262,
      "kl": 0.5498046875,
      "learning_rate": 2.9307692307692307e-07,
      "loss": 0.0649,
      "reward": 0.7733360826969147,
      "reward_std": 0.029630614444613457,
      "reward_std_all": 0.03417633846402168,
      "rewards-max/forensics_sinkhorn_reward": 0.9822382032871246,
      "rewards-max/forensics_strict_edge_reward": 0.7026884853839874,
      "rewards-max/forensics_tiou_reward": 0.9813024997711182,
      "rewards/forensics_sinkhorn_reward": 0.7384317815303802,
      "rewards/forensics_strict_edge_reward": 0.07144362106919289,
      "rewards/forensics_tiou_reward": 0.5580350756645203,
      "step": 920
    },
    {
      "completion_length": 28.78125,
      "epoch": 2.82642089093702,
      "grad_norm": 9.289445877075195,
      "kl": 0.4814453125,
      "learning_rate": 2.9230769230769234e-07,
      "loss": 0.0284,
      "reward": 0.7591943442821503,
      "reward_std": 0.040003351867198944,
      "reward_std_all": 0.042768267914652824,
      "rewards-max/forensics_sinkhorn_reward": 0.977876603603363,
      "rewards-max/forensics_strict_edge_reward": 0.43117815256118774,
      "rewards-max/forensics_tiou_reward": 0.9704727232456207,
      "rewards/forensics_sinkhorn_reward": 0.7173182964324951,
      "rewards/forensics_strict_edge_reward": 0.06800578534603119,
      "rewards/forensics_tiou_reward": 0.5753993988037109,
      "step": 921
    },
    {
      "completion_length": 23.640625,
      "epoch": 2.8294930875576036,
      "grad_norm": 9.36133861541748,
      "kl": 0.638671875,
      "learning_rate": 2.915384615384615e-07,
      "loss": -0.0517,
      "reward": 0.7669022083282471,
      "reward_std": 0.057542890310287476,
      "reward_std_all": 0.05260268226265907,
      "rewards-max/forensics_sinkhorn_reward": 0.9887281954288483,
      "rewards-max/forensics_strict_edge_reward": 0.9524013698101044,
      "rewards-max/forensics_tiou_reward": 0.9981772601604462,
      "rewards/forensics_sinkhorn_reward": 0.763850063085556,
      "rewards/forensics_strict_edge_reward": 0.09182357788085938,
      "rewards/forensics_tiou_reward": 0.6006966829299927,
      "step": 922
    },
    {
      "completion_length": 26.140625,
      "epoch": 2.8325652841781874,
      "grad_norm": 7.888514041900635,
      "kl": 0.548828125,
      "learning_rate": 2.9076923076923076e-07,
      "loss": 0.0293,
      "reward": 0.7266245484352112,
      "reward_std": 0.03947881329804659,
      "reward_std_all": 0.062079502269625664,
      "rewards-max/forensics_sinkhorn_reward": 0.9811137914657593,
      "rewards-max/forensics_strict_edge_reward": 0.6703200340270996,
      "rewards-max/forensics_tiou_reward": 0.9807909727096558,
      "rewards/forensics_sinkhorn_reward": 0.6888040602207184,
      "rewards/forensics_strict_edge_reward": 0.07887348532676697,
      "rewards/forensics_tiou_reward": 0.5518182516098022,
      "step": 923
    },
    {
      "completion_length": 26.578125,
      "epoch": 2.835637480798771,
      "grad_norm": 11.961081504821777,
      "kl": 0.544921875,
      "learning_rate": 2.9e-07,
      "loss": 0.0062,
      "reward": 0.7722028493881226,
      "reward_std": 0.04469837807118893,
      "reward_std_all": 0.04273239057511091,
      "rewards-max/forensics_sinkhorn_reward": 0.9723745584487915,
      "rewards-max/forensics_strict_edge_reward": 0.3536384403705597,
      "rewards-max/forensics_tiou_reward": 0.962494820356369,
      "rewards/forensics_sinkhorn_reward": 0.6883053481578827,
      "rewards/forensics_strict_edge_reward": 0.034483009949326515,
      "rewards/forensics_tiou_reward": 0.5428401082754135,
      "step": 924
    },
    {
      "completion_length": 22.203125,
      "epoch": 2.838709677419355,
      "grad_norm": 12.287701606750488,
      "kl": 0.55078125,
      "learning_rate": 2.8923076923076924e-07,
      "loss": 0.3236,
      "reward": 0.7557700574398041,
      "reward_std": 0.06945176422595978,
      "reward_std_all": 0.06411192007362843,
      "rewards-max/forensics_sinkhorn_reward": 0.9821499288082123,
      "rewards-max/forensics_strict_edge_reward": 0.7311047911643982,
      "rewards-max/forensics_tiou_reward": 0.9808081984519958,
      "rewards/forensics_sinkhorn_reward": 0.7225108742713928,
      "rewards/forensics_strict_edge_reward": 0.11351728066802025,
      "rewards/forensics_tiou_reward": 0.5462397933006287,
      "step": 925
    },
    {
      "completion_length": 24.34375,
      "epoch": 2.8417818740399383,
      "grad_norm": 17.32102394104004,
      "kl": 1.0458984375,
      "learning_rate": 2.884615384615384e-07,
      "loss": 0.2265,
      "reward": 0.7675108313560486,
      "reward_std": 0.016878814669325948,
      "reward_std_all": 0.029853275045752525,
      "rewards-max/forensics_sinkhorn_reward": 0.9690963327884674,
      "rewards-max/forensics_strict_edge_reward": 0.37208256125450134,
      "rewards-max/forensics_tiou_reward": 0.9506854116916656,
      "rewards/forensics_sinkhorn_reward": 0.719514399766922,
      "rewards/forensics_strict_edge_reward": 0.039160444401204586,
      "rewards/forensics_tiou_reward": 0.49317798018455505,
      "step": 926
    },
    {
      "completion_length": 27.90625,
      "epoch": 2.8448540706605225,
      "grad_norm": 7.718536376953125,
      "kl": 0.578125,
      "learning_rate": 2.8769230769230767e-07,
      "loss": 0.1749,
      "reward": 0.7583517134189606,
      "reward_std": 0.04856795445084572,
      "reward_std_all": 0.048691101372241974,
      "rewards-max/forensics_sinkhorn_reward": 0.985187828540802,
      "rewards-max/forensics_strict_edge_reward": 0.7229099273681641,
      "rewards-max/forensics_tiou_reward": 0.990574300289154,
      "rewards/forensics_sinkhorn_reward": 0.6628094911575317,
      "rewards/forensics_strict_edge_reward": 0.07881490513682365,
      "rewards/forensics_tiou_reward": 0.5524841248989105,
      "step": 927
    },
    {
      "completion_length": 23.25,
      "epoch": 2.847926267281106,
      "grad_norm": 8.490215301513672,
      "kl": 0.57421875,
      "learning_rate": 2.8692307692307694e-07,
      "loss": -0.0312,
      "reward": 0.7514429986476898,
      "reward_std": 0.041033294051885605,
      "reward_std_all": 0.04685427434742451,
      "rewards-max/forensics_sinkhorn_reward": 0.9777770042419434,
      "rewards-max/forensics_strict_edge_reward": 0.5811558067798615,
      "rewards-max/forensics_tiou_reward": 0.9717502593994141,
      "rewards/forensics_sinkhorn_reward": 0.7300917506217957,
      "rewards/forensics_strict_edge_reward": 0.0811283327639103,
      "rewards/forensics_tiou_reward": 0.5412507355213165,
      "step": 928
    },
    {
      "completion_length": 24.125,
      "epoch": 2.8509984639016897,
      "grad_norm": 9.909815788269043,
      "kl": 0.607421875,
      "learning_rate": 2.8615384615384615e-07,
      "loss": -0.0722,
      "reward": 0.7700599431991577,
      "reward_std": 0.02226998098194599,
      "reward_std_all": 0.029000836424529552,
      "rewards-max/forensics_sinkhorn_reward": 0.9720670580863953,
      "rewards-max/forensics_strict_edge_reward": 0.41346001625061035,
      "rewards-max/forensics_tiou_reward": 0.96045783162117,
      "rewards/forensics_sinkhorn_reward": 0.7526146173477173,
      "rewards/forensics_strict_edge_reward": 0.04812993109226227,
      "rewards/forensics_tiou_reward": 0.5786646008491516,
      "step": 929
    },
    {
      "completion_length": 24.25,
      "epoch": 2.8540706605222734,
      "grad_norm": 9.391889572143555,
      "kl": 0.4814453125,
      "learning_rate": 2.8538461538461536e-07,
      "loss": -0.012,
      "reward": 0.7702782452106476,
      "reward_std": 0.03077587205916643,
      "reward_std_all": 0.03418294992297888,
      "rewards-max/forensics_sinkhorn_reward": 0.9821699559688568,
      "rewards-max/forensics_strict_edge_reward": 0.7659632563591003,
      "rewards-max/forensics_tiou_reward": 0.9825958013534546,
      "rewards/forensics_sinkhorn_reward": 0.7687131762504578,
      "rewards/forensics_strict_edge_reward": 0.09812391549348831,
      "rewards/forensics_tiou_reward": 0.5774495303630829,
      "step": 930
    },
    {
      "completion_length": 22.484375,
      "epoch": 2.857142857142857,
      "grad_norm": 9.812190055847168,
      "kl": 0.58203125,
      "learning_rate": 2.846153846153846e-07,
      "loss": 0.1614,
      "reward": 0.769899994134903,
      "reward_std": 0.04360434412956238,
      "reward_std_all": 0.04313823580741882,
      "rewards-max/forensics_sinkhorn_reward": 0.9537607729434967,
      "rewards-max/forensics_strict_edge_reward": 0.332917258143425,
      "rewards-max/forensics_tiou_reward": 0.9134353399276733,
      "rewards/forensics_sinkhorn_reward": 0.7384136617183685,
      "rewards/forensics_strict_edge_reward": 0.026318627409636974,
      "rewards/forensics_tiou_reward": 0.4703930914402008,
      "step": 931
    },
    {
      "completion_length": 27.28125,
      "epoch": 2.860215053763441,
      "grad_norm": 11.069989204406738,
      "kl": 0.533203125,
      "learning_rate": 2.8384615384615384e-07,
      "loss": 0.1727,
      "reward": 0.7763550579547882,
      "reward_std": 0.03377543389797211,
      "reward_std_all": 0.033658239990472794,
      "rewards-max/forensics_sinkhorn_reward": 0.9524503946304321,
      "rewards-max/forensics_strict_edge_reward": 0.43677301704883575,
      "rewards-max/forensics_tiou_reward": 0.9166854918003082,
      "rewards/forensics_sinkhorn_reward": 0.6680631041526794,
      "rewards/forensics_strict_edge_reward": 0.031069396063685417,
      "rewards/forensics_tiou_reward": 0.5004504173994064,
      "step": 932
    },
    {
      "completion_length": 22.703125,
      "epoch": 2.8632872503840243,
      "grad_norm": 12.316580772399902,
      "kl": 0.478515625,
      "learning_rate": 2.8307692307692306e-07,
      "loss": 0.083,
      "reward": 0.7490143775939941,
      "reward_std": 0.02710349252447486,
      "reward_std_all": 0.04241642728447914,
      "rewards-max/forensics_sinkhorn_reward": 0.963394820690155,
      "rewards-max/forensics_strict_edge_reward": 0.45218876749277115,
      "rewards-max/forensics_tiou_reward": 0.9371106922626495,
      "rewards/forensics_sinkhorn_reward": 0.6915789544582367,
      "rewards/forensics_strict_edge_reward": 0.03621991258114576,
      "rewards/forensics_tiou_reward": 0.45619630813598633,
      "step": 933
    },
    {
      "completion_length": 22.0625,
      "epoch": 2.8663594470046085,
      "grad_norm": 9.904450416564941,
      "kl": 0.4990234375,
      "learning_rate": 2.823076923076923e-07,
      "loss": -0.1286,
      "reward": 0.7932164072990417,
      "reward_std": 0.021894396282732487,
      "reward_std_all": 0.020517073571681976,
      "rewards-max/forensics_sinkhorn_reward": 0.9825182855129242,
      "rewards-max/forensics_strict_edge_reward": 0.7268245220184326,
      "rewards-max/forensics_tiou_reward": 0.9820983111858368,
      "rewards/forensics_sinkhorn_reward": 0.8220345377922058,
      "rewards/forensics_strict_edge_reward": 0.11747835576534271,
      "rewards/forensics_tiou_reward": 0.656337171792984,
      "step": 934
    },
    {
      "completion_length": 26.171875,
      "epoch": 2.869431643625192,
      "grad_norm": 11.131564140319824,
      "kl": 0.5537109375,
      "learning_rate": 2.815384615384615e-07,
      "loss": -0.0813,
      "reward": 0.7652358114719391,
      "reward_std": 0.04501770809292793,
      "reward_std_all": 0.045533547177910805,
      "rewards-max/forensics_sinkhorn_reward": 0.982538491487503,
      "rewards-max/forensics_strict_edge_reward": 0.32655003666877747,
      "rewards-max/forensics_tiou_reward": 0.9812061488628387,
      "rewards/forensics_sinkhorn_reward": 0.726752907037735,
      "rewards/forensics_strict_edge_reward": 0.029089219868183136,
      "rewards/forensics_tiou_reward": 0.5350084602832794,
      "step": 935
    },
    {
      "completion_length": 27.734375,
      "epoch": 2.8725038402457757,
      "grad_norm": 9.919241905212402,
      "kl": 1.03125,
      "learning_rate": 2.8076923076923075e-07,
      "loss": 0.3185,
      "reward": 0.7484492063522339,
      "reward_std": 0.05424013873562217,
      "reward_std_all": 0.05732705513946712,
      "rewards-max/forensics_sinkhorn_reward": 0.9415823519229889,
      "rewards-max/forensics_strict_edge_reward": 0.5012606793316081,
      "rewards-max/forensics_tiou_reward": 0.8856509327888489,
      "rewards/forensics_sinkhorn_reward": 0.6566988527774811,
      "rewards/forensics_strict_edge_reward": 0.039000059266982134,
      "rewards/forensics_tiou_reward": 0.4996907114982605,
      "step": 936
    },
    {
      "completion_length": 25.5625,
      "epoch": 2.8755760368663594,
      "grad_norm": 10.70608139038086,
      "kl": 0.3876953125,
      "learning_rate": 2.8e-07,
      "loss": -0.0338,
      "reward": 0.7498178780078888,
      "reward_std": 0.05081337038427591,
      "reward_std_all": 0.05330284312367439,
      "rewards-max/forensics_sinkhorn_reward": 0.9858421683311462,
      "rewards-max/forensics_strict_edge_reward": 0.5846472382545471,
      "rewards-max/forensics_tiou_reward": 0.990452915430069,
      "rewards/forensics_sinkhorn_reward": 0.7360946834087372,
      "rewards/forensics_strict_edge_reward": 0.09376433864235878,
      "rewards/forensics_tiou_reward": 0.5750828087329865,
      "step": 937
    },
    {
      "completion_length": 23.75,
      "epoch": 2.878648233486943,
      "grad_norm": 12.177741050720215,
      "kl": 0.427734375,
      "learning_rate": 2.7923076923076923e-07,
      "loss": -0.2307,
      "reward": 0.7905344665050507,
      "reward_std": 0.028468727599829435,
      "reward_std_all": 0.02612194954417646,
      "rewards-max/forensics_sinkhorn_reward": 0.9803373217582703,
      "rewards-max/forensics_strict_edge_reward": 0.6326714754104614,
      "rewards-max/forensics_tiou_reward": 0.9742227792739868,
      "rewards/forensics_sinkhorn_reward": 0.7842592000961304,
      "rewards/forensics_strict_edge_reward": 0.1324007660150528,
      "rewards/forensics_tiou_reward": 0.6581440269947052,
      "step": 938
    },
    {
      "completion_length": 24.796875,
      "epoch": 2.881720430107527,
      "grad_norm": 9.618566513061523,
      "kl": 0.4462890625,
      "learning_rate": 2.7846153846153845e-07,
      "loss": -0.1836,
      "reward": 0.7719365954399109,
      "reward_std": 0.03516603074967861,
      "reward_std_all": 0.03657376021146774,
      "rewards-max/forensics_sinkhorn_reward": 0.9826787114143372,
      "rewards-max/forensics_strict_edge_reward": 0.5918685793876648,
      "rewards-max/forensics_tiou_reward": 0.9837054908275604,
      "rewards/forensics_sinkhorn_reward": 0.7885725498199463,
      "rewards/forensics_strict_edge_reward": 0.10054317861795425,
      "rewards/forensics_tiou_reward": 0.6019819676876068,
      "step": 939
    },
    {
      "completion_length": 28.5625,
      "epoch": 2.8847926267281108,
      "grad_norm": 10.786904335021973,
      "kl": 0.4091796875,
      "learning_rate": 2.7769230769230766e-07,
      "loss": 0.3639,
      "reward": 0.7246838510036469,
      "reward_std": 0.06665828451514244,
      "reward_std_all": 0.07536023110151291,
      "rewards-max/forensics_sinkhorn_reward": 0.9798794090747833,
      "rewards-max/forensics_strict_edge_reward": 0.6209261119365692,
      "rewards-max/forensics_tiou_reward": 0.9769083559513092,
      "rewards/forensics_sinkhorn_reward": 0.658330887556076,
      "rewards/forensics_strict_edge_reward": 0.11881370097398758,
      "rewards/forensics_tiou_reward": 0.5510478019714355,
      "step": 940
    },
    {
      "completion_length": 23.0625,
      "epoch": 2.8878648233486945,
      "grad_norm": 8.552440643310547,
      "kl": 0.42578125,
      "learning_rate": 2.7692307692307693e-07,
      "loss": -0.2758,
      "reward": 0.7571738362312317,
      "reward_std": 0.04089703410863876,
      "reward_std_all": 0.045209964737296104,
      "rewards-max/forensics_sinkhorn_reward": 0.9748763740062714,
      "rewards-max/forensics_strict_edge_reward": 0.6266564428806305,
      "rewards-max/forensics_tiou_reward": 0.9660735428333282,
      "rewards/forensics_sinkhorn_reward": 0.7122431099414825,
      "rewards/forensics_strict_edge_reward": 0.07216440513730049,
      "rewards/forensics_tiou_reward": 0.5468455255031586,
      "step": 941
    },
    {
      "completion_length": 24.265625,
      "epoch": 2.890937019969278,
      "grad_norm": 12.290319442749023,
      "kl": 0.4599609375,
      "learning_rate": 2.7615384615384614e-07,
      "loss": -0.1081,
      "reward": 0.779076874256134,
      "reward_std": 0.03260456211864948,
      "reward_std_all": 0.031902678310871124,
      "rewards-max/forensics_sinkhorn_reward": 0.9786718487739563,
      "rewards-max/forensics_strict_edge_reward": 0.647892102599144,
      "rewards-max/forensics_tiou_reward": 0.9724427759647369,
      "rewards/forensics_sinkhorn_reward": 0.7824823558330536,
      "rewards/forensics_strict_edge_reward": 0.12455195188522339,
      "rewards/forensics_tiou_reward": 0.5780336558818817,
      "step": 942
    },
    {
      "completion_length": 25.015625,
      "epoch": 2.8940092165898617,
      "grad_norm": 10.611507415771484,
      "kl": 0.4814453125,
      "learning_rate": 2.753846153846154e-07,
      "loss": -0.1275,
      "reward": 0.7835782468318939,
      "reward_std": 0.028684012591838837,
      "reward_std_all": 0.02808383759111166,
      "rewards-max/forensics_sinkhorn_reward": 0.9842984676361084,
      "rewards-max/forensics_strict_edge_reward": 0.6934256851673126,
      "rewards-max/forensics_tiou_reward": 0.9855886399745941,
      "rewards/forensics_sinkhorn_reward": 0.7230840921401978,
      "rewards/forensics_strict_edge_reward": 0.09758235886693001,
      "rewards/forensics_tiou_reward": 0.5741818249225616,
      "step": 943
    },
    {
      "completion_length": 23.21875,
      "epoch": 2.8970814132104454,
      "grad_norm": 9.544378280639648,
      "kl": 0.521484375,
      "learning_rate": 2.7461538461538457e-07,
      "loss": -0.0407,
      "reward": 0.7421979606151581,
      "reward_std": 0.03139159735292196,
      "reward_std_all": 0.04907442070543766,
      "rewards-max/forensics_sinkhorn_reward": 0.9764618575572968,
      "rewards-max/forensics_strict_edge_reward": 0.6762872636318207,
      "rewards-max/forensics_tiou_reward": 0.969241738319397,
      "rewards/forensics_sinkhorn_reward": 0.7191800475120544,
      "rewards/forensics_strict_edge_reward": 0.09279579482972622,
      "rewards/forensics_tiou_reward": 0.5446644872426987,
      "step": 944
    },
    {
      "completion_length": 24.0,
      "epoch": 2.900153609831029,
      "grad_norm": 17.632720947265625,
      "kl": 0.509765625,
      "learning_rate": 2.7384615384615383e-07,
      "loss": 0.146,
      "reward": 0.7018864154815674,
      "reward_std": 0.06146521307528019,
      "reward_std_all": 0.07806968316435814,
      "rewards-max/forensics_sinkhorn_reward": 0.9692901968955994,
      "rewards-max/forensics_strict_edge_reward": 0.38535140454769135,
      "rewards-max/forensics_tiou_reward": 0.9490390717983246,
      "rewards/forensics_sinkhorn_reward": 0.630545973777771,
      "rewards/forensics_strict_edge_reward": 0.0434429906308651,
      "rewards/forensics_tiou_reward": 0.4169186055660248,
      "step": 945
    },
    {
      "completion_length": 26.046875,
      "epoch": 2.903225806451613,
      "grad_norm": 15.813990592956543,
      "kl": 0.48828125,
      "learning_rate": 2.7307692307692305e-07,
      "loss": -0.2171,
      "reward": 0.7794135212898254,
      "reward_std": 0.03138023801147938,
      "reward_std_all": 0.031068000942468643,
      "rewards-max/forensics_sinkhorn_reward": 0.9733850657939911,
      "rewards-max/forensics_strict_edge_reward": 0.563819408416748,
      "rewards-max/forensics_tiou_reward": 0.9611567556858063,
      "rewards/forensics_sinkhorn_reward": 0.7317152619361877,
      "rewards/forensics_strict_edge_reward": 0.07720134034752846,
      "rewards/forensics_tiou_reward": 0.5806207656860352,
      "step": 946
    },
    {
      "completion_length": 27.25,
      "epoch": 2.9062980030721968,
      "grad_norm": 11.942281723022461,
      "kl": 0.59375,
      "learning_rate": 2.723076923076923e-07,
      "loss": 0.0541,
      "reward": 0.7656396627426147,
      "reward_std": 0.028889021836221218,
      "reward_std_all": 0.03653416782617569,
      "rewards-max/forensics_sinkhorn_reward": 0.9867448806762695,
      "rewards-max/forensics_strict_edge_reward": 0.8582656681537628,
      "rewards-max/forensics_tiou_reward": 0.9940246045589447,
      "rewards/forensics_sinkhorn_reward": 0.7121118903160095,
      "rewards/forensics_strict_edge_reward": 0.09660225734114647,
      "rewards/forensics_tiou_reward": 0.5345619916915894,
      "step": 947
    },
    {
      "completion_length": 24.015625,
      "epoch": 2.9093701996927805,
      "grad_norm": 9.484954833984375,
      "kl": 0.4208984375,
      "learning_rate": 2.7153846153846153e-07,
      "loss": 0.0139,
      "reward": 0.7653911411762238,
      "reward_std": 0.039061352610588074,
      "reward_std_all": 0.04173840023577213,
      "rewards-max/forensics_sinkhorn_reward": 0.9812077581882477,
      "rewards-max/forensics_strict_edge_reward": 0.7366199493408203,
      "rewards-max/forensics_tiou_reward": 0.9793870747089386,
      "rewards/forensics_sinkhorn_reward": 0.7633743286132812,
      "rewards/forensics_strict_edge_reward": 0.12032990157604218,
      "rewards/forensics_tiou_reward": 0.5926918983459473,
      "step": 948
    },
    {
      "completion_length": 24.953125,
      "epoch": 2.912442396313364,
      "grad_norm": 17.359622955322266,
      "kl": 0.4267578125,
      "learning_rate": 2.7076923076923074e-07,
      "loss": 0.097,
      "reward": 0.7730999290943146,
      "reward_std": 0.03934408724308014,
      "reward_std_all": 0.038360778242349625,
      "rewards-max/forensics_sinkhorn_reward": 0.9507086873054504,
      "rewards-max/forensics_strict_edge_reward": 0.4778955429792404,
      "rewards-max/forensics_tiou_reward": 0.9105920791625977,
      "rewards/forensics_sinkhorn_reward": 0.7411614656448364,
      "rewards/forensics_strict_edge_reward": 0.06376818753778934,
      "rewards/forensics_tiou_reward": 0.5307316780090332,
      "step": 949
    },
    {
      "completion_length": 24.453125,
      "epoch": 2.9155145929339477,
      "grad_norm": 11.622408866882324,
      "kl": 0.861328125,
      "learning_rate": 2.7e-07,
      "loss": 0.4494,
      "reward": 0.7230115532875061,
      "reward_std": 0.04139767028391361,
      "reward_std_all": 0.06363939121365547,
      "rewards-max/forensics_sinkhorn_reward": 0.9748384654521942,
      "rewards-max/forensics_strict_edge_reward": 0.5702531039714813,
      "rewards-max/forensics_tiou_reward": 0.9615518152713776,
      "rewards/forensics_sinkhorn_reward": 0.6628485321998596,
      "rewards/forensics_strict_edge_reward": 0.07705355063080788,
      "rewards/forensics_tiou_reward": 0.4701828807592392,
      "step": 950
    },
    {
      "completion_length": 26.890625,
      "epoch": 2.9185867895545314,
      "grad_norm": 8.611618995666504,
      "kl": 0.5244140625,
      "learning_rate": 2.692307692307692e-07,
      "loss": 0.1635,
      "reward": 0.7439724504947662,
      "reward_std": 0.04505380429327488,
      "reward_std_all": 0.05496493726968765,
      "rewards-max/forensics_sinkhorn_reward": 0.9743199348449707,
      "rewards-max/forensics_strict_edge_reward": 0.4770301729440689,
      "rewards-max/forensics_tiou_reward": 0.9601851403713226,
      "rewards/forensics_sinkhorn_reward": 0.662541002035141,
      "rewards/forensics_strict_edge_reward": 0.0631072074174881,
      "rewards/forensics_tiou_reward": 0.5008540600538254,
      "step": 951
    },
    {
      "completion_length": 22.109375,
      "epoch": 2.921658986175115,
      "grad_norm": 10.834259033203125,
      "kl": 0.591796875,
      "learning_rate": 2.6846153846153844e-07,
      "loss": 0.0239,
      "reward": 0.7681376039981842,
      "reward_std": 0.04711326211690903,
      "reward_std_all": 0.046415941789746284,
      "rewards-max/forensics_sinkhorn_reward": 0.9789533317089081,
      "rewards-max/forensics_strict_edge_reward": 0.7201169431209564,
      "rewards-max/forensics_tiou_reward": 0.976186990737915,
      "rewards/forensics_sinkhorn_reward": 0.7384676039218903,
      "rewards/forensics_strict_edge_reward": 0.08714966103434563,
      "rewards/forensics_tiou_reward": 0.5794784426689148,
      "step": 952
    },
    {
      "completion_length": 27.1875,
      "epoch": 2.924731182795699,
      "grad_norm": 8.40487289428711,
      "kl": 0.63671875,
      "learning_rate": 2.6769230769230765e-07,
      "loss": -0.1986,
      "reward": 0.7977675199508667,
      "reward_std": 0.0156767126172781,
      "reward_std_all": 0.015192166436463594,
      "rewards-max/forensics_sinkhorn_reward": 0.9820088148117065,
      "rewards-max/forensics_strict_edge_reward": 0.568440854549408,
      "rewards-max/forensics_tiou_reward": 0.9808700680732727,
      "rewards/forensics_sinkhorn_reward": 0.7863112688064575,
      "rewards/forensics_strict_edge_reward": 0.11321685835719109,
      "rewards/forensics_tiou_reward": 0.6627056300640106,
      "step": 953
    },
    {
      "completion_length": 24.75,
      "epoch": 2.9278033794162828,
      "grad_norm": 9.440825462341309,
      "kl": 0.4931640625,
      "learning_rate": 2.669230769230769e-07,
      "loss": -0.0198,
      "reward": 0.7049130499362946,
      "reward_std": 0.06449029967188835,
      "reward_std_all": 0.08067601174116135,
      "rewards-max/forensics_sinkhorn_reward": 0.9866138696670532,
      "rewards-max/forensics_strict_edge_reward": 0.6071805953979492,
      "rewards-max/forensics_tiou_reward": 0.9925285279750824,
      "rewards/forensics_sinkhorn_reward": 0.5944833159446716,
      "rewards/forensics_strict_edge_reward": 0.06459285318851471,
      "rewards/forensics_tiou_reward": 0.5102412551641464,
      "step": 954
    },
    {
      "completion_length": 25.71875,
      "epoch": 2.9308755760368665,
      "grad_norm": 12.073810577392578,
      "kl": 0.888671875,
      "learning_rate": 2.6615384615384613e-07,
      "loss": 0.0243,
      "reward": 0.7778075337409973,
      "reward_std": 0.03462303429841995,
      "reward_std_all": 0.033783841878175735,
      "rewards-max/forensics_sinkhorn_reward": 0.9802514314651489,
      "rewards-max/forensics_strict_edge_reward": 0.5539274513721466,
      "rewards-max/forensics_tiou_reward": 0.9773809313774109,
      "rewards/forensics_sinkhorn_reward": 0.7514959275722504,
      "rewards/forensics_strict_edge_reward": 0.062452686950564384,
      "rewards/forensics_tiou_reward": 0.5543748438358307,
      "step": 955
    },
    {
      "completion_length": 26.890625,
      "epoch": 2.93394777265745,
      "grad_norm": 8.838958740234375,
      "kl": 0.3779296875,
      "learning_rate": 2.653846153846154e-07,
      "loss": 0.0175,
      "reward": 0.755401074886322,
      "reward_std": 0.032681324519217014,
      "reward_std_all": 0.04295487143099308,
      "rewards-max/forensics_sinkhorn_reward": 0.9681885242462158,
      "rewards-max/forensics_strict_edge_reward": 0.4273769110441208,
      "rewards-max/forensics_tiou_reward": 0.948039710521698,
      "rewards/forensics_sinkhorn_reward": 0.7584525644779205,
      "rewards/forensics_strict_edge_reward": 0.07992113381624222,
      "rewards/forensics_tiou_reward": 0.6145732998847961,
      "step": 956
    },
    {
      "completion_length": 24.25,
      "epoch": 2.9370199692780337,
      "grad_norm": 11.471098899841309,
      "kl": 0.4140625,
      "learning_rate": 2.646153846153846e-07,
      "loss": -0.0671,
      "reward": 0.7659767270088196,
      "reward_std": 0.045711856335401535,
      "reward_std_all": 0.044801438227295876,
      "rewards-max/forensics_sinkhorn_reward": 0.9557272791862488,
      "rewards-max/forensics_strict_edge_reward": 0.36298468708992004,
      "rewards-max/forensics_tiou_reward": 0.9139071702957153,
      "rewards/forensics_sinkhorn_reward": 0.7143636345863342,
      "rewards/forensics_strict_edge_reward": 0.02575475163757801,
      "rewards/forensics_tiou_reward": 0.49365048110485077,
      "step": 957
    },
    {
      "completion_length": 24.890625,
      "epoch": 2.9400921658986174,
      "grad_norm": 11.141971588134766,
      "kl": 0.4765625,
      "learning_rate": 2.638461538461538e-07,
      "loss": 0.1409,
      "reward": 0.7521951794624329,
      "reward_std": 0.030262688174843788,
      "reward_std_all": 0.043627929873764515,
      "rewards-max/forensics_sinkhorn_reward": 0.9730455279350281,
      "rewards-max/forensics_strict_edge_reward": 0.5110019147396088,
      "rewards-max/forensics_tiou_reward": 0.9582183659076691,
      "rewards/forensics_sinkhorn_reward": 0.7201690673828125,
      "rewards/forensics_strict_edge_reward": 0.07836712524294853,
      "rewards/forensics_tiou_reward": 0.5477010607719421,
      "step": 958
    },
    {
      "completion_length": 23.765625,
      "epoch": 2.943164362519201,
      "grad_norm": 8.889922142028809,
      "kl": 0.708984375,
      "learning_rate": 2.630769230769231e-07,
      "loss": 0.0146,
      "reward": 0.7763248383998871,
      "reward_std": 0.015986846759915352,
      "reward_std_all": 0.02300208806991577,
      "rewards-max/forensics_sinkhorn_reward": 0.9814203679561615,
      "rewards-max/forensics_strict_edge_reward": 0.765928328037262,
      "rewards-max/forensics_tiou_reward": 0.9814484417438507,
      "rewards/forensics_sinkhorn_reward": 0.727135419845581,
      "rewards/forensics_strict_edge_reward": 0.14622515439987183,
      "rewards/forensics_tiou_reward": 0.5522408783435822,
      "step": 959
    },
    {
      "completion_length": 27.453125,
      "epoch": 2.946236559139785,
      "grad_norm": 20.215456008911133,
      "kl": 1.169921875,
      "learning_rate": 2.623076923076923e-07,
      "loss": -0.1079,
      "reward": 0.7323879599571228,
      "reward_std": 0.007772269658744335,
      "reward_std_all": 0.04138938710093498,
      "rewards-max/forensics_sinkhorn_reward": 0.9865045547485352,
      "rewards-max/forensics_strict_edge_reward": 0.5900419056415558,
      "rewards-max/forensics_tiou_reward": 0.9939983189105988,
      "rewards/forensics_sinkhorn_reward": 0.6357397437095642,
      "rewards/forensics_strict_edge_reward": 0.08209671825170517,
      "rewards/forensics_tiou_reward": 0.5182621628046036,
      "step": 960
    },
    {
      "completion_length": 24.78125,
      "epoch": 2.9493087557603688,
      "grad_norm": 10.27651596069336,
      "kl": 0.4521484375,
      "learning_rate": 2.615384615384615e-07,
      "loss": -0.153,
      "reward": 0.724279135465622,
      "reward_std": 0.03853087034076452,
      "reward_std_all": 0.055223917588591576,
      "rewards-max/forensics_sinkhorn_reward": 0.9760634005069733,
      "rewards-max/forensics_strict_edge_reward": 0.7246644794940948,
      "rewards-max/forensics_tiou_reward": 0.9698066711425781,
      "rewards/forensics_sinkhorn_reward": 0.6550813913345337,
      "rewards/forensics_strict_edge_reward": 0.07394556887447834,
      "rewards/forensics_tiou_reward": 0.44217798113822937,
      "step": 961
    },
    {
      "completion_length": 22.625,
      "epoch": 2.9523809523809526,
      "grad_norm": 9.409781455993652,
      "kl": 0.458984375,
      "learning_rate": 2.6076923076923073e-07,
      "loss": -0.0301,
      "reward": 0.7735579013824463,
      "reward_std": 0.03836425393819809,
      "reward_std_all": 0.03739559277892113,
      "rewards-max/forensics_sinkhorn_reward": 0.984259843826294,
      "rewards-max/forensics_strict_edge_reward": 0.7457418143749237,
      "rewards-max/forensics_tiou_reward": 0.9864294230937958,
      "rewards/forensics_sinkhorn_reward": 0.7259647250175476,
      "rewards/forensics_strict_edge_reward": 0.11235266923904419,
      "rewards/forensics_tiou_reward": 0.5855158567428589,
      "step": 962
    },
    {
      "completion_length": 26.015625,
      "epoch": 2.955453149001536,
      "grad_norm": 7.794765949249268,
      "kl": 0.603515625,
      "learning_rate": 2.6e-07,
      "loss": 0.1133,
      "reward": 0.7693426609039307,
      "reward_std": 0.029976770281791687,
      "reward_std_all": 0.03498490899801254,
      "rewards-max/forensics_sinkhorn_reward": 0.9800372421741486,
      "rewards-max/forensics_strict_edge_reward": 0.3970619887113571,
      "rewards-max/forensics_tiou_reward": 0.976129949092865,
      "rewards/forensics_sinkhorn_reward": 0.7203890681266785,
      "rewards/forensics_strict_edge_reward": 0.054087741300463676,
      "rewards/forensics_tiou_reward": 0.5858092904090881,
      "step": 963
    },
    {
      "completion_length": 27.171875,
      "epoch": 2.9585253456221197,
      "grad_norm": 10.735869407653809,
      "kl": 0.5048828125,
      "learning_rate": 2.592307692307692e-07,
      "loss": 0.033,
      "reward": 0.7981854677200317,
      "reward_std": 0.012101585045456886,
      "reward_std_all": 0.012406199239194393,
      "rewards-max/forensics_sinkhorn_reward": 0.9786411821842194,
      "rewards-max/forensics_strict_edge_reward": 0.5152659118175507,
      "rewards-max/forensics_tiou_reward": 0.9765013158321381,
      "rewards/forensics_sinkhorn_reward": 0.8056158423423767,
      "rewards/forensics_strict_edge_reward": 0.08193812519311905,
      "rewards/forensics_tiou_reward": 0.600024402141571,
      "step": 964
    },
    {
      "completion_length": 27.0625,
      "epoch": 2.9615975422427034,
      "grad_norm": 7.933253765106201,
      "kl": 0.4951171875,
      "learning_rate": 2.584615384615385e-07,
      "loss": 0.1031,
      "reward": 0.7682093977928162,
      "reward_std": 0.042375389486551285,
      "reward_std_all": 0.041926633566617966,
      "rewards-max/forensics_sinkhorn_reward": 0.9846581518650055,
      "rewards-max/forensics_strict_edge_reward": 0.7203241437673569,
      "rewards-max/forensics_tiou_reward": 0.9876132607460022,
      "rewards/forensics_sinkhorn_reward": 0.771878570318222,
      "rewards/forensics_strict_edge_reward": 0.06614071689546108,
      "rewards/forensics_tiou_reward": 0.6323583126068115,
      "step": 965
    },
    {
      "completion_length": 26.75,
      "epoch": 2.964669738863287,
      "grad_norm": 8.921941757202148,
      "kl": 0.373046875,
      "learning_rate": 2.5769230769230764e-07,
      "loss": -0.1704,
      "reward": 0.7756495475769043,
      "reward_std": 0.024100229376927018,
      "reward_std_all": 0.030560781247913837,
      "rewards-max/forensics_sinkhorn_reward": 0.9805651009082794,
      "rewards-max/forensics_strict_edge_reward": 0.5242990553379059,
      "rewards-max/forensics_tiou_reward": 0.9774589836597443,
      "rewards/forensics_sinkhorn_reward": 0.7789593935012817,
      "rewards/forensics_strict_edge_reward": 0.06929527409374714,
      "rewards/forensics_tiou_reward": 0.6019579470157623,
      "step": 966
    },
    {
      "completion_length": 23.84375,
      "epoch": 2.967741935483871,
      "grad_norm": 10.742571830749512,
      "kl": 0.50390625,
      "learning_rate": 2.569230769230769e-07,
      "loss": -0.0049,
      "reward": 0.7589408159255981,
      "reward_std": 0.04937826283276081,
      "reward_std_all": 0.050499994307756424,
      "rewards-max/forensics_sinkhorn_reward": 0.9701192677021027,
      "rewards-max/forensics_strict_edge_reward": 0.4152248799800873,
      "rewards-max/forensics_tiou_reward": 0.9550464153289795,
      "rewards/forensics_sinkhorn_reward": 0.6980925500392914,
      "rewards/forensics_strict_edge_reward": 0.03622066415846348,
      "rewards/forensics_tiou_reward": 0.4743584394454956,
      "step": 967
    },
    {
      "completion_length": 22.453125,
      "epoch": 2.9708141321044548,
      "grad_norm": 12.447490692138672,
      "kl": 0.6494140625,
      "learning_rate": 2.5615384615384617e-07,
      "loss": -0.1287,
      "reward": 0.7721168994903564,
      "reward_std": 0.05153435654938221,
      "reward_std_all": 0.04714861884713173,
      "rewards-max/forensics_sinkhorn_reward": 0.9857443869113922,
      "rewards-max/forensics_strict_edge_reward": 0.824172168970108,
      "rewards-max/forensics_tiou_reward": 0.9917426109313965,
      "rewards/forensics_sinkhorn_reward": 0.7656108736991882,
      "rewards/forensics_strict_edge_reward": 0.09735628589987755,
      "rewards/forensics_tiou_reward": 0.5833121538162231,
      "step": 968
    },
    {
      "completion_length": 25.78125,
      "epoch": 2.9738863287250386,
      "grad_norm": 10.266227722167969,
      "kl": 0.7578125,
      "learning_rate": 2.553846153846154e-07,
      "loss": 0.0296,
      "reward": 0.7872261703014374,
      "reward_std": 0.02197308308677748,
      "reward_std_all": 0.023329342948272824,
      "rewards-max/forensics_sinkhorn_reward": 0.9773244261741638,
      "rewards-max/forensics_strict_edge_reward": 0.6737824082374573,
      "rewards-max/forensics_tiou_reward": 0.9727108776569366,
      "rewards/forensics_sinkhorn_reward": 0.7006320953369141,
      "rewards/forensics_strict_edge_reward": 0.07212459668517113,
      "rewards/forensics_tiou_reward": 0.5632020235061646,
      "step": 969
    },
    {
      "completion_length": 27.34375,
      "epoch": 2.976958525345622,
      "grad_norm": 10.370847702026367,
      "kl": 0.599609375,
      "learning_rate": 2.546153846153846e-07,
      "loss": 0.2636,
      "reward": 0.7647178769111633,
      "reward_std": 0.024887241423130035,
      "reward_std_all": 0.035703908652067184,
      "rewards-max/forensics_sinkhorn_reward": 0.9790386855602264,
      "rewards-max/forensics_strict_edge_reward": 0.3885071277618408,
      "rewards-max/forensics_tiou_reward": 0.9748875200748444,
      "rewards/forensics_sinkhorn_reward": 0.6868186593055725,
      "rewards/forensics_strict_edge_reward": 0.039916492998600006,
      "rewards/forensics_tiou_reward": 0.5517803132534027,
      "step": 970
    },
    {
      "completion_length": 27.359375,
      "epoch": 2.980030721966206,
      "grad_norm": 11.14004898071289,
      "kl": 0.5791015625,
      "learning_rate": 2.538461538461538e-07,
      "loss": 0.0018,
      "reward": 0.7253421247005463,
      "reward_std": 0.03527451306581497,
      "reward_std_all": 0.05610848590731621,
      "rewards-max/forensics_sinkhorn_reward": 0.9844083487987518,
      "rewards-max/forensics_strict_edge_reward": 0.47467146813869476,
      "rewards-max/forensics_tiou_reward": 0.9893281161785126,
      "rewards/forensics_sinkhorn_reward": 0.6977195739746094,
      "rewards/forensics_strict_edge_reward": 0.08047070354223251,
      "rewards/forensics_tiou_reward": 0.5256796330213547,
      "step": 971
    },
    {
      "completion_length": 25.734375,
      "epoch": 2.9831029185867894,
      "grad_norm": 15.057692527770996,
      "kl": 0.51171875,
      "learning_rate": 2.530769230769231e-07,
      "loss": 0.0569,
      "reward": 0.7599830031394958,
      "reward_std": 0.05038708448410034,
      "reward_std_all": 0.05385865829885006,
      "rewards-max/forensics_sinkhorn_reward": 0.9860331416130066,
      "rewards-max/forensics_strict_edge_reward": 0.723625510931015,
      "rewards-max/forensics_tiou_reward": 0.9907840490341187,
      "rewards/forensics_sinkhorn_reward": 0.728145182132721,
      "rewards/forensics_strict_edge_reward": 0.09446460753679276,
      "rewards/forensics_tiou_reward": 0.5386532545089722,
      "step": 972
    },
    {
      "completion_length": 24.84375,
      "epoch": 2.986175115207373,
      "grad_norm": 11.995010375976562,
      "kl": 0.673828125,
      "learning_rate": 2.523076923076923e-07,
      "loss": -0.2677,
      "reward": 0.7987357378005981,
      "reward_std": 0.013600517297163606,
      "reward_std_all": 0.01339307555463165,
      "rewards-max/forensics_sinkhorn_reward": 0.9797740578651428,
      "rewards-max/forensics_strict_edge_reward": 0.6770831942558289,
      "rewards-max/forensics_tiou_reward": 0.9763577282428741,
      "rewards/forensics_sinkhorn_reward": 0.7717169523239136,
      "rewards/forensics_strict_edge_reward": 0.11469604820013046,
      "rewards/forensics_tiou_reward": 0.6638993322849274,
      "step": 973
    },
    {
      "completion_length": 22.25,
      "epoch": 2.989247311827957,
      "grad_norm": 12.66617488861084,
      "kl": 0.619140625,
      "learning_rate": 2.515384615384615e-07,
      "loss": -0.1727,
      "reward": 0.7681352198123932,
      "reward_std": 0.030880508944392204,
      "reward_std_all": 0.03468786831945181,
      "rewards-max/forensics_sinkhorn_reward": 0.9820065498352051,
      "rewards-max/forensics_strict_edge_reward": 0.8465203046798706,
      "rewards-max/forensics_tiou_reward": 0.9791655540466309,
      "rewards/forensics_sinkhorn_reward": 0.6906934678554535,
      "rewards/forensics_strict_edge_reward": 0.06447476521134377,
      "rewards/forensics_tiou_reward": 0.5173044502735138,
      "step": 974
    },
    {
      "completion_length": 23.578125,
      "epoch": 2.9923195084485408,
      "grad_norm": 16.858633041381836,
      "kl": 0.9267578125,
      "learning_rate": 2.507692307692307e-07,
      "loss": -0.27,
      "reward": 0.7757607102394104,
      "reward_std": 0.04690672643482685,
      "reward_std_all": 0.042852047830820084,
      "rewards-max/forensics_sinkhorn_reward": 0.9842406809329987,
      "rewards-max/forensics_strict_edge_reward": 0.4770301729440689,
      "rewards-max/forensics_tiou_reward": 0.98635134100914,
      "rewards/forensics_sinkhorn_reward": 0.7879537045955658,
      "rewards/forensics_strict_edge_reward": 0.10869896411895752,
      "rewards/forensics_tiou_reward": 0.5977553129196167,
      "step": 975
    },
    {
      "completion_length": 25.234375,
      "epoch": 2.9953917050691246,
      "grad_norm": 8.555970191955566,
      "kl": 0.56640625,
      "learning_rate": 2.5e-07,
      "loss": 0.0369,
      "reward": 0.7955431044101715,
      "reward_std": 0.015600265003740788,
      "reward_std_all": 0.01570231094956398,
      "rewards-max/forensics_sinkhorn_reward": 0.9778717160224915,
      "rewards-max/forensics_strict_edge_reward": 0.6718102097511292,
      "rewards-max/forensics_tiou_reward": 0.9726369976997375,
      "rewards/forensics_sinkhorn_reward": 0.7437772750854492,
      "rewards/forensics_strict_edge_reward": 0.08607938885688782,
      "rewards/forensics_tiou_reward": 0.5968925654888153,
      "step": 976
    },
    {
      "completion_length": 23.109375,
      "epoch": 2.998463901689708,
      "grad_norm": 10.938156127929688,
      "kl": 0.48828125,
      "learning_rate": 2.492307692307692e-07,
      "loss": 0.0163,
      "reward": 0.7819216847419739,
      "reward_std": 0.03568902611732483,
      "reward_std_all": 0.03331438824534416,
      "rewards-max/forensics_sinkhorn_reward": 0.9786960482597351,
      "rewards-max/forensics_strict_edge_reward": 0.5925651341676712,
      "rewards-max/forensics_tiou_reward": 0.9686842262744904,
      "rewards/forensics_sinkhorn_reward": 0.7587977945804596,
      "rewards/forensics_strict_edge_reward": 0.08876035735011101,
      "rewards/forensics_tiou_reward": 0.5916312634944916,
      "step": 977
    },
    {
      "completion_length": 32.60000228881836,
      "epoch": 3.0,
      "grad_norm": 7.230753421783447,
      "kl": 0.462890625,
      "learning_rate": 2.4846153846153847e-07,
      "loss": 0.0905,
      "reward": 0.8046265840530396,
      "reward_std": 0.018837258219718933,
      "reward_std_all": 0.003916326444596052,
      "rewards-max/forensics_sinkhorn_reward": 0.9303784370422363,
      "rewards-max/forensics_strict_edge_reward": 0.2654097378253937,
      "rewards-max/forensics_tiou_reward": 0.8651568293571472,
      "rewards/forensics_sinkhorn_reward": 0.6645282506942749,
      "rewards/forensics_strict_edge_reward": 0.05387416481971741,
      "rewards/forensics_tiou_reward": 0.7239088416099548,
      "step": 978
    },
    {
      "completion_length": 23.140625,
      "epoch": 3.003072196620584,
      "grad_norm": 10.690479278564453,
      "kl": 0.5693359375,
      "learning_rate": 2.476923076923077e-07,
      "loss": -0.1074,
      "reward": 0.7898562550544739,
      "reward_std": 0.026759416796267033,
      "reward_std_all": 0.025222443975508213,
      "rewards-max/forensics_sinkhorn_reward": 0.9862648248672485,
      "rewards-max/forensics_strict_edge_reward": 0.6154412180185318,
      "rewards-max/forensics_tiou_reward": 0.992215633392334,
      "rewards/forensics_sinkhorn_reward": 0.7703473567962646,
      "rewards/forensics_strict_edge_reward": 0.1156046986579895,
      "rewards/forensics_tiou_reward": 0.6140235662460327,
      "step": 979
    },
    {
      "completion_length": 26.15625,
      "epoch": 3.0061443932411676,
      "grad_norm": 8.983124732971191,
      "kl": 0.494140625,
      "learning_rate": 2.469230769230769e-07,
      "loss": 0.045,
      "reward": 0.7538906037807465,
      "reward_std": 0.03509399387985468,
      "reward_std_all": 0.0444320198148489,
      "rewards-max/forensics_sinkhorn_reward": 0.9794028103351593,
      "rewards-max/forensics_strict_edge_reward": 0.5702531039714813,
      "rewards-max/forensics_tiou_reward": 0.9745588004589081,
      "rewards/forensics_sinkhorn_reward": 0.6720267236232758,
      "rewards/forensics_strict_edge_reward": 0.061130499467253685,
      "rewards/forensics_tiou_reward": 0.5032064020633698,
      "step": 980
    },
    {
      "completion_length": 23.25,
      "epoch": 3.0092165898617513,
      "grad_norm": 8.971339225769043,
      "kl": 0.591796875,
      "learning_rate": 2.4615384615384616e-07,
      "loss": -0.2291,
      "reward": 0.7846387624740601,
      "reward_std": 0.013396383496001363,
      "reward_std_all": 0.020206094719469547,
      "rewards-max/forensics_sinkhorn_reward": 0.9770355522632599,
      "rewards-max/forensics_strict_edge_reward": 0.684698224067688,
      "rewards-max/forensics_tiou_reward": 0.9702345728874207,
      "rewards/forensics_sinkhorn_reward": 0.7573105692863464,
      "rewards/forensics_strict_edge_reward": 0.08106276765465736,
      "rewards/forensics_tiou_reward": 0.5444389581680298,
      "step": 981
    },
    {
      "completion_length": 25.171875,
      "epoch": 3.0122887864823347,
      "grad_norm": 9.70051383972168,
      "kl": 0.603515625,
      "learning_rate": 2.453846153846154e-07,
      "loss": -0.1406,
      "reward": 0.7760397791862488,
      "reward_std": 0.04471941792871803,
      "reward_std_all": 0.041191898635588586,
      "rewards-max/forensics_sinkhorn_reward": 0.9885607659816742,
      "rewards-max/forensics_strict_edge_reward": 0.7933230996131897,
      "rewards-max/forensics_tiou_reward": 0.9987761676311493,
      "rewards/forensics_sinkhorn_reward": 0.8017351031303406,
      "rewards/forensics_strict_edge_reward": 0.1197570264339447,
      "rewards/forensics_tiou_reward": 0.6676109731197357,
      "step": 982
    },
    {
      "completion_length": 24.234375,
      "epoch": 3.0153609831029184,
      "grad_norm": 18.082189559936523,
      "kl": 0.4755859375,
      "learning_rate": 2.446153846153846e-07,
      "loss": 0.0229,
      "reward": 0.7903757989406586,
      "reward_std": 0.019452818436548114,
      "reward_std_all": 0.01988345105201006,
      "rewards-max/forensics_sinkhorn_reward": 0.9803225994110107,
      "rewards-max/forensics_strict_edge_reward": 0.5194980204105377,
      "rewards-max/forensics_tiou_reward": 0.9770381450653076,
      "rewards/forensics_sinkhorn_reward": 0.7931615710258484,
      "rewards/forensics_strict_edge_reward": 0.08318430930376053,
      "rewards/forensics_tiou_reward": 0.6320535242557526,
      "step": 983
    },
    {
      "completion_length": 22.8125,
      "epoch": 3.0184331797235022,
      "grad_norm": 12.995241165161133,
      "kl": 0.62890625,
      "learning_rate": 2.4384615384615386e-07,
      "loss": 0.0363,
      "reward": 0.7604157328605652,
      "reward_std": 0.05843932554125786,
      "reward_std_all": 0.05487408488988876,
      "rewards-max/forensics_sinkhorn_reward": 0.9844056367874146,
      "rewards-max/forensics_strict_edge_reward": 0.6706924736499786,
      "rewards-max/forensics_tiou_reward": 0.9862534701824188,
      "rewards/forensics_sinkhorn_reward": 0.733238160610199,
      "rewards/forensics_strict_edge_reward": 0.08274559676647186,
      "rewards/forensics_tiou_reward": 0.5331549048423767,
      "step": 984
    },
    {
      "completion_length": 24.140625,
      "epoch": 3.021505376344086,
      "grad_norm": 11.276689529418945,
      "kl": 0.59765625,
      "learning_rate": 2.4307692307692307e-07,
      "loss": 0.2769,
      "reward": 0.749945878982544,
      "reward_std": 0.0395629471167922,
      "reward_std_all": 0.04912140592932701,
      "rewards-max/forensics_sinkhorn_reward": 0.9691583514213562,
      "rewards-max/forensics_strict_edge_reward": 0.4727507531642914,
      "rewards-max/forensics_tiou_reward": 0.9512182176113129,
      "rewards/forensics_sinkhorn_reward": 0.6911678910255432,
      "rewards/forensics_strict_edge_reward": 0.0464343037456274,
      "rewards/forensics_tiou_reward": 0.5454756915569305,
      "step": 985
    },
    {
      "completion_length": 23.515625,
      "epoch": 3.02457757296467,
      "grad_norm": 11.540037155151367,
      "kl": 0.4833984375,
      "learning_rate": 2.423076923076923e-07,
      "loss": -0.0863,
      "reward": 0.7716320753097534,
      "reward_std": 0.017548253759741783,
      "reward_std_all": 0.026530887000262737,
      "rewards-max/forensics_sinkhorn_reward": 0.9882201552391052,
      "rewards-max/forensics_strict_edge_reward": 0.8469520509243011,
      "rewards-max/forensics_tiou_reward": 0.9979165196418762,
      "rewards/forensics_sinkhorn_reward": 0.783663272857666,
      "rewards/forensics_strict_edge_reward": 0.10791082680225372,
      "rewards/forensics_tiou_reward": 0.6727094650268555,
      "step": 986
    },
    {
      "completion_length": 24.265625,
      "epoch": 3.0276497695852536,
      "grad_norm": 9.115386009216309,
      "kl": 0.548828125,
      "learning_rate": 2.415384615384615e-07,
      "loss": -0.0088,
      "reward": 0.7796683013439178,
      "reward_std": 0.018495609750971198,
      "reward_std_all": 0.024827458313666284,
      "rewards-max/forensics_sinkhorn_reward": 0.9774862825870514,
      "rewards-max/forensics_strict_edge_reward": 0.6593653857707977,
      "rewards-max/forensics_tiou_reward": 0.9704994261264801,
      "rewards/forensics_sinkhorn_reward": 0.755680650472641,
      "rewards/forensics_strict_edge_reward": 0.13001668080687523,
      "rewards/forensics_tiou_reward": 0.5617152452468872,
      "step": 987
    },
    {
      "completion_length": 23.125,
      "epoch": 3.0307219662058373,
      "grad_norm": 10.958996772766113,
      "kl": 0.5869140625,
      "learning_rate": 2.4076923076923076e-07,
      "loss": 0.0432,
      "reward": 0.7405243217945099,
      "reward_std": 0.047776712104678154,
      "reward_std_all": 0.055452799424529076,
      "rewards-max/forensics_sinkhorn_reward": 0.9786822497844696,
      "rewards-max/forensics_strict_edge_reward": 0.5851611196994781,
      "rewards-max/forensics_tiou_reward": 0.9747814834117889,
      "rewards/forensics_sinkhorn_reward": 0.7268278896808624,
      "rewards/forensics_strict_edge_reward": 0.10984878987073898,
      "rewards/forensics_tiou_reward": 0.5871558487415314,
      "step": 988
    },
    {
      "completion_length": 28.4375,
      "epoch": 3.033794162826421,
      "grad_norm": 8.571742057800293,
      "kl": 0.568359375,
      "learning_rate": 2.4e-07,
      "loss": -0.1907,
      "reward": 0.7304782569408417,
      "reward_std": 0.042595358565449715,
      "reward_std_all": 0.05416822200641036,
      "rewards-max/forensics_sinkhorn_reward": 0.9761990904808044,
      "rewards-max/forensics_strict_edge_reward": 0.5702531039714813,
      "rewards-max/forensics_tiou_reward": 0.9684596359729767,
      "rewards/forensics_sinkhorn_reward": 0.6485376358032227,
      "rewards/forensics_strict_edge_reward": 0.08111533895134926,
      "rewards/forensics_tiou_reward": 0.4904448390007019,
      "step": 989
    },
    {
      "completion_length": 24.03125,
      "epoch": 3.0368663594470044,
      "grad_norm": 11.125514030456543,
      "kl": 0.76953125,
      "learning_rate": 2.392307692307692e-07,
      "loss": -0.0026,
      "reward": 0.7809417247772217,
      "reward_std": 0.01686283014714718,
      "reward_std_all": 0.02233025711029768,
      "rewards-max/forensics_sinkhorn_reward": 0.9661809802055359,
      "rewards-max/forensics_strict_edge_reward": 0.6852045506238937,
      "rewards-max/forensics_tiou_reward": 0.9461567401885986,
      "rewards/forensics_sinkhorn_reward": 0.7181055843830109,
      "rewards/forensics_strict_edge_reward": 0.06713381968438625,
      "rewards/forensics_tiou_reward": 0.5158559679985046,
      "step": 990
    },
    {
      "completion_length": 24.34375,
      "epoch": 3.0399385560675882,
      "grad_norm": 15.154683113098145,
      "kl": 1.29296875,
      "learning_rate": 2.3846153846153846e-07,
      "loss": -0.0847,
      "reward": 0.7992456257343292,
      "reward_std": 0.011590803973376751,
      "reward_std_all": 0.011813861317932606,
      "rewards-max/forensics_sinkhorn_reward": 0.980532705783844,
      "rewards-max/forensics_strict_edge_reward": 0.7335375547409058,
      "rewards-max/forensics_tiou_reward": 0.9799485802650452,
      "rewards/forensics_sinkhorn_reward": 0.8400587141513824,
      "rewards/forensics_strict_edge_reward": 0.12637439742684364,
      "rewards/forensics_tiou_reward": 0.6916171014308929,
      "step": 991
    },
    {
      "completion_length": 26.4375,
      "epoch": 3.043010752688172,
      "grad_norm": 7.709336280822754,
      "kl": 0.5244140625,
      "learning_rate": 2.376923076923077e-07,
      "loss": -0.3296,
      "reward": 0.8048904538154602,
      "reward_std": 0.0029619759880006313,
      "reward_std_all": 0.004574240185320377,
      "rewards-max/forensics_sinkhorn_reward": 0.9850009083747864,
      "rewards-max/forensics_strict_edge_reward": 0.7513352334499359,
      "rewards-max/forensics_tiou_reward": 0.9881626069545746,
      "rewards/forensics_sinkhorn_reward": 0.8117298483848572,
      "rewards/forensics_strict_edge_reward": 0.12863007187843323,
      "rewards/forensics_tiou_reward": 0.6692193746566772,
      "step": 992
    },
    {
      "completion_length": 25.6875,
      "epoch": 3.046082949308756,
      "grad_norm": 11.231226921081543,
      "kl": 0.658203125,
      "learning_rate": 2.369230769230769e-07,
      "loss": 0.1646,
      "reward": 0.7555487751960754,
      "reward_std": 0.023338228464126587,
      "reward_std_all": 0.036757102236151695,
      "rewards-max/forensics_sinkhorn_reward": 0.9847188889980316,
      "rewards-max/forensics_strict_edge_reward": 0.6034213900566101,
      "rewards-max/forensics_tiou_reward": 0.9867002665996552,
      "rewards/forensics_sinkhorn_reward": 0.6938326358795166,
      "rewards/forensics_strict_edge_reward": 0.10312589630484581,
      "rewards/forensics_tiou_reward": 0.5225640535354614,
      "step": 993
    },
    {
      "completion_length": 27.265625,
      "epoch": 3.0491551459293396,
      "grad_norm": 9.474872589111328,
      "kl": 0.4169921875,
      "learning_rate": 2.3615384615384615e-07,
      "loss": -0.2422,
      "reward": 0.7764534652233124,
      "reward_std": 0.025014689192175865,
      "reward_std_all": 0.027451283298432827,
      "rewards-max/forensics_sinkhorn_reward": 0.9771542549133301,
      "rewards-max/forensics_strict_edge_reward": 0.6089114248752594,
      "rewards-max/forensics_tiou_reward": 0.9701196551322937,
      "rewards/forensics_sinkhorn_reward": 0.712652325630188,
      "rewards/forensics_strict_edge_reward": 0.08076031506061554,
      "rewards/forensics_tiou_reward": 0.5163518041372299,
      "step": 994
    },
    {
      "completion_length": 24.609375,
      "epoch": 3.0522273425499233,
      "grad_norm": 9.85148811340332,
      "kl": 0.556640625,
      "learning_rate": 2.3538461538461537e-07,
      "loss": -0.0868,
      "reward": 0.7620203197002411,
      "reward_std": 0.03310360014438629,
      "reward_std_all": 0.03895263746380806,
      "rewards-max/forensics_sinkhorn_reward": 0.9825084805488586,
      "rewards-max/forensics_strict_edge_reward": 0.7933230996131897,
      "rewards-max/forensics_tiou_reward": 0.9831126034259796,
      "rewards/forensics_sinkhorn_reward": 0.732709139585495,
      "rewards/forensics_strict_edge_reward": 0.09862997010350227,
      "rewards/forensics_tiou_reward": 0.5802199244499207,
      "step": 995
    },
    {
      "completion_length": 24.5625,
      "epoch": 3.055299539170507,
      "grad_norm": 9.222381591796875,
      "kl": 0.58203125,
      "learning_rate": 2.346153846153846e-07,
      "loss": 0.1456,
      "reward": 0.7748778760433197,
      "reward_std": 0.026706541888415813,
      "reward_std_all": 0.03292514476925135,
      "rewards-max/forensics_sinkhorn_reward": 0.9748118221759796,
      "rewards-max/forensics_strict_edge_reward": 0.606867641210556,
      "rewards-max/forensics_tiou_reward": 0.9648146629333496,
      "rewards/forensics_sinkhorn_reward": 0.7297040522098541,
      "rewards/forensics_strict_edge_reward": 0.07179118320345879,
      "rewards/forensics_tiou_reward": 0.5456793308258057,
      "step": 996
    },
    {
      "completion_length": 24.796875,
      "epoch": 3.0583717357910905,
      "grad_norm": 11.719091415405273,
      "kl": 0.982421875,
      "learning_rate": 2.3384615384615382e-07,
      "loss": 0.0047,
      "reward": 0.7633886337280273,
      "reward_std": 0.009477059356868267,
      "reward_std_all": 0.028120139613747597,
      "rewards-max/forensics_sinkhorn_reward": 0.9823736846446991,
      "rewards-max/forensics_strict_edge_reward": 0.6419725120067596,
      "rewards-max/forensics_tiou_reward": 0.9793797731399536,
      "rewards/forensics_sinkhorn_reward": 0.6701580882072449,
      "rewards/forensics_strict_edge_reward": 0.08539911359548569,
      "rewards/forensics_tiou_reward": 0.578642338514328,
      "step": 997
    },
    {
      "completion_length": 24.65625,
      "epoch": 3.0614439324116742,
      "grad_norm": 17.682588577270508,
      "kl": 0.783203125,
      "learning_rate": 2.3307692307692306e-07,
      "loss": -0.0091,
      "reward": 0.7745504081249237,
      "reward_std": 0.011867685709148645,
      "reward_std_all": 0.02394199138507247,
      "rewards-max/forensics_sinkhorn_reward": 0.980235755443573,
      "rewards-max/forensics_strict_edge_reward": 0.6713652312755585,
      "rewards-max/forensics_tiou_reward": 0.9763874709606171,
      "rewards/forensics_sinkhorn_reward": 0.7348320484161377,
      "rewards/forensics_strict_edge_reward": 0.07902735844254494,
      "rewards/forensics_tiou_reward": 0.5089936554431915,
      "step": 998
    },
    {
      "completion_length": 22.734375,
      "epoch": 3.064516129032258,
      "grad_norm": 11.321547508239746,
      "kl": 0.814453125,
      "learning_rate": 2.323076923076923e-07,
      "loss": -0.0487,
      "reward": 0.7790433764457703,
      "reward_std": 0.03767955466173589,
      "reward_std_all": 0.0360516463406384,
      "rewards-max/forensics_sinkhorn_reward": 0.9786359667778015,
      "rewards-max/forensics_strict_edge_reward": 0.5623957812786102,
      "rewards-max/forensics_tiou_reward": 0.9705681800842285,
      "rewards/forensics_sinkhorn_reward": 0.808606892824173,
      "rewards/forensics_strict_edge_reward": 0.11371538788080215,
      "rewards/forensics_tiou_reward": 0.6757801175117493,
      "step": 999
    },
    {
      "completion_length": 26.546875,
      "epoch": 3.067588325652842,
      "grad_norm": 8.851463317871094,
      "kl": 0.53125,
      "learning_rate": 2.3153846153846151e-07,
      "loss": -0.0735,
      "reward": 0.7889774739742279,
      "reward_std": 0.019594542682170868,
      "reward_std_all": 0.020269472151994705,
      "rewards-max/forensics_sinkhorn_reward": 0.9846982955932617,
      "rewards-max/forensics_strict_edge_reward": 0.6839397251605988,
      "rewards-max/forensics_tiou_reward": 0.9910638928413391,
      "rewards/forensics_sinkhorn_reward": 0.7792534828186035,
      "rewards/forensics_strict_edge_reward": 0.11338844895362854,
      "rewards/forensics_tiou_reward": 0.6452775001525879,
      "step": 1000
    },
    {
      "completion_length": 23.8125,
      "epoch": 3.0706605222734256,
      "grad_norm": 8.686684608459473,
      "kl": 0.4599609375,
      "learning_rate": 2.3076923076923078e-07,
      "loss": -0.1032,
      "reward": 0.7818542420864105,
      "reward_std": 0.02900191769003868,
      "reward_std_all": 0.02852072287350893,
      "rewards-max/forensics_sinkhorn_reward": 0.9765941500663757,
      "rewards-max/forensics_strict_edge_reward": 0.6149742305278778,
      "rewards-max/forensics_tiou_reward": 0.969439297914505,
      "rewards/forensics_sinkhorn_reward": 0.748210608959198,
      "rewards/forensics_strict_edge_reward": 0.1322365142405033,
      "rewards/forensics_tiou_reward": 0.5922092795372009,
      "step": 1001
    },
    {
      "completion_length": 22.75,
      "epoch": 3.0737327188940093,
      "grad_norm": 11.690010070800781,
      "kl": 0.58203125,
      "learning_rate": 2.3e-07,
      "loss": 0.3821,
      "reward": 0.7381377220153809,
      "reward_std": 0.08525407686829567,
      "reward_std_all": 0.08025190234184265,
      "rewards-max/forensics_sinkhorn_reward": 0.9842254519462585,
      "rewards-max/forensics_strict_edge_reward": 0.6048777997493744,
      "rewards-max/forensics_tiou_reward": 0.9885050654411316,
      "rewards/forensics_sinkhorn_reward": 0.7021360099315643,
      "rewards/forensics_strict_edge_reward": 0.07851438038051128,
      "rewards/forensics_tiou_reward": 0.4446684718132019,
      "step": 1002
    },
    {
      "completion_length": 25.5625,
      "epoch": 3.076804915514593,
      "grad_norm": 9.80919361114502,
      "kl": 0.482421875,
      "learning_rate": 2.2923076923076924e-07,
      "loss": 0.1439,
      "reward": 0.7838813364505768,
      "reward_std": 0.02951518725603819,
      "reward_std_all": 0.02917593251913786,
      "rewards-max/forensics_sinkhorn_reward": 0.980522871017456,
      "rewards-max/forensics_strict_edge_reward": 0.756708562374115,
      "rewards-max/forensics_tiou_reward": 0.9785901308059692,
      "rewards/forensics_sinkhorn_reward": 0.75754714012146,
      "rewards/forensics_strict_edge_reward": 0.08669327013194561,
      "rewards/forensics_tiou_reward": 0.5634635388851166,
      "step": 1003
    },
    {
      "completion_length": 25.953125,
      "epoch": 3.0798771121351765,
      "grad_norm": 10.34571361541748,
      "kl": 0.5791015625,
      "learning_rate": 2.2846153846153845e-07,
      "loss": 0.121,
      "reward": 0.7723353803157806,
      "reward_std": 0.053984854370355606,
      "reward_std_all": 0.04895835742354393,
      "rewards-max/forensics_sinkhorn_reward": 0.9779509603977203,
      "rewards-max/forensics_strict_edge_reward": 0.5498771071434021,
      "rewards-max/forensics_tiou_reward": 0.9686667621135712,
      "rewards/forensics_sinkhorn_reward": 0.7110200524330139,
      "rewards/forensics_strict_edge_reward": 0.057920463383197784,
      "rewards/forensics_tiou_reward": 0.5049915015697479,
      "step": 1004
    },
    {
      "completion_length": 22.609375,
      "epoch": 3.0829493087557602,
      "grad_norm": 11.097173690795898,
      "kl": 0.6572265625,
      "learning_rate": 2.276923076923077e-07,
      "loss": 0.0875,
      "reward": 0.7424004971981049,
      "reward_std": 0.05131733417510986,
      "reward_std_all": 0.05673510208725929,
      "rewards-max/forensics_sinkhorn_reward": 0.9856413006782532,
      "rewards-max/forensics_strict_edge_reward": 0.6326714754104614,
      "rewards-max/forensics_tiou_reward": 0.9909990727901459,
      "rewards/forensics_sinkhorn_reward": 0.7488283216953278,
      "rewards/forensics_strict_edge_reward": 0.14549777656793594,
      "rewards/forensics_tiou_reward": 0.6217738091945648,
      "step": 1005
    },
    {
      "completion_length": 24.59375,
      "epoch": 3.086021505376344,
      "grad_norm": 12.690774917602539,
      "kl": 0.525390625,
      "learning_rate": 2.269230769230769e-07,
      "loss": 0.1294,
      "reward": 0.7763379812240601,
      "reward_std": 0.031078309752047062,
      "reward_std_all": 0.03234437480568886,
      "rewards-max/forensics_sinkhorn_reward": 0.9863139986991882,
      "rewards-max/forensics_strict_edge_reward": 0.5842172801494598,
      "rewards-max/forensics_tiou_reward": 0.991890162229538,
      "rewards/forensics_sinkhorn_reward": 0.737851083278656,
      "rewards/forensics_strict_edge_reward": 0.09885999001562595,
      "rewards/forensics_tiou_reward": 0.5723859667778015,
      "step": 1006
    },
    {
      "completion_length": 27.296875,
      "epoch": 3.089093701996928,
      "grad_norm": 9.317712783813477,
      "kl": 0.5341796875,
      "learning_rate": 2.2615384615384614e-07,
      "loss": -0.0467,
      "reward": 0.7951371669769287,
      "reward_std": 0.011018129996955395,
      "reward_std_all": 0.012483553029596806,
      "rewards-max/forensics_sinkhorn_reward": 0.9861955940723419,
      "rewards-max/forensics_strict_edge_reward": 0.7845968306064606,
      "rewards-max/forensics_tiou_reward": 0.9905660152435303,
      "rewards/forensics_sinkhorn_reward": 0.7776125073432922,
      "rewards/forensics_strict_edge_reward": 0.0996643640100956,
      "rewards/forensics_tiou_reward": 0.5975771844387054,
      "step": 1007
    },
    {
      "completion_length": 23.65625,
      "epoch": 3.0921658986175116,
      "grad_norm": 13.2013521194458,
      "kl": 0.47265625,
      "learning_rate": 2.2538461538461536e-07,
      "loss": -0.1907,
      "reward": 0.7276954352855682,
      "reward_std": 0.0339768948033452,
      "reward_std_all": 0.058123886585235596,
      "rewards-max/forensics_sinkhorn_reward": 0.9783595204353333,
      "rewards-max/forensics_strict_edge_reward": 0.6799494326114655,
      "rewards-max/forensics_tiou_reward": 0.9722472429275513,
      "rewards/forensics_sinkhorn_reward": 0.7346777319908142,
      "rewards/forensics_strict_edge_reward": 0.10098081454634666,
      "rewards/forensics_tiou_reward": 0.5917127728462219,
      "step": 1008
    },
    {
      "completion_length": 24.4375,
      "epoch": 3.0952380952380953,
      "grad_norm": 14.358156204223633,
      "kl": 1.095703125,
      "learning_rate": 2.246153846153846e-07,
      "loss": 0.3115,
      "reward": 0.7332040965557098,
      "reward_std": 0.08820484206080437,
      "reward_std_all": 0.08340467512607574,
      "rewards-max/forensics_sinkhorn_reward": 0.9795303046703339,
      "rewards-max/forensics_strict_edge_reward": 0.6398434042930603,
      "rewards-max/forensics_tiou_reward": 0.9790407419204712,
      "rewards/forensics_sinkhorn_reward": 0.7305903434753418,
      "rewards/forensics_strict_edge_reward": 0.10739780589938164,
      "rewards/forensics_tiou_reward": 0.5761516690254211,
      "step": 1009
    },
    {
      "completion_length": 25.65625,
      "epoch": 3.098310291858679,
      "grad_norm": 13.046225547790527,
      "kl": 0.517578125,
      "learning_rate": 2.238461538461538e-07,
      "loss": -0.3593,
      "reward": 0.8009577691555023,
      "reward_std": 0.010985754081048071,
      "reward_std_all": 0.011157271335832775,
      "rewards-max/forensics_sinkhorn_reward": 0.9884691536426544,
      "rewards-max/forensics_strict_edge_reward": 0.9513615369796753,
      "rewards-max/forensics_tiou_reward": 0.9985057413578033,
      "rewards/forensics_sinkhorn_reward": 0.8338417410850525,
      "rewards/forensics_strict_edge_reward": 0.19650252163410187,
      "rewards/forensics_tiou_reward": 0.7008049786090851,
      "step": 1010
    },
    {
      "completion_length": 26.25,
      "epoch": 3.1013824884792625,
      "grad_norm": 9.924314498901367,
      "kl": 0.4970703125,
      "learning_rate": 2.2307692307692308e-07,
      "loss": -0.2531,
      "reward": 0.7685781121253967,
      "reward_std": 0.019060999155044556,
      "reward_std_all": 0.027074376121163368,
      "rewards-max/forensics_sinkhorn_reward": 0.9751418828964233,
      "rewards-max/forensics_strict_edge_reward": 0.7426987290382385,
      "rewards-max/forensics_tiou_reward": 0.9644286036491394,
      "rewards/forensics_sinkhorn_reward": 0.7806404232978821,
      "rewards/forensics_strict_edge_reward": 0.11971231549978256,
      "rewards/forensics_tiou_reward": 0.6249465346336365,
      "step": 1011
    },
    {
      "completion_length": 21.71875,
      "epoch": 3.1044546850998462,
      "grad_norm": 12.773868560791016,
      "kl": 0.87890625,
      "learning_rate": 2.2230769230769232e-07,
      "loss": -0.1328,
      "reward": 0.7918692231178284,
      "reward_std": 0.012701806612312794,
      "reward_std_all": 0.014580035582184792,
      "rewards-max/forensics_sinkhorn_reward": 0.9837555885314941,
      "rewards-max/forensics_strict_edge_reward": 0.7584008872509003,
      "rewards-max/forensics_tiou_reward": 0.9878982305526733,
      "rewards/forensics_sinkhorn_reward": 0.8060365617275238,
      "rewards/forensics_strict_edge_reward": 0.1330556385219097,
      "rewards/forensics_tiou_reward": 0.6015640199184418,
      "step": 1012
    },
    {
      "completion_length": 24.578125,
      "epoch": 3.10752688172043,
      "grad_norm": 10.657224655151367,
      "kl": 0.609375,
      "learning_rate": 2.2153846153846153e-07,
      "loss": -0.0643,
      "reward": 0.7638933956623077,
      "reward_std": 0.005086091114208102,
      "reward_std_all": 0.024792807875201106,
      "rewards-max/forensics_sinkhorn_reward": 0.9857676923274994,
      "rewards-max/forensics_strict_edge_reward": 0.756708562374115,
      "rewards-max/forensics_tiou_reward": 0.9893669188022614,
      "rewards/forensics_sinkhorn_reward": 0.7158359587192535,
      "rewards/forensics_strict_edge_reward": 0.11114990338683128,
      "rewards/forensics_tiou_reward": 0.5766119956970215,
      "step": 1013
    },
    {
      "completion_length": 27.21875,
      "epoch": 3.110599078341014,
      "grad_norm": 8.443397521972656,
      "kl": 0.6328125,
      "learning_rate": 2.2076923076923077e-07,
      "loss": 0.0063,
      "reward": 0.7949003577232361,
      "reward_std": 0.012758678523823619,
      "reward_std_all": 0.01398683781735599,
      "rewards-max/forensics_sinkhorn_reward": 0.9859276711940765,
      "rewards-max/forensics_strict_edge_reward": 0.4303831309080124,
      "rewards-max/forensics_tiou_reward": 0.9895713031291962,
      "rewards/forensics_sinkhorn_reward": 0.7741310298442841,
      "rewards/forensics_strict_edge_reward": 0.07799148187041283,
      "rewards/forensics_tiou_reward": 0.6207358539104462,
      "step": 1014
    },
    {
      "completion_length": 26.75,
      "epoch": 3.1136712749615976,
      "grad_norm": 8.712143898010254,
      "kl": 0.486328125,
      "learning_rate": 2.1999999999999998e-07,
      "loss": 0.0607,
      "reward": 0.773267924785614,
      "reward_std": 0.052497985772788525,
      "reward_std_all": 0.047444443218410015,
      "rewards-max/forensics_sinkhorn_reward": 0.9854286313056946,
      "rewards-max/forensics_strict_edge_reward": 0.7309097647666931,
      "rewards-max/forensics_tiou_reward": 0.9909628033638,
      "rewards/forensics_sinkhorn_reward": 0.7547918856143951,
      "rewards/forensics_strict_edge_reward": 0.1180700920522213,
      "rewards/forensics_tiou_reward": 0.673438310623169,
      "step": 1015
    },
    {
      "completion_length": 25.859375,
      "epoch": 3.1167434715821813,
      "grad_norm": 11.230591773986816,
      "kl": 0.6064453125,
      "learning_rate": 2.1923076923076922e-07,
      "loss": 0.0063,
      "reward": 0.7808855175971985,
      "reward_std": 0.02158396504819393,
      "reward_std_all": 0.025406737811863422,
      "rewards-max/forensics_sinkhorn_reward": 0.980637788772583,
      "rewards-max/forensics_strict_edge_reward": 0.6969127506017685,
      "rewards-max/forensics_tiou_reward": 0.9787302613258362,
      "rewards/forensics_sinkhorn_reward": 0.801543116569519,
      "rewards/forensics_strict_edge_reward": 0.09181927517056465,
      "rewards/forensics_tiou_reward": 0.6785175502300262,
      "step": 1016
    },
    {
      "completion_length": 25.296875,
      "epoch": 3.119815668202765,
      "grad_norm": 11.036652565002441,
      "kl": 0.546875,
      "learning_rate": 2.1846153846153844e-07,
      "loss": -0.1119,
      "reward": 0.7943818271160126,
      "reward_std": 0.01865517906844616,
      "reward_std_all": 0.018238923978060484,
      "rewards-max/forensics_sinkhorn_reward": 0.9767725467681885,
      "rewards-max/forensics_strict_edge_reward": 0.42718979716300964,
      "rewards-max/forensics_tiou_reward": 0.966834306716919,
      "rewards/forensics_sinkhorn_reward": 0.7951666712760925,
      "rewards/forensics_strict_edge_reward": 0.06548207812011242,
      "rewards/forensics_tiou_reward": 0.6087374687194824,
      "step": 1017
    },
    {
      "completion_length": 26.578125,
      "epoch": 3.122887864823349,
      "grad_norm": 7.900203704833984,
      "kl": 0.419921875,
      "learning_rate": 2.1769230769230768e-07,
      "loss": -0.1234,
      "reward": 0.8019505739212036,
      "reward_std": 0.004851344681810588,
      "reward_std_all": 0.006548011209815741,
      "rewards-max/forensics_sinkhorn_reward": 0.9838016629219055,
      "rewards-max/forensics_strict_edge_reward": 0.6979724019765854,
      "rewards-max/forensics_tiou_reward": 0.9861599802970886,
      "rewards/forensics_sinkhorn_reward": 0.7923997342586517,
      "rewards/forensics_strict_edge_reward": 0.09613683074712753,
      "rewards/forensics_tiou_reward": 0.6335886418819427,
      "step": 1018
    },
    {
      "completion_length": 26.296875,
      "epoch": 3.1259600614439322,
      "grad_norm": 12.83120346069336,
      "kl": 0.458984375,
      "learning_rate": 2.169230769230769e-07,
      "loss": -0.0328,
      "reward": 0.7852828800678253,
      "reward_std": 0.02244620001874864,
      "reward_std_all": 0.024335441645234823,
      "rewards-max/forensics_sinkhorn_reward": 0.9859575927257538,
      "rewards-max/forensics_strict_edge_reward": 0.70003342628479,
      "rewards-max/forensics_tiou_reward": 0.9919241368770599,
      "rewards/forensics_sinkhorn_reward": 0.7393770813941956,
      "rewards/forensics_strict_edge_reward": 0.12148618325591087,
      "rewards/forensics_tiou_reward": 0.59817835688591,
      "step": 1019
    },
    {
      "completion_length": 27.40625,
      "epoch": 3.129032258064516,
      "grad_norm": 9.167922019958496,
      "kl": 0.650390625,
      "learning_rate": 2.1615384615384613e-07,
      "loss": 0.0474,
      "reward": 0.7730268239974976,
      "reward_std": 0.04118244722485542,
      "reward_std_all": 0.03962210565805435,
      "rewards-max/forensics_sinkhorn_reward": 0.9816903173923492,
      "rewards-max/forensics_strict_edge_reward": 0.5770299434661865,
      "rewards-max/forensics_tiou_reward": 0.9784297347068787,
      "rewards/forensics_sinkhorn_reward": 0.7444290220737457,
      "rewards/forensics_strict_edge_reward": 0.08945326507091522,
      "rewards/forensics_tiou_reward": 0.6256564557552338,
      "step": 1020
    },
    {
      "completion_length": 27.890625,
      "epoch": 3.1321044546851,
      "grad_norm": 13.08678150177002,
      "kl": 0.759765625,
      "learning_rate": 2.153846153846154e-07,
      "loss": 0.1735,
      "reward": 0.7748254239559174,
      "reward_std": 0.025532465428113937,
      "reward_std_all": 0.030024881474673748,
      "rewards-max/forensics_sinkhorn_reward": 0.9666935503482819,
      "rewards-max/forensics_strict_edge_reward": 0.4521050304174423,
      "rewards-max/forensics_tiou_reward": 0.942869246006012,
      "rewards/forensics_sinkhorn_reward": 0.7187557816505432,
      "rewards/forensics_strict_edge_reward": 0.06225511245429516,
      "rewards/forensics_tiou_reward": 0.5544276237487793,
      "step": 1021
    },
    {
      "completion_length": 25.703125,
      "epoch": 3.1351766513056836,
      "grad_norm": 12.864919662475586,
      "kl": 0.5986328125,
      "learning_rate": 2.146153846153846e-07,
      "loss": -0.0487,
      "reward": 0.7635539770126343,
      "reward_std": 0.046868106350302696,
      "reward_std_all": 0.04693254083395004,
      "rewards-max/forensics_sinkhorn_reward": 0.9777944982051849,
      "rewards-max/forensics_strict_edge_reward": 0.6074129343032837,
      "rewards-max/forensics_tiou_reward": 0.9736254811286926,
      "rewards/forensics_sinkhorn_reward": 0.6803697645664215,
      "rewards/forensics_strict_edge_reward": 0.06050818972289562,
      "rewards/forensics_tiou_reward": 0.5210807919502258,
      "step": 1022
    },
    {
      "completion_length": 25.515625,
      "epoch": 3.1382488479262673,
      "grad_norm": 11.431461334228516,
      "kl": 0.61328125,
      "learning_rate": 2.1384615384615385e-07,
      "loss": -0.26,
      "reward": 0.7844467461109161,
      "reward_std": 0.01018852461129427,
      "reward_std_all": 0.01657677232287824,
      "rewards-max/forensics_sinkhorn_reward": 0.9889904856681824,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.768378734588623,
      "rewards/forensics_strict_edge_reward": 0.12656494602560997,
      "rewards/forensics_tiou_reward": 0.5662369132041931,
      "step": 1023
    },
    {
      "completion_length": 25.140625,
      "epoch": 3.141321044546851,
      "grad_norm": 9.504521369934082,
      "kl": 0.77734375,
      "learning_rate": 2.1307692307692307e-07,
      "loss": 0.1096,
      "reward": 0.7737849950790405,
      "reward_std": 0.026023288257420063,
      "reward_std_all": 0.03259947802871466,
      "rewards-max/forensics_sinkhorn_reward": 0.988853394985199,
      "rewards-max/forensics_strict_edge_reward": 0.9687933325767517,
      "rewards-max/forensics_tiou_reward": 0.9996391236782074,
      "rewards/forensics_sinkhorn_reward": 0.7328790426254272,
      "rewards/forensics_strict_edge_reward": 0.10777921974658966,
      "rewards/forensics_tiou_reward": 0.569990485906601,
      "step": 1024
    },
    {
      "completion_length": 22.125,
      "epoch": 3.144393241167435,
      "grad_norm": 10.90342903137207,
      "kl": 0.576171875,
      "learning_rate": 2.123076923076923e-07,
      "loss": 0.0653,
      "reward": 0.7723986506462097,
      "reward_std": 0.04734366200864315,
      "reward_std_all": 0.04420273192226887,
      "rewards-max/forensics_sinkhorn_reward": 0.9867396354675293,
      "rewards-max/forensics_strict_edge_reward": 0.8751733303070068,
      "rewards-max/forensics_tiou_reward": 0.993508368730545,
      "rewards/forensics_sinkhorn_reward": 0.7990275025367737,
      "rewards/forensics_strict_edge_reward": 0.09884176775813103,
      "rewards/forensics_tiou_reward": 0.6342350840568542,
      "step": 1025
    },
    {
      "completion_length": 29.328125,
      "epoch": 3.1474654377880182,
      "grad_norm": 10.22278881072998,
      "kl": 0.5185546875,
      "learning_rate": 2.1153846153846152e-07,
      "loss": -0.0368,
      "reward": 0.7946857810020447,
      "reward_std": 0.010292315972037613,
      "reward_std_all": 0.012547065736725926,
      "rewards-max/forensics_sinkhorn_reward": 0.9633884429931641,
      "rewards-max/forensics_strict_edge_reward": 0.316214457154274,
      "rewards-max/forensics_tiou_reward": 0.936542421579361,
      "rewards/forensics_sinkhorn_reward": 0.7538422048091888,
      "rewards/forensics_strict_edge_reward": 0.06102496897801757,
      "rewards/forensics_tiou_reward": 0.5799675583839417,
      "step": 1026
    },
    {
      "completion_length": 26.453125,
      "epoch": 3.150537634408602,
      "grad_norm": 10.414627075195312,
      "kl": 0.455078125,
      "learning_rate": 2.1076923076923076e-07,
      "loss": -0.3078,
      "reward": 0.804792582988739,
      "reward_std": 0.0019140279036946595,
      "reward_std_all": 0.003935178508982062,
      "rewards-max/forensics_sinkhorn_reward": 0.9619578421115875,
      "rewards-max/forensics_strict_edge_reward": 0.6806710362434387,
      "rewards-max/forensics_tiou_reward": 0.9277725517749786,
      "rewards/forensics_sinkhorn_reward": 0.7691103518009186,
      "rewards/forensics_strict_edge_reward": 0.09674229472875595,
      "rewards/forensics_tiou_reward": 0.6131794452667236,
      "step": 1027
    },
    {
      "completion_length": 21.890625,
      "epoch": 3.153609831029186,
      "grad_norm": 14.286358833312988,
      "kl": 0.62890625,
      "learning_rate": 2.0999999999999997e-07,
      "loss": 0.1222,
      "reward": 0.759624183177948,
      "reward_std": 0.029871642589569092,
      "reward_std_all": 0.03876603953540325,
      "rewards-max/forensics_sinkhorn_reward": 0.983957827091217,
      "rewards-max/forensics_strict_edge_reward": 0.6284831166267395,
      "rewards-max/forensics_tiou_reward": 0.9841916263103485,
      "rewards/forensics_sinkhorn_reward": 0.702275961637497,
      "rewards/forensics_strict_edge_reward": 0.09061064198613167,
      "rewards/forensics_tiou_reward": 0.49535612761974335,
      "step": 1028
    },
    {
      "completion_length": 26.78125,
      "epoch": 3.1566820276497696,
      "grad_norm": 13.804088592529297,
      "kl": 0.5419921875,
      "learning_rate": 2.0923076923076921e-07,
      "loss": 0.0112,
      "reward": 0.7865538895130157,
      "reward_std": 0.026572407223284245,
      "reward_std_all": 0.02666928805410862,
      "rewards-max/forensics_sinkhorn_reward": 0.9668057858943939,
      "rewards-max/forensics_strict_edge_reward": 0.4407646059989929,
      "rewards-max/forensics_tiou_reward": 0.9485197961330414,
      "rewards/forensics_sinkhorn_reward": 0.7281638979911804,
      "rewards/forensics_strict_edge_reward": 0.04811702482402325,
      "rewards/forensics_tiou_reward": 0.5634385943412781,
      "step": 1029
    },
    {
      "completion_length": 25.328125,
      "epoch": 3.1597542242703534,
      "grad_norm": 12.805065155029297,
      "kl": 0.41015625,
      "learning_rate": 2.0846153846153843e-07,
      "loss": -0.2048,
      "reward": 0.7674314677715302,
      "reward_std": 0.03558688424527645,
      "reward_std_all": 0.0382271995767951,
      "rewards-max/forensics_sinkhorn_reward": 0.9832518100738525,
      "rewards-max/forensics_strict_edge_reward": 0.6862294971942902,
      "rewards-max/forensics_tiou_reward": 0.982595294713974,
      "rewards/forensics_sinkhorn_reward": 0.8053188621997833,
      "rewards/forensics_strict_edge_reward": 0.13198209553956985,
      "rewards/forensics_tiou_reward": 0.6914953589439392,
      "step": 1030
    },
    {
      "completion_length": 26.53125,
      "epoch": 3.162826420890937,
      "grad_norm": 8.968099594116211,
      "kl": 0.6640625,
      "learning_rate": 2.076923076923077e-07,
      "loss": 0.0014,
      "reward": 0.7825939059257507,
      "reward_std": 0.035290587693452835,
      "reward_std_all": 0.03265257924795151,
      "rewards-max/forensics_sinkhorn_reward": 0.9851895868778229,
      "rewards-max/forensics_strict_edge_reward": 0.711992472410202,
      "rewards-max/forensics_tiou_reward": 0.9886411726474762,
      "rewards/forensics_sinkhorn_reward": 0.8048542141914368,
      "rewards/forensics_strict_edge_reward": 0.19126416742801666,
      "rewards/forensics_tiou_reward": 0.6796916425228119,
      "step": 1031
    },
    {
      "completion_length": 27.328125,
      "epoch": 3.165898617511521,
      "grad_norm": 13.88664722442627,
      "kl": 1.115234375,
      "learning_rate": 2.0692307692307694e-07,
      "loss": 0.0383,
      "reward": 0.7654709219932556,
      "reward_std": 0.02986257988959551,
      "reward_std_all": 0.03654481563717127,
      "rewards-max/forensics_sinkhorn_reward": 0.9863084256649017,
      "rewards-max/forensics_strict_edge_reward": 0.6816803216934204,
      "rewards-max/forensics_tiou_reward": 0.9918996691703796,
      "rewards/forensics_sinkhorn_reward": 0.7787326872348785,
      "rewards/forensics_strict_edge_reward": 0.06584680452942848,
      "rewards/forensics_tiou_reward": 0.6237880885601044,
      "step": 1032
    },
    {
      "completion_length": 24.140625,
      "epoch": 3.1689708141321047,
      "grad_norm": 10.933161735534668,
      "kl": 0.580078125,
      "learning_rate": 2.0615384615384615e-07,
      "loss": 0.1203,
      "reward": 0.7501519024372101,
      "reward_std": 0.04074880527332425,
      "reward_std_all": 0.04819680657237768,
      "rewards-max/forensics_sinkhorn_reward": 0.9857393801212311,
      "rewards-max/forensics_strict_edge_reward": 0.6942952275276184,
      "rewards-max/forensics_tiou_reward": 0.9906535446643829,
      "rewards/forensics_sinkhorn_reward": 0.7088274359703064,
      "rewards/forensics_strict_edge_reward": 0.0755893662571907,
      "rewards/forensics_tiou_reward": 0.5193373560905457,
      "step": 1033
    },
    {
      "completion_length": 24.109375,
      "epoch": 3.172043010752688,
      "grad_norm": 11.133307456970215,
      "kl": 0.6640625,
      "learning_rate": 2.053846153846154e-07,
      "loss": 0.0102,
      "reward": 0.7526330053806305,
      "reward_std": 0.05142364092171192,
      "reward_std_all": 0.053132278844714165,
      "rewards-max/forensics_sinkhorn_reward": 0.9859499037265778,
      "rewards-max/forensics_strict_edge_reward": 0.5816259980201721,
      "rewards-max/forensics_tiou_reward": 0.990219920873642,
      "rewards/forensics_sinkhorn_reward": 0.6963511109352112,
      "rewards/forensics_strict_edge_reward": 0.09620623290538788,
      "rewards/forensics_tiou_reward": 0.5227176547050476,
      "step": 1034
    },
    {
      "completion_length": 23.96875,
      "epoch": 3.175115207373272,
      "grad_norm": 11.936077117919922,
      "kl": 0.3935546875,
      "learning_rate": 2.046153846153846e-07,
      "loss": -0.059,
      "reward": 0.788381427526474,
      "reward_std": 0.027472587302327156,
      "reward_std_all": 0.026039606891572475,
      "rewards-max/forensics_sinkhorn_reward": 0.9850491583347321,
      "rewards-max/forensics_strict_edge_reward": 0.6965087056159973,
      "rewards-max/forensics_tiou_reward": 0.9891480207443237,
      "rewards/forensics_sinkhorn_reward": 0.7624236941337585,
      "rewards/forensics_strict_edge_reward": 0.16017688810825348,
      "rewards/forensics_tiou_reward": 0.6585894525051117,
      "step": 1035
    },
    {
      "completion_length": 25.6875,
      "epoch": 3.1781874039938556,
      "grad_norm": 11.329122543334961,
      "kl": 0.87890625,
      "learning_rate": 2.0384615384615384e-07,
      "loss": 0.1334,
      "reward": 0.6893226504325867,
      "reward_std": 0.05954660661518574,
      "reward_std_all": 0.08808278292417526,
      "rewards-max/forensics_sinkhorn_reward": 0.9863787889480591,
      "rewards-max/forensics_strict_edge_reward": 0.9093653857707977,
      "rewards-max/forensics_tiou_reward": 0.9933711290359497,
      "rewards/forensics_sinkhorn_reward": 0.6300737857818604,
      "rewards/forensics_strict_edge_reward": 0.11732914671301842,
      "rewards/forensics_tiou_reward": 0.4467478543519974,
      "step": 1036
    },
    {
      "completion_length": 30.609375,
      "epoch": 3.1812596006144394,
      "grad_norm": 8.51044750213623,
      "kl": 0.609375,
      "learning_rate": 2.0307692307692306e-07,
      "loss": -0.0593,
      "reward": 0.7678152322769165,
      "reward_std": 0.054224470630288124,
      "reward_std_all": 0.050423478707671165,
      "rewards-max/forensics_sinkhorn_reward": 0.9682612121105194,
      "rewards-max/forensics_strict_edge_reward": 0.31389058381319046,
      "rewards-max/forensics_tiou_reward": 0.9464614689350128,
      "rewards/forensics_sinkhorn_reward": 0.7256855964660645,
      "rewards/forensics_strict_edge_reward": 0.06894002482295036,
      "rewards/forensics_tiou_reward": 0.5372998118400574,
      "step": 1037
    },
    {
      "completion_length": 29.5625,
      "epoch": 3.184331797235023,
      "grad_norm": 10.05550765991211,
      "kl": 0.5947265625,
      "learning_rate": 2.023076923076923e-07,
      "loss": 0.0222,
      "reward": 0.7739395797252655,
      "reward_std": 0.04058871604502201,
      "reward_std_all": 0.03940512053668499,
      "rewards-max/forensics_sinkhorn_reward": 0.9834068715572357,
      "rewards-max/forensics_strict_edge_reward": 0.4392508268356323,
      "rewards-max/forensics_tiou_reward": 0.9863285720348358,
      "rewards/forensics_sinkhorn_reward": 0.7326653003692627,
      "rewards/forensics_strict_edge_reward": 0.07485641166567802,
      "rewards/forensics_tiou_reward": 0.6001229882240295,
      "step": 1038
    },
    {
      "completion_length": 26.421875,
      "epoch": 3.187403993855607,
      "grad_norm": 9.79248046875,
      "kl": 0.736328125,
      "learning_rate": 2.015384615384615e-07,
      "loss": 0.0119,
      "reward": 0.7578538060188293,
      "reward_std": 0.034249285236001015,
      "reward_std_all": 0.040664203464984894,
      "rewards-max/forensics_sinkhorn_reward": 0.9737173616886139,
      "rewards-max/forensics_strict_edge_reward": 0.44204281270504,
      "rewards-max/forensics_tiou_reward": 0.9664308428764343,
      "rewards/forensics_sinkhorn_reward": 0.7317987382411957,
      "rewards/forensics_strict_edge_reward": 0.06236657314002514,
      "rewards/forensics_tiou_reward": 0.5869196951389313,
      "step": 1039
    },
    {
      "completion_length": 26.6875,
      "epoch": 3.1904761904761907,
      "grad_norm": 10.006636619567871,
      "kl": 0.541015625,
      "learning_rate": 2.0076923076923075e-07,
      "loss": -0.094,
      "reward": 0.7796438038349152,
      "reward_std": 0.02680684777442366,
      "reward_std_all": 0.027870773104950786,
      "rewards-max/forensics_sinkhorn_reward": 0.9818227291107178,
      "rewards-max/forensics_strict_edge_reward": 0.5338919758796692,
      "rewards-max/forensics_tiou_reward": 0.9814505279064178,
      "rewards/forensics_sinkhorn_reward": 0.7555534839630127,
      "rewards/forensics_strict_edge_reward": 0.08040553890168667,
      "rewards/forensics_tiou_reward": 0.5553500950336456,
      "step": 1040
    },
    {
      "completion_length": 26.671875,
      "epoch": 3.193548387096774,
      "grad_norm": 9.83788776397705,
      "kl": 0.49609375,
      "learning_rate": 2e-07,
      "loss": 0.0576,
      "reward": 0.7724260985851288,
      "reward_std": 0.04519714927300811,
      "reward_std_all": 0.0429132878780365,
      "rewards-max/forensics_sinkhorn_reward": 0.9846423268318176,
      "rewards-max/forensics_strict_edge_reward": 0.5904144644737244,
      "rewards-max/forensics_tiou_reward": 0.9890052378177643,
      "rewards/forensics_sinkhorn_reward": 0.7489189207553864,
      "rewards/forensics_strict_edge_reward": 0.10601243749260902,
      "rewards/forensics_tiou_reward": 0.6033157706260681,
      "step": 1041
    },
    {
      "completion_length": 28.953125,
      "epoch": 3.196620583717358,
      "grad_norm": 11.523316383361816,
      "kl": 0.5927734375,
      "learning_rate": 1.9923076923076923e-07,
      "loss": 0.092,
      "reward": 0.7184911072254181,
      "reward_std": 0.04832548648118973,
      "reward_std_all": 0.06457317620515823,
      "rewards-max/forensics_sinkhorn_reward": 0.9823199212551117,
      "rewards-max/forensics_strict_edge_reward": 0.7366199493408203,
      "rewards-max/forensics_tiou_reward": 0.9834098815917969,
      "rewards/forensics_sinkhorn_reward": 0.6843736171722412,
      "rewards/forensics_strict_edge_reward": 0.09915617853403091,
      "rewards/forensics_tiou_reward": 0.5729193389415741,
      "step": 1042
    },
    {
      "completion_length": 22.109375,
      "epoch": 3.1996927803379416,
      "grad_norm": 11.794049263000488,
      "kl": 0.53515625,
      "learning_rate": 1.9846153846153844e-07,
      "loss": -0.2381,
      "reward": 0.7973321676254272,
      "reward_std": 0.006562262773513794,
      "reward_std_all": 0.009070752072148025,
      "rewards-max/forensics_sinkhorn_reward": 0.9813359379768372,
      "rewards-max/forensics_strict_edge_reward": 0.7401526570320129,
      "rewards-max/forensics_tiou_reward": 0.9796088933944702,
      "rewards/forensics_sinkhorn_reward": 0.8060451447963715,
      "rewards/forensics_strict_edge_reward": 0.13080714270472527,
      "rewards/forensics_tiou_reward": 0.6064084768295288,
      "step": 1043
    },
    {
      "completion_length": 29.03125,
      "epoch": 3.2027649769585254,
      "grad_norm": 12.566452026367188,
      "kl": 0.3486328125,
      "learning_rate": 1.9769230769230768e-07,
      "loss": 0.0567,
      "reward": 0.7656637132167816,
      "reward_std": 0.03189203981310129,
      "reward_std_all": 0.03712217276915908,
      "rewards-max/forensics_sinkhorn_reward": 0.9816352725028992,
      "rewards-max/forensics_strict_edge_reward": 0.6316592693328857,
      "rewards-max/forensics_tiou_reward": 0.9812112152576447,
      "rewards/forensics_sinkhorn_reward": 0.7556421458721161,
      "rewards/forensics_strict_edge_reward": 0.13532817363739014,
      "rewards/forensics_tiou_reward": 0.6449772417545319,
      "step": 1044
    },
    {
      "completion_length": 22.625,
      "epoch": 3.205837173579109,
      "grad_norm": 12.84003734588623,
      "kl": 0.4833984375,
      "learning_rate": 1.9692307692307693e-07,
      "loss": -0.1188,
      "reward": 0.7266234755516052,
      "reward_std": 0.021213099360466003,
      "reward_std_all": 0.05130944959819317,
      "rewards-max/forensics_sinkhorn_reward": 0.982104480266571,
      "rewards-max/forensics_strict_edge_reward": 0.7053695321083069,
      "rewards-max/forensics_tiou_reward": 0.9810059070587158,
      "rewards/forensics_sinkhorn_reward": 0.6969864964485168,
      "rewards/forensics_strict_edge_reward": 0.11700975894927979,
      "rewards/forensics_tiou_reward": 0.5582741796970367,
      "step": 1045
    },
    {
      "completion_length": 24.6875,
      "epoch": 3.208909370199693,
      "grad_norm": 10.541098594665527,
      "kl": 0.77734375,
      "learning_rate": 1.9615384615384614e-07,
      "loss": 0.1556,
      "reward": 0.7515695989131927,
      "reward_std": 0.02663218043744564,
      "reward_std_all": 0.043190876953303814,
      "rewards-max/forensics_sinkhorn_reward": 0.9811863005161285,
      "rewards-max/forensics_strict_edge_reward": 0.49066421389579773,
      "rewards-max/forensics_tiou_reward": 0.9761652052402496,
      "rewards/forensics_sinkhorn_reward": 0.7454077005386353,
      "rewards/forensics_strict_edge_reward": 0.05680941790342331,
      "rewards/forensics_tiou_reward": 0.5797846019268036,
      "step": 1046
    },
    {
      "completion_length": 23.46875,
      "epoch": 3.2119815668202767,
      "grad_norm": 12.090337753295898,
      "kl": 0.4248046875,
      "learning_rate": 1.9538461538461538e-07,
      "loss": -0.1685,
      "reward": 0.7839131057262421,
      "reward_std": 0.026182305067777634,
      "reward_std_all": 0.026091618463397026,
      "rewards-max/forensics_sinkhorn_reward": 0.9823909401893616,
      "rewards-max/forensics_strict_edge_reward": 0.5622940063476562,
      "rewards-max/forensics_tiou_reward": 0.9807852804660797,
      "rewards/forensics_sinkhorn_reward": 0.7902420163154602,
      "rewards/forensics_strict_edge_reward": 0.08198268339037895,
      "rewards/forensics_tiou_reward": 0.6263845562934875,
      "step": 1047
    },
    {
      "completion_length": 24.015625,
      "epoch": 3.21505376344086,
      "grad_norm": 16.315174102783203,
      "kl": 0.6171875,
      "learning_rate": 1.946153846153846e-07,
      "loss": 0.0475,
      "reward": 0.7392905652523041,
      "reward_std": 0.02949559036642313,
      "reward_std_all": 0.04756708722561598,
      "rewards-max/forensics_sinkhorn_reward": 0.9873576760292053,
      "rewards-max/forensics_strict_edge_reward": 0.8191856443881989,
      "rewards-max/forensics_tiou_reward": 0.9953399300575256,
      "rewards/forensics_sinkhorn_reward": 0.7018038034439087,
      "rewards/forensics_strict_edge_reward": 0.1209670677781105,
      "rewards/forensics_tiou_reward": 0.553478866815567,
      "step": 1048
    },
    {
      "completion_length": 26.421875,
      "epoch": 3.218125960061444,
      "grad_norm": 10.409002304077148,
      "kl": 0.58203125,
      "learning_rate": 1.9384615384615383e-07,
      "loss": 0.0231,
      "reward": 0.756761223077774,
      "reward_std": 0.05335655715316534,
      "reward_std_all": 0.05294530838727951,
      "rewards-max/forensics_sinkhorn_reward": 0.9762279391288757,
      "rewards-max/forensics_strict_edge_reward": 0.5566954910755157,
      "rewards-max/forensics_tiou_reward": 0.9659262001514435,
      "rewards/forensics_sinkhorn_reward": 0.7050058245658875,
      "rewards/forensics_strict_edge_reward": 0.07007511891424656,
      "rewards/forensics_tiou_reward": 0.5131023079156876,
      "step": 1049
    },
    {
      "completion_length": 29.640625,
      "epoch": 3.2211981566820276,
      "grad_norm": 8.951743125915527,
      "kl": 0.541015625,
      "learning_rate": 1.9307692307692307e-07,
      "loss": 0.0864,
      "reward": 0.7607138752937317,
      "reward_std": 0.012859299837145954,
      "reward_std_all": 0.0282118774484843,
      "rewards-max/forensics_sinkhorn_reward": 0.983132928609848,
      "rewards-max/forensics_strict_edge_reward": 0.5913059562444687,
      "rewards-max/forensics_tiou_reward": 0.9832644760608673,
      "rewards/forensics_sinkhorn_reward": 0.7671739161014557,
      "rewards/forensics_strict_edge_reward": 0.11030208319425583,
      "rewards/forensics_tiou_reward": 0.6483198702335358,
      "step": 1050
    },
    {
      "completion_length": 25.15625,
      "epoch": 3.2242703533026114,
      "grad_norm": 28.434816360473633,
      "kl": 2.0625,
      "learning_rate": 1.9230769230769231e-07,
      "loss": 0.0372,
      "reward": 0.7553012371063232,
      "reward_std": 0.02335476316511631,
      "reward_std_all": 0.03604570962488651,
      "rewards-max/forensics_sinkhorn_reward": 0.9868365824222565,
      "rewards-max/forensics_strict_edge_reward": 0.8751733303070068,
      "rewards-max/forensics_tiou_reward": 0.9941255748271942,
      "rewards/forensics_sinkhorn_reward": 0.6873256862163544,
      "rewards/forensics_strict_edge_reward": 0.13980020582675934,
      "rewards/forensics_tiou_reward": 0.5853381752967834,
      "step": 1051
    },
    {
      "completion_length": 25.34375,
      "epoch": 3.227342549923195,
      "grad_norm": 9.202681541442871,
      "kl": 0.4794921875,
      "learning_rate": 1.9153846153846153e-07,
      "loss": -0.0862,
      "reward": 0.7874877750873566,
      "reward_std": 0.010112961754202843,
      "reward_std_all": 0.015401806915178895,
      "rewards-max/forensics_sinkhorn_reward": 0.9798099994659424,
      "rewards-max/forensics_strict_edge_reward": 0.588144987821579,
      "rewards-max/forensics_tiou_reward": 0.979939192533493,
      "rewards/forensics_sinkhorn_reward": 0.7781802415847778,
      "rewards/forensics_strict_edge_reward": 0.05882251635193825,
      "rewards/forensics_tiou_reward": 0.5830037295818329,
      "step": 1052
    },
    {
      "completion_length": 27.84375,
      "epoch": 3.230414746543779,
      "grad_norm": 10.808487892150879,
      "kl": 0.5341796875,
      "learning_rate": 1.9076923076923077e-07,
      "loss": 0.2099,
      "reward": 0.769083172082901,
      "reward_std": 0.017479603178799152,
      "reward_std_all": 0.027167372405529022,
      "rewards-max/forensics_sinkhorn_reward": 0.9427746534347534,
      "rewards-max/forensics_strict_edge_reward": 0.46152280271053314,
      "rewards-max/forensics_tiou_reward": 0.8923760652542114,
      "rewards/forensics_sinkhorn_reward": 0.6947170197963715,
      "rewards/forensics_strict_edge_reward": 0.07435917668044567,
      "rewards/forensics_tiou_reward": 0.5297241508960724,
      "step": 1053
    },
    {
      "completion_length": 24.484375,
      "epoch": 3.2334869431643627,
      "grad_norm": 10.645511627197266,
      "kl": 0.67578125,
      "learning_rate": 1.8999999999999998e-07,
      "loss": 0.2003,
      "reward": 0.7979917526245117,
      "reward_std": 0.00877540372312069,
      "reward_std_all": 0.010657336562871933,
      "rewards-max/forensics_sinkhorn_reward": 0.9773819446563721,
      "rewards-max/forensics_strict_edge_reward": 0.5846173912286758,
      "rewards-max/forensics_tiou_reward": 0.9708908498287201,
      "rewards/forensics_sinkhorn_reward": 0.7591448426246643,
      "rewards/forensics_strict_edge_reward": 0.07478048279881477,
      "rewards/forensics_tiou_reward": 0.6045563220977783,
      "step": 1054
    },
    {
      "completion_length": 24.828125,
      "epoch": 3.236559139784946,
      "grad_norm": 9.603434562683105,
      "kl": 0.5673828125,
      "learning_rate": 1.8923076923076922e-07,
      "loss": 0.2748,
      "reward": 0.7435462772846222,
      "reward_std": 0.054538071155548096,
      "reward_std_all": 0.06152229383587837,
      "rewards-max/forensics_sinkhorn_reward": 0.9866584241390228,
      "rewards-max/forensics_strict_edge_reward": 0.7933230996131897,
      "rewards-max/forensics_tiou_reward": 0.994486391544342,
      "rewards/forensics_sinkhorn_reward": 0.6980298757553101,
      "rewards/forensics_strict_edge_reward": 0.07961671054363251,
      "rewards/forensics_tiou_reward": 0.5236808657646179,
      "step": 1055
    },
    {
      "completion_length": 25.546875,
      "epoch": 3.23963133640553,
      "grad_norm": 10.353116989135742,
      "kl": 0.625,
      "learning_rate": 1.8846153846153846e-07,
      "loss": 0.1031,
      "reward": 0.727243185043335,
      "reward_std": 0.05546021740883589,
      "reward_std_all": 0.06456212513148785,
      "rewards-max/forensics_sinkhorn_reward": 0.9714190065860748,
      "rewards-max/forensics_strict_edge_reward": 0.3642989844083786,
      "rewards-max/forensics_tiou_reward": 0.9526975750923157,
      "rewards/forensics_sinkhorn_reward": 0.6491531133651733,
      "rewards/forensics_strict_edge_reward": 0.03658119589090347,
      "rewards/forensics_tiou_reward": 0.4675631821155548,
      "step": 1056
    },
    {
      "completion_length": 30.703125,
      "epoch": 3.2427035330261136,
      "grad_norm": 7.92757511138916,
      "kl": 0.515625,
      "learning_rate": 1.8769230769230767e-07,
      "loss": -0.0312,
      "reward": 0.7791972160339355,
      "reward_std": 0.026300831697881222,
      "reward_std_all": 0.028495989739894867,
      "rewards-max/forensics_sinkhorn_reward": 0.984833687543869,
      "rewards-max/forensics_strict_edge_reward": 0.6689256727695465,
      "rewards-max/forensics_tiou_reward": 0.9895782768726349,
      "rewards/forensics_sinkhorn_reward": 0.7309826612472534,
      "rewards/forensics_strict_edge_reward": 0.11281915754079819,
      "rewards/forensics_tiou_reward": 0.6083294451236725,
      "step": 1057
    },
    {
      "completion_length": 23.828125,
      "epoch": 3.2457757296466974,
      "grad_norm": 12.107780456542969,
      "kl": 0.6689453125,
      "learning_rate": 1.8692307692307692e-07,
      "loss": 0.0163,
      "reward": 0.7694286704063416,
      "reward_std": 0.042211148887872696,
      "reward_std_all": 0.04145372658967972,
      "rewards-max/forensics_sinkhorn_reward": 0.9744237065315247,
      "rewards-max/forensics_strict_edge_reward": 0.5422053933143616,
      "rewards-max/forensics_tiou_reward": 0.9644707143306732,
      "rewards/forensics_sinkhorn_reward": 0.7324880957603455,
      "rewards/forensics_strict_edge_reward": 0.09060569480061531,
      "rewards/forensics_tiou_reward": 0.5350401997566223,
      "step": 1058
    },
    {
      "completion_length": 25.09375,
      "epoch": 3.248847926267281,
      "grad_norm": 8.725454330444336,
      "kl": 0.6337890625,
      "learning_rate": 1.8615384615384613e-07,
      "loss": 0.0443,
      "reward": 0.7926886677742004,
      "reward_std": 0.01734220440266654,
      "reward_std_all": 0.01808916055597365,
      "rewards-max/forensics_sinkhorn_reward": 0.9793207943439484,
      "rewards-max/forensics_strict_edge_reward": 0.5940010249614716,
      "rewards-max/forensics_tiou_reward": 0.9736440479755402,
      "rewards/forensics_sinkhorn_reward": 0.7608185708522797,
      "rewards/forensics_strict_edge_reward": 0.05793646723031998,
      "rewards/forensics_tiou_reward": 0.5609709322452545,
      "step": 1059
    },
    {
      "completion_length": 23.8125,
      "epoch": 3.251920122887865,
      "grad_norm": 11.45996379852295,
      "kl": 0.548828125,
      "learning_rate": 1.853846153846154e-07,
      "loss": -0.0766,
      "reward": 0.7659368813037872,
      "reward_std": 0.061570351012051105,
      "reward_std_all": 0.05598472384735942,
      "rewards-max/forensics_sinkhorn_reward": 0.9859670400619507,
      "rewards-max/forensics_strict_edge_reward": 0.46924614906311035,
      "rewards-max/forensics_tiou_reward": 0.9932555854320526,
      "rewards/forensics_sinkhorn_reward": 0.7352160215377808,
      "rewards/forensics_strict_edge_reward": 0.05875750631093979,
      "rewards/forensics_tiou_reward": 0.6174070239067078,
      "step": 1060
    },
    {
      "completion_length": 24.34375,
      "epoch": 3.2549923195084487,
      "grad_norm": 10.889105796813965,
      "kl": 0.4248046875,
      "learning_rate": 1.846153846153846e-07,
      "loss": -0.0937,
      "reward": 0.7810481190681458,
      "reward_std": 0.02913463395088911,
      "reward_std_all": 0.029136648401618004,
      "rewards-max/forensics_sinkhorn_reward": 0.9757209420204163,
      "rewards-max/forensics_strict_edge_reward": 0.6398434042930603,
      "rewards-max/forensics_tiou_reward": 0.9681254625320435,
      "rewards/forensics_sinkhorn_reward": 0.7753607928752899,
      "rewards/forensics_strict_edge_reward": 0.1622835248708725,
      "rewards/forensics_tiou_reward": 0.5959343910217285,
      "step": 1061
    },
    {
      "completion_length": 25.421875,
      "epoch": 3.258064516129032,
      "grad_norm": 11.81411361694336,
      "kl": 0.5859375,
      "learning_rate": 1.8384615384615385e-07,
      "loss": 0.0007,
      "reward": 0.7452268600463867,
      "reward_std": 0.03963633719831705,
      "reward_std_all": 0.049949998036026955,
      "rewards-max/forensics_sinkhorn_reward": 0.9796152710914612,
      "rewards-max/forensics_strict_edge_reward": 0.742465078830719,
      "rewards-max/forensics_tiou_reward": 0.9771205484867096,
      "rewards/forensics_sinkhorn_reward": 0.677926629781723,
      "rewards/forensics_strict_edge_reward": 0.11086535081267357,
      "rewards/forensics_tiou_reward": 0.5140054225921631,
      "step": 1062
    },
    {
      "completion_length": 23.234375,
      "epoch": 3.261136712749616,
      "grad_norm": 12.90029239654541,
      "kl": 0.5078125,
      "learning_rate": 1.8307692307692306e-07,
      "loss": 0.3077,
      "reward": 0.7463051378726959,
      "reward_std": 0.041866617277264595,
      "reward_std_all": 0.054390549659729004,
      "rewards-max/forensics_sinkhorn_reward": 0.982519268989563,
      "rewards-max/forensics_strict_edge_reward": 0.661531001329422,
      "rewards-max/forensics_tiou_reward": 0.9817201495170593,
      "rewards/forensics_sinkhorn_reward": 0.6730666756629944,
      "rewards/forensics_strict_edge_reward": 0.09301762655377388,
      "rewards/forensics_tiou_reward": 0.5584396719932556,
      "step": 1063
    },
    {
      "completion_length": 25.796875,
      "epoch": 3.2642089093701996,
      "grad_norm": 10.500345230102539,
      "kl": 0.4814453125,
      "learning_rate": 1.823076923076923e-07,
      "loss": -0.1823,
      "reward": 0.770227700471878,
      "reward_std": 0.023895557038486004,
      "reward_std_all": 0.031633246690034866,
      "rewards-max/forensics_sinkhorn_reward": 0.9731437265872955,
      "rewards-max/forensics_strict_edge_reward": 0.5600500553846359,
      "rewards-max/forensics_tiou_reward": 0.9537522494792938,
      "rewards/forensics_sinkhorn_reward": 0.7675591707229614,
      "rewards/forensics_strict_edge_reward": 0.08343293517827988,
      "rewards/forensics_tiou_reward": 0.5764312148094177,
      "step": 1064
    },
    {
      "completion_length": 27.109375,
      "epoch": 3.2672811059907834,
      "grad_norm": 13.041093826293945,
      "kl": 0.802734375,
      "learning_rate": 1.8153846153846152e-07,
      "loss": -0.0549,
      "reward": 0.7903965413570404,
      "reward_std": 0.02854064991697669,
      "reward_std_all": 0.02635293360799551,
      "rewards-max/forensics_sinkhorn_reward": 0.9878159165382385,
      "rewards-max/forensics_strict_edge_reward": 0.9232408702373505,
      "rewards-max/forensics_tiou_reward": 0.996440976858139,
      "rewards/forensics_sinkhorn_reward": 0.7994845807552338,
      "rewards/forensics_strict_edge_reward": 0.1309134140610695,
      "rewards/forensics_tiou_reward": 0.6629931628704071,
      "step": 1065
    },
    {
      "completion_length": 24.34375,
      "epoch": 3.270353302611367,
      "grad_norm": 12.015063285827637,
      "kl": 0.611328125,
      "learning_rate": 1.8076923076923076e-07,
      "loss": 0.0727,
      "reward": 0.78069868683815,
      "reward_std": 0.034434753470122814,
      "reward_std_all": 0.03328828141093254,
      "rewards-max/forensics_sinkhorn_reward": 0.985606461763382,
      "rewards-max/forensics_strict_edge_reward": 0.8582656681537628,
      "rewards-max/forensics_tiou_reward": 0.9920602142810822,
      "rewards/forensics_sinkhorn_reward": 0.7596492171287537,
      "rewards/forensics_strict_edge_reward": 0.09183506295084953,
      "rewards/forensics_tiou_reward": 0.6372022032737732,
      "step": 1066
    },
    {
      "completion_length": 28.234375,
      "epoch": 3.273425499231951,
      "grad_norm": 10.171852111816406,
      "kl": 0.685546875,
      "learning_rate": 1.8e-07,
      "loss": 0.107,
      "reward": 0.7892045676708221,
      "reward_std": 0.012316480744630098,
      "reward_std_all": 0.015943581238389015,
      "rewards-max/forensics_sinkhorn_reward": 0.978324294090271,
      "rewards-max/forensics_strict_edge_reward": 0.5499051213264465,
      "rewards-max/forensics_tiou_reward": 0.9705860316753387,
      "rewards/forensics_sinkhorn_reward": 0.7451528906822205,
      "rewards/forensics_strict_edge_reward": 0.06682022660970688,
      "rewards/forensics_tiou_reward": 0.5603512823581696,
      "step": 1067
    },
    {
      "completion_length": 25.828125,
      "epoch": 3.2764976958525347,
      "grad_norm": 8.849101066589355,
      "kl": 0.595703125,
      "learning_rate": 1.792307692307692e-07,
      "loss": -0.1413,
      "reward": 0.7860701978206635,
      "reward_std": 0.035658450331538916,
      "reward_std_all": 0.03247113572433591,
      "rewards-max/forensics_sinkhorn_reward": 0.9756926000118256,
      "rewards-max/forensics_strict_edge_reward": 0.57895427942276,
      "rewards-max/forensics_tiou_reward": 0.9671646654605865,
      "rewards/forensics_sinkhorn_reward": 0.773285835981369,
      "rewards/forensics_strict_edge_reward": 0.12723995372653008,
      "rewards/forensics_tiou_reward": 0.6556328535079956,
      "step": 1068
    },
    {
      "completion_length": 23.78125,
      "epoch": 3.279569892473118,
      "grad_norm": 15.84067153930664,
      "kl": 0.5234375,
      "learning_rate": 1.7846153846153845e-07,
      "loss": -0.0729,
      "reward": 0.7488717138767242,
      "reward_std": 0.04056026041507721,
      "reward_std_all": 0.046404313296079636,
      "rewards-max/forensics_sinkhorn_reward": 0.9813664555549622,
      "rewards-max/forensics_strict_edge_reward": 0.7038446068763733,
      "rewards-max/forensics_tiou_reward": 0.9723600447177887,
      "rewards/forensics_sinkhorn_reward": 0.7459478974342346,
      "rewards/forensics_strict_edge_reward": 0.11662273854017258,
      "rewards/forensics_tiou_reward": 0.5697741806507111,
      "step": 1069
    },
    {
      "completion_length": 28.53125,
      "epoch": 3.282642089093702,
      "grad_norm": 14.227921485900879,
      "kl": 0.642578125,
      "learning_rate": 1.776923076923077e-07,
      "loss": 0.0233,
      "reward": 0.7859213650226593,
      "reward_std": 0.02920352201908827,
      "reward_std_all": 0.027710767462849617,
      "rewards-max/forensics_sinkhorn_reward": 0.9843239486217499,
      "rewards-max/forensics_strict_edge_reward": 0.7224559187889099,
      "rewards-max/forensics_tiou_reward": 0.9901764988899231,
      "rewards/forensics_sinkhorn_reward": 0.7478899657726288,
      "rewards/forensics_strict_edge_reward": 0.09582870453596115,
      "rewards/forensics_tiou_reward": 0.5987953841686249,
      "step": 1070
    },
    {
      "completion_length": 28.390625,
      "epoch": 3.2857142857142856,
      "grad_norm": 11.749160766601562,
      "kl": 0.517578125,
      "learning_rate": 1.7692307692307693e-07,
      "loss": -0.0711,
      "reward": 0.7926065027713776,
      "reward_std": 0.023392796982079744,
      "reward_std_all": 0.022022024728357792,
      "rewards-max/forensics_sinkhorn_reward": 0.9788734316825867,
      "rewards-max/forensics_strict_edge_reward": 0.6326714754104614,
      "rewards-max/forensics_tiou_reward": 0.9741952121257782,
      "rewards/forensics_sinkhorn_reward": 0.7412925362586975,
      "rewards/forensics_strict_edge_reward": 0.12099595740437508,
      "rewards/forensics_tiou_reward": 0.6243833005428314,
      "step": 1071
    },
    {
      "completion_length": 23.140625,
      "epoch": 3.2887864823348694,
      "grad_norm": 16.317689895629883,
      "kl": 0.775390625,
      "learning_rate": 1.7615384615384615e-07,
      "loss": -0.113,
      "reward": 0.7948192358016968,
      "reward_std": 0.014201275538653135,
      "reward_std_all": 0.014929046388715506,
      "rewards-max/forensics_sinkhorn_reward": 0.9823734760284424,
      "rewards-max/forensics_strict_edge_reward": 0.7309097647666931,
      "rewards-max/forensics_tiou_reward": 0.9820455014705658,
      "rewards/forensics_sinkhorn_reward": 0.7881660163402557,
      "rewards/forensics_strict_edge_reward": 0.0838947482407093,
      "rewards/forensics_tiou_reward": 0.6044419705867767,
      "step": 1072
    },
    {
      "completion_length": 26.3125,
      "epoch": 3.291858678955453,
      "grad_norm": 12.452414512634277,
      "kl": 0.515625,
      "learning_rate": 1.7538461538461539e-07,
      "loss": -0.1195,
      "reward": 0.7993848621845245,
      "reward_std": 0.01367725757881999,
      "reward_std_all": 0.013264550361782312,
      "rewards-max/forensics_sinkhorn_reward": 0.9859179854393005,
      "rewards-max/forensics_strict_edge_reward": 0.8582656681537628,
      "rewards-max/forensics_tiou_reward": 0.9914781153202057,
      "rewards/forensics_sinkhorn_reward": 0.8371002674102783,
      "rewards/forensics_strict_edge_reward": 0.11586900800466537,
      "rewards/forensics_tiou_reward": 0.7137022912502289,
      "step": 1073
    },
    {
      "completion_length": 28.375,
      "epoch": 3.294930875576037,
      "grad_norm": 9.135226249694824,
      "kl": 0.7763671875,
      "learning_rate": 1.746153846153846e-07,
      "loss": 0.0449,
      "reward": 0.7665435969829559,
      "reward_std": 0.025744669139385223,
      "reward_std_all": 0.033487796783447266,
      "rewards-max/forensics_sinkhorn_reward": 0.9819221496582031,
      "rewards-max/forensics_strict_edge_reward": 0.51341712474823,
      "rewards-max/forensics_tiou_reward": 0.9821005463600159,
      "rewards/forensics_sinkhorn_reward": 0.7731020748615265,
      "rewards/forensics_strict_edge_reward": 0.09578219056129456,
      "rewards/forensics_tiou_reward": 0.624138355255127,
      "step": 1074
    },
    {
      "completion_length": 32.515625,
      "epoch": 3.2980030721966207,
      "grad_norm": 10.919771194458008,
      "kl": 0.693359375,
      "learning_rate": 1.7384615384615384e-07,
      "loss": 0.2286,
      "reward": 0.7304656803607941,
      "reward_std": 0.0363646037876606,
      "reward_std_all": 0.05526041239500046,
      "rewards-max/forensics_sinkhorn_reward": 0.9862320721149445,
      "rewards-max/forensics_strict_edge_reward": 0.9375866651535034,
      "rewards-max/forensics_tiou_reward": 0.9935760796070099,
      "rewards/forensics_sinkhorn_reward": 0.6172186732292175,
      "rewards/forensics_strict_edge_reward": 0.07354104705154896,
      "rewards/forensics_tiou_reward": 0.4382372200489044,
      "step": 1075
    },
    {
      "completion_length": 22.28125,
      "epoch": 3.3010752688172045,
      "grad_norm": 9.733531951904297,
      "kl": 0.48046875,
      "learning_rate": 1.7307692307692305e-07,
      "loss": -0.0309,
      "reward": 0.7678938508033752,
      "reward_std": 0.031549818348139524,
      "reward_std_all": 0.03794613014906645,
      "rewards-max/forensics_sinkhorn_reward": 0.9860002398490906,
      "rewards-max/forensics_strict_edge_reward": 0.8665722012519836,
      "rewards-max/forensics_tiou_reward": 0.9920806288719177,
      "rewards/forensics_sinkhorn_reward": 0.7662108838558197,
      "rewards/forensics_strict_edge_reward": 0.12466364726424217,
      "rewards/forensics_tiou_reward": 0.5931687355041504,
      "step": 1076
    },
    {
      "completion_length": 23.390625,
      "epoch": 3.3041474654377883,
      "grad_norm": 10.170021057128906,
      "kl": 0.677734375,
      "learning_rate": 1.723076923076923e-07,
      "loss": -0.2378,
      "reward": 0.7761886417865753,
      "reward_std": 0.037225114181637764,
      "reward_std_all": 0.03590252809226513,
      "rewards-max/forensics_sinkhorn_reward": 0.9856425225734711,
      "rewards-max/forensics_strict_edge_reward": 0.8106843829154968,
      "rewards-max/forensics_tiou_reward": 0.990417867898941,
      "rewards/forensics_sinkhorn_reward": 0.7820412516593933,
      "rewards/forensics_strict_edge_reward": 0.13933298736810684,
      "rewards/forensics_tiou_reward": 0.6089979708194733,
      "step": 1077
    },
    {
      "completion_length": 23.890625,
      "epoch": 3.3072196620583716,
      "grad_norm": 11.516094207763672,
      "kl": 0.5625,
      "learning_rate": 1.7153846153846153e-07,
      "loss": 0.0681,
      "reward": 0.7485333681106567,
      "reward_std": 0.03669337788596749,
      "reward_std_all": 0.045753683894872665,
      "rewards-max/forensics_sinkhorn_reward": 0.9864047467708588,
      "rewards-max/forensics_strict_edge_reward": 0.6065583229064941,
      "rewards-max/forensics_tiou_reward": 0.9927198886871338,
      "rewards/forensics_sinkhorn_reward": 0.7245005965232849,
      "rewards/forensics_strict_edge_reward": 0.11075056344270706,
      "rewards/forensics_tiou_reward": 0.5410221964120865,
      "step": 1078
    },
    {
      "completion_length": 27.859375,
      "epoch": 3.3102918586789554,
      "grad_norm": 8.556251525878906,
      "kl": 0.5029296875,
      "learning_rate": 1.7076923076923075e-07,
      "loss": 0.1123,
      "reward": 0.7837764918804169,
      "reward_std": 0.03062605205923319,
      "reward_std_all": 0.029425055719912052,
      "rewards-max/forensics_sinkhorn_reward": 0.9834748804569244,
      "rewards-max/forensics_strict_edge_reward": 0.47831887006759644,
      "rewards-max/forensics_tiou_reward": 0.9845200181007385,
      "rewards/forensics_sinkhorn_reward": 0.7573001682758331,
      "rewards/forensics_strict_edge_reward": 0.10185112431645393,
      "rewards/forensics_tiou_reward": 0.640966534614563,
      "step": 1079
    },
    {
      "completion_length": 22.28125,
      "epoch": 3.313364055299539,
      "grad_norm": 9.777639389038086,
      "kl": 0.392578125,
      "learning_rate": 1.7000000000000001e-07,
      "loss": 0.1165,
      "reward": 0.7592495381832123,
      "reward_std": 0.03660164214670658,
      "reward_std_all": 0.04135235399007797,
      "rewards-max/forensics_sinkhorn_reward": 0.9871358573436737,
      "rewards-max/forensics_strict_edge_reward": 0.6515887677669525,
      "rewards-max/forensics_tiou_reward": 0.9948266744613647,
      "rewards/forensics_sinkhorn_reward": 0.6697844564914703,
      "rewards/forensics_strict_edge_reward": 0.04609784297645092,
      "rewards/forensics_tiou_reward": 0.4551985561847687,
      "step": 1080
    },
    {
      "completion_length": 27.203125,
      "epoch": 3.316436251920123,
      "grad_norm": 10.295089721679688,
      "kl": 0.607421875,
      "learning_rate": 1.6923076923076923e-07,
      "loss": 0.1472,
      "reward": 0.7336703836917877,
      "reward_std": 0.05082080978900194,
      "reward_std_all": 0.060030654072761536,
      "rewards-max/forensics_sinkhorn_reward": 0.9805356860160828,
      "rewards-max/forensics_strict_edge_reward": 0.6347748339176178,
      "rewards-max/forensics_tiou_reward": 0.9798469245433807,
      "rewards/forensics_sinkhorn_reward": 0.6842643916606903,
      "rewards/forensics_strict_edge_reward": 0.06141463667154312,
      "rewards/forensics_tiou_reward": 0.5320715606212616,
      "step": 1081
    },
    {
      "completion_length": 25.125,
      "epoch": 3.3195084485407067,
      "grad_norm": 9.160011291503906,
      "kl": 0.458984375,
      "learning_rate": 1.6846153846153847e-07,
      "loss": 0.0989,
      "reward": 0.7482636272907257,
      "reward_std": 0.027623046189546585,
      "reward_std_all": 0.04582262318581343,
      "rewards-max/forensics_sinkhorn_reward": 0.9837625920772552,
      "rewards-max/forensics_strict_edge_reward": 0.6630095839500427,
      "rewards-max/forensics_tiou_reward": 0.9860388040542603,
      "rewards/forensics_sinkhorn_reward": 0.7332856059074402,
      "rewards/forensics_strict_edge_reward": 0.09872668236494064,
      "rewards/forensics_tiou_reward": 0.5715813040733337,
      "step": 1082
    },
    {
      "completion_length": 22.765625,
      "epoch": 3.3225806451612905,
      "grad_norm": 15.20024299621582,
      "kl": 0.63671875,
      "learning_rate": 1.6769230769230768e-07,
      "loss": 0.1139,
      "reward": 0.721706360578537,
      "reward_std": 0.015899801859632134,
      "reward_std_all": 0.047467746771872044,
      "rewards-max/forensics_sinkhorn_reward": 0.9773591458797455,
      "rewards-max/forensics_strict_edge_reward": 0.4251096248626709,
      "rewards-max/forensics_tiou_reward": 0.9625926613807678,
      "rewards/forensics_sinkhorn_reward": 0.6581538319587708,
      "rewards/forensics_strict_edge_reward": 0.07181346137076616,
      "rewards/forensics_tiou_reward": 0.49404093623161316,
      "step": 1083
    },
    {
      "completion_length": 27.40625,
      "epoch": 3.3256528417818743,
      "grad_norm": 13.865413665771484,
      "kl": 0.5107421875,
      "learning_rate": 1.6692307692307692e-07,
      "loss": 0.0228,
      "reward": 0.7662186622619629,
      "reward_std": 0.017133545828983188,
      "reward_std_all": 0.029926995746791363,
      "rewards-max/forensics_sinkhorn_reward": 0.9758972525596619,
      "rewards-max/forensics_strict_edge_reward": 0.5461536049842834,
      "rewards-max/forensics_tiou_reward": 0.9664151072502136,
      "rewards/forensics_sinkhorn_reward": 0.7271126806735992,
      "rewards/forensics_strict_edge_reward": 0.07262205146253109,
      "rewards/forensics_tiou_reward": 0.531660258769989,
      "step": 1084
    },
    {
      "completion_length": 25.578125,
      "epoch": 3.3287250384024576,
      "grad_norm": 10.832976341247559,
      "kl": 0.5029296875,
      "learning_rate": 1.6615384615384613e-07,
      "loss": 0.1894,
      "reward": 0.7707688212394714,
      "reward_std": 0.022607480641454458,
      "reward_std_all": 0.029822029173374176,
      "rewards-max/forensics_sinkhorn_reward": 0.9834959506988525,
      "rewards-max/forensics_strict_edge_reward": 0.5736939311027527,
      "rewards-max/forensics_tiou_reward": 0.9830847382545471,
      "rewards/forensics_sinkhorn_reward": 0.7099185585975647,
      "rewards/forensics_strict_edge_reward": 0.09355468302965164,
      "rewards/forensics_tiou_reward": 0.5523377507925034,
      "step": 1085
    },
    {
      "completion_length": 28.078125,
      "epoch": 3.3317972350230414,
      "grad_norm": 9.785888671875,
      "kl": 0.6484375,
      "learning_rate": 1.6538461538461538e-07,
      "loss": 0.068,
      "reward": 0.7809228301048279,
      "reward_std": 0.031566014513373375,
      "reward_std_all": 0.030910979956388474,
      "rewards-max/forensics_sinkhorn_reward": 0.9769639074802399,
      "rewards-max/forensics_strict_edge_reward": 0.4178338199853897,
      "rewards-max/forensics_tiou_reward": 0.9680930078029633,
      "rewards/forensics_sinkhorn_reward": 0.7375666499137878,
      "rewards/forensics_strict_edge_reward": 0.044136494398117065,
      "rewards/forensics_tiou_reward": 0.6001645624637604,
      "step": 1086
    },
    {
      "completion_length": 27.984375,
      "epoch": 3.334869431643625,
      "grad_norm": 12.404760360717773,
      "kl": 1.037109375,
      "learning_rate": 1.646153846153846e-07,
      "loss": -0.1706,
      "reward": 0.7528164684772491,
      "reward_std": 0.032280596904456615,
      "reward_std_all": 0.04405881464481354,
      "rewards-max/forensics_sinkhorn_reward": 0.975994884967804,
      "rewards-max/forensics_strict_edge_reward": 0.5971094369888306,
      "rewards-max/forensics_tiou_reward": 0.9656887948513031,
      "rewards/forensics_sinkhorn_reward": 0.6933371424674988,
      "rewards/forensics_strict_edge_reward": 0.08905592933297157,
      "rewards/forensics_tiou_reward": 0.5185451507568359,
      "step": 1087
    },
    {
      "completion_length": 24.078125,
      "epoch": 3.337941628264209,
      "grad_norm": 10.178709983825684,
      "kl": 0.505859375,
      "learning_rate": 1.6384615384615383e-07,
      "loss": -0.0486,
      "reward": 0.7874091863632202,
      "reward_std": 0.017618926358409226,
      "reward_std_all": 0.01916706096380949,
      "rewards-max/forensics_sinkhorn_reward": 0.9746232628822327,
      "rewards-max/forensics_strict_edge_reward": 0.8333333432674408,
      "rewards-max/forensics_tiou_reward": 0.9623560011386871,
      "rewards/forensics_sinkhorn_reward": 0.7439125776290894,
      "rewards/forensics_strict_edge_reward": 0.1294129677116871,
      "rewards/forensics_tiou_reward": 0.5905574560165405,
      "step": 1088
    },
    {
      "completion_length": 22.859375,
      "epoch": 3.3410138248847927,
      "grad_norm": 9.930045127868652,
      "kl": 0.4208984375,
      "learning_rate": 1.6307692307692307e-07,
      "loss": -0.247,
      "reward": 0.8045015335083008,
      "reward_std": 0.00411355122923851,
      "reward_std_all": 0.0052896731067448854,
      "rewards-max/forensics_sinkhorn_reward": 0.9850694239139557,
      "rewards-max/forensics_strict_edge_reward": 0.8351600170135498,
      "rewards-max/forensics_tiou_reward": 0.9897692501544952,
      "rewards/forensics_sinkhorn_reward": 0.841732919216156,
      "rewards/forensics_strict_edge_reward": 0.1326792761683464,
      "rewards/forensics_tiou_reward": 0.7042504847049713,
      "step": 1089
    },
    {
      "completion_length": 27.234375,
      "epoch": 3.3440860215053765,
      "grad_norm": 45.25657653808594,
      "kl": 3.01171875,
      "learning_rate": 1.623076923076923e-07,
      "loss": -0.0245,
      "reward": 0.8008145391941071,
      "reward_std": 0.010598967201076448,
      "reward_std_all": 0.01099677081219852,
      "rewards-max/forensics_sinkhorn_reward": 0.9806921482086182,
      "rewards-max/forensics_strict_edge_reward": 0.6624195873737335,
      "rewards-max/forensics_tiou_reward": 0.9773626029491425,
      "rewards/forensics_sinkhorn_reward": 0.8370360434055328,
      "rewards/forensics_strict_edge_reward": 0.11397384107112885,
      "rewards/forensics_tiou_reward": 0.698711484670639,
      "step": 1090
    },
    {
      "completion_length": 23.546875,
      "epoch": 3.3471582181259603,
      "grad_norm": 12.996745109558105,
      "kl": 0.474609375,
      "learning_rate": 1.6153846153846155e-07,
      "loss": -0.1656,
      "reward": 0.7975975871086121,
      "reward_std": 0.011771796271204948,
      "reward_std_all": 0.012411344330757856,
      "rewards-max/forensics_sinkhorn_reward": 0.9872177541255951,
      "rewards-max/forensics_strict_edge_reward": 0.8929734230041504,
      "rewards-max/forensics_tiou_reward": 0.9949747920036316,
      "rewards/forensics_sinkhorn_reward": 0.8181912899017334,
      "rewards/forensics_strict_edge_reward": 0.12772653996944427,
      "rewards/forensics_tiou_reward": 0.669402003288269,
      "step": 1091
    },
    {
      "completion_length": 22.15625,
      "epoch": 3.3502304147465436,
      "grad_norm": 16.53929328918457,
      "kl": 0.5068359375,
      "learning_rate": 1.6076923076923076e-07,
      "loss": -0.0029,
      "reward": 0.7872734367847443,
      "reward_std": 0.014547912403941154,
      "reward_std_all": 0.01799020217731595,
      "rewards-max/forensics_sinkhorn_reward": 0.9876190423965454,
      "rewards-max/forensics_strict_edge_reward": 0.952418714761734,
      "rewards-max/forensics_tiou_reward": 0.9965516030788422,
      "rewards/forensics_sinkhorn_reward": 0.7613219320774078,
      "rewards/forensics_strict_edge_reward": 0.18094376474618912,
      "rewards/forensics_tiou_reward": 0.6084256768226624,
      "step": 1092
    },
    {
      "completion_length": 26.953125,
      "epoch": 3.3533026113671274,
      "grad_norm": 16.96415138244629,
      "kl": 0.60546875,
      "learning_rate": 1.6e-07,
      "loss": -0.0317,
      "reward": 0.7731771469116211,
      "reward_std": 0.047785963863134384,
      "reward_std_all": 0.044125666841864586,
      "rewards-max/forensics_sinkhorn_reward": 0.9772058427333832,
      "rewards-max/forensics_strict_edge_reward": 0.6824378371238708,
      "rewards-max/forensics_tiou_reward": 0.9718417227268219,
      "rewards/forensics_sinkhorn_reward": 0.7495801448822021,
      "rewards/forensics_strict_edge_reward": 0.11165882647037506,
      "rewards/forensics_tiou_reward": 0.5766474008560181,
      "step": 1093
    },
    {
      "completion_length": 23.171875,
      "epoch": 3.356374807987711,
      "grad_norm": 9.543683052062988,
      "kl": 0.7578125,
      "learning_rate": 1.5923076923076922e-07,
      "loss": -0.0238,
      "reward": 0.7915195822715759,
      "reward_std": 0.02426466252654791,
      "reward_std_all": 0.023116091266274452,
      "rewards-max/forensics_sinkhorn_reward": 0.9865463376045227,
      "rewards-max/forensics_strict_edge_reward": 0.7245311737060547,
      "rewards-max/forensics_tiou_reward": 0.994684636592865,
      "rewards/forensics_sinkhorn_reward": 0.7291793525218964,
      "rewards/forensics_strict_edge_reward": 0.10814491286873817,
      "rewards/forensics_tiou_reward": 0.5503436028957367,
      "step": 1094
    },
    {
      "completion_length": 24.046875,
      "epoch": 3.359447004608295,
      "grad_norm": 16.7696533203125,
      "kl": 0.4697265625,
      "learning_rate": 1.5846153846153846e-07,
      "loss": 0.1775,
      "reward": 0.7331258654594421,
      "reward_std": 0.08092045411467552,
      "reward_std_all": 0.07910961657762527,
      "rewards-max/forensics_sinkhorn_reward": 0.9869626462459564,
      "rewards-max/forensics_strict_edge_reward": 0.5559406280517578,
      "rewards-max/forensics_tiou_reward": 0.9944508075714111,
      "rewards/forensics_sinkhorn_reward": 0.6956603825092316,
      "rewards/forensics_strict_edge_reward": 0.04742487985640764,
      "rewards/forensics_tiou_reward": 0.5166738033294678,
      "step": 1095
    },
    {
      "completion_length": 26.4375,
      "epoch": 3.3625192012288787,
      "grad_norm": 13.862590789794922,
      "kl": 0.84765625,
      "learning_rate": 1.5769230769230767e-07,
      "loss": 0.0439,
      "reward": 0.7861178517341614,
      "reward_std": 0.025511500192806125,
      "reward_std_all": 0.025519843213260174,
      "rewards-max/forensics_sinkhorn_reward": 0.9842175543308258,
      "rewards-max/forensics_strict_edge_reward": 0.6796615123748779,
      "rewards-max/forensics_tiou_reward": 0.9865174293518066,
      "rewards/forensics_sinkhorn_reward": 0.6857545971870422,
      "rewards/forensics_strict_edge_reward": 0.08353948965668678,
      "rewards/forensics_tiou_reward": 0.5415786355733871,
      "step": 1096
    },
    {
      "completion_length": 22.796875,
      "epoch": 3.3655913978494625,
      "grad_norm": 11.426008224487305,
      "kl": 0.70703125,
      "learning_rate": 1.569230769230769e-07,
      "loss": 0.1573,
      "reward": 0.7757920324802399,
      "reward_std": 0.04740282706916332,
      "reward_std_all": 0.04307330586016178,
      "rewards-max/forensics_sinkhorn_reward": 0.9859515130519867,
      "rewards-max/forensics_strict_edge_reward": 0.7421593070030212,
      "rewards-max/forensics_tiou_reward": 0.9911015033721924,
      "rewards/forensics_sinkhorn_reward": 0.7747114300727844,
      "rewards/forensics_strict_edge_reward": 0.09548420831561089,
      "rewards/forensics_tiou_reward": 0.6269512474536896,
      "step": 1097
    },
    {
      "completion_length": 27.140625,
      "epoch": 3.3686635944700463,
      "grad_norm": 15.829987525939941,
      "kl": 0.533203125,
      "learning_rate": 1.5615384615384612e-07,
      "loss": -0.2567,
      "reward": 0.771879106760025,
      "reward_std": 0.004488682490773499,
      "reward_std_all": 0.02175048738718033,
      "rewards-max/forensics_sinkhorn_reward": 0.9821268022060394,
      "rewards-max/forensics_strict_edge_reward": 0.756708562374115,
      "rewards-max/forensics_tiou_reward": 0.9830785393714905,
      "rewards/forensics_sinkhorn_reward": 0.7696613371372223,
      "rewards/forensics_strict_edge_reward": 0.10985636711120605,
      "rewards/forensics_tiou_reward": 0.6174851953983307,
      "step": 1098
    },
    {
      "completion_length": 21.453125,
      "epoch": 3.3717357910906296,
      "grad_norm": 18.272716522216797,
      "kl": 0.4306640625,
      "learning_rate": 1.5538461538461536e-07,
      "loss": -0.2847,
      "reward": 0.8006677627563477,
      "reward_std": 0.0036312862066552043,
      "reward_std_all": 0.006302142981439829,
      "rewards-max/forensics_sinkhorn_reward": 0.9849904775619507,
      "rewards-max/forensics_strict_edge_reward": 0.6236754059791565,
      "rewards-max/forensics_tiou_reward": 0.9895915985107422,
      "rewards/forensics_sinkhorn_reward": 0.8202431499958038,
      "rewards/forensics_strict_edge_reward": 0.11287469416856766,
      "rewards/forensics_tiou_reward": 0.6613914370536804,
      "step": 1099
    },
    {
      "completion_length": 24.6875,
      "epoch": 3.3748079877112134,
      "grad_norm": 10.499566078186035,
      "kl": 0.6337890625,
      "learning_rate": 1.5461538461538463e-07,
      "loss": 0.306,
      "reward": 0.7384886145591736,
      "reward_std": 0.05386342108249664,
      "reward_std_all": 0.061044614762067795,
      "rewards-max/forensics_sinkhorn_reward": 0.9873039126396179,
      "rewards-max/forensics_strict_edge_reward": 0.7958523333072662,
      "rewards-max/forensics_tiou_reward": 0.9951704740524292,
      "rewards/forensics_sinkhorn_reward": 0.6519462168216705,
      "rewards/forensics_strict_edge_reward": 0.0742887370288372,
      "rewards/forensics_tiou_reward": 0.5262164175510406,
      "step": 1100
    },
    {
      "completion_length": 22.71875,
      "epoch": 3.377880184331797,
      "grad_norm": 16.174213409423828,
      "kl": 1.0224609375,
      "learning_rate": 1.5384615384615385e-07,
      "loss": 0.1871,
      "reward": 0.7759190499782562,
      "reward_std": 0.034140653908252716,
      "reward_std_all": 0.03637901972979307,
      "rewards-max/forensics_sinkhorn_reward": 0.9829724133014679,
      "rewards-max/forensics_strict_edge_reward": 0.45034897327423096,
      "rewards-max/forensics_tiou_reward": 0.9844627380371094,
      "rewards/forensics_sinkhorn_reward": 0.7357783019542694,
      "rewards/forensics_strict_edge_reward": 0.050291525200009346,
      "rewards/forensics_tiou_reward": 0.5422673225402832,
      "step": 1101
    },
    {
      "completion_length": 23.6875,
      "epoch": 3.380952380952381,
      "grad_norm": 20.3800048828125,
      "kl": 0.962890625,
      "learning_rate": 1.5307692307692309e-07,
      "loss": 0.0097,
      "reward": 0.7661321461200714,
      "reward_std": 0.0482458733022213,
      "reward_std_all": 0.04827482812106609,
      "rewards-max/forensics_sinkhorn_reward": 0.9780493080615997,
      "rewards-max/forensics_strict_edge_reward": 0.6741649508476257,
      "rewards-max/forensics_tiou_reward": 0.9725854098796844,
      "rewards/forensics_sinkhorn_reward": 0.7254603803157806,
      "rewards/forensics_strict_edge_reward": 0.06399309635162354,
      "rewards/forensics_tiou_reward": 0.5054240822792053,
      "step": 1102
    },
    {
      "completion_length": 25.6875,
      "epoch": 3.3840245775729647,
      "grad_norm": 11.165096282958984,
      "kl": 0.7734375,
      "learning_rate": 1.523076923076923e-07,
      "loss": 0.0185,
      "reward": 0.7868910729885101,
      "reward_std": 0.026611122069880366,
      "reward_std_all": 0.025829343823716044,
      "rewards-max/forensics_sinkhorn_reward": 0.9774608016014099,
      "rewards-max/forensics_strict_edge_reward": 0.5950735807418823,
      "rewards-max/forensics_tiou_reward": 0.9692734777927399,
      "rewards/forensics_sinkhorn_reward": 0.8122721910476685,
      "rewards/forensics_strict_edge_reward": 0.11035483330488205,
      "rewards/forensics_tiou_reward": 0.6406576633453369,
      "step": 1103
    },
    {
      "completion_length": 25.34375,
      "epoch": 3.3870967741935485,
      "grad_norm": 12.26934814453125,
      "kl": 0.5048828125,
      "learning_rate": 1.5153846153846154e-07,
      "loss": -0.0988,
      "reward": 0.7626230716705322,
      "reward_std": 0.029976502060890198,
      "reward_std_all": 0.036584507673978806,
      "rewards-max/forensics_sinkhorn_reward": 0.9820683598518372,
      "rewards-max/forensics_strict_edge_reward": 0.44064828753471375,
      "rewards-max/forensics_tiou_reward": 0.9784791767597198,
      "rewards/forensics_sinkhorn_reward": 0.766107976436615,
      "rewards/forensics_strict_edge_reward": 0.06923186779022217,
      "rewards/forensics_tiou_reward": 0.6149955987930298,
      "step": 1104
    },
    {
      "completion_length": 26.25,
      "epoch": 3.3901689708141323,
      "grad_norm": 9.597302436828613,
      "kl": 0.71875,
      "learning_rate": 1.5076923076923075e-07,
      "loss": -0.1256,
      "reward": 0.7909007668495178,
      "reward_std": 0.026822420302778482,
      "reward_std_all": 0.024828876368701458,
      "rewards-max/forensics_sinkhorn_reward": 0.9826691448688507,
      "rewards-max/forensics_strict_edge_reward": 0.5884901583194733,
      "rewards-max/forensics_tiou_reward": 0.9827454686164856,
      "rewards/forensics_sinkhorn_reward": 0.7645479142665863,
      "rewards/forensics_strict_edge_reward": 0.10181118547916412,
      "rewards/forensics_tiou_reward": 0.611009031534195,
      "step": 1105
    },
    {
      "completion_length": 27.140625,
      "epoch": 3.3932411674347156,
      "grad_norm": 8.535406112670898,
      "kl": 0.5078125,
      "learning_rate": 1.5e-07,
      "loss": -0.0838,
      "reward": 0.7679960429668427,
      "reward_std": 0.039168523624539375,
      "reward_std_all": 0.04232237488031387,
      "rewards-max/forensics_sinkhorn_reward": 0.9814222753047943,
      "rewards-max/forensics_strict_edge_reward": 0.6718102097511292,
      "rewards-max/forensics_tiou_reward": 0.9765020906925201,
      "rewards/forensics_sinkhorn_reward": 0.7685615718364716,
      "rewards/forensics_strict_edge_reward": 0.11629428341984749,
      "rewards/forensics_tiou_reward": 0.6051779389381409,
      "step": 1106
    },
    {
      "completion_length": 21.453125,
      "epoch": 3.3963133640552994,
      "grad_norm": 14.449507713317871,
      "kl": 0.4580078125,
      "learning_rate": 1.492307692307692e-07,
      "loss": -0.1313,
      "reward": 0.7737137973308563,
      "reward_std": 0.01548861083574593,
      "reward_std_all": 0.026293043745681643,
      "rewards-max/forensics_sinkhorn_reward": 0.9843608736991882,
      "rewards-max/forensics_strict_edge_reward": 0.7047859728336334,
      "rewards-max/forensics_tiou_reward": 0.9859501421451569,
      "rewards/forensics_sinkhorn_reward": 0.7794548571109772,
      "rewards/forensics_strict_edge_reward": 0.12620864063501358,
      "rewards/forensics_tiou_reward": 0.6499277055263519,
      "step": 1107
    },
    {
      "completion_length": 24.21875,
      "epoch": 3.399385560675883,
      "grad_norm": 9.692239761352539,
      "kl": 0.603515625,
      "learning_rate": 1.4846153846153845e-07,
      "loss": 0.1256,
      "reward": 0.777317613363266,
      "reward_std": 0.026513610035181046,
      "reward_std_all": 0.028838797472417355,
      "rewards-max/forensics_sinkhorn_reward": 0.9765392243862152,
      "rewards-max/forensics_strict_edge_reward": 0.661531001329422,
      "rewards-max/forensics_tiou_reward": 0.9700180292129517,
      "rewards/forensics_sinkhorn_reward": 0.7153031826019287,
      "rewards/forensics_strict_edge_reward": 0.14851657301187515,
      "rewards/forensics_tiou_reward": 0.592355340719223,
      "step": 1108
    },
    {
      "completion_length": 27.390625,
      "epoch": 3.402457757296467,
      "grad_norm": 10.11516284942627,
      "kl": 0.7421875,
      "learning_rate": 1.476923076923077e-07,
      "loss": -0.0435,
      "reward": 0.7996823787689209,
      "reward_std": 0.007941690273582935,
      "reward_std_all": 0.009163685026578605,
      "rewards-max/forensics_sinkhorn_reward": 0.9815197885036469,
      "rewards-max/forensics_strict_edge_reward": 0.7055691182613373,
      "rewards-max/forensics_tiou_reward": 0.9787670075893402,
      "rewards/forensics_sinkhorn_reward": 0.7693357765674591,
      "rewards/forensics_strict_edge_reward": 0.08918029069900513,
      "rewards/forensics_tiou_reward": 0.652391642332077,
      "step": 1109
    },
    {
      "completion_length": 22.96875,
      "epoch": 3.4055299539170507,
      "grad_norm": 12.641489028930664,
      "kl": 0.744140625,
      "learning_rate": 1.4692307692307693e-07,
      "loss": 0.0549,
      "reward": 0.7907687425613403,
      "reward_std": 0.010720501188188791,
      "reward_std_all": 0.014498825185000896,
      "rewards-max/forensics_sinkhorn_reward": 0.9775028228759766,
      "rewards-max/forensics_strict_edge_reward": 0.5422053933143616,
      "rewards-max/forensics_tiou_reward": 0.9705535471439362,
      "rewards/forensics_sinkhorn_reward": 0.7032440304756165,
      "rewards/forensics_strict_edge_reward": 0.07497815042734146,
      "rewards/forensics_tiou_reward": 0.5184545814990997,
      "step": 1110
    },
    {
      "completion_length": 27.40625,
      "epoch": 3.4086021505376345,
      "grad_norm": 10.020000457763672,
      "kl": 0.556640625,
      "learning_rate": 1.4615384615384617e-07,
      "loss": 0.0303,
      "reward": 0.7636238932609558,
      "reward_std": 0.038003044202923775,
      "reward_std_all": 0.04158158227801323,
      "rewards-max/forensics_sinkhorn_reward": 0.9837937355041504,
      "rewards-max/forensics_strict_edge_reward": 0.7169294059276581,
      "rewards-max/forensics_tiou_reward": 0.9861735999584198,
      "rewards/forensics_sinkhorn_reward": 0.7204884588718414,
      "rewards/forensics_strict_edge_reward": 0.10676733776926994,
      "rewards/forensics_tiou_reward": 0.5760080367326736,
      "step": 1111
    },
    {
      "completion_length": 23.640625,
      "epoch": 3.4116743471582183,
      "grad_norm": 13.673441886901855,
      "kl": 0.607421875,
      "learning_rate": 1.4538461538461538e-07,
      "loss": -0.0361,
      "reward": 0.7795968949794769,
      "reward_std": 0.04713106341660023,
      "reward_std_all": 0.04248024523258209,
      "rewards-max/forensics_sinkhorn_reward": 0.9783943891525269,
      "rewards-max/forensics_strict_edge_reward": 0.6515887677669525,
      "rewards-max/forensics_tiou_reward": 0.9716787338256836,
      "rewards/forensics_sinkhorn_reward": 0.8031033277511597,
      "rewards/forensics_strict_edge_reward": 0.09744544699788094,
      "rewards/forensics_tiou_reward": 0.6488824784755707,
      "step": 1112
    },
    {
      "completion_length": 23.953125,
      "epoch": 3.4147465437788016,
      "grad_norm": 10.511210441589355,
      "kl": 0.5693359375,
      "learning_rate": 1.4461538461538462e-07,
      "loss": -0.2145,
      "reward": 0.7728733718395233,
      "reward_std": 0.00648858561180532,
      "reward_std_all": 0.022252596681937575,
      "rewards-max/forensics_sinkhorn_reward": 0.9863129556179047,
      "rewards-max/forensics_strict_edge_reward": 0.8756595849990845,
      "rewards-max/forensics_tiou_reward": 0.9930212795734406,
      "rewards/forensics_sinkhorn_reward": 0.8071793615818024,
      "rewards/forensics_strict_edge_reward": 0.1481473743915558,
      "rewards/forensics_tiou_reward": 0.657272458076477,
      "step": 1113
    },
    {
      "completion_length": 27.03125,
      "epoch": 3.4178187403993854,
      "grad_norm": 9.333110809326172,
      "kl": 0.48828125,
      "learning_rate": 1.4384615384615384e-07,
      "loss": 0.1003,
      "reward": 0.7729111611843109,
      "reward_std": 0.03138504782691598,
      "reward_std_all": 0.033249299973249435,
      "rewards-max/forensics_sinkhorn_reward": 0.9859624207019806,
      "rewards-max/forensics_strict_edge_reward": 0.7151772081851959,
      "rewards-max/forensics_tiou_reward": 0.9889704585075378,
      "rewards/forensics_sinkhorn_reward": 0.7398335635662079,
      "rewards/forensics_strict_edge_reward": 0.07528645917773247,
      "rewards/forensics_tiou_reward": 0.6036023199558258,
      "step": 1114
    },
    {
      "completion_length": 26.796875,
      "epoch": 3.420890937019969,
      "grad_norm": 11.741369247436523,
      "kl": 0.64453125,
      "learning_rate": 1.4307692307692308e-07,
      "loss": -0.2154,
      "reward": 0.7965734302997589,
      "reward_std": 0.01093835721258074,
      "reward_std_all": 0.012078117346391082,
      "rewards-max/forensics_sinkhorn_reward": 0.9762517809867859,
      "rewards-max/forensics_strict_edge_reward": 0.5515579879283905,
      "rewards-max/forensics_tiou_reward": 0.9695217311382294,
      "rewards/forensics_sinkhorn_reward": 0.7919971942901611,
      "rewards/forensics_strict_edge_reward": 0.09988927096128464,
      "rewards/forensics_tiou_reward": 0.6333451271057129,
      "step": 1115
    },
    {
      "completion_length": 25.25,
      "epoch": 3.423963133640553,
      "grad_norm": 9.902020454406738,
      "kl": 0.484375,
      "learning_rate": 1.423076923076923e-07,
      "loss": 0.258,
      "reward": 0.7189384400844574,
      "reward_std": 0.08412039466202259,
      "reward_std_all": 0.08608266338706017,
      "rewards-max/forensics_sinkhorn_reward": 0.9781709909439087,
      "rewards-max/forensics_strict_edge_reward": 0.49686121940612793,
      "rewards-max/forensics_tiou_reward": 0.9705096185207367,
      "rewards/forensics_sinkhorn_reward": 0.6413498520851135,
      "rewards/forensics_strict_edge_reward": 0.07907629013061523,
      "rewards/forensics_tiou_reward": 0.49419426918029785,
      "step": 1116
    },
    {
      "completion_length": 25.734375,
      "epoch": 3.4270353302611367,
      "grad_norm": 9.17811393737793,
      "kl": 0.5048828125,
      "learning_rate": 1.4153846153846153e-07,
      "loss": 0.2729,
      "reward": 0.7112028896808624,
      "reward_std": 0.061111653223633766,
      "reward_std_all": 0.07943081855773926,
      "rewards-max/forensics_sinkhorn_reward": 0.9816422164440155,
      "rewards-max/forensics_strict_edge_reward": 0.6524136364459991,
      "rewards-max/forensics_tiou_reward": 0.9807101190090179,
      "rewards/forensics_sinkhorn_reward": 0.6817103922367096,
      "rewards/forensics_strict_edge_reward": 0.10663189738988876,
      "rewards/forensics_tiou_reward": 0.5272395759820938,
      "step": 1117
    },
    {
      "completion_length": 26.859375,
      "epoch": 3.4301075268817205,
      "grad_norm": 12.476425170898438,
      "kl": 0.548828125,
      "learning_rate": 1.4076923076923074e-07,
      "loss": 0.0016,
      "reward": 0.7682672739028931,
      "reward_std": 0.029807339422404766,
      "reward_std_all": 0.03475385345518589,
      "rewards-max/forensics_sinkhorn_reward": 0.9790631830692291,
      "rewards-max/forensics_strict_edge_reward": 0.6281587183475494,
      "rewards-max/forensics_tiou_reward": 0.9713913798332214,
      "rewards/forensics_sinkhorn_reward": 0.7799567580223083,
      "rewards/forensics_strict_edge_reward": 0.1191057562828064,
      "rewards/forensics_tiou_reward": 0.6319440007209778,
      "step": 1118
    },
    {
      "completion_length": 22.328125,
      "epoch": 3.4331797235023043,
      "grad_norm": 13.272163391113281,
      "kl": 0.6025390625,
      "learning_rate": 1.4e-07,
      "loss": 0.0199,
      "reward": 0.75888991355896,
      "reward_std": 0.017461213283240795,
      "reward_std_all": 0.03201081603765488,
      "rewards-max/forensics_sinkhorn_reward": 0.979175865650177,
      "rewards-max/forensics_strict_edge_reward": 0.6505971103906631,
      "rewards-max/forensics_tiou_reward": 0.9720612466335297,
      "rewards/forensics_sinkhorn_reward": 0.74407559633255,
      "rewards/forensics_strict_edge_reward": 0.06121605634689331,
      "rewards/forensics_tiou_reward": 0.515836775302887,
      "step": 1119
    },
    {
      "completion_length": 24.515625,
      "epoch": 3.436251920122888,
      "grad_norm": 10.49084186553955,
      "kl": 0.5556640625,
      "learning_rate": 1.3923076923076922e-07,
      "loss": -0.0716,
      "reward": 0.7523966729640961,
      "reward_std": 0.041508326306939125,
      "reward_std_all": 0.049665821716189384,
      "rewards-max/forensics_sinkhorn_reward": 0.9869234561920166,
      "rewards-max/forensics_strict_edge_reward": 0.7943885624408722,
      "rewards-max/forensics_tiou_reward": 0.9931831657886505,
      "rewards/forensics_sinkhorn_reward": 0.6979115605354309,
      "rewards/forensics_strict_edge_reward": 0.10169760510325432,
      "rewards/forensics_tiou_reward": 0.5782403945922852,
      "step": 1120
    },
    {
      "completion_length": 25.578125,
      "epoch": 3.439324116743472,
      "grad_norm": 11.355331420898438,
      "kl": 0.5546875,
      "learning_rate": 1.3846153846153846e-07,
      "loss": 0.0525,
      "reward": 0.7441486120223999,
      "reward_std": 0.037213725969195366,
      "reward_std_all": 0.046979401260614395,
      "rewards-max/forensics_sinkhorn_reward": 0.9794796407222748,
      "rewards-max/forensics_strict_edge_reward": 0.655701607465744,
      "rewards-max/forensics_tiou_reward": 0.9742739498615265,
      "rewards/forensics_sinkhorn_reward": 0.7040881812572479,
      "rewards/forensics_strict_edge_reward": 0.09759238548576832,
      "rewards/forensics_tiou_reward": 0.5816775858402252,
      "step": 1121
    },
    {
      "completion_length": 21.5,
      "epoch": 3.442396313364055,
      "grad_norm": 15.069150924682617,
      "kl": 0.5634765625,
      "learning_rate": 1.376923076923077e-07,
      "loss": 0.0471,
      "reward": 0.7435916364192963,
      "reward_std": 0.07353068701922894,
      "reward_std_all": 0.06994991935789585,
      "rewards-max/forensics_sinkhorn_reward": 0.9776008427143097,
      "rewards-max/forensics_strict_edge_reward": 0.6176230013370514,
      "rewards-max/forensics_tiou_reward": 0.9717782437801361,
      "rewards/forensics_sinkhorn_reward": 0.735367476940155,
      "rewards/forensics_strict_edge_reward": 0.10626713559031487,
      "rewards/forensics_tiou_reward": 0.5923345685005188,
      "step": 1122
    },
    {
      "completion_length": 26.015625,
      "epoch": 3.445468509984639,
      "grad_norm": 10.396523475646973,
      "kl": 0.6328125,
      "learning_rate": 1.3692307692307692e-07,
      "loss": 0.3183,
      "reward": 0.7391970753669739,
      "reward_std": 0.06110982224345207,
      "reward_std_all": 0.0632869228720665,
      "rewards-max/forensics_sinkhorn_reward": 0.9767414331436157,
      "rewards-max/forensics_strict_edge_reward": 0.5014327168464661,
      "rewards-max/forensics_tiou_reward": 0.969913512468338,
      "rewards/forensics_sinkhorn_reward": 0.6772076189517975,
      "rewards/forensics_strict_edge_reward": 0.06835039146244526,
      "rewards/forensics_tiou_reward": 0.530837893486023,
      "step": 1123
    },
    {
      "completion_length": 26.546875,
      "epoch": 3.4485407066052227,
      "grad_norm": 12.634831428527832,
      "kl": 0.6171875,
      "learning_rate": 1.3615384615384616e-07,
      "loss": 0.086,
      "reward": 0.7564745843410492,
      "reward_std": 0.04386497847735882,
      "reward_std_all": 0.046165818348526955,
      "rewards-max/forensics_sinkhorn_reward": 0.982428640127182,
      "rewards-max/forensics_strict_edge_reward": 0.8582656681537628,
      "rewards-max/forensics_tiou_reward": 0.984647274017334,
      "rewards/forensics_sinkhorn_reward": 0.7320670485496521,
      "rewards/forensics_strict_edge_reward": 0.12305454537272453,
      "rewards/forensics_tiou_reward": 0.5992640256881714,
      "step": 1124
    },
    {
      "completion_length": 22.203125,
      "epoch": 3.4516129032258065,
      "grad_norm": 12.682928085327148,
      "kl": 0.658203125,
      "learning_rate": 1.3538461538461537e-07,
      "loss": 0.025,
      "reward": 0.7873030304908752,
      "reward_std": 0.025558721274137497,
      "reward_std_all": 0.025069032795727253,
      "rewards-max/forensics_sinkhorn_reward": 0.984107494354248,
      "rewards-max/forensics_strict_edge_reward": 0.6907829642295837,
      "rewards-max/forensics_tiou_reward": 0.9880912899971008,
      "rewards/forensics_sinkhorn_reward": 0.8190127313137054,
      "rewards/forensics_strict_edge_reward": 0.07035441510379314,
      "rewards/forensics_tiou_reward": 0.6379647850990295,
      "step": 1125
    },
    {
      "completion_length": 26.625,
      "epoch": 3.4546850998463903,
      "grad_norm": 11.510215759277344,
      "kl": 0.474609375,
      "learning_rate": 1.346153846153846e-07,
      "loss": -0.0673,
      "reward": 0.797871470451355,
      "reward_std": 0.015705514117144048,
      "reward_std_all": 0.015279582235962152,
      "rewards-max/forensics_sinkhorn_reward": 0.9844216704368591,
      "rewards-max/forensics_strict_edge_reward": 0.7933230996131897,
      "rewards-max/forensics_tiou_reward": 0.9870818257331848,
      "rewards/forensics_sinkhorn_reward": 0.8110988438129425,
      "rewards/forensics_strict_edge_reward": 0.12325645610690117,
      "rewards/forensics_tiou_reward": 0.680626392364502,
      "step": 1126
    },
    {
      "completion_length": 26.5,
      "epoch": 3.457757296466974,
      "grad_norm": 10.918588638305664,
      "kl": 0.60546875,
      "learning_rate": 1.3384615384615383e-07,
      "loss": 0.1855,
      "reward": 0.7611167728900909,
      "reward_std": 0.05689265765249729,
      "reward_std_all": 0.053828563541173935,
      "rewards-max/forensics_sinkhorn_reward": 0.9702573716640472,
      "rewards-max/forensics_strict_edge_reward": 0.355170801281929,
      "rewards-max/forensics_tiou_reward": 0.9523825943470001,
      "rewards/forensics_sinkhorn_reward": 0.7087833285331726,
      "rewards/forensics_strict_edge_reward": 0.04251148924231529,
      "rewards/forensics_tiou_reward": 0.5132697224617004,
      "step": 1127
    },
    {
      "completion_length": 25.625,
      "epoch": 3.460829493087558,
      "grad_norm": 9.908411979675293,
      "kl": 0.77734375,
      "learning_rate": 1.3307692307692307e-07,
      "loss": -0.0696,
      "reward": 0.7757076025009155,
      "reward_std": 0.021760548232123256,
      "reward_std_all": 0.025648019975051284,
      "rewards-max/forensics_sinkhorn_reward": 0.9797720909118652,
      "rewards-max/forensics_strict_edge_reward": 0.5166013538837433,
      "rewards-max/forensics_tiou_reward": 0.9712315201759338,
      "rewards/forensics_sinkhorn_reward": 0.7255820333957672,
      "rewards/forensics_strict_edge_reward": 0.06615985184907913,
      "rewards/forensics_tiou_reward": 0.5762732326984406,
      "step": 1128
    },
    {
      "completion_length": 24.625,
      "epoch": 3.463901689708141,
      "grad_norm": 9.32530403137207,
      "kl": 0.3369140625,
      "learning_rate": 1.323076923076923e-07,
      "loss": -0.0615,
      "reward": 0.7941484749317169,
      "reward_std": 0.02368521597236395,
      "reward_std_all": 0.021833323873579502,
      "rewards-max/forensics_sinkhorn_reward": 0.9857366681098938,
      "rewards-max/forensics_strict_edge_reward": 0.8582656681537628,
      "rewards-max/forensics_tiou_reward": 0.9919845759868622,
      "rewards/forensics_sinkhorn_reward": 0.8221920728683472,
      "rewards/forensics_strict_edge_reward": 0.18339689075946808,
      "rewards/forensics_tiou_reward": 0.7381474375724792,
      "step": 1129
    },
    {
      "completion_length": 27.734375,
      "epoch": 3.466973886328725,
      "grad_norm": 8.88404369354248,
      "kl": 0.701171875,
      "learning_rate": 1.3153846153846155e-07,
      "loss": 0.2842,
      "reward": 0.7573105096817017,
      "reward_std": 0.04419231042265892,
      "reward_std_all": 0.049600500613451004,
      "rewards-max/forensics_sinkhorn_reward": 0.9863146245479584,
      "rewards-max/forensics_strict_edge_reward": 0.7958523333072662,
      "rewards-max/forensics_tiou_reward": 0.9931711852550507,
      "rewards/forensics_sinkhorn_reward": 0.6632798612117767,
      "rewards/forensics_strict_edge_reward": 0.09018453769385815,
      "rewards/forensics_tiou_reward": 0.5710558891296387,
      "step": 1130
    },
    {
      "completion_length": 23.484375,
      "epoch": 3.4700460829493087,
      "grad_norm": 12.121581077575684,
      "kl": 0.3974609375,
      "learning_rate": 1.3076923076923076e-07,
      "loss": -0.2237,
      "reward": 0.7956036329269409,
      "reward_std": 0.0027379296952858567,
      "reward_std_all": 0.008452476467937231,
      "rewards-max/forensics_sinkhorn_reward": 0.9853188395500183,
      "rewards-max/forensics_strict_edge_reward": 0.7032595127820969,
      "rewards-max/forensics_tiou_reward": 0.991793155670166,
      "rewards/forensics_sinkhorn_reward": 0.7969110310077667,
      "rewards/forensics_strict_edge_reward": 0.04773612506687641,
      "rewards/forensics_tiou_reward": 0.6095854043960571,
      "step": 1131
    },
    {
      "completion_length": 26.34375,
      "epoch": 3.4731182795698925,
      "grad_norm": 11.051849365234375,
      "kl": 0.5380859375,
      "learning_rate": 1.3e-07,
      "loss": 0.0294,
      "reward": 0.7464265823364258,
      "reward_std": 0.045438237488269806,
      "reward_std_all": 0.051734769716858864,
      "rewards-max/forensics_sinkhorn_reward": 0.9794802665710449,
      "rewards-max/forensics_strict_edge_reward": 0.5945845991373062,
      "rewards-max/forensics_tiou_reward": 0.9756926000118256,
      "rewards/forensics_sinkhorn_reward": 0.6934544146060944,
      "rewards/forensics_strict_edge_reward": 0.0829476211220026,
      "rewards/forensics_tiou_reward": 0.5466109812259674,
      "step": 1132
    },
    {
      "completion_length": 24.71875,
      "epoch": 3.4761904761904763,
      "grad_norm": 13.112963676452637,
      "kl": 0.49609375,
      "learning_rate": 1.2923076923076924e-07,
      "loss": -0.0717,
      "reward": 0.7777782082557678,
      "reward_std": 0.03329175477847457,
      "reward_std_all": 0.034389812499284744,
      "rewards-max/forensics_sinkhorn_reward": 0.9852594137191772,
      "rewards-max/forensics_strict_edge_reward": 0.5311143696308136,
      "rewards-max/forensics_tiou_reward": 0.9880183339118958,
      "rewards/forensics_sinkhorn_reward": 0.7646215856075287,
      "rewards/forensics_strict_edge_reward": 0.13638141378760338,
      "rewards/forensics_tiou_reward": 0.6315003335475922,
      "step": 1133
    },
    {
      "completion_length": 25.21875,
      "epoch": 3.47926267281106,
      "grad_norm": 9.54619026184082,
      "kl": 0.541015625,
      "learning_rate": 1.2846153846153845e-07,
      "loss": 0.0732,
      "reward": 0.7864454984664917,
      "reward_std": 0.03455973044037819,
      "reward_std_all": 0.031500390730798244,
      "rewards-max/forensics_sinkhorn_reward": 0.9879853129386902,
      "rewards-max/forensics_strict_edge_reward": 0.8756595849990845,
      "rewards-max/forensics_tiou_reward": 0.9972742199897766,
      "rewards/forensics_sinkhorn_reward": 0.7628181874752045,
      "rewards/forensics_strict_edge_reward": 0.10485276952385902,
      "rewards/forensics_tiou_reward": 0.6474072933197021,
      "step": 1134
    },
    {
      "completion_length": 26.765625,
      "epoch": 3.482334869431644,
      "grad_norm": 12.486318588256836,
      "kl": 0.490234375,
      "learning_rate": 1.276923076923077e-07,
      "loss": 0.3259,
      "reward": 0.736334353685379,
      "reward_std": 0.061598051339387894,
      "reward_std_all": 0.06609220430254936,
      "rewards-max/forensics_sinkhorn_reward": 0.9821443259716034,
      "rewards-max/forensics_strict_edge_reward": 0.7972188889980316,
      "rewards-max/forensics_tiou_reward": 0.9831826388835907,
      "rewards/forensics_sinkhorn_reward": 0.6905666589736938,
      "rewards/forensics_strict_edge_reward": 0.058995144441723824,
      "rewards/forensics_tiou_reward": 0.4987250417470932,
      "step": 1135
    },
    {
      "completion_length": 25.5625,
      "epoch": 3.485407066052227,
      "grad_norm": 11.689701080322266,
      "kl": 0.6484375,
      "learning_rate": 1.269230769230769e-07,
      "loss": 0.06,
      "reward": 0.7724230587482452,
      "reward_std": 0.043397649424150586,
      "reward_std_all": 0.041644308250397444,
      "rewards-max/forensics_sinkhorn_reward": 0.9637739658355713,
      "rewards-max/forensics_strict_edge_reward": 0.44948313385248184,
      "rewards-max/forensics_tiou_reward": 0.9387840926647186,
      "rewards/forensics_sinkhorn_reward": 0.6990785002708435,
      "rewards/forensics_strict_edge_reward": 0.0784520385786891,
      "rewards/forensics_tiou_reward": 0.5539190769195557,
      "step": 1136
    },
    {
      "completion_length": 21.6875,
      "epoch": 3.488479262672811,
      "grad_norm": 13.609683990478516,
      "kl": 0.87890625,
      "learning_rate": 1.2615384615384615e-07,
      "loss": -0.1997,
      "reward": 0.7776691019535065,
      "reward_std": 0.010789241059683263,
      "reward_std_all": 0.01927279308438301,
      "rewards-max/forensics_sinkhorn_reward": 0.9839054048061371,
      "rewards-max/forensics_strict_edge_reward": 0.7403165996074677,
      "rewards-max/forensics_tiou_reward": 0.9867509603500366,
      "rewards/forensics_sinkhorn_reward": 0.7710570991039276,
      "rewards/forensics_strict_edge_reward": 0.08406352251768112,
      "rewards/forensics_tiou_reward": 0.5702893435955048,
      "step": 1137
    },
    {
      "completion_length": 25.28125,
      "epoch": 3.4915514592933947,
      "grad_norm": 11.342695236206055,
      "kl": 0.455078125,
      "learning_rate": 1.2538461538461536e-07,
      "loss": -0.0499,
      "reward": 0.770444244146347,
      "reward_std": 0.0404089605435729,
      "reward_std_all": 0.04031524062156677,
      "rewards-max/forensics_sinkhorn_reward": 0.9786691963672638,
      "rewards-max/forensics_strict_edge_reward": 0.4422979950904846,
      "rewards-max/forensics_tiou_reward": 0.9773220419883728,
      "rewards/forensics_sinkhorn_reward": 0.7592131793498993,
      "rewards/forensics_strict_edge_reward": 0.06464234739542007,
      "rewards/forensics_tiou_reward": 0.5909578800201416,
      "step": 1138
    },
    {
      "completion_length": 24.78125,
      "epoch": 3.4946236559139785,
      "grad_norm": 12.432580947875977,
      "kl": 0.626953125,
      "learning_rate": 1.246153846153846e-07,
      "loss": -0.0827,
      "reward": 0.7911675274372101,
      "reward_std": 0.017863242886960506,
      "reward_std_all": 0.018429117277264595,
      "rewards-max/forensics_sinkhorn_reward": 0.9773384630680084,
      "rewards-max/forensics_strict_edge_reward": 0.7500002980232239,
      "rewards-max/forensics_tiou_reward": 0.971153050661087,
      "rewards/forensics_sinkhorn_reward": 0.7905115485191345,
      "rewards/forensics_strict_edge_reward": 0.08429661393165588,
      "rewards/forensics_tiou_reward": 0.597201943397522,
      "step": 1139
    },
    {
      "completion_length": 25.578125,
      "epoch": 3.4976958525345623,
      "grad_norm": 14.096527099609375,
      "kl": 0.4384765625,
      "learning_rate": 1.2384615384615384e-07,
      "loss": -0.0938,
      "reward": 0.7973716259002686,
      "reward_std": 0.014821643941104412,
      "reward_std_all": 0.014712423086166382,
      "rewards-max/forensics_sinkhorn_reward": 0.9749095737934113,
      "rewards-max/forensics_strict_edge_reward": 0.541615754365921,
      "rewards-max/forensics_tiou_reward": 0.9633204340934753,
      "rewards/forensics_sinkhorn_reward": 0.754557728767395,
      "rewards/forensics_strict_edge_reward": 0.09270425513386726,
      "rewards/forensics_tiou_reward": 0.5959238409996033,
      "step": 1140
    },
    {
      "completion_length": 22.125,
      "epoch": 3.500768049155146,
      "grad_norm": 9.559179306030273,
      "kl": 0.4267578125,
      "learning_rate": 1.2307692307692308e-07,
      "loss": 0.0906,
      "reward": 0.7436147332191467,
      "reward_std": 0.03056172374635935,
      "reward_std_all": 0.04851529002189636,
      "rewards-max/forensics_sinkhorn_reward": 0.9868890643119812,
      "rewards-max/forensics_strict_edge_reward": 0.882964164018631,
      "rewards-max/forensics_tiou_reward": 0.993230402469635,
      "rewards/forensics_sinkhorn_reward": 0.7265611588954926,
      "rewards/forensics_strict_edge_reward": 0.09858522564172745,
      "rewards/forensics_tiou_reward": 0.5717816352844238,
      "step": 1141
    },
    {
      "completion_length": 26.703125,
      "epoch": 3.50384024577573,
      "grad_norm": 10.443686485290527,
      "kl": 0.693359375,
      "learning_rate": 1.223076923076923e-07,
      "loss": -0.1212,
      "reward": 0.7776529788970947,
      "reward_std": 0.03657007683068514,
      "reward_std_all": 0.03511129319667816,
      "rewards-max/forensics_sinkhorn_reward": 0.9863129258155823,
      "rewards-max/forensics_strict_edge_reward": 0.8704091012477875,
      "rewards-max/forensics_tiou_reward": 0.9944343268871307,
      "rewards/forensics_sinkhorn_reward": 0.7652160823345184,
      "rewards/forensics_strict_edge_reward": 0.18059473857283592,
      "rewards/forensics_tiou_reward": 0.6705791056156158,
      "step": 1142
    },
    {
      "completion_length": 25.109375,
      "epoch": 3.506912442396313,
      "grad_norm": 11.807889938354492,
      "kl": 0.6796875,
      "learning_rate": 1.2153846153846154e-07,
      "loss": 0.0479,
      "reward": 0.7817257642745972,
      "reward_std": 0.029363995417952538,
      "reward_std_all": 0.029208393301814795,
      "rewards-max/forensics_sinkhorn_reward": 0.9826914668083191,
      "rewards-max/forensics_strict_edge_reward": 0.7709200084209442,
      "rewards-max/forensics_tiou_reward": 0.9822668433189392,
      "rewards/forensics_sinkhorn_reward": 0.761781632900238,
      "rewards/forensics_strict_edge_reward": 0.07932011783123016,
      "rewards/forensics_tiou_reward": 0.5767760872840881,
      "step": 1143
    },
    {
      "completion_length": 23.953125,
      "epoch": 3.509984639016897,
      "grad_norm": 11.424786567687988,
      "kl": 0.837890625,
      "learning_rate": 1.2076923076923075e-07,
      "loss": -0.0762,
      "reward": 0.7801632285118103,
      "reward_std": 0.03558121528476477,
      "reward_std_all": 0.03505116235464811,
      "rewards-max/forensics_sinkhorn_reward": 0.9848816394805908,
      "rewards-max/forensics_strict_edge_reward": 0.5098983347415924,
      "rewards-max/forensics_tiou_reward": 0.9872660636901855,
      "rewards/forensics_sinkhorn_reward": 0.8044893741607666,
      "rewards/forensics_strict_edge_reward": 0.09089949540793896,
      "rewards/forensics_tiou_reward": 0.6387327313423157,
      "step": 1144
    },
    {
      "completion_length": 25.6875,
      "epoch": 3.5130568356374807,
      "grad_norm": 11.44002914428711,
      "kl": 0.5361328125,
      "learning_rate": 1.2e-07,
      "loss": -0.1615,
      "reward": 0.780617356300354,
      "reward_std": 0.026029914966784418,
      "reward_std_all": 0.026902901707217097,
      "rewards-max/forensics_sinkhorn_reward": 0.9737927615642548,
      "rewards-max/forensics_strict_edge_reward": 0.4710261672735214,
      "rewards-max/forensics_tiou_reward": 0.9635888636112213,
      "rewards/forensics_sinkhorn_reward": 0.7835959792137146,
      "rewards/forensics_strict_edge_reward": 0.11110088974237442,
      "rewards/forensics_tiou_reward": 0.6559731364250183,
      "step": 1145
    },
    {
      "completion_length": 24.609375,
      "epoch": 3.5161290322580645,
      "grad_norm": 14.602139472961426,
      "kl": 1.087890625,
      "learning_rate": 1.1923076923076923e-07,
      "loss": -0.0486,
      "reward": 0.7370563745498657,
      "reward_std": 0.013359609991312027,
      "reward_std_all": 0.04048900073394179,
      "rewards-max/forensics_sinkhorn_reward": 0.9737805724143982,
      "rewards-max/forensics_strict_edge_reward": 0.48898807168006897,
      "rewards-max/forensics_tiou_reward": 0.9607247114181519,
      "rewards/forensics_sinkhorn_reward": 0.7005476951599121,
      "rewards/forensics_strict_edge_reward": 0.05632337927818298,
      "rewards/forensics_tiou_reward": 0.529996395111084,
      "step": 1146
    },
    {
      "completion_length": 28.515625,
      "epoch": 3.5192012288786483,
      "grad_norm": 11.599847793579102,
      "kl": 0.8515625,
      "learning_rate": 1.1846153846153846e-07,
      "loss": 0.1079,
      "reward": 0.7789416015148163,
      "reward_std": 0.03194059379165992,
      "reward_std_all": 0.031637536361813545,
      "rewards-max/forensics_sinkhorn_reward": 0.981567919254303,
      "rewards-max/forensics_strict_edge_reward": 0.41102899610996246,
      "rewards-max/forensics_tiou_reward": 0.9791774153709412,
      "rewards/forensics_sinkhorn_reward": 0.7247489988803864,
      "rewards/forensics_strict_edge_reward": 0.03546471148729324,
      "rewards/forensics_tiou_reward": 0.5595102310180664,
      "step": 1147
    },
    {
      "completion_length": 26.5625,
      "epoch": 3.522273425499232,
      "grad_norm": 13.769466400146484,
      "kl": 0.61328125,
      "learning_rate": 1.1769230769230768e-07,
      "loss": 0.1428,
      "reward": 0.7480158805847168,
      "reward_std": 0.0537372175604105,
      "reward_std_all": 0.057614775374531746,
      "rewards-max/forensics_sinkhorn_reward": 0.9837301075458527,
      "rewards-max/forensics_strict_edge_reward": 0.7923295497894287,
      "rewards-max/forensics_tiou_reward": 0.9862333238124847,
      "rewards/forensics_sinkhorn_reward": 0.7234897017478943,
      "rewards/forensics_strict_edge_reward": 0.10451884940266609,
      "rewards/forensics_tiou_reward": 0.5501619875431061,
      "step": 1148
    },
    {
      "completion_length": 25.265625,
      "epoch": 3.525345622119816,
      "grad_norm": 10.724016189575195,
      "kl": 0.552734375,
      "learning_rate": 1.1692307692307691e-07,
      "loss": 0.079,
      "reward": 0.7602148056030273,
      "reward_std": 0.02879963256418705,
      "reward_std_all": 0.03609631024301052,
      "rewards-max/forensics_sinkhorn_reward": 0.9825623333454132,
      "rewards-max/forensics_strict_edge_reward": 0.7286747694015503,
      "rewards-max/forensics_tiou_reward": 0.983607143163681,
      "rewards/forensics_sinkhorn_reward": 0.7449101805686951,
      "rewards/forensics_strict_edge_reward": 0.11884859949350357,
      "rewards/forensics_tiou_reward": 0.559075266122818,
      "step": 1149
    },
    {
      "completion_length": 25.0,
      "epoch": 3.528417818740399,
      "grad_norm": 13.589760780334473,
      "kl": 0.896484375,
      "learning_rate": 1.1615384615384615e-07,
      "loss": -0.02,
      "reward": 0.7650295794010162,
      "reward_std": 0.04349036607891321,
      "reward_std_all": 0.04379585385322571,
      "rewards-max/forensics_sinkhorn_reward": 0.9838763475418091,
      "rewards-max/forensics_strict_edge_reward": 0.586119145154953,
      "rewards-max/forensics_tiou_reward": 0.989432156085968,
      "rewards/forensics_sinkhorn_reward": 0.7407657504081726,
      "rewards/forensics_strict_edge_reward": 0.08925218880176544,
      "rewards/forensics_tiou_reward": 0.6135571599006653,
      "step": 1150
    },
    {
      "completion_length": 27.171875,
      "epoch": 3.531490015360983,
      "grad_norm": 10.239307403564453,
      "kl": 0.673828125,
      "learning_rate": 1.1538461538461539e-07,
      "loss": 0.1827,
      "reward": 0.7815290689468384,
      "reward_std": 0.03068607859313488,
      "reward_std_all": 0.030095945578068495,
      "rewards-max/forensics_sinkhorn_reward": 0.9814561009407043,
      "rewards-max/forensics_strict_edge_reward": 0.8120969831943512,
      "rewards-max/forensics_tiou_reward": 0.9830134212970734,
      "rewards/forensics_sinkhorn_reward": 0.7245646417140961,
      "rewards/forensics_strict_edge_reward": 0.10014490410685539,
      "rewards/forensics_tiou_reward": 0.5898266136646271,
      "step": 1151
    },
    {
      "completion_length": 22.65625,
      "epoch": 3.5345622119815667,
      "grad_norm": 10.350892066955566,
      "kl": 0.4130859375,
      "learning_rate": 1.1461538461538462e-07,
      "loss": 0.0633,
      "reward": 0.7660742402076721,
      "reward_std": 0.03552316688001156,
      "reward_std_all": 0.03953780746087432,
      "rewards-max/forensics_sinkhorn_reward": 0.9854451715946198,
      "rewards-max/forensics_strict_edge_reward": 0.8929734230041504,
      "rewards-max/forensics_tiou_reward": 0.9897316992282867,
      "rewards/forensics_sinkhorn_reward": 0.7532542049884796,
      "rewards/forensics_strict_edge_reward": 0.11461624130606651,
      "rewards/forensics_tiou_reward": 0.5845685005187988,
      "step": 1152
    },
    {
      "completion_length": 24.25,
      "epoch": 3.5376344086021505,
      "grad_norm": 13.574204444885254,
      "kl": 0.7158203125,
      "learning_rate": 1.1384615384615384e-07,
      "loss": -0.1874,
      "reward": 0.7711814343929291,
      "reward_std": 0.011809861869551241,
      "reward_std_all": 0.023128396249376237,
      "rewards-max/forensics_sinkhorn_reward": 0.9817728698253632,
      "rewards-max/forensics_strict_edge_reward": 0.7181241810321808,
      "rewards-max/forensics_tiou_reward": 0.9806186854839325,
      "rewards/forensics_sinkhorn_reward": 0.7452428340911865,
      "rewards/forensics_strict_edge_reward": 0.09615105390548706,
      "rewards/forensics_tiou_reward": 0.5785598754882812,
      "step": 1153
    },
    {
      "completion_length": 24.234375,
      "epoch": 3.5407066052227343,
      "grad_norm": 11.570791244506836,
      "kl": 0.5029296875,
      "learning_rate": 1.1307692307692307e-07,
      "loss": -0.1509,
      "reward": 0.7929271757602692,
      "reward_std": 0.013844172470271587,
      "reward_std_all": 0.01540764793753624,
      "rewards-max/forensics_sinkhorn_reward": 0.9814042150974274,
      "rewards-max/forensics_strict_edge_reward": 0.6149742305278778,
      "rewards-max/forensics_tiou_reward": 0.9792933166027069,
      "rewards/forensics_sinkhorn_reward": 0.8143115639686584,
      "rewards/forensics_strict_edge_reward": 0.10432364791631699,
      "rewards/forensics_tiou_reward": 0.6642091870307922,
      "step": 1154
    },
    {
      "completion_length": 24.21875,
      "epoch": 3.543778801843318,
      "grad_norm": 14.544415473937988,
      "kl": 0.685546875,
      "learning_rate": 1.123076923076923e-07,
      "loss": -0.0387,
      "reward": 0.770831823348999,
      "reward_std": 0.011086941231042147,
      "reward_std_all": 0.024991270154714584,
      "rewards-max/forensics_sinkhorn_reward": 0.9842748045921326,
      "rewards-max/forensics_strict_edge_reward": 0.7744058072566986,
      "rewards-max/forensics_tiou_reward": 0.9888860881328583,
      "rewards/forensics_sinkhorn_reward": 0.7489902973175049,
      "rewards/forensics_strict_edge_reward": 0.14078734070062637,
      "rewards/forensics_tiou_reward": 0.5825961530208588,
      "step": 1155
    },
    {
      "completion_length": 26.328125,
      "epoch": 3.546850998463902,
      "grad_norm": 10.969175338745117,
      "kl": 0.828125,
      "learning_rate": 1.1153846153846154e-07,
      "loss": -0.2603,
      "reward": 0.8050774335861206,
      "reward_std": 0.0016395333805121481,
      "reward_std_all": 0.0036767798010259867,
      "rewards-max/forensics_sinkhorn_reward": 0.9715954959392548,
      "rewards-max/forensics_strict_edge_reward": 0.5557698607444763,
      "rewards-max/forensics_tiou_reward": 0.9547892808914185,
      "rewards/forensics_sinkhorn_reward": 0.821320652961731,
      "rewards/forensics_strict_edge_reward": 0.05167543143033981,
      "rewards/forensics_tiou_reward": 0.6291394829750061,
      "step": 1156
    },
    {
      "completion_length": 28.65625,
      "epoch": 3.549923195084485,
      "grad_norm": 8.46511173248291,
      "kl": 0.80859375,
      "learning_rate": 1.1076923076923077e-07,
      "loss": 0.0212,
      "reward": 0.7872219681739807,
      "reward_std": 0.0243695005774498,
      "reward_std_all": 0.02405722811818123,
      "rewards-max/forensics_sinkhorn_reward": 0.9853765666484833,
      "rewards-max/forensics_strict_edge_reward": 0.5559177100658417,
      "rewards-max/forensics_tiou_reward": 0.9891618192195892,
      "rewards/forensics_sinkhorn_reward": 0.8298985064029694,
      "rewards/forensics_strict_edge_reward": 0.1021713986992836,
      "rewards/forensics_tiou_reward": 0.710197925567627,
      "step": 1157
    },
    {
      "completion_length": 24.40625,
      "epoch": 3.5529953917050694,
      "grad_norm": 10.096480369567871,
      "kl": 0.458984375,
      "learning_rate": 1.0999999999999999e-07,
      "loss": -0.1325,
      "reward": 0.791425883769989,
      "reward_std": 0.027718893252313137,
      "reward_std_all": 0.02547239325940609,
      "rewards-max/forensics_sinkhorn_reward": 0.985553652048111,
      "rewards-max/forensics_strict_edge_reward": 0.7933230996131897,
      "rewards-max/forensics_tiou_reward": 0.9900161623954773,
      "rewards/forensics_sinkhorn_reward": 0.8142213523387909,
      "rewards/forensics_strict_edge_reward": 0.1747291050851345,
      "rewards/forensics_tiou_reward": 0.7085292637348175,
      "step": 1158
    },
    {
      "completion_length": 24.0625,
      "epoch": 3.5560675883256527,
      "grad_norm": 12.232372283935547,
      "kl": 0.63671875,
      "learning_rate": 1.0923076923076922e-07,
      "loss": 0.3071,
      "reward": 0.7424595952033997,
      "reward_std": 0.07973042130470276,
      "reward_std_all": 0.07534226030111313,
      "rewards-max/forensics_sinkhorn_reward": 0.9797914326190948,
      "rewards-max/forensics_strict_edge_reward": 0.4738437533378601,
      "rewards-max/forensics_tiou_reward": 0.9739565551280975,
      "rewards/forensics_sinkhorn_reward": 0.7517452836036682,
      "rewards/forensics_strict_edge_reward": 0.0813475213944912,
      "rewards/forensics_tiou_reward": 0.5652473866939545,
      "step": 1159
    },
    {
      "completion_length": 27.5,
      "epoch": 3.5591397849462365,
      "grad_norm": 9.722888946533203,
      "kl": 0.947265625,
      "learning_rate": 1.0846153846153845e-07,
      "loss": 0.1561,
      "reward": 0.7539249956607819,
      "reward_std": 0.06248587742447853,
      "reward_std_all": 0.05926605127751827,
      "rewards-max/forensics_sinkhorn_reward": 0.9762389361858368,
      "rewards-max/forensics_strict_edge_reward": 0.585642009973526,
      "rewards-max/forensics_tiou_reward": 0.970615565776825,
      "rewards/forensics_sinkhorn_reward": 0.7375591993331909,
      "rewards/forensics_strict_edge_reward": 0.09936311841011047,
      "rewards/forensics_tiou_reward": 0.5574853420257568,
      "step": 1160
    },
    {
      "completion_length": 21.875,
      "epoch": 3.5622119815668203,
      "grad_norm": 12.826536178588867,
      "kl": 0.4755859375,
      "learning_rate": 1.076923076923077e-07,
      "loss": -0.0297,
      "reward": 0.7928566932678223,
      "reward_std": 0.022631769068539143,
      "reward_std_all": 0.02185456920415163,
      "rewards-max/forensics_sinkhorn_reward": 0.9817728698253632,
      "rewards-max/forensics_strict_edge_reward": 0.6209261119365692,
      "rewards-max/forensics_tiou_reward": 0.9803399443626404,
      "rewards/forensics_sinkhorn_reward": 0.778844803571701,
      "rewards/forensics_strict_edge_reward": 0.08340640366077423,
      "rewards/forensics_tiou_reward": 0.601335197687149,
      "step": 1161
    },
    {
      "completion_length": 23.875,
      "epoch": 3.565284178187404,
      "grad_norm": 9.03491497039795,
      "kl": 0.4814453125,
      "learning_rate": 1.0692307692307693e-07,
      "loss": -0.2569,
      "reward": 0.8036457896232605,
      "reward_std": 0.006249421508982778,
      "reward_std_all": 0.006902649067342281,
      "rewards-max/forensics_sinkhorn_reward": 0.9864700138568878,
      "rewards-max/forensics_strict_edge_reward": 0.7482926547527313,
      "rewards-max/forensics_tiou_reward": 0.9926517605781555,
      "rewards/forensics_sinkhorn_reward": 0.8618941903114319,
      "rewards/forensics_strict_edge_reward": 0.11883623898029327,
      "rewards/forensics_tiou_reward": 0.7134488523006439,
      "step": 1162
    },
    {
      "completion_length": 21.96875,
      "epoch": 3.568356374807988,
      "grad_norm": 12.461723327636719,
      "kl": 0.5029296875,
      "learning_rate": 1.0615384615384615e-07,
      "loss": -0.0504,
      "reward": 0.7760225534439087,
      "reward_std": 0.0388097376562655,
      "reward_std_all": 0.037260085344314575,
      "rewards-max/forensics_sinkhorn_reward": 0.9619259238243103,
      "rewards-max/forensics_strict_edge_reward": 0.740818202495575,
      "rewards-max/forensics_tiou_reward": 0.9377565979957581,
      "rewards/forensics_sinkhorn_reward": 0.7591622769832611,
      "rewards/forensics_strict_edge_reward": 0.08495928160846233,
      "rewards/forensics_tiou_reward": 0.5622552931308746,
      "step": 1163
    },
    {
      "completion_length": 24.46875,
      "epoch": 3.571428571428571,
      "grad_norm": 10.49669361114502,
      "kl": 0.517578125,
      "learning_rate": 1.0538461538461538e-07,
      "loss": 0.0281,
      "reward": 0.7812178432941437,
      "reward_std": 0.015714809764176607,
      "reward_std_all": 0.02055043401196599,
      "rewards-max/forensics_sinkhorn_reward": 0.9770502150058746,
      "rewards-max/forensics_strict_edge_reward": 0.6918988227844238,
      "rewards-max/forensics_tiou_reward": 0.9691205024719238,
      "rewards/forensics_sinkhorn_reward": 0.7165834009647369,
      "rewards/forensics_strict_edge_reward": 0.10096453875303268,
      "rewards/forensics_tiou_reward": 0.5906884670257568,
      "step": 1164
    },
    {
      "completion_length": 27.15625,
      "epoch": 3.5745007680491554,
      "grad_norm": 11.765921592712402,
      "kl": 0.3984375,
      "learning_rate": 1.0461538461538461e-07,
      "loss": 0.1133,
      "reward": 0.7900805771350861,
      "reward_std": 0.020479163737036288,
      "reward_std_all": 0.020730976946651936,
      "rewards-max/forensics_sinkhorn_reward": 0.9762557446956635,
      "rewards-max/forensics_strict_edge_reward": 0.3754626661539078,
      "rewards-max/forensics_tiou_reward": 0.9660048484802246,
      "rewards/forensics_sinkhorn_reward": 0.7105123996734619,
      "rewards/forensics_strict_edge_reward": 0.03414713032543659,
      "rewards/forensics_tiou_reward": 0.5063060820102692,
      "step": 1165
    },
    {
      "completion_length": 24.53125,
      "epoch": 3.5775729646697387,
      "grad_norm": 11.134239196777344,
      "kl": 0.578125,
      "learning_rate": 1.0384615384615385e-07,
      "loss": 0.0382,
      "reward": 0.7450740039348602,
      "reward_std": 0.06144073698669672,
      "reward_std_all": 0.061927422881126404,
      "rewards-max/forensics_sinkhorn_reward": 0.9835124611854553,
      "rewards-max/forensics_strict_edge_reward": 0.7936499714851379,
      "rewards-max/forensics_tiou_reward": 0.9838948249816895,
      "rewards/forensics_sinkhorn_reward": 0.7454707324504852,
      "rewards/forensics_strict_edge_reward": 0.12351399660110474,
      "rewards/forensics_tiou_reward": 0.583459734916687,
      "step": 1166
    },
    {
      "completion_length": 24.234375,
      "epoch": 3.5806451612903225,
      "grad_norm": 10.904616355895996,
      "kl": 0.5859375,
      "learning_rate": 1.0307692307692307e-07,
      "loss": 0.1255,
      "reward": 0.7605392932891846,
      "reward_std": 0.035454392433166504,
      "reward_std_all": 0.0426468700170517,
      "rewards-max/forensics_sinkhorn_reward": 0.986970067024231,
      "rewards-max/forensics_strict_edge_reward": 0.7933230996131897,
      "rewards-max/forensics_tiou_reward": 0.995958536863327,
      "rewards/forensics_sinkhorn_reward": 0.7162076532840729,
      "rewards/forensics_strict_edge_reward": 0.10727665387094021,
      "rewards/forensics_tiou_reward": 0.5759623348712921,
      "step": 1167
    },
    {
      "completion_length": 28.65625,
      "epoch": 3.5837173579109063,
      "grad_norm": 9.482207298278809,
      "kl": 0.611328125,
      "learning_rate": 1.023076923076923e-07,
      "loss": 0.0781,
      "reward": 0.7768663763999939,
      "reward_std": 0.046830954030156136,
      "reward_std_all": 0.04309456329792738,
      "rewards-max/forensics_sinkhorn_reward": 0.977365255355835,
      "rewards-max/forensics_strict_edge_reward": 0.5434702187776566,
      "rewards-max/forensics_tiou_reward": 0.9702591598033905,
      "rewards/forensics_sinkhorn_reward": 0.7445121705532074,
      "rewards/forensics_strict_edge_reward": 0.0598469078540802,
      "rewards/forensics_tiou_reward": 0.592525064945221,
      "step": 1168
    },
    {
      "completion_length": 28.296875,
      "epoch": 3.58678955453149,
      "grad_norm": 18.70382308959961,
      "kl": 1.318359375,
      "learning_rate": 1.0153846153846153e-07,
      "loss": -0.05,
      "reward": 0.7675116956233978,
      "reward_std": 0.004273375729098916,
      "reward_std_all": 0.02363694692030549,
      "rewards-max/forensics_sinkhorn_reward": 0.9886131286621094,
      "rewards-max/forensics_strict_edge_reward": 0.6824378371238708,
      "rewards-max/forensics_tiou_reward": 0.9987132251262665,
      "rewards/forensics_sinkhorn_reward": 0.7618440389633179,
      "rewards/forensics_strict_edge_reward": 0.07871289178729057,
      "rewards/forensics_tiou_reward": 0.6439692676067352,
      "step": 1169
    },
    {
      "completion_length": 24.03125,
      "epoch": 3.589861751152074,
      "grad_norm": 11.694960594177246,
      "kl": 0.4921875,
      "learning_rate": 1.0076923076923076e-07,
      "loss": -0.1699,
      "reward": 0.7970871329307556,
      "reward_std": 0.016896786633878946,
      "reward_std_all": 0.01650410331785679,
      "rewards-max/forensics_sinkhorn_reward": 0.9847809970378876,
      "rewards-max/forensics_strict_edge_reward": 0.7933230996131897,
      "rewards-max/forensics_tiou_reward": 0.9894545376300812,
      "rewards/forensics_sinkhorn_reward": 0.8267052471637726,
      "rewards/forensics_strict_edge_reward": 0.1134420596063137,
      "rewards/forensics_tiou_reward": 0.6784360408782959,
      "step": 1170
    },
    {
      "completion_length": 23.890625,
      "epoch": 3.592933947772657,
      "grad_norm": 9.710603713989258,
      "kl": 0.470703125,
      "learning_rate": 1e-07,
      "loss": 0.1401,
      "reward": 0.7743276357650757,
      "reward_std": 0.04041757062077522,
      "reward_std_all": 0.03876770846545696,
      "rewards-max/forensics_sinkhorn_reward": 0.9884518682956696,
      "rewards-max/forensics_strict_edge_reward": 0.9513615369796753,
      "rewards-max/forensics_tiou_reward": 0.9985567033290863,
      "rewards/forensics_sinkhorn_reward": 0.7696835100650787,
      "rewards/forensics_strict_edge_reward": 0.07422134093940258,
      "rewards/forensics_tiou_reward": 0.5910504162311554,
      "step": 1171
    },
    {
      "completion_length": 26.0625,
      "epoch": 3.5960061443932414,
      "grad_norm": 10.573676109313965,
      "kl": 0.60546875,
      "learning_rate": 9.923076923076922e-08,
      "loss": -0.0693,
      "reward": 0.7673884034156799,
      "reward_std": 0.016727661713957787,
      "reward_std_all": 0.02773002814501524,
      "rewards-max/forensics_sinkhorn_reward": 0.9841702282428741,
      "rewards-max/forensics_strict_edge_reward": 0.6633935272693634,
      "rewards-max/forensics_tiou_reward": 0.9878160953521729,
      "rewards/forensics_sinkhorn_reward": 0.7464562058448792,
      "rewards/forensics_strict_edge_reward": 0.06088129058480263,
      "rewards/forensics_tiou_reward": 0.5853796601295471,
      "step": 1172
    },
    {
      "completion_length": 23.453125,
      "epoch": 3.5990783410138247,
      "grad_norm": 13.601325035095215,
      "kl": 0.509765625,
      "learning_rate": 9.846153846153846e-08,
      "loss": -0.0792,
      "reward": 0.7627663612365723,
      "reward_std": 0.0325152431614697,
      "reward_std_all": 0.03693138505332172,
      "rewards-max/forensics_sinkhorn_reward": 0.9798266887664795,
      "rewards-max/forensics_strict_edge_reward": 0.5274569988250732,
      "rewards-max/forensics_tiou_reward": 0.9738565683364868,
      "rewards/forensics_sinkhorn_reward": 0.7687188982963562,
      "rewards/forensics_strict_edge_reward": 0.09581691585481167,
      "rewards/forensics_tiou_reward": 0.6280214488506317,
      "step": 1173
    },
    {
      "completion_length": 24.03125,
      "epoch": 3.6021505376344085,
      "grad_norm": 11.042792320251465,
      "kl": 0.5234375,
      "learning_rate": 9.769230769230769e-08,
      "loss": -0.0032,
      "reward": 0.7535873353481293,
      "reward_std": 0.04372503608465195,
      "reward_std_all": 0.049754729494452477,
      "rewards-max/forensics_sinkhorn_reward": 0.9862141311168671,
      "rewards-max/forensics_strict_edge_reward": 0.7815065383911133,
      "rewards-max/forensics_tiou_reward": 0.9918584227561951,
      "rewards/forensics_sinkhorn_reward": 0.733283281326294,
      "rewards/forensics_strict_edge_reward": 0.09769481793045998,
      "rewards/forensics_tiou_reward": 0.5836344659328461,
      "step": 1174
    },
    {
      "completion_length": 24.0625,
      "epoch": 3.6052227342549923,
      "grad_norm": 12.52713394165039,
      "kl": 0.6328125,
      "learning_rate": 9.692307692307692e-08,
      "loss": 0.0859,
      "reward": 0.7671723961830139,
      "reward_std": 0.023608575575053692,
      "reward_std_all": 0.0324220759794116,
      "rewards-max/forensics_sinkhorn_reward": 0.9857602417469025,
      "rewards-max/forensics_strict_edge_reward": 0.680637389421463,
      "rewards-max/forensics_tiou_reward": 0.9903383553028107,
      "rewards/forensics_sinkhorn_reward": 0.7449906170368195,
      "rewards/forensics_strict_edge_reward": 0.1157091073691845,
      "rewards/forensics_tiou_reward": 0.6312889456748962,
      "step": 1175
    },
    {
      "completion_length": 25.328125,
      "epoch": 3.608294930875576,
      "grad_norm": 20.83788299560547,
      "kl": 2.22265625,
      "learning_rate": 9.615384615384616e-08,
      "loss": -0.1032,
      "reward": 0.8015322089195251,
      "reward_std": 0.008408830035477877,
      "reward_std_all": 0.009038146119564772,
      "rewards-max/forensics_sinkhorn_reward": 0.9833728969097137,
      "rewards-max/forensics_strict_edge_reward": 0.7742200195789337,
      "rewards-max/forensics_tiou_reward": 0.9851922392845154,
      "rewards/forensics_sinkhorn_reward": 0.7901696264743805,
      "rewards/forensics_strict_edge_reward": 0.09941092878580093,
      "rewards/forensics_tiou_reward": 0.5950891375541687,
      "step": 1176
    },
    {
      "completion_length": 25.65625,
      "epoch": 3.61136712749616,
      "grad_norm": 9.851521492004395,
      "kl": 0.416015625,
      "learning_rate": 9.538461538461538e-08,
      "loss": -0.3237,
      "reward": 0.8049513101577759,
      "reward_std": 0.0022353812819346786,
      "reward_std_all": 0.003953848616220057,
      "rewards-max/forensics_sinkhorn_reward": 0.9811744093894958,
      "rewards-max/forensics_strict_edge_reward": 0.6017688363790512,
      "rewards-max/forensics_tiou_reward": 0.9792801141738892,
      "rewards/forensics_sinkhorn_reward": 0.7857466638088226,
      "rewards/forensics_strict_edge_reward": 0.11952974461019039,
      "rewards/forensics_tiou_reward": 0.6111188232898712,
      "step": 1177
    },
    {
      "completion_length": 26.234375,
      "epoch": 3.6144393241167436,
      "grad_norm": 10.825542449951172,
      "kl": 0.5478515625,
      "learning_rate": 9.461538461538461e-08,
      "loss": -0.1823,
      "reward": 0.7762156128883362,
      "reward_std": 0.01992777269333601,
      "reward_std_all": 0.025454380549490452,
      "rewards-max/forensics_sinkhorn_reward": 0.9826328456401825,
      "rewards-max/forensics_strict_edge_reward": 0.732774555683136,
      "rewards-max/forensics_tiou_reward": 0.983814924955368,
      "rewards/forensics_sinkhorn_reward": 0.7353571951389313,
      "rewards/forensics_strict_edge_reward": 0.10006393119692802,
      "rewards/forensics_tiou_reward": 0.5800060331821442,
      "step": 1178
    },
    {
      "completion_length": 26.90625,
      "epoch": 3.6175115207373274,
      "grad_norm": 10.612237930297852,
      "kl": 0.4990234375,
      "learning_rate": 9.384615384615384e-08,
      "loss": 0.0661,
      "reward": 0.7921084463596344,
      "reward_std": 0.02591097680851817,
      "reward_std_all": 0.023944132030010223,
      "rewards-max/forensics_sinkhorn_reward": 0.9877085089683533,
      "rewards-max/forensics_strict_edge_reward": 0.8815515339374542,
      "rewards-max/forensics_tiou_reward": 0.9956662356853485,
      "rewards/forensics_sinkhorn_reward": 0.7956371307373047,
      "rewards/forensics_strict_edge_reward": 0.1448952853679657,
      "rewards/forensics_tiou_reward": 0.6701377928256989,
      "step": 1179
    },
    {
      "completion_length": 25.140625,
      "epoch": 3.6205837173579107,
      "grad_norm": 10.996864318847656,
      "kl": 0.4453125,
      "learning_rate": 9.307692307692306e-08,
      "loss": -0.2486,
      "reward": 0.7826558947563171,
      "reward_std": 0.014449956361204386,
      "reward_std_all": 0.02130023157224059,
      "rewards-max/forensics_sinkhorn_reward": 0.9860597550868988,
      "rewards-max/forensics_strict_edge_reward": 0.6956532299518585,
      "rewards-max/forensics_tiou_reward": 0.9907786548137665,
      "rewards/forensics_sinkhorn_reward": 0.7891203165054321,
      "rewards/forensics_strict_edge_reward": 0.08869308605790138,
      "rewards/forensics_tiou_reward": 0.6725986897945404,
      "step": 1180
    },
    {
      "completion_length": 24.65625,
      "epoch": 3.6236559139784945,
      "grad_norm": 10.261571884155273,
      "kl": 0.5625,
      "learning_rate": 9.23076923076923e-08,
      "loss": -0.0097,
      "reward": 0.7881880104541779,
      "reward_std": 0.01844406547024846,
      "reward_std_all": 0.021253378596156836,
      "rewards-max/forensics_sinkhorn_reward": 0.9856312870979309,
      "rewards-max/forensics_strict_edge_reward": 0.8381626009941101,
      "rewards-max/forensics_tiou_reward": 0.992166668176651,
      "rewards/forensics_sinkhorn_reward": 0.7910150587558746,
      "rewards/forensics_strict_edge_reward": 0.1627931222319603,
      "rewards/forensics_tiou_reward": 0.6443325877189636,
      "step": 1181
    },
    {
      "completion_length": 23.546875,
      "epoch": 3.6267281105990783,
      "grad_norm": 10.408702850341797,
      "kl": 0.642578125,
      "learning_rate": 9.153846153846153e-08,
      "loss": 0.1172,
      "reward": 0.7701722979545593,
      "reward_std": 0.03316615615040064,
      "reward_std_all": 0.0347753306850791,
      "rewards-max/forensics_sinkhorn_reward": 0.9763854742050171,
      "rewards-max/forensics_strict_edge_reward": 0.6266644895076752,
      "rewards-max/forensics_tiou_reward": 0.9683473706245422,
      "rewards/forensics_sinkhorn_reward": 0.7365450859069824,
      "rewards/forensics_strict_edge_reward": 0.0737402644008398,
      "rewards/forensics_tiou_reward": 0.5480841249227524,
      "step": 1182
    },
    {
      "completion_length": 25.1875,
      "epoch": 3.629800307219662,
      "grad_norm": 11.171539306640625,
      "kl": 0.6474609375,
      "learning_rate": 9.076923076923076e-08,
      "loss": 0.1076,
      "reward": 0.7680054605007172,
      "reward_std": 0.02330250944942236,
      "reward_std_all": 0.03140516905114055,
      "rewards-max/forensics_sinkhorn_reward": 0.9832540452480316,
      "rewards-max/forensics_strict_edge_reward": 0.6824378371238708,
      "rewards-max/forensics_tiou_reward": 0.985252171754837,
      "rewards/forensics_sinkhorn_reward": 0.7488150894641876,
      "rewards/forensics_strict_edge_reward": 0.09590881317853928,
      "rewards/forensics_tiou_reward": 0.5812763869762421,
      "step": 1183
    },
    {
      "completion_length": 27.109375,
      "epoch": 3.632872503840246,
      "grad_norm": 9.145620346069336,
      "kl": 0.5234375,
      "learning_rate": 9e-08,
      "loss": -0.0858,
      "reward": 0.770319253206253,
      "reward_std": 0.03158582840114832,
      "reward_std_all": 0.03482450358569622,
      "rewards-max/forensics_sinkhorn_reward": 0.982109397649765,
      "rewards-max/forensics_strict_edge_reward": 0.526706799864769,
      "rewards-max/forensics_tiou_reward": 0.9826584458351135,
      "rewards/forensics_sinkhorn_reward": 0.7863986194133759,
      "rewards/forensics_strict_edge_reward": 0.09891282767057419,
      "rewards/forensics_tiou_reward": 0.6559024155139923,
      "step": 1184
    },
    {
      "completion_length": 26.734375,
      "epoch": 3.6359447004608296,
      "grad_norm": 10.121840476989746,
      "kl": 0.453125,
      "learning_rate": 8.923076923076923e-08,
      "loss": 0.0536,
      "reward": 0.7597315907478333,
      "reward_std": 0.029322061222046614,
      "reward_std_all": 0.037434171885252,
      "rewards-max/forensics_sinkhorn_reward": 0.9829087853431702,
      "rewards-max/forensics_strict_edge_reward": 0.7165313363075256,
      "rewards-max/forensics_tiou_reward": 0.9839522242546082,
      "rewards/forensics_sinkhorn_reward": 0.7066403329372406,
      "rewards/forensics_strict_edge_reward": 0.09313418716192245,
      "rewards/forensics_tiou_reward": 0.5455128252506256,
      "step": 1185
    },
    {
      "completion_length": 27.234375,
      "epoch": 3.6390168970814134,
      "grad_norm": 8.052621841430664,
      "kl": 0.6171875,
      "learning_rate": 8.846153846153847e-08,
      "loss": -0.062,
      "reward": 0.7513976693153381,
      "reward_std": 0.02681508380919695,
      "reward_std_all": 0.04058771952986717,
      "rewards-max/forensics_sinkhorn_reward": 0.980230063199997,
      "rewards-max/forensics_strict_edge_reward": 0.5602131485939026,
      "rewards-max/forensics_tiou_reward": 0.9785630404949188,
      "rewards/forensics_sinkhorn_reward": 0.7171305119991302,
      "rewards/forensics_strict_edge_reward": 0.10509716533124447,
      "rewards/forensics_tiou_reward": 0.5895744860172272,
      "step": 1186
    },
    {
      "completion_length": 26.0625,
      "epoch": 3.6420890937019967,
      "grad_norm": 9.919598579406738,
      "kl": 0.703125,
      "learning_rate": 8.769230769230769e-08,
      "loss": -0.0055,
      "reward": 0.802119642496109,
      "reward_std": 0.006903582718223333,
      "reward_std_all": 0.008003775030374527,
      "rewards-max/forensics_sinkhorn_reward": 0.9818126559257507,
      "rewards-max/forensics_strict_edge_reward": 0.7169294059276581,
      "rewards-max/forensics_tiou_reward": 0.9818827509880066,
      "rewards/forensics_sinkhorn_reward": 0.78626748919487,
      "rewards/forensics_strict_edge_reward": 0.09732365980744362,
      "rewards/forensics_tiou_reward": 0.612606942653656,
      "step": 1187
    },
    {
      "completion_length": 27.484375,
      "epoch": 3.6451612903225805,
      "grad_norm": 10.808847427368164,
      "kl": 0.5517578125,
      "learning_rate": 8.692307692307692e-08,
      "loss": -0.0764,
      "reward": 0.7682921290397644,
      "reward_std": 0.039010291919112206,
      "reward_std_all": 0.03974681720137596,
      "rewards-max/forensics_sinkhorn_reward": 0.9842977225780487,
      "rewards-max/forensics_strict_edge_reward": 0.5418168008327484,
      "rewards-max/forensics_tiou_reward": 0.9873683154582977,
      "rewards/forensics_sinkhorn_reward": 0.7272457778453827,
      "rewards/forensics_strict_edge_reward": 0.05753055773675442,
      "rewards/forensics_tiou_reward": 0.6189966797828674,
      "step": 1188
    },
    {
      "completion_length": 22.03125,
      "epoch": 3.6482334869431643,
      "grad_norm": 11.94795036315918,
      "kl": 0.4609375,
      "learning_rate": 8.615384615384615e-08,
      "loss": -0.1998,
      "reward": 0.8047245740890503,
      "reward_std": 0.003915810026228428,
      "reward_std_all": 0.004943460691720247,
      "rewards-max/forensics_sinkhorn_reward": 0.9842643737792969,
      "rewards-max/forensics_strict_edge_reward": 0.7958523333072662,
      "rewards-max/forensics_tiou_reward": 0.988928884267807,
      "rewards/forensics_sinkhorn_reward": 0.8211697936058044,
      "rewards/forensics_strict_edge_reward": 0.11959255486726761,
      "rewards/forensics_tiou_reward": 0.6609274744987488,
      "step": 1189
    },
    {
      "completion_length": 21.25,
      "epoch": 3.651305683563748,
      "grad_norm": 13.20295524597168,
      "kl": 0.689453125,
      "learning_rate": 8.538461538461537e-08,
      "loss": 0.0591,
      "reward": 0.7883411049842834,
      "reward_std": 0.025552612729370594,
      "reward_std_all": 0.02474714070558548,
      "rewards-max/forensics_sinkhorn_reward": 0.9816642105579376,
      "rewards-max/forensics_strict_edge_reward": 0.8704091012477875,
      "rewards-max/forensics_tiou_reward": 0.9833942651748657,
      "rewards/forensics_sinkhorn_reward": 0.7563033699989319,
      "rewards/forensics_strict_edge_reward": 0.08761636167764664,
      "rewards/forensics_tiou_reward": 0.5327296853065491,
      "step": 1190
    },
    {
      "completion_length": 23.046875,
      "epoch": 3.654377880184332,
      "grad_norm": 10.264158248901367,
      "kl": 0.833984375,
      "learning_rate": 8.461538461538461e-08,
      "loss": -0.036,
      "reward": 0.768876314163208,
      "reward_std": 0.043090006336569786,
      "reward_std_all": 0.04207564331591129,
      "rewards-max/forensics_sinkhorn_reward": 0.9865289330482483,
      "rewards-max/forensics_strict_edge_reward": 0.7457418143749237,
      "rewards-max/forensics_tiou_reward": 0.9951206743717194,
      "rewards/forensics_sinkhorn_reward": 0.797051340341568,
      "rewards/forensics_strict_edge_reward": 0.14742053300142288,
      "rewards/forensics_tiou_reward": 0.6421802639961243,
      "step": 1191
    },
    {
      "completion_length": 21.984375,
      "epoch": 3.6574500768049156,
      "grad_norm": 14.47128677368164,
      "kl": 0.712890625,
      "learning_rate": 8.384615384615384e-08,
      "loss": 0.0186,
      "reward": 0.7699063420295715,
      "reward_std": 0.029904109425842762,
      "reward_std_all": 0.03448310820385814,
      "rewards-max/forensics_sinkhorn_reward": 0.9859212636947632,
      "rewards-max/forensics_strict_edge_reward": 0.6892678737640381,
      "rewards-max/forensics_tiou_reward": 0.9906706511974335,
      "rewards/forensics_sinkhorn_reward": 0.7419044375419617,
      "rewards/forensics_strict_edge_reward": 0.13322055712342262,
      "rewards/forensics_tiou_reward": 0.5837089121341705,
      "step": 1192
    },
    {
      "completion_length": 26.890625,
      "epoch": 3.6605222734254994,
      "grad_norm": 14.425250053405762,
      "kl": 0.9912109375,
      "learning_rate": 8.307692307692307e-08,
      "loss": 0.1718,
      "reward": 0.7744597494602203,
      "reward_std": 0.04446026869118214,
      "reward_std_all": 0.04227039776742458,
      "rewards-max/forensics_sinkhorn_reward": 0.9827050566673279,
      "rewards-max/forensics_strict_edge_reward": 0.765928328037262,
      "rewards-max/forensics_tiou_reward": 0.9848053455352783,
      "rewards/forensics_sinkhorn_reward": 0.7226692736148834,
      "rewards/forensics_strict_edge_reward": 0.09824217110872269,
      "rewards/forensics_tiou_reward": 0.5874016880989075,
      "step": 1193
    },
    {
      "completion_length": 23.234375,
      "epoch": 3.6635944700460827,
      "grad_norm": 11.264887809753418,
      "kl": 0.791015625,
      "learning_rate": 8.23076923076923e-08,
      "loss": -0.2301,
      "reward": 0.8056602478027344,
      "reward_std": 0.002852122765034437,
      "reward_std_all": 0.003975543542765081,
      "rewards-max/forensics_sinkhorn_reward": 0.985424667596817,
      "rewards-max/forensics_strict_edge_reward": 0.7094893157482147,
      "rewards-max/forensics_tiou_reward": 0.9898569881916046,
      "rewards/forensics_sinkhorn_reward": 0.8106496334075928,
      "rewards/forensics_strict_edge_reward": 0.1675696223974228,
      "rewards/forensics_tiou_reward": 0.6710604727268219,
      "step": 1194
    },
    {
      "completion_length": 20.921875,
      "epoch": 3.6666666666666665,
      "grad_norm": 11.278204917907715,
      "kl": 0.5654296875,
      "learning_rate": 8.153846153846153e-08,
      "loss": -0.0156,
      "reward": 0.7910661101341248,
      "reward_std": 0.012015179847367108,
      "reward_std_all": 0.014926426345482469,
      "rewards-max/forensics_sinkhorn_reward": 0.9862874746322632,
      "rewards-max/forensics_strict_edge_reward": 0.8205508291721344,
      "rewards-max/forensics_tiou_reward": 0.9916915893554688,
      "rewards/forensics_sinkhorn_reward": 0.7469155788421631,
      "rewards/forensics_strict_edge_reward": 0.08015048317611217,
      "rewards/forensics_tiou_reward": 0.5049338340759277,
      "step": 1195
    },
    {
      "completion_length": 22.890625,
      "epoch": 3.6697388632872503,
      "grad_norm": 17.179895401000977,
      "kl": 0.599609375,
      "learning_rate": 8.076923076923077e-08,
      "loss": -0.0776,
      "reward": 0.7640641629695892,
      "reward_std": 0.0492772962898016,
      "reward_std_all": 0.04851076193153858,
      "rewards-max/forensics_sinkhorn_reward": 0.9784730076789856,
      "rewards-max/forensics_strict_edge_reward": 0.7367854118347168,
      "rewards-max/forensics_tiou_reward": 0.975780189037323,
      "rewards/forensics_sinkhorn_reward": 0.7177625298500061,
      "rewards/forensics_strict_edge_reward": 0.0908224955201149,
      "rewards/forensics_tiou_reward": 0.5193332135677338,
      "step": 1196
    },
    {
      "completion_length": 27.046875,
      "epoch": 3.672811059907834,
      "grad_norm": 9.191255569458008,
      "kl": 0.6259765625,
      "learning_rate": 8e-08,
      "loss": 0.0141,
      "reward": 0.7886017262935638,
      "reward_std": 0.020698870066553354,
      "reward_std_all": 0.02123286807909608,
      "rewards-max/forensics_sinkhorn_reward": 0.9859495162963867,
      "rewards-max/forensics_strict_edge_reward": 0.628287136554718,
      "rewards-max/forensics_tiou_reward": 0.9891428351402283,
      "rewards/forensics_sinkhorn_reward": 0.7491470277309418,
      "rewards/forensics_strict_edge_reward": 0.07084036245942116,
      "rewards/forensics_tiou_reward": 0.6022970825433731,
      "step": 1197
    },
    {
      "completion_length": 21.4375,
      "epoch": 3.675883256528418,
      "grad_norm": 11.820900917053223,
      "kl": 0.5263671875,
      "learning_rate": 7.923076923076923e-08,
      "loss": 0.0292,
      "reward": 0.7460386455059052,
      "reward_std": 0.02754370868206024,
      "reward_std_all": 0.04535972699522972,
      "rewards-max/forensics_sinkhorn_reward": 0.9735055565834045,
      "rewards-max/forensics_strict_edge_reward": 0.6032288670539856,
      "rewards-max/forensics_tiou_reward": 0.9622678756713867,
      "rewards/forensics_sinkhorn_reward": 0.698324590921402,
      "rewards/forensics_strict_edge_reward": 0.10532047599554062,
      "rewards/forensics_tiou_reward": 0.516595184803009,
      "step": 1198
    },
    {
      "completion_length": 24.96875,
      "epoch": 3.6789554531490016,
      "grad_norm": 10.970962524414062,
      "kl": 0.69921875,
      "learning_rate": 7.846153846153846e-08,
      "loss": -0.0917,
      "reward": 0.766026496887207,
      "reward_std": 0.050448985770344734,
      "reward_std_all": 0.04761666711419821,
      "rewards-max/forensics_sinkhorn_reward": 0.9827108085155487,
      "rewards-max/forensics_strict_edge_reward": 0.7764008641242981,
      "rewards-max/forensics_tiou_reward": 0.9853371977806091,
      "rewards/forensics_sinkhorn_reward": 0.7928045094013214,
      "rewards/forensics_strict_edge_reward": 0.10468310490250587,
      "rewards/forensics_tiou_reward": 0.6446285545825958,
      "step": 1199
    },
    {
      "completion_length": 23.078125,
      "epoch": 3.6820276497695854,
      "grad_norm": 9.859479904174805,
      "kl": 0.603515625,
      "learning_rate": 7.769230769230768e-08,
      "loss": -0.0129,
      "reward": 0.761137992143631,
      "reward_std": 0.05755997635424137,
      "reward_std_all": 0.05400216579437256,
      "rewards-max/forensics_sinkhorn_reward": 0.9878669083118439,
      "rewards-max/forensics_strict_edge_reward": 0.9232408702373505,
      "rewards-max/forensics_tiou_reward": 0.9965416789054871,
      "rewards/forensics_sinkhorn_reward": 0.7468962371349335,
      "rewards/forensics_strict_edge_reward": 0.17491628974676132,
      "rewards/forensics_tiou_reward": 0.6253988891839981,
      "step": 1200
    },
    {
      "completion_length": 24.859375,
      "epoch": 3.6850998463901687,
      "grad_norm": 11.149054527282715,
      "kl": 0.84765625,
      "learning_rate": 7.692307692307692e-08,
      "loss": 0.1832,
      "reward": 0.7806823253631592,
      "reward_std": 0.01728144148364663,
      "reward_std_all": 0.022998478263616562,
      "rewards-max/forensics_sinkhorn_reward": 0.9723276793956757,
      "rewards-max/forensics_strict_edge_reward": 0.38009122014045715,
      "rewards-max/forensics_tiou_reward": 0.9601394832134247,
      "rewards/forensics_sinkhorn_reward": 0.7566517889499664,
      "rewards/forensics_strict_edge_reward": 0.04654247872531414,
      "rewards/forensics_tiou_reward": 0.6053639054298401,
      "step": 1201
    },
    {
      "completion_length": 22.703125,
      "epoch": 3.688172043010753,
      "grad_norm": 12.415173530578613,
      "kl": 0.685546875,
      "learning_rate": 7.615384615384615e-08,
      "loss": -0.09,
      "reward": 0.7859340310096741,
      "reward_std": 0.0250631021335721,
      "reward_std_all": 0.02483724197372794,
      "rewards-max/forensics_sinkhorn_reward": 0.9824741780757904,
      "rewards-max/forensics_strict_edge_reward": 0.7633048295974731,
      "rewards-max/forensics_tiou_reward": 0.9826139807701111,
      "rewards/forensics_sinkhorn_reward": 0.7774378657341003,
      "rewards/forensics_strict_edge_reward": 0.09373888000845909,
      "rewards/forensics_tiou_reward": 0.5929753184318542,
      "step": 1202
    },
    {
      "completion_length": 26.84375,
      "epoch": 3.6912442396313363,
      "grad_norm": 11.188461303710938,
      "kl": 0.611328125,
      "learning_rate": 7.538461538461538e-08,
      "loss": -0.1425,
      "reward": 0.8010669350624084,
      "reward_std": 0.009041228564456105,
      "reward_std_all": 0.009556756354868412,
      "rewards-max/forensics_sinkhorn_reward": 0.9753184616565704,
      "rewards-max/forensics_strict_edge_reward": 0.6824378371238708,
      "rewards-max/forensics_tiou_reward": 0.9679158627986908,
      "rewards/forensics_sinkhorn_reward": 0.8072380423545837,
      "rewards/forensics_strict_edge_reward": 0.14998320490121841,
      "rewards/forensics_tiou_reward": 0.646736741065979,
      "step": 1203
    },
    {
      "completion_length": 25.140625,
      "epoch": 3.69431643625192,
      "grad_norm": 11.001903533935547,
      "kl": 0.65234375,
      "learning_rate": 7.46153846153846e-08,
      "loss": 0.0596,
      "reward": 0.7724158465862274,
      "reward_std": 0.029543448239564896,
      "reward_std_all": 0.03177380934357643,
      "rewards-max/forensics_sinkhorn_reward": 0.9773048758506775,
      "rewards-max/forensics_strict_edge_reward": 0.3743867874145508,
      "rewards-max/forensics_tiou_reward": 0.9686567485332489,
      "rewards/forensics_sinkhorn_reward": 0.7250184416770935,
      "rewards/forensics_strict_edge_reward": 0.04302556626498699,
      "rewards/forensics_tiou_reward": 0.5046547949314117,
      "step": 1204
    },
    {
      "completion_length": 21.890625,
      "epoch": 3.697388632872504,
      "grad_norm": 10.65434741973877,
      "kl": 0.875,
      "learning_rate": 7.384615384615384e-08,
      "loss": -0.1741,
      "reward": 0.7992790937423706,
      "reward_std": 0.010560242109932005,
      "reward_std_all": 0.010919296881183982,
      "rewards-max/forensics_sinkhorn_reward": 0.9889746308326721,
      "rewards-max/forensics_strict_edge_reward": 0.9836080372333527,
      "rewards-max/forensics_tiou_reward": 0.9999392628669739,
      "rewards/forensics_sinkhorn_reward": 0.8388153910636902,
      "rewards/forensics_strict_edge_reward": 0.22388283908367157,
      "rewards/forensics_tiou_reward": 0.6949939429759979,
      "step": 1205
    },
    {
      "completion_length": 26.40625,
      "epoch": 3.7004608294930876,
      "grad_norm": 10.739601135253906,
      "kl": 0.54296875,
      "learning_rate": 7.307692307692308e-08,
      "loss": -0.1026,
      "reward": 0.7845079004764557,
      "reward_std": 0.03939043264836073,
      "reward_std_all": 0.03534961771219969,
      "rewards-max/forensics_sinkhorn_reward": 0.9802680909633636,
      "rewards-max/forensics_strict_edge_reward": 0.49020934104919434,
      "rewards-max/forensics_tiou_reward": 0.9707117676734924,
      "rewards/forensics_sinkhorn_reward": 0.8175216615200043,
      "rewards/forensics_strict_edge_reward": 0.09936540201306343,
      "rewards/forensics_tiou_reward": 0.6932507157325745,
      "step": 1206
    },
    {
      "completion_length": 25.359375,
      "epoch": 3.7035330261136714,
      "grad_norm": 9.660609245300293,
      "kl": 0.72265625,
      "learning_rate": 7.230769230769231e-08,
      "loss": -0.0044,
      "reward": 0.7944458723068237,
      "reward_std": 0.019119855016469955,
      "reward_std_all": 0.018437260761857033,
      "rewards-max/forensics_sinkhorn_reward": 0.9808711111545563,
      "rewards-max/forensics_strict_edge_reward": 0.6614228188991547,
      "rewards-max/forensics_tiou_reward": 0.9764210879802704,
      "rewards/forensics_sinkhorn_reward": 0.7933460772037506,
      "rewards/forensics_strict_edge_reward": 0.11101464927196503,
      "rewards/forensics_tiou_reward": 0.6118631958961487,
      "step": 1207
    },
    {
      "completion_length": 26.265625,
      "epoch": 3.7066052227342547,
      "grad_norm": 22.546728134155273,
      "kl": 0.673828125,
      "learning_rate": 7.153846153846154e-08,
      "loss": 0.1392,
      "reward": 0.7435393035411835,
      "reward_std": 0.009100571740418673,
      "reward_std_all": 0.03592936135828495,
      "rewards-max/forensics_sinkhorn_reward": 0.9840276837348938,
      "rewards-max/forensics_strict_edge_reward": 0.7463764250278473,
      "rewards-max/forensics_tiou_reward": 0.9858288466930389,
      "rewards/forensics_sinkhorn_reward": 0.7035264074802399,
      "rewards/forensics_strict_edge_reward": 0.05238107591867447,
      "rewards/forensics_tiou_reward": 0.5247770249843597,
      "step": 1208
    },
    {
      "completion_length": 23.984375,
      "epoch": 3.709677419354839,
      "grad_norm": 13.03278923034668,
      "kl": 0.583984375,
      "learning_rate": 7.076923076923076e-08,
      "loss": -0.0337,
      "reward": 0.7621457278728485,
      "reward_std": 0.0457107899710536,
      "reward_std_all": 0.04592924099415541,
      "rewards-max/forensics_sinkhorn_reward": 0.9830855131149292,
      "rewards-max/forensics_strict_edge_reward": 0.756708562374115,
      "rewards-max/forensics_tiou_reward": 0.9836075007915497,
      "rewards/forensics_sinkhorn_reward": 0.7414904236793518,
      "rewards/forensics_strict_edge_reward": 0.14914686977863312,
      "rewards/forensics_tiou_reward": 0.6101846396923065,
      "step": 1209
    },
    {
      "completion_length": 27.765625,
      "epoch": 3.7127496159754223,
      "grad_norm": 10.653029441833496,
      "kl": 0.5224609375,
      "learning_rate": 7e-08,
      "loss": 0.1674,
      "reward": 0.7635748982429504,
      "reward_std": 0.03520748484879732,
      "reward_std_all": 0.040480537340044975,
      "rewards-max/forensics_sinkhorn_reward": 0.9812825918197632,
      "rewards-max/forensics_strict_edge_reward": 0.5912453234195709,
      "rewards-max/forensics_tiou_reward": 0.9793242514133453,
      "rewards/forensics_sinkhorn_reward": 0.716322124004364,
      "rewards/forensics_strict_edge_reward": 0.08866792917251587,
      "rewards/forensics_tiou_reward": 0.5763500928878784,
      "step": 1210
    },
    {
      "completion_length": 23.578125,
      "epoch": 3.715821812596006,
      "grad_norm": 10.688175201416016,
      "kl": 0.58984375,
      "learning_rate": 6.923076923076923e-08,
      "loss": -0.0638,
      "reward": 0.8026719689369202,
      "reward_std": 0.006324731628410518,
      "reward_std_all": 0.007408662233501673,
      "rewards-max/forensics_sinkhorn_reward": 0.9849720895290375,
      "rewards-max/forensics_strict_edge_reward": 0.5829301476478577,
      "rewards-max/forensics_tiou_reward": 0.9898042380809784,
      "rewards/forensics_sinkhorn_reward": 0.806479424238205,
      "rewards/forensics_strict_edge_reward": 0.07346051931381226,
      "rewards/forensics_tiou_reward": 0.6489940881729126,
      "step": 1211
    },
    {
      "completion_length": 27.046875,
      "epoch": 3.71889400921659,
      "grad_norm": 10.602540016174316,
      "kl": 0.525390625,
      "learning_rate": 6.846153846153846e-08,
      "loss": -0.2299,
      "reward": 0.8021904528141022,
      "reward_std": 0.006651073694229126,
      "reward_std_all": 0.007775084115564823,
      "rewards-max/forensics_sinkhorn_reward": 0.9732062518596649,
      "rewards-max/forensics_strict_edge_reward": 0.564324826002121,
      "rewards-max/forensics_tiou_reward": 0.9618920087814331,
      "rewards/forensics_sinkhorn_reward": 0.8008656203746796,
      "rewards/forensics_strict_edge_reward": 0.06096137315034866,
      "rewards/forensics_tiou_reward": 0.6289644837379456,
      "step": 1212
    },
    {
      "completion_length": 26.8125,
      "epoch": 3.7219662058371736,
      "grad_norm": 13.522607803344727,
      "kl": 0.548828125,
      "learning_rate": 6.769230769230769e-08,
      "loss": -0.2169,
      "reward": 0.774746835231781,
      "reward_std": 0.014309513848274946,
      "reward_std_all": 0.02206583833321929,
      "rewards-max/forensics_sinkhorn_reward": 0.9864881932735443,
      "rewards-max/forensics_strict_edge_reward": 0.8582656681537628,
      "rewards-max/forensics_tiou_reward": 0.9947072267532349,
      "rewards/forensics_sinkhorn_reward": 0.7813499271869659,
      "rewards/forensics_strict_edge_reward": 0.13779567927122116,
      "rewards/forensics_tiou_reward": 0.6125231385231018,
      "step": 1213
    },
    {
      "completion_length": 24.6875,
      "epoch": 3.7250384024577574,
      "grad_norm": 12.10176944732666,
      "kl": 0.6015625,
      "learning_rate": 6.692307692307691e-08,
      "loss": 0.0802,
      "reward": 0.7519065141677856,
      "reward_std": 0.02385304542258382,
      "reward_std_all": 0.04041469842195511,
      "rewards-max/forensics_sinkhorn_reward": 0.9306767880916595,
      "rewards-max/forensics_strict_edge_reward": 0.47269192337989807,
      "rewards-max/forensics_tiou_reward": 0.8789489269256592,
      "rewards/forensics_sinkhorn_reward": 0.6980071663856506,
      "rewards/forensics_strict_edge_reward": 0.04163155518472195,
      "rewards/forensics_tiou_reward": 0.49445223808288574,
      "step": 1214
    },
    {
      "completion_length": 24.1875,
      "epoch": 3.7281105990783407,
      "grad_norm": 17.317428588867188,
      "kl": 0.6328125,
      "learning_rate": 6.615384615384615e-08,
      "loss": -0.1484,
      "reward": 0.7980554401874542,
      "reward_std": 0.00904661393724382,
      "reward_std_all": 0.01077157526742667,
      "rewards-max/forensics_sinkhorn_reward": 0.9846117496490479,
      "rewards-max/forensics_strict_edge_reward": 0.6149742305278778,
      "rewards-max/forensics_tiou_reward": 0.9871382415294647,
      "rewards/forensics_sinkhorn_reward": 0.8227113485336304,
      "rewards/forensics_strict_edge_reward": 0.17072997987270355,
      "rewards/forensics_tiou_reward": 0.7086315453052521,
      "step": 1215
    },
    {
      "completion_length": 24.109375,
      "epoch": 3.731182795698925,
      "grad_norm": 10.546660423278809,
      "kl": 0.646484375,
      "learning_rate": 6.538461538461538e-08,
      "loss": 0.4216,
      "reward": 0.7456262707710266,
      "reward_std": 0.06889089941978455,
      "reward_std_all": 0.06609701178967953,
      "rewards-max/forensics_sinkhorn_reward": 0.9872556626796722,
      "rewards-max/forensics_strict_edge_reward": 0.8665722012519836,
      "rewards-max/forensics_tiou_reward": 0.9941936731338501,
      "rewards/forensics_sinkhorn_reward": 0.6913794279098511,
      "rewards/forensics_strict_edge_reward": 0.09853970259428024,
      "rewards/forensics_tiou_reward": 0.52189801633358,
      "step": 1216
    },
    {
      "completion_length": 28.53125,
      "epoch": 3.7342549923195083,
      "grad_norm": 10.405302047729492,
      "kl": 0.876953125,
      "learning_rate": 6.461538461538462e-08,
      "loss": 0.1538,
      "reward": 0.7303398251533508,
      "reward_std": 0.05503409914672375,
      "reward_std_all": 0.06894526258111,
      "rewards-max/forensics_sinkhorn_reward": 0.9582386016845703,
      "rewards-max/forensics_strict_edge_reward": 0.5383955463767052,
      "rewards-max/forensics_tiou_reward": 0.9279992580413818,
      "rewards/forensics_sinkhorn_reward": 0.7279361486434937,
      "rewards/forensics_strict_edge_reward": 0.0859765699133277,
      "rewards/forensics_tiou_reward": 0.5768081247806549,
      "step": 1217
    },
    {
      "completion_length": 23.5625,
      "epoch": 3.737327188940092,
      "grad_norm": 12.65660572052002,
      "kl": 0.57421875,
      "learning_rate": 6.384615384615385e-08,
      "loss": -0.0853,
      "reward": 0.7995316982269287,
      "reward_std": 0.013517942279577255,
      "reward_std_all": 0.013154711574316025,
      "rewards-max/forensics_sinkhorn_reward": 0.9764183163642883,
      "rewards-max/forensics_strict_edge_reward": 0.6718102097511292,
      "rewards-max/forensics_tiou_reward": 0.968782514333725,
      "rewards/forensics_sinkhorn_reward": 0.855667769908905,
      "rewards/forensics_strict_edge_reward": 0.11621984094381332,
      "rewards/forensics_tiou_reward": 0.7173195481300354,
      "step": 1218
    },
    {
      "completion_length": 22.21875,
      "epoch": 3.740399385560676,
      "grad_norm": 11.671953201293945,
      "kl": 0.66796875,
      "learning_rate": 6.307692307692307e-08,
      "loss": 0.024,
      "reward": 0.7800849378108978,
      "reward_std": 0.03811026271432638,
      "reward_std_all": 0.035750724375247955,
      "rewards-max/forensics_sinkhorn_reward": 0.9684586822986603,
      "rewards-max/forensics_strict_edge_reward": 0.655701607465744,
      "rewards-max/forensics_tiou_reward": 0.9516473114490509,
      "rewards/forensics_sinkhorn_reward": 0.7405473291873932,
      "rewards/forensics_strict_edge_reward": 0.08453232701867819,
      "rewards/forensics_tiou_reward": 0.528153195977211,
      "step": 1219
    },
    {
      "completion_length": 22.265625,
      "epoch": 3.7434715821812596,
      "grad_norm": 12.300813674926758,
      "kl": 0.5537109375,
      "learning_rate": 6.23076923076923e-08,
      "loss": -0.0009,
      "reward": 0.7891476154327393,
      "reward_std": 0.02362584974616766,
      "reward_std_all": 0.02285848930478096,
      "rewards-max/forensics_sinkhorn_reward": 0.9787154197692871,
      "rewards-max/forensics_strict_edge_reward": 0.6824378371238708,
      "rewards-max/forensics_tiou_reward": 0.9749713838100433,
      "rewards/forensics_sinkhorn_reward": 0.8044280111789703,
      "rewards/forensics_strict_edge_reward": 0.09189985319972038,
      "rewards/forensics_tiou_reward": 0.6209274530410767,
      "step": 1220
    },
    {
      "completion_length": 26.625,
      "epoch": 3.7465437788018434,
      "grad_norm": 11.178421974182129,
      "kl": 0.56640625,
      "learning_rate": 6.153846153846154e-08,
      "loss": 0.1495,
      "reward": 0.7573047876358032,
      "reward_std": 0.051959795877337456,
      "reward_std_all": 0.0530826598405838,
      "rewards-max/forensics_sinkhorn_reward": 0.9866355061531067,
      "rewards-max/forensics_strict_edge_reward": 0.8900053799152374,
      "rewards-max/forensics_tiou_reward": 0.9938836693763733,
      "rewards/forensics_sinkhorn_reward": 0.7007167041301727,
      "rewards/forensics_strict_edge_reward": 0.10173965245485306,
      "rewards/forensics_tiou_reward": 0.5398591160774231,
      "step": 1221
    },
    {
      "completion_length": 23.0,
      "epoch": 3.749615975422427,
      "grad_norm": 11.57309627532959,
      "kl": 0.642578125,
      "learning_rate": 6.076923076923077e-08,
      "loss": 0.0534,
      "reward": 0.752664715051651,
      "reward_std": 0.042076073004864156,
      "reward_std_all": 0.0528609249740839,
      "rewards-max/forensics_sinkhorn_reward": 0.9833377003669739,
      "rewards-max/forensics_strict_edge_reward": 0.6942952275276184,
      "rewards-max/forensics_tiou_reward": 0.9854368567466736,
      "rewards/forensics_sinkhorn_reward": 0.6425814926624298,
      "rewards/forensics_strict_edge_reward": 0.08334452658891678,
      "rewards/forensics_tiou_reward": 0.5353517234325409,
      "step": 1222
    },
    {
      "completion_length": 24.453125,
      "epoch": 3.752688172043011,
      "grad_norm": 12.933307647705078,
      "kl": 0.61328125,
      "learning_rate": 6e-08,
      "loss": 0.2271,
      "reward": 0.7735418081283569,
      "reward_std": 0.038177161128260195,
      "reward_std_all": 0.03730515576899052,
      "rewards-max/forensics_sinkhorn_reward": 0.97759810090065,
      "rewards-max/forensics_strict_edge_reward": 0.4673481583595276,
      "rewards-max/forensics_tiou_reward": 0.972232848405838,
      "rewards/forensics_sinkhorn_reward": 0.6891953349113464,
      "rewards/forensics_strict_edge_reward": 0.0490314494818449,
      "rewards/forensics_tiou_reward": 0.49550817906856537,
      "step": 1223
    },
    {
      "completion_length": 22.75,
      "epoch": 3.7557603686635943,
      "grad_norm": 10.346835136413574,
      "kl": 0.4794921875,
      "learning_rate": 5.923076923076923e-08,
      "loss": 0.0169,
      "reward": 0.7994271516799927,
      "reward_std": 0.012191622867248952,
      "reward_std_all": 0.01246849587187171,
      "rewards-max/forensics_sinkhorn_reward": 0.9880876541137695,
      "rewards-max/forensics_strict_edge_reward": 0.9836080372333527,
      "rewards-max/forensics_tiou_reward": 0.9972281754016876,
      "rewards/forensics_sinkhorn_reward": 0.8098210692405701,
      "rewards/forensics_strict_edge_reward": 0.1184397041797638,
      "rewards/forensics_tiou_reward": 0.6227383017539978,
      "step": 1224
    },
    {
      "completion_length": 24.984375,
      "epoch": 3.758832565284178,
      "grad_norm": 10.025264739990234,
      "kl": 0.5205078125,
      "learning_rate": 5.8461538461538455e-08,
      "loss": 0.0881,
      "reward": 0.7427646815776825,
      "reward_std": 0.04913514852523804,
      "reward_std_all": 0.05893614701926708,
      "rewards-max/forensics_sinkhorn_reward": 0.9846182763576508,
      "rewards-max/forensics_strict_edge_reward": 0.6887466907501221,
      "rewards-max/forensics_tiou_reward": 0.9868148267269135,
      "rewards/forensics_sinkhorn_reward": 0.7614041268825531,
      "rewards/forensics_strict_edge_reward": 0.1272197887301445,
      "rewards/forensics_tiou_reward": 0.6290083825588226,
      "step": 1225
    },
    {
      "completion_length": 24.609375,
      "epoch": 3.761904761904762,
      "grad_norm": 13.021565437316895,
      "kl": 0.583984375,
      "learning_rate": 5.7692307692307695e-08,
      "loss": -0.1319,
      "reward": 0.7725852727890015,
      "reward_std": 0.03228342765942216,
      "reward_std_all": 0.03339871345087886,
      "rewards-max/forensics_sinkhorn_reward": 0.9835976362228394,
      "rewards-max/forensics_strict_edge_reward": 0.7676310539245605,
      "rewards-max/forensics_tiou_reward": 0.9862600564956665,
      "rewards/forensics_sinkhorn_reward": 0.7788569927215576,
      "rewards/forensics_strict_edge_reward": 0.1243884339928627,
      "rewards/forensics_tiou_reward": 0.5882176160812378,
      "step": 1226
    },
    {
      "completion_length": 26.53125,
      "epoch": 3.7649769585253456,
      "grad_norm": 12.142919540405273,
      "kl": 0.5546875,
      "learning_rate": 5.692307692307692e-08,
      "loss": -0.1968,
      "reward": 0.773637980222702,
      "reward_std": 0.034037006087601185,
      "reward_std_all": 0.03428700007498264,
      "rewards-max/forensics_sinkhorn_reward": 0.9833586513996124,
      "rewards-max/forensics_strict_edge_reward": 0.8418737053871155,
      "rewards-max/forensics_tiou_reward": 0.9860914349555969,
      "rewards/forensics_sinkhorn_reward": 0.7300222516059875,
      "rewards/forensics_strict_edge_reward": 0.1560550555586815,
      "rewards/forensics_tiou_reward": 0.5289022624492645,
      "step": 1227
    },
    {
      "completion_length": 25.484375,
      "epoch": 3.7680491551459294,
      "grad_norm": 12.67621898651123,
      "kl": 0.4150390625,
      "learning_rate": 5.615384615384615e-08,
      "loss": -0.1827,
      "reward": 0.7847466170787811,
      "reward_std": 0.0207898038206622,
      "reward_std_all": 0.022356103057973087,
      "rewards-max/forensics_sinkhorn_reward": 0.9844562411308289,
      "rewards-max/forensics_strict_edge_reward": 0.6657527685165405,
      "rewards-max/forensics_tiou_reward": 0.9877166152000427,
      "rewards/forensics_sinkhorn_reward": 0.7680808305740356,
      "rewards/forensics_strict_edge_reward": 0.09877309016883373,
      "rewards/forensics_tiou_reward": 0.6510845720767975,
      "step": 1228
    },
    {
      "completion_length": 25.1875,
      "epoch": 3.771121351766513,
      "grad_norm": 15.094047546386719,
      "kl": 0.490234375,
      "learning_rate": 5.538461538461538e-08,
      "loss": -0.0104,
      "reward": 0.7876458466053009,
      "reward_std": 0.016491294838488102,
      "reward_std_all": 0.019150400534272194,
      "rewards-max/forensics_sinkhorn_reward": 0.9753582179546356,
      "rewards-max/forensics_strict_edge_reward": 0.4623851925134659,
      "rewards-max/forensics_tiou_reward": 0.9649753570556641,
      "rewards/forensics_sinkhorn_reward": 0.7386622130870819,
      "rewards/forensics_strict_edge_reward": 0.075182287953794,
      "rewards/forensics_tiou_reward": 0.5398685783147812,
      "step": 1229
    },
    {
      "completion_length": 23.265625,
      "epoch": 3.774193548387097,
      "grad_norm": 13.85725212097168,
      "kl": 0.5595703125,
      "learning_rate": 5.461538461538461e-08,
      "loss": 0.0184,
      "reward": 0.78148552775383,
      "reward_std": 0.030857632867991924,
      "reward_std_all": 0.029749950394034386,
      "rewards-max/forensics_sinkhorn_reward": 0.982486218214035,
      "rewards-max/forensics_strict_edge_reward": 0.6667979955673218,
      "rewards-max/forensics_tiou_reward": 0.9811222553253174,
      "rewards/forensics_sinkhorn_reward": 0.7712134122848511,
      "rewards/forensics_strict_edge_reward": 0.1121770590543747,
      "rewards/forensics_tiou_reward": 0.6279441714286804,
      "step": 1230
    },
    {
      "completion_length": 26.078125,
      "epoch": 3.7772657450076803,
      "grad_norm": 11.288777351379395,
      "kl": 0.6796875,
      "learning_rate": 5.384615384615385e-08,
      "loss": -0.2556,
      "reward": 0.7941528558731079,
      "reward_std": 0.014475739444606006,
      "reward_std_all": 0.015242729103192687,
      "rewards-max/forensics_sinkhorn_reward": 0.9852606654167175,
      "rewards-max/forensics_strict_edge_reward": 0.6594520807266235,
      "rewards-max/forensics_tiou_reward": 0.9905031323432922,
      "rewards/forensics_sinkhorn_reward": 0.7510533928871155,
      "rewards/forensics_strict_edge_reward": 0.09806392714381218,
      "rewards/forensics_tiou_reward": 0.5970051884651184,
      "step": 1231
    },
    {
      "completion_length": 22.15625,
      "epoch": 3.780337941628264,
      "grad_norm": 14.120040893554688,
      "kl": 0.697265625,
      "learning_rate": 5.307692307692308e-08,
      "loss": 0.113,
      "reward": 0.7676323056221008,
      "reward_std": 0.05284670181572437,
      "reward_std_all": 0.04937346279621124,
      "rewards-max/forensics_sinkhorn_reward": 0.9719157516956329,
      "rewards-max/forensics_strict_edge_reward": 0.6326714754104614,
      "rewards-max/forensics_tiou_reward": 0.9596826434135437,
      "rewards/forensics_sinkhorn_reward": 0.7310056388378143,
      "rewards/forensics_strict_edge_reward": 0.0616266094148159,
      "rewards/forensics_tiou_reward": 0.5118855237960815,
      "step": 1232
    },
    {
      "completion_length": 26.3125,
      "epoch": 3.783410138248848,
      "grad_norm": 9.136990547180176,
      "kl": 0.5390625,
      "learning_rate": 5.2307692307692304e-08,
      "loss": -0.2024,
      "reward": 0.7712057828903198,
      "reward_std": 0.020834566093981266,
      "reward_std_all": 0.02831408753991127,
      "rewards-max/forensics_sinkhorn_reward": 0.9852073490619659,
      "rewards-max/forensics_strict_edge_reward": 0.7815065383911133,
      "rewards-max/forensics_tiou_reward": 0.9887486100196838,
      "rewards/forensics_sinkhorn_reward": 0.7677842974662781,
      "rewards/forensics_strict_edge_reward": 0.1660018116235733,
      "rewards/forensics_tiou_reward": 0.6181781888008118,
      "step": 1233
    },
    {
      "completion_length": 22.03125,
      "epoch": 3.7864823348694316,
      "grad_norm": 13.782408714294434,
      "kl": 0.7265625,
      "learning_rate": 5.153846153846154e-08,
      "loss": -0.0381,
      "reward": 0.7933293282985687,
      "reward_std": 0.019574596313759685,
      "reward_std_all": 0.019397394615225494,
      "rewards-max/forensics_sinkhorn_reward": 0.9852294623851776,
      "rewards-max/forensics_strict_edge_reward": 0.7533732652664185,
      "rewards-max/forensics_tiou_reward": 0.9880926311016083,
      "rewards/forensics_sinkhorn_reward": 0.7811064124107361,
      "rewards/forensics_strict_edge_reward": 0.10581278428435326,
      "rewards/forensics_tiou_reward": 0.5868263244628906,
      "step": 1234
    },
    {
      "completion_length": 25.9375,
      "epoch": 3.7895545314900154,
      "grad_norm": 9.879639625549316,
      "kl": 0.591796875,
      "learning_rate": 5.0769230769230764e-08,
      "loss": 0.0525,
      "reward": 0.7481130957603455,
      "reward_std": 0.050862740725278854,
      "reward_std_all": 0.05412804335355759,
      "rewards-max/forensics_sinkhorn_reward": 0.9841752350330353,
      "rewards-max/forensics_strict_edge_reward": 0.7808269262313843,
      "rewards-max/forensics_tiou_reward": 0.9854840636253357,
      "rewards/forensics_sinkhorn_reward": 0.7494676113128662,
      "rewards/forensics_strict_edge_reward": 0.11801858991384506,
      "rewards/forensics_tiou_reward": 0.6102273762226105,
      "step": 1235
    },
    {
      "completion_length": 25.359375,
      "epoch": 3.792626728110599,
      "grad_norm": 13.126999855041504,
      "kl": 0.5888671875,
      "learning_rate": 5e-08,
      "loss": -0.1375,
      "reward": 0.7975570559501648,
      "reward_std": 0.008112905779853463,
      "reward_std_all": 0.010415361728519201,
      "rewards-max/forensics_sinkhorn_reward": 0.9817446172237396,
      "rewards-max/forensics_strict_edge_reward": 0.709127277135849,
      "rewards-max/forensics_tiou_reward": 0.9818802177906036,
      "rewards/forensics_sinkhorn_reward": 0.7618444561958313,
      "rewards/forensics_strict_edge_reward": 0.11346814036369324,
      "rewards/forensics_tiou_reward": 0.5977079570293427,
      "step": 1236
    },
    {
      "completion_length": 23.140625,
      "epoch": 3.795698924731183,
      "grad_norm": 13.142117500305176,
      "kl": 0.69140625,
      "learning_rate": 4.923076923076923e-08,
      "loss": 0.1184,
      "reward": 0.7747827172279358,
      "reward_std": 0.039272975525818765,
      "reward_std_all": 0.03795405523851514,
      "rewards-max/forensics_sinkhorn_reward": 0.9811975061893463,
      "rewards-max/forensics_strict_edge_reward": 0.6005697548389435,
      "rewards-max/forensics_tiou_reward": 0.9784137010574341,
      "rewards/forensics_sinkhorn_reward": 0.7533747553825378,
      "rewards/forensics_strict_edge_reward": 0.08241704851388931,
      "rewards/forensics_tiou_reward": 0.6042689085006714,
      "step": 1237
    },
    {
      "completion_length": 25.78125,
      "epoch": 3.7987711213517663,
      "grad_norm": 9.358445167541504,
      "kl": 0.806640625,
      "learning_rate": 4.846153846153846e-08,
      "loss": -0.0277,
      "reward": 0.7717710733413696,
      "reward_std": 0.009159274864941835,
      "reward_std_all": 0.02242556679993868,
      "rewards-max/forensics_sinkhorn_reward": 0.9858210980892181,
      "rewards-max/forensics_strict_edge_reward": 0.5422783046960831,
      "rewards-max/forensics_tiou_reward": 0.9915700852870941,
      "rewards/forensics_sinkhorn_reward": 0.7576491236686707,
      "rewards/forensics_strict_edge_reward": 0.09411047771573067,
      "rewards/forensics_tiou_reward": 0.6355483829975128,
      "step": 1238
    },
    {
      "completion_length": 27.046875,
      "epoch": 3.80184331797235,
      "grad_norm": 9.783400535583496,
      "kl": 0.607421875,
      "learning_rate": 4.769230769230769e-08,
      "loss": 0.0737,
      "reward": 0.7534025311470032,
      "reward_std": 0.05386997386813164,
      "reward_std_all": 0.05354726314544678,
      "rewards-max/forensics_sinkhorn_reward": 0.9670776724815369,
      "rewards-max/forensics_strict_edge_reward": 0.46148188412189484,
      "rewards-max/forensics_tiou_reward": 0.9480542242527008,
      "rewards/forensics_sinkhorn_reward": 0.6833666861057281,
      "rewards/forensics_strict_edge_reward": 0.061339810490608215,
      "rewards/forensics_tiou_reward": 0.4856227934360504,
      "step": 1239
    },
    {
      "completion_length": 21.953125,
      "epoch": 3.804915514592934,
      "grad_norm": 12.793787002563477,
      "kl": 0.4873046875,
      "learning_rate": 4.692307692307692e-08,
      "loss": 0.2317,
      "reward": 0.7773044109344482,
      "reward_std": 0.018871178850531578,
      "reward_std_all": 0.026265766471624374,
      "rewards-max/forensics_sinkhorn_reward": 0.9694303572177887,
      "rewards-max/forensics_strict_edge_reward": 0.35114636272192,
      "rewards-max/forensics_tiou_reward": 0.9516182839870453,
      "rewards/forensics_sinkhorn_reward": 0.7604993879795074,
      "rewards/forensics_strict_edge_reward": 0.05962579697370529,
      "rewards/forensics_tiou_reward": 0.5709611177444458,
      "step": 1240
    },
    {
      "completion_length": 24.09375,
      "epoch": 3.8079877112135176,
      "grad_norm": 13.454398155212402,
      "kl": 0.529296875,
      "learning_rate": 4.615384615384615e-08,
      "loss": -0.0981,
      "reward": 0.7606439888477325,
      "reward_std": 0.02139253099448979,
      "reward_std_all": 0.03488973784260452,
      "rewards-max/forensics_sinkhorn_reward": 0.9754550158977509,
      "rewards-max/forensics_strict_edge_reward": 0.5434179902076721,
      "rewards-max/forensics_tiou_reward": 0.9637704491615295,
      "rewards/forensics_sinkhorn_reward": 0.7820069491863251,
      "rewards/forensics_strict_edge_reward": 0.08913188427686691,
      "rewards/forensics_tiou_reward": 0.6303825676441193,
      "step": 1241
    },
    {
      "completion_length": 22.0,
      "epoch": 3.8110599078341014,
      "grad_norm": 13.78259563446045,
      "kl": 0.447265625,
      "learning_rate": 4.538461538461538e-08,
      "loss": -0.3576,
      "reward": 0.7831234335899353,
      "reward_std": 0.00773822411429137,
      "reward_std_all": 0.017646667547523975,
      "rewards-max/forensics_sinkhorn_reward": 0.9763084352016449,
      "rewards-max/forensics_strict_edge_reward": 0.5404154062271118,
      "rewards-max/forensics_tiou_reward": 0.9647824168205261,
      "rewards/forensics_sinkhorn_reward": 0.7954119741916656,
      "rewards/forensics_strict_edge_reward": 0.11826788261532784,
      "rewards/forensics_tiou_reward": 0.6139970123767853,
      "step": 1242
    },
    {
      "completion_length": 26.828125,
      "epoch": 3.814132104454685,
      "grad_norm": 17.11556053161621,
      "kl": 0.89453125,
      "learning_rate": 4.461538461538461e-08,
      "loss": 0.1264,
      "reward": 0.7725359201431274,
      "reward_std": 0.03139256872236729,
      "reward_std_all": 0.03455863147974014,
      "rewards-max/forensics_sinkhorn_reward": 0.9889904856681824,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.7233455777168274,
      "rewards/forensics_strict_edge_reward": 0.1484699733555317,
      "rewards/forensics_tiou_reward": 0.6017899513244629,
      "step": 1243
    },
    {
      "completion_length": 25.703125,
      "epoch": 3.817204301075269,
      "grad_norm": 10.451796531677246,
      "kl": 0.4638671875,
      "learning_rate": 4.3846153846153846e-08,
      "loss": -0.0363,
      "reward": 0.7443723082542419,
      "reward_std": 0.02602127834688872,
      "reward_std_all": 0.043795882258564234,
      "rewards-max/forensics_sinkhorn_reward": 0.982578694820404,
      "rewards-max/forensics_strict_edge_reward": 0.6930406093597412,
      "rewards-max/forensics_tiou_reward": 0.9819204807281494,
      "rewards/forensics_sinkhorn_reward": 0.7082009613513947,
      "rewards/forensics_strict_edge_reward": 0.10055913403630257,
      "rewards/forensics_tiou_reward": 0.5498573184013367,
      "step": 1244
    },
    {
      "completion_length": 30.265625,
      "epoch": 3.8202764976958523,
      "grad_norm": 12.476286888122559,
      "kl": 0.4970703125,
      "learning_rate": 4.307692307692307e-08,
      "loss": -0.1794,
      "reward": 0.7733106017112732,
      "reward_std": 0.031717706413473934,
      "reward_std_all": 0.035365132498554885,
      "rewards-max/forensics_sinkhorn_reward": 0.9735166728496552,
      "rewards-max/forensics_strict_edge_reward": 0.4081176519393921,
      "rewards-max/forensics_tiou_reward": 0.9639581739902496,
      "rewards/forensics_sinkhorn_reward": 0.7159585356712341,
      "rewards/forensics_strict_edge_reward": 0.06820459105074406,
      "rewards/forensics_tiou_reward": 0.5683069080114365,
      "step": 1245
    },
    {
      "completion_length": 23.46875,
      "epoch": 3.823348694316436,
      "grad_norm": 14.516366958618164,
      "kl": 0.8203125,
      "learning_rate": 4.230769230769231e-08,
      "loss": -0.1892,
      "reward": 0.7824981510639191,
      "reward_std": 0.026788318529725075,
      "reward_std_all": 0.026852315291762352,
      "rewards-max/forensics_sinkhorn_reward": 0.9813711047172546,
      "rewards-max/forensics_strict_edge_reward": 0.7676310539245605,
      "rewards-max/forensics_tiou_reward": 0.9790809154510498,
      "rewards/forensics_sinkhorn_reward": 0.7831339836120605,
      "rewards/forensics_strict_edge_reward": 0.09971446171402931,
      "rewards/forensics_tiou_reward": 0.6287696063518524,
      "step": 1246
    },
    {
      "completion_length": 26.984375,
      "epoch": 3.82642089093702,
      "grad_norm": 12.081145286560059,
      "kl": 0.6640625,
      "learning_rate": 4.1538461538461534e-08,
      "loss": 0.3159,
      "reward": 0.7548266053199768,
      "reward_std": 0.03268296550959349,
      "reward_std_all": 0.04223649203777313,
      "rewards-max/forensics_sinkhorn_reward": 0.9727338552474976,
      "rewards-max/forensics_strict_edge_reward": 0.48478010296821594,
      "rewards-max/forensics_tiou_reward": 0.9575945138931274,
      "rewards/forensics_sinkhorn_reward": 0.7110810875892639,
      "rewards/forensics_strict_edge_reward": 0.050794124603271484,
      "rewards/forensics_tiou_reward": 0.5215744376182556,
      "step": 1247
    },
    {
      "completion_length": 27.84375,
      "epoch": 3.8294930875576036,
      "grad_norm": 14.078128814697266,
      "kl": 0.537109375,
      "learning_rate": 4.076923076923077e-08,
      "loss": 0.1892,
      "reward": 0.741843581199646,
      "reward_std": 0.03399873245507479,
      "reward_std_all": 0.04836377501487732,
      "rewards-max/forensics_sinkhorn_reward": 0.9784733355045319,
      "rewards-max/forensics_strict_edge_reward": 0.571812093257904,
      "rewards-max/forensics_tiou_reward": 0.9791470766067505,
      "rewards/forensics_sinkhorn_reward": 0.6722128093242645,
      "rewards/forensics_strict_edge_reward": 0.04475048929452896,
      "rewards/forensics_tiou_reward": 0.5146128237247467,
      "step": 1248
    },
    {
      "completion_length": 23.265625,
      "epoch": 3.8325652841781874,
      "grad_norm": 14.423864364624023,
      "kl": 0.4306640625,
      "learning_rate": 4e-08,
      "loss": -0.1206,
      "reward": 0.7536188066005707,
      "reward_std": 0.009517619735561311,
      "reward_std_all": 0.030403517419472337,
      "rewards-max/forensics_sinkhorn_reward": 0.9868015348911285,
      "rewards-max/forensics_strict_edge_reward": 0.882964164018631,
      "rewards-max/forensics_tiou_reward": 0.9949827790260315,
      "rewards/forensics_sinkhorn_reward": 0.712453305721283,
      "rewards/forensics_strict_edge_reward": 0.13884372264146805,
      "rewards/forensics_tiou_reward": 0.5694754719734192,
      "step": 1249
    },
    {
      "completion_length": 26.46875,
      "epoch": 3.835637480798771,
      "grad_norm": 11.225981712341309,
      "kl": 0.5546875,
      "learning_rate": 3.923076923076923e-08,
      "loss": -0.1504,
      "reward": 0.7929150462150574,
      "reward_std": 0.013888674206100404,
      "reward_std_all": 0.01582130859605968,
      "rewards-max/forensics_sinkhorn_reward": 0.9807299673557281,
      "rewards-max/forensics_strict_edge_reward": 0.6762872636318207,
      "rewards-max/forensics_tiou_reward": 0.9786303341388702,
      "rewards/forensics_sinkhorn_reward": 0.7627094388008118,
      "rewards/forensics_strict_edge_reward": 0.08694512024521828,
      "rewards/forensics_tiou_reward": 0.6425416171550751,
      "step": 1250
    },
    {
      "completion_length": 26.09375,
      "epoch": 3.838709677419355,
      "grad_norm": 10.876737594604492,
      "kl": 0.5078125,
      "learning_rate": 3.846153846153846e-08,
      "loss": -0.1028,
      "reward": 0.754572182893753,
      "reward_std": 0.028772903606295586,
      "reward_std_all": 0.04223410598933697,
      "rewards-max/forensics_sinkhorn_reward": 0.9834102690219879,
      "rewards-max/forensics_strict_edge_reward": 0.882964164018631,
      "rewards-max/forensics_tiou_reward": 0.9870569109916687,
      "rewards/forensics_sinkhorn_reward": 0.7452708780765533,
      "rewards/forensics_strict_edge_reward": 0.19412197172641754,
      "rewards/forensics_tiou_reward": 0.6099054515361786,
      "step": 1251
    },
    {
      "completion_length": 23.40625,
      "epoch": 3.8417818740399383,
      "grad_norm": 12.183894157409668,
      "kl": 0.4287109375,
      "learning_rate": 3.769230769230769e-08,
      "loss": -0.0597,
      "reward": 0.7620440721511841,
      "reward_std": 0.05301478039473295,
      "reward_std_all": 0.05108275916427374,
      "rewards-max/forensics_sinkhorn_reward": 0.9830755591392517,
      "rewards-max/forensics_strict_edge_reward": 0.6515887677669525,
      "rewards-max/forensics_tiou_reward": 0.9840269684791565,
      "rewards/forensics_sinkhorn_reward": 0.7490206062793732,
      "rewards/forensics_strict_edge_reward": 0.06804417073726654,
      "rewards/forensics_tiou_reward": 0.5658099949359894,
      "step": 1252
    },
    {
      "completion_length": 27.140625,
      "epoch": 3.8448540706605225,
      "grad_norm": 11.108710289001465,
      "kl": 0.6171875,
      "learning_rate": 3.692307692307692e-08,
      "loss": -0.1599,
      "reward": 0.788256824016571,
      "reward_std": 0.005302684381604195,
      "reward_std_all": 0.012862547300755978,
      "rewards-max/forensics_sinkhorn_reward": 0.9835408329963684,
      "rewards-max/forensics_strict_edge_reward": 0.554103210568428,
      "rewards-max/forensics_tiou_reward": 0.9849080741405487,
      "rewards/forensics_sinkhorn_reward": 0.7801380753517151,
      "rewards/forensics_strict_edge_reward": 0.10626393184065819,
      "rewards/forensics_tiou_reward": 0.6247994303703308,
      "step": 1253
    },
    {
      "completion_length": 20.703125,
      "epoch": 3.847926267281106,
      "grad_norm": 12.946954727172852,
      "kl": 0.552734375,
      "learning_rate": 3.6153846153846155e-08,
      "loss": -0.182,
      "reward": 0.777775913476944,
      "reward_std": 0.03902278933674097,
      "reward_std_all": 0.037024556659162045,
      "rewards-max/forensics_sinkhorn_reward": 0.9840609729290009,
      "rewards-max/forensics_strict_edge_reward": 0.9093653857707977,
      "rewards-max/forensics_tiou_reward": 0.9888224601745605,
      "rewards/forensics_sinkhorn_reward": 0.7553404271602631,
      "rewards/forensics_strict_edge_reward": 0.11018813028931618,
      "rewards/forensics_tiou_reward": 0.5324594378471375,
      "step": 1254
    },
    {
      "completion_length": 27.578125,
      "epoch": 3.8509984639016897,
      "grad_norm": 17.790000915527344,
      "kl": 0.580078125,
      "learning_rate": 3.538461538461538e-08,
      "loss": -0.1921,
      "reward": 0.7625064849853516,
      "reward_std": 0.024772929027676582,
      "reward_std_all": 0.03698069369420409,
      "rewards-max/forensics_sinkhorn_reward": 0.9884460270404816,
      "rewards-max/forensics_strict_edge_reward": 0.8333354890346527,
      "rewards-max/forensics_tiou_reward": 0.9977462589740753,
      "rewards/forensics_sinkhorn_reward": 0.7072434723377228,
      "rewards/forensics_strict_edge_reward": 0.08807359635829926,
      "rewards/forensics_tiou_reward": 0.5693105757236481,
      "step": 1255
    },
    {
      "completion_length": 23.71875,
      "epoch": 3.8540706605222734,
      "grad_norm": 11.435504913330078,
      "kl": 0.4365234375,
      "learning_rate": 3.4615384615384616e-08,
      "loss": -0.0275,
      "reward": 0.7790156900882721,
      "reward_std": 0.03521496616303921,
      "reward_std_all": 0.033828520216047764,
      "rewards-max/forensics_sinkhorn_reward": 0.9830718040466309,
      "rewards-max/forensics_strict_edge_reward": 0.5500316619873047,
      "rewards-max/forensics_tiou_reward": 0.9831455051898956,
      "rewards/forensics_sinkhorn_reward": 0.7391215562820435,
      "rewards/forensics_strict_edge_reward": 0.08600674569606781,
      "rewards/forensics_tiou_reward": 0.6233378648757935,
      "step": 1256
    },
    {
      "completion_length": 24.515625,
      "epoch": 3.857142857142857,
      "grad_norm": 11.681398391723633,
      "kl": 0.548828125,
      "learning_rate": 3.384615384615384e-08,
      "loss": -0.1392,
      "reward": 0.7660717964172363,
      "reward_std": 0.040364207699894905,
      "reward_std_all": 0.04289734549820423,
      "rewards-max/forensics_sinkhorn_reward": 0.9889904856681824,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.7463043630123138,
      "rewards/forensics_strict_edge_reward": 0.1685515120625496,
      "rewards/forensics_tiou_reward": 0.6131042540073395,
      "step": 1257
    },
    {
      "completion_length": 29.03125,
      "epoch": 3.860215053763441,
      "grad_norm": 10.556049346923828,
      "kl": 0.814453125,
      "learning_rate": 3.3076923076923076e-08,
      "loss": 0.0838,
      "reward": 0.748516857624054,
      "reward_std": 0.05822194740176201,
      "reward_std_all": 0.05755232088267803,
      "rewards-max/forensics_sinkhorn_reward": 0.9758649170398712,
      "rewards-max/forensics_strict_edge_reward": 0.6168098747730255,
      "rewards-max/forensics_tiou_reward": 0.9693413376808167,
      "rewards/forensics_sinkhorn_reward": 0.6890420913696289,
      "rewards/forensics_strict_edge_reward": 0.10291656665503979,
      "rewards/forensics_tiou_reward": 0.5625174641609192,
      "step": 1258
    },
    {
      "completion_length": 23.90625,
      "epoch": 3.8632872503840243,
      "grad_norm": 11.20127010345459,
      "kl": 0.3916015625,
      "learning_rate": 3.230769230769231e-08,
      "loss": -0.0557,
      "reward": 0.7836668491363525,
      "reward_std": 0.029906756011769176,
      "reward_std_all": 0.02970903948880732,
      "rewards-max/forensics_sinkhorn_reward": 0.9832400381565094,
      "rewards-max/forensics_strict_edge_reward": 0.5521624386310577,
      "rewards-max/forensics_tiou_reward": 0.982592761516571,
      "rewards/forensics_sinkhorn_reward": 0.8025211691856384,
      "rewards/forensics_strict_edge_reward": 0.09706304594874382,
      "rewards/forensics_tiou_reward": 0.6578211188316345,
      "step": 1259
    },
    {
      "completion_length": 21.828125,
      "epoch": 3.8663594470046085,
      "grad_norm": 12.296835899353027,
      "kl": 0.591796875,
      "learning_rate": 3.153846153846154e-08,
      "loss": -0.0001,
      "reward": 0.778100997209549,
      "reward_std": 0.030170679092407227,
      "reward_std_all": 0.03030336182564497,
      "rewards-max/forensics_sinkhorn_reward": 0.9727183282375336,
      "rewards-max/forensics_strict_edge_reward": 0.656868189573288,
      "rewards-max/forensics_tiou_reward": 0.9533762335777283,
      "rewards/forensics_sinkhorn_reward": 0.7457018494606018,
      "rewards/forensics_strict_edge_reward": 0.09636415541172028,
      "rewards/forensics_tiou_reward": 0.5222199261188507,
      "step": 1260
    },
    {
      "completion_length": 27.390625,
      "epoch": 3.869431643625192,
      "grad_norm": 10.891508102416992,
      "kl": 0.826171875,
      "learning_rate": 3.076923076923077e-08,
      "loss": -0.0705,
      "reward": 0.7835109233856201,
      "reward_std": 0.014308801386505365,
      "reward_std_all": 0.019434326328337193,
      "rewards-max/forensics_sinkhorn_reward": 0.9843932688236237,
      "rewards-max/forensics_strict_edge_reward": 0.6519230306148529,
      "rewards-max/forensics_tiou_reward": 0.988388180732727,
      "rewards/forensics_sinkhorn_reward": 0.7788745760917664,
      "rewards/forensics_strict_edge_reward": 0.1037021204829216,
      "rewards/forensics_tiou_reward": 0.6433489322662354,
      "step": 1261
    },
    {
      "completion_length": 26.125,
      "epoch": 3.8725038402457757,
      "grad_norm": 10.185404777526855,
      "kl": 0.5810546875,
      "learning_rate": 3e-08,
      "loss": 0.0292,
      "reward": 0.7764663100242615,
      "reward_std": 0.031197838485240936,
      "reward_std_all": 0.032586000859737396,
      "rewards-max/forensics_sinkhorn_reward": 0.9860754013061523,
      "rewards-max/forensics_strict_edge_reward": 0.7556840479373932,
      "rewards-max/forensics_tiou_reward": 0.9926312863826752,
      "rewards/forensics_sinkhorn_reward": 0.7677873373031616,
      "rewards/forensics_strict_edge_reward": 0.12859007716178894,
      "rewards/forensics_tiou_reward": 0.6262495219707489,
      "step": 1262
    },
    {
      "completion_length": 22.53125,
      "epoch": 3.8755760368663594,
      "grad_norm": 14.567856788635254,
      "kl": 0.640625,
      "learning_rate": 2.9230769230769228e-08,
      "loss": -0.0371,
      "reward": 0.8047217428684235,
      "reward_std": 0.0016026757657527924,
      "reward_std_all": 0.0038087544962763786,
      "rewards-max/forensics_sinkhorn_reward": 0.9835674464702606,
      "rewards-max/forensics_strict_edge_reward": 0.4389830380678177,
      "rewards-max/forensics_tiou_reward": 0.9827541708946228,
      "rewards/forensics_sinkhorn_reward": 0.7798133790493011,
      "rewards/forensics_strict_edge_reward": 0.07996408175677061,
      "rewards/forensics_tiou_reward": 0.6040723025798798,
      "step": 1263
    },
    {
      "completion_length": 25.65625,
      "epoch": 3.878648233486943,
      "grad_norm": 9.969252586364746,
      "kl": 0.9609375,
      "learning_rate": 2.846153846153846e-08,
      "loss": 0.0731,
      "reward": 0.7743664681911469,
      "reward_std": 0.038644105195999146,
      "reward_std_all": 0.037376197054982185,
      "rewards-max/forensics_sinkhorn_reward": 0.9816218316555023,
      "rewards-max/forensics_strict_edge_reward": 0.8010868430137634,
      "rewards-max/forensics_tiou_reward": 0.9805588722229004,
      "rewards/forensics_sinkhorn_reward": 0.7557516694068909,
      "rewards/forensics_strict_edge_reward": 0.126240536570549,
      "rewards/forensics_tiou_reward": 0.6144852936267853,
      "step": 1264
    },
    {
      "completion_length": 25.59375,
      "epoch": 3.881720430107527,
      "grad_norm": 12.13481616973877,
      "kl": 0.6015625,
      "learning_rate": 2.769230769230769e-08,
      "loss": -0.0389,
      "reward": 0.7861567735671997,
      "reward_std": 0.022974110674113035,
      "reward_std_all": 0.0232092859223485,
      "rewards-max/forensics_sinkhorn_reward": 0.9830185174942017,
      "rewards-max/forensics_strict_edge_reward": 0.5247012078762054,
      "rewards-max/forensics_tiou_reward": 0.9833244681358337,
      "rewards/forensics_sinkhorn_reward": 0.8106629252433777,
      "rewards/forensics_strict_edge_reward": 0.0868937335908413,
      "rewards/forensics_tiou_reward": 0.6635096371173859,
      "step": 1265
    },
    {
      "completion_length": 21.28125,
      "epoch": 3.8847926267281108,
      "grad_norm": 14.350813865661621,
      "kl": 0.681640625,
      "learning_rate": 2.6923076923076925e-08,
      "loss": -0.0201,
      "reward": 0.7789939641952515,
      "reward_std": 0.033887202385813,
      "reward_std_all": 0.03354113595560193,
      "rewards-max/forensics_sinkhorn_reward": 0.9819231033325195,
      "rewards-max/forensics_strict_edge_reward": 0.6934256851673126,
      "rewards-max/forensics_tiou_reward": 0.9806628227233887,
      "rewards/forensics_sinkhorn_reward": 0.7550191283226013,
      "rewards/forensics_strict_edge_reward": 0.06867234408855438,
      "rewards/forensics_tiou_reward": 0.5594519972801208,
      "step": 1266
    },
    {
      "completion_length": 23.90625,
      "epoch": 3.8878648233486945,
      "grad_norm": 11.220230102539062,
      "kl": 0.5400390625,
      "learning_rate": 2.6153846153846152e-08,
      "loss": -0.0669,
      "reward": 0.7965913712978363,
      "reward_std": 0.017089329077862203,
      "reward_std_all": 0.016682101180776954,
      "rewards-max/forensics_sinkhorn_reward": 0.9806186258792877,
      "rewards-max/forensics_strict_edge_reward": 0.6559408754110336,
      "rewards-max/forensics_tiou_reward": 0.9757835865020752,
      "rewards/forensics_sinkhorn_reward": 0.8115920424461365,
      "rewards/forensics_strict_edge_reward": 0.09958072006702423,
      "rewards/forensics_tiou_reward": 0.6167729496955872,
      "step": 1267
    },
    {
      "completion_length": 26.5,
      "epoch": 3.890937019969278,
      "grad_norm": 11.221651077270508,
      "kl": 0.4033203125,
      "learning_rate": 2.5384615384615382e-08,
      "loss": 0.0371,
      "reward": 0.7993616759777069,
      "reward_std": 0.0074837044812738895,
      "reward_std_all": 0.00905290013179183,
      "rewards-max/forensics_sinkhorn_reward": 0.9757648408412933,
      "rewards-max/forensics_strict_edge_reward": 0.7408519983291626,
      "rewards-max/forensics_tiou_reward": 0.9658978581428528,
      "rewards/forensics_sinkhorn_reward": 0.7464607656002045,
      "rewards/forensics_strict_edge_reward": 0.12744299322366714,
      "rewards/forensics_tiou_reward": 0.634170651435852,
      "step": 1268
    },
    {
      "completion_length": 24.328125,
      "epoch": 3.8940092165898617,
      "grad_norm": 22.86222267150879,
      "kl": 0.48046875,
      "learning_rate": 2.4615384615384616e-08,
      "loss": -0.0933,
      "reward": 0.7965124249458313,
      "reward_std": 0.008996298070997,
      "reward_std_all": 0.011300528421998024,
      "rewards-max/forensics_sinkhorn_reward": 0.9750460684299469,
      "rewards-max/forensics_strict_edge_reward": 0.5905451774597168,
      "rewards-max/forensics_tiou_reward": 0.9666953980922699,
      "rewards/forensics_sinkhorn_reward": 0.7800132930278778,
      "rewards/forensics_strict_edge_reward": 0.07860256172716618,
      "rewards/forensics_tiou_reward": 0.5946878790855408,
      "step": 1269
    },
    {
      "completion_length": 24.90625,
      "epoch": 3.8970814132104454,
      "grad_norm": 12.402047157287598,
      "kl": 0.6728515625,
      "learning_rate": 2.3846153846153846e-08,
      "loss": -0.0315,
      "reward": 0.7541896104812622,
      "reward_std": 0.04810159280896187,
      "reward_std_all": 0.04986879415810108,
      "rewards-max/forensics_sinkhorn_reward": 0.9836817681789398,
      "rewards-max/forensics_strict_edge_reward": 0.7229099273681641,
      "rewards-max/forensics_tiou_reward": 0.9837023615837097,
      "rewards/forensics_sinkhorn_reward": 0.7576616406440735,
      "rewards/forensics_strict_edge_reward": 0.13123152405023575,
      "rewards/forensics_tiou_reward": 0.6141605377197266,
      "step": 1270
    },
    {
      "completion_length": 27.703125,
      "epoch": 3.900153609831029,
      "grad_norm": 10.884673118591309,
      "kl": 0.4736328125,
      "learning_rate": 2.3076923076923076e-08,
      "loss": -0.2034,
      "reward": 0.749613344669342,
      "reward_std": 0.02325283596292138,
      "reward_std_all": 0.04151183366775513,
      "rewards-max/forensics_sinkhorn_reward": 0.9864862859249115,
      "rewards-max/forensics_strict_edge_reward": 0.726351723074913,
      "rewards-max/forensics_tiou_reward": 0.9947253465652466,
      "rewards/forensics_sinkhorn_reward": 0.7486250400543213,
      "rewards/forensics_strict_edge_reward": 0.12233797833323479,
      "rewards/forensics_tiou_reward": 0.6327337622642517,
      "step": 1271
    },
    {
      "completion_length": 26.65625,
      "epoch": 3.903225806451613,
      "grad_norm": 12.39938735961914,
      "kl": 0.515625,
      "learning_rate": 2.2307692307692306e-08,
      "loss": -0.1305,
      "reward": 0.7667796909809113,
      "reward_std": 0.029164589941501617,
      "reward_std_all": 0.03588049300014973,
      "rewards-max/forensics_sinkhorn_reward": 0.9718818664550781,
      "rewards-max/forensics_strict_edge_reward": 0.48839980363845825,
      "rewards-max/forensics_tiou_reward": 0.9530557692050934,
      "rewards/forensics_sinkhorn_reward": 0.7522136569023132,
      "rewards/forensics_strict_edge_reward": 0.07360531948506832,
      "rewards/forensics_tiou_reward": 0.5992449522018433,
      "step": 1272
    },
    {
      "completion_length": 26.53125,
      "epoch": 3.9062980030721968,
      "grad_norm": 9.696115493774414,
      "kl": 0.734375,
      "learning_rate": 2.1538461538461537e-08,
      "loss": 0.0199,
      "reward": 0.7986178994178772,
      "reward_std": 0.01078785303980112,
      "reward_std_all": 0.01200762391090393,
      "rewards-max/forensics_sinkhorn_reward": 0.9854376912117004,
      "rewards-max/forensics_strict_edge_reward": 0.5801982581615448,
      "rewards-max/forensics_tiou_reward": 0.9932997524738312,
      "rewards/forensics_sinkhorn_reward": 0.780071347951889,
      "rewards/forensics_strict_edge_reward": 0.13646329939365387,
      "rewards/forensics_tiou_reward": 0.6450276970863342,
      "step": 1273
    },
    {
      "completion_length": 25.546875,
      "epoch": 3.9093701996927805,
      "grad_norm": 8.672510147094727,
      "kl": 0.494140625,
      "learning_rate": 2.0769230769230767e-08,
      "loss": 0.0727,
      "reward": 0.7713899910449982,
      "reward_std": 0.02982653072103858,
      "reward_std_all": 0.034974245354533195,
      "rewards-max/forensics_sinkhorn_reward": 0.9842382073402405,
      "rewards-max/forensics_strict_edge_reward": 0.7617588341236115,
      "rewards-max/forensics_tiou_reward": 0.9886210858821869,
      "rewards/forensics_sinkhorn_reward": 0.7539336085319519,
      "rewards/forensics_strict_edge_reward": 0.10579367354512215,
      "rewards/forensics_tiou_reward": 0.5952102243900299,
      "step": 1274
    },
    {
      "completion_length": 23.65625,
      "epoch": 3.912442396313364,
      "grad_norm": 9.220300674438477,
      "kl": 0.71875,
      "learning_rate": 2e-08,
      "loss": 0.1161,
      "reward": 0.777083158493042,
      "reward_std": 0.03971690125763416,
      "reward_std_all": 0.03732117637991905,
      "rewards-max/forensics_sinkhorn_reward": 0.9846823215484619,
      "rewards-max/forensics_strict_edge_reward": 0.7482926547527313,
      "rewards-max/forensics_tiou_reward": 0.9868985712528229,
      "rewards/forensics_sinkhorn_reward": 0.7782160341739655,
      "rewards/forensics_strict_edge_reward": 0.16039561480283737,
      "rewards/forensics_tiou_reward": 0.6469365358352661,
      "step": 1275
    },
    {
      "completion_length": 26.75,
      "epoch": 3.9155145929339477,
      "grad_norm": 11.120304107666016,
      "kl": 0.51953125,
      "learning_rate": 1.923076923076923e-08,
      "loss": 0.0999,
      "reward": 0.7842997908592224,
      "reward_std": 0.02045180695131421,
      "reward_std_all": 0.022048178361728787,
      "rewards-max/forensics_sinkhorn_reward": 0.9832710921764374,
      "rewards-max/forensics_strict_edge_reward": 0.5771172791719437,
      "rewards-max/forensics_tiou_reward": 0.9862519800662994,
      "rewards/forensics_sinkhorn_reward": 0.7471132576465607,
      "rewards/forensics_strict_edge_reward": 0.09819376468658447,
      "rewards/forensics_tiou_reward": 0.5844533443450928,
      "step": 1276
    },
    {
      "completion_length": 22.265625,
      "epoch": 3.9185867895545314,
      "grad_norm": 13.133655548095703,
      "kl": 0.451171875,
      "learning_rate": 1.846153846153846e-08,
      "loss": -0.1667,
      "reward": 0.7917567789554596,
      "reward_std": 0.0025255203945562243,
      "reward_std_all": 0.010688761249184608,
      "rewards-max/forensics_sinkhorn_reward": 0.9855255484580994,
      "rewards-max/forensics_strict_edge_reward": 0.9375866651535034,
      "rewards-max/forensics_tiou_reward": 0.9921880960464478,
      "rewards/forensics_sinkhorn_reward": 0.7963935732841492,
      "rewards/forensics_strict_edge_reward": 0.1949976682662964,
      "rewards/forensics_tiou_reward": 0.6784176528453827,
      "step": 1277
    },
    {
      "completion_length": 26.375,
      "epoch": 3.921658986175115,
      "grad_norm": 14.45852279663086,
      "kl": 0.9580078125,
      "learning_rate": 1.769230769230769e-08,
      "loss": -0.0351,
      "reward": 0.7758950591087341,
      "reward_std": 0.034658804070204496,
      "reward_std_all": 0.03478762088343501,
      "rewards-max/forensics_sinkhorn_reward": 0.9832858741283417,
      "rewards-max/forensics_strict_edge_reward": 0.5770992785692215,
      "rewards-max/forensics_tiou_reward": 0.9861456751823425,
      "rewards/forensics_sinkhorn_reward": 0.6971204876899719,
      "rewards/forensics_strict_edge_reward": 0.10314378514885902,
      "rewards/forensics_tiou_reward": 0.5781028568744659,
      "step": 1278
    },
    {
      "completion_length": 23.046875,
      "epoch": 3.924731182795699,
      "grad_norm": 11.748353958129883,
      "kl": 0.580078125,
      "learning_rate": 1.692307692307692e-08,
      "loss": 0.1087,
      "reward": 0.7645336985588074,
      "reward_std": 0.065324692055583,
      "reward_std_all": 0.05893315002322197,
      "rewards-max/forensics_sinkhorn_reward": 0.9874409437179565,
      "rewards-max/forensics_strict_edge_reward": 0.7933230996131897,
      "rewards-max/forensics_tiou_reward": 0.9963687360286713,
      "rewards/forensics_sinkhorn_reward": 0.7203534841537476,
      "rewards/forensics_strict_edge_reward": 0.14342395216226578,
      "rewards/forensics_tiou_reward": 0.5863332450389862,
      "step": 1279
    },
    {
      "completion_length": 26.921875,
      "epoch": 3.9278033794162828,
      "grad_norm": 10.356389999389648,
      "kl": 0.5859375,
      "learning_rate": 1.6153846153846155e-08,
      "loss": 0.1032,
      "reward": 0.7889716327190399,
      "reward_std": 0.01081043086014688,
      "reward_std_all": 0.014856071677058935,
      "rewards-max/forensics_sinkhorn_reward": 0.9857814311981201,
      "rewards-max/forensics_strict_edge_reward": 0.7744058072566986,
      "rewards-max/forensics_tiou_reward": 0.9902372360229492,
      "rewards/forensics_sinkhorn_reward": 0.7536338567733765,
      "rewards/forensics_strict_edge_reward": 0.09889159351587296,
      "rewards/forensics_tiou_reward": 0.6539579033851624,
      "step": 1280
    },
    {
      "completion_length": 29.453125,
      "epoch": 3.9308755760368665,
      "grad_norm": 9.546055793762207,
      "kl": 0.65625,
      "learning_rate": 1.5384615384615385e-08,
      "loss": 0.1005,
      "reward": 0.7586086094379425,
      "reward_std": 0.014446087647229433,
      "reward_std_all": 0.032085723243653774,
      "rewards-max/forensics_sinkhorn_reward": 0.9825008809566498,
      "rewards-max/forensics_strict_edge_reward": 0.6222288608551025,
      "rewards-max/forensics_tiou_reward": 0.9839734435081482,
      "rewards/forensics_sinkhorn_reward": 0.7255672812461853,
      "rewards/forensics_strict_edge_reward": 0.1078347247093916,
      "rewards/forensics_tiou_reward": 0.5939099192619324,
      "step": 1281
    },
    {
      "completion_length": 27.625,
      "epoch": 3.93394777265745,
      "grad_norm": 12.259971618652344,
      "kl": 0.76171875,
      "learning_rate": 1.4615384615384614e-08,
      "loss": -0.1195,
      "reward": 0.7824648916721344,
      "reward_std": 0.021154449903406203,
      "reward_std_all": 0.02302985591813922,
      "rewards-max/forensics_sinkhorn_reward": 0.9857868850231171,
      "rewards-max/forensics_strict_edge_reward": 0.6910748332738876,
      "rewards-max/forensics_tiou_reward": 0.9926360249519348,
      "rewards/forensics_sinkhorn_reward": 0.7868374586105347,
      "rewards/forensics_strict_edge_reward": 0.06647337973117828,
      "rewards/forensics_tiou_reward": 0.6262785494327545,
      "step": 1282
    },
    {
      "completion_length": 21.953125,
      "epoch": 3.9370199692780337,
      "grad_norm": 14.181256294250488,
      "kl": 0.5390625,
      "learning_rate": 1.3846153846153846e-08,
      "loss": -0.1351,
      "reward": 0.7933585345745087,
      "reward_std": 0.024605765007436275,
      "reward_std_all": 0.022608229890465736,
      "rewards-max/forensics_sinkhorn_reward": 0.980079174041748,
      "rewards-max/forensics_strict_edge_reward": 0.4966079294681549,
      "rewards-max/forensics_tiou_reward": 0.9738763272762299,
      "rewards/forensics_sinkhorn_reward": 0.8442082107067108,
      "rewards/forensics_strict_edge_reward": 0.12471464276313782,
      "rewards/forensics_tiou_reward": 0.7025166451931,
      "step": 1283
    },
    {
      "completion_length": 24.640625,
      "epoch": 3.9400921658986174,
      "grad_norm": 14.864885330200195,
      "kl": 0.51171875,
      "learning_rate": 1.3076923076923076e-08,
      "loss": -0.121,
      "reward": 0.7834285199642181,
      "reward_std": 0.032171749509871006,
      "reward_std_all": 0.03142764000222087,
      "rewards-max/forensics_sinkhorn_reward": 0.9855102598667145,
      "rewards-max/forensics_strict_edge_reward": 0.7068770229816437,
      "rewards-max/forensics_tiou_reward": 0.9922563135623932,
      "rewards/forensics_sinkhorn_reward": 0.7759533524513245,
      "rewards/forensics_strict_edge_reward": 0.11385393887758255,
      "rewards/forensics_tiou_reward": 0.6223427653312683,
      "step": 1284
    },
    {
      "completion_length": 23.203125,
      "epoch": 3.943164362519201,
      "grad_norm": 10.687092781066895,
      "kl": 0.61328125,
      "learning_rate": 1.2307692307692308e-08,
      "loss": 0.0196,
      "reward": 0.77763232588768,
      "reward_std": 0.03569864481687546,
      "reward_std_all": 0.03443656861782074,
      "rewards-max/forensics_sinkhorn_reward": 0.9771672487258911,
      "rewards-max/forensics_strict_edge_reward": 0.7475712299346924,
      "rewards-max/forensics_tiou_reward": 0.9650472402572632,
      "rewards/forensics_sinkhorn_reward": 0.7420582175254822,
      "rewards/forensics_strict_edge_reward": 0.0766131691634655,
      "rewards/forensics_tiou_reward": 0.5296759009361267,
      "step": 1285
    },
    {
      "completion_length": 25.703125,
      "epoch": 3.946236559139785,
      "grad_norm": 13.301726341247559,
      "kl": 0.64453125,
      "learning_rate": 1.1538461538461538e-08,
      "loss": 0.07,
      "reward": 0.7747381627559662,
      "reward_std": 0.016525388695299625,
      "reward_std_all": 0.0265238955616951,
      "rewards-max/forensics_sinkhorn_reward": 0.9873312413692474,
      "rewards-max/forensics_strict_edge_reward": 0.7923295497894287,
      "rewards-max/forensics_tiou_reward": 0.9943425953388214,
      "rewards/forensics_sinkhorn_reward": 0.7852974236011505,
      "rewards/forensics_strict_edge_reward": 0.0863448865711689,
      "rewards/forensics_tiou_reward": 0.6176743507385254,
      "step": 1286
    },
    {
      "completion_length": 24.453125,
      "epoch": 3.9493087557603688,
      "grad_norm": 17.225637435913086,
      "kl": 0.5322265625,
      "learning_rate": 1.0769230769230768e-08,
      "loss": -0.2976,
      "reward": 0.7965904772281647,
      "reward_std": 0.010022893315181136,
      "reward_std_all": 0.011526962043717504,
      "rewards-max/forensics_sinkhorn_reward": 0.9817573726177216,
      "rewards-max/forensics_strict_edge_reward": 0.6620501279830933,
      "rewards-max/forensics_tiou_reward": 0.982483834028244,
      "rewards/forensics_sinkhorn_reward": 0.8166784644126892,
      "rewards/forensics_strict_edge_reward": 0.11503247544169426,
      "rewards/forensics_tiou_reward": 0.6961460113525391,
      "step": 1287
    },
    {
      "completion_length": 24.109375,
      "epoch": 3.9523809523809526,
      "grad_norm": 13.7335205078125,
      "kl": 0.521484375,
      "learning_rate": 1e-08,
      "loss": -0.165,
      "reward": 0.7752509415149689,
      "reward_std": 0.04010299500077963,
      "reward_std_all": 0.03819822985678911,
      "rewards-max/forensics_sinkhorn_reward": 0.9857544898986816,
      "rewards-max/forensics_strict_edge_reward": 0.7488664835691452,
      "rewards-max/forensics_tiou_reward": 0.9906542003154755,
      "rewards/forensics_sinkhorn_reward": 0.7711034119129181,
      "rewards/forensics_strict_edge_reward": 0.13002120703458786,
      "rewards/forensics_tiou_reward": 0.6259284317493439,
      "step": 1288
    },
    {
      "completion_length": 23.21875,
      "epoch": 3.955453149001536,
      "grad_norm": 12.820817947387695,
      "kl": 0.65625,
      "learning_rate": 9.23076923076923e-09,
      "loss": 0.109,
      "reward": 0.752103865146637,
      "reward_std": 0.02707710675895214,
      "reward_std_all": 0.04320947267115116,
      "rewards-max/forensics_sinkhorn_reward": 0.9787775278091431,
      "rewards-max/forensics_strict_edge_reward": 0.561550498008728,
      "rewards-max/forensics_tiou_reward": 0.9700824916362762,
      "rewards/forensics_sinkhorn_reward": 0.7632201910018921,
      "rewards/forensics_strict_edge_reward": 0.11268393322825432,
      "rewards/forensics_tiou_reward": 0.5997701287269592,
      "step": 1289
    },
    {
      "completion_length": 26.578125,
      "epoch": 3.9585253456221197,
      "grad_norm": 12.198567390441895,
      "kl": 0.404296875,
      "learning_rate": 8.46153846153846e-09,
      "loss": -0.0396,
      "reward": 0.7166403234004974,
      "reward_std": 0.07445911131799221,
      "reward_std_all": 0.08144033700227737,
      "rewards-max/forensics_sinkhorn_reward": 0.9777397811412811,
      "rewards-max/forensics_strict_edge_reward": 0.6195405721664429,
      "rewards-max/forensics_tiou_reward": 0.9695935547351837,
      "rewards/forensics_sinkhorn_reward": 0.7112513482570648,
      "rewards/forensics_strict_edge_reward": 0.09935292974114418,
      "rewards/forensics_tiou_reward": 0.5819691717624664,
      "step": 1290
    },
    {
      "completion_length": 23.953125,
      "epoch": 3.9615975422427034,
      "grad_norm": 12.757083892822266,
      "kl": 0.5439453125,
      "learning_rate": 7.692307692307693e-09,
      "loss": -0.1923,
      "reward": 0.8021873831748962,
      "reward_std": 0.009288901055697352,
      "reward_std_all": 0.009471896570175886,
      "rewards-max/forensics_sinkhorn_reward": 0.980867862701416,
      "rewards-max/forensics_strict_edge_reward": 0.6515887677669525,
      "rewards-max/forensics_tiou_reward": 0.9782769680023193,
      "rewards/forensics_sinkhorn_reward": 0.86553755402565,
      "rewards/forensics_strict_edge_reward": 0.11954829096794128,
      "rewards/forensics_tiou_reward": 0.7380503118038177,
      "step": 1291
    },
    {
      "completion_length": 25.125,
      "epoch": 3.964669738863287,
      "grad_norm": 10.74061107635498,
      "kl": 0.5361328125,
      "learning_rate": 6.923076923076923e-09,
      "loss": 0.0107,
      "reward": 0.7791552245616913,
      "reward_std": 0.019242385402321815,
      "reward_std_all": 0.026065651327371597,
      "rewards-max/forensics_sinkhorn_reward": 0.9878706932067871,
      "rewards-max/forensics_strict_edge_reward": 0.5316829085350037,
      "rewards-max/forensics_tiou_reward": 0.9954460561275482,
      "rewards/forensics_sinkhorn_reward": 0.7235795557498932,
      "rewards/forensics_strict_edge_reward": 0.1057894378900528,
      "rewards/forensics_tiou_reward": 0.5697344839572906,
      "step": 1292
    },
    {
      "completion_length": 25.6875,
      "epoch": 3.967741935483871,
      "grad_norm": 13.464608192443848,
      "kl": 0.673828125,
      "learning_rate": 6.153846153846154e-09,
      "loss": 0.105,
      "reward": 0.7210240066051483,
      "reward_std": 0.057946834713220596,
      "reward_std_all": 0.06838526763021946,
      "rewards-max/forensics_sinkhorn_reward": 0.980416864156723,
      "rewards-max/forensics_strict_edge_reward": 0.5059049725532532,
      "rewards-max/forensics_tiou_reward": 0.9795194566249847,
      "rewards/forensics_sinkhorn_reward": 0.6415240168571472,
      "rewards/forensics_strict_edge_reward": 0.06578974239528179,
      "rewards/forensics_tiou_reward": 0.4510860741138458,
      "step": 1293
    },
    {
      "completion_length": 24.859375,
      "epoch": 3.9708141321044548,
      "grad_norm": 10.12903881072998,
      "kl": 0.5546875,
      "learning_rate": 5.384615384615384e-09,
      "loss": -0.2296,
      "reward": 0.7384087443351746,
      "reward_std": 0.02861933410167694,
      "reward_std_all": 0.04912709258496761,
      "rewards-max/forensics_sinkhorn_reward": 0.980357438325882,
      "rewards-max/forensics_strict_edge_reward": 0.7201169431209564,
      "rewards-max/forensics_tiou_reward": 0.9781030714511871,
      "rewards/forensics_sinkhorn_reward": 0.68647700548172,
      "rewards/forensics_strict_edge_reward": 0.08073082938790321,
      "rewards/forensics_tiou_reward": 0.4831397980451584,
      "step": 1294
    },
    {
      "completion_length": 28.21875,
      "epoch": 3.9738863287250386,
      "grad_norm": 10.618188858032227,
      "kl": 0.5498046875,
      "learning_rate": 4.615384615384615e-09,
      "loss": -0.3303,
      "reward": 0.8037760853767395,
      "reward_std": 0.004790679435245693,
      "reward_std_all": 0.006166461855173111,
      "rewards-max/forensics_sinkhorn_reward": 0.9802369177341461,
      "rewards-max/forensics_strict_edge_reward": 0.5902654826641083,
      "rewards-max/forensics_tiou_reward": 0.9779553711414337,
      "rewards/forensics_sinkhorn_reward": 0.8237404525279999,
      "rewards/forensics_strict_edge_reward": 0.08816629275679588,
      "rewards/forensics_tiou_reward": 0.6801879405975342,
      "step": 1295
    },
    {
      "completion_length": 27.953125,
      "epoch": 3.976958525345622,
      "grad_norm": 10.245548248291016,
      "kl": 0.609375,
      "learning_rate": 3.846153846153846e-09,
      "loss": 0.078,
      "reward": 0.8017397522926331,
      "reward_std": 0.005706864059902728,
      "reward_std_all": 0.007065423182211816,
      "rewards-max/forensics_sinkhorn_reward": 0.9807412624359131,
      "rewards-max/forensics_strict_edge_reward": 0.47910328209400177,
      "rewards-max/forensics_tiou_reward": 0.9786608517169952,
      "rewards/forensics_sinkhorn_reward": 0.7774777412414551,
      "rewards/forensics_strict_edge_reward": 0.09122936055064201,
      "rewards/forensics_tiou_reward": 0.6673425436019897,
      "step": 1296
    },
    {
      "completion_length": 28.203125,
      "epoch": 3.980030721966206,
      "grad_norm": 8.320113182067871,
      "kl": 0.794921875,
      "learning_rate": 3.076923076923077e-09,
      "loss": -0.0166,
      "reward": 0.7294159829616547,
      "reward_std": 0.062312180176377296,
      "reward_std_all": 0.06743366457521915,
      "rewards-max/forensics_sinkhorn_reward": 0.9849268198013306,
      "rewards-max/forensics_strict_edge_reward": 0.6378968358039856,
      "rewards-max/forensics_tiou_reward": 0.9877196550369263,
      "rewards/forensics_sinkhorn_reward": 0.6934154629707336,
      "rewards/forensics_strict_edge_reward": 0.09927313588559628,
      "rewards/forensics_tiou_reward": 0.5609768331050873,
      "step": 1297
    },
    {
      "completion_length": 24.234375,
      "epoch": 3.9831029185867894,
      "grad_norm": 11.331542015075684,
      "kl": 0.755859375,
      "learning_rate": 2.3076923076923076e-09,
      "loss": -0.08,
      "reward": 0.748887687921524,
      "reward_std": 0.04842417873442173,
      "reward_std_all": 0.0530749075114727,
      "rewards-max/forensics_sinkhorn_reward": 0.982761800289154,
      "rewards-max/forensics_strict_edge_reward": 0.6793906837701797,
      "rewards-max/forensics_tiou_reward": 0.9829095005989075,
      "rewards/forensics_sinkhorn_reward": 0.7226720452308655,
      "rewards/forensics_strict_edge_reward": 0.1315978579223156,
      "rewards/forensics_tiou_reward": 0.603747084736824,
      "step": 1298
    },
    {
      "completion_length": 24.484375,
      "epoch": 3.986175115207373,
      "grad_norm": 9.824557304382324,
      "kl": 0.5126953125,
      "learning_rate": 1.5384615384615385e-09,
      "loss": 0.2085,
      "reward": 0.7829464972019196,
      "reward_std": 0.030324206687510014,
      "reward_std_all": 0.030278686434030533,
      "rewards-max/forensics_sinkhorn_reward": 0.976248174905777,
      "rewards-max/forensics_strict_edge_reward": 0.5304600298404694,
      "rewards-max/forensics_tiou_reward": 0.962367832660675,
      "rewards/forensics_sinkhorn_reward": 0.7295227944850922,
      "rewards/forensics_strict_edge_reward": 0.07776358723640442,
      "rewards/forensics_tiou_reward": 0.541921079158783,
      "step": 1299
    },
    {
      "completion_length": 25.09375,
      "epoch": 3.989247311827957,
      "grad_norm": 13.50718879699707,
      "kl": 0.74609375,
      "learning_rate": 7.692307692307692e-10,
      "loss": 0.0934,
      "reward": 0.7841616570949554,
      "reward_std": 0.032588085159659386,
      "reward_std_all": 0.030710188671946526,
      "rewards-max/forensics_sinkhorn_reward": 0.9847172796726227,
      "rewards-max/forensics_strict_edge_reward": 0.5708207190036774,
      "rewards-max/forensics_tiou_reward": 0.9895649552345276,
      "rewards/forensics_sinkhorn_reward": 0.7797514498233795,
      "rewards/forensics_strict_edge_reward": 0.08770668134093285,
      "rewards/forensics_tiou_reward": 0.592295378446579,
      "step": 1300
    }
  ],
  "logging_steps": 1.0,
  "max_steps": 1300,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 4,
  "save_steps": 1300,
  "stateful_callbacks": {
    "TrainerControl": {
      "args": {
        "should_epoch_stop": false,
        "should_evaluate": false,
        "should_log": false,
        "should_save": true,
        "should_training_stop": true
      },
      "attributes": {}
    }
  },
  "total_flos": 0.0,
  "train_batch_size": 1,
  "trial_name": null,
  "trial_params": null
}