{
  "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": 26.0,
      "epoch": 0.0030721966205837174,
      "grad_norm": 8.558785438537598,
      "kl": 0.0,
      "learning_rate": 1e-06,
      "loss": -0.0408,
      "reward": 0.6489152610301971,
      "reward_std": 0.5033352822065353,
      "reward_std_all": 0.46224114298820496,
      "rewards-max/forensics_sinkhorn_reward": 0.8948954939842224,
      "rewards-max/forensics_strict_edge_reward": 0.0034832387515990604,
      "rewards-max/forensics_tiou_reward": 0.7112037837505341,
      "rewards/forensics_sinkhorn_reward": 0.4323672205209732,
      "rewards/forensics_strict_edge_reward": 0.00020890026155351293,
      "rewards/forensics_tiou_reward": 0.21633916348218918,
      "step": 1
    },
    {
      "completion_length": 22.203125,
      "epoch": 0.006144393241167435,
      "grad_norm": 5.826648712158203,
      "kl": 0.0003123283386230469,
      "learning_rate": 9.992307692307693e-07,
      "loss": -0.0241,
      "reward": 0.6987107694149017,
      "reward_std": 0.5074449777603149,
      "reward_std_all": 0.44501686096191406,
      "rewards-max/forensics_sinkhorn_reward": 0.9358151257038116,
      "rewards-max/forensics_strict_edge_reward": 0.10339117422699928,
      "rewards-max/forensics_tiou_reward": 0.8631332814693451,
      "rewards/forensics_sinkhorn_reward": 0.47036004066467285,
      "rewards/forensics_strict_edge_reward": 0.0061918029095977545,
      "rewards/forensics_tiou_reward": 0.22215894609689713,
      "step": 2
    },
    {
      "completion_length": 21.375,
      "epoch": 0.009216589861751152,
      "grad_norm": 5.102816104888916,
      "kl": 0.0008487701416015625,
      "learning_rate": 9.984615384615384e-07,
      "loss": -0.1543,
      "reward": 0.7325699925422668,
      "reward_std": 0.4703468680381775,
      "reward_std_all": 0.4468201845884323,
      "rewards-max/forensics_sinkhorn_reward": 0.9335746467113495,
      "rewards-max/forensics_strict_edge_reward": 0.3926364965736866,
      "rewards-max/forensics_tiou_reward": 0.858588457107544,
      "rewards/forensics_sinkhorn_reward": 0.48983049392700195,
      "rewards/forensics_strict_edge_reward": 0.020481302868574858,
      "rewards/forensics_tiou_reward": 0.22225818783044815,
      "step": 3
    },
    {
      "completion_length": 20.34375,
      "epoch": 0.01228878648233487,
      "grad_norm": 4.853639602661133,
      "kl": 0.00046062469482421875,
      "learning_rate": 9.976923076923077e-07,
      "loss": -0.0604,
      "reward": 0.745880126953125,
      "reward_std": 0.38190561532974243,
      "reward_std_all": 0.42270542681217194,
      "rewards-max/forensics_sinkhorn_reward": 0.9153572618961334,
      "rewards-max/forensics_strict_edge_reward": 0.07904881902504712,
      "rewards-max/forensics_tiou_reward": 0.8131700754165649,
      "rewards/forensics_sinkhorn_reward": 0.5223603248596191,
      "rewards/forensics_strict_edge_reward": 0.0038403154467232525,
      "rewards/forensics_tiou_reward": 0.21967948973178864,
      "step": 4
    },
    {
      "completion_length": 19.859375,
      "epoch": 0.015360983102918587,
      "grad_norm": 5.442771911621094,
      "kl": 0.004669189453125,
      "learning_rate": 9.969230769230768e-07,
      "loss": -0.0895,
      "reward": 0.7238772809505463,
      "reward_std": 0.4919976592063904,
      "reward_std_all": 0.47938962280750275,
      "rewards-max/forensics_sinkhorn_reward": 0.8943866193294525,
      "rewards-max/forensics_strict_edge_reward": 0.02375007583759725,
      "rewards-max/forensics_tiou_reward": 0.7760106325149536,
      "rewards/forensics_sinkhorn_reward": 0.5135886669158936,
      "rewards/forensics_strict_edge_reward": 0.0012353916827123612,
      "rewards/forensics_tiou_reward": 0.20905321091413498,
      "step": 5
    },
    {
      "completion_length": 20.6875,
      "epoch": 0.018433179723502304,
      "grad_norm": 3.95896053314209,
      "kl": 0.005035400390625,
      "learning_rate": 9.961538461538461e-07,
      "loss": -0.061,
      "reward": 0.7362069189548492,
      "reward_std": 0.4049859493970871,
      "reward_std_all": 0.35104145109653473,
      "rewards-max/forensics_sinkhorn_reward": 0.8367486298084259,
      "rewards-max/forensics_strict_edge_reward": 0.05933381989598274,
      "rewards-max/forensics_tiou_reward": 0.6093680262565613,
      "rewards/forensics_sinkhorn_reward": 0.5043180584907532,
      "rewards/forensics_strict_edge_reward": 0.003015678026713431,
      "rewards/forensics_tiou_reward": 0.22887317836284637,
      "step": 6
    },
    {
      "completion_length": 19.921875,
      "epoch": 0.021505376344086023,
      "grad_norm": 4.132191181182861,
      "kl": 0.0082855224609375,
      "learning_rate": 9.953846153846154e-07,
      "loss": -0.1119,
      "reward": 0.7262179255485535,
      "reward_std": 0.41417618095874786,
      "reward_std_all": 0.3959553986787796,
      "rewards-max/forensics_sinkhorn_reward": 0.8502029180526733,
      "rewards-max/forensics_strict_edge_reward": 0.04148180969059467,
      "rewards-max/forensics_tiou_reward": 0.658677726984024,
      "rewards/forensics_sinkhorn_reward": 0.5082278698682785,
      "rewards/forensics_strict_edge_reward": 0.0020774893928319216,
      "rewards/forensics_tiou_reward": 0.2159125730395317,
      "step": 7
    },
    {
      "completion_length": 21.828125,
      "epoch": 0.02457757296466974,
      "grad_norm": 4.976407051086426,
      "kl": 0.009490966796875,
      "learning_rate": 9.946153846153846e-07,
      "loss": -0.0362,
      "reward": 0.627858579158783,
      "reward_std": 0.4634048789739609,
      "reward_std_all": 0.40056703984737396,
      "rewards-max/forensics_sinkhorn_reward": 0.8655683696269989,
      "rewards-max/forensics_strict_edge_reward": 0.15063580486093997,
      "rewards-max/forensics_tiou_reward": 0.7111949622631073,
      "rewards/forensics_sinkhorn_reward": 0.44321757555007935,
      "rewards/forensics_strict_edge_reward": 0.006287165123467275,
      "rewards/forensics_tiou_reward": 0.17835382372140884,
      "step": 8
    },
    {
      "completion_length": 20.46875,
      "epoch": 0.027649769585253458,
      "grad_norm": 4.552184104919434,
      "kl": 0.011138916015625,
      "learning_rate": 9.938461538461539e-07,
      "loss": -0.0274,
      "reward": 0.8325291275978088,
      "reward_std": 0.3125668615102768,
      "reward_std_all": 0.29496045410633087,
      "rewards-max/forensics_sinkhorn_reward": 0.8297591805458069,
      "rewards-max/forensics_strict_edge_reward": 0.18930562224704772,
      "rewards-max/forensics_tiou_reward": 0.6156750619411469,
      "rewards/forensics_sinkhorn_reward": 0.5811546444892883,
      "rewards/forensics_strict_edge_reward": 0.007923284279968357,
      "rewards/forensics_tiou_reward": 0.24345124512910843,
      "step": 9
    },
    {
      "completion_length": 20.46875,
      "epoch": 0.030721966205837174,
      "grad_norm": 4.8582634925842285,
      "kl": 0.02764892578125,
      "learning_rate": 9.93076923076923e-07,
      "loss": -0.0912,
      "reward": 0.8474940657615662,
      "reward_std": 0.3519132435321808,
      "reward_std_all": 0.32495853304862976,
      "rewards-max/forensics_sinkhorn_reward": 0.9331088960170746,
      "rewards-max/forensics_strict_edge_reward": 0.409512179743615,
      "rewards-max/forensics_tiou_reward": 0.842156171798706,
      "rewards/forensics_sinkhorn_reward": 0.5868945717811584,
      "rewards/forensics_strict_edge_reward": 0.01712154385586473,
      "rewards/forensics_tiou_reward": 0.2434779852628708,
      "step": 10
    },
    {
      "completion_length": 20.09375,
      "epoch": 0.03379416282642089,
      "grad_norm": 4.4294304847717285,
      "kl": 0.043701171875,
      "learning_rate": 9.923076923076923e-07,
      "loss": -0.093,
      "reward": 0.8704609274864197,
      "reward_std": 0.32754214107990265,
      "reward_std_all": 0.34804442524909973,
      "rewards-max/forensics_sinkhorn_reward": 0.9208440780639648,
      "rewards-max/forensics_strict_edge_reward": 0.014695167541503906,
      "rewards-max/forensics_tiou_reward": 0.7954417765140533,
      "rewards/forensics_sinkhorn_reward": 0.614234983921051,
      "rewards/forensics_strict_edge_reward": 0.0006128065288066864,
      "rewards/forensics_tiou_reward": 0.25561314076185226,
      "step": 11
    },
    {
      "completion_length": 19.609375,
      "epoch": 0.03686635944700461,
      "grad_norm": 4.047691822052002,
      "kl": 0.0484619140625,
      "learning_rate": 9.915384615384616e-07,
      "loss": -0.0309,
      "reward": 0.7797723710536957,
      "reward_std": 0.39836208522319794,
      "reward_std_all": 0.3484674245119095,
      "rewards-max/forensics_sinkhorn_reward": 0.8849934637546539,
      "rewards-max/forensics_strict_edge_reward": 0.29066096246242523,
      "rewards-max/forensics_tiou_reward": 0.7890922725200653,
      "rewards/forensics_sinkhorn_reward": 0.5288772583007812,
      "rewards/forensics_strict_edge_reward": 0.012981719803065062,
      "rewards/forensics_tiou_reward": 0.23791342228651047,
      "step": 12
    },
    {
      "completion_length": 19.390625,
      "epoch": 0.039938556067588324,
      "grad_norm": 3.7042763233184814,
      "kl": 0.057373046875,
      "learning_rate": 9.907692307692307e-07,
      "loss": -0.0061,
      "reward": 0.7321389615535736,
      "reward_std": 0.27807383984327316,
      "reward_std_all": 0.30824604630470276,
      "rewards-max/forensics_sinkhorn_reward": 0.86156365275383,
      "rewards-max/forensics_strict_edge_reward": 0.0027531535597518086,
      "rewards-max/forensics_tiou_reward": 0.706723541021347,
      "rewards/forensics_sinkhorn_reward": 0.544201672077179,
      "rewards/forensics_strict_edge_reward": 0.00011988170444965363,
      "rewards/forensics_tiou_reward": 0.18781740218400955,
      "step": 13
    },
    {
      "completion_length": 20.59375,
      "epoch": 0.043010752688172046,
      "grad_norm": 4.653467178344727,
      "kl": 0.056640625,
      "learning_rate": 9.9e-07,
      "loss": -0.018,
      "reward": 0.8976566195487976,
      "reward_std": 0.3197324275970459,
      "reward_std_all": 0.3367404639720917,
      "rewards-max/forensics_sinkhorn_reward": 0.9590798020362854,
      "rewards-max/forensics_strict_edge_reward": 0.009202033168548951,
      "rewards-max/forensics_tiou_reward": 0.9133019149303436,
      "rewards/forensics_sinkhorn_reward": 0.6129853129386902,
      "rewards/forensics_strict_edge_reward": 0.0009375988975079963,
      "rewards/forensics_tiou_reward": 0.28373362869024277,
      "step": 14
    },
    {
      "completion_length": 19.546875,
      "epoch": 0.04608294930875576,
      "grad_norm": 4.68034553527832,
      "kl": 0.07373046875,
      "learning_rate": 9.892307692307692e-07,
      "loss": -0.0525,
      "reward": 0.8637139201164246,
      "reward_std": 0.3781227320432663,
      "reward_std_all": 0.36807334423065186,
      "rewards-max/forensics_sinkhorn_reward": 0.9381068050861359,
      "rewards-max/forensics_strict_edge_reward": 0.0728039859677665,
      "rewards-max/forensics_tiou_reward": 0.8668831586837769,
      "rewards/forensics_sinkhorn_reward": 0.6017611026763916,
      "rewards/forensics_strict_edge_reward": 0.0038808063854958164,
      "rewards/forensics_tiou_reward": 0.25807200372219086,
      "step": 15
    },
    {
      "completion_length": 19.203125,
      "epoch": 0.04915514592933948,
      "grad_norm": 4.312374114990234,
      "kl": 0.0587158203125,
      "learning_rate": 9.884615384615385e-07,
      "loss": -0.1026,
      "reward": 0.8592056035995483,
      "reward_std": 0.2657603994011879,
      "reward_std_all": 0.24574915319681168,
      "rewards-max/forensics_sinkhorn_reward": 0.8741466701030731,
      "rewards-max/forensics_strict_edge_reward": 0.16635468223830685,
      "rewards-max/forensics_tiou_reward": 0.7431319952011108,
      "rewards/forensics_sinkhorn_reward": 0.6243565082550049,
      "rewards/forensics_strict_edge_reward": 0.007242200223117834,
      "rewards/forensics_tiou_reward": 0.2276068478822708,
      "step": 16
    },
    {
      "completion_length": 20.1875,
      "epoch": 0.05222734254992319,
      "grad_norm": 3.6067636013031006,
      "kl": 0.0660400390625,
      "learning_rate": 9.876923076923076e-07,
      "loss": -0.0264,
      "reward": 0.7069980502128601,
      "reward_std": 0.25390221178531647,
      "reward_std_all": 0.24319495260715485,
      "rewards-max/forensics_sinkhorn_reward": 0.8531520664691925,
      "rewards-max/forensics_strict_edge_reward": 1.5232095559980507e-05,
      "rewards-max/forensics_tiou_reward": 0.6576937139034271,
      "rewards/forensics_sinkhorn_reward": 0.5296090841293335,
      "rewards/forensics_strict_edge_reward": 6.461589853179461e-07,
      "rewards/forensics_tiou_reward": 0.17738836258649826,
      "step": 17
    },
    {
      "completion_length": 20.078125,
      "epoch": 0.055299539170506916,
      "grad_norm": 5.145208835601807,
      "kl": 0.108642578125,
      "learning_rate": 9.86923076923077e-07,
      "loss": -0.0812,
      "reward": 0.9139134883880615,
      "reward_std": 0.30817416310310364,
      "reward_std_all": 0.2870287522673607,
      "rewards-max/forensics_sinkhorn_reward": 0.9715335965156555,
      "rewards-max/forensics_strict_edge_reward": 0.019066933017398924,
      "rewards-max/forensics_tiou_reward": 0.9400639235973358,
      "rewards/forensics_sinkhorn_reward": 0.6153468489646912,
      "rewards/forensics_strict_edge_reward": 0.0011906668925503183,
      "rewards/forensics_tiou_reward": 0.29737597703933716,
      "step": 18
    },
    {
      "completion_length": 19.609375,
      "epoch": 0.05837173579109063,
      "grad_norm": 4.0636701583862305,
      "kl": 0.14599609375,
      "learning_rate": 9.86153846153846e-07,
      "loss": 0.0628,
      "reward": 0.8065270483493805,
      "reward_std": 0.3525824099779129,
      "reward_std_all": 0.32060183584690094,
      "rewards-max/forensics_sinkhorn_reward": 0.9682606160640717,
      "rewards-max/forensics_strict_edge_reward": 0.07139093219302595,
      "rewards-max/forensics_tiou_reward": 0.9372183084487915,
      "rewards/forensics_sinkhorn_reward": 0.5684181451797485,
      "rewards/forensics_strict_edge_reward": 0.0032142994459718466,
      "rewards/forensics_tiou_reward": 0.23489461839199066,
      "step": 19
    },
    {
      "completion_length": 21.046875,
      "epoch": 0.06144393241167435,
      "grad_norm": 3.704202890396118,
      "kl": 0.099609375,
      "learning_rate": 9.853846153846153e-07,
      "loss": 0.005,
      "reward": 0.9185802638530731,
      "reward_std": 0.36118432879447937,
      "reward_std_all": 0.354316383600235,
      "rewards-max/forensics_sinkhorn_reward": 0.9599528908729553,
      "rewards-max/forensics_strict_edge_reward": 0.025429228437133133,
      "rewards-max/forensics_tiou_reward": 0.9113918542861938,
      "rewards/forensics_sinkhorn_reward": 0.6170839965343475,
      "rewards/forensics_strict_edge_reward": 0.0010621565070323413,
      "rewards/forensics_tiou_reward": 0.3004341721534729,
      "step": 20
    },
    {
      "completion_length": 19.78125,
      "epoch": 0.06451612903225806,
      "grad_norm": 6.422260284423828,
      "kl": 0.13232421875,
      "learning_rate": 9.846153846153847e-07,
      "loss": -0.021,
      "reward": 0.8247807323932648,
      "reward_std": 0.2946697920560837,
      "reward_std_all": 0.26010436564683914,
      "rewards-max/forensics_sinkhorn_reward": 0.9381836950778961,
      "rewards-max/forensics_strict_edge_reward": 0.024758148370011668,
      "rewards-max/forensics_tiou_reward": 0.8449395596981049,
      "rewards/forensics_sinkhorn_reward": 0.5831574201583862,
      "rewards/forensics_strict_edge_reward": 0.0013208964756463715,
      "rewards/forensics_tiou_reward": 0.24030238389968872,
      "step": 21
    },
    {
      "completion_length": 19.0625,
      "epoch": 0.06758832565284179,
      "grad_norm": 4.83114767074585,
      "kl": 0.103515625,
      "learning_rate": 9.838461538461538e-07,
      "loss": 0.0137,
      "reward": 0.8585453033447266,
      "reward_std": 0.2506757080554962,
      "reward_std_all": 0.2508486807346344,
      "rewards-max/forensics_sinkhorn_reward": 0.8942334353923798,
      "rewards-max/forensics_strict_edge_reward": 0.009496610902715474,
      "rewards-max/forensics_tiou_reward": 0.7598086297512054,
      "rewards/forensics_sinkhorn_reward": 0.6152245402336121,
      "rewards/forensics_strict_edge_reward": 0.00043155581079190597,
      "rewards/forensics_tiou_reward": 0.24288921803236008,
      "step": 22
    },
    {
      "completion_length": 19.15625,
      "epoch": 0.0706605222734255,
      "grad_norm": 4.884471416473389,
      "kl": 0.16796875,
      "learning_rate": 9.83076923076923e-07,
      "loss": 0.0151,
      "reward": 0.9606408774852753,
      "reward_std": 0.2394006848335266,
      "reward_std_all": 0.21563907712697983,
      "rewards-max/forensics_sinkhorn_reward": 0.9430445432662964,
      "rewards-max/forensics_strict_edge_reward": 0.02339149545878172,
      "rewards-max/forensics_tiou_reward": 0.8761608302593231,
      "rewards/forensics_sinkhorn_reward": 0.6707760095596313,
      "rewards/forensics_strict_edge_reward": 0.0009796530212042853,
      "rewards/forensics_tiou_reward": 0.2888852208852768,
      "step": 23
    },
    {
      "completion_length": 19.234375,
      "epoch": 0.07373271889400922,
      "grad_norm": 5.409490585327148,
      "kl": 0.17529296875,
      "learning_rate": 9.823076923076922e-07,
      "loss": -0.018,
      "reward": 0.8078778386116028,
      "reward_std": 0.2139895111322403,
      "reward_std_all": 0.21164482831954956,
      "rewards-max/forensics_sinkhorn_reward": 0.8099269866943359,
      "rewards-max/forensics_strict_edge_reward": 4.271993202564772e-05,
      "rewards-max/forensics_tiou_reward": 0.5791734158992767,
      "rewards/forensics_sinkhorn_reward": 0.6029402315616608,
      "rewards/forensics_strict_edge_reward": 1.8593250956655538e-06,
      "rewards/forensics_tiou_reward": 0.20493578165769577,
      "step": 24
    },
    {
      "completion_length": 19.203125,
      "epoch": 0.07680491551459294,
      "grad_norm": 4.951542377471924,
      "kl": 0.1962890625,
      "learning_rate": 9.815384615384615e-07,
      "loss": 0.1023,
      "reward": 0.7482871413230896,
      "reward_std": 0.19431743770837784,
      "reward_std_all": 0.17534756660461426,
      "rewards-max/forensics_sinkhorn_reward": 0.9180215001106262,
      "rewards-max/forensics_strict_edge_reward": 0.00803894383534498,
      "rewards-max/forensics_tiou_reward": 0.8173397481441498,
      "rewards/forensics_sinkhorn_reward": 0.5659304559230804,
      "rewards/forensics_strict_edge_reward": 0.0003349888992261185,
      "rewards/forensics_tiou_reward": 0.18202172964811325,
      "step": 25
    },
    {
      "completion_length": 19.09375,
      "epoch": 0.07987711213517665,
      "grad_norm": 4.834479808807373,
      "kl": 0.1630859375,
      "learning_rate": 9.807692307692306e-07,
      "loss": -0.0214,
      "reward": 0.886256992816925,
      "reward_std": 0.2296781837940216,
      "reward_std_all": 0.2941068857908249,
      "rewards-max/forensics_sinkhorn_reward": 0.8997719883918762,
      "rewards-max/forensics_strict_edge_reward": 0.3626402444206178,
      "rewards-max/forensics_tiou_reward": 0.8057336211204529,
      "rewards/forensics_sinkhorn_reward": 0.6242946684360504,
      "rewards/forensics_strict_edge_reward": 0.028626059443922713,
      "rewards/forensics_tiou_reward": 0.23333626985549927,
      "step": 26
    },
    {
      "completion_length": 19.375,
      "epoch": 0.08294930875576037,
      "grad_norm": 4.957857608795166,
      "kl": 0.19677734375,
      "learning_rate": 9.8e-07,
      "loss": -0.0405,
      "reward": 0.9564214050769806,
      "reward_std": 0.2161746844649315,
      "reward_std_all": 0.225605770945549,
      "rewards-max/forensics_sinkhorn_reward": 0.91444531083107,
      "rewards-max/forensics_strict_edge_reward": 0.08039581301272847,
      "rewards-max/forensics_tiou_reward": 0.8126180171966553,
      "rewards/forensics_sinkhorn_reward": 0.6331685483455658,
      "rewards/forensics_strict_edge_reward": 0.003894672277965583,
      "rewards/forensics_tiou_reward": 0.3193581849336624,
      "step": 27
    },
    {
      "completion_length": 19.140625,
      "epoch": 0.08602150537634409,
      "grad_norm": 4.5504536628723145,
      "kl": 0.13623046875,
      "learning_rate": 9.792307692307693e-07,
      "loss": -0.042,
      "reward": 0.8152008652687073,
      "reward_std": 0.15701495110988617,
      "reward_std_all": 0.16952265799045563,
      "rewards-max/forensics_sinkhorn_reward": 0.8063660860061646,
      "rewards-max/forensics_strict_edge_reward": 1.3829927099084927e-06,
      "rewards-max/forensics_tiou_reward": 0.5227465033531189,
      "rewards/forensics_sinkhorn_reward": 0.611156702041626,
      "rewards/forensics_strict_edge_reward": 5.9217745304596114e-08,
      "rewards/forensics_tiou_reward": 0.2040441408753395,
      "step": 28
    },
    {
      "completion_length": 19.328125,
      "epoch": 0.0890937019969278,
      "grad_norm": 5.669538974761963,
      "kl": 0.1845703125,
      "learning_rate": 9.784615384615384e-07,
      "loss": 0.046,
      "reward": 0.9968467354774475,
      "reward_std": 0.266853965818882,
      "reward_std_all": 0.26012779772281647,
      "rewards-max/forensics_sinkhorn_reward": 0.9035694599151611,
      "rewards-max/forensics_strict_edge_reward": 0.0019842077699649963,
      "rewards-max/forensics_tiou_reward": 0.7658933699131012,
      "rewards/forensics_sinkhorn_reward": 0.6768273413181305,
      "rewards/forensics_strict_edge_reward": 0.0001611301133550569,
      "rewards/forensics_tiou_reward": 0.3198583200573921,
      "step": 29
    },
    {
      "completion_length": 20.28125,
      "epoch": 0.09216589861751152,
      "grad_norm": 4.694350242614746,
      "kl": 0.14208984375,
      "learning_rate": 9.776923076923077e-07,
      "loss": 0.0153,
      "reward": 0.9344806671142578,
      "reward_std": 0.2375185638666153,
      "reward_std_all": 0.21913577616214752,
      "rewards-max/forensics_sinkhorn_reward": 0.9565164744853973,
      "rewards-max/forensics_strict_edge_reward": 0.3674234878271818,
      "rewards-max/forensics_tiou_reward": 0.9076099693775177,
      "rewards/forensics_sinkhorn_reward": 0.6358278393745422,
      "rewards/forensics_strict_edge_reward": 0.015372635389212519,
      "rewards/forensics_tiou_reward": 0.2832801192998886,
      "step": 30
    },
    {
      "completion_length": 19.125,
      "epoch": 0.09523809523809523,
      "grad_norm": 4.951654434204102,
      "kl": 0.19482421875,
      "learning_rate": 9.769230769230768e-07,
      "loss": -0.1329,
      "reward": 0.8683719336986542,
      "reward_std": 0.26656727492809296,
      "reward_std_all": 0.27278606593608856,
      "rewards-max/forensics_sinkhorn_reward": 0.9311478137969971,
      "rewards-max/forensics_strict_edge_reward": 0.17308608070015907,
      "rewards-max/forensics_tiou_reward": 0.8371435105800629,
      "rewards/forensics_sinkhorn_reward": 0.6060026586055756,
      "rewards/forensics_strict_edge_reward": 0.007264591429702705,
      "rewards/forensics_tiou_reward": 0.2551046684384346,
      "step": 31
    },
    {
      "completion_length": 19.03125,
      "epoch": 0.09831029185867896,
      "grad_norm": 5.369050979614258,
      "kl": 0.16455078125,
      "learning_rate": 9.761538461538461e-07,
      "loss": 0.1421,
      "reward": 0.8277509212493896,
      "reward_std": 0.22990470379590988,
      "reward_std_all": 0.215962253510952,
      "rewards-max/forensics_sinkhorn_reward": 0.8956437408924103,
      "rewards-max/forensics_strict_edge_reward": 0.0029349473188631237,
      "rewards-max/forensics_tiou_reward": 0.7795713841915131,
      "rewards/forensics_sinkhorn_reward": 0.608214795589447,
      "rewards/forensics_strict_edge_reward": 0.00013665080768987536,
      "rewards/forensics_tiou_reward": 0.21939948201179504,
      "step": 32
    },
    {
      "completion_length": 19.21875,
      "epoch": 0.10138248847926268,
      "grad_norm": 5.3632659912109375,
      "kl": 0.14794921875,
      "learning_rate": 9.753846153846154e-07,
      "loss": -0.0306,
      "reward": 0.8307405114173889,
      "reward_std": 0.18438317626714706,
      "reward_std_all": 0.20512080192565918,
      "rewards-max/forensics_sinkhorn_reward": 0.804746687412262,
      "rewards-max/forensics_strict_edge_reward": 0.012039921398354636,
      "rewards-max/forensics_tiou_reward": 0.574292004108429,
      "rewards/forensics_sinkhorn_reward": 0.600720077753067,
      "rewards/forensics_strict_edge_reward": 0.0010293510502776826,
      "rewards/forensics_tiou_reward": 0.2289910688996315,
      "step": 33
    },
    {
      "completion_length": 20.0,
      "epoch": 0.10445468509984639,
      "grad_norm": 5.353480339050293,
      "kl": 0.14208984375,
      "learning_rate": 9.746153846153845e-07,
      "loss": -0.0337,
      "reward": 1.0426818132400513,
      "reward_std": 0.2966972291469574,
      "reward_std_all": 0.2702902853488922,
      "rewards-max/forensics_sinkhorn_reward": 0.9104199707508087,
      "rewards-max/forensics_strict_edge_reward": 0.04421532154083252,
      "rewards-max/forensics_tiou_reward": 0.7575789988040924,
      "rewards/forensics_sinkhorn_reward": 0.6847296059131622,
      "rewards/forensics_strict_edge_reward": 0.004437216965015978,
      "rewards/forensics_tiou_reward": 0.35351499915122986,
      "step": 34
    },
    {
      "completion_length": 19.9375,
      "epoch": 0.10752688172043011,
      "grad_norm": 5.14887809753418,
      "kl": 0.18505859375,
      "learning_rate": 9.738461538461539e-07,
      "loss": 0.0637,
      "reward": 0.8242895007133484,
      "reward_std": 0.22100578993558884,
      "reward_std_all": 0.22390542179346085,
      "rewards-max/forensics_sinkhorn_reward": 0.7770891487598419,
      "rewards-max/forensics_strict_edge_reward": 0.0038220062851905823,
      "rewards-max/forensics_tiou_reward": 0.5666883289813995,
      "rewards/forensics_sinkhorn_reward": 0.5916140079498291,
      "rewards/forensics_strict_edge_reward": 0.00023394118761643767,
      "rewards/forensics_tiou_reward": 0.23244158923625946,
      "step": 35
    },
    {
      "completion_length": 19.28125,
      "epoch": 0.11059907834101383,
      "grad_norm": 5.905935764312744,
      "kl": 0.15185546875,
      "learning_rate": 9.73076923076923e-07,
      "loss": 0.1685,
      "reward": 0.8585297465324402,
      "reward_std": 0.19994689524173737,
      "reward_std_all": 0.18508072197437286,
      "rewards-max/forensics_sinkhorn_reward": 0.9070494174957275,
      "rewards-max/forensics_strict_edge_reward": 0.0007039923834781803,
      "rewards-max/forensics_tiou_reward": 0.7787026464939117,
      "rewards/forensics_sinkhorn_reward": 0.6068915128707886,
      "rewards/forensics_strict_edge_reward": 2.977227781642e-05,
      "rewards/forensics_tiou_reward": 0.2516084834933281,
      "step": 36
    },
    {
      "completion_length": 19.140625,
      "epoch": 0.11367127496159754,
      "grad_norm": 4.5906195640563965,
      "kl": 0.14697265625,
      "learning_rate": 9.723076923076923e-07,
      "loss": -0.0227,
      "reward": 0.8429761826992035,
      "reward_std": 0.23241007328033447,
      "reward_std_all": 0.24841301888227463,
      "rewards-max/forensics_sinkhorn_reward": 0.9595628678798676,
      "rewards-max/forensics_strict_edge_reward": 0.503368973499164,
      "rewards-max/forensics_tiou_reward": 0.9204715490341187,
      "rewards/forensics_sinkhorn_reward": 0.5672642886638641,
      "rewards/forensics_strict_edge_reward": 0.021017703867983073,
      "rewards/forensics_tiou_reward": 0.2546941712498665,
      "step": 37
    },
    {
      "completion_length": 19.078125,
      "epoch": 0.11674347158218126,
      "grad_norm": 4.658812999725342,
      "kl": 0.249755859375,
      "learning_rate": 9.715384615384616e-07,
      "loss": 0.0841,
      "reward": 1.0285819470882416,
      "reward_std": 0.32981185615062714,
      "reward_std_all": 0.33040569722652435,
      "rewards-max/forensics_sinkhorn_reward": 0.9343140125274658,
      "rewards-max/forensics_strict_edge_reward": 0.051529765129089355,
      "rewards-max/forensics_tiou_reward": 0.8740149438381195,
      "rewards/forensics_sinkhorn_reward": 0.6798595786094666,
      "rewards/forensics_strict_edge_reward": 0.0031834993278607726,
      "rewards/forensics_tiou_reward": 0.3455388993024826,
      "step": 38
    },
    {
      "completion_length": 19.171875,
      "epoch": 0.11981566820276497,
      "grad_norm": 4.764664173126221,
      "kl": 0.12255859375,
      "learning_rate": 9.707692307692307e-07,
      "loss": -0.1179,
      "reward": 1.0156236290931702,
      "reward_std": 0.2015138790011406,
      "reward_std_all": 0.2052026391029358,
      "rewards-max/forensics_sinkhorn_reward": 0.8976016342639923,
      "rewards-max/forensics_strict_edge_reward": 0.16327794641256332,
      "rewards-max/forensics_tiou_reward": 0.7466294467449188,
      "rewards/forensics_sinkhorn_reward": 0.6724447906017303,
      "rewards/forensics_strict_edge_reward": 0.007868582324590534,
      "rewards/forensics_tiou_reward": 0.335310235619545,
      "step": 39
    },
    {
      "completion_length": 19.046875,
      "epoch": 0.1228878648233487,
      "grad_norm": 5.313331604003906,
      "kl": 0.15869140625,
      "learning_rate": 9.7e-07,
      "loss": -0.0591,
      "reward": 0.9478538334369659,
      "reward_std": 0.21361741423606873,
      "reward_std_all": 0.20763930678367615,
      "rewards-max/forensics_sinkhorn_reward": 0.9547297954559326,
      "rewards-max/forensics_strict_edge_reward": 0.048390737269073725,
      "rewards-max/forensics_tiou_reward": 0.8965029418468475,
      "rewards/forensics_sinkhorn_reward": 0.644230306148529,
      "rewards/forensics_strict_edge_reward": 0.002077209879644215,
      "rewards/forensics_tiou_reward": 0.3015463352203369,
      "step": 40
    },
    {
      "completion_length": 20.890625,
      "epoch": 0.1259600614439324,
      "grad_norm": 6.065368175506592,
      "kl": 0.1748046875,
      "learning_rate": 9.692307692307691e-07,
      "loss": -0.0172,
      "reward": 0.8553125262260437,
      "reward_std": 0.29145829379558563,
      "reward_std_all": 0.25659775733947754,
      "rewards-max/forensics_sinkhorn_reward": 0.9066718220710754,
      "rewards-max/forensics_strict_edge_reward": 0.3868401274085045,
      "rewards-max/forensics_tiou_reward": 0.7836086750030518,
      "rewards/forensics_sinkhorn_reward": 0.5870758295059204,
      "rewards/forensics_strict_edge_reward": 0.02197613571479451,
      "rewards/forensics_tiou_reward": 0.24626056849956512,
      "step": 41
    },
    {
      "completion_length": 19.03125,
      "epoch": 0.12903225806451613,
      "grad_norm": 5.24353551864624,
      "kl": 0.140625,
      "learning_rate": 9.684615384615385e-07,
      "loss": -0.1146,
      "reward": 1.044357806444168,
      "reward_std": 0.28008291125297546,
      "reward_std_all": 0.2864759564399719,
      "rewards-max/forensics_sinkhorn_reward": 0.9242747128009796,
      "rewards-max/forensics_strict_edge_reward": 0.11481890268623829,
      "rewards-max/forensics_tiou_reward": 0.8318684995174408,
      "rewards/forensics_sinkhorn_reward": 0.6854090094566345,
      "rewards/forensics_strict_edge_reward": 0.006524569820612669,
      "rewards/forensics_tiou_reward": 0.3524242639541626,
      "step": 42
    },
    {
      "completion_length": 19.328125,
      "epoch": 0.13210445468509985,
      "grad_norm": 4.885237693786621,
      "kl": 0.17041015625,
      "learning_rate": 9.676923076923078e-07,
      "loss": -0.0505,
      "reward": 0.9116848707199097,
      "reward_std": 0.22217541933059692,
      "reward_std_all": 0.19993788748979568,
      "rewards-max/forensics_sinkhorn_reward": 0.8179903626441956,
      "rewards-max/forensics_strict_edge_reward": 0.05066506378352642,
      "rewards-max/forensics_tiou_reward": 0.6933289170265198,
      "rewards/forensics_sinkhorn_reward": 0.6207060217857361,
      "rewards/forensics_strict_edge_reward": 0.0030348835280165076,
      "rewards/forensics_tiou_reward": 0.28794391453266144,
      "step": 43
    },
    {
      "completion_length": 19.296875,
      "epoch": 0.13517665130568357,
      "grad_norm": 5.458635330200195,
      "kl": 0.109619140625,
      "learning_rate": 9.669230769230769e-07,
      "loss": -0.1303,
      "reward": 0.881513774394989,
      "reward_std": 0.29807303845882416,
      "reward_std_all": 0.292207732796669,
      "rewards-max/forensics_sinkhorn_reward": 0.9350305199623108,
      "rewards-max/forensics_strict_edge_reward": 0.042580432098475285,
      "rewards-max/forensics_tiou_reward": 0.8360885679721832,
      "rewards/forensics_sinkhorn_reward": 0.6130360960960388,
      "rewards/forensics_strict_edge_reward": 0.0024220748082370847,
      "rewards/forensics_tiou_reward": 0.2660555988550186,
      "step": 44
    },
    {
      "completion_length": 19.15625,
      "epoch": 0.1382488479262673,
      "grad_norm": 4.243495464324951,
      "kl": 0.125,
      "learning_rate": 9.66153846153846e-07,
      "loss": -0.0217,
      "reward": 1.0244876742362976,
      "reward_std": 0.22278174757957458,
      "reward_std_all": 0.22436872124671936,
      "rewards-max/forensics_sinkhorn_reward": 0.9192622005939484,
      "rewards-max/forensics_strict_edge_reward": 0.009518958628177643,
      "rewards-max/forensics_tiou_reward": 0.7969870567321777,
      "rewards/forensics_sinkhorn_reward": 0.6776541173458099,
      "rewards/forensics_strict_edge_reward": 0.0010360133019275963,
      "rewards/forensics_tiou_reward": 0.3457975387573242,
      "step": 45
    },
    {
      "completion_length": 19.21875,
      "epoch": 0.141321044546851,
      "grad_norm": 7.028076648712158,
      "kl": 0.2109375,
      "learning_rate": 9.653846153846153e-07,
      "loss": -0.0854,
      "reward": 0.8871306478977203,
      "reward_std": 0.32480786740779877,
      "reward_std_all": 0.29230256378650665,
      "rewards-max/forensics_sinkhorn_reward": 0.924386590719223,
      "rewards-max/forensics_strict_edge_reward": 0.11130063980817795,
      "rewards-max/forensics_tiou_reward": 0.8317457735538483,
      "rewards/forensics_sinkhorn_reward": 0.6090422570705414,
      "rewards/forensics_strict_edge_reward": 0.005357485031709075,
      "rewards/forensics_tiou_reward": 0.27273086458444595,
      "step": 46
    },
    {
      "completion_length": 19.359375,
      "epoch": 0.1443932411674347,
      "grad_norm": 5.42559814453125,
      "kl": 0.17041015625,
      "learning_rate": 9.646153846153846e-07,
      "loss": -0.0991,
      "reward": 0.9489152729511261,
      "reward_std": 0.34742239117622375,
      "reward_std_all": 0.3055984079837799,
      "rewards-max/forensics_sinkhorn_reward": 0.9664946496486664,
      "rewards-max/forensics_strict_edge_reward": 0.21625080704689026,
      "rewards-max/forensics_tiou_reward": 0.9348981082439423,
      "rewards/forensics_sinkhorn_reward": 0.6095836758613586,
      "rewards/forensics_strict_edge_reward": 0.018606911413371563,
      "rewards/forensics_tiou_reward": 0.32072463631629944,
      "step": 47
    },
    {
      "completion_length": 19.234375,
      "epoch": 0.14746543778801843,
      "grad_norm": 5.769023418426514,
      "kl": 0.13916015625,
      "learning_rate": 9.638461538461537e-07,
      "loss": -0.1138,
      "reward": 0.9344372153282166,
      "reward_std": 0.2339879870414734,
      "reward_std_all": 0.25539158284664154,
      "rewards-max/forensics_sinkhorn_reward": 0.8827491700649261,
      "rewards-max/forensics_strict_edge_reward": 0.0102200866676867,
      "rewards-max/forensics_tiou_reward": 0.7283811569213867,
      "rewards/forensics_sinkhorn_reward": 0.6416096687316895,
      "rewards/forensics_strict_edge_reward": 0.000530346849700436,
      "rewards/forensics_tiou_reward": 0.2922971844673157,
      "step": 48
    },
    {
      "completion_length": 19.546875,
      "epoch": 0.15053763440860216,
      "grad_norm": 6.825312614440918,
      "kl": 0.113525390625,
      "learning_rate": 9.63076923076923e-07,
      "loss": 0.0347,
      "reward": 0.8111413717269897,
      "reward_std": 0.23028089106082916,
      "reward_std_all": 0.20442970097064972,
      "rewards-max/forensics_sinkhorn_reward": 0.8974554240703583,
      "rewards-max/forensics_strict_edge_reward": 0.18444087158422917,
      "rewards-max/forensics_tiou_reward": 0.7633512020111084,
      "rewards/forensics_sinkhorn_reward": 0.5740707814693451,
      "rewards/forensics_strict_edge_reward": 0.007685282042075414,
      "rewards/forensics_tiou_reward": 0.22938527911901474,
      "step": 49
    },
    {
      "completion_length": 19.484375,
      "epoch": 0.15360983102918588,
      "grad_norm": 5.11923885345459,
      "kl": 0.23291015625,
      "learning_rate": 9.623076923076922e-07,
      "loss": 0.0202,
      "reward": 0.8476293683052063,
      "reward_std": 0.33114269375801086,
      "reward_std_all": 0.29986003041267395,
      "rewards-max/forensics_sinkhorn_reward": 0.9365185499191284,
      "rewards-max/forensics_strict_edge_reward": 0.027307508513331413,
      "rewards-max/forensics_tiou_reward": 0.8575373589992523,
      "rewards/forensics_sinkhorn_reward": 0.5647765398025513,
      "rewards/forensics_strict_edge_reward": 0.0015921647427603602,
      "rewards/forensics_tiou_reward": 0.2812606319785118,
      "step": 50
    },
    {
      "completion_length": 19.296875,
      "epoch": 0.15668202764976957,
      "grad_norm": 5.026761054992676,
      "kl": 0.162109375,
      "learning_rate": 9.615384615384615e-07,
      "loss": -0.0786,
      "reward": 0.7638880610466003,
      "reward_std": 0.21814942359924316,
      "reward_std_all": 0.1997159942984581,
      "rewards-max/forensics_sinkhorn_reward": 0.8363581299781799,
      "rewards-max/forensics_strict_edge_reward": 0.016621627259155503,
      "rewards-max/forensics_tiou_reward": 0.6152920424938202,
      "rewards/forensics_sinkhorn_reward": 0.5671547055244446,
      "rewards/forensics_strict_edge_reward": 0.00070141305945981,
      "rewards/forensics_tiou_reward": 0.19603195041418076,
      "step": 51
    },
    {
      "completion_length": 19.078125,
      "epoch": 0.1597542242703533,
      "grad_norm": 5.138802528381348,
      "kl": 0.1748046875,
      "learning_rate": 9.607692307692308e-07,
      "loss": -0.1432,
      "reward": 0.9450220763683319,
      "reward_std": 0.23975758999586105,
      "reward_std_all": 0.2400246039032936,
      "rewards-max/forensics_sinkhorn_reward": 0.9472363293170929,
      "rewards-max/forensics_strict_edge_reward": 0.15833484381437302,
      "rewards-max/forensics_tiou_reward": 0.859541267156601,
      "rewards/forensics_sinkhorn_reward": 0.6188942790031433,
      "rewards/forensics_strict_edge_reward": 0.007220983970910311,
      "rewards/forensics_tiou_reward": 0.31890687346458435,
      "step": 52
    },
    {
      "completion_length": 19.140625,
      "epoch": 0.16282642089093702,
      "grad_norm": 4.7364630699157715,
      "kl": 0.110595703125,
      "learning_rate": 9.6e-07,
      "loss": -0.0228,
      "reward": 1.115693747997284,
      "reward_std": 0.29475128650665283,
      "reward_std_all": 0.25620799511671066,
      "rewards-max/forensics_sinkhorn_reward": 0.9731894731521606,
      "rewards-max/forensics_strict_edge_reward": 0.6499901115894318,
      "rewards-max/forensics_tiou_reward": 0.9506785571575165,
      "rewards/forensics_sinkhorn_reward": 0.6848466396331787,
      "rewards/forensics_strict_edge_reward": 0.044081903994083405,
      "rewards/forensics_tiou_reward": 0.3867652118206024,
      "step": 53
    },
    {
      "completion_length": 19.1875,
      "epoch": 0.16589861751152074,
      "grad_norm": 5.031655311584473,
      "kl": 0.135498046875,
      "learning_rate": 9.592307692307692e-07,
      "loss": 0.1164,
      "reward": 0.9303083419799805,
      "reward_std": 0.301890566945076,
      "reward_std_all": 0.2816984951496124,
      "rewards-max/forensics_sinkhorn_reward": 0.8701339662075043,
      "rewards-max/forensics_strict_edge_reward": 0.002136149210855365,
      "rewards-max/forensics_tiou_reward": 0.6786919236183167,
      "rewards/forensics_sinkhorn_reward": 0.6396964192390442,
      "rewards/forensics_strict_edge_reward": 0.00010780482261907309,
      "rewards/forensics_tiou_reward": 0.2905040979385376,
      "step": 54
    },
    {
      "completion_length": 19.75,
      "epoch": 0.16897081413210446,
      "grad_norm": 4.330155372619629,
      "kl": 0.11083984375,
      "learning_rate": 9.584615384615383e-07,
      "loss": 0.0164,
      "reward": 0.8526396155357361,
      "reward_std": 0.2790137827396393,
      "reward_std_all": 0.25478648394346237,
      "rewards-max/forensics_sinkhorn_reward": 0.872437596321106,
      "rewards-max/forensics_strict_edge_reward": 0.08426333032548428,
      "rewards-max/forensics_tiou_reward": 0.7598047852516174,
      "rewards/forensics_sinkhorn_reward": 0.5941063463687897,
      "rewards/forensics_strict_edge_reward": 0.003824868123047054,
      "rewards/forensics_tiou_reward": 0.2547083720564842,
      "step": 55
    },
    {
      "completion_length": 19.3125,
      "epoch": 0.17204301075268819,
      "grad_norm": 5.363908767700195,
      "kl": 0.1865234375,
      "learning_rate": 9.576923076923077e-07,
      "loss": 0.0357,
      "reward": 0.9571260213851929,
      "reward_std": 0.28013336658477783,
      "reward_std_all": 0.28033825755119324,
      "rewards-max/forensics_sinkhorn_reward": 0.9342572093009949,
      "rewards-max/forensics_strict_edge_reward": 0.010494248213944957,
      "rewards-max/forensics_tiou_reward": 0.8359178304672241,
      "rewards/forensics_sinkhorn_reward": 0.6362133026123047,
      "rewards/forensics_strict_edge_reward": 0.00045828252405044623,
      "rewards/forensics_tiou_reward": 0.3204544931650162,
      "step": 56
    },
    {
      "completion_length": 19.0625,
      "epoch": 0.17511520737327188,
      "grad_norm": 5.390691757202148,
      "kl": 0.080810546875,
      "learning_rate": 9.569230769230768e-07,
      "loss": 0.0671,
      "reward": 0.9842076301574707,
      "reward_std": 0.26435981690883636,
      "reward_std_all": 0.24719099700450897,
      "rewards-max/forensics_sinkhorn_reward": 0.8633989095687866,
      "rewards-max/forensics_strict_edge_reward": 0.08967849239706993,
      "rewards-max/forensics_tiou_reward": 0.7295335531234741,
      "rewards/forensics_sinkhorn_reward": 0.6554993093013763,
      "rewards/forensics_strict_edge_reward": 0.004381208564154804,
      "rewards/forensics_tiou_reward": 0.32432714104652405,
      "step": 57
    },
    {
      "completion_length": 19.484375,
      "epoch": 0.1781874039938556,
      "grad_norm": 5.104494094848633,
      "kl": 0.082275390625,
      "learning_rate": 9.56153846153846e-07,
      "loss": -0.0513,
      "reward": 0.9309871196746826,
      "reward_std": 0.26362794637680054,
      "reward_std_all": 0.25390128046274185,
      "rewards-max/forensics_sinkhorn_reward": 0.9245500564575195,
      "rewards-max/forensics_strict_edge_reward": 0.0883494276415604,
      "rewards-max/forensics_tiou_reward": 0.7856815457344055,
      "rewards/forensics_sinkhorn_reward": 0.6207777559757233,
      "rewards/forensics_strict_edge_reward": 0.010378134321989307,
      "rewards/forensics_tiou_reward": 0.29983121156692505,
      "step": 58
    },
    {
      "completion_length": 19.25,
      "epoch": 0.18125960061443933,
      "grad_norm": 6.342766284942627,
      "kl": 0.11328125,
      "learning_rate": 9.553846153846154e-07,
      "loss": 0.1152,
      "reward": 0.8551617562770844,
      "reward_std": 0.3627282381057739,
      "reward_std_all": 0.312929704785347,
      "rewards-max/forensics_sinkhorn_reward": 0.9595339894294739,
      "rewards-max/forensics_strict_edge_reward": 0.017482548020780087,
      "rewards-max/forensics_tiou_reward": 0.9138956964015961,
      "rewards/forensics_sinkhorn_reward": 0.5722448229789734,
      "rewards/forensics_strict_edge_reward": 0.001217572600580752,
      "rewards/forensics_tiou_reward": 0.28169938921928406,
      "step": 59
    },
    {
      "completion_length": 19.46875,
      "epoch": 0.18433179723502305,
      "grad_norm": 5.436454772949219,
      "kl": 0.122314453125,
      "learning_rate": 9.546153846153845e-07,
      "loss": -0.0103,
      "reward": 0.7586705684661865,
      "reward_std": 0.23342455923557281,
      "reward_std_all": 0.22296389937400818,
      "rewards-max/forensics_sinkhorn_reward": 0.7855973541736603,
      "rewards-max/forensics_strict_edge_reward": 0.18293729022201433,
      "rewards-max/forensics_tiou_reward": 0.5230916291475296,
      "rewards/forensics_sinkhorn_reward": 0.5219547748565674,
      "rewards/forensics_strict_edge_reward": 0.019870372359226618,
      "rewards/forensics_tiou_reward": 0.21684543415904045,
      "step": 60
    },
    {
      "completion_length": 19.3125,
      "epoch": 0.18740399385560677,
      "grad_norm": 4.848219871520996,
      "kl": 0.167236328125,
      "learning_rate": 9.538461538461538e-07,
      "loss": 0.0522,
      "reward": 0.8099565804004669,
      "reward_std": 0.20747268199920654,
      "reward_std_all": 0.19082709401845932,
      "rewards-max/forensics_sinkhorn_reward": 0.8590027689933777,
      "rewards-max/forensics_strict_edge_reward": 0.026489595429666224,
      "rewards-max/forensics_tiou_reward": 0.7122114300727844,
      "rewards/forensics_sinkhorn_reward": 0.5765396952629089,
      "rewards/forensics_strict_edge_reward": 0.0016753427285038924,
      "rewards/forensics_tiou_reward": 0.2317415177822113,
      "step": 61
    },
    {
      "completion_length": 19.40625,
      "epoch": 0.19047619047619047,
      "grad_norm": 5.346248626708984,
      "kl": 0.21044921875,
      "learning_rate": 9.53076923076923e-07,
      "loss": -0.0128,
      "reward": 0.9342429339885712,
      "reward_std": 0.2429877147078514,
      "reward_std_all": 0.23872444033622742,
      "rewards-max/forensics_sinkhorn_reward": 0.9261718690395355,
      "rewards-max/forensics_strict_edge_reward": 0.032208616845309734,
      "rewards-max/forensics_tiou_reward": 0.8333982825279236,
      "rewards/forensics_sinkhorn_reward": 0.6294356286525726,
      "rewards/forensics_strict_edge_reward": 0.0019101412617601454,
      "rewards/forensics_tiou_reward": 0.302897185087204,
      "step": 62
    },
    {
      "completion_length": 19.1875,
      "epoch": 0.1935483870967742,
      "grad_norm": 4.506579399108887,
      "kl": 0.203369140625,
      "learning_rate": 9.523076923076923e-07,
      "loss": -0.085,
      "reward": 0.9389093220233917,
      "reward_std": 0.2748565152287483,
      "reward_std_all": 0.25504227727651596,
      "rewards-max/forensics_sinkhorn_reward": 0.9492859244346619,
      "rewards-max/forensics_strict_edge_reward": 0.301801897585392,
      "rewards-max/forensics_tiou_reward": 0.9068838357925415,
      "rewards/forensics_sinkhorn_reward": 0.6106933355331421,
      "rewards/forensics_strict_edge_reward": 0.020919273607432842,
      "rewards/forensics_tiou_reward": 0.30729664862155914,
      "step": 63
    },
    {
      "completion_length": 19.265625,
      "epoch": 0.1966205837173579,
      "grad_norm": 6.438655853271484,
      "kl": 0.113525390625,
      "learning_rate": 9.515384615384615e-07,
      "loss": 0.0316,
      "reward": 1.1378422975540161,
      "reward_std": 0.36753590404987335,
      "reward_std_all": 0.339749276638031,
      "rewards-max/forensics_sinkhorn_reward": 0.9596495032310486,
      "rewards-max/forensics_strict_edge_reward": 0.0933635625988245,
      "rewards-max/forensics_tiou_reward": 0.9042367339134216,
      "rewards/forensics_sinkhorn_reward": 0.7120653986930847,
      "rewards/forensics_strict_edge_reward": 0.006062930915504694,
      "rewards/forensics_tiou_reward": 0.41971398890018463,
      "step": 64
    },
    {
      "completion_length": 19.1875,
      "epoch": 0.19969278033794163,
      "grad_norm": 4.909124851226807,
      "kl": 0.198486328125,
      "learning_rate": 9.507692307692308e-07,
      "loss": -0.0205,
      "reward": 1.1545263528823853,
      "reward_std": 0.2634302154183388,
      "reward_std_all": 0.24275141209363937,
      "rewards-max/forensics_sinkhorn_reward": 0.9168272614479065,
      "rewards-max/forensics_strict_edge_reward": 0.335579325677827,
      "rewards-max/forensics_tiou_reward": 0.7991918921470642,
      "rewards/forensics_sinkhorn_reward": 0.7261722683906555,
      "rewards/forensics_strict_edge_reward": 0.015043876192066818,
      "rewards/forensics_tiou_reward": 0.4133101999759674,
      "step": 65
    },
    {
      "completion_length": 19.328125,
      "epoch": 0.20276497695852536,
      "grad_norm": 5.798024654388428,
      "kl": 0.150390625,
      "learning_rate": 9.499999999999999e-07,
      "loss": 0.01,
      "reward": 0.9100905358791351,
      "reward_std": 0.21614310145378113,
      "reward_std_all": 0.19403119385242462,
      "rewards-max/forensics_sinkhorn_reward": 0.8722825050354004,
      "rewards-max/forensics_strict_edge_reward": 0.014245092403143644,
      "rewards-max/forensics_tiou_reward": 0.7471856772899628,
      "rewards/forensics_sinkhorn_reward": 0.6257389485836029,
      "rewards/forensics_strict_edge_reward": 0.0013456355081871152,
      "rewards/forensics_tiou_reward": 0.2830059379339218,
      "step": 66
    },
    {
      "completion_length": 19.046875,
      "epoch": 0.20583717357910905,
      "grad_norm": 5.23105001449585,
      "kl": 0.142822265625,
      "learning_rate": 9.492307692307691e-07,
      "loss": -0.0522,
      "reward": 0.8978174924850464,
      "reward_std": 0.2716006934642792,
      "reward_std_all": 0.28049346059560776,
      "rewards-max/forensics_sinkhorn_reward": 0.8975782692432404,
      "rewards-max/forensics_strict_edge_reward": 0.11986838193843141,
      "rewards-max/forensics_tiou_reward": 0.760513961315155,
      "rewards/forensics_sinkhorn_reward": 0.6185930371284485,
      "rewards/forensics_strict_edge_reward": 0.005029190155255492,
      "rewards/forensics_tiou_reward": 0.27419526875019073,
      "step": 67
    },
    {
      "completion_length": 19.140625,
      "epoch": 0.20890937019969277,
      "grad_norm": 5.751281261444092,
      "kl": 0.1416015625,
      "learning_rate": 9.484615384615384e-07,
      "loss": -0.028,
      "reward": 1.009433388710022,
      "reward_std": 0.2366953194141388,
      "reward_std_all": 0.21699056774377823,
      "rewards-max/forensics_sinkhorn_reward": 0.9569561183452606,
      "rewards-max/forensics_strict_edge_reward": 0.1751270443201065,
      "rewards-max/forensics_tiou_reward": 0.9069613516330719,
      "rewards/forensics_sinkhorn_reward": 0.6574613153934479,
      "rewards/forensics_strict_edge_reward": 0.008966323919594288,
      "rewards/forensics_tiou_reward": 0.34300582110881805,
      "step": 68
    },
    {
      "completion_length": 19.015625,
      "epoch": 0.2119815668202765,
      "grad_norm": 5.098416328430176,
      "kl": 0.10986328125,
      "learning_rate": 9.476923076923077e-07,
      "loss": 0.0882,
      "reward": 1.011414796113968,
      "reward_std": 0.33056697249412537,
      "reward_std_all": 0.3187883347272873,
      "rewards-max/forensics_sinkhorn_reward": 0.9675859212875366,
      "rewards-max/forensics_strict_edge_reward": 0.345868531614542,
      "rewards-max/forensics_tiou_reward": 0.9407563209533691,
      "rewards/forensics_sinkhorn_reward": 0.6572649776935577,
      "rewards/forensics_strict_edge_reward": 0.015724516240879893,
      "rewards/forensics_tiou_reward": 0.33842532336711884,
      "step": 69
    },
    {
      "completion_length": 19.984375,
      "epoch": 0.21505376344086022,
      "grad_norm": 5.095010280609131,
      "kl": 0.14990234375,
      "learning_rate": 9.46923076923077e-07,
      "loss": -0.1095,
      "reward": 0.9269570410251617,
      "reward_std": 0.292869508266449,
      "reward_std_all": 0.2710065692663193,
      "rewards-max/forensics_sinkhorn_reward": 0.9193789958953857,
      "rewards-max/forensics_strict_edge_reward": 0.3359076395863667,
      "rewards-max/forensics_tiou_reward": 0.797895222902298,
      "rewards/forensics_sinkhorn_reward": 0.6173526048660278,
      "rewards/forensics_strict_edge_reward": 0.014569931925507262,
      "rewards/forensics_tiou_reward": 0.2950344830751419,
      "step": 70
    },
    {
      "completion_length": 19.21875,
      "epoch": 0.21812596006144394,
      "grad_norm": 4.83811616897583,
      "kl": 0.22265625,
      "learning_rate": 9.461538461538461e-07,
      "loss": -0.0382,
      "reward": 0.8986694812774658,
      "reward_std": 0.2852112352848053,
      "reward_std_all": 0.2700361832976341,
      "rewards-max/forensics_sinkhorn_reward": 0.9521690905094147,
      "rewards-max/forensics_strict_edge_reward": 0.03406575135886669,
      "rewards-max/forensics_tiou_reward": 0.9122794568538666,
      "rewards/forensics_sinkhorn_reward": 0.6152835488319397,
      "rewards/forensics_strict_edge_reward": 0.00204790556745138,
      "rewards/forensics_tiou_reward": 0.28133805096149445,
      "step": 71
    },
    {
      "completion_length": 19.453125,
      "epoch": 0.22119815668202766,
      "grad_norm": 4.204245090484619,
      "kl": 0.1181640625,
      "learning_rate": 9.453846153846153e-07,
      "loss": -0.0684,
      "reward": 0.9176974296569824,
      "reward_std": 0.24449359625577927,
      "reward_std_all": 0.21727200597524643,
      "rewards-max/forensics_sinkhorn_reward": 0.9059486389160156,
      "rewards-max/forensics_strict_edge_reward": 0.8016124665737152,
      "rewards-max/forensics_tiou_reward": 0.831986278295517,
      "rewards/forensics_sinkhorn_reward": 0.5899870097637177,
      "rewards/forensics_strict_edge_reward": 0.05127178877592087,
      "rewards/forensics_tiou_reward": 0.2764386385679245,
      "step": 72
    },
    {
      "completion_length": 19.1875,
      "epoch": 0.22427035330261136,
      "grad_norm": 5.216245651245117,
      "kl": 0.215087890625,
      "learning_rate": 9.446153846153846e-07,
      "loss": 0.0211,
      "reward": 0.9908809661865234,
      "reward_std": 0.3063606917858124,
      "reward_std_all": 0.29860591888427734,
      "rewards-max/forensics_sinkhorn_reward": 0.9619231522083282,
      "rewards-max/forensics_strict_edge_reward": 0.3981185480952263,
      "rewards-max/forensics_tiou_reward": 0.916305273771286,
      "rewards/forensics_sinkhorn_reward": 0.652643084526062,
      "rewards/forensics_strict_edge_reward": 0.01900824811309576,
      "rewards/forensics_tiou_reward": 0.3192295879125595,
      "step": 73
    },
    {
      "completion_length": 18.953125,
      "epoch": 0.22734254992319508,
      "grad_norm": 4.990727424621582,
      "kl": 0.1357421875,
      "learning_rate": 9.438461538461538e-07,
      "loss": -0.0125,
      "reward": 0.9101403057575226,
      "reward_std": 0.2470199316740036,
      "reward_std_all": 0.22297830879688263,
      "rewards-max/forensics_sinkhorn_reward": 0.9066725671291351,
      "rewards-max/forensics_strict_edge_reward": 0.07248864942812361,
      "rewards-max/forensics_tiou_reward": 0.793015331029892,
      "rewards/forensics_sinkhorn_reward": 0.6047999858856201,
      "rewards/forensics_strict_edge_reward": 0.00562110456667142,
      "rewards/forensics_tiou_reward": 0.2997192293405533,
      "step": 74
    },
    {
      "completion_length": 19.0,
      "epoch": 0.2304147465437788,
      "grad_norm": 5.407229423522949,
      "kl": 0.150390625,
      "learning_rate": 9.43076923076923e-07,
      "loss": -0.0228,
      "reward": 0.9526098370552063,
      "reward_std": 0.21731212735176086,
      "reward_std_all": 0.243439182639122,
      "rewards-max/forensics_sinkhorn_reward": 0.9204513132572174,
      "rewards-max/forensics_strict_edge_reward": 0.014757162425667048,
      "rewards-max/forensics_tiou_reward": 0.7941985726356506,
      "rewards/forensics_sinkhorn_reward": 0.6445048749446869,
      "rewards/forensics_strict_edge_reward": 0.0008211873209802434,
      "rewards/forensics_tiou_reward": 0.30728374421596527,
      "step": 75
    },
    {
      "completion_length": 19.015625,
      "epoch": 0.23348694316436253,
      "grad_norm": 7.5841569900512695,
      "kl": 0.13134765625,
      "learning_rate": 9.423076923076923e-07,
      "loss": -0.0927,
      "reward": 0.9241547882556915,
      "reward_std": 0.3614913374185562,
      "reward_std_all": 0.3250711262226105,
      "rewards-max/forensics_sinkhorn_reward": 0.985633134841919,
      "rewards-max/forensics_strict_edge_reward": 0.586972177028656,
      "rewards-max/forensics_tiou_reward": 0.9878042042255402,
      "rewards/forensics_sinkhorn_reward": 0.615020751953125,
      "rewards/forensics_strict_edge_reward": 0.03739806264638901,
      "rewards/forensics_tiou_reward": 0.2717360258102417,
      "step": 76
    },
    {
      "completion_length": 19.3125,
      "epoch": 0.23655913978494625,
      "grad_norm": 5.612007141113281,
      "kl": 0.11376953125,
      "learning_rate": 9.415384615384615e-07,
      "loss": -0.1475,
      "reward": 0.8385519981384277,
      "reward_std": 0.41030609607696533,
      "reward_std_all": 0.3628003001213074,
      "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": 0.5510081946849823,
      "rewards/forensics_strict_edge_reward": 0.05664115212857723,
      "rewards/forensics_tiou_reward": 0.23090268671512604,
      "step": 77
    },
    {
      "completion_length": 19.1875,
      "epoch": 0.23963133640552994,
      "grad_norm": 5.885386943817139,
      "kl": 0.109375,
      "learning_rate": 9.407692307692308e-07,
      "loss": -0.0514,
      "reward": 1.1282707452774048,
      "reward_std": 0.3970319628715515,
      "reward_std_all": 0.3546053618192673,
      "rewards-max/forensics_sinkhorn_reward": 0.9626392126083374,
      "rewards-max/forensics_strict_edge_reward": 0.03775855712592602,
      "rewards-max/forensics_tiou_reward": 0.9214553534984589,
      "rewards/forensics_sinkhorn_reward": 0.7110689580440521,
      "rewards/forensics_strict_edge_reward": 0.003435261663980782,
      "rewards/forensics_tiou_reward": 0.4137665331363678,
      "step": 78
    },
    {
      "completion_length": 19.40625,
      "epoch": 0.24270353302611367,
      "grad_norm": 5.059311389923096,
      "kl": 0.136474609375,
      "learning_rate": 9.399999999999999e-07,
      "loss": 0.0402,
      "reward": 0.8847492337226868,
      "reward_std": 0.3111356198787689,
      "reward_std_all": 0.2795242965221405,
      "rewards-max/forensics_sinkhorn_reward": 0.9172548353672028,
      "rewards-max/forensics_strict_edge_reward": 0.030213441234081984,
      "rewards-max/forensics_tiou_reward": 0.8036067187786102,
      "rewards/forensics_sinkhorn_reward": 0.609281599521637,
      "rewards/forensics_strict_edge_reward": 0.002152335728169419,
      "rewards/forensics_tiou_reward": 0.2733153700828552,
      "step": 79
    },
    {
      "completion_length": 19.125,
      "epoch": 0.2457757296466974,
      "grad_norm": 4.246232986450195,
      "kl": 0.10791015625,
      "learning_rate": 9.392307692307691e-07,
      "loss": -0.1282,
      "reward": 1.1581348776817322,
      "reward_std": 0.23887043446302414,
      "reward_std_all": 0.22825326025485992,
      "rewards-max/forensics_sinkhorn_reward": 0.9712150394916534,
      "rewards-max/forensics_strict_edge_reward": 0.2747939005494118,
      "rewards-max/forensics_tiou_reward": 0.9482986629009247,
      "rewards/forensics_sinkhorn_reward": 0.706951379776001,
      "rewards/forensics_strict_edge_reward": 0.013219952583312988,
      "rewards/forensics_tiou_reward": 0.4379635453224182,
      "step": 80
    },
    {
      "completion_length": 19.375,
      "epoch": 0.2488479262672811,
      "grad_norm": 7.227939605712891,
      "kl": 0.14453125,
      "learning_rate": 9.384615384615384e-07,
      "loss": 0.043,
      "reward": 0.9796050488948822,
      "reward_std": 0.33386892080307007,
      "reward_std_all": 0.34608694911003113,
      "rewards-max/forensics_sinkhorn_reward": 0.9652935564517975,
      "rewards-max/forensics_strict_edge_reward": 0.16732843965291977,
      "rewards-max/forensics_tiou_reward": 0.9354350566864014,
      "rewards/forensics_sinkhorn_reward": 0.6425445973873138,
      "rewards/forensics_strict_edge_reward": 0.010117985773831606,
      "rewards/forensics_tiou_reward": 0.3269423842430115,
      "step": 81
    },
    {
      "completion_length": 19.265625,
      "epoch": 0.2519201228878648,
      "grad_norm": 5.572041034698486,
      "kl": 0.13525390625,
      "learning_rate": 9.376923076923076e-07,
      "loss": -0.0054,
      "reward": 0.9723597168922424,
      "reward_std": 0.2526901960372925,
      "reward_std_all": 0.27732478082180023,
      "rewards-max/forensics_sinkhorn_reward": 0.9624883830547333,
      "rewards-max/forensics_strict_edge_reward": 0.0231928308676288,
      "rewards-max/forensics_tiou_reward": 0.9299704432487488,
      "rewards/forensics_sinkhorn_reward": 0.6328075528144836,
      "rewards/forensics_strict_edge_reward": 0.0009930193668878928,
      "rewards/forensics_tiou_reward": 0.3385591208934784,
      "step": 82
    },
    {
      "completion_length": 19.0625,
      "epoch": 0.25499231950844853,
      "grad_norm": 5.16548490524292,
      "kl": 0.10400390625,
      "learning_rate": 9.36923076923077e-07,
      "loss": -0.0017,
      "reward": 0.9485182166099548,
      "reward_std": 0.25032587349414825,
      "reward_std_all": 0.2468234971165657,
      "rewards-max/forensics_sinkhorn_reward": 0.8649631142616272,
      "rewards-max/forensics_strict_edge_reward": 0.5000012843228205,
      "rewards-max/forensics_tiou_reward": 0.7953786849975586,
      "rewards/forensics_sinkhorn_reward": 0.6011195480823517,
      "rewards/forensics_strict_edge_reward": 0.038942539207312166,
      "rewards/forensics_tiou_reward": 0.308456115424633,
      "step": 83
    },
    {
      "completion_length": 19.078125,
      "epoch": 0.25806451612903225,
      "grad_norm": 5.953417778015137,
      "kl": 0.14208984375,
      "learning_rate": 9.361538461538461e-07,
      "loss": -0.0013,
      "reward": 0.932685136795044,
      "reward_std": 0.33560582995414734,
      "reward_std_all": 0.3200185298919678,
      "rewards-max/forensics_sinkhorn_reward": 0.9771774113178253,
      "rewards-max/forensics_strict_edge_reward": 0.20560714229941368,
      "rewards-max/forensics_tiou_reward": 0.9644850194454193,
      "rewards/forensics_sinkhorn_reward": 0.634329080581665,
      "rewards/forensics_strict_edge_reward": 0.014193225186318159,
      "rewards/forensics_tiou_reward": 0.28416289389133453,
      "step": 84
    },
    {
      "completion_length": 18.859375,
      "epoch": 0.261136712749616,
      "grad_norm": 5.371397018432617,
      "kl": 0.15380859375,
      "learning_rate": 9.353846153846153e-07,
      "loss": 0.0486,
      "reward": 0.94362011551857,
      "reward_std": 0.24585320055484772,
      "reward_std_all": 0.2450777217745781,
      "rewards-max/forensics_sinkhorn_reward": 0.8771450817584991,
      "rewards-max/forensics_strict_edge_reward": 0.34321367740631104,
      "rewards-max/forensics_tiou_reward": 0.7726679146289825,
      "rewards/forensics_sinkhorn_reward": 0.6191577613353729,
      "rewards/forensics_strict_edge_reward": 0.0159237552434206,
      "rewards/forensics_tiou_reward": 0.30853863060474396,
      "step": 85
    },
    {
      "completion_length": 19.40625,
      "epoch": 0.2642089093701997,
      "grad_norm": 5.343618392944336,
      "kl": 0.13037109375,
      "learning_rate": 9.346153846153846e-07,
      "loss": 0.0788,
      "reward": 1.1042668223381042,
      "reward_std": 0.19392997026443481,
      "reward_std_all": 0.1721450239419937,
      "rewards-max/forensics_sinkhorn_reward": 0.8804748952388763,
      "rewards-max/forensics_strict_edge_reward": 0.07773343566805124,
      "rewards-max/forensics_tiou_reward": 0.7893716394901276,
      "rewards/forensics_sinkhorn_reward": 0.6923770308494568,
      "rewards/forensics_strict_edge_reward": 0.003765359695535153,
      "rewards/forensics_tiou_reward": 0.4081243574619293,
      "step": 86
    },
    {
      "completion_length": 19.25,
      "epoch": 0.2672811059907834,
      "grad_norm": 6.169074535369873,
      "kl": 0.299072265625,
      "learning_rate": 9.338461538461538e-07,
      "loss": 0.0237,
      "reward": 1.0148341655731201,
      "reward_std": 0.19333531707525253,
      "reward_std_all": 0.17973137646913528,
      "rewards-max/forensics_sinkhorn_reward": 0.9534726738929749,
      "rewards-max/forensics_strict_edge_reward": 0.4277711436152458,
      "rewards-max/forensics_tiou_reward": 0.8971284329891205,
      "rewards/forensics_sinkhorn_reward": 0.6305919885635376,
      "rewards/forensics_strict_edge_reward": 0.022808031179010868,
      "rewards/forensics_tiou_reward": 0.36143407225608826,
      "step": 87
    },
    {
      "completion_length": 19.0625,
      "epoch": 0.27035330261136714,
      "grad_norm": 6.739053726196289,
      "kl": 0.121337890625,
      "learning_rate": 9.33076923076923e-07,
      "loss": 0.015,
      "reward": 1.0310188829898834,
      "reward_std": 0.31068848818540573,
      "reward_std_all": 0.2778128311038017,
      "rewards-max/forensics_sinkhorn_reward": 0.9793905913829803,
      "rewards-max/forensics_strict_edge_reward": 0.28814417496323586,
      "rewards-max/forensics_tiou_reward": 0.9738164842128754,
      "rewards/forensics_sinkhorn_reward": 0.6443840265274048,
      "rewards/forensics_strict_edge_reward": 0.014385964488610625,
      "rewards/forensics_tiou_reward": 0.37224890291690826,
      "step": 88
    },
    {
      "completion_length": 19.03125,
      "epoch": 0.27342549923195086,
      "grad_norm": 5.286235332489014,
      "kl": 0.219970703125,
      "learning_rate": 9.323076923076922e-07,
      "loss": 0.148,
      "reward": 1.0048745274543762,
      "reward_std": 0.3104122579097748,
      "reward_std_all": 0.27369359135627747,
      "rewards-max/forensics_sinkhorn_reward": 0.956335723400116,
      "rewards-max/forensics_strict_edge_reward": 0.3371981307864189,
      "rewards-max/forensics_tiou_reward": 0.9134330153465271,
      "rewards/forensics_sinkhorn_reward": 0.6300609707832336,
      "rewards/forensics_strict_edge_reward": 0.03601721627637744,
      "rewards/forensics_tiou_reward": 0.33879633247852325,
      "step": 89
    },
    {
      "completion_length": 19.015625,
      "epoch": 0.2764976958525346,
      "grad_norm": 6.6271491050720215,
      "kl": 0.16064453125,
      "learning_rate": 9.315384615384615e-07,
      "loss": 0.057,
      "reward": 1.0023368895053864,
      "reward_std": 0.2607201635837555,
      "reward_std_all": 0.2334292083978653,
      "rewards-max/forensics_sinkhorn_reward": 0.9052378535270691,
      "rewards-max/forensics_strict_edge_reward": 0.2763826996088028,
      "rewards-max/forensics_tiou_reward": 0.8111026585102081,
      "rewards/forensics_sinkhorn_reward": 0.6377447247505188,
      "rewards/forensics_strict_edge_reward": 0.0203274330124259,
      "rewards/forensics_tiou_reward": 0.34426476061344147,
      "step": 90
    },
    {
      "completion_length": 19.03125,
      "epoch": 0.27956989247311825,
      "grad_norm": 6.33404541015625,
      "kl": 0.17236328125,
      "learning_rate": 9.307692307692308e-07,
      "loss": -0.0631,
      "reward": 1.0961567163467407,
      "reward_std": 0.20810484886169434,
      "reward_std_all": 0.22305752336978912,
      "rewards-max/forensics_sinkhorn_reward": 0.9664648175239563,
      "rewards-max/forensics_strict_edge_reward": 0.06318584457039833,
      "rewards-max/forensics_tiou_reward": 0.9308746755123138,
      "rewards/forensics_sinkhorn_reward": 0.7050102949142456,
      "rewards/forensics_strict_edge_reward": 0.0039395412895828485,
      "rewards/forensics_tiou_reward": 0.38720688223838806,
      "step": 91
    },
    {
      "completion_length": 19.3125,
      "epoch": 0.282642089093702,
      "grad_norm": 6.68434476852417,
      "kl": 0.195068359375,
      "learning_rate": 9.3e-07,
      "loss": -0.0892,
      "reward": 0.9043582379817963,
      "reward_std": 0.2109789252281189,
      "reward_std_all": 0.18743257224559784,
      "rewards-max/forensics_sinkhorn_reward": 0.9508930146694183,
      "rewards-max/forensics_strict_edge_reward": 0.17492867028340697,
      "rewards-max/forensics_tiou_reward": 0.8838011622428894,
      "rewards/forensics_sinkhorn_reward": 0.5974376499652863,
      "rewards/forensics_strict_edge_reward": 0.010336996972910129,
      "rewards/forensics_tiou_reward": 0.2965835779905319,
      "step": 92
    },
    {
      "completion_length": 19.140625,
      "epoch": 0.2857142857142857,
      "grad_norm": 5.556509017944336,
      "kl": 0.18603515625,
      "learning_rate": 9.292307692307692e-07,
      "loss": -0.0832,
      "reward": 0.9000044763088226,
      "reward_std": 0.27468280494213104,
      "reward_std_all": 0.25033819675445557,
      "rewards-max/forensics_sinkhorn_reward": 0.9713300168514252,
      "rewards-max/forensics_strict_edge_reward": 0.07842736225575209,
      "rewards-max/forensics_tiou_reward": 0.9528107941150665,
      "rewards/forensics_sinkhorn_reward": 0.612089067697525,
      "rewards/forensics_strict_edge_reward": 0.004981175879947841,
      "rewards/forensics_tiou_reward": 0.2829342186450958,
      "step": 93
    },
    {
      "completion_length": 19.015625,
      "epoch": 0.2887864823348694,
      "grad_norm": 5.607344627380371,
      "kl": 0.17626953125,
      "learning_rate": 9.284615384615384e-07,
      "loss": -0.017,
      "reward": 0.9364007413387299,
      "reward_std": 0.2778007835149765,
      "reward_std_all": 0.24373354017734528,
      "rewards-max/forensics_sinkhorn_reward": 0.910840630531311,
      "rewards-max/forensics_strict_edge_reward": 0.05605424381792545,
      "rewards-max/forensics_tiou_reward": 0.8057403266429901,
      "rewards/forensics_sinkhorn_reward": 0.6246724426746368,
      "rewards/forensics_strict_edge_reward": 0.0032289018854498863,
      "rewards/forensics_tiou_reward": 0.3084994703531265,
      "step": 94
    },
    {
      "completion_length": 19.140625,
      "epoch": 0.29185867895545314,
      "grad_norm": 6.405000686645508,
      "kl": 0.125244140625,
      "learning_rate": 9.276923076923076e-07,
      "loss": 0.1239,
      "reward": 0.9648754894733429,
      "reward_std": 0.23072335124015808,
      "reward_std_all": 0.20362213253974915,
      "rewards-max/forensics_sinkhorn_reward": 0.8956080973148346,
      "rewards-max/forensics_strict_edge_reward": 0.015879779239185154,
      "rewards-max/forensics_tiou_reward": 0.7889309525489807,
      "rewards/forensics_sinkhorn_reward": 0.6523380875587463,
      "rewards/forensics_strict_edge_reward": 0.0008979966514743865,
      "rewards/forensics_tiou_reward": 0.3116394281387329,
      "step": 95
    },
    {
      "completion_length": 19.15625,
      "epoch": 0.29493087557603687,
      "grad_norm": 6.005867004394531,
      "kl": 0.14404296875,
      "learning_rate": 9.26923076923077e-07,
      "loss": -0.0488,
      "reward": 0.8920484483242035,
      "reward_std": 0.2727344334125519,
      "reward_std_all": 0.26352445781230927,
      "rewards-max/forensics_sinkhorn_reward": 0.9431682825088501,
      "rewards-max/forensics_strict_edge_reward": 0.3754238709807396,
      "rewards-max/forensics_tiou_reward": 0.8627500832080841,
      "rewards/forensics_sinkhorn_reward": 0.5954826772212982,
      "rewards/forensics_strict_edge_reward": 0.017250283737666905,
      "rewards/forensics_tiou_reward": 0.27931544184684753,
      "step": 96
    },
    {
      "completion_length": 19.171875,
      "epoch": 0.2980030721966206,
      "grad_norm": 5.366506576538086,
      "kl": 0.1787109375,
      "learning_rate": 9.261538461538461e-07,
      "loss": -0.0719,
      "reward": 1.172013521194458,
      "reward_std": 0.3129006028175354,
      "reward_std_all": 0.2788335382938385,
      "rewards-max/forensics_sinkhorn_reward": 0.9458170235157013,
      "rewards-max/forensics_strict_edge_reward": 0.15473409369587898,
      "rewards-max/forensics_tiou_reward": 0.8868560791015625,
      "rewards/forensics_sinkhorn_reward": 0.7170080244541168,
      "rewards/forensics_strict_edge_reward": 0.013755455380305648,
      "rewards/forensics_tiou_reward": 0.4412500560283661,
      "step": 97
    },
    {
      "completion_length": 19.03125,
      "epoch": 0.3010752688172043,
      "grad_norm": 6.070049285888672,
      "kl": 0.104248046875,
      "learning_rate": 9.253846153846153e-07,
      "loss": -0.0059,
      "reward": 0.9403181374073029,
      "reward_std": 0.21539296209812164,
      "reward_std_all": 0.19711323827505112,
      "rewards-max/forensics_sinkhorn_reward": 0.8367884457111359,
      "rewards-max/forensics_strict_edge_reward": 0.45930445194244385,
      "rewards-max/forensics_tiou_reward": 0.6528144925832748,
      "rewards/forensics_sinkhorn_reward": 0.6116916239261627,
      "rewards/forensics_strict_edge_reward": 0.023839603178203106,
      "rewards/forensics_tiou_reward": 0.30478687584400177,
      "step": 98
    },
    {
      "completion_length": 19.515625,
      "epoch": 0.30414746543778803,
      "grad_norm": 5.682151794433594,
      "kl": 0.240234375,
      "learning_rate": 9.246153846153846e-07,
      "loss": -0.0478,
      "reward": 0.9272012114524841,
      "reward_std": 0.18439310789108276,
      "reward_std_all": 0.17962007969617844,
      "rewards-max/forensics_sinkhorn_reward": 0.9541292488574982,
      "rewards-max/forensics_strict_edge_reward": 0.08866321199457161,
      "rewards-max/forensics_tiou_reward": 0.9050864279270172,
      "rewards/forensics_sinkhorn_reward": 0.6193450391292572,
      "rewards/forensics_strict_edge_reward": 0.004130512694246136,
      "rewards/forensics_tiou_reward": 0.30372560024261475,
      "step": 99
    },
    {
      "completion_length": 19.0,
      "epoch": 0.30721966205837176,
      "grad_norm": 5.219364643096924,
      "kl": 0.21875,
      "learning_rate": 9.238461538461538e-07,
      "loss": 0.0947,
      "reward": 0.9360751807689667,
      "reward_std": 0.2309674471616745,
      "reward_std_all": 0.2206253483891487,
      "rewards-max/forensics_sinkhorn_reward": 0.957131952047348,
      "rewards-max/forensics_strict_edge_reward": 0.2449169121682644,
      "rewards-max/forensics_tiou_reward": 0.9202726185321808,
      "rewards/forensics_sinkhorn_reward": 0.631866842508316,
      "rewards/forensics_strict_edge_reward": 0.01131347962655127,
      "rewards/forensics_tiou_reward": 0.2928948253393173,
      "step": 100
    },
    {
      "completion_length": 19.15625,
      "epoch": 0.3102918586789555,
      "grad_norm": 9.521459579467773,
      "kl": 0.31103515625,
      "learning_rate": 9.230769230769231e-07,
      "loss": -0.0595,
      "reward": 0.8580002784729004,
      "reward_std": 0.25572778284549713,
      "reward_std_all": 0.2333662211894989,
      "rewards-max/forensics_sinkhorn_reward": 0.8596631288528442,
      "rewards-max/forensics_strict_edge_reward": 0.07738083520234795,
      "rewards-max/forensics_tiou_reward": 0.6931422054767609,
      "rewards/forensics_sinkhorn_reward": 0.5848348140716553,
      "rewards/forensics_strict_edge_reward": 0.0075373519539425615,
      "rewards/forensics_tiou_reward": 0.2656281441450119,
      "step": 101
    },
    {
      "completion_length": 18.890625,
      "epoch": 0.31336405529953915,
      "grad_norm": 6.198897838592529,
      "kl": 0.136474609375,
      "learning_rate": 9.223076923076922e-07,
      "loss": 0.0694,
      "reward": 0.9967775642871857,
      "reward_std": 0.23073160648345947,
      "reward_std_all": 0.22909530997276306,
      "rewards-max/forensics_sinkhorn_reward": 0.9386425912380219,
      "rewards-max/forensics_strict_edge_reward": 0.10106844397523673,
      "rewards-max/forensics_tiou_reward": 0.8395865261554718,
      "rewards/forensics_sinkhorn_reward": 0.6405297219753265,
      "rewards/forensics_strict_edge_reward": 0.010436339406624029,
      "rewards/forensics_tiou_reward": 0.34581147134304047,
      "step": 102
    },
    {
      "completion_length": 18.9375,
      "epoch": 0.31643625192012287,
      "grad_norm": 5.357475280761719,
      "kl": 0.123779296875,
      "learning_rate": 9.215384615384614e-07,
      "loss": -0.0913,
      "reward": 0.9103232324123383,
      "reward_std": 0.17867138236761093,
      "reward_std_all": 0.21229980885982513,
      "rewards-max/forensics_sinkhorn_reward": 0.8324035704135895,
      "rewards-max/forensics_strict_edge_reward": 0.04499996663071215,
      "rewards-max/forensics_tiou_reward": 0.7050040364265442,
      "rewards/forensics_sinkhorn_reward": 0.6183197498321533,
      "rewards/forensics_strict_edge_reward": 0.0019697198731591925,
      "rewards/forensics_tiou_reward": 0.290033757686615,
      "step": 103
    },
    {
      "completion_length": 19.1875,
      "epoch": 0.3195084485407066,
      "grad_norm": 6.618043422698975,
      "kl": 0.14501953125,
      "learning_rate": 9.207692307692308e-07,
      "loss": -0.0881,
      "reward": 0.8699685335159302,
      "reward_std": 0.2566346824169159,
      "reward_std_all": 0.2561202943325043,
      "rewards-max/forensics_sinkhorn_reward": 0.9213335812091827,
      "rewards-max/forensics_strict_edge_reward": 0.12609066464938223,
      "rewards-max/forensics_tiou_reward": 0.79073166847229,
      "rewards/forensics_sinkhorn_reward": 0.6092631220817566,
      "rewards/forensics_strict_edge_reward": 0.006273817038163543,
      "rewards/forensics_tiou_reward": 0.2544315904378891,
      "step": 104
    },
    {
      "completion_length": 19.25,
      "epoch": 0.3225806451612903,
      "grad_norm": 5.426355838775635,
      "kl": 0.134765625,
      "learning_rate": 9.2e-07,
      "loss": -0.1371,
      "reward": 1.057445228099823,
      "reward_std": 0.20286690443754196,
      "reward_std_all": 0.20854083448648453,
      "rewards-max/forensics_sinkhorn_reward": 0.9464547038078308,
      "rewards-max/forensics_strict_edge_reward": 0.3668328719213605,
      "rewards-max/forensics_tiou_reward": 0.8886255621910095,
      "rewards/forensics_sinkhorn_reward": 0.673149049282074,
      "rewards/forensics_strict_edge_reward": 0.017823343514464796,
      "rewards/forensics_tiou_reward": 0.3664728254079819,
      "step": 105
    },
    {
      "completion_length": 20.46875,
      "epoch": 0.32565284178187404,
      "grad_norm": 5.414739608764648,
      "kl": 0.22216796875,
      "learning_rate": 9.192307692307692e-07,
      "loss": 0.0392,
      "reward": 0.8872845768928528,
      "reward_std": 0.28650428354740143,
      "reward_std_all": 0.25728651881217957,
      "rewards-max/forensics_sinkhorn_reward": 0.9771691858768463,
      "rewards-max/forensics_strict_edge_reward": 0.36576345656067133,
      "rewards-max/forensics_tiou_reward": 0.9668279886245728,
      "rewards/forensics_sinkhorn_reward": 0.6038046181201935,
      "rewards/forensics_strict_edge_reward": 0.015298484649974853,
      "rewards/forensics_tiou_reward": 0.2681814655661583,
      "step": 106
    },
    {
      "completion_length": 21.046875,
      "epoch": 0.32872503840245776,
      "grad_norm": 5.285342216491699,
      "kl": 0.16162109375,
      "learning_rate": 9.184615384615384e-07,
      "loss": -0.0432,
      "reward": 1.0828341841697693,
      "reward_std": 0.3429826647043228,
      "reward_std_all": 0.3085297644138336,
      "rewards-max/forensics_sinkhorn_reward": 0.9877024590969086,
      "rewards-max/forensics_strict_edge_reward": 0.8053101599216461,
      "rewards-max/forensics_tiou_reward": 0.9953836798667908,
      "rewards/forensics_sinkhorn_reward": 0.6428065001964569,
      "rewards/forensics_strict_edge_reward": 0.038507526740431786,
      "rewards/forensics_tiou_reward": 0.40152016282081604,
      "step": 107
    },
    {
      "completion_length": 19.25,
      "epoch": 0.3317972350230415,
      "grad_norm": 5.309720039367676,
      "kl": 0.11767578125,
      "learning_rate": 9.176923076923076e-07,
      "loss": -0.0873,
      "reward": 1.1410881876945496,
      "reward_std": 0.2862361967563629,
      "reward_std_all": 0.2596883624792099,
      "rewards-max/forensics_sinkhorn_reward": 0.9412653148174286,
      "rewards-max/forensics_strict_edge_reward": 0.1516507789492607,
      "rewards-max/forensics_tiou_reward": 0.8806153535842896,
      "rewards/forensics_sinkhorn_reward": 0.7062397599220276,
      "rewards/forensics_strict_edge_reward": 0.008051601238548756,
      "rewards/forensics_tiou_reward": 0.4267967939376831,
      "step": 108
    },
    {
      "completion_length": 19.640625,
      "epoch": 0.3348694316436252,
      "grad_norm": 6.4041218757629395,
      "kl": 0.13232421875,
      "learning_rate": 9.169230769230769e-07,
      "loss": -0.0081,
      "reward": 0.8911643922328949,
      "reward_std": 0.2829720824956894,
      "reward_std_all": 0.25886739790439606,
      "rewards-max/forensics_sinkhorn_reward": 0.952325850725174,
      "rewards-max/forensics_strict_edge_reward": 0.3598749848315492,
      "rewards-max/forensics_tiou_reward": 0.8855545818805695,
      "rewards/forensics_sinkhorn_reward": 0.5882969498634338,
      "rewards/forensics_strict_edge_reward": 0.015350530666182749,
      "rewards/forensics_tiou_reward": 0.28751689195632935,
      "step": 109
    },
    {
      "completion_length": 19.25,
      "epoch": 0.3379416282642089,
      "grad_norm": 7.194064140319824,
      "kl": 0.135986328125,
      "learning_rate": 9.161538461538462e-07,
      "loss": 0.0222,
      "reward": 1.00270277261734,
      "reward_std": 0.21110796928405762,
      "reward_std_all": 0.21029691398143768,
      "rewards-max/forensics_sinkhorn_reward": 0.934584230184555,
      "rewards-max/forensics_strict_edge_reward": 0.13865773589350283,
      "rewards-max/forensics_tiou_reward": 0.8624293804168701,
      "rewards/forensics_sinkhorn_reward": 0.6581851243972778,
      "rewards/forensics_strict_edge_reward": 0.007524957065470517,
      "rewards/forensics_tiou_reward": 0.336992546916008,
      "step": 110
    },
    {
      "completion_length": 19.140625,
      "epoch": 0.34101382488479265,
      "grad_norm": 8.165924072265625,
      "kl": 0.158447265625,
      "learning_rate": 9.153846153846153e-07,
      "loss": -0.0458,
      "reward": 1.0800963342189789,
      "reward_std": 0.29808154702186584,
      "reward_std_all": 0.27820412814617157,
      "rewards-max/forensics_sinkhorn_reward": 0.9802142977714539,
      "rewards-max/forensics_strict_edge_reward": 0.5247012078762054,
      "rewards-max/forensics_tiou_reward": 0.9730556309223175,
      "rewards/forensics_sinkhorn_reward": 0.6739360094070435,
      "rewards/forensics_strict_edge_reward": 0.034266313537955284,
      "rewards/forensics_tiou_reward": 0.371893972158432,
      "step": 111
    },
    {
      "completion_length": 19.21875,
      "epoch": 0.34408602150537637,
      "grad_norm": 5.348781585693359,
      "kl": 0.13916015625,
      "learning_rate": 9.146153846153846e-07,
      "loss": -0.1457,
      "reward": 1.1887706518173218,
      "reward_std": 0.28522129356861115,
      "reward_std_all": 0.24657711386680603,
      "rewards-max/forensics_sinkhorn_reward": 0.9434261322021484,
      "rewards-max/forensics_strict_edge_reward": 0.4690144043415785,
      "rewards-max/forensics_tiou_reward": 0.8768547177314758,
      "rewards/forensics_sinkhorn_reward": 0.705710768699646,
      "rewards/forensics_strict_edge_reward": 0.035194510128349066,
      "rewards/forensics_tiou_reward": 0.4478653073310852,
      "step": 112
    },
    {
      "completion_length": 19.40625,
      "epoch": 0.34715821812596004,
      "grad_norm": 5.8183512687683105,
      "kl": 0.19140625,
      "learning_rate": 9.138461538461538e-07,
      "loss": -0.0258,
      "reward": 1.1037973463535309,
      "reward_std": 0.22900722175836563,
      "reward_std_all": 0.21535298228263855,
      "rewards-max/forensics_sinkhorn_reward": 0.9818207025527954,
      "rewards-max/forensics_strict_edge_reward": 0.12386709917336702,
      "rewards-max/forensics_tiou_reward": 0.9792949557304382,
      "rewards/forensics_sinkhorn_reward": 0.6932863891124725,
      "rewards/forensics_strict_edge_reward": 0.007341539196204394,
      "rewards/forensics_tiou_reward": 0.4031693637371063,
      "step": 113
    },
    {
      "completion_length": 19.234375,
      "epoch": 0.35023041474654376,
      "grad_norm": 6.744665145874023,
      "kl": 0.16943359375,
      "learning_rate": 9.130769230769231e-07,
      "loss": -0.0592,
      "reward": 0.8609116077423096,
      "reward_std": 0.1993100866675377,
      "reward_std_all": 0.19737914949655533,
      "rewards-max/forensics_sinkhorn_reward": 0.8603391945362091,
      "rewards-max/forensics_strict_edge_reward": 0.06830165069550276,
      "rewards-max/forensics_tiou_reward": 0.7029723227024078,
      "rewards/forensics_sinkhorn_reward": 0.5926028192043304,
      "rewards/forensics_strict_edge_reward": 0.002963915641885251,
      "rewards/forensics_tiou_reward": 0.265344962477684,
      "step": 114
    },
    {
      "completion_length": 19.390625,
      "epoch": 0.3533026113671275,
      "grad_norm": 4.790302753448486,
      "kl": 0.12890625,
      "learning_rate": 9.123076923076922e-07,
      "loss": -0.0439,
      "reward": 0.8427823185920715,
      "reward_std": 0.20294032990932465,
      "reward_std_all": 0.19253776967525482,
      "rewards-max/forensics_sinkhorn_reward": 0.7973790168762207,
      "rewards-max/forensics_strict_edge_reward": 0.24158136546611786,
      "rewards-max/forensics_tiou_reward": 0.6430414915084839,
      "rewards/forensics_sinkhorn_reward": 0.5851653516292572,
      "rewards/forensics_strict_edge_reward": 0.017924337182193995,
      "rewards/forensics_tiou_reward": 0.23969262093305588,
      "step": 115
    },
    {
      "completion_length": 19.25,
      "epoch": 0.3563748079877112,
      "grad_norm": 5.603209972381592,
      "kl": 0.260498046875,
      "learning_rate": 9.115384615384614e-07,
      "loss": 0.0557,
      "reward": 0.9650189280509949,
      "reward_std": 0.1701780930161476,
      "reward_std_all": 0.1642649993300438,
      "rewards-max/forensics_sinkhorn_reward": 0.9477307498455048,
      "rewards-max/forensics_strict_edge_reward": 0.05298168899025768,
      "rewards-max/forensics_tiou_reward": 0.8937591910362244,
      "rewards/forensics_sinkhorn_reward": 0.6320716142654419,
      "rewards/forensics_strict_edge_reward": 0.003919664261047728,
      "rewards/forensics_tiou_reward": 0.3290276974439621,
      "step": 116
    },
    {
      "completion_length": 18.890625,
      "epoch": 0.35944700460829493,
      "grad_norm": 16.22139549255371,
      "kl": 0.72412109375,
      "learning_rate": 9.107692307692308e-07,
      "loss": 0.0991,
      "reward": 0.7877576351165771,
      "reward_std": 0.24497489631175995,
      "reward_std_all": 0.21082878857851028,
      "rewards-max/forensics_sinkhorn_reward": 0.9054366946220398,
      "rewards-max/forensics_strict_edge_reward": 0.07958457246422768,
      "rewards-max/forensics_tiou_reward": 0.7884410321712494,
      "rewards/forensics_sinkhorn_reward": 0.5505239963531494,
      "rewards/forensics_strict_edge_reward": 0.0036154407425783575,
      "rewards/forensics_tiou_reward": 0.23361817002296448,
      "step": 117
    },
    {
      "completion_length": 19.421875,
      "epoch": 0.36251920122887865,
      "grad_norm": 6.536604404449463,
      "kl": 0.12744140625,
      "learning_rate": 9.1e-07,
      "loss": 0.0764,
      "reward": 1.030934602022171,
      "reward_std": 0.20331746339797974,
      "reward_std_all": 0.18054219335317612,
      "rewards-max/forensics_sinkhorn_reward": 0.9258135259151459,
      "rewards-max/forensics_strict_edge_reward": 0.0452746357768774,
      "rewards-max/forensics_tiou_reward": 0.8520755767822266,
      "rewards/forensics_sinkhorn_reward": 0.6584377884864807,
      "rewards/forensics_strict_edge_reward": 0.001969014178030193,
      "rewards/forensics_tiou_reward": 0.37052787840366364,
      "step": 118
    },
    {
      "completion_length": 19.125,
      "epoch": 0.3655913978494624,
      "grad_norm": 6.30157470703125,
      "kl": 0.25048828125,
      "learning_rate": 9.092307692307692e-07,
      "loss": 0.0308,
      "reward": 1.0839614868164062,
      "reward_std": 0.3029961585998535,
      "reward_std_all": 0.29494985938072205,
      "rewards-max/forensics_sinkhorn_reward": 0.9741857945919037,
      "rewards-max/forensics_strict_edge_reward": 0.3885071277618408,
      "rewards-max/forensics_tiou_reward": 0.9618492126464844,
      "rewards/forensics_sinkhorn_reward": 0.6848989129066467,
      "rewards/forensics_strict_edge_reward": 0.025761359371244907,
      "rewards/forensics_tiou_reward": 0.3733011782169342,
      "step": 119
    },
    {
      "completion_length": 19.359375,
      "epoch": 0.3686635944700461,
      "grad_norm": 8.620915412902832,
      "kl": 0.20068359375,
      "learning_rate": 9.084615384615384e-07,
      "loss": 0.0545,
      "reward": 0.9049635529518127,
      "reward_std": 0.29573188722133636,
      "reward_std_all": 0.2845572680234909,
      "rewards-max/forensics_sinkhorn_reward": 0.9217961132526398,
      "rewards-max/forensics_strict_edge_reward": 0.08529012836515903,
      "rewards-max/forensics_tiou_reward": 0.8069508373737335,
      "rewards/forensics_sinkhorn_reward": 0.6183691918849945,
      "rewards/forensics_strict_edge_reward": 0.003835746174445376,
      "rewards/forensics_tiou_reward": 0.28275860846042633,
      "step": 120
    },
    {
      "completion_length": 19.265625,
      "epoch": 0.3717357910906298,
      "grad_norm": 5.650326728820801,
      "kl": 0.1328125,
      "learning_rate": 9.076923076923076e-07,
      "loss": 0.1616,
      "reward": 0.9151427149772644,
      "reward_std": 0.20642579346895218,
      "reward_std_all": 0.18233665078878403,
      "rewards-max/forensics_sinkhorn_reward": 0.9344210922718048,
      "rewards-max/forensics_strict_edge_reward": 0.19689108431339264,
      "rewards-max/forensics_tiou_reward": 0.8629539608955383,
      "rewards/forensics_sinkhorn_reward": 0.6062424182891846,
      "rewards/forensics_strict_edge_reward": 0.009881343692541122,
      "rewards/forensics_tiou_reward": 0.2990189492702484,
      "step": 121
    },
    {
      "completion_length": 19.078125,
      "epoch": 0.37480798771121354,
      "grad_norm": 5.237515926361084,
      "kl": 0.122314453125,
      "learning_rate": 9.069230769230769e-07,
      "loss": -0.0309,
      "reward": 1.1784392595291138,
      "reward_std": 0.33406467735767365,
      "reward_std_all": 0.31359831988811493,
      "rewards-max/forensics_sinkhorn_reward": 0.9391234815120697,
      "rewards-max/forensics_strict_edge_reward": 0.49555233120918274,
      "rewards-max/forensics_tiou_reward": 0.886853814125061,
      "rewards/forensics_sinkhorn_reward": 0.6943714618682861,
      "rewards/forensics_strict_edge_reward": 0.04796487092971802,
      "rewards/forensics_tiou_reward": 0.43610289692878723,
      "step": 122
    },
    {
      "completion_length": 19.671875,
      "epoch": 0.3778801843317972,
      "grad_norm": 6.096288681030273,
      "kl": 0.3017578125,
      "learning_rate": 9.061538461538461e-07,
      "loss": 0.0835,
      "reward": 1.053637683391571,
      "reward_std": 0.2805280461907387,
      "reward_std_all": 0.25740962475538254,
      "rewards-max/forensics_sinkhorn_reward": 0.9067163169384003,
      "rewards-max/forensics_strict_edge_reward": 0.3467242270708084,
      "rewards-max/forensics_tiou_reward": 0.7684359550476074,
      "rewards/forensics_sinkhorn_reward": 0.6709196865558624,
      "rewards/forensics_strict_edge_reward": 0.01608571643009782,
      "rewards/forensics_tiou_reward": 0.36663223803043365,
      "step": 123
    },
    {
      "completion_length": 19.328125,
      "epoch": 0.38095238095238093,
      "grad_norm": 9.438824653625488,
      "kl": 0.19580078125,
      "learning_rate": 9.053846153846153e-07,
      "loss": -0.0193,
      "reward": 1.0675503313541412,
      "reward_std": 0.26255880296230316,
      "reward_std_all": 0.23676641285419464,
      "rewards-max/forensics_sinkhorn_reward": 0.9644650220870972,
      "rewards-max/forensics_strict_edge_reward": 0.20242567360401154,
      "rewards-max/forensics_tiou_reward": 0.9269275367259979,
      "rewards/forensics_sinkhorn_reward": 0.6720563769340515,
      "rewards/forensics_strict_edge_reward": 0.012248799204826355,
      "rewards/forensics_tiou_reward": 0.3832451403141022,
      "step": 124
    },
    {
      "completion_length": 21.046875,
      "epoch": 0.38402457757296465,
      "grad_norm": 20.444913864135742,
      "kl": 1.68701171875,
      "learning_rate": 9.046153846153846e-07,
      "loss": 0.0063,
      "reward": 0.9561331272125244,
      "reward_std": 0.314999595284462,
      "reward_std_all": 0.2918657884001732,
      "rewards-max/forensics_sinkhorn_reward": 0.9651855826377869,
      "rewards-max/forensics_strict_edge_reward": 0.09536146000027657,
      "rewards-max/forensics_tiou_reward": 0.9352549910545349,
      "rewards/forensics_sinkhorn_reward": 0.6231791377067566,
      "rewards/forensics_strict_edge_reward": 0.009525951638352126,
      "rewards/forensics_tiou_reward": 0.32342806458473206,
      "step": 125
    },
    {
      "completion_length": 19.484375,
      "epoch": 0.3870967741935484,
      "grad_norm": 6.481964111328125,
      "kl": 0.181640625,
      "learning_rate": 9.038461538461538e-07,
      "loss": -0.0162,
      "reward": 0.8316461443901062,
      "reward_std": 0.2429802119731903,
      "reward_std_all": 0.22400273382663727,
      "rewards-max/forensics_sinkhorn_reward": 0.9038352072238922,
      "rewards-max/forensics_strict_edge_reward": 0.09446288768958766,
      "rewards-max/forensics_tiou_reward": 0.7793098092079163,
      "rewards/forensics_sinkhorn_reward": 0.5781872272491455,
      "rewards/forensics_strict_edge_reward": 0.007684397907496532,
      "rewards/forensics_tiou_reward": 0.245774507522583,
      "step": 126
    },
    {
      "completion_length": 19.359375,
      "epoch": 0.3901689708141321,
      "grad_norm": 5.265381336212158,
      "kl": 0.1943359375,
      "learning_rate": 9.030769230769231e-07,
      "loss": -0.0209,
      "reward": 0.9864177703857422,
      "reward_std": 0.2252417355775833,
      "reward_std_all": 0.20301757752895355,
      "rewards-max/forensics_sinkhorn_reward": 0.9476462006568909,
      "rewards-max/forensics_strict_edge_reward": 0.11941153556108475,
      "rewards-max/forensics_tiou_reward": 0.8894854187965393,
      "rewards/forensics_sinkhorn_reward": 0.652275413274765,
      "rewards/forensics_strict_edge_reward": 0.008087260997854173,
      "rewards/forensics_tiou_reward": 0.32605506479740143,
      "step": 127
    },
    {
      "completion_length": 19.640625,
      "epoch": 0.3932411674347158,
      "grad_norm": 5.994120121002197,
      "kl": 0.13525390625,
      "learning_rate": 9.023076923076922e-07,
      "loss": 0.0404,
      "reward": 1.0073354840278625,
      "reward_std": 0.28333529829978943,
      "reward_std_all": 0.2549983561038971,
      "rewards-max/forensics_sinkhorn_reward": 0.9158836305141449,
      "rewards-max/forensics_strict_edge_reward": 0.1049298383295536,
      "rewards-max/forensics_tiou_reward": 0.794139951467514,
      "rewards/forensics_sinkhorn_reward": 0.6428742408752441,
      "rewards/forensics_strict_edge_reward": 0.006792309693992138,
      "rewards/forensics_tiou_reward": 0.3576688915491104,
      "step": 128
    },
    {
      "completion_length": 19.265625,
      "epoch": 0.39631336405529954,
      "grad_norm": 5.289714813232422,
      "kl": 0.18505859375,
      "learning_rate": 9.015384615384614e-07,
      "loss": -0.0187,
      "reward": 0.9029532670974731,
      "reward_std": 0.19896753132343292,
      "reward_std_all": 0.2096891850233078,
      "rewards-max/forensics_sinkhorn_reward": 0.8385738432407379,
      "rewards-max/forensics_strict_edge_reward": 0.30180229991674423,
      "rewards-max/forensics_tiou_reward": 0.7274810373783112,
      "rewards/forensics_sinkhorn_reward": 0.6069163978099823,
      "rewards/forensics_strict_edge_reward": 0.020949636353179812,
      "rewards/forensics_tiou_reward": 0.27508723735809326,
      "step": 129
    },
    {
      "completion_length": 20.15625,
      "epoch": 0.39938556067588327,
      "grad_norm": 5.290317058563232,
      "kl": 0.22802734375,
      "learning_rate": 9.007692307692307e-07,
      "loss": -0.0141,
      "reward": 0.8888687193393707,
      "reward_std": 0.22751843929290771,
      "reward_std_all": 0.21707090735435486,
      "rewards-max/forensics_sinkhorn_reward": 0.9129361808300018,
      "rewards-max/forensics_strict_edge_reward": 0.4043722003698349,
      "rewards-max/forensics_tiou_reward": 0.8161151707172394,
      "rewards/forensics_sinkhorn_reward": 0.594380646944046,
      "rewards/forensics_strict_edge_reward": 0.02462536795064807,
      "rewards/forensics_tiou_reward": 0.26986271142959595,
      "step": 130
    },
    {
      "completion_length": 19.578125,
      "epoch": 0.402457757296467,
      "grad_norm": 5.978974342346191,
      "kl": 0.13818359375,
      "learning_rate": 9e-07,
      "loss": -0.0462,
      "reward": 0.9559390544891357,
      "reward_std": 0.20452116429805756,
      "reward_std_all": 0.2190144583582878,
      "rewards-max/forensics_sinkhorn_reward": 0.9341867566108704,
      "rewards-max/forensics_strict_edge_reward": 0.10736460611224174,
      "rewards-max/forensics_tiou_reward": 0.8621941804885864,
      "rewards/forensics_sinkhorn_reward": 0.6079558730125427,
      "rewards/forensics_strict_edge_reward": 0.006288354750722647,
      "rewards/forensics_tiou_reward": 0.34169481694698334,
      "step": 131
    },
    {
      "completion_length": 19.6875,
      "epoch": 0.4055299539170507,
      "grad_norm": 6.176400661468506,
      "kl": 0.109130859375,
      "learning_rate": 8.992307692307693e-07,
      "loss": -0.0562,
      "reward": 0.9908311367034912,
      "reward_std": 0.29210300743579865,
      "reward_std_all": 0.270823173224926,
      "rewards-max/forensics_sinkhorn_reward": 0.9377115070819855,
      "rewards-max/forensics_strict_edge_reward": 0.25700416913605295,
      "rewards-max/forensics_tiou_reward": 0.8658766448497772,
      "rewards/forensics_sinkhorn_reward": 0.6432420015335083,
      "rewards/forensics_strict_edge_reward": 0.019102055390249006,
      "rewards/forensics_tiou_reward": 0.3284871131181717,
      "step": 132
    },
    {
      "completion_length": 21.078125,
      "epoch": 0.40860215053763443,
      "grad_norm": 6.358539581298828,
      "kl": 0.185302734375,
      "learning_rate": 8.984615384615384e-07,
      "loss": -0.0457,
      "reward": 0.9882842004299164,
      "reward_std": 0.350761741399765,
      "reward_std_all": 0.3167107254266739,
      "rewards-max/forensics_sinkhorn_reward": 0.956445187330246,
      "rewards-max/forensics_strict_edge_reward": 0.4078576974570751,
      "rewards-max/forensics_tiou_reward": 0.9102253615856171,
      "rewards/forensics_sinkhorn_reward": 0.619669497013092,
      "rewards/forensics_strict_edge_reward": 0.03705605526920408,
      "rewards/forensics_tiou_reward": 0.33155862987041473,
      "step": 133
    },
    {
      "completion_length": 21.375,
      "epoch": 0.4116743471582181,
      "grad_norm": 6.772880554199219,
      "kl": 0.3837890625,
      "learning_rate": 8.976923076923076e-07,
      "loss": -0.1136,
      "reward": 0.9878324866294861,
      "reward_std": 0.21592959016561508,
      "reward_std_all": 0.2183963805437088,
      "rewards-max/forensics_sinkhorn_reward": 0.927154541015625,
      "rewards-max/forensics_strict_edge_reward": 0.05527026252821088,
      "rewards-max/forensics_tiou_reward": 0.8383505046367645,
      "rewards/forensics_sinkhorn_reward": 0.6539273858070374,
      "rewards/forensics_strict_edge_reward": 0.0028582660015672445,
      "rewards/forensics_tiou_reward": 0.33104680478572845,
      "step": 134
    },
    {
      "completion_length": 19.984375,
      "epoch": 0.4147465437788018,
      "grad_norm": 8.100847244262695,
      "kl": 0.1533203125,
      "learning_rate": 8.969230769230769e-07,
      "loss": 0.1265,
      "reward": 0.9886124432086945,
      "reward_std": 0.3423630893230438,
      "reward_std_all": 0.3018413931131363,
      "rewards-max/forensics_sinkhorn_reward": 0.9603699445724487,
      "rewards-max/forensics_strict_edge_reward": 0.2353852316737175,
      "rewards-max/forensics_tiou_reward": 0.9158341884613037,
      "rewards/forensics_sinkhorn_reward": 0.6314481794834137,
      "rewards/forensics_strict_edge_reward": 0.019274471327662468,
      "rewards/forensics_tiou_reward": 0.3378898501396179,
      "step": 135
    },
    {
      "completion_length": 19.921875,
      "epoch": 0.41781874039938555,
      "grad_norm": 6.842417240142822,
      "kl": 0.17724609375,
      "learning_rate": 8.961538461538461e-07,
      "loss": -0.009,
      "reward": 1.0526572167873383,
      "reward_std": 0.28160546720027924,
      "reward_std_all": 0.26344890892505646,
      "rewards-max/forensics_sinkhorn_reward": 0.9078228175640106,
      "rewards-max/forensics_strict_edge_reward": 0.01661082288956095,
      "rewards-max/forensics_tiou_reward": 0.7722923755645752,
      "rewards/forensics_sinkhorn_reward": 0.6790774762630463,
      "rewards/forensics_strict_edge_reward": 0.0008914952019836164,
      "rewards/forensics_tiou_reward": 0.3726882040500641,
      "step": 136
    },
    {
      "completion_length": 20.53125,
      "epoch": 0.42089093701996927,
      "grad_norm": 8.08124828338623,
      "kl": 0.212890625,
      "learning_rate": 8.953846153846152e-07,
      "loss": -0.1115,
      "reward": 1.054709106683731,
      "reward_std": 0.3470572829246521,
      "reward_std_all": 0.30420199036598206,
      "rewards-max/forensics_sinkhorn_reward": 0.9628360867500305,
      "rewards-max/forensics_strict_edge_reward": 0.33540203935990576,
      "rewards-max/forensics_tiou_reward": 0.9099223017692566,
      "rewards/forensics_sinkhorn_reward": 0.6628009974956512,
      "rewards/forensics_strict_edge_reward": 0.01824553489768732,
      "rewards/forensics_tiou_reward": 0.3736625909805298,
      "step": 137
    },
    {
      "completion_length": 20.40625,
      "epoch": 0.423963133640553,
      "grad_norm": 6.214811325073242,
      "kl": 0.17431640625,
      "learning_rate": 8.946153846153846e-07,
      "loss": -0.1167,
      "reward": 0.8771660327911377,
      "reward_std": 0.295628622174263,
      "reward_std_all": 0.28206369280815125,
      "rewards-max/forensics_sinkhorn_reward": 0.9778718054294586,
      "rewards-max/forensics_strict_edge_reward": 0.3214932158589363,
      "rewards-max/forensics_tiou_reward": 0.9666609466075897,
      "rewards/forensics_sinkhorn_reward": 0.5807015597820282,
      "rewards/forensics_strict_edge_reward": 0.016330121085047722,
      "rewards/forensics_tiou_reward": 0.2801343500614166,
      "step": 138
    },
    {
      "completion_length": 19.375,
      "epoch": 0.4270353302611367,
      "grad_norm": 6.673334121704102,
      "kl": 0.19677734375,
      "learning_rate": 8.938461538461538e-07,
      "loss": 0.1034,
      "reward": 1.1404204964637756,
      "reward_std": 0.2445855364203453,
      "reward_std_all": 0.21446546912193298,
      "rewards-max/forensics_sinkhorn_reward": 0.9549613893032074,
      "rewards-max/forensics_strict_edge_reward": 0.36005355045199394,
      "rewards-max/forensics_tiou_reward": 0.9036625623703003,
      "rewards/forensics_sinkhorn_reward": 0.6976217031478882,
      "rewards/forensics_strict_edge_reward": 0.01624167920090258,
      "rewards/forensics_tiou_reward": 0.42655718326568604,
      "step": 139
    },
    {
      "completion_length": 20.078125,
      "epoch": 0.43010752688172044,
      "grad_norm": 7.792646884918213,
      "kl": 0.47705078125,
      "learning_rate": 8.930769230769231e-07,
      "loss": -0.038,
      "reward": 1.1009512543678284,
      "reward_std": 0.29822373390197754,
      "reward_std_all": 0.2897961139678955,
      "rewards-max/forensics_sinkhorn_reward": 0.9293943643569946,
      "rewards-max/forensics_strict_edge_reward": 0.09515612758696079,
      "rewards-max/forensics_tiou_reward": 0.853155106306076,
      "rewards/forensics_sinkhorn_reward": 0.6947937607765198,
      "rewards/forensics_strict_edge_reward": 0.006404676008969545,
      "rewards/forensics_tiou_reward": 0.39975278079509735,
      "step": 140
    },
    {
      "completion_length": 19.734375,
      "epoch": 0.43317972350230416,
      "grad_norm": 5.956088542938232,
      "kl": 0.17626953125,
      "learning_rate": 8.923076923076923e-07,
      "loss": -0.1401,
      "reward": 1.1039998531341553,
      "reward_std": 0.30296090245246887,
      "reward_std_all": 0.32856321334838867,
      "rewards-max/forensics_sinkhorn_reward": 0.9775969386100769,
      "rewards-max/forensics_strict_edge_reward": 0.08881983137689531,
      "rewards-max/forensics_tiou_reward": 0.9587568044662476,
      "rewards/forensics_sinkhorn_reward": 0.6830728948116302,
      "rewards/forensics_strict_edge_reward": 0.009134190273471177,
      "rewards/forensics_tiou_reward": 0.41179272532463074,
      "step": 141
    },
    {
      "completion_length": 19.3125,
      "epoch": 0.4362519201228879,
      "grad_norm": 5.799772262573242,
      "kl": 0.1650390625,
      "learning_rate": 8.915384615384614e-07,
      "loss": -0.0159,
      "reward": 1.109261691570282,
      "reward_std": 0.17683784663677216,
      "reward_std_all": 0.16812869906425476,
      "rewards-max/forensics_sinkhorn_reward": 0.973167896270752,
      "rewards-max/forensics_strict_edge_reward": 0.30970802903175354,
      "rewards-max/forensics_tiou_reward": 0.9513788223266602,
      "rewards/forensics_sinkhorn_reward": 0.6938313245773315,
      "rewards/forensics_strict_edge_reward": 0.015064852312207222,
      "rewards/forensics_tiou_reward": 0.4003655016422272,
      "step": 142
    },
    {
      "completion_length": 19.671875,
      "epoch": 0.4393241167434716,
      "grad_norm": 4.917923450469971,
      "kl": 0.17626953125,
      "learning_rate": 8.907692307692307e-07,
      "loss": -0.0308,
      "reward": 1.1148725748062134,
      "reward_std": 0.2633684426546097,
      "reward_std_all": 0.2617831379175186,
      "rewards-max/forensics_sinkhorn_reward": 0.8982143700122833,
      "rewards-max/forensics_strict_edge_reward": 0.07377285370603204,
      "rewards-max/forensics_tiou_reward": 0.7819603085517883,
      "rewards/forensics_sinkhorn_reward": 0.7041671872138977,
      "rewards/forensics_strict_edge_reward": 0.0037205332191661,
      "rewards/forensics_tiou_reward": 0.4069848507642746,
      "step": 143
    },
    {
      "completion_length": 19.421875,
      "epoch": 0.4423963133640553,
      "grad_norm": 8.803635597229004,
      "kl": 0.541015625,
      "learning_rate": 8.9e-07,
      "loss": -0.0376,
      "reward": 1.0032639503479004,
      "reward_std": 0.17689978331327438,
      "reward_std_all": 0.2065856233239174,
      "rewards-max/forensics_sinkhorn_reward": 0.916705310344696,
      "rewards-max/forensics_strict_edge_reward": 0.01413731393404305,
      "rewards-max/forensics_tiou_reward": 0.7960748076438904,
      "rewards/forensics_sinkhorn_reward": 0.6559317708015442,
      "rewards/forensics_strict_edge_reward": 0.0011235904530622065,
      "rewards/forensics_tiou_reward": 0.3462085947394371,
      "step": 144
    },
    {
      "completion_length": 20.953125,
      "epoch": 0.445468509984639,
      "grad_norm": 4.846534729003906,
      "kl": 0.162109375,
      "learning_rate": 8.892307692307693e-07,
      "loss": -0.1028,
      "reward": 1.1368792653083801,
      "reward_std": 0.2458425611257553,
      "reward_std_all": 0.24083247035741806,
      "rewards-max/forensics_sinkhorn_reward": 0.9801691174507141,
      "rewards-max/forensics_strict_edge_reward": 0.21059804409742355,
      "rewards-max/forensics_tiou_reward": 0.9697275161743164,
      "rewards/forensics_sinkhorn_reward": 0.6906865835189819,
      "rewards/forensics_strict_edge_reward": 0.015551699791103601,
      "rewards/forensics_tiou_reward": 0.43064092099666595,
      "step": 145
    },
    {
      "completion_length": 19.515625,
      "epoch": 0.4485407066052227,
      "grad_norm": 6.035407066345215,
      "kl": 0.259765625,
      "learning_rate": 8.884615384615384e-07,
      "loss": 0.0205,
      "reward": 0.9615515768527985,
      "reward_std": 0.26724401116371155,
      "reward_std_all": 0.2569497525691986,
      "rewards-max/forensics_sinkhorn_reward": 0.9172960519790649,
      "rewards-max/forensics_strict_edge_reward": 0.13635169342160225,
      "rewards-max/forensics_tiou_reward": 0.8200290501117706,
      "rewards/forensics_sinkhorn_reward": 0.6254070997238159,
      "rewards/forensics_strict_edge_reward": 0.006182268727570772,
      "rewards/forensics_tiou_reward": 0.32996220886707306,
      "step": 146
    },
    {
      "completion_length": 19.984375,
      "epoch": 0.45161290322580644,
      "grad_norm": 6.7163825035095215,
      "kl": 0.310546875,
      "learning_rate": 8.876923076923076e-07,
      "loss": 0.008,
      "reward": 1.0971247553825378,
      "reward_std": 0.29812854528427124,
      "reward_std_all": 0.32688166201114655,
      "rewards-max/forensics_sinkhorn_reward": 0.9823060035705566,
      "rewards-max/forensics_strict_edge_reward": 0.6065583229064941,
      "rewards-max/forensics_tiou_reward": 0.979914128780365,
      "rewards/forensics_sinkhorn_reward": 0.6768780648708344,
      "rewards/forensics_strict_edge_reward": 0.03311985731124878,
      "rewards/forensics_tiou_reward": 0.3871268630027771,
      "step": 147
    },
    {
      "completion_length": 19.453125,
      "epoch": 0.45468509984639016,
      "grad_norm": 5.663802623748779,
      "kl": 0.14990234375,
      "learning_rate": 8.869230769230769e-07,
      "loss": 0.0408,
      "reward": 1.042338341474533,
      "reward_std": 0.21287181228399277,
      "reward_std_all": 0.19624296575784683,
      "rewards-max/forensics_sinkhorn_reward": 0.9075575470924377,
      "rewards-max/forensics_strict_edge_reward": 0.18994817696511745,
      "rewards-max/forensics_tiou_reward": 0.8030007481575012,
      "rewards/forensics_sinkhorn_reward": 0.6770815551280975,
      "rewards/forensics_strict_edge_reward": 0.00849946599919349,
      "rewards/forensics_tiou_reward": 0.3567572981119156,
      "step": 148
    },
    {
      "completion_length": 19.34375,
      "epoch": 0.4577572964669739,
      "grad_norm": 5.53862190246582,
      "kl": 0.15380859375,
      "learning_rate": 8.861538461538461e-07,
      "loss": 0.0248,
      "reward": 1.1175625920295715,
      "reward_std": 0.3226577490568161,
      "reward_std_all": 0.3178836852312088,
      "rewards-max/forensics_sinkhorn_reward": 0.9744125604629517,
      "rewards-max/forensics_strict_edge_reward": 0.5413877367973328,
      "rewards-max/forensics_tiou_reward": 0.9689441919326782,
      "rewards/forensics_sinkhorn_reward": 0.6745476126670837,
      "rewards/forensics_strict_edge_reward": 0.05166151747107506,
      "rewards/forensics_tiou_reward": 0.39135344326496124,
      "step": 149
    },
    {
      "completion_length": 19.953125,
      "epoch": 0.4608294930875576,
      "grad_norm": 6.748775005340576,
      "kl": 0.26318359375,
      "learning_rate": 8.853846153846154e-07,
      "loss": -0.0124,
      "reward": 1.0431628823280334,
      "reward_std": 0.2422940582036972,
      "reward_std_all": 0.27054162323474884,
      "rewards-max/forensics_sinkhorn_reward": 0.9152542352676392,
      "rewards-max/forensics_strict_edge_reward": 0.03293501678854227,
      "rewards-max/forensics_tiou_reward": 0.8151524662971497,
      "rewards/forensics_sinkhorn_reward": 0.6462368965148926,
      "rewards/forensics_strict_edge_reward": 0.003362217976246029,
      "rewards/forensics_tiou_reward": 0.39356379210948944,
      "step": 150
    },
    {
      "completion_length": 19.78125,
      "epoch": 0.46390168970814133,
      "grad_norm": 6.192536354064941,
      "kl": 0.144775390625,
      "learning_rate": 8.846153846153846e-07,
      "loss": 0.1347,
      "reward": 1.1110173463821411,
      "reward_std": 0.18557216227054596,
      "reward_std_all": 0.17412541061639786,
      "rewards-max/forensics_sinkhorn_reward": 0.9342280626296997,
      "rewards-max/forensics_strict_edge_reward": 0.41928549110889435,
      "rewards-max/forensics_tiou_reward": 0.8699315190315247,
      "rewards/forensics_sinkhorn_reward": 0.6782847940921783,
      "rewards/forensics_strict_edge_reward": 0.02969823032617569,
      "rewards/forensics_tiou_reward": 0.4030342996120453,
      "step": 151
    },
    {
      "completion_length": 20.046875,
      "epoch": 0.46697388632872505,
      "grad_norm": 7.048968315124512,
      "kl": 0.1845703125,
      "learning_rate": 8.838461538461538e-07,
      "loss": -0.0497,
      "reward": 1.1398966908454895,
      "reward_std": 0.2405060976743698,
      "reward_std_all": 0.23305942863225937,
      "rewards-max/forensics_sinkhorn_reward": 0.9570326209068298,
      "rewards-max/forensics_strict_edge_reward": 0.16433284990489483,
      "rewards-max/forensics_tiou_reward": 0.9101540148258209,
      "rewards/forensics_sinkhorn_reward": 0.6899699866771698,
      "rewards/forensics_strict_edge_reward": 0.010941580287180841,
      "rewards/forensics_tiou_reward": 0.4389851540327072,
      "step": 152
    },
    {
      "completion_length": 20.0625,
      "epoch": 0.4700460829493088,
      "grad_norm": 5.4523024559021,
      "kl": 0.21142578125,
      "learning_rate": 8.830769230769231e-07,
      "loss": 0.0059,
      "reward": 1.2278839945793152,
      "reward_std": 0.2762480601668358,
      "reward_std_all": 0.2727179750800133,
      "rewards-max/forensics_sinkhorn_reward": 0.9431179165840149,
      "rewards-max/forensics_strict_edge_reward": 0.25687629368621856,
      "rewards-max/forensics_tiou_reward": 0.8831491470336914,
      "rewards/forensics_sinkhorn_reward": 0.7247904241085052,
      "rewards/forensics_strict_edge_reward": 0.019531267298589228,
      "rewards/forensics_tiou_reward": 0.48356229066848755,
      "step": 153
    },
    {
      "completion_length": 20.09375,
      "epoch": 0.4731182795698925,
      "grad_norm": 5.432553291320801,
      "kl": 0.1826171875,
      "learning_rate": 8.823076923076923e-07,
      "loss": -0.0126,
      "reward": 1.011293739080429,
      "reward_std": 0.3160855174064636,
      "reward_std_all": 0.29860930144786835,
      "rewards-max/forensics_sinkhorn_reward": 0.863378494977951,
      "rewards-max/forensics_strict_edge_reward": 0.35027457773685455,
      "rewards-max/forensics_tiou_reward": 0.7785235047340393,
      "rewards/forensics_sinkhorn_reward": 0.6287200450897217,
      "rewards/forensics_strict_edge_reward": 0.026207849383354187,
      "rewards/forensics_tiou_reward": 0.35636579990386963,
      "step": 154
    },
    {
      "completion_length": 19.734375,
      "epoch": 0.47619047619047616,
      "grad_norm": 7.579434871673584,
      "kl": 0.16552734375,
      "learning_rate": 8.815384615384614e-07,
      "loss": 0.0137,
      "reward": 1.1345415711402893,
      "reward_std": 0.27338775992393494,
      "reward_std_all": 0.2659362331032753,
      "rewards-max/forensics_sinkhorn_reward": 0.9798468947410583,
      "rewards-max/forensics_strict_edge_reward": 0.23775557428598404,
      "rewards-max/forensics_tiou_reward": 0.9706624448299408,
      "rewards/forensics_sinkhorn_reward": 0.7022494375705719,
      "rewards/forensics_strict_edge_reward": 0.022743343841284513,
      "rewards/forensics_tiou_reward": 0.4095488041639328,
      "step": 155
    },
    {
      "completion_length": 19.6875,
      "epoch": 0.4792626728110599,
      "grad_norm": 6.724539756774902,
      "kl": 0.14111328125,
      "learning_rate": 8.807692307692307e-07,
      "loss": 0.0421,
      "reward": 1.039831280708313,
      "reward_std": 0.23299090564250946,
      "reward_std_all": 0.21570508182048798,
      "rewards-max/forensics_sinkhorn_reward": 0.9556683003902435,
      "rewards-max/forensics_strict_edge_reward": 0.3599386508576572,
      "rewards-max/forensics_tiou_reward": 0.8858682215213776,
      "rewards/forensics_sinkhorn_reward": 0.6455452144145966,
      "rewards/forensics_strict_edge_reward": 0.017769990976375993,
      "rewards/forensics_tiou_reward": 0.37651611864566803,
      "step": 156
    },
    {
      "completion_length": 20.328125,
      "epoch": 0.4823348694316436,
      "grad_norm": 6.95663595199585,
      "kl": 0.1318359375,
      "learning_rate": 8.799999999999999e-07,
      "loss": -0.0564,
      "reward": 1.2497820258140564,
      "reward_std": 0.29664769768714905,
      "reward_std_all": 0.2632759064435959,
      "rewards-max/forensics_sinkhorn_reward": 0.9540372490882874,
      "rewards-max/forensics_strict_edge_reward": 0.2393038645386696,
      "rewards-max/forensics_tiou_reward": 0.8984278440475464,
      "rewards/forensics_sinkhorn_reward": 0.7203052639961243,
      "rewards/forensics_strict_edge_reward": 0.021018109284341335,
      "rewards/forensics_tiou_reward": 0.50845867395401,
      "step": 157
    },
    {
      "completion_length": 21.046875,
      "epoch": 0.48540706605222733,
      "grad_norm": 6.656193733215332,
      "kl": 0.11962890625,
      "learning_rate": 8.792307692307693e-07,
      "loss": 0.0875,
      "reward": 0.9413456916809082,
      "reward_std": 0.21068592369556427,
      "reward_std_all": 0.22965173423290253,
      "rewards-max/forensics_sinkhorn_reward": 0.9449227154254913,
      "rewards-max/forensics_strict_edge_reward": 0.25244972109794617,
      "rewards-max/forensics_tiou_reward": 0.8920353949069977,
      "rewards/forensics_sinkhorn_reward": 0.6057535409927368,
      "rewards/forensics_strict_edge_reward": 0.014562634285539389,
      "rewards/forensics_tiou_reward": 0.32102957367897034,
      "step": 158
    },
    {
      "completion_length": 21.125,
      "epoch": 0.48847926267281105,
      "grad_norm": 6.852994441986084,
      "kl": 0.131103515625,
      "learning_rate": 8.784615384615385e-07,
      "loss": 0.0397,
      "reward": 0.8735234141349792,
      "reward_std": 0.23663592338562012,
      "reward_std_all": 0.23672320693731308,
      "rewards-max/forensics_sinkhorn_reward": 0.961775153875351,
      "rewards-max/forensics_strict_edge_reward": 0.1944793276488781,
      "rewards-max/forensics_tiou_reward": 0.9216828048229218,
      "rewards/forensics_sinkhorn_reward": 0.5746017694473267,
      "rewards/forensics_strict_edge_reward": 0.008514291141182184,
      "rewards/forensics_tiou_reward": 0.2904074192047119,
      "step": 159
    },
    {
      "completion_length": 21.140625,
      "epoch": 0.4915514592933948,
      "grad_norm": 8.649259567260742,
      "kl": 0.1484375,
      "learning_rate": 8.776923076923076e-07,
      "loss": 0.0297,
      "reward": 1.0000522136688232,
      "reward_std": 0.2630322128534317,
      "reward_std_all": 0.2373691201210022,
      "rewards-max/forensics_sinkhorn_reward": 0.8791629076004028,
      "rewards-max/forensics_strict_edge_reward": 0.13787522038910538,
      "rewards-max/forensics_tiou_reward": 0.7477156817913055,
      "rewards/forensics_sinkhorn_reward": 0.639481246471405,
      "rewards/forensics_strict_edge_reward": 0.01187315775314346,
      "rewards/forensics_tiou_reward": 0.3486977070569992,
      "step": 160
    },
    {
      "completion_length": 19.71875,
      "epoch": 0.4946236559139785,
      "grad_norm": 8.253763198852539,
      "kl": 0.114013671875,
      "learning_rate": 8.769230769230769e-07,
      "loss": -0.0418,
      "reward": 0.9932677149772644,
      "reward_std": 0.2516491115093231,
      "reward_std_all": 0.2527138739824295,
      "rewards-max/forensics_sinkhorn_reward": 0.9244272410869598,
      "rewards-max/forensics_strict_edge_reward": 0.029022383503615856,
      "rewards-max/forensics_tiou_reward": 0.8286836445331573,
      "rewards/forensics_sinkhorn_reward": 0.6263748407363892,
      "rewards/forensics_strict_edge_reward": 0.003099748631939292,
      "rewards/forensics_tiou_reward": 0.36379316449165344,
      "step": 161
    },
    {
      "completion_length": 20.21875,
      "epoch": 0.4976958525345622,
      "grad_norm": 6.1616058349609375,
      "kl": 0.1826171875,
      "learning_rate": 8.761538461538461e-07,
      "loss": -0.0605,
      "reward": 1.2057544589042664,
      "reward_std": 0.27346497774124146,
      "reward_std_all": 0.23999565839767456,
      "rewards-max/forensics_sinkhorn_reward": 0.9745433926582336,
      "rewards-max/forensics_strict_edge_reward": 0.4233581870794296,
      "rewards-max/forensics_tiou_reward": 0.966566264629364,
      "rewards/forensics_sinkhorn_reward": 0.7062837481498718,
      "rewards/forensics_strict_edge_reward": 0.04986536782234907,
      "rewards/forensics_tiou_reward": 0.4496053606271744,
      "step": 162
    },
    {
      "completion_length": 19.75,
      "epoch": 0.500768049155146,
      "grad_norm": 7.558534622192383,
      "kl": 0.330078125,
      "learning_rate": 8.753846153846154e-07,
      "loss": -0.1431,
      "reward": 0.8784870505332947,
      "reward_std": 0.17255692183971405,
      "reward_std_all": 0.17578159272670746,
      "rewards-max/forensics_sinkhorn_reward": 0.9106163084506989,
      "rewards-max/forensics_strict_edge_reward": 0.025844454765319824,
      "rewards-max/forensics_tiou_reward": 0.8090875744819641,
      "rewards/forensics_sinkhorn_reward": 0.6040102541446686,
      "rewards/forensics_strict_edge_reward": 0.0014187937777023762,
      "rewards/forensics_tiou_reward": 0.2730579823255539,
      "step": 163
    },
    {
      "completion_length": 19.8125,
      "epoch": 0.5038402457757296,
      "grad_norm": 7.470381259918213,
      "kl": 0.127197265625,
      "learning_rate": 8.746153846153845e-07,
      "loss": 0.0108,
      "reward": 0.9386872053146362,
      "reward_std": 0.1702033132314682,
      "reward_std_all": 0.14573854207992554,
      "rewards-max/forensics_sinkhorn_reward": 0.9269427359104156,
      "rewards-max/forensics_strict_edge_reward": 0.04619642940815538,
      "rewards-max/forensics_tiou_reward": 0.8257598876953125,
      "rewards/forensics_sinkhorn_reward": 0.6307464241981506,
      "rewards/forensics_strict_edge_reward": 0.0029352785568335094,
      "rewards/forensics_tiou_reward": 0.3050055652856827,
      "step": 164
    },
    {
      "completion_length": 20.796875,
      "epoch": 0.5069124423963134,
      "grad_norm": 5.652077674865723,
      "kl": 0.134765625,
      "learning_rate": 8.738461538461538e-07,
      "loss": 0.0113,
      "reward": 1.2119665741920471,
      "reward_std": 0.4244651794433594,
      "reward_std_all": 0.36737361550331116,
      "rewards-max/forensics_sinkhorn_reward": 0.9817086160182953,
      "rewards-max/forensics_strict_edge_reward": 0.6660739481449127,
      "rewards-max/forensics_tiou_reward": 0.9805321097373962,
      "rewards/forensics_sinkhorn_reward": 0.6819136738777161,
      "rewards/forensics_strict_edge_reward": 0.07741216570138931,
      "rewards/forensics_tiou_reward": 0.45264074206352234,
      "step": 165
    },
    {
      "completion_length": 20.3125,
      "epoch": 0.5099846390168971,
      "grad_norm": 6.10890531539917,
      "kl": 0.1796875,
      "learning_rate": 8.730769230769231e-07,
      "loss": 0.0847,
      "reward": 1.1673910021781921,
      "reward_std": 0.32346637547016144,
      "reward_std_all": 0.28778813779354095,
      "rewards-max/forensics_sinkhorn_reward": 0.9698314368724823,
      "rewards-max/forensics_strict_edge_reward": 0.3885071277618408,
      "rewards-max/forensics_tiou_reward": 0.9494181871414185,
      "rewards/forensics_sinkhorn_reward": 0.6798957586288452,
      "rewards/forensics_strict_edge_reward": 0.03322456497699022,
      "rewards/forensics_tiou_reward": 0.4542706608772278,
      "step": 166
    },
    {
      "completion_length": 21.09375,
      "epoch": 0.5130568356374808,
      "grad_norm": 7.7104010581970215,
      "kl": 0.2421875,
      "learning_rate": 8.723076923076923e-07,
      "loss": -0.0461,
      "reward": 1.210854411125183,
      "reward_std": 0.30777084827423096,
      "reward_std_all": 0.31184022128582,
      "rewards-max/forensics_sinkhorn_reward": 0.9696025252342224,
      "rewards-max/forensics_strict_edge_reward": 0.22929870337247849,
      "rewards-max/forensics_tiou_reward": 0.9502771496772766,
      "rewards/forensics_sinkhorn_reward": 0.7170141339302063,
      "rewards/forensics_strict_edge_reward": 0.017736160196363926,
      "rewards/forensics_tiou_reward": 0.4761042296886444,
      "step": 167
    },
    {
      "completion_length": 22.046875,
      "epoch": 0.5161290322580645,
      "grad_norm": 5.757166385650635,
      "kl": 0.345703125,
      "learning_rate": 8.715384615384614e-07,
      "loss": 0.0884,
      "reward": 1.0290198922157288,
      "reward_std": 0.30253028124570847,
      "reward_std_all": 0.26805177330970764,
      "rewards-max/forensics_sinkhorn_reward": 0.9687390327453613,
      "rewards-max/forensics_strict_edge_reward": 0.06650098785758018,
      "rewards-max/forensics_tiou_reward": 0.9366882145404816,
      "rewards/forensics_sinkhorn_reward": 0.6349034905433655,
      "rewards/forensics_strict_edge_reward": 0.005046307807788253,
      "rewards/forensics_tiou_reward": 0.3890700340270996,
      "step": 168
    },
    {
      "completion_length": 20.546875,
      "epoch": 0.5192012288786483,
      "grad_norm": 6.132126808166504,
      "kl": 0.27978515625,
      "learning_rate": 8.707692307692307e-07,
      "loss": 0.1545,
      "reward": 1.0141641199588776,
      "reward_std": 0.21647507697343826,
      "reward_std_all": 0.23747969418764114,
      "rewards-max/forensics_sinkhorn_reward": 0.8855878710746765,
      "rewards-max/forensics_strict_edge_reward": 0.15063559396730852,
      "rewards-max/forensics_tiou_reward": 0.7795741558074951,
      "rewards/forensics_sinkhorn_reward": 0.639641135931015,
      "rewards/forensics_strict_edge_reward": 0.016758969658440037,
      "rewards/forensics_tiou_reward": 0.35776402056217194,
      "step": 169
    },
    {
      "completion_length": 20.03125,
      "epoch": 0.522273425499232,
      "grad_norm": 8.310827255249023,
      "kl": 0.416015625,
      "learning_rate": 8.699999999999999e-07,
      "loss": -0.028,
      "reward": 1.058998465538025,
      "reward_std": 0.1799987033009529,
      "reward_std_all": 0.18778646737337112,
      "rewards-max/forensics_sinkhorn_reward": 0.9644583463668823,
      "rewards-max/forensics_strict_edge_reward": 0.1799445990473032,
      "rewards-max/forensics_tiou_reward": 0.9342924654483795,
      "rewards/forensics_sinkhorn_reward": 0.6420229077339172,
      "rewards/forensics_strict_edge_reward": 0.017362122889608145,
      "rewards/forensics_tiou_reward": 0.39961352944374084,
      "step": 170
    },
    {
      "completion_length": 23.140625,
      "epoch": 0.5253456221198156,
      "grad_norm": 5.638839244842529,
      "kl": 0.19921875,
      "learning_rate": 8.692307692307692e-07,
      "loss": -0.0054,
      "reward": 0.9165335297584534,
      "reward_std": 0.2135583609342575,
      "reward_std_all": 0.19227945059537888,
      "rewards-max/forensics_sinkhorn_reward": 0.9277281761169434,
      "rewards-max/forensics_strict_edge_reward": 0.24915537610650063,
      "rewards-max/forensics_tiou_reward": 0.8431264460086823,
      "rewards/forensics_sinkhorn_reward": 0.5754968822002411,
      "rewards/forensics_strict_edge_reward": 0.014973882585763931,
      "rewards/forensics_tiou_reward": 0.32606276869773865,
      "step": 171
    },
    {
      "completion_length": 22.3125,
      "epoch": 0.5284178187403994,
      "grad_norm": 7.059197902679443,
      "kl": 0.21240234375,
      "learning_rate": 8.684615384615385e-07,
      "loss": 0.0646,
      "reward": 1.0513836741447449,
      "reward_std": 0.24360938370227814,
      "reward_std_all": 0.25157182663679123,
      "rewards-max/forensics_sinkhorn_reward": 0.961467832326889,
      "rewards-max/forensics_strict_edge_reward": 0.3829674479129608,
      "rewards-max/forensics_tiou_reward": 0.9233094155788422,
      "rewards/forensics_sinkhorn_reward": 0.641455888748169,
      "rewards/forensics_strict_edge_reward": 0.018409785584367455,
      "rewards/forensics_tiou_reward": 0.3915180414915085,
      "step": 172
    },
    {
      "completion_length": 19.65625,
      "epoch": 0.5314900153609831,
      "grad_norm": 4.880054473876953,
      "kl": 0.125,
      "learning_rate": 8.676923076923076e-07,
      "loss": -0.0071,
      "reward": 1.2287487983703613,
      "reward_std": 0.2127072513103485,
      "reward_std_all": 0.18602292239665985,
      "rewards-max/forensics_sinkhorn_reward": 0.9697396159172058,
      "rewards-max/forensics_strict_edge_reward": 0.12424845062196255,
      "rewards-max/forensics_tiou_reward": 0.9379960596561432,
      "rewards/forensics_sinkhorn_reward": 0.7205245792865753,
      "rewards/forensics_strict_edge_reward": 0.015908190165646374,
      "rewards/forensics_tiou_reward": 0.49231603741645813,
      "step": 173
    },
    {
      "completion_length": 21.765625,
      "epoch": 0.5345622119815668,
      "grad_norm": 8.77233600616455,
      "kl": 0.1220703125,
      "learning_rate": 8.669230769230769e-07,
      "loss": -0.0711,
      "reward": 1.2095165252685547,
      "reward_std": 0.29928432404994965,
      "reward_std_all": 0.288596510887146,
      "rewards-max/forensics_sinkhorn_reward": 0.951062947511673,
      "rewards-max/forensics_strict_edge_reward": 0.4005978852510452,
      "rewards-max/forensics_tiou_reward": 0.9051630795001984,
      "rewards/forensics_sinkhorn_reward": 0.7210308909416199,
      "rewards/forensics_strict_edge_reward": 0.024466454051434994,
      "rewards/forensics_tiou_reward": 0.4640190899372101,
      "step": 174
    },
    {
      "completion_length": 21.421875,
      "epoch": 0.5376344086021505,
      "grad_norm": 8.954930305480957,
      "kl": 0.17041015625,
      "learning_rate": 8.661538461538461e-07,
      "loss": 0.022,
      "reward": 1.0430516600608826,
      "reward_std": 0.21829676628112793,
      "reward_std_all": 0.22042877972126007,
      "rewards-max/forensics_sinkhorn_reward": 0.9638349413871765,
      "rewards-max/forensics_strict_edge_reward": 0.36650875210762024,
      "rewards-max/forensics_tiou_reward": 0.9310238659381866,
      "rewards/forensics_sinkhorn_reward": 0.6191787719726562,
      "rewards/forensics_strict_edge_reward": 0.031576504930853844,
      "rewards/forensics_tiou_reward": 0.3922963887453079,
      "step": 175
    },
    {
      "completion_length": 20.546875,
      "epoch": 0.5407066052227343,
      "grad_norm": 7.687387943267822,
      "kl": 0.228515625,
      "learning_rate": 8.653846153846154e-07,
      "loss": -0.033,
      "reward": 0.8940120339393616,
      "reward_std": 0.16069264709949493,
      "reward_std_all": 0.19499673694372177,
      "rewards-max/forensics_sinkhorn_reward": 0.876222163438797,
      "rewards-max/forensics_strict_edge_reward": 0.3078978843986988,
      "rewards-max/forensics_tiou_reward": 0.7706705629825592,
      "rewards/forensics_sinkhorn_reward": 0.5918347537517548,
      "rewards/forensics_strict_edge_reward": 0.014312824234366417,
      "rewards/forensics_tiou_reward": 0.28786447644233704,
      "step": 176
    },
    {
      "completion_length": 20.9375,
      "epoch": 0.543778801843318,
      "grad_norm": 6.113035202026367,
      "kl": 0.1279296875,
      "learning_rate": 8.646153846153845e-07,
      "loss": 0.0771,
      "reward": 1.2274397611618042,
      "reward_std": 0.3605844974517822,
      "reward_std_all": 0.3447287231683731,
      "rewards-max/forensics_sinkhorn_reward": 0.9748791754245758,
      "rewards-max/forensics_strict_edge_reward": 0.3961939886212349,
      "rewards-max/forensics_tiou_reward": 0.9622163772583008,
      "rewards/forensics_sinkhorn_reward": 0.7172899842262268,
      "rewards/forensics_strict_edge_reward": 0.024633216205984354,
      "rewards/forensics_tiou_reward": 0.48551662266254425,
      "step": 177
    },
    {
      "completion_length": 20.59375,
      "epoch": 0.5468509984639017,
      "grad_norm": 5.7559733390808105,
      "kl": 0.4130859375,
      "learning_rate": 8.638461538461537e-07,
      "loss": -0.0309,
      "reward": 1.0224274098873138,
      "reward_std": 0.23953549563884735,
      "reward_std_all": 0.22939318418502808,
      "rewards-max/forensics_sinkhorn_reward": 0.9168953001499176,
      "rewards-max/forensics_strict_edge_reward": 0.0422375313937664,
      "rewards-max/forensics_tiou_reward": 0.8137273788452148,
      "rewards/forensics_sinkhorn_reward": 0.6450713276863098,
      "rewards/forensics_strict_edge_reward": 0.002751719322986901,
      "rewards/forensics_tiou_reward": 0.37460434436798096,
      "step": 178
    },
    {
      "completion_length": 21.09375,
      "epoch": 0.5499231950844854,
      "grad_norm": 11.0872220993042,
      "kl": 0.1171875,
      "learning_rate": 8.630769230769231e-07,
      "loss": -0.0246,
      "reward": 1.1308422088623047,
      "reward_std": 0.27451203763484955,
      "reward_std_all": 0.25372354686260223,
      "rewards-max/forensics_sinkhorn_reward": 0.9636462926864624,
      "rewards-max/forensics_strict_edge_reward": 0.18100214190781116,
      "rewards-max/forensics_tiou_reward": 0.9293774664402008,
      "rewards/forensics_sinkhorn_reward": 0.7006673812866211,
      "rewards/forensics_strict_edge_reward": 0.014319465495646,
      "rewards/forensics_tiou_reward": 0.41585543751716614,
      "step": 179
    },
    {
      "completion_length": 19.65625,
      "epoch": 0.5529953917050692,
      "grad_norm": 7.743005275726318,
      "kl": 0.133544921875,
      "learning_rate": 8.623076923076923e-07,
      "loss": 0.0117,
      "reward": 0.9978492856025696,
      "reward_std": 0.27234014868736267,
      "reward_std_all": 0.2441403567790985,
      "rewards-max/forensics_sinkhorn_reward": 0.954237699508667,
      "rewards-max/forensics_strict_edge_reward": 0.3184485696256161,
      "rewards-max/forensics_tiou_reward": 0.9091455042362213,
      "rewards/forensics_sinkhorn_reward": 0.6229220330715179,
      "rewards/forensics_strict_edge_reward": 0.016800356097519398,
      "rewards/forensics_tiou_reward": 0.3581268638372421,
      "step": 180
    },
    {
      "completion_length": 19.46875,
      "epoch": 0.5560675883256528,
      "grad_norm": 5.571047782897949,
      "kl": 0.1650390625,
      "learning_rate": 8.615384615384616e-07,
      "loss": -0.0342,
      "reward": 1.1076748371124268,
      "reward_std": 0.2172357738018036,
      "reward_std_all": 0.22688451409339905,
      "rewards-max/forensics_sinkhorn_reward": 0.9744445979595184,
      "rewards-max/forensics_strict_edge_reward": 0.3157382905483246,
      "rewards-max/forensics_tiou_reward": 0.9634271264076233,
      "rewards/forensics_sinkhorn_reward": 0.6786801815032959,
      "rewards/forensics_strict_edge_reward": 0.014867182355374098,
      "rewards/forensics_tiou_reward": 0.41412755846977234,
      "step": 181
    },
    {
      "completion_length": 20.109375,
      "epoch": 0.5591397849462365,
      "grad_norm": 6.236608028411865,
      "kl": 0.1982421875,
      "learning_rate": 8.607692307692307e-07,
      "loss": 0.0099,
      "reward": 1.1731195449829102,
      "reward_std": 0.3043360263109207,
      "reward_std_all": 0.26521432399749756,
      "rewards-max/forensics_sinkhorn_reward": 0.952421098947525,
      "rewards-max/forensics_strict_edge_reward": 0.478572185151279,
      "rewards-max/forensics_tiou_reward": 0.9038486480712891,
      "rewards/forensics_sinkhorn_reward": 0.6861177980899811,
      "rewards/forensics_strict_edge_reward": 0.032811757759191096,
      "rewards/forensics_tiou_reward": 0.4541899561882019,
      "step": 182
    },
    {
      "completion_length": 20.109375,
      "epoch": 0.5622119815668203,
      "grad_norm": 6.167573928833008,
      "kl": 0.1376953125,
      "learning_rate": 8.599999999999999e-07,
      "loss": -0.0114,
      "reward": 1.0549663305282593,
      "reward_std": 0.2172318994998932,
      "reward_std_all": 0.2031576707959175,
      "rewards-max/forensics_sinkhorn_reward": 0.9469181299209595,
      "rewards-max/forensics_strict_edge_reward": 0.0525545934215188,
      "rewards-max/forensics_tiou_reward": 0.8840171098709106,
      "rewards/forensics_sinkhorn_reward": 0.6695787012577057,
      "rewards/forensics_strict_edge_reward": 0.0048347399570047855,
      "rewards/forensics_tiou_reward": 0.38055285811424255,
      "step": 183
    },
    {
      "completion_length": 19.484375,
      "epoch": 0.565284178187404,
      "grad_norm": 5.771642208099365,
      "kl": 0.15673828125,
      "learning_rate": 8.592307692307692e-07,
      "loss": -0.0127,
      "reward": 1.0500319600105286,
      "reward_std": 0.22646702826023102,
      "reward_std_all": 0.20384127646684647,
      "rewards-max/forensics_sinkhorn_reward": 0.9408278167247772,
      "rewards-max/forensics_strict_edge_reward": 0.21599593851715326,
      "rewards-max/forensics_tiou_reward": 0.8737291395664215,
      "rewards/forensics_sinkhorn_reward": 0.6641461253166199,
      "rewards/forensics_strict_edge_reward": 0.011634299531579018,
      "rewards/forensics_tiou_reward": 0.3742515444755554,
      "step": 184
    },
    {
      "completion_length": 20.125,
      "epoch": 0.5683563748079877,
      "grad_norm": 8.373784065246582,
      "kl": 0.43017578125,
      "learning_rate": 8.584615384615385e-07,
      "loss": -0.0482,
      "reward": 1.0553726553916931,
      "reward_std": 0.3045947998762131,
      "reward_std_all": 0.2781805470585823,
      "rewards-max/forensics_sinkhorn_reward": 0.9365101754665375,
      "rewards-max/forensics_strict_edge_reward": 0.20731645077466965,
      "rewards-max/forensics_tiou_reward": 0.8499695658683777,
      "rewards/forensics_sinkhorn_reward": 0.6511255204677582,
      "rewards/forensics_strict_edge_reward": 0.013147582300007343,
      "rewards/forensics_tiou_reward": 0.39109960198402405,
      "step": 185
    },
    {
      "completion_length": 19.34375,
      "epoch": 0.5714285714285714,
      "grad_norm": 6.21439266204834,
      "kl": 0.14990234375,
      "learning_rate": 8.576923076923076e-07,
      "loss": -0.0678,
      "reward": 1.0673664808273315,
      "reward_std": 0.18326665461063385,
      "reward_std_all": 0.15835320949554443,
      "rewards-max/forensics_sinkhorn_reward": 0.9155329167842865,
      "rewards-max/forensics_strict_edge_reward": 0.38022976741194725,
      "rewards-max/forensics_tiou_reward": 0.8147512674331665,
      "rewards/forensics_sinkhorn_reward": 0.6630463004112244,
      "rewards/forensics_strict_edge_reward": 0.02763105323538184,
      "rewards/forensics_tiou_reward": 0.3766890615224838,
      "step": 186
    },
    {
      "completion_length": 19.71875,
      "epoch": 0.5745007680491552,
      "grad_norm": 8.716621398925781,
      "kl": 0.203125,
      "learning_rate": 8.569230769230769e-07,
      "loss": -0.0164,
      "reward": 0.9709804058074951,
      "reward_std": 0.24653872847557068,
      "reward_std_all": 0.22097942233085632,
      "rewards-max/forensics_sinkhorn_reward": 0.9167070388793945,
      "rewards-max/forensics_strict_edge_reward": 0.06081283465027809,
      "rewards-max/forensics_tiou_reward": 0.8317941427230835,
      "rewards/forensics_sinkhorn_reward": 0.6364383101463318,
      "rewards/forensics_strict_edge_reward": 0.0031770410714671016,
      "rewards/forensics_tiou_reward": 0.33136501908302307,
      "step": 187
    },
    {
      "completion_length": 19.0625,
      "epoch": 0.5775729646697388,
      "grad_norm": 6.66377067565918,
      "kl": 0.1357421875,
      "learning_rate": 8.561538461538461e-07,
      "loss": 0.0289,
      "reward": 1.1003738641738892,
      "reward_std": 0.2799859046936035,
      "reward_std_all": 0.26087982952594757,
      "rewards-max/forensics_sinkhorn_reward": 0.9624603688716888,
      "rewards-max/forensics_strict_edge_reward": 0.4968075789511204,
      "rewards-max/forensics_tiou_reward": 0.9323046505451202,
      "rewards/forensics_sinkhorn_reward": 0.6748879253864288,
      "rewards/forensics_strict_edge_reward": 0.027427385561168194,
      "rewards/forensics_tiou_reward": 0.39805854856967926,
      "step": 188
    },
    {
      "completion_length": 19.78125,
      "epoch": 0.5806451612903226,
      "grad_norm": 6.6778178215026855,
      "kl": 0.15576171875,
      "learning_rate": 8.553846153846154e-07,
      "loss": -0.0357,
      "reward": 1.136844277381897,
      "reward_std": 0.22860683500766754,
      "reward_std_all": 0.21459074318408966,
      "rewards-max/forensics_sinkhorn_reward": 0.9706047177314758,
      "rewards-max/forensics_strict_edge_reward": 0.3196670934557915,
      "rewards-max/forensics_tiou_reward": 0.9460224211215973,
      "rewards/forensics_sinkhorn_reward": 0.6853673160076141,
      "rewards/forensics_strict_edge_reward": 0.01979718077927828,
      "rewards/forensics_tiou_reward": 0.43167978525161743,
      "step": 189
    },
    {
      "completion_length": 19.25,
      "epoch": 0.5837173579109063,
      "grad_norm": 7.289783477783203,
      "kl": 0.197265625,
      "learning_rate": 8.546153846153846e-07,
      "loss": 0.0119,
      "reward": 1.0633885264396667,
      "reward_std": 0.2646891102194786,
      "reward_std_all": 0.241511270403862,
      "rewards-max/forensics_sinkhorn_reward": 0.9149303138256073,
      "rewards-max/forensics_strict_edge_reward": 0.23053402453660965,
      "rewards-max/forensics_tiou_reward": 0.8212811648845673,
      "rewards/forensics_sinkhorn_reward": 0.6655019223690033,
      "rewards/forensics_strict_edge_reward": 0.015045605599880219,
      "rewards/forensics_tiou_reward": 0.3828410655260086,
      "step": 190
    },
    {
      "completion_length": 20.109375,
      "epoch": 0.5867895545314901,
      "grad_norm": 7.276024341583252,
      "kl": 0.22998046875,
      "learning_rate": 8.538461538461537e-07,
      "loss": -0.0346,
      "reward": 1.1146600246429443,
      "reward_std": 0.2651011794805527,
      "reward_std_all": 0.24880339950323105,
      "rewards-max/forensics_sinkhorn_reward": 0.9213087558746338,
      "rewards-max/forensics_strict_edge_reward": 0.2799966372549534,
      "rewards-max/forensics_tiou_reward": 0.8462398946285248,
      "rewards/forensics_sinkhorn_reward": 0.6859302520751953,
      "rewards/forensics_strict_edge_reward": 0.018199006095528603,
      "rewards/forensics_tiou_reward": 0.4105307310819626,
      "step": 191
    },
    {
      "completion_length": 19.640625,
      "epoch": 0.5898617511520737,
      "grad_norm": 7.964563846588135,
      "kl": 0.15478515625,
      "learning_rate": 8.530769230769231e-07,
      "loss": -0.0863,
      "reward": 1.239334225654602,
      "reward_std": 0.24518005549907684,
      "reward_std_all": 0.22264843434095383,
      "rewards-max/forensics_sinkhorn_reward": 0.9754282832145691,
      "rewards-max/forensics_strict_edge_reward": 0.21105021238327026,
      "rewards-max/forensics_tiou_reward": 0.9606500267982483,
      "rewards/forensics_sinkhorn_reward": 0.7313124239444733,
      "rewards/forensics_strict_edge_reward": 0.020303075201809406,
      "rewards/forensics_tiou_reward": 0.48771870136260986,
      "step": 192
    },
    {
      "completion_length": 20.0625,
      "epoch": 0.5929339477726574,
      "grad_norm": 6.178680896759033,
      "kl": 0.18505859375,
      "learning_rate": 8.523076923076923e-07,
      "loss": -0.0105,
      "reward": 1.0857120752334595,
      "reward_std": 0.3316560387611389,
      "reward_std_all": 0.2964085265994072,
      "rewards-max/forensics_sinkhorn_reward": 0.952980250120163,
      "rewards-max/forensics_strict_edge_reward": 0.3184485696256161,
      "rewards-max/forensics_tiou_reward": 0.9111136198043823,
      "rewards/forensics_sinkhorn_reward": 0.6752219200134277,
      "rewards/forensics_strict_edge_reward": 0.026911256602033973,
      "rewards/forensics_tiou_reward": 0.38357892632484436,
      "step": 193
    },
    {
      "completion_length": 20.453125,
      "epoch": 0.5960061443932412,
      "grad_norm": 6.665913105010986,
      "kl": 0.30517578125,
      "learning_rate": 8.515384615384616e-07,
      "loss": 0.0335,
      "reward": 1.093560367822647,
      "reward_std": 0.18554046750068665,
      "reward_std_all": 0.19773341715335846,
      "rewards-max/forensics_sinkhorn_reward": 0.941678911447525,
      "rewards-max/forensics_strict_edge_reward": 0.05067023797892034,
      "rewards-max/forensics_tiou_reward": 0.8729842603206635,
      "rewards/forensics_sinkhorn_reward": 0.6794516742229462,
      "rewards/forensics_strict_edge_reward": 0.004233836472849362,
      "rewards/forensics_tiou_reward": 0.409874826669693,
      "step": 194
    },
    {
      "completion_length": 19.875,
      "epoch": 0.5990783410138248,
      "grad_norm": 7.522073745727539,
      "kl": 0.2138671875,
      "learning_rate": 8.507692307692307e-07,
      "loss": -0.0549,
      "reward": 1.0780934691429138,
      "reward_std": 0.24227182567119598,
      "reward_std_all": 0.21546247601509094,
      "rewards-max/forensics_sinkhorn_reward": 0.9269751906394958,
      "rewards-max/forensics_strict_edge_reward": 0.06664478126913309,
      "rewards-max/forensics_tiou_reward": 0.8378656506538391,
      "rewards/forensics_sinkhorn_reward": 0.6580113172531128,
      "rewards/forensics_strict_edge_reward": 0.0059722441947087646,
      "rewards/forensics_tiou_reward": 0.4141099154949188,
      "step": 195
    },
    {
      "completion_length": 20.90625,
      "epoch": 0.6021505376344086,
      "grad_norm": 8.243013381958008,
      "kl": 0.1708984375,
      "learning_rate": 8.499999999999999e-07,
      "loss": 0.1325,
      "reward": 0.8957524597644806,
      "reward_std": 0.19306257367134094,
      "reward_std_all": 0.18241803348064423,
      "rewards-max/forensics_sinkhorn_reward": 0.922459214925766,
      "rewards-max/forensics_strict_edge_reward": 0.07955422066152096,
      "rewards-max/forensics_tiou_reward": 0.7961633503437042,
      "rewards/forensics_sinkhorn_reward": 0.5685087144374847,
      "rewards/forensics_strict_edge_reward": 0.009234416182152927,
      "rewards/forensics_tiou_reward": 0.3180093914270401,
      "step": 196
    },
    {
      "completion_length": 20.90625,
      "epoch": 0.6052227342549923,
      "grad_norm": 7.903574466705322,
      "kl": 0.24365234375,
      "learning_rate": 8.492307692307692e-07,
      "loss": -0.0599,
      "reward": 0.9816498756408691,
      "reward_std": 0.25236430764198303,
      "reward_std_all": 0.26695726811885834,
      "rewards-max/forensics_sinkhorn_reward": 0.9605794847011566,
      "rewards-max/forensics_strict_edge_reward": 0.18669800739735365,
      "rewards-max/forensics_tiou_reward": 0.9185743629932404,
      "rewards/forensics_sinkhorn_reward": 0.6157990992069244,
      "rewards/forensics_strict_edge_reward": 0.010841542389243841,
      "rewards/forensics_tiou_reward": 0.3550092577934265,
      "step": 197
    },
    {
      "completion_length": 20.296875,
      "epoch": 0.6082949308755761,
      "grad_norm": 7.83865213394165,
      "kl": 0.181640625,
      "learning_rate": 8.484615384615384e-07,
      "loss": 0.0369,
      "reward": 0.9726820290088654,
      "reward_std": 0.2581886053085327,
      "reward_std_all": 0.2379942536354065,
      "rewards-max/forensics_sinkhorn_reward": 0.9576283693313599,
      "rewards-max/forensics_strict_edge_reward": 0.07423840020783246,
      "rewards-max/forensics_tiou_reward": 0.9122278690338135,
      "rewards/forensics_sinkhorn_reward": 0.6314103901386261,
      "rewards/forensics_strict_edge_reward": 0.006168937106849626,
      "rewards/forensics_tiou_reward": 0.33510276675224304,
      "step": 198
    },
    {
      "completion_length": 19.4375,
      "epoch": 0.6113671274961597,
      "grad_norm": 4.87379264831543,
      "kl": 0.212890625,
      "learning_rate": 8.476923076923077e-07,
      "loss": -0.0433,
      "reward": 1.239975094795227,
      "reward_std": 0.2894885465502739,
      "reward_std_all": 0.26166799664497375,
      "rewards-max/forensics_sinkhorn_reward": 0.9704789221286774,
      "rewards-max/forensics_strict_edge_reward": 0.4307430237531662,
      "rewards-max/forensics_tiou_reward": 0.9502265751361847,
      "rewards/forensics_sinkhorn_reward": 0.7269569635391235,
      "rewards/forensics_strict_edge_reward": 0.036422424018383026,
      "rewards/forensics_tiou_reward": 0.4765956997871399,
      "step": 199
    },
    {
      "completion_length": 19.78125,
      "epoch": 0.6144393241167435,
      "grad_norm": 7.326886177062988,
      "kl": 0.1904296875,
      "learning_rate": 8.469230769230769e-07,
      "loss": -0.0326,
      "reward": 1.2174378633499146,
      "reward_std": 0.253757543861866,
      "reward_std_all": 0.28823432326316833,
      "rewards-max/forensics_sinkhorn_reward": 0.9640413224697113,
      "rewards-max/forensics_strict_edge_reward": 0.4263888746500015,
      "rewards-max/forensics_tiou_reward": 0.9363939464092255,
      "rewards/forensics_sinkhorn_reward": 0.6995802223682404,
      "rewards/forensics_strict_edge_reward": 0.026669122278690338,
      "rewards/forensics_tiou_reward": 0.49118858575820923,
      "step": 200
    },
    {
      "completion_length": 21.234375,
      "epoch": 0.6175115207373272,
      "grad_norm": 6.2343220710754395,
      "kl": 0.130859375,
      "learning_rate": 8.461538461538461e-07,
      "loss": -0.038,
      "reward": 1.1361756324768066,
      "reward_std": 0.29026931524276733,
      "reward_std_all": 0.2673271968960762,
      "rewards-max/forensics_sinkhorn_reward": 0.9641131162643433,
      "rewards-max/forensics_strict_edge_reward": 0.2972450405359268,
      "rewards-max/forensics_tiou_reward": 0.9372246563434601,
      "rewards/forensics_sinkhorn_reward": 0.6874806582927704,
      "rewards/forensics_strict_edge_reward": 0.023512191139161587,
      "rewards/forensics_tiou_reward": 0.4251827597618103,
      "step": 201
    },
    {
      "completion_length": 22.203125,
      "epoch": 0.620583717357911,
      "grad_norm": 8.220473289489746,
      "kl": 0.453125,
      "learning_rate": 8.453846153846154e-07,
      "loss": 0.0125,
      "reward": 0.9108411967754364,
      "reward_std": 0.2229568362236023,
      "reward_std_all": 0.20667319744825363,
      "rewards-max/forensics_sinkhorn_reward": 0.9396184682846069,
      "rewards-max/forensics_strict_edge_reward": 0.010124027402071079,
      "rewards-max/forensics_tiou_reward": 0.8697703182697296,
      "rewards/forensics_sinkhorn_reward": 0.601071834564209,
      "rewards/forensics_strict_edge_reward": 0.0004741543017701133,
      "rewards/forensics_tiou_reward": 0.30929526686668396,
      "step": 202
    },
    {
      "completion_length": 20.515625,
      "epoch": 0.6236559139784946,
      "grad_norm": 6.168328762054443,
      "kl": 0.142578125,
      "learning_rate": 8.446153846153846e-07,
      "loss": -0.0037,
      "reward": 1.060707151889801,
      "reward_std": 0.22978537529706955,
      "reward_std_all": 0.23421762883663177,
      "rewards-max/forensics_sinkhorn_reward": 0.9359557926654816,
      "rewards-max/forensics_strict_edge_reward": 0.3176007568836212,
      "rewards-max/forensics_tiou_reward": 0.8711283802986145,
      "rewards/forensics_sinkhorn_reward": 0.6637193560600281,
      "rewards/forensics_strict_edge_reward": 0.015227177180349827,
      "rewards/forensics_tiou_reward": 0.38176071643829346,
      "step": 203
    },
    {
      "completion_length": 20.9375,
      "epoch": 0.6267281105990783,
      "grad_norm": 11.574390411376953,
      "kl": 0.2509765625,
      "learning_rate": 8.438461538461537e-07,
      "loss": -0.029,
      "reward": 1.085101306438446,
      "reward_std": 0.24846727401018143,
      "reward_std_all": 0.25001125782728195,
      "rewards-max/forensics_sinkhorn_reward": 0.9255022406578064,
      "rewards-max/forensics_strict_edge_reward": 0.23862625006586313,
      "rewards-max/forensics_tiou_reward": 0.8501511514186859,
      "rewards/forensics_sinkhorn_reward": 0.6371733546257019,
      "rewards/forensics_strict_edge_reward": 0.0234757088473998,
      "rewards/forensics_tiou_reward": 0.4244523048400879,
      "step": 204
    },
    {
      "completion_length": 21.34375,
      "epoch": 0.6298003072196621,
      "grad_norm": 7.518248081207275,
      "kl": 0.21435546875,
      "learning_rate": 8.43076923076923e-07,
      "loss": -0.0191,
      "reward": 1.150402307510376,
      "reward_std": 0.3099987357854843,
      "reward_std_all": 0.2858316972851753,
      "rewards-max/forensics_sinkhorn_reward": 0.9587107002735138,
      "rewards-max/forensics_strict_edge_reward": 0.18334465473890305,
      "rewards-max/forensics_tiou_reward": 0.9241927266120911,
      "rewards/forensics_sinkhorn_reward": 0.679846465587616,
      "rewards/forensics_strict_edge_reward": 0.009753059130162,
      "rewards/forensics_tiou_reward": 0.46080282330513,
      "step": 205
    },
    {
      "completion_length": 20.109375,
      "epoch": 0.6328725038402457,
      "grad_norm": 8.408854484558105,
      "kl": 0.20751953125,
      "learning_rate": 8.423076923076923e-07,
      "loss": -0.0627,
      "reward": 1.179326057434082,
      "reward_std": 0.3017275333404541,
      "reward_std_all": 0.295999214053154,
      "rewards-max/forensics_sinkhorn_reward": 0.9796804785728455,
      "rewards-max/forensics_strict_edge_reward": 0.34317469596862793,
      "rewards-max/forensics_tiou_reward": 0.9731220006942749,
      "rewards/forensics_sinkhorn_reward": 0.7043595016002655,
      "rewards/forensics_strict_edge_reward": 0.0370585824130103,
      "rewards/forensics_tiou_reward": 0.4379079192876816,
      "step": 206
    },
    {
      "completion_length": 20.125,
      "epoch": 0.6359447004608295,
      "grad_norm": 6.9705281257629395,
      "kl": 0.16162109375,
      "learning_rate": 8.415384615384616e-07,
      "loss": 0.0927,
      "reward": 1.1796373128890991,
      "reward_std": 0.20162997394800186,
      "reward_std_all": 0.1898781582713127,
      "rewards-max/forensics_sinkhorn_reward": 0.9511536955833435,
      "rewards-max/forensics_strict_edge_reward": 0.03328432235866785,
      "rewards-max/forensics_tiou_reward": 0.8946061432361603,
      "rewards/forensics_sinkhorn_reward": 0.7199934124946594,
      "rewards/forensics_strict_edge_reward": 0.0034232084872201085,
      "rewards/forensics_tiou_reward": 0.45622068643569946,
      "step": 207
    },
    {
      "completion_length": 19.4375,
      "epoch": 0.6390168970814132,
      "grad_norm": 7.0497941970825195,
      "kl": 0.1552734375,
      "learning_rate": 8.407692307692308e-07,
      "loss": -0.0096,
      "reward": 1.265533208847046,
      "reward_std": 0.24018559604883194,
      "reward_std_all": 0.27326375246047974,
      "rewards-max/forensics_sinkhorn_reward": 0.9484151899814606,
      "rewards-max/forensics_strict_edge_reward": 0.23236726224422455,
      "rewards-max/forensics_tiou_reward": 0.8914314210414886,
      "rewards/forensics_sinkhorn_reward": 0.7402569949626923,
      "rewards/forensics_strict_edge_reward": 0.024287311360239983,
      "rewards/forensics_tiou_reward": 0.5009889304637909,
      "step": 208
    },
    {
      "completion_length": 19.953125,
      "epoch": 0.642089093701997,
      "grad_norm": 6.203027248382568,
      "kl": 0.1650390625,
      "learning_rate": 8.399999999999999e-07,
      "loss": 0.0396,
      "reward": 0.8367346227169037,
      "reward_std": 0.16814864426851273,
      "reward_std_all": 0.17313405126333237,
      "rewards-max/forensics_sinkhorn_reward": 0.8595321476459503,
      "rewards-max/forensics_strict_edge_reward": 0.02409369187807897,
      "rewards-max/forensics_tiou_reward": 0.6924143433570862,
      "rewards/forensics_sinkhorn_reward": 0.5657656788825989,
      "rewards/forensics_strict_edge_reward": 0.0017847574830511803,
      "rewards/forensics_tiou_reward": 0.2691841796040535,
      "step": 209
    },
    {
      "completion_length": 20.015625,
      "epoch": 0.6451612903225806,
      "grad_norm": 11.668509483337402,
      "kl": 0.337890625,
      "learning_rate": 8.392307692307692e-07,
      "loss": 0.0205,
      "reward": 0.8997438549995422,
      "reward_std": 0.23716361075639725,
      "reward_std_all": 0.21422307938337326,
      "rewards-max/forensics_sinkhorn_reward": 0.8246447145938873,
      "rewards-max/forensics_strict_edge_reward": 0.13922217302024364,
      "rewards-max/forensics_tiou_reward": 0.7030816078186035,
      "rewards/forensics_sinkhorn_reward": 0.571878045797348,
      "rewards/forensics_strict_edge_reward": 0.010262800613418221,
      "rewards/forensics_tiou_reward": 0.3176030218601227,
      "step": 210
    },
    {
      "completion_length": 20.015625,
      "epoch": 0.6482334869431644,
      "grad_norm": 6.911232948303223,
      "kl": 0.1533203125,
      "learning_rate": 8.384615384615384e-07,
      "loss": 0.0812,
      "reward": 1.1285844445228577,
      "reward_std": 0.22473633289337158,
      "reward_std_all": 0.23478107899427414,
      "rewards-max/forensics_sinkhorn_reward": 0.9564495086669922,
      "rewards-max/forensics_strict_edge_reward": 0.2544358931481838,
      "rewards-max/forensics_tiou_reward": 0.9167895317077637,
      "rewards/forensics_sinkhorn_reward": 0.6932522356510162,
      "rewards/forensics_strict_edge_reward": 0.01431635464541614,
      "rewards/forensics_tiou_reward": 0.42101579904556274,
      "step": 211
    },
    {
      "completion_length": 20.140625,
      "epoch": 0.6513056835637481,
      "grad_norm": 7.821379661560059,
      "kl": 0.234375,
      "learning_rate": 8.376923076923076e-07,
      "loss": 0.0632,
      "reward": 1.1428831815719604,
      "reward_std": 0.3066203147172928,
      "reward_std_all": 0.27883361279964447,
      "rewards-max/forensics_sinkhorn_reward": 0.9596489369869232,
      "rewards-max/forensics_strict_edge_reward": 0.20137850940227509,
      "rewards-max/forensics_tiou_reward": 0.9162658452987671,
      "rewards/forensics_sinkhorn_reward": 0.6821742057800293,
      "rewards/forensics_strict_edge_reward": 0.017102353274822235,
      "rewards/forensics_tiou_reward": 0.44360658526420593,
      "step": 212
    },
    {
      "completion_length": 20.828125,
      "epoch": 0.6543778801843319,
      "grad_norm": 5.991766452789307,
      "kl": 0.158447265625,
      "learning_rate": 8.369230769230769e-07,
      "loss": 0.0482,
      "reward": 0.9924529194831848,
      "reward_std": 0.23695964366197586,
      "reward_std_all": 0.21071387827396393,
      "rewards-max/forensics_sinkhorn_reward": 0.9581696391105652,
      "rewards-max/forensics_strict_edge_reward": 0.021485736593604088,
      "rewards-max/forensics_tiou_reward": 0.9264779388904572,
      "rewards/forensics_sinkhorn_reward": 0.6389707624912262,
      "rewards/forensics_strict_edge_reward": 0.0014864570694044232,
      "rewards/forensics_tiou_reward": 0.35199567675590515,
      "step": 213
    },
    {
      "completion_length": 19.96875,
      "epoch": 0.6574500768049155,
      "grad_norm": 5.937542915344238,
      "kl": 0.2578125,
      "learning_rate": 8.361538461538461e-07,
      "loss": 0.044,
      "reward": 1.057811290025711,
      "reward_std": 0.27623216807842255,
      "reward_std_all": 0.2594356834888458,
      "rewards-max/forensics_sinkhorn_reward": 0.9178709387779236,
      "rewards-max/forensics_strict_edge_reward": 0.12037250585854053,
      "rewards-max/forensics_tiou_reward": 0.8224874138832092,
      "rewards/forensics_sinkhorn_reward": 0.6577591001987457,
      "rewards/forensics_strict_edge_reward": 0.020801521051907912,
      "rewards/forensics_tiou_reward": 0.379250705242157,
      "step": 214
    },
    {
      "completion_length": 21.03125,
      "epoch": 0.6605222734254992,
      "grad_norm": 6.083658695220947,
      "kl": 0.2177734375,
      "learning_rate": 8.353846153846154e-07,
      "loss": -0.0419,
      "reward": 1.1441907286643982,
      "reward_std": 0.2309955209493637,
      "reward_std_all": 0.26047227531671524,
      "rewards-max/forensics_sinkhorn_reward": 0.9211911261081696,
      "rewards-max/forensics_strict_edge_reward": 0.4012645184993744,
      "rewards-max/forensics_tiou_reward": 0.8196170032024384,
      "rewards/forensics_sinkhorn_reward": 0.6841279566287994,
      "rewards/forensics_strict_edge_reward": 0.024751615710556507,
      "rewards/forensics_tiou_reward": 0.43531110882759094,
      "step": 215
    },
    {
      "completion_length": 20.015625,
      "epoch": 0.663594470046083,
      "grad_norm": 7.698914527893066,
      "kl": 0.15234375,
      "learning_rate": 8.346153846153846e-07,
      "loss": -0.0137,
      "reward": 1.09803307056427,
      "reward_std": 0.19290512800216675,
      "reward_std_all": 0.18619810044765472,
      "rewards-max/forensics_sinkhorn_reward": 0.9693824946880341,
      "rewards-max/forensics_strict_edge_reward": 0.11933133378624916,
      "rewards-max/forensics_tiou_reward": 0.949015349149704,
      "rewards/forensics_sinkhorn_reward": 0.6941657364368439,
      "rewards/forensics_strict_edge_reward": 0.007286942098289728,
      "rewards/forensics_tiou_reward": 0.3965803533792496,
      "step": 216
    },
    {
      "completion_length": 19.90625,
      "epoch": 0.6666666666666666,
      "grad_norm": 6.858330249786377,
      "kl": 0.27978515625,
      "learning_rate": 8.338461538461537e-07,
      "loss": 0.0379,
      "reward": 1.244979977607727,
      "reward_std": 0.2756797969341278,
      "reward_std_all": 0.2640867978334427,
      "rewards-max/forensics_sinkhorn_reward": 0.9454298913478851,
      "rewards-max/forensics_strict_edge_reward": 0.28422017209231853,
      "rewards-max/forensics_tiou_reward": 0.8877534866333008,
      "rewards/forensics_sinkhorn_reward": 0.7360211908817291,
      "rewards/forensics_strict_edge_reward": 0.025760240852832794,
      "rewards/forensics_tiou_reward": 0.4831985533237457,
      "step": 217
    },
    {
      "completion_length": 22.015625,
      "epoch": 0.6697388632872504,
      "grad_norm": 7.949079513549805,
      "kl": 0.130859375,
      "learning_rate": 8.33076923076923e-07,
      "loss": -0.0701,
      "reward": 1.1560010313987732,
      "reward_std": 0.28197574615478516,
      "reward_std_all": 0.26699769496917725,
      "rewards-max/forensics_sinkhorn_reward": 0.9326403439044952,
      "rewards-max/forensics_strict_edge_reward": 0.0910189300775528,
      "rewards-max/forensics_tiou_reward": 0.8641828000545502,
      "rewards/forensics_sinkhorn_reward": 0.6828080415725708,
      "rewards/forensics_strict_edge_reward": 0.007107424549758434,
      "rewards/forensics_tiou_reward": 0.4660855680704117,
      "step": 218
    },
    {
      "completion_length": 20.25,
      "epoch": 0.6728110599078341,
      "grad_norm": 8.885168075561523,
      "kl": 0.2021484375,
      "learning_rate": 8.323076923076922e-07,
      "loss": 0.0651,
      "reward": 1.0410871505737305,
      "reward_std": 0.24470921605825424,
      "reward_std_all": 0.22319460660219193,
      "rewards-max/forensics_sinkhorn_reward": 0.9079444110393524,
      "rewards-max/forensics_strict_edge_reward": 0.05878341570496559,
      "rewards-max/forensics_tiou_reward": 0.8044801950454712,
      "rewards/forensics_sinkhorn_reward": 0.6763817369937897,
      "rewards/forensics_strict_edge_reward": 0.004296577535569668,
      "rewards/forensics_tiou_reward": 0.36040888726711273,
      "step": 219
    },
    {
      "completion_length": 20.265625,
      "epoch": 0.6758832565284179,
      "grad_norm": 9.453909873962402,
      "kl": 0.25732421875,
      "learning_rate": 8.315384615384616e-07,
      "loss": 0.0335,
      "reward": 0.9893579483032227,
      "reward_std": 0.2977096736431122,
      "reward_std_all": 0.2619529664516449,
      "rewards-max/forensics_sinkhorn_reward": 0.9243277609348297,
      "rewards-max/forensics_strict_edge_reward": 0.3704701326787472,
      "rewards-max/forensics_tiou_reward": 0.8454222083091736,
      "rewards/forensics_sinkhorn_reward": 0.6295120120048523,
      "rewards/forensics_strict_edge_reward": 0.016993718221783638,
      "rewards/forensics_tiou_reward": 0.34285223484039307,
      "step": 220
    },
    {
      "completion_length": 19.234375,
      "epoch": 0.6789554531490015,
      "grad_norm": 8.034399032592773,
      "kl": 0.384765625,
      "learning_rate": 8.307692307692308e-07,
      "loss": -0.0093,
      "reward": 1.1064980030059814,
      "reward_std": 0.2471105381846428,
      "reward_std_all": 0.22587507963180542,
      "rewards-max/forensics_sinkhorn_reward": 0.957838624715805,
      "rewards-max/forensics_strict_edge_reward": 0.26087334007024765,
      "rewards-max/forensics_tiou_reward": 0.9209162592887878,
      "rewards/forensics_sinkhorn_reward": 0.683376669883728,
      "rewards/forensics_strict_edge_reward": 0.022241147235035896,
      "rewards/forensics_tiou_reward": 0.4008801132440567,
      "step": 221
    },
    {
      "completion_length": 20.734375,
      "epoch": 0.6820276497695853,
      "grad_norm": 8.186382293701172,
      "kl": 0.1806640625,
      "learning_rate": 8.299999999999999e-07,
      "loss": -0.1191,
      "reward": 1.0284669399261475,
      "reward_std": 0.1247178167104721,
      "reward_std_all": 0.126458328217268,
      "rewards-max/forensics_sinkhorn_reward": 0.9658809900283813,
      "rewards-max/forensics_strict_edge_reward": 0.09709111507982016,
      "rewards-max/forensics_tiou_reward": 0.9369331896305084,
      "rewards/forensics_sinkhorn_reward": 0.6387580931186676,
      "rewards/forensics_strict_edge_reward": 0.004160217009484768,
      "rewards/forensics_tiou_reward": 0.3855486661195755,
      "step": 222
    },
    {
      "completion_length": 19.59375,
      "epoch": 0.685099846390169,
      "grad_norm": 9.858696937561035,
      "kl": 0.220703125,
      "learning_rate": 8.292307692307692e-07,
      "loss": -0.0034,
      "reward": 1.0825817584991455,
      "reward_std": 0.21818770468235016,
      "reward_std_all": 0.19870369136333466,
      "rewards-max/forensics_sinkhorn_reward": 0.9208627343177795,
      "rewards-max/forensics_strict_edge_reward": 0.035673994570970535,
      "rewards-max/forensics_tiou_reward": 0.8364569842815399,
      "rewards/forensics_sinkhorn_reward": 0.671848714351654,
      "rewards/forensics_strict_edge_reward": 0.003664657473564148,
      "rewards/forensics_tiou_reward": 0.40706847608089447,
      "step": 223
    },
    {
      "completion_length": 19.984375,
      "epoch": 0.6881720430107527,
      "grad_norm": 7.205101013183594,
      "kl": 0.40234375,
      "learning_rate": 8.284615384615384e-07,
      "loss": 0.0929,
      "reward": 1.084884762763977,
      "reward_std": 0.23472149670124054,
      "reward_std_all": 0.21027635037899017,
      "rewards-max/forensics_sinkhorn_reward": 0.9543707966804504,
      "rewards-max/forensics_strict_edge_reward": 0.46020403504371643,
      "rewards-max/forensics_tiou_reward": 0.9115542769432068,
      "rewards/forensics_sinkhorn_reward": 0.6394184231758118,
      "rewards/forensics_strict_edge_reward": 0.031619906425476074,
      "rewards/forensics_tiou_reward": 0.4138465076684952,
      "step": 224
    },
    {
      "completion_length": 20.265625,
      "epoch": 0.6912442396313364,
      "grad_norm": 6.6835479736328125,
      "kl": 0.18505859375,
      "learning_rate": 8.276923076923077e-07,
      "loss": 0.081,
      "reward": 1.1404955387115479,
      "reward_std": 0.26083941757678986,
      "reward_std_all": 0.22947368025779724,
      "rewards-max/forensics_sinkhorn_reward": 0.9799984097480774,
      "rewards-max/forensics_strict_edge_reward": 0.6149742305278778,
      "rewards-max/forensics_tiou_reward": 0.9755249619483948,
      "rewards/forensics_sinkhorn_reward": 0.6831494569778442,
      "rewards/forensics_strict_edge_reward": 0.026901873759925365,
      "rewards/forensics_tiou_reward": 0.4304441511631012,
      "step": 225
    },
    {
      "completion_length": 20.21875,
      "epoch": 0.6943164362519201,
      "grad_norm": 7.537506580352783,
      "kl": 0.15869140625,
      "learning_rate": 8.269230769230768e-07,
      "loss": -0.0874,
      "reward": 1.0032515227794647,
      "reward_std": 0.19549575075507164,
      "reward_std_all": 0.17875409126281738,
      "rewards-max/forensics_sinkhorn_reward": 0.9443368017673492,
      "rewards-max/forensics_strict_edge_reward": 0.04958093259483576,
      "rewards-max/forensics_tiou_reward": 0.8880270421504974,
      "rewards/forensics_sinkhorn_reward": 0.6299907565116882,
      "rewards/forensics_strict_edge_reward": 0.003992733312770724,
      "rewards/forensics_tiou_reward": 0.369268000125885,
      "step": 226
    },
    {
      "completion_length": 19.609375,
      "epoch": 0.6973886328725039,
      "grad_norm": 7.65195894241333,
      "kl": 0.29931640625,
      "learning_rate": 8.261538461538461e-07,
      "loss": -0.0104,
      "reward": 1.2146709561347961,
      "reward_std": 0.26916657388210297,
      "reward_std_all": 0.23840030282735825,
      "rewards-max/forensics_sinkhorn_reward": 0.9614145755767822,
      "rewards-max/forensics_strict_edge_reward": 0.4277711436152458,
      "rewards-max/forensics_tiou_reward": 0.9235826432704926,
      "rewards/forensics_sinkhorn_reward": 0.7119105458259583,
      "rewards/forensics_strict_edge_reward": 0.0375896580517292,
      "rewards/forensics_tiou_reward": 0.4651707708835602,
      "step": 227
    },
    {
      "completion_length": 19.203125,
      "epoch": 0.7004608294930875,
      "grad_norm": 9.059178352355957,
      "kl": 0.3037109375,
      "learning_rate": 8.253846153846154e-07,
      "loss": -0.0635,
      "reward": 1.0199699997901917,
      "reward_std": 0.1801837980747223,
      "reward_std_all": 0.17970744520425797,
      "rewards-max/forensics_sinkhorn_reward": 0.9669459164142609,
      "rewards-max/forensics_strict_edge_reward": 0.33328156918287277,
      "rewards-max/forensics_tiou_reward": 0.9380422830581665,
      "rewards/forensics_sinkhorn_reward": 0.6297298073768616,
      "rewards/forensics_strict_edge_reward": 0.022656451910734177,
      "rewards/forensics_tiou_reward": 0.3675837069749832,
      "step": 228
    },
    {
      "completion_length": 19.546875,
      "epoch": 0.7035330261136713,
      "grad_norm": 7.1806254386901855,
      "kl": 0.18212890625,
      "learning_rate": 8.246153846153846e-07,
      "loss": -0.0681,
      "reward": 1.1275964975357056,
      "reward_std": 0.1971782147884369,
      "reward_std_all": 0.20599797368049622,
      "rewards-max/forensics_sinkhorn_reward": 0.9651416540145874,
      "rewards-max/forensics_strict_edge_reward": 0.5028069615364075,
      "rewards-max/forensics_tiou_reward": 0.9249891340732574,
      "rewards/forensics_sinkhorn_reward": 0.6754557192325592,
      "rewards/forensics_strict_edge_reward": 0.02709181047976017,
      "rewards/forensics_tiou_reward": 0.42504899203777313,
      "step": 229
    },
    {
      "completion_length": 20.015625,
      "epoch": 0.706605222734255,
      "grad_norm": 9.264419555664062,
      "kl": 0.30078125,
      "learning_rate": 8.238461538461538e-07,
      "loss": -0.1106,
      "reward": 1.0546886920928955,
      "reward_std": 0.28679537773132324,
      "reward_std_all": 0.2876502200961113,
      "rewards-max/forensics_sinkhorn_reward": 0.8926992118358612,
      "rewards-max/forensics_strict_edge_reward": 0.3375745862722397,
      "rewards-max/forensics_tiou_reward": 0.8054089248180389,
      "rewards/forensics_sinkhorn_reward": 0.6519205570220947,
      "rewards/forensics_strict_edge_reward": 0.03240938484668732,
      "rewards/forensics_tiou_reward": 0.37035873532295227,
      "step": 230
    },
    {
      "completion_length": 20.375,
      "epoch": 0.7096774193548387,
      "grad_norm": 8.093185424804688,
      "kl": 0.25341796875,
      "learning_rate": 8.23076923076923e-07,
      "loss": -0.0493,
      "reward": 1.0698301196098328,
      "reward_std": 0.17442023009061813,
      "reward_std_all": 0.16644082963466644,
      "rewards-max/forensics_sinkhorn_reward": 0.9436911642551422,
      "rewards-max/forensics_strict_edge_reward": 0.1759628690779209,
      "rewards-max/forensics_tiou_reward": 0.8901375830173492,
      "rewards/forensics_sinkhorn_reward": 0.6444480717182159,
      "rewards/forensics_strict_edge_reward": 0.011718481313437223,
      "rewards/forensics_tiou_reward": 0.41366352140903473,
      "step": 231
    },
    {
      "completion_length": 20.125,
      "epoch": 0.7127496159754224,
      "grad_norm": 8.19355297088623,
      "kl": 0.22314453125,
      "learning_rate": 8.223076923076922e-07,
      "loss": 0.0498,
      "reward": 1.1702206134796143,
      "reward_std": 0.31563399732112885,
      "reward_std_all": 0.28123752772808075,
      "rewards-max/forensics_sinkhorn_reward": 0.9678794145584106,
      "rewards-max/forensics_strict_edge_reward": 0.36461610347032547,
      "rewards-max/forensics_tiou_reward": 0.9446841776371002,
      "rewards/forensics_sinkhorn_reward": 0.6928583979606628,
      "rewards/forensics_strict_edge_reward": 0.018380722030997276,
      "rewards/forensics_tiou_reward": 0.4589815139770508,
      "step": 232
    },
    {
      "completion_length": 19.65625,
      "epoch": 0.7158218125960062,
      "grad_norm": 7.551730632781982,
      "kl": 0.2802734375,
      "learning_rate": 8.215384615384616e-07,
      "loss": -0.0394,
      "reward": 1.0394138097763062,
      "reward_std": 0.25041721761226654,
      "reward_std_all": 0.23303744941949844,
      "rewards-max/forensics_sinkhorn_reward": 0.9256546199321747,
      "rewards-max/forensics_strict_edge_reward": 0.4416893757879734,
      "rewards-max/forensics_tiou_reward": 0.8424073755741119,
      "rewards/forensics_sinkhorn_reward": 0.649158775806427,
      "rewards/forensics_strict_edge_reward": 0.03476916113868356,
      "rewards/forensics_tiou_reward": 0.35548587143421173,
      "step": 233
    },
    {
      "completion_length": 21.609375,
      "epoch": 0.7188940092165899,
      "grad_norm": 9.793098449707031,
      "kl": 0.16943359375,
      "learning_rate": 8.207692307692308e-07,
      "loss": -0.0374,
      "reward": 1.0615778267383575,
      "reward_std": 0.1982327178120613,
      "reward_std_all": 0.18556282669305801,
      "rewards-max/forensics_sinkhorn_reward": 0.914003312587738,
      "rewards-max/forensics_strict_edge_reward": 0.14081952348351479,
      "rewards-max/forensics_tiou_reward": 0.8126190006732941,
      "rewards/forensics_sinkhorn_reward": 0.642797440290451,
      "rewards/forensics_strict_edge_reward": 0.0073648064862936735,
      "rewards/forensics_tiou_reward": 0.41141557693481445,
      "step": 234
    },
    {
      "completion_length": 22.703125,
      "epoch": 0.7219662058371735,
      "grad_norm": 8.767109870910645,
      "kl": 0.214599609375,
      "learning_rate": 8.199999999999999e-07,
      "loss": 0.2105,
      "reward": 0.9597481489181519,
      "reward_std": 0.22074318677186966,
      "reward_std_all": 0.21285858005285263,
      "rewards-max/forensics_sinkhorn_reward": 0.9281772673130035,
      "rewards-max/forensics_strict_edge_reward": 0.09079281613230705,
      "rewards-max/forensics_tiou_reward": 0.850839376449585,
      "rewards/forensics_sinkhorn_reward": 0.6169110834598541,
      "rewards/forensics_strict_edge_reward": 0.005077243549749255,
      "rewards/forensics_tiou_reward": 0.3377598524093628,
      "step": 235
    },
    {
      "completion_length": 20.203125,
      "epoch": 0.7250384024577573,
      "grad_norm": 7.873169898986816,
      "kl": 0.14599609375,
      "learning_rate": 8.192307692307692e-07,
      "loss": -0.001,
      "reward": 1.0678898990154266,
      "reward_std": 0.1686365231871605,
      "reward_std_all": 0.16608501225709915,
      "rewards-max/forensics_sinkhorn_reward": 0.9517285525798798,
      "rewards-max/forensics_strict_edge_reward": 0.41770588979125023,
      "rewards-max/forensics_tiou_reward": 0.9101574718952179,
      "rewards/forensics_sinkhorn_reward": 0.6735145449638367,
      "rewards/forensics_strict_edge_reward": 0.02162932022474706,
      "rewards/forensics_tiou_reward": 0.372746080160141,
      "step": 236
    },
    {
      "completion_length": 22.765625,
      "epoch": 0.728110599078341,
      "grad_norm": 7.437342166900635,
      "kl": 0.16552734375,
      "learning_rate": 8.184615384615384e-07,
      "loss": -0.0809,
      "reward": 1.0399607419967651,
      "reward_std": 0.3388911932706833,
      "reward_std_all": 0.3590400516986847,
      "rewards-max/forensics_sinkhorn_reward": 0.9514362812042236,
      "rewards-max/forensics_strict_edge_reward": 0.20290684700012207,
      "rewards-max/forensics_tiou_reward": 0.9040705263614655,
      "rewards/forensics_sinkhorn_reward": 0.6349385678768158,
      "rewards/forensics_strict_edge_reward": 0.011698517482727766,
      "rewards/forensics_tiou_reward": 0.3933236449956894,
      "step": 237
    },
    {
      "completion_length": 19.828125,
      "epoch": 0.7311827956989247,
      "grad_norm": 8.531716346740723,
      "kl": 0.17041015625,
      "learning_rate": 8.176923076923077e-07,
      "loss": 0.011,
      "reward": 1.0731070041656494,
      "reward_std": 0.26319098472595215,
      "reward_std_all": 0.23132194578647614,
      "rewards-max/forensics_sinkhorn_reward": 0.9443538784980774,
      "rewards-max/forensics_strict_edge_reward": 0.43425891920924187,
      "rewards-max/forensics_tiou_reward": 0.8860384523868561,
      "rewards/forensics_sinkhorn_reward": 0.6642812788486481,
      "rewards/forensics_strict_edge_reward": 0.030182582908309996,
      "rewards/forensics_tiou_reward": 0.37864306569099426,
      "step": 238
    },
    {
      "completion_length": 25.265625,
      "epoch": 0.7342549923195084,
      "grad_norm": 7.471730709075928,
      "kl": 0.19677734375,
      "learning_rate": 8.169230769230769e-07,
      "loss": 0.0325,
      "reward": 1.1911477446556091,
      "reward_std": 0.27889376878738403,
      "reward_std_all": 0.2740000784397125,
      "rewards-max/forensics_sinkhorn_reward": 0.9640804827213287,
      "rewards-max/forensics_strict_edge_reward": 0.28077706694602966,
      "rewards-max/forensics_tiou_reward": 0.9313132464885712,
      "rewards/forensics_sinkhorn_reward": 0.6680390238761902,
      "rewards/forensics_strict_edge_reward": 0.027768593281507492,
      "rewards/forensics_tiou_reward": 0.49534013867378235,
      "step": 239
    },
    {
      "completion_length": 24.453125,
      "epoch": 0.7373271889400922,
      "grad_norm": 6.282108306884766,
      "kl": 0.20263671875,
      "learning_rate": 8.16153846153846e-07,
      "loss": -0.1098,
      "reward": 1.0736931562423706,
      "reward_std": 0.23631644248962402,
      "reward_std_all": 0.21945448964834213,
      "rewards-max/forensics_sinkhorn_reward": 0.9474853575229645,
      "rewards-max/forensics_strict_edge_reward": 0.21835794672369957,
      "rewards-max/forensics_tiou_reward": 0.8932444155216217,
      "rewards/forensics_sinkhorn_reward": 0.6209670305252075,
      "rewards/forensics_strict_edge_reward": 0.01879064249806106,
      "rewards/forensics_tiou_reward": 0.43393556773662567,
      "step": 240
    },
    {
      "completion_length": 22.421875,
      "epoch": 0.7403993855606759,
      "grad_norm": 8.299549102783203,
      "kl": 0.136474609375,
      "learning_rate": 8.153846153846154e-07,
      "loss": 0.0144,
      "reward": 1.1709542870521545,
      "reward_std": 0.24369575083255768,
      "reward_std_all": 0.21837124973535538,
      "rewards-max/forensics_sinkhorn_reward": 0.947643518447876,
      "rewards-max/forensics_strict_edge_reward": 0.0845158938318491,
      "rewards-max/forensics_tiou_reward": 0.9008306264877319,
      "rewards/forensics_sinkhorn_reward": 0.706682026386261,
      "rewards/forensics_strict_edge_reward": 0.005242125247605145,
      "rewards/forensics_tiou_reward": 0.4590301513671875,
      "step": 241
    },
    {
      "completion_length": 20.65625,
      "epoch": 0.7434715821812596,
      "grad_norm": 10.333229064941406,
      "kl": 0.19921875,
      "learning_rate": 8.146153846153846e-07,
      "loss": -0.01,
      "reward": 1.2105854749679565,
      "reward_std": 0.2956916391849518,
      "reward_std_all": 0.32763662934303284,
      "rewards-max/forensics_sinkhorn_reward": 0.9171724021434784,
      "rewards-max/forensics_strict_edge_reward": 0.4850485920906067,
      "rewards-max/forensics_tiou_reward": 0.8341173827648163,
      "rewards/forensics_sinkhorn_reward": 0.7006322145462036,
      "rewards/forensics_strict_edge_reward": 0.025983690284192562,
      "rewards/forensics_tiou_reward": 0.4839695245027542,
      "step": 242
    },
    {
      "completion_length": 21.40625,
      "epoch": 0.7465437788018433,
      "grad_norm": 7.74158239364624,
      "kl": 0.19140625,
      "learning_rate": 8.138461538461538e-07,
      "loss": 0.0383,
      "reward": 1.3461707830429077,
      "reward_std": 0.1581442505121231,
      "reward_std_all": 0.16068976745009422,
      "rewards-max/forensics_sinkhorn_reward": 0.9684934020042419,
      "rewards-max/forensics_strict_edge_reward": 0.09807267226278782,
      "rewards-max/forensics_tiou_reward": 0.9479595124721527,
      "rewards/forensics_sinkhorn_reward": 0.7684176862239838,
      "rewards/forensics_strict_edge_reward": 0.010732903145253658,
      "rewards/forensics_tiou_reward": 0.5670201480388641,
      "step": 243
    },
    {
      "completion_length": 24.03125,
      "epoch": 0.7496159754224271,
      "grad_norm": 9.741495132446289,
      "kl": 0.2802734375,
      "learning_rate": 8.13076923076923e-07,
      "loss": 0.085,
      "reward": 1.1490835845470428,
      "reward_std": 0.35427461564540863,
      "reward_std_all": 0.3169509768486023,
      "rewards-max/forensics_sinkhorn_reward": 0.9855606853961945,
      "rewards-max/forensics_strict_edge_reward": 0.8582656681537628,
      "rewards-max/forensics_tiou_reward": 0.9917255640029907,
      "rewards/forensics_sinkhorn_reward": 0.6655098795890808,
      "rewards/forensics_strict_edge_reward": 0.04472743347287178,
      "rewards/forensics_tiou_reward": 0.43884626030921936,
      "step": 244
    },
    {
      "completion_length": 20.3125,
      "epoch": 0.7526881720430108,
      "grad_norm": 7.471555233001709,
      "kl": 0.18359375,
      "learning_rate": 8.123076923076922e-07,
      "loss": 0.017,
      "reward": 1.0498390197753906,
      "reward_std": 0.2065093219280243,
      "reward_std_all": 0.22913139313459396,
      "rewards-max/forensics_sinkhorn_reward": 0.9426684975624084,
      "rewards-max/forensics_strict_edge_reward": 0.2227302473038435,
      "rewards-max/forensics_tiou_reward": 0.8824627995491028,
      "rewards/forensics_sinkhorn_reward": 0.662718802690506,
      "rewards/forensics_strict_edge_reward": 0.01827357872389257,
      "rewards/forensics_tiou_reward": 0.3688466101884842,
      "step": 245
    },
    {
      "completion_length": 20.78125,
      "epoch": 0.7557603686635944,
      "grad_norm": 10.874828338623047,
      "kl": 0.19091796875,
      "learning_rate": 8.115384615384615e-07,
      "loss": -0.0554,
      "reward": 1.123966097831726,
      "reward_std": 0.24807217717170715,
      "reward_std_all": 0.22235504537820816,
      "rewards-max/forensics_sinkhorn_reward": 0.9760829508304596,
      "rewards-max/forensics_strict_edge_reward": 0.3141753673553467,
      "rewards-max/forensics_tiou_reward": 0.9639665484428406,
      "rewards/forensics_sinkhorn_reward": 0.6930056214332581,
      "rewards/forensics_strict_edge_reward": 0.022137343883514404,
      "rewards/forensics_tiou_reward": 0.4088231474161148,
      "step": 246
    },
    {
      "completion_length": 21.71875,
      "epoch": 0.7588325652841782,
      "grad_norm": 8.423761367797852,
      "kl": 0.23828125,
      "learning_rate": 8.107692307692308e-07,
      "loss": -0.0005,
      "reward": 1.1131943464279175,
      "reward_std": 0.2351491004228592,
      "reward_std_all": 0.23894894123077393,
      "rewards-max/forensics_sinkhorn_reward": 0.9470195770263672,
      "rewards-max/forensics_strict_edge_reward": 0.07713586557656527,
      "rewards-max/forensics_tiou_reward": 0.8811467587947845,
      "rewards/forensics_sinkhorn_reward": 0.6555048525333405,
      "rewards/forensics_strict_edge_reward": 0.007081757532432675,
      "rewards/forensics_tiou_reward": 0.4506077319383621,
      "step": 247
    },
    {
      "completion_length": 23.078125,
      "epoch": 0.7619047619047619,
      "grad_norm": 5.669068813323975,
      "kl": 0.18798828125,
      "learning_rate": 8.1e-07,
      "loss": 0.0489,
      "reward": 1.119743674993515,
      "reward_std": 0.18600696325302124,
      "reward_std_all": 0.1687219962477684,
      "rewards-max/forensics_sinkhorn_reward": 0.9079897701740265,
      "rewards-max/forensics_strict_edge_reward": 0.13325238227844238,
      "rewards-max/forensics_tiou_reward": 0.8036381304264069,
      "rewards/forensics_sinkhorn_reward": 0.6741793155670166,
      "rewards/forensics_strict_edge_reward": 0.008707791334018111,
      "rewards/forensics_tiou_reward": 0.43685653805732727,
      "step": 248
    },
    {
      "completion_length": 22.078125,
      "epoch": 0.7649769585253456,
      "grad_norm": 8.74802017211914,
      "kl": 0.3994140625,
      "learning_rate": 8.092307692307692e-07,
      "loss": -0.0771,
      "reward": 1.1809147000312805,
      "reward_std": 0.21842342615127563,
      "reward_std_all": 0.19849778711795807,
      "rewards-max/forensics_sinkhorn_reward": 0.9219236969947815,
      "rewards-max/forensics_strict_edge_reward": 0.07786010578274727,
      "rewards-max/forensics_tiou_reward": 0.8355780243873596,
      "rewards/forensics_sinkhorn_reward": 0.7116504013538361,
      "rewards/forensics_strict_edge_reward": 0.005358283640816808,
      "rewards/forensics_tiou_reward": 0.46390610933303833,
      "step": 249
    },
    {
      "completion_length": 22.640625,
      "epoch": 0.7680491551459293,
      "grad_norm": 6.7117018699646,
      "kl": 0.19775390625,
      "learning_rate": 8.084615384615384e-07,
      "loss": -0.058,
      "reward": 1.1636712551116943,
      "reward_std": 0.13735001161694527,
      "reward_std_all": 0.14024442434310913,
      "rewards-max/forensics_sinkhorn_reward": 0.9211638271808624,
      "rewards-max/forensics_strict_edge_reward": 0.13194618560373783,
      "rewards-max/forensics_tiou_reward": 0.8383214771747589,
      "rewards/forensics_sinkhorn_reward": 0.7045428156852722,
      "rewards/forensics_strict_edge_reward": 0.015391683438792825,
      "rewards/forensics_tiou_reward": 0.4437367767095566,
      "step": 250
    },
    {
      "completion_length": 21.078125,
      "epoch": 0.7711213517665131,
      "grad_norm": 7.569887638092041,
      "kl": 0.22412109375,
      "learning_rate": 8.076923076923077e-07,
      "loss": -0.0432,
      "reward": 1.0299885869026184,
      "reward_std": 0.15845837444067,
      "reward_std_all": 0.1481829658150673,
      "rewards-max/forensics_sinkhorn_reward": 0.9276413023471832,
      "rewards-max/forensics_strict_edge_reward": 0.1138949766755104,
      "rewards-max/forensics_tiou_reward": 0.8495673835277557,
      "rewards/forensics_sinkhorn_reward": 0.6512826085090637,
      "rewards/forensics_strict_edge_reward": 0.004968828638084233,
      "rewards/forensics_tiou_reward": 0.3737372010946274,
      "step": 251
    },
    {
      "completion_length": 20.1875,
      "epoch": 0.7741935483870968,
      "grad_norm": 8.595390319824219,
      "kl": 0.35791015625,
      "learning_rate": 8.069230769230769e-07,
      "loss": -0.0358,
      "reward": 1.0315903425216675,
      "reward_std": 0.21560388058423996,
      "reward_std_all": 0.2540455386042595,
      "rewards-max/forensics_sinkhorn_reward": 0.9170425832271576,
      "rewards-max/forensics_strict_edge_reward": 0.042667726025683805,
      "rewards-max/forensics_tiou_reward": 0.8057222962379456,
      "rewards/forensics_sinkhorn_reward": 0.6583840548992157,
      "rewards/forensics_strict_edge_reward": 0.005138456450367812,
      "rewards/forensics_tiou_reward": 0.36806778609752655,
      "step": 252
    },
    {
      "completion_length": 22.375,
      "epoch": 0.7772657450076805,
      "grad_norm": 8.337164878845215,
      "kl": 0.208984375,
      "learning_rate": 8.06153846153846e-07,
      "loss": 0.0445,
      "reward": 1.1547868847846985,
      "reward_std": 0.2570101395249367,
      "reward_std_all": 0.23467287421226501,
      "rewards-max/forensics_sinkhorn_reward": 0.9428238570690155,
      "rewards-max/forensics_strict_edge_reward": 0.2731337174773216,
      "rewards-max/forensics_tiou_reward": 0.8831482231616974,
      "rewards/forensics_sinkhorn_reward": 0.701909601688385,
      "rewards/forensics_strict_edge_reward": 0.01508576306514442,
      "rewards/forensics_tiou_reward": 0.43779149651527405,
      "step": 253
    },
    {
      "completion_length": 20.171875,
      "epoch": 0.7803379416282642,
      "grad_norm": 7.898599147796631,
      "kl": 0.3671875,
      "learning_rate": 8.053846153846154e-07,
      "loss": 0.0943,
      "reward": 1.000393569469452,
      "reward_std": 0.1920340359210968,
      "reward_std_all": 0.16585512086749077,
      "rewards-max/forensics_sinkhorn_reward": 0.9788781702518463,
      "rewards-max/forensics_strict_edge_reward": 0.5422053933143616,
      "rewards-max/forensics_tiou_reward": 0.9712117612361908,
      "rewards/forensics_sinkhorn_reward": 0.6306719183921814,
      "rewards/forensics_strict_edge_reward": 0.03647589683532715,
      "rewards/forensics_tiou_reward": 0.3332457095384598,
      "step": 254
    },
    {
      "completion_length": 20.5,
      "epoch": 0.783410138248848,
      "grad_norm": 7.975688457489014,
      "kl": 0.220703125,
      "learning_rate": 8.046153846153846e-07,
      "loss": -0.096,
      "reward": 1.06838858127594,
      "reward_std": 0.17655367404222488,
      "reward_std_all": 0.16801048815250397,
      "rewards-max/forensics_sinkhorn_reward": 0.9089421629905701,
      "rewards-max/forensics_strict_edge_reward": 0.030044551996979862,
      "rewards-max/forensics_tiou_reward": 0.7986912727355957,
      "rewards/forensics_sinkhorn_reward": 0.6872978210449219,
      "rewards/forensics_strict_edge_reward": 0.002810266025335295,
      "rewards/forensics_tiou_reward": 0.37828047573566437,
      "step": 255
    },
    {
      "completion_length": 21.171875,
      "epoch": 0.7864823348694316,
      "grad_norm": 9.048943519592285,
      "kl": 0.62109375,
      "learning_rate": 8.038461538461538e-07,
      "loss": 0.0383,
      "reward": 1.0126630365848541,
      "reward_std": 0.18295373022556305,
      "reward_std_all": 0.1881985291838646,
      "rewards-max/forensics_sinkhorn_reward": 0.8965411484241486,
      "rewards-max/forensics_strict_edge_reward": 0.12515248451381922,
      "rewards-max/forensics_tiou_reward": 0.7726847231388092,
      "rewards/forensics_sinkhorn_reward": 0.6578738987445831,
      "rewards/forensics_strict_edge_reward": 0.010932270844932646,
      "rewards/forensics_tiou_reward": 0.3438568562269211,
      "step": 256
    },
    {
      "completion_length": 23.125,
      "epoch": 0.7895545314900153,
      "grad_norm": 10.995834350585938,
      "kl": 0.1640625,
      "learning_rate": 8.03076923076923e-07,
      "loss": -0.0075,
      "reward": 1.086348831653595,
      "reward_std": 0.27032676339149475,
      "reward_std_all": 0.2710036337375641,
      "rewards-max/forensics_sinkhorn_reward": 0.9771382510662079,
      "rewards-max/forensics_strict_edge_reward": 0.21355018205940723,
      "rewards-max/forensics_tiou_reward": 0.9698202908039093,
      "rewards/forensics_sinkhorn_reward": 0.6440598964691162,
      "rewards/forensics_strict_edge_reward": 0.009781864820979536,
      "rewards/forensics_tiou_reward": 0.43250706791877747,
      "step": 257
    },
    {
      "completion_length": 21.21875,
      "epoch": 0.7926267281105991,
      "grad_norm": 6.045054912567139,
      "kl": 0.1962890625,
      "learning_rate": 8.023076923076922e-07,
      "loss": 0.0667,
      "reward": 0.956709086894989,
      "reward_std": 0.3049141466617584,
      "reward_std_all": 0.2708599418401718,
      "rewards-max/forensics_sinkhorn_reward": 0.9331951439380646,
      "rewards-max/forensics_strict_edge_reward": 0.1926657184958458,
      "rewards-max/forensics_tiou_reward": 0.8616599142551422,
      "rewards/forensics_sinkhorn_reward": 0.5984058678150177,
      "rewards/forensics_strict_edge_reward": 0.012111779768019915,
      "rewards/forensics_tiou_reward": 0.3461914509534836,
      "step": 258
    },
    {
      "completion_length": 21.765625,
      "epoch": 0.7956989247311828,
      "grad_norm": 8.67448902130127,
      "kl": 0.22607421875,
      "learning_rate": 8.015384615384615e-07,
      "loss": -0.0547,
      "reward": 1.129256010055542,
      "reward_std": 0.24315479397773743,
      "reward_std_all": 0.2252410501241684,
      "rewards-max/forensics_sinkhorn_reward": 0.9703242480754852,
      "rewards-max/forensics_strict_edge_reward": 0.4589669406414032,
      "rewards-max/forensics_tiou_reward": 0.9466919898986816,
      "rewards/forensics_sinkhorn_reward": 0.6783452928066254,
      "rewards/forensics_strict_edge_reward": 0.027759303338825703,
      "rewards/forensics_tiou_reward": 0.4231513887643814,
      "step": 259
    },
    {
      "completion_length": 20.84375,
      "epoch": 0.7987711213517665,
      "grad_norm": 8.611160278320312,
      "kl": 0.19287109375,
      "learning_rate": 8.007692307692308e-07,
      "loss": -0.0452,
      "reward": 1.0767487287521362,
      "reward_std": 0.1412975825369358,
      "reward_std_all": 0.1270517036318779,
      "rewards-max/forensics_sinkhorn_reward": 0.9509114921092987,
      "rewards-max/forensics_strict_edge_reward": 0.22390545904636383,
      "rewards-max/forensics_tiou_reward": 0.8888471126556396,
      "rewards/forensics_sinkhorn_reward": 0.6569990515708923,
      "rewards/forensics_strict_edge_reward": 0.010334366699680686,
      "rewards/forensics_tiou_reward": 0.4094153344631195,
      "step": 260
    },
    {
      "completion_length": 21.203125,
      "epoch": 0.8018433179723502,
      "grad_norm": 7.5516839027404785,
      "kl": 0.16455078125,
      "learning_rate": 8e-07,
      "loss": 0.0151,
      "reward": 1.2493996620178223,
      "reward_std": 0.22929969429969788,
      "reward_std_all": 0.22193197906017303,
      "rewards-max/forensics_sinkhorn_reward": 0.9253762066364288,
      "rewards-max/forensics_strict_edge_reward": 0.2864522561430931,
      "rewards-max/forensics_tiou_reward": 0.8430386185646057,
      "rewards/forensics_sinkhorn_reward": 0.7333990335464478,
      "rewards/forensics_strict_edge_reward": 0.03016447275876999,
      "rewards/forensics_tiou_reward": 0.4858361631631851,
      "step": 261
    },
    {
      "completion_length": 19.40625,
      "epoch": 0.804915514592934,
      "grad_norm": 13.479561805725098,
      "kl": 0.4453125,
      "learning_rate": 7.992307692307692e-07,
      "loss": -0.0215,
      "reward": 1.1257745027542114,
      "reward_std": 0.24741669744253159,
      "reward_std_all": 0.23223596811294556,
      "rewards-max/forensics_sinkhorn_reward": 0.8767302334308624,
      "rewards-max/forensics_strict_edge_reward": 0.31055257469415665,
      "rewards-max/forensics_tiou_reward": 0.7690332233905792,
      "rewards/forensics_sinkhorn_reward": 0.6755674779415131,
      "rewards/forensics_strict_edge_reward": 0.0244410065934062,
      "rewards/forensics_tiou_reward": 0.42576606571674347,
      "step": 262
    },
    {
      "completion_length": 21.484375,
      "epoch": 0.8079877112135176,
      "grad_norm": 9.505818367004395,
      "kl": 0.36474609375,
      "learning_rate": 7.984615384615384e-07,
      "loss": -0.0584,
      "reward": 1.0452378988265991,
      "reward_std": 0.22248585522174835,
      "reward_std_all": 0.23958025872707367,
      "rewards-max/forensics_sinkhorn_reward": 0.957694411277771,
      "rewards-max/forensics_strict_edge_reward": 0.09379712492227554,
      "rewards-max/forensics_tiou_reward": 0.9095408916473389,
      "rewards/forensics_sinkhorn_reward": 0.6739877462387085,
      "rewards/forensics_strict_edge_reward": 0.004957212600857019,
      "rewards/forensics_tiou_reward": 0.36629293859004974,
      "step": 263
    },
    {
      "completion_length": 22.0625,
      "epoch": 0.8110599078341014,
      "grad_norm": 8.712757110595703,
      "kl": 0.1728515625,
      "learning_rate": 7.976923076923077e-07,
      "loss": -0.1118,
      "reward": 1.1986194252967834,
      "reward_std": 0.18851043283939362,
      "reward_std_all": 0.19795167446136475,
      "rewards-max/forensics_sinkhorn_reward": 0.9371321201324463,
      "rewards-max/forensics_strict_edge_reward": 0.30374059826135635,
      "rewards-max/forensics_tiou_reward": 0.8704949617385864,
      "rewards/forensics_sinkhorn_reward": 0.7161820232868195,
      "rewards/forensics_strict_edge_reward": 0.018012091517448425,
      "rewards/forensics_tiou_reward": 0.46442538499832153,
      "step": 264
    },
    {
      "completion_length": 21.859375,
      "epoch": 0.8141321044546851,
      "grad_norm": 5.830319404602051,
      "kl": 0.2890625,
      "learning_rate": 7.969230769230769e-07,
      "loss": -0.0185,
      "reward": 0.8654733598232269,
      "reward_std": 0.15069570764899254,
      "reward_std_all": 0.1455925554037094,
      "rewards-max/forensics_sinkhorn_reward": 0.873553991317749,
      "rewards-max/forensics_strict_edge_reward": 0.06248272955417633,
      "rewards-max/forensics_tiou_reward": 0.7125070691108704,
      "rewards/forensics_sinkhorn_reward": 0.5787367224693298,
      "rewards/forensics_strict_edge_reward": 0.002905165543779731,
      "rewards/forensics_tiou_reward": 0.283831462264061,
      "step": 265
    },
    {
      "completion_length": 21.0,
      "epoch": 0.8172043010752689,
      "grad_norm": 9.021906852722168,
      "kl": 0.2431640625,
      "learning_rate": 7.96153846153846e-07,
      "loss": 0.0312,
      "reward": 1.0124963521957397,
      "reward_std": 0.283282071352005,
      "reward_std_all": 0.2655939757823944,
      "rewards-max/forensics_sinkhorn_reward": 0.9755941927433014,
      "rewards-max/forensics_strict_edge_reward": 0.3922314904630184,
      "rewards-max/forensics_tiou_reward": 0.9548742771148682,
      "rewards/forensics_sinkhorn_reward": 0.6302514672279358,
      "rewards/forensics_strict_edge_reward": 0.02429731236770749,
      "rewards/forensics_tiou_reward": 0.35794752836227417,
      "step": 266
    },
    {
      "completion_length": 24.40625,
      "epoch": 0.8202764976958525,
      "grad_norm": 7.009968280792236,
      "kl": 0.21728515625,
      "learning_rate": 7.953846153846154e-07,
      "loss": 0.0097,
      "reward": 0.8930897116661072,
      "reward_std": 0.18390041589736938,
      "reward_std_all": 0.20305554568767548,
      "rewards-max/forensics_sinkhorn_reward": 0.8475892245769501,
      "rewards-max/forensics_strict_edge_reward": 0.02095708013607389,
      "rewards-max/forensics_tiou_reward": 0.7084470093250275,
      "rewards/forensics_sinkhorn_reward": 0.5847755074501038,
      "rewards/forensics_strict_edge_reward": 0.0013524435782422128,
      "rewards/forensics_tiou_reward": 0.3069617301225662,
      "step": 267
    },
    {
      "completion_length": 21.34375,
      "epoch": 0.8233486943164362,
      "grad_norm": 11.117053031921387,
      "kl": 0.24169921875,
      "learning_rate": 7.946153846153846e-07,
      "loss": 0.061,
      "reward": 0.9752101004123688,
      "reward_std": 0.2528003007173538,
      "reward_std_all": 0.2308136373758316,
      "rewards-max/forensics_sinkhorn_reward": 0.9333387017250061,
      "rewards-max/forensics_strict_edge_reward": 0.1990319082979113,
      "rewards-max/forensics_tiou_reward": 0.8560196757316589,
      "rewards/forensics_sinkhorn_reward": 0.6146343946456909,
      "rewards/forensics_strict_edge_reward": 0.01789797851233743,
      "rewards/forensics_tiou_reward": 0.3426777273416519,
      "step": 268
    },
    {
      "completion_length": 21.25,
      "epoch": 0.82642089093702,
      "grad_norm": 8.69359302520752,
      "kl": 0.1328125,
      "learning_rate": 7.938461538461538e-07,
      "loss": -0.0849,
      "reward": 1.194653034210205,
      "reward_std": 0.2176860347390175,
      "reward_std_all": 0.19188274443149567,
      "rewards-max/forensics_sinkhorn_reward": 0.9757356941699982,
      "rewards-max/forensics_strict_edge_reward": 0.42878545820713043,
      "rewards-max/forensics_tiou_reward": 0.964717835187912,
      "rewards/forensics_sinkhorn_reward": 0.7116150259971619,
      "rewards/forensics_strict_edge_reward": 0.03476725518703461,
      "rewards/forensics_tiou_reward": 0.4482707679271698,
      "step": 269
    },
    {
      "completion_length": 23.953125,
      "epoch": 0.8294930875576036,
      "grad_norm": 8.159682273864746,
      "kl": 0.30126953125,
      "learning_rate": 7.930769230769231e-07,
      "loss": 0.0639,
      "reward": 1.095299780368805,
      "reward_std": 0.2884667217731476,
      "reward_std_all": 0.25893476605415344,
      "rewards-max/forensics_sinkhorn_reward": 0.9612824618816376,
      "rewards-max/forensics_strict_edge_reward": 0.2499256581068039,
      "rewards-max/forensics_tiou_reward": 0.9330158233642578,
      "rewards/forensics_sinkhorn_reward": 0.6290009319782257,
      "rewards/forensics_strict_edge_reward": 0.02003592485561967,
      "rewards/forensics_tiou_reward": 0.4462628662586212,
      "step": 270
    },
    {
      "completion_length": 22.40625,
      "epoch": 0.8325652841781874,
      "grad_norm": 7.262181282043457,
      "kl": 0.21875,
      "learning_rate": 7.923076923076922e-07,
      "loss": -0.0536,
      "reward": 1.0545792877674103,
      "reward_std": 0.214489184319973,
      "reward_std_all": 0.20045123249292374,
      "rewards-max/forensics_sinkhorn_reward": 0.9608547389507294,
      "rewards-max/forensics_strict_edge_reward": 0.23560881987214088,
      "rewards-max/forensics_tiou_reward": 0.9092950522899628,
      "rewards/forensics_sinkhorn_reward": 0.6521562933921814,
      "rewards/forensics_strict_edge_reward": 0.017539491411298513,
      "rewards/forensics_tiou_reward": 0.38488343358039856,
      "step": 271
    },
    {
      "completion_length": 21.40625,
      "epoch": 0.8356374807987711,
      "grad_norm": 8.353940963745117,
      "kl": 0.1357421875,
      "learning_rate": 7.915384615384615e-07,
      "loss": -0.1662,
      "reward": 1.2470248937606812,
      "reward_std": 0.21792984008789062,
      "reward_std_all": 0.23759403824806213,
      "rewards-max/forensics_sinkhorn_reward": 0.9760487973690033,
      "rewards-max/forensics_strict_edge_reward": 0.24188234575558454,
      "rewards-max/forensics_tiou_reward": 0.959555983543396,
      "rewards/forensics_sinkhorn_reward": 0.7244083285331726,
      "rewards/forensics_strict_edge_reward": 0.025912476201483514,
      "rewards/forensics_tiou_reward": 0.4967040866613388,
      "step": 272
    },
    {
      "completion_length": 23.078125,
      "epoch": 0.8387096774193549,
      "grad_norm": 9.962955474853516,
      "kl": 0.18359375,
      "learning_rate": 7.907692307692307e-07,
      "loss": 0.0484,
      "reward": 1.0466077029705048,
      "reward_std": 0.19775225967168808,
      "reward_std_all": 0.1690468043088913,
      "rewards-max/forensics_sinkhorn_reward": 0.9198048412799835,
      "rewards-max/forensics_strict_edge_reward": 0.19769802689552307,
      "rewards-max/forensics_tiou_reward": 0.825882077217102,
      "rewards/forensics_sinkhorn_reward": 0.6434611678123474,
      "rewards/forensics_strict_edge_reward": 0.011299411300569773,
      "rewards/forensics_tiou_reward": 0.39184707403182983,
      "step": 273
    },
    {
      "completion_length": 24.96875,
      "epoch": 0.8417818740399385,
      "grad_norm": 8.714496612548828,
      "kl": 0.236328125,
      "learning_rate": 7.9e-07,
      "loss": -0.084,
      "reward": 1.059462547302246,
      "reward_std": 0.22326727211475372,
      "reward_std_all": 0.19384222477674484,
      "rewards-max/forensics_sinkhorn_reward": 0.9396823644638062,
      "rewards-max/forensics_strict_edge_reward": 0.17030530516058207,
      "rewards-max/forensics_tiou_reward": 0.8640748858451843,
      "rewards/forensics_sinkhorn_reward": 0.6297011077404022,
      "rewards/forensics_strict_edge_reward": 0.007734586368314922,
      "rewards/forensics_tiou_reward": 0.42202678322792053,
      "step": 274
    },
    {
      "completion_length": 21.609375,
      "epoch": 0.8448540706605223,
      "grad_norm": 9.25124740600586,
      "kl": 0.183349609375,
      "learning_rate": 7.892307692307692e-07,
      "loss": 0.0466,
      "reward": 1.2324098944664001,
      "reward_std": 0.16770468652248383,
      "reward_std_all": 0.16140137612819672,
      "rewards-max/forensics_sinkhorn_reward": 0.9430018663406372,
      "rewards-max/forensics_strict_edge_reward": 0.265866382047534,
      "rewards-max/forensics_tiou_reward": 0.8857716023921967,
      "rewards/forensics_sinkhorn_reward": 0.7227489948272705,
      "rewards/forensics_strict_edge_reward": 0.023506090976297855,
      "rewards/forensics_tiou_reward": 0.48615485429763794,
      "step": 275
    },
    {
      "completion_length": 22.25,
      "epoch": 0.847926267281106,
      "grad_norm": 9.261551856994629,
      "kl": 0.2685546875,
      "learning_rate": 7.884615384615384e-07,
      "loss": -0.0328,
      "reward": 1.1648313999176025,
      "reward_std": 0.27922891825437546,
      "reward_std_all": 0.26058731228113174,
      "rewards-max/forensics_sinkhorn_reward": 0.9017457067966461,
      "rewards-max/forensics_strict_edge_reward": 0.26141034346073866,
      "rewards-max/forensics_tiou_reward": 0.8023904860019684,
      "rewards/forensics_sinkhorn_reward": 0.6960657238960266,
      "rewards/forensics_strict_edge_reward": 0.02851615205872804,
      "rewards/forensics_tiou_reward": 0.44024956226348877,
      "step": 276
    },
    {
      "completion_length": 22.5625,
      "epoch": 0.8509984639016898,
      "grad_norm": 6.961018085479736,
      "kl": 0.2509765625,
      "learning_rate": 7.876923076923077e-07,
      "loss": -0.0114,
      "reward": 1.1245884597301483,
      "reward_std": 0.27486132085323334,
      "reward_std_all": 0.23772358894348145,
      "rewards-max/forensics_sinkhorn_reward": 0.9606059491634369,
      "rewards-max/forensics_strict_edge_reward": 0.5062990635633469,
      "rewards-max/forensics_tiou_reward": 0.9296295642852783,
      "rewards/forensics_sinkhorn_reward": 0.6910044252872467,
      "rewards/forensics_strict_edge_reward": 0.025673137977719307,
      "rewards/forensics_tiou_reward": 0.40791092813014984,
      "step": 277
    },
    {
      "completion_length": 23.125,
      "epoch": 0.8540706605222734,
      "grad_norm": 9.217570304870605,
      "kl": 0.14990234375,
      "learning_rate": 7.869230769230769e-07,
      "loss": -0.0328,
      "reward": 0.998620867729187,
      "reward_std": 0.24737988412380219,
      "reward_std_all": 0.2257489264011383,
      "rewards-max/forensics_sinkhorn_reward": 0.9224681258201599,
      "rewards-max/forensics_strict_edge_reward": 0.28705035150051117,
      "rewards-max/forensics_tiou_reward": 0.8132308423519135,
      "rewards/forensics_sinkhorn_reward": 0.6180504262447357,
      "rewards/forensics_strict_edge_reward": 0.018837206065654755,
      "rewards/forensics_tiou_reward": 0.36173322796821594,
      "step": 278
    },
    {
      "completion_length": 23.421875,
      "epoch": 0.8571428571428571,
      "grad_norm": 7.942857265472412,
      "kl": 0.171875,
      "learning_rate": 7.861538461538461e-07,
      "loss": -0.0149,
      "reward": 1.0511631965637207,
      "reward_std": 0.3272942453622818,
      "reward_std_all": 0.3002912402153015,
      "rewards-max/forensics_sinkhorn_reward": 0.9464280605316162,
      "rewards-max/forensics_strict_edge_reward": 0.08596969489008188,
      "rewards-max/forensics_tiou_reward": 0.8941896557807922,
      "rewards/forensics_sinkhorn_reward": 0.626489907503128,
      "rewards/forensics_strict_edge_reward": 0.006694726413115859,
      "rewards/forensics_tiou_reward": 0.41797854006290436,
      "step": 279
    },
    {
      "completion_length": 24.40625,
      "epoch": 0.8602150537634409,
      "grad_norm": 5.856282711029053,
      "kl": 0.16455078125,
      "learning_rate": 7.853846153846153e-07,
      "loss": -0.0352,
      "reward": 1.2568358182907104,
      "reward_std": 0.2568046450614929,
      "reward_std_all": 0.23419124633073807,
      "rewards-max/forensics_sinkhorn_reward": 0.942993551492691,
      "rewards-max/forensics_strict_edge_reward": 0.3419768214225769,
      "rewards-max/forensics_tiou_reward": 0.8731246888637543,
      "rewards/forensics_sinkhorn_reward": 0.7099796235561371,
      "rewards/forensics_strict_edge_reward": 0.04015221260488033,
      "rewards/forensics_tiou_reward": 0.5067039132118225,
      "step": 280
    },
    {
      "completion_length": 20.765625,
      "epoch": 0.8632872503840245,
      "grad_norm": 9.614285469055176,
      "kl": 0.15771484375,
      "learning_rate": 7.846153846153846e-07,
      "loss": -0.0005,
      "reward": 1.141631305217743,
      "reward_std": 0.14855030924081802,
      "reward_std_all": 0.1638011485338211,
      "rewards-max/forensics_sinkhorn_reward": 0.963556170463562,
      "rewards-max/forensics_strict_edge_reward": 0.3144784942269325,
      "rewards-max/forensics_tiou_reward": 0.9390410780906677,
      "rewards/forensics_sinkhorn_reward": 0.6797847747802734,
      "rewards/forensics_strict_edge_reward": 0.015372799709439278,
      "rewards/forensics_tiou_reward": 0.44647374749183655,
      "step": 281
    },
    {
      "completion_length": 21.5625,
      "epoch": 0.8663594470046083,
      "grad_norm": 6.654672622680664,
      "kl": 0.291015625,
      "learning_rate": 7.838461538461539e-07,
      "loss": -0.0121,
      "reward": 1.0068949162960052,
      "reward_std": 0.23201756924390793,
      "reward_std_all": 0.2085975706577301,
      "rewards-max/forensics_sinkhorn_reward": 0.9269982874393463,
      "rewards-max/forensics_strict_edge_reward": 0.3325132876634598,
      "rewards-max/forensics_tiou_reward": 0.8352627158164978,
      "rewards/forensics_sinkhorn_reward": 0.6093716323375702,
      "rewards/forensics_strict_edge_reward": 0.02675968548282981,
      "rewards/forensics_tiou_reward": 0.3707636296749115,
      "step": 282
    },
    {
      "completion_length": 23.796875,
      "epoch": 0.869431643625192,
      "grad_norm": 11.745622634887695,
      "kl": 0.20556640625,
      "learning_rate": 7.830769230769231e-07,
      "loss": 0.037,
      "reward": 1.036502480506897,
      "reward_std": 0.2656548023223877,
      "reward_std_all": 0.24562812596559525,
      "rewards-max/forensics_sinkhorn_reward": 0.9459617137908936,
      "rewards-max/forensics_strict_edge_reward": 0.039443506859242916,
      "rewards-max/forensics_tiou_reward": 0.8885388970375061,
      "rewards/forensics_sinkhorn_reward": 0.6332782804965973,
      "rewards/forensics_strict_edge_reward": 0.0021664741216227412,
      "rewards/forensics_tiou_reward": 0.401057705283165,
      "step": 283
    },
    {
      "completion_length": 20.859375,
      "epoch": 0.8725038402457758,
      "grad_norm": 10.716646194458008,
      "kl": 0.164794921875,
      "learning_rate": 7.823076923076922e-07,
      "loss": 0.1419,
      "reward": 1.2124063968658447,
      "reward_std": 0.24728277325630188,
      "reward_std_all": 0.26206544041633606,
      "rewards-max/forensics_sinkhorn_reward": 0.9615591764450073,
      "rewards-max/forensics_strict_edge_reward": 0.3353709802031517,
      "rewards-max/forensics_tiou_reward": 0.9291525185108185,
      "rewards/forensics_sinkhorn_reward": 0.734668493270874,
      "rewards/forensics_strict_edge_reward": 0.017499472480267286,
      "rewards/forensics_tiou_reward": 0.46023842692375183,
      "step": 284
    },
    {
      "completion_length": 23.3125,
      "epoch": 0.8755760368663594,
      "grad_norm": 8.009867668151855,
      "kl": 0.19921875,
      "learning_rate": 7.815384615384615e-07,
      "loss": 0.0027,
      "reward": 1.0994118452072144,
      "reward_std": 0.30738620460033417,
      "reward_std_all": 0.2817564010620117,
      "rewards-max/forensics_sinkhorn_reward": 0.9389869570732117,
      "rewards-max/forensics_strict_edge_reward": 0.3677594065666199,
      "rewards-max/forensics_tiou_reward": 0.8746560513973236,
      "rewards/forensics_sinkhorn_reward": 0.6686483323574066,
      "rewards/forensics_strict_edge_reward": 0.030745147727429867,
      "rewards/forensics_tiou_reward": 0.4000183790922165,
      "step": 285
    },
    {
      "completion_length": 21.8125,
      "epoch": 0.8786482334869432,
      "grad_norm": 10.442387580871582,
      "kl": 0.17333984375,
      "learning_rate": 7.807692307692307e-07,
      "loss": -0.0227,
      "reward": 1.0036545991897583,
      "reward_std": 0.254051610827446,
      "reward_std_all": 0.22527916729450226,
      "rewards-max/forensics_sinkhorn_reward": 0.9654397368431091,
      "rewards-max/forensics_strict_edge_reward": 0.2242445945739746,
      "rewards-max/forensics_tiou_reward": 0.9326504170894623,
      "rewards/forensics_sinkhorn_reward": 0.6208761930465698,
      "rewards/forensics_strict_edge_reward": 0.03033863683231175,
      "rewards/forensics_tiou_reward": 0.3524397611618042,
      "step": 286
    },
    {
      "completion_length": 20.296875,
      "epoch": 0.8817204301075269,
      "grad_norm": 6.127209186553955,
      "kl": 0.23046875,
      "learning_rate": 7.799999999999999e-07,
      "loss": 0.0361,
      "reward": 1.0536051988601685,
      "reward_std": 0.2515765577554703,
      "reward_std_all": 0.22444648295640945,
      "rewards-max/forensics_sinkhorn_reward": 0.9511329829692841,
      "rewards-max/forensics_strict_edge_reward": 0.10239118337631226,
      "rewards-max/forensics_tiou_reward": 0.8866287469863892,
      "rewards/forensics_sinkhorn_reward": 0.6416107714176178,
      "rewards/forensics_strict_edge_reward": 0.011062588077038527,
      "rewards/forensics_tiou_reward": 0.40093185007572174,
      "step": 287
    },
    {
      "completion_length": 21.609375,
      "epoch": 0.8847926267281107,
      "grad_norm": 8.636290550231934,
      "kl": 0.22802734375,
      "learning_rate": 7.792307692307693e-07,
      "loss": 0.0987,
      "reward": 0.9463670253753662,
      "reward_std": 0.17025425285100937,
      "reward_std_all": 0.15874582529067993,
      "rewards-max/forensics_sinkhorn_reward": 0.8896462619304657,
      "rewards-max/forensics_strict_edge_reward": 0.19709588214755058,
      "rewards-max/forensics_tiou_reward": 0.7651137709617615,
      "rewards/forensics_sinkhorn_reward": 0.6127243041992188,
      "rewards/forensics_strict_edge_reward": 0.01121049770154059,
      "rewards/forensics_tiou_reward": 0.3224322646856308,
      "step": 288
    },
    {
      "completion_length": 20.875,
      "epoch": 0.8878648233486943,
      "grad_norm": 9.79089069366455,
      "kl": 0.228515625,
      "learning_rate": 7.784615384615384e-07,
      "loss": -0.0278,
      "reward": 1.088763415813446,
      "reward_std": 0.3018733114004135,
      "reward_std_all": 0.27598198503255844,
      "rewards-max/forensics_sinkhorn_reward": 0.9553775787353516,
      "rewards-max/forensics_strict_edge_reward": 0.2471451535820961,
      "rewards-max/forensics_tiou_reward": 0.9092212021350861,
      "rewards/forensics_sinkhorn_reward": 0.6513403952121735,
      "rewards/forensics_strict_edge_reward": 0.019325765781104565,
      "rewards/forensics_tiou_reward": 0.41809725761413574,
      "step": 289
    },
    {
      "completion_length": 26.328125,
      "epoch": 0.890937019969278,
      "grad_norm": 7.504724025726318,
      "kl": 0.255859375,
      "learning_rate": 7.776923076923077e-07,
      "loss": 0.0062,
      "reward": 1.1063460111618042,
      "reward_std": 0.15676283836364746,
      "reward_std_all": 0.17966244369745255,
      "rewards-max/forensics_sinkhorn_reward": 0.9487336874008179,
      "rewards-max/forensics_strict_edge_reward": 0.09275638498365879,
      "rewards-max/forensics_tiou_reward": 0.9033868908882141,
      "rewards/forensics_sinkhorn_reward": 0.6601445376873016,
      "rewards/forensics_strict_edge_reward": 0.006241606781259179,
      "rewards/forensics_tiou_reward": 0.43995991349220276,
      "step": 290
    },
    {
      "completion_length": 20.328125,
      "epoch": 0.8940092165898618,
      "grad_norm": 10.667698860168457,
      "kl": 0.19921875,
      "learning_rate": 7.769230769230769e-07,
      "loss": 0.0016,
      "reward": 1.1213610172271729,
      "reward_std": 0.1321830227971077,
      "reward_std_all": 0.11716289818286896,
      "rewards-max/forensics_sinkhorn_reward": 0.9534969627857208,
      "rewards-max/forensics_strict_edge_reward": 0.011872438015416265,
      "rewards-max/forensics_tiou_reward": 0.8928779363632202,
      "rewards/forensics_sinkhorn_reward": 0.6789708733558655,
      "rewards/forensics_strict_edge_reward": 0.0008110814669635147,
      "rewards/forensics_tiou_reward": 0.44157904386520386,
      "step": 291
    },
    {
      "completion_length": 20.265625,
      "epoch": 0.8970814132104454,
      "grad_norm": 9.662022590637207,
      "kl": 0.171875,
      "learning_rate": 7.761538461538461e-07,
      "loss": -0.0763,
      "reward": 0.9715360403060913,
      "reward_std": 0.2518957704305649,
      "reward_std_all": 0.2229408696293831,
      "rewards-max/forensics_sinkhorn_reward": 0.9668095707893372,
      "rewards-max/forensics_strict_edge_reward": 0.3800070434808731,
      "rewards-max/forensics_tiou_reward": 0.9446820318698883,
      "rewards/forensics_sinkhorn_reward": 0.6223989427089691,
      "rewards/forensics_strict_edge_reward": 0.01822528149932623,
      "rewards/forensics_tiou_reward": 0.33091185986995697,
      "step": 292
    },
    {
      "completion_length": 20.890625,
      "epoch": 0.9001536098310292,
      "grad_norm": 7.861318111419678,
      "kl": 0.1953125,
      "learning_rate": 7.753846153846153e-07,
      "loss": -0.011,
      "reward": 1.2151581048965454,
      "reward_std": 0.23010414093732834,
      "reward_std_all": 0.20813338458538055,
      "rewards-max/forensics_sinkhorn_reward": 0.932733803987503,
      "rewards-max/forensics_strict_edge_reward": 0.21621999144554138,
      "rewards-max/forensics_tiou_reward": 0.831041693687439,
      "rewards/forensics_sinkhorn_reward": 0.7258618175983429,
      "rewards/forensics_strict_edge_reward": 0.015608087182044983,
      "rewards/forensics_tiou_reward": 0.47368811070919037,
      "step": 293
    },
    {
      "completion_length": 22.875,
      "epoch": 0.9032258064516129,
      "grad_norm": 7.925800800323486,
      "kl": 0.16943359375,
      "learning_rate": 7.746153846153845e-07,
      "loss": 0.0799,
      "reward": 0.9523175358772278,
      "reward_std": 0.19476701319217682,
      "reward_std_all": 0.19455213099718094,
      "rewards-max/forensics_sinkhorn_reward": 0.9463759362697601,
      "rewards-max/forensics_strict_edge_reward": 0.17374406196177006,
      "rewards-max/forensics_tiou_reward": 0.901190996170044,
      "rewards/forensics_sinkhorn_reward": 0.613290399312973,
      "rewards/forensics_strict_edge_reward": 0.008910464763175696,
      "rewards/forensics_tiou_reward": 0.3301166892051697,
      "step": 294
    },
    {
      "completion_length": 20.03125,
      "epoch": 0.9062980030721967,
      "grad_norm": 10.257216453552246,
      "kl": 0.158203125,
      "learning_rate": 7.738461538461539e-07,
      "loss": -0.0208,
      "reward": 1.1223703622817993,
      "reward_std": 0.17966046929359436,
      "reward_std_all": 0.2068917378783226,
      "rewards-max/forensics_sinkhorn_reward": 0.9619533717632294,
      "rewards-max/forensics_strict_edge_reward": 0.2010706141591072,
      "rewards-max/forensics_tiou_reward": 0.904924064874649,
      "rewards/forensics_sinkhorn_reward": 0.7058207094669342,
      "rewards/forensics_strict_edge_reward": 0.010248889680951834,
      "rewards/forensics_tiou_reward": 0.40630075335502625,
      "step": 295
    },
    {
      "completion_length": 20.546875,
      "epoch": 0.9093701996927803,
      "grad_norm": 8.531110763549805,
      "kl": 0.1875,
      "learning_rate": 7.730769230769231e-07,
      "loss": -0.1252,
      "reward": 1.1611728072166443,
      "reward_std": 0.24326062202453613,
      "reward_std_all": 0.23146961629390717,
      "rewards-max/forensics_sinkhorn_reward": 0.9711572229862213,
      "rewards-max/forensics_strict_edge_reward": 0.24787070602178574,
      "rewards-max/forensics_tiou_reward": 0.9433547854423523,
      "rewards/forensics_sinkhorn_reward": 0.6843513250350952,
      "rewards/forensics_strict_edge_reward": 0.034190508304163814,
      "rewards/forensics_tiou_reward": 0.4426310360431671,
      "step": 296
    },
    {
      "completion_length": 20.96875,
      "epoch": 0.9124423963133641,
      "grad_norm": 12.713372230529785,
      "kl": 0.24169921875,
      "learning_rate": 7.723076923076923e-07,
      "loss": -0.0523,
      "reward": 1.1701282858848572,
      "reward_std": 0.2889661490917206,
      "reward_std_all": 0.2588914632797241,
      "rewards-max/forensics_sinkhorn_reward": 0.9590179324150085,
      "rewards-max/forensics_strict_edge_reward": 0.25439786771312356,
      "rewards-max/forensics_tiou_reward": 0.923557460308075,
      "rewards/forensics_sinkhorn_reward": 0.7004460692405701,
      "rewards/forensics_strict_edge_reward": 0.02553189219906926,
      "rewards/forensics_tiou_reward": 0.4441503584384918,
      "step": 297
    },
    {
      "completion_length": 22.28125,
      "epoch": 0.9155145929339478,
      "grad_norm": 11.354487419128418,
      "kl": 0.1865234375,
      "learning_rate": 7.715384615384615e-07,
      "loss": 0.0031,
      "reward": 1.0751924216747284,
      "reward_std": 0.2500069737434387,
      "reward_std_all": 0.2135384902358055,
      "rewards-max/forensics_sinkhorn_reward": 0.9394853711128235,
      "rewards-max/forensics_strict_edge_reward": 0.2596282288432121,
      "rewards-max/forensics_tiou_reward": 0.8668342232704163,
      "rewards/forensics_sinkhorn_reward": 0.6794303357601166,
      "rewards/forensics_strict_edge_reward": 0.018714554142206907,
      "rewards/forensics_tiou_reward": 0.3770475536584854,
      "step": 298
    },
    {
      "completion_length": 20.75,
      "epoch": 0.9185867895545314,
      "grad_norm": 6.5368757247924805,
      "kl": 0.150390625,
      "learning_rate": 7.707692307692307e-07,
      "loss": 0.0432,
      "reward": 1.1695102453231812,
      "reward_std": 0.19342417269945145,
      "reward_std_all": 0.19763827323913574,
      "rewards-max/forensics_sinkhorn_reward": 0.9675437211990356,
      "rewards-max/forensics_strict_edge_reward": 0.17621293663978577,
      "rewards-max/forensics_tiou_reward": 0.9292292594909668,
      "rewards/forensics_sinkhorn_reward": 0.7037501335144043,
      "rewards/forensics_strict_edge_reward": 0.012909685261547565,
      "rewards/forensics_tiou_reward": 0.4528503715991974,
      "step": 299
    },
    {
      "completion_length": 19.796875,
      "epoch": 0.9216589861751152,
      "grad_norm": 14.425507545471191,
      "kl": 0.6220703125,
      "learning_rate": 7.699999999999999e-07,
      "loss": 0.1364,
      "reward": 1.0923432111740112,
      "reward_std": 0.23132477700710297,
      "reward_std_all": 0.21976502984762192,
      "rewards-max/forensics_sinkhorn_reward": 0.9631685614585876,
      "rewards-max/forensics_strict_edge_reward": 0.3169572949409485,
      "rewards-max/forensics_tiou_reward": 0.9340332746505737,
      "rewards/forensics_sinkhorn_reward": 0.6707879602909088,
      "rewards/forensics_strict_edge_reward": 0.020175286568701267,
      "rewards/forensics_tiou_reward": 0.40137992799282074,
      "step": 300
    },
    {
      "completion_length": 19.953125,
      "epoch": 0.9247311827956989,
      "grad_norm": 6.43282413482666,
      "kl": 0.27392578125,
      "learning_rate": 7.692307692307693e-07,
      "loss": 0.0338,
      "reward": 1.1630443930625916,
      "reward_std": 0.15776659548282623,
      "reward_std_all": 0.14593090116977692,
      "rewards-max/forensics_sinkhorn_reward": 0.922261655330658,
      "rewards-max/forensics_strict_edge_reward": 0.05623254366219044,
      "rewards-max/forensics_tiou_reward": 0.8143399059772491,
      "rewards/forensics_sinkhorn_reward": 0.7088655531406403,
      "rewards/forensics_strict_edge_reward": 0.005034274887293577,
      "rewards/forensics_tiou_reward": 0.449144572019577,
      "step": 301
    },
    {
      "completion_length": 19.4375,
      "epoch": 0.9278033794162827,
      "grad_norm": 9.847257614135742,
      "kl": 0.291015625,
      "learning_rate": 7.684615384615384e-07,
      "loss": 0.0138,
      "reward": 1.1467739343643188,
      "reward_std": 0.1981208398938179,
      "reward_std_all": 0.20029912889003754,
      "rewards-max/forensics_sinkhorn_reward": 0.9728361666202545,
      "rewards-max/forensics_strict_edge_reward": 0.37697096541523933,
      "rewards-max/forensics_tiou_reward": 0.9478849172592163,
      "rewards/forensics_sinkhorn_reward": 0.6812898814678192,
      "rewards/forensics_strict_edge_reward": 0.0411463042255491,
      "rewards/forensics_tiou_reward": 0.42433783411979675,
      "step": 302
    },
    {
      "completion_length": 19.90625,
      "epoch": 0.9308755760368663,
      "grad_norm": 6.002472877502441,
      "kl": 0.20947265625,
      "learning_rate": 7.676923076923077e-07,
      "loss": 0.0201,
      "reward": 1.149944543838501,
      "reward_std": 0.1735910102725029,
      "reward_std_all": 0.16823212057352066,
      "rewards-max/forensics_sinkhorn_reward": 0.9496829211711884,
      "rewards-max/forensics_strict_edge_reward": 0.029042456299066544,
      "rewards-max/forensics_tiou_reward": 0.8800031244754791,
      "rewards/forensics_sinkhorn_reward": 0.7185300588607788,
      "rewards/forensics_strict_edge_reward": 0.001438126782886684,
      "rewards/forensics_tiou_reward": 0.42997635900974274,
      "step": 303
    },
    {
      "completion_length": 19.34375,
      "epoch": 0.9339477726574501,
      "grad_norm": 11.035322189331055,
      "kl": 0.17578125,
      "learning_rate": 7.669230769230769e-07,
      "loss": -0.0392,
      "reward": 1.2212949395179749,
      "reward_std": 0.22159407287836075,
      "reward_std_all": 0.20590712875127792,
      "rewards-max/forensics_sinkhorn_reward": 0.9819167852401733,
      "rewards-max/forensics_strict_edge_reward": 0.38782531255856156,
      "rewards-max/forensics_tiou_reward": 0.9798087179660797,
      "rewards/forensics_sinkhorn_reward": 0.7262018322944641,
      "rewards/forensics_strict_edge_reward": 0.02200493379496038,
      "rewards/forensics_tiou_reward": 0.47308817505836487,
      "step": 304
    },
    {
      "completion_length": 20.578125,
      "epoch": 0.9370199692780338,
      "grad_norm": 10.086763381958008,
      "kl": 0.18603515625,
      "learning_rate": 7.661538461538461e-07,
      "loss": -0.0836,
      "reward": 1.0581908822059631,
      "reward_std": 0.2616780623793602,
      "reward_std_all": 0.2380274087190628,
      "rewards-max/forensics_sinkhorn_reward": 0.9389192759990692,
      "rewards-max/forensics_strict_edge_reward": 0.3254753649234772,
      "rewards-max/forensics_tiou_reward": 0.8618314862251282,
      "rewards/forensics_sinkhorn_reward": 0.641083151102066,
      "rewards/forensics_strict_edge_reward": 0.030387612991034985,
      "rewards/forensics_tiou_reward": 0.38672007620334625,
      "step": 305
    },
    {
      "completion_length": 20.328125,
      "epoch": 0.9400921658986175,
      "grad_norm": 7.986942291259766,
      "kl": 0.16455078125,
      "learning_rate": 7.653846153846153e-07,
      "loss": -0.0532,
      "reward": 1.2146449089050293,
      "reward_std": 0.20184414088726044,
      "reward_std_all": 0.18810324370861053,
      "rewards-max/forensics_sinkhorn_reward": 0.9432135224342346,
      "rewards-max/forensics_strict_edge_reward": 0.33328156918287277,
      "rewards-max/forensics_tiou_reward": 0.8761173784732819,
      "rewards/forensics_sinkhorn_reward": 0.7278759181499481,
      "rewards/forensics_strict_edge_reward": 0.019587131217122078,
      "rewards/forensics_tiou_reward": 0.46718187630176544,
      "step": 306
    },
    {
      "completion_length": 19.1875,
      "epoch": 0.9431643625192012,
      "grad_norm": 9.404058456420898,
      "kl": 0.169189453125,
      "learning_rate": 7.646153846153845e-07,
      "loss": -0.0043,
      "reward": 1.1684364676475525,
      "reward_std": 0.1592409461736679,
      "reward_std_all": 0.1616360917687416,
      "rewards-max/forensics_sinkhorn_reward": 0.9026331603527069,
      "rewards-max/forensics_strict_edge_reward": 0.1734738820232451,
      "rewards-max/forensics_tiou_reward": 0.7937654852867126,
      "rewards/forensics_sinkhorn_reward": 0.7216529548168182,
      "rewards/forensics_strict_edge_reward": 0.009483859117608517,
      "rewards/forensics_tiou_reward": 0.43729962408542633,
      "step": 307
    },
    {
      "completion_length": 20.3125,
      "epoch": 0.946236559139785,
      "grad_norm": 11.763525009155273,
      "kl": 0.1884765625,
      "learning_rate": 7.638461538461539e-07,
      "loss": -0.0643,
      "reward": 0.993390291929245,
      "reward_std": 0.14332862198352814,
      "reward_std_all": 0.18434254080057144,
      "rewards-max/forensics_sinkhorn_reward": 0.9420949518680573,
      "rewards-max/forensics_strict_edge_reward": 0.2555304039269686,
      "rewards-max/forensics_tiou_reward": 0.8563916981220245,
      "rewards/forensics_sinkhorn_reward": 0.6227614283561707,
      "rewards/forensics_strict_edge_reward": 0.013177052256651223,
      "rewards/forensics_tiou_reward": 0.35745178163051605,
      "step": 308
    },
    {
      "completion_length": 19.9375,
      "epoch": 0.9493087557603687,
      "grad_norm": 13.506754875183105,
      "kl": 1.06298828125,
      "learning_rate": 7.630769230769231e-07,
      "loss": 0.0307,
      "reward": 1.0189821422100067,
      "reward_std": 0.15989738702774048,
      "reward_std_all": 0.1614900901913643,
      "rewards-max/forensics_sinkhorn_reward": 0.8945117294788361,
      "rewards-max/forensics_strict_edge_reward": 0.0594215951859951,
      "rewards-max/forensics_tiou_reward": 0.8008507192134857,
      "rewards/forensics_sinkhorn_reward": 0.6413615047931671,
      "rewards/forensics_strict_edge_reward": 0.0033578422153368592,
      "rewards/forensics_tiou_reward": 0.3742627501487732,
      "step": 309
    },
    {
      "completion_length": 19.9375,
      "epoch": 0.9523809523809523,
      "grad_norm": 7.4458465576171875,
      "kl": 0.28857421875,
      "learning_rate": 7.623076923076923e-07,
      "loss": -0.0181,
      "reward": 1.194675326347351,
      "reward_std": 0.25965800136327744,
      "reward_std_all": 0.2212381213903427,
      "rewards-max/forensics_sinkhorn_reward": 0.9571115970611572,
      "rewards-max/forensics_strict_edge_reward": 0.5850475430488586,
      "rewards-max/forensics_tiou_reward": 0.9123424887657166,
      "rewards/forensics_sinkhorn_reward": 0.6979973018169403,
      "rewards/forensics_strict_edge_reward": 0.03904152289032936,
      "rewards/forensics_tiou_reward": 0.4576365053653717,
      "step": 310
    },
    {
      "completion_length": 20.109375,
      "epoch": 0.9554531490015361,
      "grad_norm": 6.076689720153809,
      "kl": 0.2978515625,
      "learning_rate": 7.615384615384615e-07,
      "loss": -0.0081,
      "reward": 0.979576587677002,
      "reward_std": 0.16476890444755554,
      "reward_std_all": 0.14595654979348183,
      "rewards-max/forensics_sinkhorn_reward": 0.8696512281894684,
      "rewards-max/forensics_strict_edge_reward": 0.14292298909276724,
      "rewards-max/forensics_tiou_reward": 0.7562919557094574,
      "rewards/forensics_sinkhorn_reward": 0.641179233789444,
      "rewards/forensics_strict_edge_reward": 0.007047370309010148,
      "rewards/forensics_tiou_reward": 0.3313499987125397,
      "step": 311
    },
    {
      "completion_length": 20.328125,
      "epoch": 0.9585253456221198,
      "grad_norm": 6.978910446166992,
      "kl": 0.3486328125,
      "learning_rate": 7.607692307692307e-07,
      "loss": 0.0253,
      "reward": 1.1063141822814941,
      "reward_std": 0.13701127097010612,
      "reward_std_all": 0.14682377502322197,
      "rewards-max/forensics_sinkhorn_reward": 0.9149279594421387,
      "rewards-max/forensics_strict_edge_reward": 0.09397092321887612,
      "rewards-max/forensics_tiou_reward": 0.7809372842311859,
      "rewards/forensics_sinkhorn_reward": 0.6849184632301331,
      "rewards/forensics_strict_edge_reward": 0.005394160805735737,
      "rewards/forensics_tiou_reward": 0.41600148379802704,
      "step": 312
    },
    {
      "completion_length": 19.6875,
      "epoch": 0.9615975422427036,
      "grad_norm": 10.407896995544434,
      "kl": 0.220703125,
      "learning_rate": 7.599999999999999e-07,
      "loss": 0.0194,
      "reward": 1.0708210468292236,
      "reward_std": 0.2340337634086609,
      "reward_std_all": 0.2036196067929268,
      "rewards-max/forensics_sinkhorn_reward": 0.9434762001037598,
      "rewards-max/forensics_strict_edge_reward": 0.11653948575258255,
      "rewards-max/forensics_tiou_reward": 0.8827558159828186,
      "rewards/forensics_sinkhorn_reward": 0.6725631654262543,
      "rewards/forensics_strict_edge_reward": 0.007284114370122552,
      "rewards/forensics_tiou_reward": 0.3909738063812256,
      "step": 313
    },
    {
      "completion_length": 19.6875,
      "epoch": 0.9646697388632872,
      "grad_norm": 9.129220008850098,
      "kl": 0.26953125,
      "learning_rate": 7.592307692307692e-07,
      "loss": -0.0212,
      "reward": 1.1626200079917908,
      "reward_std": 0.20023201406002045,
      "reward_std_all": 0.1741848662495613,
      "rewards-max/forensics_sinkhorn_reward": 0.9317967593669891,
      "rewards-max/forensics_strict_edge_reward": 0.0578472837805748,
      "rewards-max/forensics_tiou_reward": 0.8497599959373474,
      "rewards/forensics_sinkhorn_reward": 0.7280260026454926,
      "rewards/forensics_strict_edge_reward": 0.004685999592766166,
      "rewards/forensics_tiou_reward": 0.42990800738334656,
      "step": 314
    },
    {
      "completion_length": 21.5625,
      "epoch": 0.967741935483871,
      "grad_norm": 9.28159236907959,
      "kl": 0.2119140625,
      "learning_rate": 7.584615384615384e-07,
      "loss": 0.0631,
      "reward": 1.1882532835006714,
      "reward_std": 0.13796642422676086,
      "reward_std_all": 0.140682190656662,
      "rewards-max/forensics_sinkhorn_reward": 0.9639245867729187,
      "rewards-max/forensics_strict_edge_reward": 0.17388518527150154,
      "rewards-max/forensics_tiou_reward": 0.9402864575386047,
      "rewards/forensics_sinkhorn_reward": 0.7225594520568848,
      "rewards/forensics_strict_edge_reward": 0.012751839589327574,
      "rewards/forensics_tiou_reward": 0.45294196903705597,
      "step": 315
    },
    {
      "completion_length": 20.28125,
      "epoch": 0.9708141321044547,
      "grad_norm": 7.306722640991211,
      "kl": 0.4521484375,
      "learning_rate": 7.576923076923077e-07,
      "loss": 0.0269,
      "reward": 1.1198424100875854,
      "reward_std": 0.20970787107944489,
      "reward_std_all": 0.18169168382883072,
      "rewards-max/forensics_sinkhorn_reward": 0.9506092667579651,
      "rewards-max/forensics_strict_edge_reward": 0.32331598550081253,
      "rewards-max/forensics_tiou_reward": 0.8987183570861816,
      "rewards/forensics_sinkhorn_reward": 0.6748879253864288,
      "rewards/forensics_strict_edge_reward": 0.0174033734947443,
      "rewards/forensics_tiou_reward": 0.4275510609149933,
      "step": 316
    },
    {
      "completion_length": 19.921875,
      "epoch": 0.9738863287250384,
      "grad_norm": 7.6611433029174805,
      "kl": 0.2861328125,
      "learning_rate": 7.569230769230769e-07,
      "loss": 0.0036,
      "reward": 1.093949019908905,
      "reward_std": 0.16907277703285217,
      "reward_std_all": 0.15512463077902794,
      "rewards-max/forensics_sinkhorn_reward": 0.9350313246250153,
      "rewards-max/forensics_strict_edge_reward": 0.11293345876038074,
      "rewards-max/forensics_tiou_reward": 0.8475819826126099,
      "rewards/forensics_sinkhorn_reward": 0.6618775129318237,
      "rewards/forensics_strict_edge_reward": 0.006621147389523685,
      "rewards/forensics_tiou_reward": 0.4254503846168518,
      "step": 317
    },
    {
      "completion_length": 20.109375,
      "epoch": 0.9769585253456221,
      "grad_norm": 10.377766609191895,
      "kl": 0.248046875,
      "learning_rate": 7.561538461538461e-07,
      "loss": -0.1462,
      "reward": 1.0717853009700775,
      "reward_std": 0.12874817103147507,
      "reward_std_all": 0.13732775300741196,
      "rewards-max/forensics_sinkhorn_reward": 0.9002485275268555,
      "rewards-max/forensics_strict_edge_reward": 0.03025755099952221,
      "rewards-max/forensics_tiou_reward": 0.7787624299526215,
      "rewards/forensics_sinkhorn_reward": 0.6575988829135895,
      "rewards/forensics_strict_edge_reward": 0.00350572494789958,
      "rewards/forensics_tiou_reward": 0.4106806814670563,
      "step": 318
    },
    {
      "completion_length": 20.28125,
      "epoch": 0.9800307219662059,
      "grad_norm": 8.200782775878906,
      "kl": 0.216796875,
      "learning_rate": 7.553846153846154e-07,
      "loss": 0.0579,
      "reward": 0.9940164685249329,
      "reward_std": 0.16797137260437012,
      "reward_std_all": 0.17039624601602554,
      "rewards-max/forensics_sinkhorn_reward": 0.9400881230831146,
      "rewards-max/forensics_strict_edge_reward": 0.18344930466264486,
      "rewards-max/forensics_tiou_reward": 0.8729272782802582,
      "rewards/forensics_sinkhorn_reward": 0.6400894820690155,
      "rewards/forensics_strict_edge_reward": 0.010094537836266682,
      "rewards/forensics_tiou_reward": 0.34383244812488556,
      "step": 319
    },
    {
      "completion_length": 20.8125,
      "epoch": 0.9831029185867896,
      "grad_norm": 6.682837963104248,
      "kl": 0.2041015625,
      "learning_rate": 7.546153846153845e-07,
      "loss": -0.0508,
      "reward": 1.3219424486160278,
      "reward_std": 0.2531324326992035,
      "reward_std_all": 0.23627031594514847,
      "rewards-max/forensics_sinkhorn_reward": 0.9841972291469574,
      "rewards-max/forensics_strict_edge_reward": 0.7032848298549652,
      "rewards-max/forensics_tiou_reward": 0.983946681022644,
      "rewards/forensics_sinkhorn_reward": 0.7540999948978424,
      "rewards/forensics_strict_edge_reward": 0.038237886503338814,
      "rewards/forensics_tiou_reward": 0.5296045690774918,
      "step": 320
    },
    {
      "completion_length": 19.59375,
      "epoch": 0.9861751152073732,
      "grad_norm": 9.66579818725586,
      "kl": 0.53515625,
      "learning_rate": 7.538461538461538e-07,
      "loss": 0.0416,
      "reward": 1.088993787765503,
      "reward_std": 0.24366192519664764,
      "reward_std_all": 0.22660133242607117,
      "rewards-max/forensics_sinkhorn_reward": 0.9547257423400879,
      "rewards-max/forensics_strict_edge_reward": 0.33659186866134405,
      "rewards-max/forensics_tiou_reward": 0.9032056033611298,
      "rewards/forensics_sinkhorn_reward": 0.6539222598075867,
      "rewards/forensics_strict_edge_reward": 0.02525264848372899,
      "rewards/forensics_tiou_reward": 0.4098187983036041,
      "step": 321
    },
    {
      "completion_length": 20.21875,
      "epoch": 0.989247311827957,
      "grad_norm": 10.54845905303955,
      "kl": 0.2197265625,
      "learning_rate": 7.530769230769231e-07,
      "loss": -0.0347,
      "reward": 1.0660294592380524,
      "reward_std": 0.228775504976511,
      "reward_std_all": 0.20764752477407455,
      "rewards-max/forensics_sinkhorn_reward": 0.9308495819568634,
      "rewards-max/forensics_strict_edge_reward": 0.378445103764534,
      "rewards-max/forensics_tiou_reward": 0.8545367419719696,
      "rewards/forensics_sinkhorn_reward": 0.6553238332271576,
      "rewards/forensics_strict_edge_reward": 0.032701654359698296,
      "rewards/forensics_tiou_reward": 0.3780040442943573,
      "step": 322
    },
    {
      "completion_length": 20.59375,
      "epoch": 0.9923195084485407,
      "grad_norm": 8.474227905273438,
      "kl": 0.17333984375,
      "learning_rate": 7.523076923076923e-07,
      "loss": -0.0154,
      "reward": 1.0416680574417114,
      "reward_std": 0.17957991361618042,
      "reward_std_all": 0.16899757087230682,
      "rewards-max/forensics_sinkhorn_reward": 0.9242610037326813,
      "rewards-max/forensics_strict_edge_reward": 0.0257661291398108,
      "rewards-max/forensics_tiou_reward": 0.8245218098163605,
      "rewards/forensics_sinkhorn_reward": 0.6603865027427673,
      "rewards/forensics_strict_edge_reward": 0.0016721200954634696,
      "rewards/forensics_tiou_reward": 0.3796093761920929,
      "step": 323
    },
    {
      "completion_length": 20.078125,
      "epoch": 0.9953917050691244,
      "grad_norm": 11.243912696838379,
      "kl": 0.1884765625,
      "learning_rate": 7.515384615384615e-07,
      "loss": 0.0141,
      "reward": 1.2067126035690308,
      "reward_std": 0.2399412840604782,
      "reward_std_all": 0.21626726537942886,
      "rewards-max/forensics_sinkhorn_reward": 0.9502066373825073,
      "rewards-max/forensics_strict_edge_reward": 0.08196062594652176,
      "rewards-max/forensics_tiou_reward": 0.8948597609996796,
      "rewards/forensics_sinkhorn_reward": 0.7265220582485199,
      "rewards/forensics_strict_edge_reward": 0.00575527991168201,
      "rewards/forensics_tiou_reward": 0.4744352698326111,
      "step": 324
    },
    {
      "completion_length": 22.34375,
      "epoch": 0.9984639016897081,
      "grad_norm": 8.794547080993652,
      "kl": 0.3349609375,
      "learning_rate": 7.507692307692307e-07,
      "loss": -0.0523,
      "reward": 1.159641683101654,
      "reward_std": 0.2550933137536049,
      "reward_std_all": 0.23275870084762573,
      "rewards-max/forensics_sinkhorn_reward": 0.9428018033504486,
      "rewards-max/forensics_strict_edge_reward": 0.29233212023973465,
      "rewards-max/forensics_tiou_reward": 0.890495777130127,
      "rewards/forensics_sinkhorn_reward": 0.6782707571983337,
      "rewards/forensics_strict_edge_reward": 0.019851332064718008,
      "rewards/forensics_tiou_reward": 0.46151961386203766,
      "step": 325
    },
    {
      "completion_length": 19.200000762939453,
      "epoch": 1.0,
      "grad_norm": 5.086738586425781,
      "kl": 0.1796875,
      "learning_rate": 7.5e-07,
      "loss": -0.0241,
      "reward": 1.2149256467819214,
      "reward_std": 0.22371283173561096,
      "reward_std_all": 0.18668751418590546,
      "rewards-max/forensics_sinkhorn_reward": 0.9848296046257019,
      "rewards-max/forensics_strict_edge_reward": 0.3558189272880554,
      "rewards-max/forensics_tiou_reward": 0.9877390265464783,
      "rewards/forensics_sinkhorn_reward": 0.7231308817863464,
      "rewards/forensics_strict_edge_reward": 0.07116378843784332,
      "rewards/forensics_tiou_reward": 0.42063093185424805,
      "step": 326
    },
    {
      "completion_length": 22.953125,
      "epoch": 1.0030721966205838,
      "grad_norm": 8.507364273071289,
      "kl": 0.22119140625,
      "learning_rate": 7.492307692307692e-07,
      "loss": -0.0328,
      "reward": 1.1631531119346619,
      "reward_std": 0.2293170467019081,
      "reward_std_all": 0.22992753982543945,
      "rewards-max/forensics_sinkhorn_reward": 0.9608497321605682,
      "rewards-max/forensics_strict_edge_reward": 0.29493413120508194,
      "rewards-max/forensics_tiou_reward": 0.9241732060909271,
      "rewards/forensics_sinkhorn_reward": 0.6821890771389008,
      "rewards/forensics_strict_edge_reward": 0.02904164302162826,
      "rewards/forensics_tiou_reward": 0.45192235708236694,
      "step": 327
    },
    {
      "completion_length": 21.609375,
      "epoch": 1.0061443932411673,
      "grad_norm": 6.950249671936035,
      "kl": 0.306640625,
      "learning_rate": 7.484615384615384e-07,
      "loss": -0.0095,
      "reward": 1.34581458568573,
      "reward_std": 0.2817128002643585,
      "reward_std_all": 0.2562016546726227,
      "rewards-max/forensics_sinkhorn_reward": 0.9782715141773224,
      "rewards-max/forensics_strict_edge_reward": 0.3894408792257309,
      "rewards-max/forensics_tiou_reward": 0.9681523144245148,
      "rewards/forensics_sinkhorn_reward": 0.7637516558170319,
      "rewards/forensics_strict_edge_reward": 0.03935872204601765,
      "rewards/forensics_tiou_reward": 0.542704164981842,
      "step": 328
    },
    {
      "completion_length": 22.3125,
      "epoch": 1.0092165898617511,
      "grad_norm": 7.762089729309082,
      "kl": 0.2568359375,
      "learning_rate": 7.476923076923077e-07,
      "loss": 0.0582,
      "reward": 1.1461085677146912,
      "reward_std": 0.27261775732040405,
      "reward_std_all": 0.24819360673427582,
      "rewards-max/forensics_sinkhorn_reward": 0.9644580185413361,
      "rewards-max/forensics_strict_edge_reward": 0.2554169371724129,
      "rewards-max/forensics_tiou_reward": 0.9379132986068726,
      "rewards/forensics_sinkhorn_reward": 0.6852881908416748,
      "rewards/forensics_strict_edge_reward": 0.015692250803112984,
      "rewards/forensics_tiou_reward": 0.4451281428337097,
      "step": 329
    },
    {
      "completion_length": 21.75,
      "epoch": 1.012288786482335,
      "grad_norm": 8.424525260925293,
      "kl": 0.287109375,
      "learning_rate": 7.469230769230769e-07,
      "loss": 0.0242,
      "reward": 1.267612874507904,
      "reward_std": 0.2004832923412323,
      "reward_std_all": 0.19057806581258774,
      "rewards-max/forensics_sinkhorn_reward": 0.9757124781608582,
      "rewards-max/forensics_strict_edge_reward": 0.30387546122074127,
      "rewards-max/forensics_tiou_reward": 0.9651798903942108,
      "rewards/forensics_sinkhorn_reward": 0.7419022023677826,
      "rewards/forensics_strict_edge_reward": 0.029144251719117165,
      "rewards/forensics_tiou_reward": 0.496566504240036,
      "step": 330
    },
    {
      "completion_length": 21.359375,
      "epoch": 1.0153609831029187,
      "grad_norm": 9.461750030517578,
      "kl": 0.20556640625,
      "learning_rate": 7.461538461538461e-07,
      "loss": 0.0193,
      "reward": 1.2155092358589172,
      "reward_std": 0.20191840827465057,
      "reward_std_all": 0.18258707970380783,
      "rewards-max/forensics_sinkhorn_reward": 0.9621948897838593,
      "rewards-max/forensics_strict_edge_reward": 0.3859708905220032,
      "rewards-max/forensics_tiou_reward": 0.9327269792556763,
      "rewards/forensics_sinkhorn_reward": 0.7114246785640717,
      "rewards/forensics_strict_edge_reward": 0.027449597604572773,
      "rewards/forensics_tiou_reward": 0.4766349494457245,
      "step": 331
    },
    {
      "completion_length": 22.203125,
      "epoch": 1.0184331797235022,
      "grad_norm": 9.105042457580566,
      "kl": 0.2998046875,
      "learning_rate": 7.453846153846154e-07,
      "loss": 0.0653,
      "reward": 1.136249840259552,
      "reward_std": 0.19600585103034973,
      "reward_std_all": 0.18825282901525497,
      "rewards-max/forensics_sinkhorn_reward": 0.9606281220912933,
      "rewards-max/forensics_strict_edge_reward": 0.4047088027000427,
      "rewards-max/forensics_tiou_reward": 0.9308620095252991,
      "rewards/forensics_sinkhorn_reward": 0.6592019498348236,
      "rewards/forensics_strict_edge_reward": 0.03237797133624554,
      "rewards/forensics_tiou_reward": 0.44467002153396606,
      "step": 332
    },
    {
      "completion_length": 20.375,
      "epoch": 1.021505376344086,
      "grad_norm": 6.478119850158691,
      "kl": 0.3779296875,
      "learning_rate": 7.446153846153845e-07,
      "loss": 0.0862,
      "reward": 1.2355851531028748,
      "reward_std": 0.22925814986228943,
      "reward_std_all": 0.20585619658231735,
      "rewards-max/forensics_sinkhorn_reward": 0.9712669551372528,
      "rewards-max/forensics_strict_edge_reward": 0.30007198452949524,
      "rewards-max/forensics_tiou_reward": 0.9485511779785156,
      "rewards/forensics_sinkhorn_reward": 0.7269938290119171,
      "rewards/forensics_strict_edge_reward": 0.025009713135659695,
      "rewards/forensics_tiou_reward": 0.48358166217803955,
      "step": 333
    },
    {
      "completion_length": 23.484375,
      "epoch": 1.0245775729646698,
      "grad_norm": 8.216143608093262,
      "kl": 0.21435546875,
      "learning_rate": 7.438461538461538e-07,
      "loss": -0.0435,
      "reward": 1.1620039343833923,
      "reward_std": 0.2249264270067215,
      "reward_std_all": 0.2208845466375351,
      "rewards-max/forensics_sinkhorn_reward": 0.9475880861282349,
      "rewards-max/forensics_strict_edge_reward": 0.4427416920661926,
      "rewards-max/forensics_tiou_reward": 0.9045083224773407,
      "rewards/forensics_sinkhorn_reward": 0.6804170906543732,
      "rewards/forensics_strict_edge_reward": 0.025705129839479923,
      "rewards/forensics_tiou_reward": 0.4558817595243454,
      "step": 334
    },
    {
      "completion_length": 20.9375,
      "epoch": 1.0276497695852536,
      "grad_norm": 10.865671157836914,
      "kl": 0.2705078125,
      "learning_rate": 7.43076923076923e-07,
      "loss": 0.0481,
      "reward": 1.1002418994903564,
      "reward_std": 0.21965602040290833,
      "reward_std_all": 0.1883881539106369,
      "rewards-max/forensics_sinkhorn_reward": 0.9375995695590973,
      "rewards-max/forensics_strict_edge_reward": 0.15664700604975224,
      "rewards-max/forensics_tiou_reward": 0.8697832524776459,
      "rewards/forensics_sinkhorn_reward": 0.6529262065887451,
      "rewards/forensics_strict_edge_reward": 0.009350572829134762,
      "rewards/forensics_tiou_reward": 0.4379650354385376,
      "step": 335
    },
    {
      "completion_length": 24.21875,
      "epoch": 1.0307219662058371,
      "grad_norm": 10.187251091003418,
      "kl": 0.2353515625,
      "learning_rate": 7.423076923076923e-07,
      "loss": -0.0382,
      "reward": 1.2702349424362183,
      "reward_std": 0.26818765699863434,
      "reward_std_all": 0.2544950842857361,
      "rewards-max/forensics_sinkhorn_reward": 0.9732778966426849,
      "rewards-max/forensics_strict_edge_reward": 0.2710481956601143,
      "rewards-max/forensics_tiou_reward": 0.9548081159591675,
      "rewards/forensics_sinkhorn_reward": 0.6964791715145111,
      "rewards/forensics_strict_edge_reward": 0.02293882891535759,
      "rewards/forensics_tiou_reward": 0.550816997885704,
      "step": 336
    },
    {
      "completion_length": 20.328125,
      "epoch": 1.033794162826421,
      "grad_norm": 8.854069709777832,
      "kl": 0.2333984375,
      "learning_rate": 7.415384615384616e-07,
      "loss": -0.0647,
      "reward": 1.1749500036239624,
      "reward_std": 0.17430628836154938,
      "reward_std_all": 0.22226862609386444,
      "rewards-max/forensics_sinkhorn_reward": 0.935355931520462,
      "rewards-max/forensics_strict_edge_reward": 0.0733795166015625,
      "rewards-max/forensics_tiou_reward": 0.8690758645534515,
      "rewards/forensics_sinkhorn_reward": 0.7195149660110474,
      "rewards/forensics_strict_edge_reward": 0.007297284202650189,
      "rewards/forensics_tiou_reward": 0.44813771545886993,
      "step": 337
    },
    {
      "completion_length": 21.484375,
      "epoch": 1.0368663594470047,
      "grad_norm": 8.48971176147461,
      "kl": 0.2744140625,
      "learning_rate": 7.407692307692307e-07,
      "loss": 0.0224,
      "reward": 1.1233460307121277,
      "reward_std": 0.2160046398639679,
      "reward_std_all": 0.19754396378993988,
      "rewards-max/forensics_sinkhorn_reward": 0.9477084875106812,
      "rewards-max/forensics_strict_edge_reward": 0.1959612954815384,
      "rewards-max/forensics_tiou_reward": 0.8725080192089081,
      "rewards/forensics_sinkhorn_reward": 0.69691601395607,
      "rewards/forensics_strict_edge_reward": 0.010284166732162703,
      "rewards/forensics_tiou_reward": 0.41614581644535065,
      "step": 338
    },
    {
      "completion_length": 24.0,
      "epoch": 1.0399385560675882,
      "grad_norm": 9.964242935180664,
      "kl": 0.193359375,
      "learning_rate": 7.4e-07,
      "loss": 0.0482,
      "reward": 1.223469853401184,
      "reward_std": 0.2685698792338371,
      "reward_std_all": 0.2331305518746376,
      "rewards-max/forensics_sinkhorn_reward": 0.9859246611595154,
      "rewards-max/forensics_strict_edge_reward": 0.6405399739742279,
      "rewards-max/forensics_tiou_reward": 0.988789439201355,
      "rewards/forensics_sinkhorn_reward": 0.695559173822403,
      "rewards/forensics_strict_edge_reward": 0.03652677498757839,
      "rewards/forensics_tiou_reward": 0.491383895277977,
      "step": 339
    },
    {
      "completion_length": 23.609375,
      "epoch": 1.043010752688172,
      "grad_norm": 5.913001537322998,
      "kl": 0.19580078125,
      "learning_rate": 7.392307692307692e-07,
      "loss": 0.0811,
      "reward": 1.2067844867706299,
      "reward_std": 0.1958761066198349,
      "reward_std_all": 0.1737028881907463,
      "rewards-max/forensics_sinkhorn_reward": 0.9203566312789917,
      "rewards-max/forensics_strict_edge_reward": 0.40824079513549805,
      "rewards-max/forensics_tiou_reward": 0.834070473909378,
      "rewards/forensics_sinkhorn_reward": 0.716234415769577,
      "rewards/forensics_strict_edge_reward": 0.029100721701979637,
      "rewards/forensics_tiou_reward": 0.46144936978816986,
      "step": 340
    },
    {
      "completion_length": 21.640625,
      "epoch": 1.0460829493087558,
      "grad_norm": 8.754416465759277,
      "kl": 0.31005859375,
      "learning_rate": 7.384615384615384e-07,
      "loss": 0.1188,
      "reward": 1.1619902849197388,
      "reward_std": 0.309808149933815,
      "reward_std_all": 0.2721398174762726,
      "rewards-max/forensics_sinkhorn_reward": 0.9641371071338654,
      "rewards-max/forensics_strict_edge_reward": 0.15508664026856422,
      "rewards-max/forensics_tiou_reward": 0.926781177520752,
      "rewards/forensics_sinkhorn_reward": 0.6960135102272034,
      "rewards/forensics_strict_edge_reward": 0.007931085187010467,
      "rewards/forensics_tiou_reward": 0.4580456614494324,
      "step": 341
    },
    {
      "completion_length": 23.109375,
      "epoch": 1.0491551459293396,
      "grad_norm": 9.089052200317383,
      "kl": 0.20458984375,
      "learning_rate": 7.376923076923076e-07,
      "loss": -0.0137,
      "reward": 1.1071144938468933,
      "reward_std": 0.16347283869981766,
      "reward_std_all": 0.1578163504600525,
      "rewards-max/forensics_sinkhorn_reward": 0.9407894611358643,
      "rewards-max/forensics_strict_edge_reward": 0.07367527484893799,
      "rewards-max/forensics_tiou_reward": 0.875519186258316,
      "rewards/forensics_sinkhorn_reward": 0.6814257800579071,
      "rewards/forensics_strict_edge_reward": 0.004823342896997929,
      "rewards/forensics_tiou_reward": 0.42086540162563324,
      "step": 342
    },
    {
      "completion_length": 23.296875,
      "epoch": 1.0522273425499231,
      "grad_norm": 8.686434745788574,
      "kl": 0.16259765625,
      "learning_rate": 7.369230769230769e-07,
      "loss": 0.0673,
      "reward": 1.0864461660385132,
      "reward_std": 0.2560608685016632,
      "reward_std_all": 0.2633008435368538,
      "rewards-max/forensics_sinkhorn_reward": 0.9889523088932037,
      "rewards-max/forensics_strict_edge_reward": 0.9677534997463226,
      "rewards-max/forensics_tiou_reward": 0.9998709559440613,
      "rewards/forensics_sinkhorn_reward": 0.6191026568412781,
      "rewards/forensics_strict_edge_reward": 0.08396251127123833,
      "rewards/forensics_tiou_reward": 0.3833809942007065,
      "step": 343
    },
    {
      "completion_length": 25.46875,
      "epoch": 1.055299539170507,
      "grad_norm": 8.563838958740234,
      "kl": 0.21337890625,
      "learning_rate": 7.361538461538461e-07,
      "loss": 0.0121,
      "reward": 1.219710648059845,
      "reward_std": 0.2770746946334839,
      "reward_std_all": 0.2566698417067528,
      "rewards-max/forensics_sinkhorn_reward": 0.9705241024494171,
      "rewards-max/forensics_strict_edge_reward": 0.5373985767364502,
      "rewards-max/forensics_tiou_reward": 0.946328341960907,
      "rewards/forensics_sinkhorn_reward": 0.6802875101566315,
      "rewards/forensics_strict_edge_reward": 0.040527600795030594,
      "rewards/forensics_tiou_reward": 0.49889563024044037,
      "step": 344
    },
    {
      "completion_length": 24.0625,
      "epoch": 1.0583717357910907,
      "grad_norm": 7.394922256469727,
      "kl": 0.2578125,
      "learning_rate": 7.353846153846154e-07,
      "loss": -0.0464,
      "reward": 1.3395636677742004,
      "reward_std": 0.2512587532401085,
      "reward_std_all": 0.2227802351117134,
      "rewards-max/forensics_sinkhorn_reward": 0.9840532839298248,
      "rewards-max/forensics_strict_edge_reward": 0.6149742305278778,
      "rewards-max/forensics_tiou_reward": 0.9863326251506805,
      "rewards/forensics_sinkhorn_reward": 0.7319489121437073,
      "rewards/forensics_strict_edge_reward": 0.05716104805469513,
      "rewards/forensics_tiou_reward": 0.5504537224769592,
      "step": 345
    },
    {
      "completion_length": 24.53125,
      "epoch": 1.0614439324116745,
      "grad_norm": 7.866548538208008,
      "kl": 0.20654296875,
      "learning_rate": 7.346153846153846e-07,
      "loss": -0.0594,
      "reward": 1.1032223105430603,
      "reward_std": 0.2295580878853798,
      "reward_std_all": 0.20377005636692047,
      "rewards-max/forensics_sinkhorn_reward": 0.9302274882793427,
      "rewards-max/forensics_strict_edge_reward": 0.06890701688826084,
      "rewards-max/forensics_tiou_reward": 0.8498712182044983,
      "rewards/forensics_sinkhorn_reward": 0.6712505221366882,
      "rewards/forensics_strict_edge_reward": 0.005462253539008088,
      "rewards/forensics_tiou_reward": 0.4265096038579941,
      "step": 346
    },
    {
      "completion_length": 21.296875,
      "epoch": 1.064516129032258,
      "grad_norm": 7.47238302230835,
      "kl": 0.2265625,
      "learning_rate": 7.338461538461538e-07,
      "loss": 0.1073,
      "reward": 1.251043438911438,
      "reward_std": 0.2003503441810608,
      "reward_std_all": 0.18174002319574356,
      "rewards-max/forensics_sinkhorn_reward": 0.9539846479892731,
      "rewards-max/forensics_strict_edge_reward": 0.04513075388967991,
      "rewards-max/forensics_tiou_reward": 0.8985658288002014,
      "rewards/forensics_sinkhorn_reward": 0.7400085926055908,
      "rewards/forensics_strict_edge_reward": 0.006419135257601738,
      "rewards/forensics_tiou_reward": 0.5046157091856003,
      "step": 347
    },
    {
      "completion_length": 23.34375,
      "epoch": 1.0675883256528418,
      "grad_norm": 8.76345157623291,
      "kl": 0.2080078125,
      "learning_rate": 7.33076923076923e-07,
      "loss": -0.0224,
      "reward": 1.245202362537384,
      "reward_std": 0.2853686809539795,
      "reward_std_all": 0.2616772949695587,
      "rewards-max/forensics_sinkhorn_reward": 0.9707030653953552,
      "rewards-max/forensics_strict_edge_reward": 0.5221183001995087,
      "rewards-max/forensics_tiou_reward": 0.9551535248756409,
      "rewards/forensics_sinkhorn_reward": 0.7190852761268616,
      "rewards/forensics_strict_edge_reward": 0.04089144803583622,
      "rewards/forensics_tiou_reward": 0.4852256178855896,
      "step": 348
    },
    {
      "completion_length": 21.421875,
      "epoch": 1.0706605222734256,
      "grad_norm": 8.707110404968262,
      "kl": 0.435546875,
      "learning_rate": 7.323076923076923e-07,
      "loss": -0.0268,
      "reward": 1.3396052718162537,
      "reward_std": 0.2269831895828247,
      "reward_std_all": 0.20604582875967026,
      "rewards-max/forensics_sinkhorn_reward": 0.9747450351715088,
      "rewards-max/forensics_strict_edge_reward": 0.4745236784219742,
      "rewards-max/forensics_tiou_reward": 0.9615511000156403,
      "rewards/forensics_sinkhorn_reward": 0.7398900985717773,
      "rewards/forensics_strict_edge_reward": 0.05413376912474632,
      "rewards/forensics_tiou_reward": 0.5455813854932785,
      "step": 349
    },
    {
      "completion_length": 22.375,
      "epoch": 1.0737327188940091,
      "grad_norm": 8.603501319885254,
      "kl": 0.2060546875,
      "learning_rate": 7.315384615384616e-07,
      "loss": -0.0067,
      "reward": 1.2230606079101562,
      "reward_std": 0.28275713324546814,
      "reward_std_all": 0.24632877856492996,
      "rewards-max/forensics_sinkhorn_reward": 0.9677044749259949,
      "rewards-max/forensics_strict_edge_reward": 0.5629585310816765,
      "rewards-max/forensics_tiou_reward": 0.9504220187664032,
      "rewards/forensics_sinkhorn_reward": 0.7105843424797058,
      "rewards/forensics_strict_edge_reward": 0.03337868023663759,
      "rewards/forensics_tiou_reward": 0.4790976047515869,
      "step": 350
    },
    {
      "completion_length": 22.28125,
      "epoch": 1.076804915514593,
      "grad_norm": 10.913789749145508,
      "kl": 0.28662109375,
      "learning_rate": 7.307692307692307e-07,
      "loss": -0.0323,
      "reward": 1.0657468438148499,
      "reward_std": 0.2638644874095917,
      "reward_std_all": 0.2667464017868042,
      "rewards-max/forensics_sinkhorn_reward": 0.9606557190418243,
      "rewards-max/forensics_strict_edge_reward": 0.17111246660351753,
      "rewards-max/forensics_tiou_reward": 0.9234909415245056,
      "rewards/forensics_sinkhorn_reward": 0.6446837782859802,
      "rewards/forensics_strict_edge_reward": 0.007862514816224575,
      "rewards/forensics_tiou_reward": 0.4132004678249359,
      "step": 351
    },
    {
      "completion_length": 21.984375,
      "epoch": 1.0798771121351767,
      "grad_norm": 11.698018074035645,
      "kl": 0.20849609375,
      "learning_rate": 7.3e-07,
      "loss": -0.0876,
      "reward": 1.0322181582450867,
      "reward_std": 0.21604017913341522,
      "reward_std_all": 0.2169335037469864,
      "rewards-max/forensics_sinkhorn_reward": 0.9747186005115509,
      "rewards-max/forensics_strict_edge_reward": 0.13095705211162567,
      "rewards-max/forensics_tiou_reward": 0.9553414583206177,
      "rewards/forensics_sinkhorn_reward": 0.6532596945762634,
      "rewards/forensics_strict_edge_reward": 0.007610791770275682,
      "rewards/forensics_tiou_reward": 0.37134766578674316,
      "step": 352
    },
    {
      "completion_length": 19.640625,
      "epoch": 1.0829493087557605,
      "grad_norm": 11.0132417678833,
      "kl": 0.298828125,
      "learning_rate": 7.292307692307692e-07,
      "loss": 0.0403,
      "reward": 1.128217875957489,
      "reward_std": 0.14819781482219696,
      "reward_std_all": 0.19214804470539093,
      "rewards-max/forensics_sinkhorn_reward": 0.9595929682254791,
      "rewards-max/forensics_strict_edge_reward": 0.15931017510592937,
      "rewards-max/forensics_tiou_reward": 0.924811065196991,
      "rewards/forensics_sinkhorn_reward": 0.6959660649299622,
      "rewards/forensics_strict_edge_reward": 0.011871166061609983,
      "rewards/forensics_tiou_reward": 0.4203806221485138,
      "step": 353
    },
    {
      "completion_length": 21.375,
      "epoch": 1.086021505376344,
      "grad_norm": 8.654870986938477,
      "kl": 0.24462890625,
      "learning_rate": 7.284615384615384e-07,
      "loss": 0.0182,
      "reward": 1.0237433314323425,
      "reward_std": 0.20416728407144547,
      "reward_std_all": 0.19035182893276215,
      "rewards-max/forensics_sinkhorn_reward": 0.9091728031635284,
      "rewards-max/forensics_strict_edge_reward": 0.13823694735765457,
      "rewards-max/forensics_tiou_reward": 0.7941935360431671,
      "rewards/forensics_sinkhorn_reward": 0.6216495335102081,
      "rewards/forensics_strict_edge_reward": 0.006270183948799968,
      "rewards/forensics_tiou_reward": 0.3958236128091812,
      "step": 354
    },
    {
      "completion_length": 22.375,
      "epoch": 1.0890937019969278,
      "grad_norm": 8.380929946899414,
      "kl": 0.21337890625,
      "learning_rate": 7.276923076923076e-07,
      "loss": 0.0814,
      "reward": 1.1686241626739502,
      "reward_std": 0.19538137316703796,
      "reward_std_all": 0.1922006607055664,
      "rewards-max/forensics_sinkhorn_reward": 0.971444696187973,
      "rewards-max/forensics_strict_edge_reward": 0.4412408843636513,
      "rewards-max/forensics_tiou_reward": 0.9498420655727386,
      "rewards/forensics_sinkhorn_reward": 0.6930161714553833,
      "rewards/forensics_strict_edge_reward": 0.02574639767408371,
      "rewards/forensics_tiou_reward": 0.4498615115880966,
      "step": 355
    },
    {
      "completion_length": 21.234375,
      "epoch": 1.0921658986175116,
      "grad_norm": 9.9564847946167,
      "kl": 0.24267578125,
      "learning_rate": 7.269230769230769e-07,
      "loss": -0.0594,
      "reward": 1.060649424791336,
      "reward_std": 0.17964737117290497,
      "reward_std_all": 0.15762070193886757,
      "rewards-max/forensics_sinkhorn_reward": 0.9169218242168427,
      "rewards-max/forensics_strict_edge_reward": 0.09066696651279926,
      "rewards-max/forensics_tiou_reward": 0.8098838627338409,
      "rewards/forensics_sinkhorn_reward": 0.6688573062419891,
      "rewards/forensics_strict_edge_reward": 0.005610813153907657,
      "rewards/forensics_tiou_reward": 0.38618141412734985,
      "step": 356
    },
    {
      "completion_length": 23.796875,
      "epoch": 1.0952380952380953,
      "grad_norm": 10.477092742919922,
      "kl": 0.31640625,
      "learning_rate": 7.261538461538461e-07,
      "loss": -0.0514,
      "reward": 1.2697123885154724,
      "reward_std": 0.24432892352342606,
      "reward_std_all": 0.2317807823419571,
      "rewards-max/forensics_sinkhorn_reward": 0.9669291079044342,
      "rewards-max/forensics_strict_edge_reward": 0.43728881794959307,
      "rewards-max/forensics_tiou_reward": 0.9384308159351349,
      "rewards/forensics_sinkhorn_reward": 0.7090480029582977,
      "rewards/forensics_strict_edge_reward": 0.030906643252819777,
      "rewards/forensics_tiou_reward": 0.5297577679157257,
      "step": 357
    },
    {
      "completion_length": 23.984375,
      "epoch": 1.098310291858679,
      "grad_norm": 7.683534622192383,
      "kl": 0.416015625,
      "learning_rate": 7.253846153846154e-07,
      "loss": 0.0718,
      "reward": 1.0911137461662292,
      "reward_std": 0.18351136520504951,
      "reward_std_all": 0.16198968142271042,
      "rewards-max/forensics_sinkhorn_reward": 0.9266164898872375,
      "rewards-max/forensics_strict_edge_reward": 0.029793493449687958,
      "rewards-max/forensics_tiou_reward": 0.8348208069801331,
      "rewards/forensics_sinkhorn_reward": 0.6643957197666168,
      "rewards/forensics_strict_edge_reward": 0.0024777494836598635,
      "rewards/forensics_tiou_reward": 0.42424020171165466,
      "step": 358
    },
    {
      "completion_length": 22.34375,
      "epoch": 1.1013824884792627,
      "grad_norm": 9.614656448364258,
      "kl": 0.189453125,
      "learning_rate": 7.246153846153846e-07,
      "loss": -0.0355,
      "reward": 1.2241398096084595,
      "reward_std": 0.12598233669996262,
      "reward_std_all": 0.14249039441347122,
      "rewards-max/forensics_sinkhorn_reward": 0.9223180413246155,
      "rewards-max/forensics_strict_edge_reward": 0.18404191732406616,
      "rewards-max/forensics_tiou_reward": 0.8244657516479492,
      "rewards/forensics_sinkhorn_reward": 0.7222886383533478,
      "rewards/forensics_strict_edge_reward": 0.01464269682765007,
      "rewards/forensics_tiou_reward": 0.4872085154056549,
      "step": 359
    },
    {
      "completion_length": 20.53125,
      "epoch": 1.1044546850998465,
      "grad_norm": 10.693521499633789,
      "kl": 0.23876953125,
      "learning_rate": 7.238461538461538e-07,
      "loss": 0.149,
      "reward": 1.3650614023208618,
      "reward_std": 0.19855770468711853,
      "reward_std_all": 0.21702422946691513,
      "rewards-max/forensics_sinkhorn_reward": 0.9824813306331635,
      "rewards-max/forensics_strict_edge_reward": 0.45164985954761505,
      "rewards-max/forensics_tiou_reward": 0.97831591963768,
      "rewards/forensics_sinkhorn_reward": 0.7739178240299225,
      "rewards/forensics_strict_edge_reward": 0.026968346908688545,
      "rewards/forensics_tiou_reward": 0.5641751885414124,
      "step": 360
    },
    {
      "completion_length": 21.65625,
      "epoch": 1.10752688172043,
      "grad_norm": 14.32019329071045,
      "kl": 0.16552734375,
      "learning_rate": 7.23076923076923e-07,
      "loss": 0.0408,
      "reward": 1.2248879671096802,
      "reward_std": 0.2843654677271843,
      "reward_std_all": 0.27846934646368027,
      "rewards-max/forensics_sinkhorn_reward": 0.9670009315013885,
      "rewards-max/forensics_strict_edge_reward": 0.39888976514339447,
      "rewards-max/forensics_tiou_reward": 0.9449582099914551,
      "rewards/forensics_sinkhorn_reward": 0.7116355299949646,
      "rewards/forensics_strict_edge_reward": 0.02037563221529126,
      "rewards/forensics_tiou_reward": 0.4928768277168274,
      "step": 361
    },
    {
      "completion_length": 21.28125,
      "epoch": 1.1105990783410138,
      "grad_norm": 10.029119491577148,
      "kl": 0.587890625,
      "learning_rate": 7.223076923076922e-07,
      "loss": 0.0704,
      "reward": 1.292137086391449,
      "reward_std": 0.24061137437820435,
      "reward_std_all": 0.22920388728380203,
      "rewards-max/forensics_sinkhorn_reward": 0.9807115495204926,
      "rewards-max/forensics_strict_edge_reward": 0.49658530950546265,
      "rewards-max/forensics_tiou_reward": 0.9754151701927185,
      "rewards/forensics_sinkhorn_reward": 0.7492566704750061,
      "rewards/forensics_strict_edge_reward": 0.03315080050379038,
      "rewards/forensics_tiou_reward": 0.5097295492887497,
      "step": 362
    },
    {
      "completion_length": 24.0625,
      "epoch": 1.1136712749615976,
      "grad_norm": 6.664681911468506,
      "kl": 0.18798828125,
      "learning_rate": 7.215384615384616e-07,
      "loss": 0.0553,
      "reward": 1.331157922744751,
      "reward_std": 0.23314248025417328,
      "reward_std_all": 0.2504969984292984,
      "rewards-max/forensics_sinkhorn_reward": 0.9675873517990112,
      "rewards-max/forensics_strict_edge_reward": 0.30460435152053833,
      "rewards-max/forensics_tiou_reward": 0.9493195712566376,
      "rewards/forensics_sinkhorn_reward": 0.749805361032486,
      "rewards/forensics_strict_edge_reward": 0.028917310759425163,
      "rewards/forensics_tiou_reward": 0.5524351596832275,
      "step": 363
    },
    {
      "completion_length": 21.609375,
      "epoch": 1.1167434715821813,
      "grad_norm": 8.997255325317383,
      "kl": 0.357421875,
      "learning_rate": 7.207692307692307e-07,
      "loss": 0.0514,
      "reward": 1.1539511680603027,
      "reward_std": 0.2267455980181694,
      "reward_std_all": 0.20488235354423523,
      "rewards-max/forensics_sinkhorn_reward": 0.9559088945388794,
      "rewards-max/forensics_strict_edge_reward": 0.35662665218114853,
      "rewards-max/forensics_tiou_reward": 0.9173563718795776,
      "rewards/forensics_sinkhorn_reward": 0.6751638352870941,
      "rewards/forensics_strict_edge_reward": 0.030509494245052338,
      "rewards/forensics_tiou_reward": 0.4482778310775757,
      "step": 364
    },
    {
      "completion_length": 21.640625,
      "epoch": 1.119815668202765,
      "grad_norm": 8.674544334411621,
      "kl": 0.3212890625,
      "learning_rate": 7.2e-07,
      "loss": -0.0128,
      "reward": 1.1884440183639526,
      "reward_std": 0.21542474627494812,
      "reward_std_all": 0.20346584916114807,
      "rewards-max/forensics_sinkhorn_reward": 0.9573847651481628,
      "rewards-max/forensics_strict_edge_reward": 0.2475353181362152,
      "rewards-max/forensics_tiou_reward": 0.9214784204959869,
      "rewards/forensics_sinkhorn_reward": 0.7078144252300262,
      "rewards/forensics_strict_edge_reward": 0.028419330716133118,
      "rewards/forensics_tiou_reward": 0.45221027731895447,
      "step": 365
    },
    {
      "completion_length": 21.34375,
      "epoch": 1.1228878648233487,
      "grad_norm": 9.976686477661133,
      "kl": 0.3056640625,
      "learning_rate": 7.192307692307692e-07,
      "loss": -0.1412,
      "reward": 1.2442159056663513,
      "reward_std": 0.1519678384065628,
      "reward_std_all": 0.1358398050069809,
      "rewards-max/forensics_sinkhorn_reward": 0.953496903181076,
      "rewards-max/forensics_strict_edge_reward": 0.15541210025548935,
      "rewards-max/forensics_tiou_reward": 0.9043993353843689,
      "rewards/forensics_sinkhorn_reward": 0.7368818521499634,
      "rewards/forensics_strict_edge_reward": 0.01936138141900301,
      "rewards/forensics_tiou_reward": 0.48797276616096497,
      "step": 366
    },
    {
      "completion_length": 23.15625,
      "epoch": 1.1259600614439325,
      "grad_norm": 9.815890312194824,
      "kl": 0.39697265625,
      "learning_rate": 7.184615384615384e-07,
      "loss": 0.0363,
      "reward": 1.0138572454452515,
      "reward_std": 0.2482338473200798,
      "reward_std_all": 0.2229461520910263,
      "rewards-max/forensics_sinkhorn_reward": 0.9257266223430634,
      "rewards-max/forensics_strict_edge_reward": 0.18184739723801613,
      "rewards-max/forensics_tiou_reward": 0.8492957651615143,
      "rewards/forensics_sinkhorn_reward": 0.6117965579032898,
      "rewards/forensics_strict_edge_reward": 0.015949618071317673,
      "rewards/forensics_tiou_reward": 0.3861110359430313,
      "step": 367
    },
    {
      "completion_length": 22.5,
      "epoch": 1.129032258064516,
      "grad_norm": 9.739683151245117,
      "kl": 0.36572265625,
      "learning_rate": 7.176923076923077e-07,
      "loss": -0.0106,
      "reward": 1.07567298412323,
      "reward_std": 0.19135165214538574,
      "reward_std_all": 0.17519481480121613,
      "rewards-max/forensics_sinkhorn_reward": 0.8580040633678436,
      "rewards-max/forensics_strict_edge_reward": 0.09808346256613731,
      "rewards-max/forensics_tiou_reward": 0.7516637444496155,
      "rewards/forensics_sinkhorn_reward": 0.6814577579498291,
      "rewards/forensics_strict_edge_reward": 0.004465494712349027,
      "rewards/forensics_tiou_reward": 0.3897497355937958,
      "step": 368
    },
    {
      "completion_length": 21.109375,
      "epoch": 1.1321044546850998,
      "grad_norm": 16.90286636352539,
      "kl": 0.3330078125,
      "learning_rate": 7.169230769230768e-07,
      "loss": 0.0328,
      "reward": 1.153347134590149,
      "reward_std": 0.20491256564855576,
      "reward_std_all": 0.21929724514484406,
      "rewards-max/forensics_sinkhorn_reward": 0.9780372381210327,
      "rewards-max/forensics_strict_edge_reward": 0.15495047345757484,
      "rewards-max/forensics_tiou_reward": 0.9612210988998413,
      "rewards/forensics_sinkhorn_reward": 0.6608643531799316,
      "rewards/forensics_strict_edge_reward": 0.013864463893696666,
      "rewards/forensics_tiou_reward": 0.47861823439598083,
      "step": 369
    },
    {
      "completion_length": 21.59375,
      "epoch": 1.1351766513056836,
      "grad_norm": 6.801283836364746,
      "kl": 0.2021484375,
      "learning_rate": 7.161538461538461e-07,
      "loss": 0.0823,
      "reward": 1.1509526371955872,
      "reward_std": 0.2635088264942169,
      "reward_std_all": 0.24332277476787567,
      "rewards-max/forensics_sinkhorn_reward": 0.9672479033470154,
      "rewards-max/forensics_strict_edge_reward": 0.3513744845986366,
      "rewards-max/forensics_tiou_reward": 0.9439926743507385,
      "rewards/forensics_sinkhorn_reward": 0.6653224527835846,
      "rewards/forensics_strict_edge_reward": 0.03444222640246153,
      "rewards/forensics_tiou_reward": 0.4511880427598953,
      "step": 370
    },
    {
      "completion_length": 19.890625,
      "epoch": 1.1382488479262673,
      "grad_norm": 10.445049285888672,
      "kl": 0.220703125,
      "learning_rate": 7.153846153846154e-07,
      "loss": 0.0452,
      "reward": 1.1235309839248657,
      "reward_std": 0.18539678305387497,
      "reward_std_all": 0.18768421560525894,
      "rewards-max/forensics_sinkhorn_reward": 0.9497463703155518,
      "rewards-max/forensics_strict_edge_reward": 0.2097843885421753,
      "rewards-max/forensics_tiou_reward": 0.9029296338558197,
      "rewards/forensics_sinkhorn_reward": 0.6623339354991913,
      "rewards/forensics_strict_edge_reward": 0.016844901721924543,
      "rewards/forensics_tiou_reward": 0.4443521499633789,
      "step": 371
    },
    {
      "completion_length": 22.5625,
      "epoch": 1.141321044546851,
      "grad_norm": 10.556557655334473,
      "kl": 0.1669921875,
      "learning_rate": 7.146153846153846e-07,
      "loss": -0.0305,
      "reward": 1.3142256140708923,
      "reward_std": 0.1963515356183052,
      "reward_std_all": 0.18692617118358612,
      "rewards-max/forensics_sinkhorn_reward": 0.9375019669532776,
      "rewards-max/forensics_strict_edge_reward": 0.6575653851032257,
      "rewards-max/forensics_tiou_reward": 0.8667797744274139,
      "rewards/forensics_sinkhorn_reward": 0.7433057427406311,
      "rewards/forensics_strict_edge_reward": 0.032665468752384186,
      "rewards/forensics_tiou_reward": 0.5382544696331024,
      "step": 372
    },
    {
      "completion_length": 23.296875,
      "epoch": 1.1443932411674347,
      "grad_norm": 9.65697956085205,
      "kl": 0.2236328125,
      "learning_rate": 7.138461538461538e-07,
      "loss": 0.0438,
      "reward": 1.1889402866363525,
      "reward_std": 0.16679013520479202,
      "reward_std_all": 0.16367851197719574,
      "rewards-max/forensics_sinkhorn_reward": 0.9138439297676086,
      "rewards-max/forensics_strict_edge_reward": 0.3130370005965233,
      "rewards-max/forensics_tiou_reward": 0.8167032599449158,
      "rewards/forensics_sinkhorn_reward": 0.6939007937908173,
      "rewards/forensics_strict_edge_reward": 0.02491224091500044,
      "rewards/forensics_tiou_reward": 0.4701271802186966,
      "step": 373
    },
    {
      "completion_length": 21.28125,
      "epoch": 1.1474654377880185,
      "grad_norm": 7.172101974487305,
      "kl": 0.20361328125,
      "learning_rate": 7.13076923076923e-07,
      "loss": -0.0706,
      "reward": 1.177746295928955,
      "reward_std": 0.24439629912376404,
      "reward_std_all": 0.21084599196910858,
      "rewards-max/forensics_sinkhorn_reward": 0.9732622504234314,
      "rewards-max/forensics_strict_edge_reward": 0.24689825624227524,
      "rewards-max/forensics_tiou_reward": 0.9530794024467468,
      "rewards/forensics_sinkhorn_reward": 0.702081024646759,
      "rewards/forensics_strict_edge_reward": 0.027966320514678955,
      "rewards/forensics_tiou_reward": 0.4476989656686783,
      "step": 374
    },
    {
      "completion_length": 21.390625,
      "epoch": 1.1505376344086022,
      "grad_norm": 11.680517196655273,
      "kl": 0.17919921875,
      "learning_rate": 7.123076923076922e-07,
      "loss": 0.0547,
      "reward": 1.116364598274231,
      "reward_std": 0.2598021477460861,
      "reward_std_all": 0.234918013215065,
      "rewards-max/forensics_sinkhorn_reward": 0.9392489790916443,
      "rewards-max/forensics_strict_edge_reward": 0.049225513357669115,
      "rewards-max/forensics_tiou_reward": 0.878202497959137,
      "rewards/forensics_sinkhorn_reward": 0.6683762967586517,
      "rewards/forensics_strict_edge_reward": 0.005904559686314315,
      "rewards/forensics_tiou_reward": 0.44208380579948425,
      "step": 375
    },
    {
      "completion_length": 21.15625,
      "epoch": 1.1536098310291858,
      "grad_norm": 14.13349437713623,
      "kl": 0.20751953125,
      "learning_rate": 7.115384615384616e-07,
      "loss": 0.0076,
      "reward": 1.135426104068756,
      "reward_std": 0.2205529361963272,
      "reward_std_all": 0.19211483746767044,
      "rewards-max/forensics_sinkhorn_reward": 0.9633220136165619,
      "rewards-max/forensics_strict_edge_reward": 0.3517177402973175,
      "rewards-max/forensics_tiou_reward": 0.9374674558639526,
      "rewards/forensics_sinkhorn_reward": 0.6636547446250916,
      "rewards/forensics_strict_edge_reward": 0.018845975399017334,
      "rewards/forensics_tiou_reward": 0.45292529463768005,
      "step": 376
    },
    {
      "completion_length": 19.609375,
      "epoch": 1.1566820276497696,
      "grad_norm": 8.76567268371582,
      "kl": 0.24658203125,
      "learning_rate": 7.107692307692308e-07,
      "loss": 0.019,
      "reward": 1.2527234554290771,
      "reward_std": 0.13636349514126778,
      "reward_std_all": 0.1400478407740593,
      "rewards-max/forensics_sinkhorn_reward": 0.9370767176151276,
      "rewards-max/forensics_strict_edge_reward": 0.45888544619083405,
      "rewards-max/forensics_tiou_reward": 0.8597164452075958,
      "rewards/forensics_sinkhorn_reward": 0.7316352725028992,
      "rewards/forensics_strict_edge_reward": 0.02632926218211651,
      "rewards/forensics_tiou_reward": 0.49475888907909393,
      "step": 377
    },
    {
      "completion_length": 20.71875,
      "epoch": 1.1597542242703534,
      "grad_norm": 10.787839889526367,
      "kl": 0.283203125,
      "learning_rate": 7.1e-07,
      "loss": 0.0568,
      "reward": 1.2967519760131836,
      "reward_std": 0.1993057280778885,
      "reward_std_all": 0.19604333490133286,
      "rewards-max/forensics_sinkhorn_reward": 0.9654929339885712,
      "rewards-max/forensics_strict_edge_reward": 0.41861192137002945,
      "rewards-max/forensics_tiou_reward": 0.937917560338974,
      "rewards/forensics_sinkhorn_reward": 0.7458512783050537,
      "rewards/forensics_strict_edge_reward": 0.039441962726414204,
      "rewards/forensics_tiou_reward": 0.511458694934845,
      "step": 378
    },
    {
      "completion_length": 19.609375,
      "epoch": 1.1628264208909371,
      "grad_norm": 10.237565040588379,
      "kl": 0.1767578125,
      "learning_rate": 7.092307692307692e-07,
      "loss": 0.0583,
      "reward": 1.1872769594192505,
      "reward_std": 0.26716112345457077,
      "reward_std_all": 0.25197382271289825,
      "rewards-max/forensics_sinkhorn_reward": 0.9295462965965271,
      "rewards-max/forensics_strict_edge_reward": 0.5031345209572464,
      "rewards-max/forensics_tiou_reward": 0.8563416600227356,
      "rewards/forensics_sinkhorn_reward": 0.6922597587108612,
      "rewards/forensics_strict_edge_reward": 0.04393184218497481,
      "rewards/forensics_tiou_reward": 0.4510853439569473,
      "step": 379
    },
    {
      "completion_length": 20.171875,
      "epoch": 1.1658986175115207,
      "grad_norm": 8.507623672485352,
      "kl": 0.2958984375,
      "learning_rate": 7.084615384615384e-07,
      "loss": 0.023,
      "reward": 1.2300562262535095,
      "reward_std": 0.17623884975910187,
      "reward_std_all": 0.1654478870332241,
      "rewards-max/forensics_sinkhorn_reward": 0.9710472226142883,
      "rewards-max/forensics_strict_edge_reward": 0.07766272500157356,
      "rewards-max/forensics_tiou_reward": 0.945451408624649,
      "rewards/forensics_sinkhorn_reward": 0.7143512666225433,
      "rewards/forensics_strict_edge_reward": 0.00477514520753175,
      "rewards/forensics_tiou_reward": 0.5109298080205917,
      "step": 380
    },
    {
      "completion_length": 20.953125,
      "epoch": 1.1689708141321045,
      "grad_norm": 8.797562599182129,
      "kl": 0.2451171875,
      "learning_rate": 7.076923076923077e-07,
      "loss": -0.0626,
      "reward": 1.2753131985664368,
      "reward_std": 0.2915848642587662,
      "reward_std_all": 0.2804594337940216,
      "rewards-max/forensics_sinkhorn_reward": 0.9755893349647522,
      "rewards-max/forensics_strict_edge_reward": 0.5415481477975845,
      "rewards-max/forensics_tiou_reward": 0.96352618932724,
      "rewards/forensics_sinkhorn_reward": 0.7259710729122162,
      "rewards/forensics_strict_edge_reward": 0.05063159577548504,
      "rewards/forensics_tiou_reward": 0.498710498213768,
      "step": 381
    },
    {
      "completion_length": 19.921875,
      "epoch": 1.1720430107526882,
      "grad_norm": 9.696013450622559,
      "kl": 0.19482421875,
      "learning_rate": 7.069230769230768e-07,
      "loss": -0.0687,
      "reward": 1.1526829600334167,
      "reward_std": 0.20066602528095245,
      "reward_std_all": 0.18098825216293335,
      "rewards-max/forensics_sinkhorn_reward": 0.9631471931934357,
      "rewards-max/forensics_strict_edge_reward": 0.35388194024562836,
      "rewards-max/forensics_tiou_reward": 0.9342660307884216,
      "rewards/forensics_sinkhorn_reward": 0.6939344108104706,
      "rewards/forensics_strict_edge_reward": 0.031053979881107807,
      "rewards/forensics_tiou_reward": 0.42769448459148407,
      "step": 382
    },
    {
      "completion_length": 20.125,
      "epoch": 1.1751152073732718,
      "grad_norm": 9.243277549743652,
      "kl": 0.2041015625,
      "learning_rate": 7.061538461538462e-07,
      "loss": 0.015,
      "reward": 1.2405118346214294,
      "reward_std": 0.2205675169825554,
      "reward_std_all": 0.250449076294899,
      "rewards-max/forensics_sinkhorn_reward": 0.9280705451965332,
      "rewards-max/forensics_strict_edge_reward": 0.5719383582472801,
      "rewards-max/forensics_tiou_reward": 0.8532935678958893,
      "rewards/forensics_sinkhorn_reward": 0.7180555760860443,
      "rewards/forensics_strict_edge_reward": 0.05709675420075655,
      "rewards/forensics_tiou_reward": 0.4653594493865967,
      "step": 383
    },
    {
      "completion_length": 21.703125,
      "epoch": 1.1781874039938556,
      "grad_norm": 12.346782684326172,
      "kl": 1.009765625,
      "learning_rate": 7.053846153846154e-07,
      "loss": 0.0868,
      "reward": 1.331617832183838,
      "reward_std": 0.2215939611196518,
      "reward_std_all": 0.22183366864919662,
      "rewards-max/forensics_sinkhorn_reward": 0.9617349207401276,
      "rewards-max/forensics_strict_edge_reward": 0.3134205862879753,
      "rewards-max/forensics_tiou_reward": 0.9289945662021637,
      "rewards/forensics_sinkhorn_reward": 0.7593128085136414,
      "rewards/forensics_strict_edge_reward": 0.02350472193211317,
      "rewards/forensics_tiou_reward": 0.5488002896308899,
      "step": 384
    },
    {
      "completion_length": 19.65625,
      "epoch": 1.1812596006144394,
      "grad_norm": 12.036484718322754,
      "kl": 0.3154296875,
      "learning_rate": 7.046153846153846e-07,
      "loss": 0.0422,
      "reward": 1.2099780440330505,
      "reward_std": 0.22038312256336212,
      "reward_std_all": 0.21611743420362473,
      "rewards-max/forensics_sinkhorn_reward": 0.9393330514431,
      "rewards-max/forensics_strict_edge_reward": 0.3437973503023386,
      "rewards-max/forensics_tiou_reward": 0.8696674406528473,
      "rewards/forensics_sinkhorn_reward": 0.7130398452281952,
      "rewards/forensics_strict_edge_reward": 0.04255822277627885,
      "rewards/forensics_tiou_reward": 0.45438000559806824,
      "step": 385
    },
    {
      "completion_length": 19.953125,
      "epoch": 1.1843317972350231,
      "grad_norm": 11.358874320983887,
      "kl": 0.15625,
      "learning_rate": 7.038461538461539e-07,
      "loss": -0.0236,
      "reward": 1.163735032081604,
      "reward_std": 0.1765778586268425,
      "reward_std_all": 0.19042464345693588,
      "rewards-max/forensics_sinkhorn_reward": 0.9566828608512878,
      "rewards-max/forensics_strict_edge_reward": 0.31179623305797577,
      "rewards-max/forensics_tiou_reward": 0.9209908246994019,
      "rewards/forensics_sinkhorn_reward": 0.7095431685447693,
      "rewards/forensics_strict_edge_reward": 0.01845630072057247,
      "rewards/forensics_tiou_reward": 0.43573547899723053,
      "step": 386
    },
    {
      "completion_length": 19.65625,
      "epoch": 1.1874039938556067,
      "grad_norm": 13.641672134399414,
      "kl": 0.21484375,
      "learning_rate": 7.03076923076923e-07,
      "loss": 0.1596,
      "reward": 1.1912768483161926,
      "reward_std": 0.18315812945365906,
      "reward_std_all": 0.1907094419002533,
      "rewards-max/forensics_sinkhorn_reward": 0.9475845098495483,
      "rewards-max/forensics_strict_edge_reward": 0.07532307133078575,
      "rewards-max/forensics_tiou_reward": 0.8870872557163239,
      "rewards/forensics_sinkhorn_reward": 0.736714631319046,
      "rewards/forensics_strict_edge_reward": 0.003413926810026169,
      "rewards/forensics_tiou_reward": 0.45114825665950775,
      "step": 387
    },
    {
      "completion_length": 21.328125,
      "epoch": 1.1904761904761905,
      "grad_norm": 8.423064231872559,
      "kl": 0.255859375,
      "learning_rate": 7.023076923076922e-07,
      "loss": 0.0221,
      "reward": 1.3627700209617615,
      "reward_std": 0.27809569239616394,
      "reward_std_all": 0.24725179374217987,
      "rewards-max/forensics_sinkhorn_reward": 0.9884018898010254,
      "rewards-max/forensics_strict_edge_reward": 0.9232408702373505,
      "rewards-max/forensics_tiou_reward": 0.9978390336036682,
      "rewards/forensics_sinkhorn_reward": 0.7648700773715973,
      "rewards/forensics_strict_edge_reward": 0.06025143153965473,
      "rewards/forensics_tiou_reward": 0.5376485139131546,
      "step": 388
    },
    {
      "completion_length": 20.9375,
      "epoch": 1.1935483870967742,
      "grad_norm": 9.380431175231934,
      "kl": 0.2080078125,
      "learning_rate": 7.015384615384615e-07,
      "loss": -0.0941,
      "reward": 1.150484561920166,
      "reward_std": 0.17784685641527176,
      "reward_std_all": 0.198469877243042,
      "rewards-max/forensics_sinkhorn_reward": 0.9503861665725708,
      "rewards-max/forensics_strict_edge_reward": 0.34755249321460724,
      "rewards-max/forensics_tiou_reward": 0.904024749994278,
      "rewards/forensics_sinkhorn_reward": 0.6849683225154877,
      "rewards/forensics_strict_edge_reward": 0.020005049649626017,
      "rewards/forensics_tiou_reward": 0.44551125168800354,
      "step": 389
    },
    {
      "completion_length": 20.75,
      "epoch": 1.1966205837173578,
      "grad_norm": 10.667387008666992,
      "kl": 0.21484375,
      "learning_rate": 7.007692307692308e-07,
      "loss": -0.0294,
      "reward": 1.189446210861206,
      "reward_std": 0.1926208883523941,
      "reward_std_all": 0.16689811646938324,
      "rewards-max/forensics_sinkhorn_reward": 0.9432705640792847,
      "rewards-max/forensics_strict_edge_reward": 0.15063884109258652,
      "rewards-max/forensics_tiou_reward": 0.8914822638034821,
      "rewards/forensics_sinkhorn_reward": 0.7206553518772125,
      "rewards/forensics_strict_edge_reward": 0.014126773457974195,
      "rewards/forensics_tiou_reward": 0.45466409623622894,
      "step": 390
    },
    {
      "completion_length": 21.484375,
      "epoch": 1.1996927803379416,
      "grad_norm": 8.221112251281738,
      "kl": 0.1787109375,
      "learning_rate": 7e-07,
      "loss": 0.0442,
      "reward": 1.1169018149375916,
      "reward_std": 0.16345009207725525,
      "reward_std_all": 0.16002189368009567,
      "rewards-max/forensics_sinkhorn_reward": 0.9241827428340912,
      "rewards-max/forensics_strict_edge_reward": 0.12023795396089554,
      "rewards-max/forensics_tiou_reward": 0.8488498330116272,
      "rewards/forensics_sinkhorn_reward": 0.6817402243614197,
      "rewards/forensics_strict_edge_reward": 0.009691189508885145,
      "rewards/forensics_tiou_reward": 0.42547033727169037,
      "step": 391
    },
    {
      "completion_length": 23.109375,
      "epoch": 1.2027649769585254,
      "grad_norm": 9.523177146911621,
      "kl": 0.19970703125,
      "learning_rate": 6.992307692307692e-07,
      "loss": 0.0346,
      "reward": 1.2150533199310303,
      "reward_std": 0.23692362755537033,
      "reward_std_all": 0.22149693965911865,
      "rewards-max/forensics_sinkhorn_reward": 0.9452805519104004,
      "rewards-max/forensics_strict_edge_reward": 0.2120999328326434,
      "rewards-max/forensics_tiou_reward": 0.8863604664802551,
      "rewards/forensics_sinkhorn_reward": 0.7249749004840851,
      "rewards/forensics_strict_edge_reward": 0.01700897014234215,
      "rewards/forensics_tiou_reward": 0.4730694442987442,
      "step": 392
    },
    {
      "completion_length": 22.90625,
      "epoch": 1.2058371735791091,
      "grad_norm": 7.583122253417969,
      "kl": 0.23291015625,
      "learning_rate": 6.984615384615384e-07,
      "loss": 0.0191,
      "reward": 1.1056694388389587,
      "reward_std": 0.1955263316631317,
      "reward_std_all": 0.17306405305862427,
      "rewards-max/forensics_sinkhorn_reward": 0.9754997193813324,
      "rewards-max/forensics_strict_edge_reward": 0.5521267205476761,
      "rewards-max/forensics_tiou_reward": 0.9620121419429779,
      "rewards/forensics_sinkhorn_reward": 0.6482212543487549,
      "rewards/forensics_strict_edge_reward": 0.04165230132639408,
      "rewards/forensics_tiou_reward": 0.4157959520816803,
      "step": 393
    },
    {
      "completion_length": 19.5625,
      "epoch": 1.2089093701996927,
      "grad_norm": 7.9867706298828125,
      "kl": 0.369140625,
      "learning_rate": 6.976923076923077e-07,
      "loss": 0.0945,
      "reward": 1.2940419912338257,
      "reward_std": 0.25847285985946655,
      "reward_std_all": 0.23931845277547836,
      "rewards-max/forensics_sinkhorn_reward": 0.959970086812973,
      "rewards-max/forensics_strict_edge_reward": 0.41550540924072266,
      "rewards-max/forensics_tiou_reward": 0.933404415845871,
      "rewards/forensics_sinkhorn_reward": 0.7380395531654358,
      "rewards/forensics_strict_edge_reward": 0.040196485817432404,
      "rewards/forensics_tiou_reward": 0.5158058851957321,
      "step": 394
    },
    {
      "completion_length": 21.0625,
      "epoch": 1.2119815668202765,
      "grad_norm": 11.095050811767578,
      "kl": 0.28564453125,
      "learning_rate": 6.969230769230769e-07,
      "loss": 0.0114,
      "reward": 1.1711182594299316,
      "reward_std": 0.2004593014717102,
      "reward_std_all": 0.18232667073607445,
      "rewards-max/forensics_sinkhorn_reward": 0.9680221378803253,
      "rewards-max/forensics_strict_edge_reward": 0.5165956895798445,
      "rewards-max/forensics_tiou_reward": 0.9405800402164459,
      "rewards/forensics_sinkhorn_reward": 0.708706796169281,
      "rewards/forensics_strict_edge_reward": 0.023739506490528584,
      "rewards/forensics_tiou_reward": 0.4386718422174454,
      "step": 395
    },
    {
      "completion_length": 21.75,
      "epoch": 1.2150537634408602,
      "grad_norm": 8.772191047668457,
      "kl": 0.3154296875,
      "learning_rate": 6.961538461538461e-07,
      "loss": 0.104,
      "reward": 1.1141047477722168,
      "reward_std": 0.18079252541065216,
      "reward_std_all": 0.16714022308588028,
      "rewards-max/forensics_sinkhorn_reward": 0.9300813972949982,
      "rewards-max/forensics_strict_edge_reward": 0.15472418069839478,
      "rewards-max/forensics_tiou_reward": 0.8588034212589264,
      "rewards/forensics_sinkhorn_reward": 0.6894792914390564,
      "rewards/forensics_strict_edge_reward": 0.0145826512016356,
      "rewards/forensics_tiou_reward": 0.4100428819656372,
      "step": 396
    },
    {
      "completion_length": 22.203125,
      "epoch": 1.218125960061444,
      "grad_norm": 16.429359436035156,
      "kl": 0.44970703125,
      "learning_rate": 6.953846153846154e-07,
      "loss": 0.0265,
      "reward": 1.2333248257637024,
      "reward_std": 0.17787779867649078,
      "reward_std_all": 0.1700209379196167,
      "rewards-max/forensics_sinkhorn_reward": 0.9200001955032349,
      "rewards-max/forensics_strict_edge_reward": 0.2420913651585579,
      "rewards-max/forensics_tiou_reward": 0.8222715556621552,
      "rewards/forensics_sinkhorn_reward": 0.7406787276268005,
      "rewards/forensics_strict_edge_reward": 0.01413867436349392,
      "rewards/forensics_tiou_reward": 0.4785074293613434,
      "step": 397
    },
    {
      "completion_length": 20.171875,
      "epoch": 1.2211981566820276,
      "grad_norm": 12.412775993347168,
      "kl": 0.3564453125,
      "learning_rate": 6.946153846153846e-07,
      "loss": -0.0239,
      "reward": 1.2738868594169617,
      "reward_std": 0.25647401809692383,
      "reward_std_all": 0.26022104918956757,
      "rewards-max/forensics_sinkhorn_reward": 0.9804361462593079,
      "rewards-max/forensics_strict_edge_reward": 0.37312997691333294,
      "rewards-max/forensics_tiou_reward": 0.9737971425056458,
      "rewards/forensics_sinkhorn_reward": 0.7179957330226898,
      "rewards/forensics_strict_edge_reward": 0.03594451071694493,
      "rewards/forensics_tiou_reward": 0.5199465751647949,
      "step": 398
    },
    {
      "completion_length": 20.65625,
      "epoch": 1.2242703533026114,
      "grad_norm": 10.9586763381958,
      "kl": 0.197265625,
      "learning_rate": 6.938461538461539e-07,
      "loss": 0.0065,
      "reward": 1.2440215349197388,
      "reward_std": 0.22268151491880417,
      "reward_std_all": 0.20666122436523438,
      "rewards-max/forensics_sinkhorn_reward": 0.9617433249950409,
      "rewards-max/forensics_strict_edge_reward": 0.45888544619083405,
      "rewards-max/forensics_tiou_reward": 0.9335775077342987,
      "rewards/forensics_sinkhorn_reward": 0.7031123042106628,
      "rewards/forensics_strict_edge_reward": 0.04562980681657791,
      "rewards/forensics_tiou_reward": 0.4952794909477234,
      "step": 399
    },
    {
      "completion_length": 24.515625,
      "epoch": 1.2273425499231951,
      "grad_norm": 8.536907196044922,
      "kl": 0.25634765625,
      "learning_rate": 6.93076923076923e-07,
      "loss": 0.0755,
      "reward": 1.3619517087936401,
      "reward_std": 0.23167936503887177,
      "reward_std_all": 0.21414105594158173,
      "rewards-max/forensics_sinkhorn_reward": 0.9644202291965485,
      "rewards-max/forensics_strict_edge_reward": 0.2832633974030614,
      "rewards-max/forensics_tiou_reward": 0.9296739995479584,
      "rewards/forensics_sinkhorn_reward": 0.7575878500938416,
      "rewards/forensics_strict_edge_reward": 0.036443007295019925,
      "rewards/forensics_tiou_reward": 0.5679208636283875,
      "step": 400
    },
    {
      "completion_length": 23.1875,
      "epoch": 1.230414746543779,
      "grad_norm": 12.73568058013916,
      "kl": 0.21630859375,
      "learning_rate": 6.923076923076922e-07,
      "loss": 0.0148,
      "reward": 1.2771884202957153,
      "reward_std": 0.20604223012924194,
      "reward_std_all": 0.21423035115003586,
      "rewards-max/forensics_sinkhorn_reward": 0.973051905632019,
      "rewards-max/forensics_strict_edge_reward": 0.2144480124115944,
      "rewards-max/forensics_tiou_reward": 0.952835738658905,
      "rewards/forensics_sinkhorn_reward": 0.7121952176094055,
      "rewards/forensics_strict_edge_reward": 0.02324589528143406,
      "rewards/forensics_tiou_reward": 0.5417471528053284,
      "step": 401
    },
    {
      "completion_length": 22.484375,
      "epoch": 1.2334869431643625,
      "grad_norm": 8.655784606933594,
      "kl": 0.22314453125,
      "learning_rate": 6.915384615384615e-07,
      "loss": 0.022,
      "reward": 1.1371884047985077,
      "reward_std": 0.28536272048950195,
      "reward_std_all": 0.27563872933387756,
      "rewards-max/forensics_sinkhorn_reward": 0.956423431634903,
      "rewards-max/forensics_strict_edge_reward": 0.3652799017727375,
      "rewards-max/forensics_tiou_reward": 0.9245871603488922,
      "rewards/forensics_sinkhorn_reward": 0.6688977479934692,
      "rewards/forensics_strict_edge_reward": 0.020616266352590173,
      "rewards/forensics_tiou_reward": 0.44767436385154724,
      "step": 402
    },
    {
      "completion_length": 22.421875,
      "epoch": 1.2365591397849462,
      "grad_norm": 19.611814498901367,
      "kl": 0.251953125,
      "learning_rate": 6.907692307692307e-07,
      "loss": 0.0072,
      "reward": 1.130616009235382,
      "reward_std": 0.17153601348400116,
      "reward_std_all": 0.1710723340511322,
      "rewards-max/forensics_sinkhorn_reward": 0.9429240822792053,
      "rewards-max/forensics_strict_edge_reward": 0.20574250631034374,
      "rewards-max/forensics_tiou_reward": 0.8939144313335419,
      "rewards/forensics_sinkhorn_reward": 0.6914944052696228,
      "rewards/forensics_strict_edge_reward": 0.013420168543234468,
      "rewards/forensics_tiou_reward": 0.425701379776001,
      "step": 403
    },
    {
      "completion_length": 21.546875,
      "epoch": 1.23963133640553,
      "grad_norm": 11.294374465942383,
      "kl": 0.59521484375,
      "learning_rate": 6.9e-07,
      "loss": 0.0522,
      "reward": 1.227684736251831,
      "reward_std": 0.23842941224575043,
      "reward_std_all": 0.2188156694173813,
      "rewards-max/forensics_sinkhorn_reward": 0.9593942761421204,
      "rewards-max/forensics_strict_edge_reward": 0.2969970703125,
      "rewards-max/forensics_tiou_reward": 0.9307029247283936,
      "rewards/forensics_sinkhorn_reward": 0.705746591091156,
      "rewards/forensics_strict_edge_reward": 0.0334056606516242,
      "rewards/forensics_tiou_reward": 0.48853252828121185,
      "step": 404
    },
    {
      "completion_length": 22.9375,
      "epoch": 1.2427035330261136,
      "grad_norm": 11.397086143493652,
      "kl": 0.32470703125,
      "learning_rate": 6.892307692307692e-07,
      "loss": -0.1115,
      "reward": 1.2860270738601685,
      "reward_std": 0.16358908265829086,
      "reward_std_all": 0.1638699620962143,
      "rewards-max/forensics_sinkhorn_reward": 0.9634639918804169,
      "rewards-max/forensics_strict_edge_reward": 0.225993312895298,
      "rewards-max/forensics_tiou_reward": 0.9358038902282715,
      "rewards/forensics_sinkhorn_reward": 0.7376992702484131,
      "rewards/forensics_strict_edge_reward": 0.026436379179358482,
      "rewards/forensics_tiou_reward": 0.5218914449214935,
      "step": 405
    },
    {
      "completion_length": 20.375,
      "epoch": 1.2457757296466974,
      "grad_norm": 8.292098999023438,
      "kl": 0.18212890625,
      "learning_rate": 6.884615384615384e-07,
      "loss": 0.096,
      "reward": 1.2637922763824463,
      "reward_std": 0.1860973834991455,
      "reward_std_all": 0.16456767171621323,
      "rewards-max/forensics_sinkhorn_reward": 0.9715120494365692,
      "rewards-max/forensics_strict_edge_reward": 0.1930476324632764,
      "rewards-max/forensics_tiou_reward": 0.9492076933383942,
      "rewards/forensics_sinkhorn_reward": 0.7415197491645813,
      "rewards/forensics_strict_edge_reward": 0.017947725486010313,
      "rewards/forensics_tiou_reward": 0.5043248385190964,
      "step": 406
    },
    {
      "completion_length": 21.4375,
      "epoch": 1.2488479262672811,
      "grad_norm": 9.99050521850586,
      "kl": 0.35693359375,
      "learning_rate": 6.876923076923077e-07,
      "loss": 0.0207,
      "reward": 1.2067744731903076,
      "reward_std": 0.22512978315353394,
      "reward_std_all": 0.20869093388319016,
      "rewards-max/forensics_sinkhorn_reward": 0.937962144613266,
      "rewards-max/forensics_strict_edge_reward": 0.30881428718566895,
      "rewards-max/forensics_tiou_reward": 0.8564966320991516,
      "rewards/forensics_sinkhorn_reward": 0.7130023539066315,
      "rewards/forensics_strict_edge_reward": 0.027025052811950445,
      "rewards/forensics_tiou_reward": 0.46674707531929016,
      "step": 407
    },
    {
      "completion_length": 20.828125,
      "epoch": 1.251920122887865,
      "grad_norm": 10.53099250793457,
      "kl": 0.21728515625,
      "learning_rate": 6.869230769230769e-07,
      "loss": 0.109,
      "reward": 1.1479913592338562,
      "reward_std": 0.2560461461544037,
      "reward_std_all": 0.22887635231018066,
      "rewards-max/forensics_sinkhorn_reward": 0.9566320776939392,
      "rewards-max/forensics_strict_edge_reward": 0.09727882780134678,
      "rewards-max/forensics_tiou_reward": 0.9132861196994781,
      "rewards/forensics_sinkhorn_reward": 0.6825208365917206,
      "rewards/forensics_strict_edge_reward": 0.015057077631354332,
      "rewards/forensics_tiou_reward": 0.45041339099407196,
      "step": 408
    },
    {
      "completion_length": 19.640625,
      "epoch": 1.2549923195084485,
      "grad_norm": 8.662386894226074,
      "kl": 0.37451171875,
      "learning_rate": 6.861538461538461e-07,
      "loss": 0.0278,
      "reward": 1.2221991419792175,
      "reward_std": 0.14212510362267494,
      "reward_std_all": 0.16311542689800262,
      "rewards-max/forensics_sinkhorn_reward": 0.9381651878356934,
      "rewards-max/forensics_strict_edge_reward": 0.23242570832371712,
      "rewards-max/forensics_tiou_reward": 0.8623563349246979,
      "rewards/forensics_sinkhorn_reward": 0.7393933236598969,
      "rewards/forensics_strict_edge_reward": 0.01709756231866777,
      "rewards/forensics_tiou_reward": 0.46570827066898346,
      "step": 409
    },
    {
      "completion_length": 19.953125,
      "epoch": 1.2580645161290323,
      "grad_norm": 10.198529243469238,
      "kl": 0.2294921875,
      "learning_rate": 6.853846153846153e-07,
      "loss": -0.0164,
      "reward": 1.204779863357544,
      "reward_std": 0.22982942312955856,
      "reward_std_all": 0.20251988619565964,
      "rewards-max/forensics_sinkhorn_reward": 0.9742830693721771,
      "rewards-max/forensics_strict_edge_reward": 0.5918685793876648,
      "rewards-max/forensics_tiou_reward": 0.9601510465145111,
      "rewards/forensics_sinkhorn_reward": 0.7015671730041504,
      "rewards/forensics_strict_edge_reward": 0.050934089347720146,
      "rewards/forensics_tiou_reward": 0.45227861404418945,
      "step": 410
    },
    {
      "completion_length": 23.078125,
      "epoch": 1.261136712749616,
      "grad_norm": 11.944246292114258,
      "kl": 0.1728515625,
      "learning_rate": 6.846153846153846e-07,
      "loss": 0.036,
      "reward": 1.2050774097442627,
      "reward_std": 0.17181334644556046,
      "reward_std_all": 0.1462271735072136,
      "rewards-max/forensics_sinkhorn_reward": 0.969172865152359,
      "rewards-max/forensics_strict_edge_reward": 0.14612476155161858,
      "rewards-max/forensics_tiou_reward": 0.9461847841739655,
      "rewards/forensics_sinkhorn_reward": 0.7136013507843018,
      "rewards/forensics_strict_edge_reward": 0.009329559747129679,
      "rewards/forensics_tiou_reward": 0.48214656114578247,
      "step": 411
    },
    {
      "completion_length": 22.171875,
      "epoch": 1.2642089093701996,
      "grad_norm": 11.820347785949707,
      "kl": 0.2548828125,
      "learning_rate": 6.838461538461539e-07,
      "loss": 0.0277,
      "reward": 1.2278298735618591,
      "reward_std": 0.22981609404087067,
      "reward_std_all": 0.21150989085435867,
      "rewards-max/forensics_sinkhorn_reward": 0.9587503671646118,
      "rewards-max/forensics_strict_edge_reward": 0.218681450933218,
      "rewards-max/forensics_tiou_reward": 0.9239187836647034,
      "rewards/forensics_sinkhorn_reward": 0.7258391380310059,
      "rewards/forensics_strict_edge_reward": 0.015131466090679169,
      "rewards/forensics_tiou_reward": 0.4868592619895935,
      "step": 412
    },
    {
      "completion_length": 22.234375,
      "epoch": 1.2672811059907834,
      "grad_norm": 13.942300796508789,
      "kl": 0.2734375,
      "learning_rate": 6.83076923076923e-07,
      "loss": -0.0266,
      "reward": 1.162227749824524,
      "reward_std": 0.3170166015625,
      "reward_std_all": 0.2823241651058197,
      "rewards-max/forensics_sinkhorn_reward": 0.9687222838401794,
      "rewards-max/forensics_strict_edge_reward": 0.4346180260181427,
      "rewards-max/forensics_tiou_reward": 0.9463821947574615,
      "rewards/forensics_sinkhorn_reward": 0.6884300112724304,
      "rewards/forensics_strict_edge_reward": 0.036943888291716576,
      "rewards/forensics_tiou_reward": 0.4368537664413452,
      "step": 413
    },
    {
      "completion_length": 21.40625,
      "epoch": 1.2703533026113671,
      "grad_norm": 11.804401397705078,
      "kl": 0.2373046875,
      "learning_rate": 6.823076923076922e-07,
      "loss": 0.0787,
      "reward": 1.1494269967079163,
      "reward_std": 0.19994744658470154,
      "reward_std_all": 0.18229975551366806,
      "rewards-max/forensics_sinkhorn_reward": 0.9402866363525391,
      "rewards-max/forensics_strict_edge_reward": 0.15963750332593918,
      "rewards-max/forensics_tiou_reward": 0.8858685195446014,
      "rewards/forensics_sinkhorn_reward": 0.6910861432552338,
      "rewards/forensics_strict_edge_reward": 0.008786556776612997,
      "rewards/forensics_tiou_reward": 0.44955427944660187,
      "step": 414
    },
    {
      "completion_length": 23.578125,
      "epoch": 1.273425499231951,
      "grad_norm": 10.00080680847168,
      "kl": 0.27001953125,
      "learning_rate": 6.815384615384615e-07,
      "loss": -0.0322,
      "reward": 1.2663779258728027,
      "reward_std": 0.21401873230934143,
      "reward_std_all": 0.19114899635314941,
      "rewards-max/forensics_sinkhorn_reward": 0.9628907144069672,
      "rewards-max/forensics_strict_edge_reward": 0.17530392855405807,
      "rewards-max/forensics_tiou_reward": 0.929167777299881,
      "rewards/forensics_sinkhorn_reward": 0.7120425403118134,
      "rewards/forensics_strict_edge_reward": 0.016989171039313078,
      "rewards/forensics_tiou_reward": 0.5373462438583374,
      "step": 415
    },
    {
      "completion_length": 22.140625,
      "epoch": 1.2764976958525347,
      "grad_norm": 8.449216842651367,
      "kl": 0.32421875,
      "learning_rate": 6.807692307692307e-07,
      "loss": 0.0442,
      "reward": 1.1156880855560303,
      "reward_std": 0.19710586220026016,
      "reward_std_all": 0.20695611089468002,
      "rewards-max/forensics_sinkhorn_reward": 0.9316717088222504,
      "rewards-max/forensics_strict_edge_reward": 0.07142229564487934,
      "rewards-max/forensics_tiou_reward": 0.8589804470539093,
      "rewards/forensics_sinkhorn_reward": 0.6714290082454681,
      "rewards/forensics_strict_edge_reward": 0.004147325875237584,
      "rewards/forensics_tiou_reward": 0.4401116371154785,
      "step": 416
    },
    {
      "completion_length": 21.0625,
      "epoch": 1.2795698924731183,
      "grad_norm": 12.065027236938477,
      "kl": 0.32470703125,
      "learning_rate": 6.800000000000001e-07,
      "loss": -0.0033,
      "reward": 0.9812802672386169,
      "reward_std": 0.18921086192131042,
      "reward_std_all": 0.17122907936573029,
      "rewards-max/forensics_sinkhorn_reward": 0.8897137343883514,
      "rewards-max/forensics_strict_edge_reward": 0.02287689922377467,
      "rewards-max/forensics_tiou_reward": 0.7573220431804657,
      "rewards/forensics_sinkhorn_reward": 0.6479766368865967,
      "rewards/forensics_strict_edge_reward": 0.0014488348897430114,
      "rewards/forensics_tiou_reward": 0.33185479044914246,
      "step": 417
    },
    {
      "completion_length": 21.734375,
      "epoch": 1.282642089093702,
      "grad_norm": 8.3430757522583,
      "kl": 0.416015625,
      "learning_rate": 6.792307692307692e-07,
      "loss": 0.1218,
      "reward": 1.2548263669013977,
      "reward_std": 0.2181776911020279,
      "reward_std_all": 0.2229590117931366,
      "rewards-max/forensics_sinkhorn_reward": 0.9355756640434265,
      "rewards-max/forensics_strict_edge_reward": 0.026843760133488104,
      "rewards-max/forensics_tiou_reward": 0.861383467912674,
      "rewards/forensics_sinkhorn_reward": 0.7309249937534332,
      "rewards/forensics_strict_edge_reward": 0.003971060068579391,
      "rewards/forensics_tiou_reward": 0.51993028819561,
      "step": 418
    },
    {
      "completion_length": 20.296875,
      "epoch": 1.2857142857142856,
      "grad_norm": 9.547204971313477,
      "kl": 0.2900390625,
      "learning_rate": 6.784615384615384e-07,
      "loss": 0.0104,
      "reward": 1.4062389731407166,
      "reward_std": 0.20600059628486633,
      "reward_std_all": 0.18600666522979736,
      "rewards-max/forensics_sinkhorn_reward": 0.9627137184143066,
      "rewards-max/forensics_strict_edge_reward": 0.4365939050912857,
      "rewards-max/forensics_tiou_reward": 0.9330098330974579,
      "rewards/forensics_sinkhorn_reward": 0.7861883640289307,
      "rewards/forensics_strict_edge_reward": 0.05188343487679958,
      "rewards/forensics_tiou_reward": 0.5681671500205994,
      "step": 419
    },
    {
      "completion_length": 23.25,
      "epoch": 1.2887864823348694,
      "grad_norm": 7.436178684234619,
      "kl": 0.212890625,
      "learning_rate": 6.776923076923077e-07,
      "loss": 0.0798,
      "reward": 1.1436690092086792,
      "reward_std": 0.18538271635770798,
      "reward_std_all": 0.18597562611103058,
      "rewards-max/forensics_sinkhorn_reward": 0.9509209096431732,
      "rewards-max/forensics_strict_edge_reward": 0.2068582996726036,
      "rewards-max/forensics_tiou_reward": 0.8943358659744263,
      "rewards/forensics_sinkhorn_reward": 0.6704963147640228,
      "rewards/forensics_strict_edge_reward": 0.02421504817903042,
      "rewards/forensics_tiou_reward": 0.448957622051239,
      "step": 420
    },
    {
      "completion_length": 23.0625,
      "epoch": 1.2918586789554531,
      "grad_norm": 13.743010520935059,
      "kl": 0.2734375,
      "learning_rate": 6.769230769230769e-07,
      "loss": -0.0473,
      "reward": 1.2250585556030273,
      "reward_std": 0.33265385776758194,
      "reward_std_all": 0.2956592068076134,
      "rewards-max/forensics_sinkhorn_reward": 0.9610419273376465,
      "rewards-max/forensics_strict_edge_reward": 0.293709484860301,
      "rewards-max/forensics_tiou_reward": 0.9252871572971344,
      "rewards/forensics_sinkhorn_reward": 0.7035070955753326,
      "rewards/forensics_strict_edge_reward": 0.02645289374049753,
      "rewards/forensics_tiou_reward": 0.49509865045547485,
      "step": 421
    },
    {
      "completion_length": 22.640625,
      "epoch": 1.294930875576037,
      "grad_norm": 7.4582414627075195,
      "kl": 0.3935546875,
      "learning_rate": 6.761538461538461e-07,
      "loss": 0.0188,
      "reward": 1.283388078212738,
      "reward_std": 0.2067306488752365,
      "reward_std_all": 0.20099925994873047,
      "rewards-max/forensics_sinkhorn_reward": 0.96671923995018,
      "rewards-max/forensics_strict_edge_reward": 0.48574335873126984,
      "rewards-max/forensics_tiou_reward": 0.944866806268692,
      "rewards/forensics_sinkhorn_reward": 0.7397390902042389,
      "rewards/forensics_strict_edge_reward": 0.05238964594900608,
      "rewards/forensics_tiou_reward": 0.4912593513727188,
      "step": 422
    },
    {
      "completion_length": 22.65625,
      "epoch": 1.2980030721966207,
      "grad_norm": 9.97001838684082,
      "kl": 0.26171875,
      "learning_rate": 6.753846153846153e-07,
      "loss": -0.0622,
      "reward": 1.4296217560768127,
      "reward_std": 0.31790710985660553,
      "reward_std_all": 0.32576093077659607,
      "rewards-max/forensics_sinkhorn_reward": 0.9804074764251709,
      "rewards-max/forensics_strict_edge_reward": 0.8260191679000854,
      "rewards-max/forensics_tiou_reward": 0.9801734685897827,
      "rewards/forensics_sinkhorn_reward": 0.7748753130435944,
      "rewards/forensics_strict_edge_reward": 0.06067730858922005,
      "rewards/forensics_tiou_reward": 0.5940691232681274,
      "step": 423
    },
    {
      "completion_length": 21.984375,
      "epoch": 1.3010752688172043,
      "grad_norm": 9.637837409973145,
      "kl": 0.2470703125,
      "learning_rate": 6.746153846153845e-07,
      "loss": -0.0111,
      "reward": 1.283283293247223,
      "reward_std": 0.18446269258856773,
      "reward_std_all": 0.19676226004958153,
      "rewards-max/forensics_sinkhorn_reward": 0.9702883064746857,
      "rewards-max/forensics_strict_edge_reward": 0.2117384374141693,
      "rewards-max/forensics_tiou_reward": 0.9477485716342926,
      "rewards/forensics_sinkhorn_reward": 0.7402991652488708,
      "rewards/forensics_strict_edge_reward": 0.025198866613209248,
      "rewards/forensics_tiou_reward": 0.5177853107452393,
      "step": 424
    },
    {
      "completion_length": 20.28125,
      "epoch": 1.304147465437788,
      "grad_norm": 10.713502883911133,
      "kl": 0.369140625,
      "learning_rate": 6.738461538461539e-07,
      "loss": -0.0364,
      "reward": 1.4321873784065247,
      "reward_std": 0.19492889940738678,
      "reward_std_all": 0.21082764118909836,
      "rewards-max/forensics_sinkhorn_reward": 0.9424369037151337,
      "rewards-max/forensics_strict_edge_reward": 0.7753324806690216,
      "rewards-max/forensics_tiou_reward": 0.885297566652298,
      "rewards/forensics_sinkhorn_reward": 0.7756258845329285,
      "rewards/forensics_strict_edge_reward": 0.08646145462989807,
      "rewards/forensics_tiou_reward": 0.570099949836731,
      "step": 425
    },
    {
      "completion_length": 21.453125,
      "epoch": 1.3072196620583718,
      "grad_norm": 11.837827682495117,
      "kl": 0.3828125,
      "learning_rate": 6.730769230769231e-07,
      "loss": 0.037,
      "reward": 1.3147621154785156,
      "reward_std": 0.2989480048418045,
      "reward_std_all": 0.27608343958854675,
      "rewards-max/forensics_sinkhorn_reward": 0.9693508446216583,
      "rewards-max/forensics_strict_edge_reward": 0.36869725584983826,
      "rewards-max/forensics_tiou_reward": 0.9469853639602661,
      "rewards/forensics_sinkhorn_reward": 0.7559563517570496,
      "rewards/forensics_strict_edge_reward": 0.0293018976226449,
      "rewards/forensics_tiou_reward": 0.5295038819313049,
      "step": 426
    },
    {
      "completion_length": 24.234375,
      "epoch": 1.3102918586789554,
      "grad_norm": 7.722384452819824,
      "kl": 0.3203125,
      "learning_rate": 6.723076923076922e-07,
      "loss": 0.0789,
      "reward": 1.3042048215866089,
      "reward_std": 0.3565337657928467,
      "reward_std_all": 0.3078286498785019,
      "rewards-max/forensics_sinkhorn_reward": 0.9749087393283844,
      "rewards-max/forensics_strict_edge_reward": 0.1729524084366858,
      "rewards-max/forensics_tiou_reward": 0.9554308652877808,
      "rewards/forensics_sinkhorn_reward": 0.7346315979957581,
      "rewards/forensics_strict_edge_reward": 0.021728052699472755,
      "rewards/forensics_tiou_reward": 0.5478451251983643,
      "step": 427
    },
    {
      "completion_length": 21.625,
      "epoch": 1.3133640552995391,
      "grad_norm": 12.50629997253418,
      "kl": 0.1806640625,
      "learning_rate": 6.715384615384615e-07,
      "loss": 0.1086,
      "reward": 1.3132410645484924,
      "reward_std": 0.2736186683177948,
      "reward_std_all": 0.24627161026000977,
      "rewards-max/forensics_sinkhorn_reward": 0.9465204775333405,
      "rewards-max/forensics_strict_edge_reward": 0.35240000672638416,
      "rewards-max/forensics_tiou_reward": 0.8938053846359253,
      "rewards/forensics_sinkhorn_reward": 0.7399837672710419,
      "rewards/forensics_strict_edge_reward": 0.053381857462227345,
      "rewards/forensics_tiou_reward": 0.5198754072189331,
      "step": 428
    },
    {
      "completion_length": 21.640625,
      "epoch": 1.316436251920123,
      "grad_norm": 7.766979217529297,
      "kl": 0.2265625,
      "learning_rate": 6.707692307692307e-07,
      "loss": 0.0864,
      "reward": 1.1886385083198547,
      "reward_std": 0.19508058577775955,
      "reward_std_all": 0.20500313490629196,
      "rewards-max/forensics_sinkhorn_reward": 0.9280780851840973,
      "rewards-max/forensics_strict_edge_reward": 0.117774847894907,
      "rewards-max/forensics_tiou_reward": 0.8461987376213074,
      "rewards/forensics_sinkhorn_reward": 0.7133131623268127,
      "rewards/forensics_strict_edge_reward": 0.009832094423472881,
      "rewards/forensics_tiou_reward": 0.4654932916164398,
      "step": 429
    },
    {
      "completion_length": 25.09375,
      "epoch": 1.3195084485407067,
      "grad_norm": 9.217371940612793,
      "kl": 0.21240234375,
      "learning_rate": 6.7e-07,
      "loss": 0.0757,
      "reward": 1.1675317883491516,
      "reward_std": 0.21037742495536804,
      "reward_std_all": 0.1985936015844345,
      "rewards-max/forensics_sinkhorn_reward": 0.9456774592399597,
      "rewards-max/forensics_strict_edge_reward": 0.2605111859738827,
      "rewards-max/forensics_tiou_reward": 0.8942369222640991,
      "rewards/forensics_sinkhorn_reward": 0.6915456354618073,
      "rewards/forensics_strict_edge_reward": 0.013276549056172371,
      "rewards/forensics_tiou_reward": 0.4627096503973007,
      "step": 430
    },
    {
      "completion_length": 23.78125,
      "epoch": 1.3225806451612903,
      "grad_norm": 11.324387550354004,
      "kl": 0.51171875,
      "learning_rate": 6.692307692307692e-07,
      "loss": 0.0496,
      "reward": 1.1671879291534424,
      "reward_std": 0.16216883808374405,
      "reward_std_all": 0.15110985934734344,
      "rewards-max/forensics_sinkhorn_reward": 0.8890307247638702,
      "rewards-max/forensics_strict_edge_reward": 0.25142066925764084,
      "rewards-max/forensics_tiou_reward": 0.8119129836559296,
      "rewards/forensics_sinkhorn_reward": 0.6884605884552002,
      "rewards/forensics_strict_edge_reward": 0.03110212506726384,
      "rewards/forensics_tiou_reward": 0.44762513041496277,
      "step": 431
    },
    {
      "completion_length": 22.265625,
      "epoch": 1.325652841781874,
      "grad_norm": 9.525300979614258,
      "kl": 0.37841796875,
      "learning_rate": 6.684615384615384e-07,
      "loss": 0.0171,
      "reward": 1.244005799293518,
      "reward_std": 0.2316371574997902,
      "reward_std_all": 0.24849414080381393,
      "rewards-max/forensics_sinkhorn_reward": 0.9475518465042114,
      "rewards-max/forensics_strict_edge_reward": 0.08896535262465477,
      "rewards-max/forensics_tiou_reward": 0.8853485286235809,
      "rewards/forensics_sinkhorn_reward": 0.7337512373924255,
      "rewards/forensics_strict_edge_reward": 0.005606382852420211,
      "rewards/forensics_tiou_reward": 0.5046482086181641,
      "step": 432
    },
    {
      "completion_length": 23.03125,
      "epoch": 1.3287250384024578,
      "grad_norm": 7.885390758514404,
      "kl": 0.21142578125,
      "learning_rate": 6.676923076923077e-07,
      "loss": 0.0754,
      "reward": 1.2817119359970093,
      "reward_std": 0.19623715430498123,
      "reward_std_all": 0.23325417935848236,
      "rewards-max/forensics_sinkhorn_reward": 0.9479142129421234,
      "rewards-max/forensics_strict_edge_reward": 0.3707124348729849,
      "rewards-max/forensics_tiou_reward": 0.8804580569267273,
      "rewards/forensics_sinkhorn_reward": 0.7257850170135498,
      "rewards/forensics_strict_edge_reward": 0.02739399520214647,
      "rewards/forensics_tiou_reward": 0.5285329222679138,
      "step": 433
    },
    {
      "completion_length": 25.0625,
      "epoch": 1.3317972350230414,
      "grad_norm": 11.192765235900879,
      "kl": 0.2158203125,
      "learning_rate": 6.669230769230769e-07,
      "loss": 0.0186,
      "reward": 1.1823468804359436,
      "reward_std": 0.29410024732351303,
      "reward_std_all": 0.29421430826187134,
      "rewards-max/forensics_sinkhorn_reward": 0.9588628709316254,
      "rewards-max/forensics_strict_edge_reward": 0.0459998122532852,
      "rewards-max/forensics_tiou_reward": 0.9276683330535889,
      "rewards/forensics_sinkhorn_reward": 0.703618198633194,
      "rewards/forensics_strict_edge_reward": 0.00209899367109756,
      "rewards/forensics_tiou_reward": 0.47662967443466187,
      "step": 434
    },
    {
      "completion_length": 23.046875,
      "epoch": 1.3348694316436251,
      "grad_norm": 7.658980369567871,
      "kl": 0.185546875,
      "learning_rate": 6.661538461538462e-07,
      "loss": 0.1245,
      "reward": 1.2601311802864075,
      "reward_std": 0.2361445352435112,
      "reward_std_all": 0.21339862793684006,
      "rewards-max/forensics_sinkhorn_reward": 0.9656562507152557,
      "rewards-max/forensics_strict_edge_reward": 0.509157819673419,
      "rewards-max/forensics_tiou_reward": 0.9372659921646118,
      "rewards/forensics_sinkhorn_reward": 0.7317067384719849,
      "rewards/forensics_strict_edge_reward": 0.05369291501119733,
      "rewards/forensics_tiou_reward": 0.47473154962062836,
      "step": 435
    },
    {
      "completion_length": 22.46875,
      "epoch": 1.337941628264209,
      "grad_norm": 9.422367095947266,
      "kl": 0.6806640625,
      "learning_rate": 6.653846153846153e-07,
      "loss": 0.0452,
      "reward": 1.2903512716293335,
      "reward_std": 0.3221071884036064,
      "reward_std_all": 0.2750665470957756,
      "rewards-max/forensics_sinkhorn_reward": 0.9687007963657379,
      "rewards-max/forensics_strict_edge_reward": 0.3800070434808731,
      "rewards-max/forensics_tiou_reward": 0.9469785392284393,
      "rewards/forensics_sinkhorn_reward": 0.7149856984615326,
      "rewards/forensics_strict_edge_reward": 0.05653511919081211,
      "rewards/forensics_tiou_reward": 0.5188304781913757,
      "step": 436
    },
    {
      "completion_length": 24.234375,
      "epoch": 1.3410138248847927,
      "grad_norm": 8.754499435424805,
      "kl": 0.28857421875,
      "learning_rate": 6.646153846153845e-07,
      "loss": -0.0579,
      "reward": 1.3809483051300049,
      "reward_std": 0.21668080240488052,
      "reward_std_all": 0.22198589891195297,
      "rewards-max/forensics_sinkhorn_reward": 0.9684410691261292,
      "rewards-max/forensics_strict_edge_reward": 0.326717309653759,
      "rewards-max/forensics_tiou_reward": 0.9457965195178986,
      "rewards/forensics_sinkhorn_reward": 0.7804518043994904,
      "rewards/forensics_strict_edge_reward": 0.02978052105754614,
      "rewards/forensics_tiou_reward": 0.5707160234451294,
      "step": 437
    },
    {
      "completion_length": 20.25,
      "epoch": 1.3440860215053765,
      "grad_norm": 9.390589714050293,
      "kl": 0.3701171875,
      "learning_rate": 6.638461538461539e-07,
      "loss": 0.0417,
      "reward": 0.8938367962837219,
      "reward_std": 0.15747933834791183,
      "reward_std_all": 0.15750883519649506,
      "rewards-max/forensics_sinkhorn_reward": 0.9178833067417145,
      "rewards-max/forensics_strict_edge_reward": 0.22467467924616358,
      "rewards-max/forensics_tiou_reward": 0.8429654836654663,
      "rewards/forensics_sinkhorn_reward": 0.602297306060791,
      "rewards/forensics_strict_edge_reward": 0.009566623119894757,
      "rewards/forensics_tiou_reward": 0.28197284042835236,
      "step": 438
    },
    {
      "completion_length": 21.65625,
      "epoch": 1.34715821812596,
      "grad_norm": 6.5189290046691895,
      "kl": 0.18798828125,
      "learning_rate": 6.630769230769231e-07,
      "loss": -0.0915,
      "reward": 1.111283540725708,
      "reward_std": 0.18859706819057465,
      "reward_std_all": 0.16621649265289307,
      "rewards-max/forensics_sinkhorn_reward": 0.934858113527298,
      "rewards-max/forensics_strict_edge_reward": 0.39086367189884186,
      "rewards-max/forensics_tiou_reward": 0.8776098787784576,
      "rewards/forensics_sinkhorn_reward": 0.6851813197135925,
      "rewards/forensics_strict_edge_reward": 0.028385099722072482,
      "rewards/forensics_tiou_reward": 0.39771707355976105,
      "step": 439
    },
    {
      "completion_length": 25.421875,
      "epoch": 1.3502304147465438,
      "grad_norm": 10.238957405090332,
      "kl": 0.3046875,
      "learning_rate": 6.623076923076923e-07,
      "loss": 0.0984,
      "reward": 1.1595577597618103,
      "reward_std": 0.27122239768505096,
      "reward_std_all": 0.23865669965744019,
      "rewards-max/forensics_sinkhorn_reward": 0.9678674638271332,
      "rewards-max/forensics_strict_edge_reward": 0.11752384901046753,
      "rewards-max/forensics_tiou_reward": 0.9415440857410431,
      "rewards/forensics_sinkhorn_reward": 0.6602033376693726,
      "rewards/forensics_strict_edge_reward": 0.007536035031080246,
      "rewards/forensics_tiou_reward": 0.4918184280395508,
      "step": 440
    },
    {
      "completion_length": 25.15625,
      "epoch": 1.3533026113671274,
      "grad_norm": 9.025731086730957,
      "kl": 0.25830078125,
      "learning_rate": 6.615384615384615e-07,
      "loss": -0.0494,
      "reward": 1.1132122874259949,
      "reward_std": 0.23248856514692307,
      "reward_std_all": 0.24162891507148743,
      "rewards-max/forensics_sinkhorn_reward": 0.9519290626049042,
      "rewards-max/forensics_strict_edge_reward": 0.14731938857585192,
      "rewards-max/forensics_tiou_reward": 0.8957212567329407,
      "rewards/forensics_sinkhorn_reward": 0.6349467039108276,
      "rewards/forensics_strict_edge_reward": 0.020559276541462168,
      "rewards/forensics_tiou_reward": 0.4577062875032425,
      "step": 441
    },
    {
      "completion_length": 21.8125,
      "epoch": 1.3563748079877112,
      "grad_norm": 6.601945400238037,
      "kl": 0.375,
      "learning_rate": 6.607692307692307e-07,
      "loss": 0.0001,
      "reward": 1.2771654725074768,
      "reward_std": 0.21997994184494019,
      "reward_std_all": 0.2164580300450325,
      "rewards-max/forensics_sinkhorn_reward": 0.9728026688098907,
      "rewards-max/forensics_strict_edge_reward": 0.6365986764431,
      "rewards-max/forensics_tiou_reward": 0.9574104845523834,
      "rewards/forensics_sinkhorn_reward": 0.7183136641979218,
      "rewards/forensics_strict_edge_reward": 0.06554026156663895,
      "rewards/forensics_tiou_reward": 0.4933115541934967,
      "step": 442
    },
    {
      "completion_length": 23.203125,
      "epoch": 1.359447004608295,
      "grad_norm": 10.183173179626465,
      "kl": 0.21923828125,
      "learning_rate": 6.6e-07,
      "loss": -0.0456,
      "reward": 1.281443178653717,
      "reward_std": 0.2674383223056793,
      "reward_std_all": 0.245619535446167,
      "rewards-max/forensics_sinkhorn_reward": 0.9641861617565155,
      "rewards-max/forensics_strict_edge_reward": 0.506708562374115,
      "rewards-max/forensics_tiou_reward": 0.9387041330337524,
      "rewards/forensics_sinkhorn_reward": 0.7327814102172852,
      "rewards/forensics_strict_edge_reward": 0.06211177259683609,
      "rewards/forensics_tiou_reward": 0.48655004799366,
      "step": 443
    },
    {
      "completion_length": 20.984375,
      "epoch": 1.3625192012288787,
      "grad_norm": 8.63659381866455,
      "kl": 0.2158203125,
      "learning_rate": 6.592307692307691e-07,
      "loss": 0.053,
      "reward": 1.4407222270965576,
      "reward_std": 0.36337482184171677,
      "reward_std_all": 0.32719291746616364,
      "rewards-max/forensics_sinkhorn_reward": 0.9725478887557983,
      "rewards-max/forensics_strict_edge_reward": 0.6516932249069214,
      "rewards-max/forensics_tiou_reward": 0.9567541182041168,
      "rewards/forensics_sinkhorn_reward": 0.7809837758541107,
      "rewards/forensics_strict_edge_reward": 0.09374850988388062,
      "rewards/forensics_tiou_reward": 0.5659898221492767,
      "step": 444
    },
    {
      "completion_length": 22.75,
      "epoch": 1.3655913978494625,
      "grad_norm": 6.094111919403076,
      "kl": 0.25830078125,
      "learning_rate": 6.584615384615384e-07,
      "loss": -0.0708,
      "reward": 1.3596143126487732,
      "reward_std": 0.2024795040488243,
      "reward_std_all": 0.18376317620277405,
      "rewards-max/forensics_sinkhorn_reward": 0.9684979617595673,
      "rewards-max/forensics_strict_edge_reward": 0.3396413326263428,
      "rewards-max/forensics_tiou_reward": 0.9413243234157562,
      "rewards/forensics_sinkhorn_reward": 0.7531585693359375,
      "rewards/forensics_strict_edge_reward": 0.030605400912463665,
      "rewards/forensics_tiou_reward": 0.5758504271507263,
      "step": 445
    },
    {
      "completion_length": 22.09375,
      "epoch": 1.368663594470046,
      "grad_norm": 9.60615348815918,
      "kl": 0.4091796875,
      "learning_rate": 6.576923076923077e-07,
      "loss": -0.075,
      "reward": 1.5279933214187622,
      "reward_std": 0.37499383091926575,
      "reward_std_all": 0.34003493189811707,
      "rewards-max/forensics_sinkhorn_reward": 0.9755555391311646,
      "rewards-max/forensics_strict_edge_reward": 0.5210062116384506,
      "rewards-max/forensics_tiou_reward": 0.9647426903247833,
      "rewards/forensics_sinkhorn_reward": 0.7788969874382019,
      "rewards/forensics_strict_edge_reward": 0.09006860759109259,
      "rewards/forensics_tiou_reward": 0.6590277552604675,
      "step": 446
    },
    {
      "completion_length": 23.5625,
      "epoch": 1.3717357910906298,
      "grad_norm": 8.52184772491455,
      "kl": 0.25146484375,
      "learning_rate": 6.569230769230769e-07,
      "loss": -0.0353,
      "reward": 1.2559112906455994,
      "reward_std": 0.31192417442798615,
      "reward_std_all": 0.29733988642692566,
      "rewards-max/forensics_sinkhorn_reward": 0.973842054605484,
      "rewards-max/forensics_strict_edge_reward": 0.24510634317994118,
      "rewards-max/forensics_tiou_reward": 0.9564524590969086,
      "rewards/forensics_sinkhorn_reward": 0.7134863138198853,
      "rewards/forensics_strict_edge_reward": 0.03370187012478709,
      "rewards/forensics_tiou_reward": 0.5087231993675232,
      "step": 447
    },
    {
      "completion_length": 20.0625,
      "epoch": 1.3748079877112136,
      "grad_norm": 22.369367599487305,
      "kl": 0.69921875,
      "learning_rate": 6.561538461538462e-07,
      "loss": -0.0373,
      "reward": 1.1402716934680939,
      "reward_std": 0.21800199151039124,
      "reward_std_all": 0.2026078701019287,
      "rewards-max/forensics_sinkhorn_reward": 0.949682354927063,
      "rewards-max/forensics_strict_edge_reward": 0.47385871270671487,
      "rewards-max/forensics_tiou_reward": 0.8939457535743713,
      "rewards/forensics_sinkhorn_reward": 0.6989820301532745,
      "rewards/forensics_strict_edge_reward": 0.02210861723870039,
      "rewards/forensics_tiou_reward": 0.4191810190677643,
      "step": 448
    },
    {
      "completion_length": 21.625,
      "epoch": 1.3778801843317972,
      "grad_norm": 12.961442947387695,
      "kl": 0.490234375,
      "learning_rate": 6.553846153846153e-07,
      "loss": -0.0101,
      "reward": 1.2846871614456177,
      "reward_std": 0.3400196433067322,
      "reward_std_all": 0.2975914031267166,
      "rewards-max/forensics_sinkhorn_reward": 0.9877427518367767,
      "rewards-max/forensics_strict_edge_reward": 0.5304600298404694,
      "rewards-max/forensics_tiou_reward": 0.9959416687488556,
      "rewards/forensics_sinkhorn_reward": 0.7106202244758606,
      "rewards/forensics_strict_edge_reward": 0.058558568358421326,
      "rewards/forensics_tiou_reward": 0.515508383512497,
      "step": 449
    },
    {
      "completion_length": 24.96875,
      "epoch": 1.380952380952381,
      "grad_norm": 10.848255157470703,
      "kl": 0.20751953125,
      "learning_rate": 6.546153846153845e-07,
      "loss": -0.0222,
      "reward": 1.1195642352104187,
      "reward_std": 0.2342844158411026,
      "reward_std_all": 0.20549052953720093,
      "rewards-max/forensics_sinkhorn_reward": 0.9446855187416077,
      "rewards-max/forensics_strict_edge_reward": 0.3191914036870003,
      "rewards-max/forensics_tiou_reward": 0.8923301994800568,
      "rewards/forensics_sinkhorn_reward": 0.6599286496639252,
      "rewards/forensics_strict_edge_reward": 0.031645646784454584,
      "rewards/forensics_tiou_reward": 0.42798982560634613,
      "step": 450
    },
    {
      "completion_length": 23.21875,
      "epoch": 1.3840245775729647,
      "grad_norm": 9.421950340270996,
      "kl": 0.408203125,
      "learning_rate": 6.538461538461538e-07,
      "loss": 0.0786,
      "reward": 1.2458714246749878,
      "reward_std": 0.2525847405195236,
      "reward_std_all": 0.25793036818504333,
      "rewards-max/forensics_sinkhorn_reward": 0.9619814157485962,
      "rewards-max/forensics_strict_edge_reward": 0.29233212023973465,
      "rewards-max/forensics_tiou_reward": 0.9265826642513275,
      "rewards/forensics_sinkhorn_reward": 0.7352473735809326,
      "rewards/forensics_strict_edge_reward": 0.020512965507805347,
      "rewards/forensics_tiou_reward": 0.4901110380887985,
      "step": 451
    },
    {
      "completion_length": 20.90625,
      "epoch": 1.3870967741935485,
      "grad_norm": 8.787131309509277,
      "kl": 0.3037109375,
      "learning_rate": 6.530769230769231e-07,
      "loss": -0.0168,
      "reward": 1.2459401488304138,
      "reward_std": 0.31480370461940765,
      "reward_std_all": 0.2755688428878784,
      "rewards-max/forensics_sinkhorn_reward": 0.9788693487644196,
      "rewards-max/forensics_strict_edge_reward": 0.35175131633877754,
      "rewards-max/forensics_tiou_reward": 0.969942033290863,
      "rewards/forensics_sinkhorn_reward": 0.713767021894455,
      "rewards/forensics_strict_edge_reward": 0.04221503157168627,
      "rewards/forensics_tiou_reward": 0.48995813727378845,
      "step": 452
    },
    {
      "completion_length": 21.375,
      "epoch": 1.390168970814132,
      "grad_norm": 7.050997257232666,
      "kl": 0.322265625,
      "learning_rate": 6.523076923076923e-07,
      "loss": 0.0968,
      "reward": 1.2769695520401,
      "reward_std": 0.2268117070198059,
      "reward_std_all": 0.195955291390419,
      "rewards-max/forensics_sinkhorn_reward": 0.9499919712543488,
      "rewards-max/forensics_strict_edge_reward": 0.22026493772864342,
      "rewards-max/forensics_tiou_reward": 0.9027562737464905,
      "rewards/forensics_sinkhorn_reward": 0.7456396520137787,
      "rewards/forensics_strict_edge_reward": 0.019254797836765647,
      "rewards/forensics_tiou_reward": 0.5120751559734344,
      "step": 453
    },
    {
      "completion_length": 20.796875,
      "epoch": 1.3932411674347158,
      "grad_norm": 13.545931816101074,
      "kl": 0.2724609375,
      "learning_rate": 6.515384615384615e-07,
      "loss": -0.0995,
      "reward": 1.511989951133728,
      "reward_std": 0.2688918337225914,
      "reward_std_all": 0.2731965556740761,
      "rewards-max/forensics_sinkhorn_reward": 0.9570901393890381,
      "rewards-max/forensics_strict_edge_reward": 0.4901042878627777,
      "rewards-max/forensics_tiou_reward": 0.9274949431419373,
      "rewards/forensics_sinkhorn_reward": 0.8104263246059418,
      "rewards/forensics_strict_edge_reward": 0.0683278776705265,
      "rewards/forensics_tiou_reward": 0.6332356929779053,
      "step": 454
    },
    {
      "completion_length": 21.0625,
      "epoch": 1.3963133640552996,
      "grad_norm": 8.859350204467773,
      "kl": 0.4482421875,
      "learning_rate": 6.507692307692307e-07,
      "loss": 0.051,
      "reward": 1.248164415359497,
      "reward_std": 0.30808021128177643,
      "reward_std_all": 0.2642481327056885,
      "rewards-max/forensics_sinkhorn_reward": 0.9756504893302917,
      "rewards-max/forensics_strict_edge_reward": 0.484806090593338,
      "rewards-max/forensics_tiou_reward": 0.9627414047718048,
      "rewards/forensics_sinkhorn_reward": 0.6871306300163269,
      "rewards/forensics_strict_edge_reward": 0.060790304094552994,
      "rewards/forensics_tiou_reward": 0.5002434998750687,
      "step": 455
    },
    {
      "completion_length": 20.53125,
      "epoch": 1.3993855606758832,
      "grad_norm": 10.658343315124512,
      "kl": 0.39990234375,
      "learning_rate": 6.5e-07,
      "loss": -0.0152,
      "reward": 1.2704519629478455,
      "reward_std": 0.24145804345607758,
      "reward_std_all": 0.25736144930124283,
      "rewards-max/forensics_sinkhorn_reward": 0.9758962392807007,
      "rewards-max/forensics_strict_edge_reward": 0.480892151594162,
      "rewards-max/forensics_tiou_reward": 0.9661474823951721,
      "rewards/forensics_sinkhorn_reward": 0.7381244003772736,
      "rewards/forensics_strict_edge_reward": 0.02503308653831482,
      "rewards/forensics_tiou_reward": 0.5072944909334183,
      "step": 456
    },
    {
      "completion_length": 19.84375,
      "epoch": 1.402457757296467,
      "grad_norm": 16.093578338623047,
      "kl": 0.294921875,
      "learning_rate": 6.492307692307692e-07,
      "loss": -0.0541,
      "reward": 1.1025660037994385,
      "reward_std": 0.17406971007585526,
      "reward_std_all": 0.15735262632369995,
      "rewards-max/forensics_sinkhorn_reward": 0.9503791034221649,
      "rewards-max/forensics_strict_edge_reward": 0.3469466418027878,
      "rewards-max/forensics_tiou_reward": 0.8959159851074219,
      "rewards/forensics_sinkhorn_reward": 0.6652092635631561,
      "rewards/forensics_strict_edge_reward": 0.025285476818680763,
      "rewards/forensics_tiou_reward": 0.4120713025331497,
      "step": 457
    },
    {
      "completion_length": 21.1875,
      "epoch": 1.4055299539170507,
      "grad_norm": 9.592914581298828,
      "kl": 0.25,
      "learning_rate": 6.484615384615383e-07,
      "loss": 0.017,
      "reward": 1.1386725306510925,
      "reward_std": 0.23347389698028564,
      "reward_std_all": 0.202541284263134,
      "rewards-max/forensics_sinkhorn_reward": 0.9238344132900238,
      "rewards-max/forensics_strict_edge_reward": 0.4807496815919876,
      "rewards-max/forensics_tiou_reward": 0.8429220616817474,
      "rewards/forensics_sinkhorn_reward": 0.6775570213794708,
      "rewards/forensics_strict_edge_reward": 0.029221689677797258,
      "rewards/forensics_tiou_reward": 0.43189384043216705,
      "step": 458
    },
    {
      "completion_length": 20.453125,
      "epoch": 1.4086021505376345,
      "grad_norm": 9.117100715637207,
      "kl": 0.24853515625,
      "learning_rate": 6.476923076923077e-07,
      "loss": -0.003,
      "reward": 1.527501106262207,
      "reward_std": 0.2739865183830261,
      "reward_std_all": 0.26315176486968994,
      "rewards-max/forensics_sinkhorn_reward": 0.9853528141975403,
      "rewards-max/forensics_strict_edge_reward": 0.6419725120067596,
      "rewards-max/forensics_tiou_reward": 0.9897935390472412,
      "rewards/forensics_sinkhorn_reward": 0.8240436613559723,
      "rewards/forensics_strict_edge_reward": 0.0741623006761074,
      "rewards/forensics_tiou_reward": 0.6292952001094818,
      "step": 459
    },
    {
      "completion_length": 20.78125,
      "epoch": 1.411674347158218,
      "grad_norm": 9.7948637008667,
      "kl": 0.33203125,
      "learning_rate": 6.469230769230769e-07,
      "loss": 0.0855,
      "reward": 1.145746111869812,
      "reward_std": 0.15502329170703888,
      "reward_std_all": 0.17677751183509827,
      "rewards-max/forensics_sinkhorn_reward": 0.9755372107028961,
      "rewards-max/forensics_strict_edge_reward": 0.3386820778250694,
      "rewards-max/forensics_tiou_reward": 0.9619477987289429,
      "rewards/forensics_sinkhorn_reward": 0.6708943247795105,
      "rewards/forensics_strict_edge_reward": 0.02123142685741186,
      "rewards/forensics_tiou_reward": 0.4536203294992447,
      "step": 460
    },
    {
      "completion_length": 19.890625,
      "epoch": 1.4147465437788018,
      "grad_norm": 12.400056838989258,
      "kl": 0.2822265625,
      "learning_rate": 6.461538461538462e-07,
      "loss": 0.0433,
      "reward": 1.1682363748550415,
      "reward_std": 0.16705255955457687,
      "reward_std_all": 0.17678191512823105,
      "rewards-max/forensics_sinkhorn_reward": 0.9758075773715973,
      "rewards-max/forensics_strict_edge_reward": 0.21618319302797318,
      "rewards-max/forensics_tiou_reward": 0.9597710967063904,
      "rewards/forensics_sinkhorn_reward": 0.7013436555862427,
      "rewards/forensics_strict_edge_reward": 0.01223185146227479,
      "rewards/forensics_tiou_reward": 0.45466089248657227,
      "step": 461
    },
    {
      "completion_length": 20.765625,
      "epoch": 1.4178187403993856,
      "grad_norm": 10.160968780517578,
      "kl": 0.20556640625,
      "learning_rate": 6.453846153846153e-07,
      "loss": 0.0849,
      "reward": 1.4376073479652405,
      "reward_std": 0.19620535522699356,
      "reward_std_all": 0.17825567722320557,
      "rewards-max/forensics_sinkhorn_reward": 0.9749939739704132,
      "rewards-max/forensics_strict_edge_reward": 0.36884360760450363,
      "rewards-max/forensics_tiou_reward": 0.9559610188007355,
      "rewards/forensics_sinkhorn_reward": 0.7972771525382996,
      "rewards/forensics_strict_edge_reward": 0.03346751816570759,
      "rewards/forensics_tiou_reward": 0.6068626344203949,
      "step": 462
    },
    {
      "completion_length": 21.84375,
      "epoch": 1.4208909370199692,
      "grad_norm": 10.58541488647461,
      "kl": 0.244140625,
      "learning_rate": 6.446153846153845e-07,
      "loss": -0.0117,
      "reward": 1.1321164965629578,
      "reward_std": 0.1380465179681778,
      "reward_std_all": 0.15481334924697876,
      "rewards-max/forensics_sinkhorn_reward": 0.8928524553775787,
      "rewards-max/forensics_strict_edge_reward": 0.16166765615344048,
      "rewards-max/forensics_tiou_reward": 0.8170634806156158,
      "rewards/forensics_sinkhorn_reward": 0.6963797807693481,
      "rewards/forensics_strict_edge_reward": 0.007270896108821034,
      "rewards/forensics_tiou_reward": 0.4284658133983612,
      "step": 463
    },
    {
      "completion_length": 20.28125,
      "epoch": 1.423963133640553,
      "grad_norm": 12.239574432373047,
      "kl": 0.21826171875,
      "learning_rate": 6.438461538461538e-07,
      "loss": -0.1156,
      "reward": 1.088561862707138,
      "reward_std": 0.2043471708893776,
      "reward_std_all": 0.18582309037446976,
      "rewards-max/forensics_sinkhorn_reward": 0.9151178002357483,
      "rewards-max/forensics_strict_edge_reward": 0.3877888023853302,
      "rewards-max/forensics_tiou_reward": 0.8233323991298676,
      "rewards/forensics_sinkhorn_reward": 0.6765875220298767,
      "rewards/forensics_strict_edge_reward": 0.021148206666111946,
      "rewards/forensics_tiou_reward": 0.3908260762691498,
      "step": 464
    },
    {
      "completion_length": 22.359375,
      "epoch": 1.4270353302611367,
      "grad_norm": 7.536533832550049,
      "kl": 0.400390625,
      "learning_rate": 6.43076923076923e-07,
      "loss": 0.1165,
      "reward": 1.310667872428894,
      "reward_std": 0.2169315069913864,
      "reward_std_all": 0.19037093967199326,
      "rewards-max/forensics_sinkhorn_reward": 0.9701993465423584,
      "rewards-max/forensics_strict_edge_reward": 0.22297906875610352,
      "rewards-max/forensics_tiou_reward": 0.9436931312084198,
      "rewards/forensics_sinkhorn_reward": 0.7411945462226868,
      "rewards/forensics_strict_edge_reward": 0.013515571132302284,
      "rewards/forensics_tiou_reward": 0.5559577643871307,
      "step": 465
    },
    {
      "completion_length": 20.453125,
      "epoch": 1.4301075268817205,
      "grad_norm": 8.825799942016602,
      "kl": 0.2373046875,
      "learning_rate": 6.423076923076924e-07,
      "loss": -0.0213,
      "reward": 1.1364361643791199,
      "reward_std": 0.2573787271976471,
      "reward_std_all": 0.22229349613189697,
      "rewards-max/forensics_sinkhorn_reward": 0.962775468826294,
      "rewards-max/forensics_strict_edge_reward": 0.16316426545381546,
      "rewards-max/forensics_tiou_reward": 0.9244408309459686,
      "rewards/forensics_sinkhorn_reward": 0.6642242670059204,
      "rewards/forensics_strict_edge_reward": 0.019124118611216545,
      "rewards/forensics_tiou_reward": 0.4530876874923706,
      "step": 466
    },
    {
      "completion_length": 21.953125,
      "epoch": 1.4331797235023043,
      "grad_norm": 10.741941452026367,
      "kl": 0.23828125,
      "learning_rate": 6.415384615384615e-07,
      "loss": -0.0343,
      "reward": 1.1798378229141235,
      "reward_std": 0.23167597502470016,
      "reward_std_all": 0.2061602845788002,
      "rewards-max/forensics_sinkhorn_reward": 0.9331095218658447,
      "rewards-max/forensics_strict_edge_reward": 0.05609549954533577,
      "rewards-max/forensics_tiou_reward": 0.850265234708786,
      "rewards/forensics_sinkhorn_reward": 0.7323605120182037,
      "rewards/forensics_strict_edge_reward": 0.003356363973580301,
      "rewards/forensics_tiou_reward": 0.44412101805210114,
      "step": 467
    },
    {
      "completion_length": 19.25,
      "epoch": 1.4362519201228878,
      "grad_norm": 12.920519828796387,
      "kl": 0.2373046875,
      "learning_rate": 6.407692307692307e-07,
      "loss": 0.0206,
      "reward": 0.9372741878032684,
      "reward_std": 0.15921270102262497,
      "reward_std_all": 0.14091575145721436,
      "rewards-max/forensics_sinkhorn_reward": 0.8878638744354248,
      "rewards-max/forensics_strict_edge_reward": 0.22915149852633476,
      "rewards-max/forensics_tiou_reward": 0.7311270236968994,
      "rewards/forensics_sinkhorn_reward": 0.6180101633071899,
      "rewards/forensics_strict_edge_reward": 0.01677926490083337,
      "rewards/forensics_tiou_reward": 0.30248476564884186,
      "step": 468
    },
    {
      "completion_length": 20.96875,
      "epoch": 1.4393241167434716,
      "grad_norm": 10.727717399597168,
      "kl": 0.185546875,
      "learning_rate": 6.4e-07,
      "loss": 0.0196,
      "reward": 1.1507561206817627,
      "reward_std": 0.2037607952952385,
      "reward_std_all": 0.18542461097240448,
      "rewards-max/forensics_sinkhorn_reward": 0.9275187849998474,
      "rewards-max/forensics_strict_edge_reward": 0.34084393084049225,
      "rewards-max/forensics_tiou_reward": 0.8446626663208008,
      "rewards/forensics_sinkhorn_reward": 0.690143346786499,
      "rewards/forensics_strict_edge_reward": 0.03095345012843609,
      "rewards/forensics_tiou_reward": 0.4296593964099884,
      "step": 469
    },
    {
      "completion_length": 19.8125,
      "epoch": 1.4423963133640554,
      "grad_norm": 10.12091064453125,
      "kl": 0.20458984375,
      "learning_rate": 6.392307692307692e-07,
      "loss": 0.0207,
      "reward": 1.4174919724464417,
      "reward_std": 0.22601962834596634,
      "reward_std_all": 0.1984826698899269,
      "rewards-max/forensics_sinkhorn_reward": 0.96140056848526,
      "rewards-max/forensics_strict_edge_reward": 0.5635445415973663,
      "rewards-max/forensics_tiou_reward": 0.9319658577442169,
      "rewards/forensics_sinkhorn_reward": 0.779353529214859,
      "rewards/forensics_strict_edge_reward": 0.05139452405273914,
      "rewards/forensics_tiou_reward": 0.5867439210414886,
      "step": 470
    },
    {
      "completion_length": 21.234375,
      "epoch": 1.445468509984639,
      "grad_norm": 15.108663558959961,
      "kl": 0.2060546875,
      "learning_rate": 6.384615384615383e-07,
      "loss": 0.0698,
      "reward": 1.3557539582252502,
      "reward_std": 0.2850178927183151,
      "reward_std_all": 0.2608643099665642,
      "rewards-max/forensics_sinkhorn_reward": 0.9744204878807068,
      "rewards-max/forensics_strict_edge_reward": 0.3197415918111801,
      "rewards-max/forensics_tiou_reward": 0.9596894085407257,
      "rewards/forensics_sinkhorn_reward": 0.7453547418117523,
      "rewards/forensics_strict_edge_reward": 0.048644695430994034,
      "rewards/forensics_tiou_reward": 0.5617544651031494,
      "step": 471
    },
    {
      "completion_length": 22.1875,
      "epoch": 1.4485407066052227,
      "grad_norm": 14.286282539367676,
      "kl": 0.2373046875,
      "learning_rate": 6.376923076923077e-07,
      "loss": -0.0398,
      "reward": 1.3716176748275757,
      "reward_std": 0.1892135739326477,
      "reward_std_all": 0.17364714294672012,
      "rewards-max/forensics_sinkhorn_reward": 0.9811334609985352,
      "rewards-max/forensics_strict_edge_reward": 0.054623812437057495,
      "rewards-max/forensics_tiou_reward": 0.9771104454994202,
      "rewards/forensics_sinkhorn_reward": 0.7942187488079071,
      "rewards/forensics_strict_edge_reward": 0.005491488380357623,
      "rewards/forensics_tiou_reward": 0.5719074904918671,
      "step": 472
    },
    {
      "completion_length": 21.890625,
      "epoch": 1.4516129032258065,
      "grad_norm": 11.662568092346191,
      "kl": 0.27294921875,
      "learning_rate": 6.369230769230769e-07,
      "loss": -0.1312,
      "reward": 1.2425729632377625,
      "reward_std": 0.17989394813776016,
      "reward_std_all": 0.1849907487630844,
      "rewards-max/forensics_sinkhorn_reward": 0.9650456011295319,
      "rewards-max/forensics_strict_edge_reward": 0.42901813983917236,
      "rewards-max/forensics_tiou_reward": 0.9390884041786194,
      "rewards/forensics_sinkhorn_reward": 0.714306503534317,
      "rewards/forensics_strict_edge_reward": 0.041981352493166924,
      "rewards/forensics_tiou_reward": 0.48628516495227814,
      "step": 473
    },
    {
      "completion_length": 21.140625,
      "epoch": 1.4546850998463903,
      "grad_norm": 12.418452262878418,
      "kl": 0.24462890625,
      "learning_rate": 6.361538461538462e-07,
      "loss": -0.0293,
      "reward": 1.2746409177780151,
      "reward_std": 0.16375527530908585,
      "reward_std_all": 0.15768073499202728,
      "rewards-max/forensics_sinkhorn_reward": 0.9543455839157104,
      "rewards-max/forensics_strict_edge_reward": 0.21280770003795624,
      "rewards-max/forensics_tiou_reward": 0.9072048664093018,
      "rewards/forensics_sinkhorn_reward": 0.7382052540779114,
      "rewards/forensics_strict_edge_reward": 0.027272209525108337,
      "rewards/forensics_tiou_reward": 0.509163424372673,
      "step": 474
    },
    {
      "completion_length": 21.359375,
      "epoch": 1.4577572964669738,
      "grad_norm": 13.17961311340332,
      "kl": 0.16796875,
      "learning_rate": 6.353846153846154e-07,
      "loss": 0.0753,
      "reward": 1.1343316435813904,
      "reward_std": 0.1584794670343399,
      "reward_std_all": 0.14621786773204803,
      "rewards-max/forensics_sinkhorn_reward": 0.977485865354538,
      "rewards-max/forensics_strict_edge_reward": 0.2944649010896683,
      "rewards-max/forensics_tiou_reward": 0.9672795534133911,
      "rewards/forensics_sinkhorn_reward": 0.6838069260120392,
      "rewards/forensics_strict_edge_reward": 0.019062294624745846,
      "rewards/forensics_tiou_reward": 0.4314623773097992,
      "step": 475
    },
    {
      "completion_length": 20.140625,
      "epoch": 1.4608294930875576,
      "grad_norm": 10.20948600769043,
      "kl": 0.2080078125,
      "learning_rate": 6.346153846153845e-07,
      "loss": -0.0826,
      "reward": 1.3097471594810486,
      "reward_std": 0.17443162575364113,
      "reward_std_all": 0.20998452603816986,
      "rewards-max/forensics_sinkhorn_reward": 0.9784288108348846,
      "rewards-max/forensics_strict_edge_reward": 0.3188999593257904,
      "rewards-max/forensics_tiou_reward": 0.9617023766040802,
      "rewards/forensics_sinkhorn_reward": 0.7399225831031799,
      "rewards/forensics_strict_edge_reward": 0.030952636152505875,
      "rewards/forensics_tiou_reward": 0.5388719141483307,
      "step": 476
    },
    {
      "completion_length": 20.765625,
      "epoch": 1.4639016897081414,
      "grad_norm": 10.390693664550781,
      "kl": 0.22705078125,
      "learning_rate": 6.338461538461538e-07,
      "loss": -0.0314,
      "reward": 1.2726298570632935,
      "reward_std": 0.14572128280997276,
      "reward_std_all": 0.18831365555524826,
      "rewards-max/forensics_sinkhorn_reward": 0.9631133079528809,
      "rewards-max/forensics_strict_edge_reward": 0.41361973341554403,
      "rewards-max/forensics_tiou_reward": 0.9310822188854218,
      "rewards/forensics_sinkhorn_reward": 0.7127735316753387,
      "rewards/forensics_strict_edge_reward": 0.027474424801766872,
      "rewards/forensics_tiou_reward": 0.5323818325996399,
      "step": 477
    },
    {
      "completion_length": 21.609375,
      "epoch": 1.466973886328725,
      "grad_norm": 9.98986530303955,
      "kl": 0.26318359375,
      "learning_rate": 6.33076923076923e-07,
      "loss": -0.0543,
      "reward": 1.2926418781280518,
      "reward_std": 0.24961815774440765,
      "reward_std_all": 0.23633618652820587,
      "rewards-max/forensics_sinkhorn_reward": 0.9706928431987762,
      "rewards-max/forensics_strict_edge_reward": 0.4217956215143204,
      "rewards-max/forensics_tiou_reward": 0.9491155445575714,
      "rewards/forensics_sinkhorn_reward": 0.7321699559688568,
      "rewards/forensics_strict_edge_reward": 0.05229797586798668,
      "rewards/forensics_tiou_reward": 0.5081739276647568,
      "step": 478
    },
    {
      "completion_length": 20.96875,
      "epoch": 1.4700460829493087,
      "grad_norm": 14.604934692382812,
      "kl": 0.4248046875,
      "learning_rate": 6.323076923076924e-07,
      "loss": 0.1001,
      "reward": 1.2660879492759705,
      "reward_std": 0.24900483340024948,
      "reward_std_all": 0.22082950174808502,
      "rewards-max/forensics_sinkhorn_reward": 0.94936603307724,
      "rewards-max/forensics_strict_edge_reward": 0.24370086938142776,
      "rewards-max/forensics_tiou_reward": 0.9001503586769104,
      "rewards/forensics_sinkhorn_reward": 0.7504245340824127,
      "rewards/forensics_strict_edge_reward": 0.023419595323503017,
      "rewards/forensics_tiou_reward": 0.49224384129047394,
      "step": 479
    },
    {
      "completion_length": 20.21875,
      "epoch": 1.4731182795698925,
      "grad_norm": 9.42983627319336,
      "kl": 0.25244140625,
      "learning_rate": 6.315384615384615e-07,
      "loss": 0.0473,
      "reward": 1.1461264491081238,
      "reward_std": 0.1746901571750641,
      "reward_std_all": 0.16830170899629593,
      "rewards-max/forensics_sinkhorn_reward": 0.9487311840057373,
      "rewards-max/forensics_strict_edge_reward": 0.15639352425932884,
      "rewards-max/forensics_tiou_reward": 0.901270866394043,
      "rewards/forensics_sinkhorn_reward": 0.684821218252182,
      "rewards/forensics_strict_edge_reward": 0.019938616082072258,
      "rewards/forensics_tiou_reward": 0.4413665682077408,
      "step": 480
    },
    {
      "completion_length": 21.953125,
      "epoch": 1.4761904761904763,
      "grad_norm": 9.1539888381958,
      "kl": 0.2216796875,
      "learning_rate": 6.307692307692307e-07,
      "loss": -0.0962,
      "reward": 1.2574861645698547,
      "reward_std": 0.2050919383764267,
      "reward_std_all": 0.18426896631717682,
      "rewards-max/forensics_sinkhorn_reward": 0.9580437242984772,
      "rewards-max/forensics_strict_edge_reward": 0.25889281718991697,
      "rewards-max/forensics_tiou_reward": 0.922328382730484,
      "rewards/forensics_sinkhorn_reward": 0.7306056618690491,
      "rewards/forensics_strict_edge_reward": 0.02181693966849707,
      "rewards/forensics_tiou_reward": 0.5050634294748306,
      "step": 481
    },
    {
      "completion_length": 19.96875,
      "epoch": 1.4792626728110598,
      "grad_norm": 10.729413032531738,
      "kl": 0.23828125,
      "learning_rate": 6.3e-07,
      "loss": 0.0635,
      "reward": 1.0988832414150238,
      "reward_std": 0.1502503976225853,
      "reward_std_all": 0.1533830687403679,
      "rewards-max/forensics_sinkhorn_reward": 0.9287371933460236,
      "rewards-max/forensics_strict_edge_reward": 0.22296760976314545,
      "rewards-max/forensics_tiou_reward": 0.8532622456550598,
      "rewards/forensics_sinkhorn_reward": 0.6755887269973755,
      "rewards/forensics_strict_edge_reward": 0.01118610380217433,
      "rewards/forensics_tiou_reward": 0.4121083766222,
      "step": 482
    },
    {
      "completion_length": 21.09375,
      "epoch": 1.4823348694316436,
      "grad_norm": 9.20195198059082,
      "kl": 0.3408203125,
      "learning_rate": 6.292307692307692e-07,
      "loss": -0.0605,
      "reward": 1.114815890789032,
      "reward_std": 0.19055065512657166,
      "reward_std_all": 0.17759931832551956,
      "rewards-max/forensics_sinkhorn_reward": 0.9782923460006714,
      "rewards-max/forensics_strict_edge_reward": 0.3236654968932271,
      "rewards-max/forensics_tiou_reward": 0.9665441513061523,
      "rewards/forensics_sinkhorn_reward": 0.6691851317882538,
      "rewards/forensics_strict_edge_reward": 0.019951176131144166,
      "rewards/forensics_tiou_reward": 0.42567962408065796,
      "step": 483
    },
    {
      "completion_length": 20.546875,
      "epoch": 1.4854070660522274,
      "grad_norm": 15.583022117614746,
      "kl": 0.26806640625,
      "learning_rate": 6.284615384615385e-07,
      "loss": 0.0731,
      "reward": 1.2796242237091064,
      "reward_std": 0.2327856793999672,
      "reward_std_all": 0.26531197875738144,
      "rewards-max/forensics_sinkhorn_reward": 0.979760617017746,
      "rewards-max/forensics_strict_edge_reward": 0.4268413633108139,
      "rewards-max/forensics_tiou_reward": 0.9748063087463379,
      "rewards/forensics_sinkhorn_reward": 0.7319879829883575,
      "rewards/forensics_strict_edge_reward": 0.02940200548619032,
      "rewards/forensics_tiou_reward": 0.5182342827320099,
      "step": 484
    },
    {
      "completion_length": 21.515625,
      "epoch": 1.488479262672811,
      "grad_norm": 14.748218536376953,
      "kl": 0.20068359375,
      "learning_rate": 6.276923076923076e-07,
      "loss": 0.1159,
      "reward": 1.134368121623993,
      "reward_std": 0.20319579541683197,
      "reward_std_all": 0.2454613447189331,
      "rewards-max/forensics_sinkhorn_reward": 0.8813227415084839,
      "rewards-max/forensics_strict_edge_reward": 0.6666666865348816,
      "rewards-max/forensics_tiou_reward": 0.7432241141796112,
      "rewards/forensics_sinkhorn_reward": 0.6605829298496246,
      "rewards/forensics_strict_edge_reward": 0.03458630945533514,
      "rewards/forensics_tiou_reward": 0.4391988515853882,
      "step": 485
    },
    {
      "completion_length": 19.421875,
      "epoch": 1.4915514592933947,
      "grad_norm": 9.856803894042969,
      "kl": 0.21435546875,
      "learning_rate": 6.269230769230769e-07,
      "loss": 0.0004,
      "reward": 1.0771465301513672,
      "reward_std": 0.1299300417304039,
      "reward_std_all": 0.11681383103132248,
      "rewards-max/forensics_sinkhorn_reward": 0.9296235144138336,
      "rewards-max/forensics_strict_edge_reward": 0.31604033848270774,
      "rewards-max/forensics_tiou_reward": 0.8534830808639526,
      "rewards/forensics_sinkhorn_reward": 0.6445950865745544,
      "rewards/forensics_strict_edge_reward": 0.02187623367353808,
      "rewards/forensics_tiou_reward": 0.4106752574443817,
      "step": 486
    },
    {
      "completion_length": 21.640625,
      "epoch": 1.4946236559139785,
      "grad_norm": 11.968062400817871,
      "kl": 0.53759765625,
      "learning_rate": 6.261538461538462e-07,
      "loss": -0.0199,
      "reward": 1.3241128325462341,
      "reward_std": 0.19994627684354782,
      "reward_std_all": 0.19221743196249008,
      "rewards-max/forensics_sinkhorn_reward": 0.9573220908641815,
      "rewards-max/forensics_strict_edge_reward": 0.2550462782382965,
      "rewards-max/forensics_tiou_reward": 0.9206777811050415,
      "rewards/forensics_sinkhorn_reward": 0.7614161670207977,
      "rewards/forensics_strict_edge_reward": 0.027298863045871258,
      "rewards/forensics_tiou_reward": 0.5353977382183075,
      "step": 487
    },
    {
      "completion_length": 21.875,
      "epoch": 1.4976958525345623,
      "grad_norm": 9.775039672851562,
      "kl": 0.3388671875,
      "learning_rate": 6.253846153846154e-07,
      "loss": -0.0427,
      "reward": 1.3908011317253113,
      "reward_std": 0.25610876083374023,
      "reward_std_all": 0.220694400370121,
      "rewards-max/forensics_sinkhorn_reward": 0.9582465589046478,
      "rewards-max/forensics_strict_edge_reward": 0.489345908164978,
      "rewards-max/forensics_tiou_reward": 0.9285736382007599,
      "rewards/forensics_sinkhorn_reward": 0.7669638693332672,
      "rewards/forensics_strict_edge_reward": 0.05454161576926708,
      "rewards/forensics_tiou_reward": 0.5692955553531647,
      "step": 488
    },
    {
      "completion_length": 23.0625,
      "epoch": 1.500768049155146,
      "grad_norm": 10.906124114990234,
      "kl": 0.24267578125,
      "learning_rate": 6.246153846153845e-07,
      "loss": 0.1112,
      "reward": 1.1569639444351196,
      "reward_std": 0.12844769656658173,
      "reward_std_all": 0.13251614198088646,
      "rewards-max/forensics_sinkhorn_reward": 0.9492497146129608,
      "rewards-max/forensics_strict_edge_reward": 0.07908300310373306,
      "rewards-max/forensics_tiou_reward": 0.9027390480041504,
      "rewards/forensics_sinkhorn_reward": 0.6758648157119751,
      "rewards/forensics_strict_edge_reward": 0.007943780859932303,
      "rewards/forensics_tiou_reward": 0.4731554388999939,
      "step": 489
    },
    {
      "completion_length": 21.75,
      "epoch": 1.5038402457757296,
      "grad_norm": 10.407751083374023,
      "kl": 0.54638671875,
      "learning_rate": 6.238461538461538e-07,
      "loss": -0.0073,
      "reward": 1.2618923783302307,
      "reward_std": 0.217268168926239,
      "reward_std_all": 0.19753696769475937,
      "rewards-max/forensics_sinkhorn_reward": 0.9583361446857452,
      "rewards-max/forensics_strict_edge_reward": 0.3373643383383751,
      "rewards-max/forensics_tiou_reward": 0.923285037279129,
      "rewards/forensics_sinkhorn_reward": 0.7102850675582886,
      "rewards/forensics_strict_edge_reward": 0.034660002682358027,
      "rewards/forensics_tiou_reward": 0.5169473737478256,
      "step": 490
    },
    {
      "completion_length": 18.984375,
      "epoch": 1.5069124423963134,
      "grad_norm": 25.236310958862305,
      "kl": 0.38818359375,
      "learning_rate": 6.23076923076923e-07,
      "loss": -0.0145,
      "reward": 1.313997507095337,
      "reward_std": 0.12513615190982819,
      "reward_std_all": 0.15231766179203987,
      "rewards-max/forensics_sinkhorn_reward": 0.9718007147312164,
      "rewards-max/forensics_strict_edge_reward": 0.3421594202518463,
      "rewards-max/forensics_tiou_reward": 0.9547234773635864,
      "rewards/forensics_sinkhorn_reward": 0.7605648636817932,
      "rewards/forensics_strict_edge_reward": 0.024581579491496086,
      "rewards/forensics_tiou_reward": 0.5288510322570801,
      "step": 491
    },
    {
      "completion_length": 21.578125,
      "epoch": 1.509984639016897,
      "grad_norm": 12.347187995910645,
      "kl": 0.57421875,
      "learning_rate": 6.223076923076924e-07,
      "loss": -0.0486,
      "reward": 1.3239707946777344,
      "reward_std": 0.21869388967752457,
      "reward_std_all": 0.21063469350337982,
      "rewards-max/forensics_sinkhorn_reward": 0.9775355160236359,
      "rewards-max/forensics_strict_edge_reward": 0.5605834722518921,
      "rewards-max/forensics_tiou_reward": 0.9698548913002014,
      "rewards/forensics_sinkhorn_reward": 0.7328763604164124,
      "rewards/forensics_strict_edge_reward": 0.04413481429219246,
      "rewards/forensics_tiou_reward": 0.5469596087932587,
      "step": 492
    },
    {
      "completion_length": 22.0,
      "epoch": 1.5130568356374807,
      "grad_norm": 10.017979621887207,
      "kl": 0.39501953125,
      "learning_rate": 6.215384615384615e-07,
      "loss": 0.0218,
      "reward": 1.4884717464447021,
      "reward_std": 0.2617281526327133,
      "reward_std_all": 0.23369626700878143,
      "rewards-max/forensics_sinkhorn_reward": 0.9669399857521057,
      "rewards-max/forensics_strict_edge_reward": 0.3678794503211975,
      "rewards-max/forensics_tiou_reward": 0.9373852908611298,
      "rewards/forensics_sinkhorn_reward": 0.8163228929042816,
      "rewards/forensics_strict_edge_reward": 0.03926266636699438,
      "rewards/forensics_tiou_reward": 0.6328861713409424,
      "step": 493
    },
    {
      "completion_length": 22.65625,
      "epoch": 1.5161290322580645,
      "grad_norm": 13.911900520324707,
      "kl": 0.33984375,
      "learning_rate": 6.207692307692307e-07,
      "loss": 0.0682,
      "reward": 1.4692556262016296,
      "reward_std": 0.2184278815984726,
      "reward_std_all": 0.2058429941534996,
      "rewards-max/forensics_sinkhorn_reward": 0.9780024886131287,
      "rewards-max/forensics_strict_edge_reward": 0.170194111764431,
      "rewards-max/forensics_tiou_reward": 0.9670609533786774,
      "rewards/forensics_sinkhorn_reward": 0.8258381187915802,
      "rewards/forensics_strict_edge_reward": 0.019858751446008682,
      "rewards/forensics_tiou_reward": 0.6235587000846863,
      "step": 494
    },
    {
      "completion_length": 22.4375,
      "epoch": 1.5192012288786483,
      "grad_norm": 18.320669174194336,
      "kl": 0.375,
      "learning_rate": 6.2e-07,
      "loss": 0.1531,
      "reward": 1.360410988330841,
      "reward_std": 0.21559783816337585,
      "reward_std_all": 0.22876106947660446,
      "rewards-max/forensics_sinkhorn_reward": 0.9678080677986145,
      "rewards-max/forensics_strict_edge_reward": 0.36725103110074997,
      "rewards-max/forensics_tiou_reward": 0.9355977475643158,
      "rewards/forensics_sinkhorn_reward": 0.76797816157341,
      "rewards/forensics_strict_edge_reward": 0.0365236543584615,
      "rewards/forensics_tiou_reward": 0.5559091866016388,
      "step": 495
    },
    {
      "completion_length": 19.53125,
      "epoch": 1.522273425499232,
      "grad_norm": 27.023574829101562,
      "kl": 0.25439453125,
      "learning_rate": 6.192307692307692e-07,
      "loss": -0.0165,
      "reward": 1.337058663368225,
      "reward_std": 0.1521497331559658,
      "reward_std_all": 0.14729534834623337,
      "rewards-max/forensics_sinkhorn_reward": 0.9547362327575684,
      "rewards-max/forensics_strict_edge_reward": 0.3510731253772974,
      "rewards-max/forensics_tiou_reward": 0.9079771339893341,
      "rewards/forensics_sinkhorn_reward": 0.7572006583213806,
      "rewards/forensics_strict_edge_reward": 0.06287168990820646,
      "rewards/forensics_tiou_reward": 0.5169863104820251,
      "step": 496
    },
    {
      "completion_length": 21.234375,
      "epoch": 1.5253456221198156,
      "grad_norm": 17.644939422607422,
      "kl": 0.8388671875,
      "learning_rate": 6.184615384615385e-07,
      "loss": -0.0271,
      "reward": 1.0931205749511719,
      "reward_std": 0.20541595667600632,
      "reward_std_all": 0.19050581753253937,
      "rewards-max/forensics_sinkhorn_reward": 0.9448310136795044,
      "rewards-max/forensics_strict_edge_reward": 0.3675718456506729,
      "rewards-max/forensics_tiou_reward": 0.8886628746986389,
      "rewards/forensics_sinkhorn_reward": 0.6682659387588501,
      "rewards/forensics_strict_edge_reward": 0.025196385569870472,
      "rewards/forensics_tiou_reward": 0.399658203125,
      "step": 497
    },
    {
      "completion_length": 22.1875,
      "epoch": 1.5284178187403994,
      "grad_norm": 7.802219390869141,
      "kl": 0.25927734375,
      "learning_rate": 6.176923076923076e-07,
      "loss": -0.0144,
      "reward": 1.2220049798488617,
      "reward_std": 0.17476020008325577,
      "reward_std_all": 0.1970406323671341,
      "rewards-max/forensics_sinkhorn_reward": 0.951175719499588,
      "rewards-max/forensics_strict_edge_reward": 0.29067233204841614,
      "rewards-max/forensics_tiou_reward": 0.9030197858810425,
      "rewards/forensics_sinkhorn_reward": 0.6995905041694641,
      "rewards/forensics_strict_edge_reward": 0.022273576818406582,
      "rewards/forensics_tiou_reward": 0.5001408457756042,
      "step": 498
    },
    {
      "completion_length": 20.46875,
      "epoch": 1.531490015360983,
      "grad_norm": 13.743690490722656,
      "kl": 0.28466796875,
      "learning_rate": 6.169230769230768e-07,
      "loss": -0.0458,
      "reward": 1.0810023248195648,
      "reward_std": 0.136513352394104,
      "reward_std_all": 0.12635993212461472,
      "rewards-max/forensics_sinkhorn_reward": 0.9099813997745514,
      "rewards-max/forensics_strict_edge_reward": 0.1800208858330734,
      "rewards-max/forensics_tiou_reward": 0.7897676825523376,
      "rewards/forensics_sinkhorn_reward": 0.6490888893604279,
      "rewards/forensics_strict_edge_reward": 0.019112079076876398,
      "rewards/forensics_tiou_reward": 0.4128013700246811,
      "step": 499
    },
    {
      "completion_length": 22.96875,
      "epoch": 1.5345622119815667,
      "grad_norm": 10.538108825683594,
      "kl": 0.32421875,
      "learning_rate": 6.161538461538462e-07,
      "loss": 0.0779,
      "reward": 1.1576806902885437,
      "reward_std": 0.17447033524513245,
      "reward_std_all": 0.16256528347730637,
      "rewards-max/forensics_sinkhorn_reward": 0.9548419713973999,
      "rewards-max/forensics_strict_edge_reward": 0.14895004034042358,
      "rewards-max/forensics_tiou_reward": 0.9049136936664581,
      "rewards/forensics_sinkhorn_reward": 0.7010202407836914,
      "rewards/forensics_strict_edge_reward": 0.01543593592941761,
      "rewards/forensics_tiou_reward": 0.44122448563575745,
      "step": 500
    },
    {
      "completion_length": 20.90625,
      "epoch": 1.5376344086021505,
      "grad_norm": 10.620134353637695,
      "kl": 0.46875,
      "learning_rate": 6.153846153846154e-07,
      "loss": 0.0532,
      "reward": 1.1330663561820984,
      "reward_std": 0.21362370252609253,
      "reward_std_all": 0.20505119115114212,
      "rewards-max/forensics_sinkhorn_reward": 0.9475798010826111,
      "rewards-max/forensics_strict_edge_reward": 0.19414592161774635,
      "rewards-max/forensics_tiou_reward": 0.8913578987121582,
      "rewards/forensics_sinkhorn_reward": 0.6586280167102814,
      "rewards/forensics_strict_edge_reward": 0.01955655822530389,
      "rewards/forensics_tiou_reward": 0.45488180220127106,
      "step": 501
    },
    {
      "completion_length": 23.34375,
      "epoch": 1.5407066052227343,
      "grad_norm": 14.412861824035645,
      "kl": 0.20458984375,
      "learning_rate": 6.146153846153845e-07,
      "loss": -0.0357,
      "reward": 1.165528118610382,
      "reward_std": 0.32805484533309937,
      "reward_std_all": 0.27948759496212006,
      "rewards-max/forensics_sinkhorn_reward": 0.9552718102931976,
      "rewards-max/forensics_strict_edge_reward": 0.5740037858486176,
      "rewards-max/forensics_tiou_reward": 0.9206673502922058,
      "rewards/forensics_sinkhorn_reward": 0.6772407293319702,
      "rewards/forensics_strict_edge_reward": 0.030359099619090557,
      "rewards/forensics_tiou_reward": 0.4579283148050308,
      "step": 502
    },
    {
      "completion_length": 23.546875,
      "epoch": 1.543778801843318,
      "grad_norm": 15.459839820861816,
      "kl": 0.1650390625,
      "learning_rate": 6.138461538461538e-07,
      "loss": -0.0317,
      "reward": 1.2738066911697388,
      "reward_std": 0.19336548447608948,
      "reward_std_all": 0.17512403428554535,
      "rewards-max/forensics_sinkhorn_reward": 0.9490817487239838,
      "rewards-max/forensics_strict_edge_reward": 0.30740495026111603,
      "rewards-max/forensics_tiou_reward": 0.8879601359367371,
      "rewards/forensics_sinkhorn_reward": 0.7245217263698578,
      "rewards/forensics_strict_edge_reward": 0.02676130272448063,
      "rewards/forensics_tiou_reward": 0.5225235819816589,
      "step": 503
    },
    {
      "completion_length": 23.1875,
      "epoch": 1.5468509984639018,
      "grad_norm": 12.595788955688477,
      "kl": 0.3857421875,
      "learning_rate": 6.13076923076923e-07,
      "loss": 0.0116,
      "reward": 1.1078209578990936,
      "reward_std": 0.2543361634016037,
      "reward_std_all": 0.23708035796880722,
      "rewards-max/forensics_sinkhorn_reward": 0.9233500361442566,
      "rewards-max/forensics_strict_edge_reward": 0.23790093138813972,
      "rewards-max/forensics_tiou_reward": 0.818170964717865,
      "rewards/forensics_sinkhorn_reward": 0.6738016903400421,
      "rewards/forensics_strict_edge_reward": 0.012270884355530143,
      "rewards/forensics_tiou_reward": 0.42174847424030304,
      "step": 504
    },
    {
      "completion_length": 21.140625,
      "epoch": 1.5499231950844854,
      "grad_norm": 15.549936294555664,
      "kl": 0.21630859375,
      "learning_rate": 6.123076923076923e-07,
      "loss": -0.0387,
      "reward": 1.2237553596496582,
      "reward_std": 0.1834341585636139,
      "reward_std_all": 0.1790766417980194,
      "rewards-max/forensics_sinkhorn_reward": 0.9265526831150055,
      "rewards-max/forensics_strict_edge_reward": 0.40184471383690834,
      "rewards-max/forensics_tiou_reward": 0.844671905040741,
      "rewards/forensics_sinkhorn_reward": 0.7135514318943024,
      "rewards/forensics_strict_edge_reward": 0.02733530104160309,
      "rewards/forensics_tiou_reward": 0.4828687459230423,
      "step": 505
    },
    {
      "completion_length": 20.90625,
      "epoch": 1.5529953917050692,
      "grad_norm": 12.245729446411133,
      "kl": 0.216796875,
      "learning_rate": 6.115384615384616e-07,
      "loss": -0.0677,
      "reward": 1.415494978427887,
      "reward_std": 0.23006313294172287,
      "reward_std_all": 0.211070254445076,
      "rewards-max/forensics_sinkhorn_reward": 0.9589565694332123,
      "rewards-max/forensics_strict_edge_reward": 0.4194317013025284,
      "rewards-max/forensics_tiou_reward": 0.9233607649803162,
      "rewards/forensics_sinkhorn_reward": 0.7762571275234222,
      "rewards/forensics_strict_edge_reward": 0.05524395778775215,
      "rewards/forensics_tiou_reward": 0.5839938074350357,
      "step": 506
    },
    {
      "completion_length": 23.28125,
      "epoch": 1.5560675883256527,
      "grad_norm": 10.145088195800781,
      "kl": 0.259765625,
      "learning_rate": 6.107692307692307e-07,
      "loss": -0.0003,
      "reward": 1.1393388509750366,
      "reward_std": 0.15708191692829132,
      "reward_std_all": 0.16253402084112167,
      "rewards-max/forensics_sinkhorn_reward": 0.9561820924282074,
      "rewards-max/forensics_strict_edge_reward": 0.15248337760567665,
      "rewards-max/forensics_tiou_reward": 0.912432074546814,
      "rewards/forensics_sinkhorn_reward": 0.6528217792510986,
      "rewards/forensics_strict_edge_reward": 0.012537606991827488,
      "rewards/forensics_tiou_reward": 0.4739794135093689,
      "step": 507
    },
    {
      "completion_length": 23.59375,
      "epoch": 1.5591397849462365,
      "grad_norm": 10.774646759033203,
      "kl": 0.3408203125,
      "learning_rate": 6.1e-07,
      "loss": 0.0145,
      "reward": 1.2186440229415894,
      "reward_std": 0.23784637451171875,
      "reward_std_all": 0.23183146864175797,
      "rewards-max/forensics_sinkhorn_reward": 0.8790186643600464,
      "rewards-max/forensics_strict_edge_reward": 0.26192018017172813,
      "rewards-max/forensics_tiou_reward": 0.7545689642429352,
      "rewards/forensics_sinkhorn_reward": 0.6883969902992249,
      "rewards/forensics_strict_edge_reward": 0.04449078382458538,
      "rewards/forensics_tiou_reward": 0.4857562631368637,
      "step": 508
    },
    {
      "completion_length": 22.03125,
      "epoch": 1.5622119815668203,
      "grad_norm": 8.765377044677734,
      "kl": 0.30712890625,
      "learning_rate": 6.092307692307692e-07,
      "loss": -0.0117,
      "reward": 1.3089193105697632,
      "reward_std": 0.2936377450823784,
      "reward_std_all": 0.2948858439922333,
      "rewards-max/forensics_sinkhorn_reward": 0.9726421236991882,
      "rewards-max/forensics_strict_edge_reward": 0.49388203024864197,
      "rewards-max/forensics_tiou_reward": 0.9536155164241791,
      "rewards/forensics_sinkhorn_reward": 0.7309975624084473,
      "rewards/forensics_strict_edge_reward": 0.04351171292364597,
      "rewards/forensics_tiou_reward": 0.5344101786613464,
      "step": 509
    },
    {
      "completion_length": 20.859375,
      "epoch": 1.565284178187404,
      "grad_norm": 13.163751602172852,
      "kl": 0.20458984375,
      "learning_rate": 6.084615384615385e-07,
      "loss": -0.0154,
      "reward": 1.2984651923179626,
      "reward_std": 0.1814587190747261,
      "reward_std_all": 0.17620689421892166,
      "rewards-max/forensics_sinkhorn_reward": 0.9506383538246155,
      "rewards-max/forensics_strict_edge_reward": 0.4399845600128174,
      "rewards-max/forensics_tiou_reward": 0.9113486409187317,
      "rewards/forensics_sinkhorn_reward": 0.7263664305210114,
      "rewards/forensics_strict_edge_reward": 0.05063055083155632,
      "rewards/forensics_tiou_reward": 0.5214682221412659,
      "step": 510
    },
    {
      "completion_length": 22.4375,
      "epoch": 1.5683563748079878,
      "grad_norm": 9.621005058288574,
      "kl": 0.26318359375,
      "learning_rate": 6.076923076923076e-07,
      "loss": -0.0011,
      "reward": 1.2346991300582886,
      "reward_std": 0.2156239002943039,
      "reward_std_all": 0.21090500801801682,
      "rewards-max/forensics_sinkhorn_reward": 0.9711408317089081,
      "rewards-max/forensics_strict_edge_reward": 0.15748992410954088,
      "rewards-max/forensics_tiou_reward": 0.9413226842880249,
      "rewards/forensics_sinkhorn_reward": 0.7290220856666565,
      "rewards/forensics_strict_edge_reward": 0.009617005809559487,
      "rewards/forensics_tiou_reward": 0.4960600584745407,
      "step": 511
    },
    {
      "completion_length": 19.609375,
      "epoch": 1.5714285714285714,
      "grad_norm": 23.611875534057617,
      "kl": 0.23486328125,
      "learning_rate": 6.069230769230768e-07,
      "loss": -0.1041,
      "reward": 1.2319672107696533,
      "reward_std": 0.25199056416749954,
      "reward_std_all": 0.22621184587478638,
      "rewards-max/forensics_sinkhorn_reward": 0.9419965147972107,
      "rewards-max/forensics_strict_edge_reward": 0.117774847894907,
      "rewards-max/forensics_tiou_reward": 0.8873976171016693,
      "rewards/forensics_sinkhorn_reward": 0.7216283977031708,
      "rewards/forensics_strict_edge_reward": 0.01081992220133543,
      "rewards/forensics_tiou_reward": 0.49951887130737305,
      "step": 512
    },
    {
      "completion_length": 20.484375,
      "epoch": 1.5745007680491552,
      "grad_norm": 16.98284149169922,
      "kl": 0.23388671875,
      "learning_rate": 6.061538461538462e-07,
      "loss": 0.1375,
      "reward": 1.1116089224815369,
      "reward_std": 0.20941230654716492,
      "reward_std_all": 0.197811558842659,
      "rewards-max/forensics_sinkhorn_reward": 0.8952511847019196,
      "rewards-max/forensics_strict_edge_reward": 0.16677918657660484,
      "rewards-max/forensics_tiou_reward": 0.7700204849243164,
      "rewards/forensics_sinkhorn_reward": 0.6515493988990784,
      "rewards/forensics_strict_edge_reward": 0.013993658940307796,
      "rewards/forensics_tiou_reward": 0.44606584310531616,
      "step": 513
    },
    {
      "completion_length": 22.78125,
      "epoch": 1.5775729646697387,
      "grad_norm": 9.854836463928223,
      "kl": 0.1865234375,
      "learning_rate": 6.053846153846154e-07,
      "loss": -0.0254,
      "reward": 1.0950818061828613,
      "reward_std": 0.23803502321243286,
      "reward_std_all": 0.2238319367170334,
      "rewards-max/forensics_sinkhorn_reward": 0.9793860018253326,
      "rewards-max/forensics_strict_edge_reward": 0.268596775829792,
      "rewards-max/forensics_tiou_reward": 0.9737639129161835,
      "rewards/forensics_sinkhorn_reward": 0.6534618139266968,
      "rewards/forensics_strict_edge_reward": 0.016712689655832946,
      "rewards/forensics_tiou_reward": 0.4249073266983032,
      "step": 514
    },
    {
      "completion_length": 23.28125,
      "epoch": 1.5806451612903225,
      "grad_norm": 10.411685943603516,
      "kl": 0.236328125,
      "learning_rate": 6.046153846153847e-07,
      "loss": -0.0172,
      "reward": 1.1002243161201477,
      "reward_std": 0.17362593114376068,
      "reward_std_all": 0.17512069270014763,
      "rewards-max/forensics_sinkhorn_reward": 0.9652658104896545,
      "rewards-max/forensics_strict_edge_reward": 0.31501996517181396,
      "rewards-max/forensics_tiou_reward": 0.9364388287067413,
      "rewards/forensics_sinkhorn_reward": 0.6589441001415253,
      "rewards/forensics_strict_edge_reward": 0.01629102323204279,
      "rewards/forensics_tiou_reward": 0.42498910427093506,
      "step": 515
    },
    {
      "completion_length": 23.375,
      "epoch": 1.5837173579109063,
      "grad_norm": 9.404751777648926,
      "kl": 0.3671875,
      "learning_rate": 6.038461538461538e-07,
      "loss": 0.0117,
      "reward": 1.4077009558677673,
      "reward_std": 0.26597078889608383,
      "reward_std_all": 0.24169348925352097,
      "rewards-max/forensics_sinkhorn_reward": 0.9810940325260162,
      "rewards-max/forensics_strict_edge_reward": 0.6285537928342819,
      "rewards-max/forensics_tiou_reward": 0.9820482432842255,
      "rewards/forensics_sinkhorn_reward": 0.7482607066631317,
      "rewards/forensics_strict_edge_reward": 0.09065477550029755,
      "rewards/forensics_tiou_reward": 0.5687854588031769,
      "step": 516
    },
    {
      "completion_length": 19.953125,
      "epoch": 1.58678955453149,
      "grad_norm": 10.78134536743164,
      "kl": 0.4716796875,
      "learning_rate": 6.03076923076923e-07,
      "loss": 0.0154,
      "reward": 1.266646385192871,
      "reward_std": 0.2479756772518158,
      "reward_std_all": 0.22172128409147263,
      "rewards-max/forensics_sinkhorn_reward": 0.9443074762821198,
      "rewards-max/forensics_strict_edge_reward": 0.44867993146181107,
      "rewards-max/forensics_tiou_reward": 0.898041307926178,
      "rewards/forensics_sinkhorn_reward": 0.731070339679718,
      "rewards/forensics_strict_edge_reward": 0.022863379679620266,
      "rewards/forensics_tiou_reward": 0.5127127170562744,
      "step": 517
    },
    {
      "completion_length": 21.515625,
      "epoch": 1.5898617511520738,
      "grad_norm": 9.093104362487793,
      "kl": 0.25146484375,
      "learning_rate": 6.023076923076923e-07,
      "loss": -0.0206,
      "reward": 1.0348658561706543,
      "reward_std": 0.22618557512760162,
      "reward_std_all": 0.20786834508180618,
      "rewards-max/forensics_sinkhorn_reward": 0.9655823111534119,
      "rewards-max/forensics_strict_edge_reward": 0.43239868991076946,
      "rewards-max/forensics_tiou_reward": 0.9151565432548523,
      "rewards/forensics_sinkhorn_reward": 0.6440733075141907,
      "rewards/forensics_strict_edge_reward": 0.019023898261366412,
      "rewards/forensics_tiou_reward": 0.37176869809627533,
      "step": 518
    },
    {
      "completion_length": 20.046875,
      "epoch": 1.5929339477726574,
      "grad_norm": 9.376124382019043,
      "kl": 0.21484375,
      "learning_rate": 6.015384615384615e-07,
      "loss": 0.1056,
      "reward": 1.1832271218299866,
      "reward_std": 0.18080775439739227,
      "reward_std_all": 0.15819590538740158,
      "rewards-max/forensics_sinkhorn_reward": 0.9647790491580963,
      "rewards-max/forensics_strict_edge_reward": 0.09100023121573031,
      "rewards-max/forensics_tiou_reward": 0.9279188811779022,
      "rewards/forensics_sinkhorn_reward": 0.7031856775283813,
      "rewards/forensics_strict_edge_reward": 0.009779770305613056,
      "rewards/forensics_tiou_reward": 0.47026166319847107,
      "step": 519
    },
    {
      "completion_length": 22.328125,
      "epoch": 1.5960061443932412,
      "grad_norm": 11.873438835144043,
      "kl": 0.28173828125,
      "learning_rate": 6.007692307692307e-07,
      "loss": 0.1475,
      "reward": 0.9427313208580017,
      "reward_std": 0.2245497703552246,
      "reward_std_all": 0.21146316081285477,
      "rewards-max/forensics_sinkhorn_reward": 0.9362269937992096,
      "rewards-max/forensics_strict_edge_reward": 0.2960040234029293,
      "rewards-max/forensics_tiou_reward": 0.872622162103653,
      "rewards/forensics_sinkhorn_reward": 0.6077515184879303,
      "rewards/forensics_strict_edge_reward": 0.01482343440875411,
      "rewards/forensics_tiou_reward": 0.32015635073184967,
      "step": 520
    },
    {
      "completion_length": 22.28125,
      "epoch": 1.5990783410138247,
      "grad_norm": 10.33852481842041,
      "kl": 0.353515625,
      "learning_rate": 6e-07,
      "loss": 0.1178,
      "reward": 0.9917675256729126,
      "reward_std": 0.2050793692469597,
      "reward_std_all": 0.2207036241889,
      "rewards-max/forensics_sinkhorn_reward": 0.9008928835391998,
      "rewards-max/forensics_strict_edge_reward": 0.3087383806705475,
      "rewards-max/forensics_tiou_reward": 0.7974729537963867,
      "rewards/forensics_sinkhorn_reward": 0.606680154800415,
      "rewards/forensics_strict_edge_reward": 0.03063889779150486,
      "rewards/forensics_tiou_reward": 0.35444845259189606,
      "step": 521
    },
    {
      "completion_length": 19.890625,
      "epoch": 1.6021505376344085,
      "grad_norm": 13.068949699401855,
      "kl": 0.2060546875,
      "learning_rate": 5.992307692307692e-07,
      "loss": 0.016,
      "reward": 1.1550803780555725,
      "reward_std": 0.166810505092144,
      "reward_std_all": 0.17766761034727097,
      "rewards-max/forensics_sinkhorn_reward": 0.9500811100006104,
      "rewards-max/forensics_strict_edge_reward": 0.06517261825501919,
      "rewards-max/forensics_tiou_reward": 0.8848908543586731,
      "rewards/forensics_sinkhorn_reward": 0.6925579011440277,
      "rewards/forensics_strict_edge_reward": 0.005473710130900145,
      "rewards/forensics_tiou_reward": 0.45704881846904755,
      "step": 522
    },
    {
      "completion_length": 22.953125,
      "epoch": 1.6052227342549923,
      "grad_norm": 10.16849136352539,
      "kl": 0.4189453125,
      "learning_rate": 5.984615384615385e-07,
      "loss": -0.0318,
      "reward": 1.252382755279541,
      "reward_std": 0.2783624082803726,
      "reward_std_all": 0.256881944835186,
      "rewards-max/forensics_sinkhorn_reward": 0.9700296223163605,
      "rewards-max/forensics_strict_edge_reward": 0.31535862013697624,
      "rewards-max/forensics_tiou_reward": 0.9504742920398712,
      "rewards/forensics_sinkhorn_reward": 0.7062767148017883,
      "rewards/forensics_strict_edge_reward": 0.029301579110324383,
      "rewards/forensics_tiou_reward": 0.5168044567108154,
      "step": 523
    },
    {
      "completion_length": 21.1875,
      "epoch": 1.608294930875576,
      "grad_norm": 10.941844940185547,
      "kl": 0.357421875,
      "learning_rate": 5.976923076923077e-07,
      "loss": -0.0044,
      "reward": 1.2641863822937012,
      "reward_std": 0.27076996117830276,
      "reward_std_all": 0.28455668687820435,
      "rewards-max/forensics_sinkhorn_reward": 0.9484951496124268,
      "rewards-max/forensics_strict_edge_reward": 0.3408282548189163,
      "rewards-max/forensics_tiou_reward": 0.8866868615150452,
      "rewards/forensics_sinkhorn_reward": 0.7345790565013885,
      "rewards/forensics_strict_edge_reward": 0.030517471954226494,
      "rewards/forensics_tiou_reward": 0.49908994138240814,
      "step": 524
    },
    {
      "completion_length": 20.75,
      "epoch": 1.6113671274961598,
      "grad_norm": 11.944144248962402,
      "kl": 0.19921875,
      "learning_rate": 5.969230769230768e-07,
      "loss": -0.0019,
      "reward": 1.0712591409683228,
      "reward_std": 0.13567021116614342,
      "reward_std_all": 0.17614369094371796,
      "rewards-max/forensics_sinkhorn_reward": 0.954298734664917,
      "rewards-max/forensics_strict_edge_reward": 0.16434098035097122,
      "rewards-max/forensics_tiou_reward": 0.9255762696266174,
      "rewards/forensics_sinkhorn_reward": 0.6596543490886688,
      "rewards/forensics_strict_edge_reward": 0.020594049943611026,
      "rewards/forensics_tiou_reward": 0.39101073145866394,
      "step": 525
    },
    {
      "completion_length": 23.078125,
      "epoch": 1.6144393241167436,
      "grad_norm": 12.15312385559082,
      "kl": 0.5341796875,
      "learning_rate": 5.961538461538461e-07,
      "loss": 0.0582,
      "reward": 1.2438649535179138,
      "reward_std": 0.26230818033218384,
      "reward_std_all": 0.25841179490089417,
      "rewards-max/forensics_sinkhorn_reward": 0.9406732320785522,
      "rewards-max/forensics_strict_edge_reward": 0.3886200934648514,
      "rewards-max/forensics_tiou_reward": 0.8747038841247559,
      "rewards/forensics_sinkhorn_reward": 0.721014529466629,
      "rewards/forensics_strict_edge_reward": 0.03726332634687424,
      "rewards/forensics_tiou_reward": 0.48558710515499115,
      "step": 526
    },
    {
      "completion_length": 21.03125,
      "epoch": 1.6175115207373272,
      "grad_norm": 10.581786155700684,
      "kl": 0.5126953125,
      "learning_rate": 5.953846153846154e-07,
      "loss": -0.1525,
      "reward": 1.4278141260147095,
      "reward_std": 0.34799759089946747,
      "reward_std_all": 0.2981245368719101,
      "rewards-max/forensics_sinkhorn_reward": 0.9768854081630707,
      "rewards-max/forensics_strict_edge_reward": 0.5424859970808029,
      "rewards-max/forensics_tiou_reward": 0.9675743579864502,
      "rewards/forensics_sinkhorn_reward": 0.7636292576789856,
      "rewards/forensics_strict_edge_reward": 0.07093942165374756,
      "rewards/forensics_tiou_reward": 0.5932454466819763,
      "step": 527
    },
    {
      "completion_length": 19.625,
      "epoch": 1.620583717357911,
      "grad_norm": 12.739923477172852,
      "kl": 0.2119140625,
      "learning_rate": 5.946153846153847e-07,
      "loss": 0.036,
      "reward": 1.2464459538459778,
      "reward_std": 0.24720560014247894,
      "reward_std_all": 0.2270696684718132,
      "rewards-max/forensics_sinkhorn_reward": 0.967763215303421,
      "rewards-max/forensics_strict_edge_reward": 0.4086042046546936,
      "rewards-max/forensics_tiou_reward": 0.9465148448944092,
      "rewards/forensics_sinkhorn_reward": 0.7201306223869324,
      "rewards/forensics_strict_edge_reward": 0.048211922869086266,
      "rewards/forensics_tiou_reward": 0.478103369474411,
      "step": 528
    },
    {
      "completion_length": 20.359375,
      "epoch": 1.6236559139784945,
      "grad_norm": 17.938270568847656,
      "kl": 0.2421875,
      "learning_rate": 5.938461538461538e-07,
      "loss": 0.0719,
      "reward": 1.2063761353492737,
      "reward_std": 0.1533328779041767,
      "reward_std_all": 0.15919344127178192,
      "rewards-max/forensics_sinkhorn_reward": 0.9243444800376892,
      "rewards-max/forensics_strict_edge_reward": 0.08198325778357685,
      "rewards-max/forensics_tiou_reward": 0.8317241966724396,
      "rewards/forensics_sinkhorn_reward": 0.7125316560268402,
      "rewards/forensics_strict_edge_reward": 0.006412848655600101,
      "rewards/forensics_tiou_reward": 0.4874316304922104,
      "step": 529
    },
    {
      "completion_length": 19.515625,
      "epoch": 1.6267281105990783,
      "grad_norm": 8.16577434539795,
      "kl": 0.27099609375,
      "learning_rate": 5.93076923076923e-07,
      "loss": 0.1192,
      "reward": 1.3018345832824707,
      "reward_std": 0.2994176000356674,
      "reward_std_all": 0.26141151785850525,
      "rewards-max/forensics_sinkhorn_reward": 0.9781320095062256,
      "rewards-max/forensics_strict_edge_reward": 0.3975932002067566,
      "rewards-max/forensics_tiou_reward": 0.970177412033081,
      "rewards/forensics_sinkhorn_reward": 0.7435807883739471,
      "rewards/forensics_strict_edge_reward": 0.03233805485069752,
      "rewards/forensics_tiou_reward": 0.5259156376123428,
      "step": 530
    },
    {
      "completion_length": 21.328125,
      "epoch": 1.629800307219662,
      "grad_norm": 7.218894958496094,
      "kl": 0.25390625,
      "learning_rate": 5.923076923076923e-07,
      "loss": 0.0098,
      "reward": 1.1315318942070007,
      "reward_std": 0.29022228717803955,
      "reward_std_all": 0.2518443092703819,
      "rewards-max/forensics_sinkhorn_reward": 0.9548200964927673,
      "rewards-max/forensics_strict_edge_reward": 0.3483770936727524,
      "rewards-max/forensics_tiou_reward": 0.8905685842037201,
      "rewards/forensics_sinkhorn_reward": 0.6620894074440002,
      "rewards/forensics_strict_edge_reward": 0.031587358098477125,
      "rewards/forensics_tiou_reward": 0.437855064868927,
      "step": 531
    },
    {
      "completion_length": 22.3125,
      "epoch": 1.6328725038402458,
      "grad_norm": 8.083393096923828,
      "kl": 0.251953125,
      "learning_rate": 5.915384615384615e-07,
      "loss": 0.0535,
      "reward": 1.1788126826286316,
      "reward_std": 0.16574358940124512,
      "reward_std_all": 0.17348363250494003,
      "rewards-max/forensics_sinkhorn_reward": 0.9389488399028778,
      "rewards-max/forensics_strict_edge_reward": 0.18477749824523926,
      "rewards-max/forensics_tiou_reward": 0.8779317140579224,
      "rewards/forensics_sinkhorn_reward": 0.703342467546463,
      "rewards/forensics_strict_edge_reward": 0.014536004979163408,
      "rewards/forensics_tiou_reward": 0.4609341472387314,
      "step": 532
    },
    {
      "completion_length": 20.9375,
      "epoch": 1.6359447004608296,
      "grad_norm": 10.50774097442627,
      "kl": 0.25927734375,
      "learning_rate": 5.907692307692307e-07,
      "loss": 0.1146,
      "reward": 1.2555232048034668,
      "reward_std": 0.3101675733923912,
      "reward_std_all": 0.2758825272321701,
      "rewards-max/forensics_sinkhorn_reward": 0.958746999502182,
      "rewards-max/forensics_strict_edge_reward": 0.5961077809333801,
      "rewards-max/forensics_tiou_reward": 0.9248098134994507,
      "rewards/forensics_sinkhorn_reward": 0.7128961682319641,
      "rewards/forensics_strict_edge_reward": 0.05879840441048145,
      "rewards/forensics_tiou_reward": 0.4838285744190216,
      "step": 533
    },
    {
      "completion_length": 19.28125,
      "epoch": 1.6390168970814132,
      "grad_norm": 13.520086288452148,
      "kl": 0.22998046875,
      "learning_rate": 5.9e-07,
      "loss": 0.0306,
      "reward": 1.136740803718567,
      "reward_std": 0.18375566601753235,
      "reward_std_all": 0.18112344294786453,
      "rewards-max/forensics_sinkhorn_reward": 0.951974630355835,
      "rewards-max/forensics_strict_edge_reward": 0.2644407991319895,
      "rewards-max/forensics_tiou_reward": 0.8712061047554016,
      "rewards/forensics_sinkhorn_reward": 0.7033355832099915,
      "rewards/forensics_strict_edge_reward": 0.016017587622627616,
      "rewards/forensics_tiou_reward": 0.4173875004053116,
      "step": 534
    },
    {
      "completion_length": 20.5,
      "epoch": 1.642089093701997,
      "grad_norm": 15.745689392089844,
      "kl": 0.29638671875,
      "learning_rate": 5.892307692307692e-07,
      "loss": -0.0293,
      "reward": 1.137219250202179,
      "reward_std": 0.2620953172445297,
      "reward_std_all": 0.23247270286083221,
      "rewards-max/forensics_sinkhorn_reward": 0.9736090302467346,
      "rewards-max/forensics_strict_edge_reward": 0.2522459824103862,
      "rewards-max/forensics_tiou_reward": 0.953399121761322,
      "rewards/forensics_sinkhorn_reward": 0.6959372460842133,
      "rewards/forensics_strict_edge_reward": 0.026763569796457887,
      "rewards/forensics_tiou_reward": 0.41451847553253174,
      "step": 535
    },
    {
      "completion_length": 19.59375,
      "epoch": 1.6451612903225805,
      "grad_norm": 9.620757102966309,
      "kl": 0.35498046875,
      "learning_rate": 5.884615384615385e-07,
      "loss": 0.0393,
      "reward": 1.0895508527755737,
      "reward_std": 0.2828332558274269,
      "reward_std_all": 0.23963763564825058,
      "rewards-max/forensics_sinkhorn_reward": 0.8529053926467896,
      "rewards-max/forensics_strict_edge_reward": 0.4900130182504654,
      "rewards-max/forensics_tiou_reward": 0.8179589211940765,
      "rewards/forensics_sinkhorn_reward": 0.6335588991641998,
      "rewards/forensics_strict_edge_reward": 0.03066061995923519,
      "rewards/forensics_tiou_reward": 0.42533135414123535,
      "step": 536
    },
    {
      "completion_length": 21.75,
      "epoch": 1.6482334869431643,
      "grad_norm": 10.633550643920898,
      "kl": 0.24658203125,
      "learning_rate": 5.876923076923077e-07,
      "loss": 0.0579,
      "reward": 1.3088518977165222,
      "reward_std": 0.2832607477903366,
      "reward_std_all": 0.25877250730991364,
      "rewards-max/forensics_sinkhorn_reward": 0.9443126618862152,
      "rewards-max/forensics_strict_edge_reward": 0.2075769528746605,
      "rewards-max/forensics_tiou_reward": 0.8787614107131958,
      "rewards/forensics_sinkhorn_reward": 0.7586058676242828,
      "rewards/forensics_strict_edge_reward": 0.02154265041463077,
      "rewards/forensics_tiou_reward": 0.5287034064531326,
      "step": 537
    },
    {
      "completion_length": 21.34375,
      "epoch": 1.651305683563748,
      "grad_norm": 9.32331371307373,
      "kl": 0.2626953125,
      "learning_rate": 5.869230769230768e-07,
      "loss": 0.1457,
      "reward": 1.22303706407547,
      "reward_std": 0.28907962143421173,
      "reward_std_all": 0.25442859530448914,
      "rewards-max/forensics_sinkhorn_reward": 0.9529632925987244,
      "rewards-max/forensics_strict_edge_reward": 0.33223120123147964,
      "rewards-max/forensics_tiou_reward": 0.9235785901546478,
      "rewards/forensics_sinkhorn_reward": 0.7213878035545349,
      "rewards/forensics_strict_edge_reward": 0.029543557669967413,
      "rewards/forensics_tiou_reward": 0.4721057862043381,
      "step": 538
    },
    {
      "completion_length": 22.71875,
      "epoch": 1.6543778801843319,
      "grad_norm": 5.873033046722412,
      "kl": 0.26708984375,
      "learning_rate": 5.861538461538461e-07,
      "loss": 0.0202,
      "reward": 1.1347864270210266,
      "reward_std": 0.2368001788854599,
      "reward_std_all": 0.213196761906147,
      "rewards-max/forensics_sinkhorn_reward": 0.9735862612724304,
      "rewards-max/forensics_strict_edge_reward": 0.4827088415622711,
      "rewards-max/forensics_tiou_reward": 0.9603069424629211,
      "rewards/forensics_sinkhorn_reward": 0.6652555465698242,
      "rewards/forensics_strict_edge_reward": 0.04636272042989731,
      "rewards/forensics_tiou_reward": 0.4231681823730469,
      "step": 539
    },
    {
      "completion_length": 22.078125,
      "epoch": 1.6574500768049156,
      "grad_norm": 14.922197341918945,
      "kl": 0.23681640625,
      "learning_rate": 5.853846153846154e-07,
      "loss": 0.0612,
      "reward": 1.2126555442810059,
      "reward_std": 0.22905675321817398,
      "reward_std_all": 0.19518860429525375,
      "rewards-max/forensics_sinkhorn_reward": 0.9586296081542969,
      "rewards-max/forensics_strict_edge_reward": 0.11832905001938343,
      "rewards-max/forensics_tiou_reward": 0.9303722679615021,
      "rewards/forensics_sinkhorn_reward": 0.7290668189525604,
      "rewards/forensics_strict_edge_reward": 0.010929750511422753,
      "rewards/forensics_tiou_reward": 0.47265900671482086,
      "step": 540
    },
    {
      "completion_length": 21.1875,
      "epoch": 1.6605222734254992,
      "grad_norm": 13.120329856872559,
      "kl": 0.306640625,
      "learning_rate": 5.846153846153847e-07,
      "loss": -0.0822,
      "reward": 1.402809500694275,
      "reward_std": 0.24940834939479828,
      "reward_std_all": 0.22914718836545944,
      "rewards-max/forensics_sinkhorn_reward": 0.9837201535701752,
      "rewards-max/forensics_strict_edge_reward": 0.49006423354148865,
      "rewards-max/forensics_tiou_reward": 0.9830417931079865,
      "rewards/forensics_sinkhorn_reward": 0.7768717408180237,
      "rewards/forensics_strict_edge_reward": 0.05629153735935688,
      "rewards/forensics_tiou_reward": 0.5696462094783783,
      "step": 541
    },
    {
      "completion_length": 22.203125,
      "epoch": 1.663594470046083,
      "grad_norm": 9.437113761901855,
      "kl": 0.25390625,
      "learning_rate": 5.838461538461538e-07,
      "loss": 0.0539,
      "reward": 1.4179590344429016,
      "reward_std": 0.27060094475746155,
      "reward_std_all": 0.2679833397269249,
      "rewards-max/forensics_sinkhorn_reward": 0.9778863787651062,
      "rewards-max/forensics_strict_edge_reward": 0.45597073435783386,
      "rewards-max/forensics_tiou_reward": 0.9732355773448944,
      "rewards/forensics_sinkhorn_reward": 0.7931120991706848,
      "rewards/forensics_strict_edge_reward": 0.03649954590946436,
      "rewards/forensics_tiou_reward": 0.5883474051952362,
      "step": 542
    },
    {
      "completion_length": 20.234375,
      "epoch": 1.6666666666666665,
      "grad_norm": 8.261310577392578,
      "kl": 0.2158203125,
      "learning_rate": 5.83076923076923e-07,
      "loss": 0.0881,
      "reward": 1.016453117132187,
      "reward_std": 0.2736287973821163,
      "reward_std_all": 0.24915676936507225,
      "rewards-max/forensics_sinkhorn_reward": 0.8921495079994202,
      "rewards-max/forensics_strict_edge_reward": 0.44391776248812675,
      "rewards-max/forensics_tiou_reward": 0.8112824261188507,
      "rewards/forensics_sinkhorn_reward": 0.6382209062576294,
      "rewards/forensics_strict_edge_reward": 0.02459246793296188,
      "rewards/forensics_tiou_reward": 0.35363972187042236,
      "step": 543
    },
    {
      "completion_length": 21.15625,
      "epoch": 1.6697388632872503,
      "grad_norm": 8.321229934692383,
      "kl": 0.22509765625,
      "learning_rate": 5.823076923076923e-07,
      "loss": 0.0072,
      "reward": 1.1724481582641602,
      "reward_std": 0.17937619984149933,
      "reward_std_all": 0.16288234293460846,
      "rewards-max/forensics_sinkhorn_reward": 0.9645018577575684,
      "rewards-max/forensics_strict_edge_reward": 0.099364441819489,
      "rewards-max/forensics_tiou_reward": 0.917212724685669,
      "rewards/forensics_sinkhorn_reward": 0.7088345885276794,
      "rewards/forensics_strict_edge_reward": 0.0090253398520872,
      "rewards/forensics_tiou_reward": 0.45458829402923584,
      "step": 544
    },
    {
      "completion_length": 20.140625,
      "epoch": 1.672811059907834,
      "grad_norm": 9.124037742614746,
      "kl": 0.26904296875,
      "learning_rate": 5.815384615384615e-07,
      "loss": 0.092,
      "reward": 1.133045732975006,
      "reward_std": 0.2485533505678177,
      "reward_std_all": 0.2186301425099373,
      "rewards-max/forensics_sinkhorn_reward": 0.8980592787265778,
      "rewards-max/forensics_strict_edge_reward": 0.10581059753894806,
      "rewards-max/forensics_tiou_reward": 0.7851063013076782,
      "rewards/forensics_sinkhorn_reward": 0.6857411563396454,
      "rewards/forensics_strict_edge_reward": 0.008459286647848785,
      "rewards/forensics_tiou_reward": 0.43884530663490295,
      "step": 545
    },
    {
      "completion_length": 20.59375,
      "epoch": 1.6758832565284179,
      "grad_norm": 15.354032516479492,
      "kl": 0.5068359375,
      "learning_rate": 5.807692307692307e-07,
      "loss": 0.0388,
      "reward": 1.4594656825065613,
      "reward_std": 0.30041807889938354,
      "reward_std_all": 0.27194972336292267,
      "rewards-max/forensics_sinkhorn_reward": 0.956404060125351,
      "rewards-max/forensics_strict_edge_reward": 0.3242585361003876,
      "rewards-max/forensics_tiou_reward": 0.9196287393569946,
      "rewards/forensics_sinkhorn_reward": 0.7923720180988312,
      "rewards/forensics_strict_edge_reward": 0.04438010323792696,
      "rewards/forensics_tiou_reward": 0.6227136850357056,
      "step": 546
    },
    {
      "completion_length": 21.125,
      "epoch": 1.6789554531490016,
      "grad_norm": 12.88104248046875,
      "kl": 0.24267578125,
      "learning_rate": 5.8e-07,
      "loss": -0.1287,
      "reward": 1.305091917514801,
      "reward_std": 0.2829591631889343,
      "reward_std_all": 0.2606077492237091,
      "rewards-max/forensics_sinkhorn_reward": 0.9755787253379822,
      "rewards-max/forensics_strict_edge_reward": 0.3332277685403824,
      "rewards-max/forensics_tiou_reward": 0.9576939344406128,
      "rewards/forensics_sinkhorn_reward": 0.7354705929756165,
      "rewards/forensics_strict_edge_reward": 0.04797344096004963,
      "rewards/forensics_tiou_reward": 0.5216479152441025,
      "step": 547
    },
    {
      "completion_length": 19.28125,
      "epoch": 1.6820276497695854,
      "grad_norm": 9.514456748962402,
      "kl": 0.330078125,
      "learning_rate": 5.792307692307692e-07,
      "loss": 0.0463,
      "reward": 1.105611503124237,
      "reward_std": 0.26422594487667084,
      "reward_std_all": 0.22640755772590637,
      "rewards-max/forensics_sinkhorn_reward": 0.9365506172180176,
      "rewards-max/forensics_strict_edge_reward": 0.1591176213696599,
      "rewards-max/forensics_tiou_reward": 0.862817794084549,
      "rewards/forensics_sinkhorn_reward": 0.6885176301002502,
      "rewards/forensics_strict_edge_reward": 0.007222860935144126,
      "rewards/forensics_tiou_reward": 0.40987108647823334,
      "step": 548
    },
    {
      "completion_length": 21.359375,
      "epoch": 1.685099846390169,
      "grad_norm": 14.723575592041016,
      "kl": 1.2783203125,
      "learning_rate": 5.784615384615385e-07,
      "loss": 0.0976,
      "reward": 1.234482228755951,
      "reward_std": 0.2634405791759491,
      "reward_std_all": 0.2637695372104645,
      "rewards-max/forensics_sinkhorn_reward": 0.9562912881374359,
      "rewards-max/forensics_strict_edge_reward": 0.34355300664901733,
      "rewards-max/forensics_tiou_reward": 0.9078639149665833,
      "rewards/forensics_sinkhorn_reward": 0.719034343957901,
      "rewards/forensics_strict_edge_reward": 0.02793254517018795,
      "rewards/forensics_tiou_reward": 0.4875153452157974,
      "step": 549
    },
    {
      "completion_length": 20.328125,
      "epoch": 1.6881720430107527,
      "grad_norm": 9.666937828063965,
      "kl": 0.18017578125,
      "learning_rate": 5.776923076923077e-07,
      "loss": -0.063,
      "reward": 1.2120141983032227,
      "reward_std": 0.21017514169216156,
      "reward_std_all": 0.18660513311624527,
      "rewards-max/forensics_sinkhorn_reward": 0.9580493867397308,
      "rewards-max/forensics_strict_edge_reward": 0.3157382905483246,
      "rewards-max/forensics_tiou_reward": 0.9171637296676636,
      "rewards/forensics_sinkhorn_reward": 0.7219112813472748,
      "rewards/forensics_strict_edge_reward": 0.020368915051221848,
      "rewards/forensics_tiou_reward": 0.46973390877246857,
      "step": 550
    },
    {
      "completion_length": 21.15625,
      "epoch": 1.6912442396313363,
      "grad_norm": 11.19339656829834,
      "kl": 0.603515625,
      "learning_rate": 5.769230769230768e-07,
      "loss": 0.0773,
      "reward": 1.2077504396438599,
      "reward_std": 0.3487227261066437,
      "reward_std_all": 0.32727910578250885,
      "rewards-max/forensics_sinkhorn_reward": 0.9833482205867767,
      "rewards-max/forensics_strict_edge_reward": 0.8582656681537628,
      "rewards-max/forensics_tiou_reward": 0.9860415458679199,
      "rewards/forensics_sinkhorn_reward": 0.6906084716320038,
      "rewards/forensics_strict_edge_reward": 0.07943546399474144,
      "rewards/forensics_tiou_reward": 0.43770651519298553,
      "step": 551
    },
    {
      "completion_length": 21.46875,
      "epoch": 1.69431643625192,
      "grad_norm": 12.108168601989746,
      "kl": 0.3291015625,
      "learning_rate": 5.761538461538461e-07,
      "loss": 0.0711,
      "reward": 1.3951780200004578,
      "reward_std": 0.24671697616577148,
      "reward_std_all": 0.2496504783630371,
      "rewards-max/forensics_sinkhorn_reward": 0.9721595942974091,
      "rewards-max/forensics_strict_edge_reward": 0.49686121940612793,
      "rewards-max/forensics_tiou_reward": 0.9567400515079498,
      "rewards/forensics_sinkhorn_reward": 0.7647398710250854,
      "rewards/forensics_strict_edge_reward": 0.0402109008282423,
      "rewards/forensics_tiou_reward": 0.5902273058891296,
      "step": 552
    },
    {
      "completion_length": 19.140625,
      "epoch": 1.6973886328725039,
      "grad_norm": 8.714859962463379,
      "kl": 0.26123046875,
      "learning_rate": 5.753846153846153e-07,
      "loss": -0.0538,
      "reward": 1.2263503074645996,
      "reward_std": 0.1693691536784172,
      "reward_std_all": 0.1824743002653122,
      "rewards-max/forensics_sinkhorn_reward": 0.8724387586116791,
      "rewards-max/forensics_strict_edge_reward": 0.18526397831737995,
      "rewards-max/forensics_tiou_reward": 0.7204306721687317,
      "rewards/forensics_sinkhorn_reward": 0.7273425459861755,
      "rewards/forensics_strict_edge_reward": 0.014631813857704401,
      "rewards/forensics_tiou_reward": 0.48437587916851044,
      "step": 553
    },
    {
      "completion_length": 22.609375,
      "epoch": 1.7004608294930876,
      "grad_norm": 12.657821655273438,
      "kl": 0.3544921875,
      "learning_rate": 5.746153846153847e-07,
      "loss": 0.1,
      "reward": 1.2758118510246277,
      "reward_std": 0.2349991500377655,
      "reward_std_all": 0.2275349348783493,
      "rewards-max/forensics_sinkhorn_reward": 0.9608483016490936,
      "rewards-max/forensics_strict_edge_reward": 0.41768769919872284,
      "rewards-max/forensics_tiou_reward": 0.9319528341293335,
      "rewards/forensics_sinkhorn_reward": 0.741776168346405,
      "rewards/forensics_strict_edge_reward": 0.028106585144996643,
      "rewards/forensics_tiou_reward": 0.5059292018413544,
      "step": 554
    },
    {
      "completion_length": 20.265625,
      "epoch": 1.7035330261136714,
      "grad_norm": 9.183070182800293,
      "kl": 0.24951171875,
      "learning_rate": 5.738461538461539e-07,
      "loss": -0.0004,
      "reward": 1.0197568535804749,
      "reward_std": 0.25492148101329803,
      "reward_std_all": 0.236307293176651,
      "rewards-max/forensics_sinkhorn_reward": 0.9211511015892029,
      "rewards-max/forensics_strict_edge_reward": 0.1364837010332849,
      "rewards-max/forensics_tiou_reward": 0.8349514007568359,
      "rewards/forensics_sinkhorn_reward": 0.639604926109314,
      "rewards/forensics_strict_edge_reward": 0.012096121990907704,
      "rewards/forensics_tiou_reward": 0.3680557906627655,
      "step": 555
    },
    {
      "completion_length": 21.78125,
      "epoch": 1.706605222734255,
      "grad_norm": 10.543098449707031,
      "kl": 0.484375,
      "learning_rate": 5.73076923076923e-07,
      "loss": -0.0463,
      "reward": 1.0957575738430023,
      "reward_std": 0.23369742184877396,
      "reward_std_all": 0.22273088991641998,
      "rewards-max/forensics_sinkhorn_reward": 0.9750208556652069,
      "rewards-max/forensics_strict_edge_reward": 0.5120998322963715,
      "rewards-max/forensics_tiou_reward": 0.9593198299407959,
      "rewards/forensics_sinkhorn_reward": 0.6624837219715118,
      "rewards/forensics_strict_edge_reward": 0.021987290121614933,
      "rewards/forensics_tiou_reward": 0.4112865626811981,
      "step": 556
    },
    {
      "completion_length": 20.0625,
      "epoch": 1.7096774193548387,
      "grad_norm": 9.8167724609375,
      "kl": 0.28515625,
      "learning_rate": 5.723076923076923e-07,
      "loss": -0.092,
      "reward": 1.2292101979255676,
      "reward_std": 0.21531835198402405,
      "reward_std_all": 0.2011844590306282,
      "rewards-max/forensics_sinkhorn_reward": 0.973796546459198,
      "rewards-max/forensics_strict_edge_reward": 0.545335203409195,
      "rewards-max/forensics_tiou_reward": 0.9609464108943939,
      "rewards/forensics_sinkhorn_reward": 0.728399008512497,
      "rewards/forensics_strict_edge_reward": 0.04547447711229324,
      "rewards/forensics_tiou_reward": 0.4553366005420685,
      "step": 557
    },
    {
      "completion_length": 19.625,
      "epoch": 1.7127496159754223,
      "grad_norm": 13.568242073059082,
      "kl": 0.3212890625,
      "learning_rate": 5.715384615384615e-07,
      "loss": -0.0524,
      "reward": 1.314577341079712,
      "reward_std": 0.2283487468957901,
      "reward_std_all": 0.20506297051906586,
      "rewards-max/forensics_sinkhorn_reward": 0.9598386585712433,
      "rewards-max/forensics_strict_edge_reward": 0.36972036957740784,
      "rewards-max/forensics_tiou_reward": 0.9286021292209625,
      "rewards/forensics_sinkhorn_reward": 0.7540941536426544,
      "rewards/forensics_strict_edge_reward": 0.04380020126700401,
      "rewards/forensics_tiou_reward": 0.5166830718517303,
      "step": 558
    },
    {
      "completion_length": 22.828125,
      "epoch": 1.715821812596006,
      "grad_norm": 12.285850524902344,
      "kl": 0.64453125,
      "learning_rate": 5.707692307692307e-07,
      "loss": -0.0251,
      "reward": 1.2944833636283875,
      "reward_std": 0.2510232776403427,
      "reward_std_all": 0.21353407204151154,
      "rewards-max/forensics_sinkhorn_reward": 0.961763858795166,
      "rewards-max/forensics_strict_edge_reward": 0.2874375730752945,
      "rewards-max/forensics_tiou_reward": 0.927619218826294,
      "rewards/forensics_sinkhorn_reward": 0.7274457812309265,
      "rewards/forensics_strict_edge_reward": 0.0337749132886529,
      "rewards/forensics_tiou_reward": 0.5332626700401306,
      "step": 559
    },
    {
      "completion_length": 21.84375,
      "epoch": 1.7188940092165899,
      "grad_norm": 11.35721492767334,
      "kl": 0.23583984375,
      "learning_rate": 5.699999999999999e-07,
      "loss": -0.0809,
      "reward": 1.071687936782837,
      "reward_std": 0.2544906362891197,
      "reward_std_all": 0.24178611487150192,
      "rewards-max/forensics_sinkhorn_reward": 0.8640164434909821,
      "rewards-max/forensics_strict_edge_reward": 0.24173212051391602,
      "rewards-max/forensics_tiou_reward": 0.7628694474697113,
      "rewards/forensics_sinkhorn_reward": 0.633318305015564,
      "rewards/forensics_strict_edge_reward": 0.025432898430153728,
      "rewards/forensics_tiou_reward": 0.4129367470741272,
      "step": 560
    },
    {
      "completion_length": 21.078125,
      "epoch": 1.7219662058371736,
      "grad_norm": 10.980023384094238,
      "kl": 0.275390625,
      "learning_rate": 5.692307692307692e-07,
      "loss": 0.0596,
      "reward": 1.2615074515342712,
      "reward_std": 0.24244870990514755,
      "reward_std_all": 0.21324114501476288,
      "rewards-max/forensics_sinkhorn_reward": 0.9655808508396149,
      "rewards-max/forensics_strict_edge_reward": 0.17178446426987648,
      "rewards-max/forensics_tiou_reward": 0.9412169754505157,
      "rewards/forensics_sinkhorn_reward": 0.7316148281097412,
      "rewards/forensics_strict_edge_reward": 0.016187147237360477,
      "rewards/forensics_tiou_reward": 0.5137054920196533,
      "step": 561
    },
    {
      "completion_length": 23.0625,
      "epoch": 1.7250384024577574,
      "grad_norm": 11.494178771972656,
      "kl": 0.23046875,
      "learning_rate": 5.684615384615385e-07,
      "loss": 0.133,
      "reward": 1.2195849418640137,
      "reward_std": 0.2680398225784302,
      "reward_std_all": 0.2382178157567978,
      "rewards-max/forensics_sinkhorn_reward": 0.9702983796596527,
      "rewards-max/forensics_strict_edge_reward": 0.4645538330078125,
      "rewards-max/forensics_tiou_reward": 0.9530609250068665,
      "rewards/forensics_sinkhorn_reward": 0.7053801417350769,
      "rewards/forensics_strict_edge_reward": 0.03814856521785259,
      "rewards/forensics_tiou_reward": 0.47605618834495544,
      "step": 562
    },
    {
      "completion_length": 20.8125,
      "epoch": 1.728110599078341,
      "grad_norm": 13.25209903717041,
      "kl": 0.4287109375,
      "learning_rate": 5.676923076923077e-07,
      "loss": 0.0459,
      "reward": 1.1614700555801392,
      "reward_std": 0.2222558632493019,
      "reward_std_all": 0.19630615413188934,
      "rewards-max/forensics_sinkhorn_reward": 0.923023134469986,
      "rewards-max/forensics_strict_edge_reward": 0.36154662049375474,
      "rewards-max/forensics_tiou_reward": 0.8423061370849609,
      "rewards/forensics_sinkhorn_reward": 0.7048362791538239,
      "rewards/forensics_strict_edge_reward": 0.01938766810053494,
      "rewards/forensics_tiou_reward": 0.4372462034225464,
      "step": 563
    },
    {
      "completion_length": 20.546875,
      "epoch": 1.7311827956989247,
      "grad_norm": 13.294026374816895,
      "kl": 0.20263671875,
      "learning_rate": 5.669230769230769e-07,
      "loss": 0.0128,
      "reward": 1.194872498512268,
      "reward_std": 0.25508079677820206,
      "reward_std_all": 0.25687285512685776,
      "rewards-max/forensics_sinkhorn_reward": 0.9753209054470062,
      "rewards-max/forensics_strict_edge_reward": 0.35741647332906723,
      "rewards-max/forensics_tiou_reward": 0.9674250781536102,
      "rewards/forensics_sinkhorn_reward": 0.6943327188491821,
      "rewards/forensics_strict_edge_reward": 0.046668888768181205,
      "rewards/forensics_tiou_reward": 0.45387090742588043,
      "step": 564
    },
    {
      "completion_length": 21.75,
      "epoch": 1.7342549923195083,
      "grad_norm": 9.295943260192871,
      "kl": 0.2724609375,
      "learning_rate": 5.661538461538461e-07,
      "loss": -0.0419,
      "reward": 1.4140412211418152,
      "reward_std": 0.2531517893075943,
      "reward_std_all": 0.2244681939482689,
      "rewards-max/forensics_sinkhorn_reward": 0.9826856851577759,
      "rewards-max/forensics_strict_edge_reward": 0.5866729617118835,
      "rewards-max/forensics_tiou_reward": 0.9817420542240143,
      "rewards/forensics_sinkhorn_reward": 0.7904540002346039,
      "rewards/forensics_strict_edge_reward": 0.037729956209659576,
      "rewards/forensics_tiou_reward": 0.5858573317527771,
      "step": 565
    },
    {
      "completion_length": 22.109375,
      "epoch": 1.737327188940092,
      "grad_norm": 12.58309555053711,
      "kl": 0.3671875,
      "learning_rate": 5.653846153846153e-07,
      "loss": 0.1159,
      "reward": 1.1362922191619873,
      "reward_std": 0.2389613837003708,
      "reward_std_all": 0.21815059334039688,
      "rewards-max/forensics_sinkhorn_reward": 0.962489664554596,
      "rewards-max/forensics_strict_edge_reward": 0.21333196386694908,
      "rewards-max/forensics_tiou_reward": 0.9354656636714935,
      "rewards/forensics_sinkhorn_reward": 0.67540043592453,
      "rewards/forensics_strict_edge_reward": 0.012920824810862541,
      "rewards/forensics_tiou_reward": 0.44797107577323914,
      "step": 566
    },
    {
      "completion_length": 22.171875,
      "epoch": 1.7403993855606759,
      "grad_norm": 27.057615280151367,
      "kl": 0.3916015625,
      "learning_rate": 5.646153846153847e-07,
      "loss": 0.0367,
      "reward": 1.3244214057922363,
      "reward_std": 0.3391869068145752,
      "reward_std_all": 0.35861995816230774,
      "rewards-max/forensics_sinkhorn_reward": 0.9788047671318054,
      "rewards-max/forensics_strict_edge_reward": 0.3581901043653488,
      "rewards-max/forensics_tiou_reward": 0.9720105528831482,
      "rewards/forensics_sinkhorn_reward": 0.740089476108551,
      "rewards/forensics_strict_edge_reward": 0.04387279227375984,
      "rewards/forensics_tiou_reward": 0.5404591262340546,
      "step": 567
    },
    {
      "completion_length": 21.609375,
      "epoch": 1.7434715821812596,
      "grad_norm": 15.8077392578125,
      "kl": 0.32958984375,
      "learning_rate": 5.638461538461539e-07,
      "loss": -0.031,
      "reward": 1.3376780152320862,
      "reward_std": 0.16284210234880447,
      "reward_std_all": 0.1812710165977478,
      "rewards-max/forensics_sinkhorn_reward": 0.9748731553554535,
      "rewards-max/forensics_strict_edge_reward": 0.08384165808092803,
      "rewards-max/forensics_tiou_reward": 0.9606974422931671,
      "rewards/forensics_sinkhorn_reward": 0.7685692310333252,
      "rewards/forensics_strict_edge_reward": 0.00824983062921092,
      "rewards/forensics_tiou_reward": 0.5608588755130768,
      "step": 568
    },
    {
      "completion_length": 23.5625,
      "epoch": 1.7465437788018434,
      "grad_norm": 8.771392822265625,
      "kl": 0.32763671875,
      "learning_rate": 5.63076923076923e-07,
      "loss": 0.0584,
      "reward": 1.1889837384223938,
      "reward_std": 0.3152094781398773,
      "reward_std_all": 0.28157271444797516,
      "rewards-max/forensics_sinkhorn_reward": 0.9439662098884583,
      "rewards-max/forensics_strict_edge_reward": 0.4033643454313278,
      "rewards-max/forensics_tiou_reward": 0.891561895608902,
      "rewards/forensics_sinkhorn_reward": 0.6800087094306946,
      "rewards/forensics_strict_edge_reward": 0.03306904435157776,
      "rewards/forensics_tiou_reward": 0.4759059101343155,
      "step": 569
    },
    {
      "completion_length": 21.71875,
      "epoch": 1.7496159754224272,
      "grad_norm": 8.668499946594238,
      "kl": 0.5625,
      "learning_rate": 5.623076923076923e-07,
      "loss": 0.0802,
      "reward": 1.3624972701072693,
      "reward_std": 0.2882898300886154,
      "reward_std_all": 0.27726636826992035,
      "rewards-max/forensics_sinkhorn_reward": 0.9821642339229584,
      "rewards-max/forensics_strict_edge_reward": 0.5702531039714813,
      "rewards-max/forensics_tiou_reward": 0.9809802174568176,
      "rewards/forensics_sinkhorn_reward": 0.7572037279605865,
      "rewards/forensics_strict_edge_reward": 0.04506521858274937,
      "rewards/forensics_tiou_reward": 0.5602284073829651,
      "step": 570
    },
    {
      "completion_length": 21.5,
      "epoch": 1.7526881720430108,
      "grad_norm": 7.454664707183838,
      "kl": 0.20263671875,
      "learning_rate": 5.615384615384615e-07,
      "loss": -0.1761,
      "reward": 1.4699703454971313,
      "reward_std": 0.15235278755426407,
      "reward_std_all": 0.15539202839136124,
      "rewards-max/forensics_sinkhorn_reward": 0.9796273708343506,
      "rewards-max/forensics_strict_edge_reward": 0.402368500828743,
      "rewards-max/forensics_tiou_reward": 0.9706422090530396,
      "rewards/forensics_sinkhorn_reward": 0.8100517988204956,
      "rewards/forensics_strict_edge_reward": 0.028680812567472458,
      "rewards/forensics_tiou_reward": 0.6312376856803894,
      "step": 571
    },
    {
      "completion_length": 21.96875,
      "epoch": 1.7557603686635943,
      "grad_norm": 7.855717658996582,
      "kl": 0.28125,
      "learning_rate": 5.607692307692308e-07,
      "loss": -0.0118,
      "reward": 1.0068728923797607,
      "reward_std": 0.22861891239881516,
      "reward_std_all": 0.20076975226402283,
      "rewards-max/forensics_sinkhorn_reward": 0.8920119106769562,
      "rewards-max/forensics_strict_edge_reward": 0.17543138563632965,
      "rewards-max/forensics_tiou_reward": 0.7961170375347137,
      "rewards/forensics_sinkhorn_reward": 0.6269280314445496,
      "rewards/forensics_strict_edge_reward": 0.011255415854975581,
      "rewards/forensics_tiou_reward": 0.36868949234485626,
      "step": 572
    },
    {
      "completion_length": 24.953125,
      "epoch": 1.758832565284178,
      "grad_norm": 7.407950401306152,
      "kl": 0.3037109375,
      "learning_rate": 5.6e-07,
      "loss": 0.0751,
      "reward": 1.1470287442207336,
      "reward_std": 0.23532280325889587,
      "reward_std_all": 0.22176988422870636,
      "rewards-max/forensics_sinkhorn_reward": 0.9614503085613251,
      "rewards-max/forensics_strict_edge_reward": 0.12831610813736916,
      "rewards-max/forensics_tiou_reward": 0.9269130229949951,
      "rewards/forensics_sinkhorn_reward": 0.6782933473587036,
      "rewards/forensics_strict_edge_reward": 0.00976989115588367,
      "rewards/forensics_tiou_reward": 0.4589654952287674,
      "step": 573
    },
    {
      "completion_length": 20.5625,
      "epoch": 1.7619047619047619,
      "grad_norm": 9.587746620178223,
      "kl": 0.42578125,
      "learning_rate": 5.592307692307691e-07,
      "loss": 0.1125,
      "reward": 1.2243416905403137,
      "reward_std": 0.1208152249455452,
      "reward_std_all": 0.16408412158489227,
      "rewards-max/forensics_sinkhorn_reward": 0.9323102533817291,
      "rewards-max/forensics_strict_edge_reward": 0.15223555080592632,
      "rewards-max/forensics_tiou_reward": 0.8534116446971893,
      "rewards/forensics_sinkhorn_reward": 0.7206970453262329,
      "rewards/forensics_strict_edge_reward": 0.013444914948195219,
      "rewards/forensics_tiou_reward": 0.4901996850967407,
      "step": 574
    },
    {
      "completion_length": 21.4375,
      "epoch": 1.7649769585253456,
      "grad_norm": 8.714221000671387,
      "kl": 0.3017578125,
      "learning_rate": 5.584615384615385e-07,
      "loss": 0.1211,
      "reward": 1.174146592617035,
      "reward_std": 0.17349840700626373,
      "reward_std_all": 0.19103089720010757,
      "rewards-max/forensics_sinkhorn_reward": 0.9199000298976898,
      "rewards-max/forensics_strict_edge_reward": 0.20528285205364227,
      "rewards-max/forensics_tiou_reward": 0.8256758153438568,
      "rewards/forensics_sinkhorn_reward": 0.7171351909637451,
      "rewards/forensics_strict_edge_reward": 0.013315421529114246,
      "rewards/forensics_tiou_reward": 0.443695992231369,
      "step": 575
    },
    {
      "completion_length": 20.890625,
      "epoch": 1.7680491551459294,
      "grad_norm": 11.324472427368164,
      "kl": 0.2236328125,
      "learning_rate": 5.576923076923077e-07,
      "loss": -0.0841,
      "reward": 1.2593923807144165,
      "reward_std": 0.26863349974155426,
      "reward_std_all": 0.23843688517808914,
      "rewards-max/forensics_sinkhorn_reward": 0.9778585433959961,
      "rewards-max/forensics_strict_edge_reward": 0.5190997421741486,
      "rewards-max/forensics_tiou_reward": 0.9678481817245483,
      "rewards/forensics_sinkhorn_reward": 0.7081734240055084,
      "rewards/forensics_strict_edge_reward": 0.058811599388718605,
      "rewards/forensics_tiou_reward": 0.4924074709415436,
      "step": 576
    },
    {
      "completion_length": 21.765625,
      "epoch": 1.7711213517665132,
      "grad_norm": 11.452052116394043,
      "kl": 0.189453125,
      "learning_rate": 5.569230769230769e-07,
      "loss": 0.0267,
      "reward": 1.1820634007453918,
      "reward_std": 0.2965780794620514,
      "reward_std_all": 0.29181575775146484,
      "rewards-max/forensics_sinkhorn_reward": 0.9771362543106079,
      "rewards-max/forensics_strict_edge_reward": 0.5311143696308136,
      "rewards-max/forensics_tiou_reward": 0.9690501391887665,
      "rewards/forensics_sinkhorn_reward": 0.6893416345119476,
      "rewards/forensics_strict_edge_reward": 0.026430298574268818,
      "rewards/forensics_tiou_reward": 0.46629147231578827,
      "step": 577
    },
    {
      "completion_length": 22.640625,
      "epoch": 1.7741935483870968,
      "grad_norm": 7.752413272857666,
      "kl": 0.27880859375,
      "learning_rate": 5.561538461538461e-07,
      "loss": 0.036,
      "reward": 1.2837011218070984,
      "reward_std": 0.240411639213562,
      "reward_std_all": 0.21151261031627655,
      "rewards-max/forensics_sinkhorn_reward": 0.958228588104248,
      "rewards-max/forensics_strict_edge_reward": 0.3715725541114807,
      "rewards-max/forensics_tiou_reward": 0.9196462631225586,
      "rewards/forensics_sinkhorn_reward": 0.7251915633678436,
      "rewards/forensics_strict_edge_reward": 0.0441170334815979,
      "rewards/forensics_tiou_reward": 0.5143924951553345,
      "step": 578
    },
    {
      "completion_length": 20.453125,
      "epoch": 1.7772657450076805,
      "grad_norm": 8.42982006072998,
      "kl": 0.4130859375,
      "learning_rate": 5.553846153846153e-07,
      "loss": -0.0374,
      "reward": 1.1968985795974731,
      "reward_std": 0.25601330399513245,
      "reward_std_all": 0.2280040755867958,
      "rewards-max/forensics_sinkhorn_reward": 0.9488186538219452,
      "rewards-max/forensics_strict_edge_reward": 0.32097239792346954,
      "rewards-max/forensics_tiou_reward": 0.886984795331955,
      "rewards/forensics_sinkhorn_reward": 0.710430771112442,
      "rewards/forensics_strict_edge_reward": 0.024652603431604803,
      "rewards/forensics_tiou_reward": 0.46181511878967285,
      "step": 579
    },
    {
      "completion_length": 21.4375,
      "epoch": 1.780337941628264,
      "grad_norm": 7.586916923522949,
      "kl": 0.25146484375,
      "learning_rate": 5.546153846153846e-07,
      "loss": 0.0285,
      "reward": 1.224054217338562,
      "reward_std": 0.24201080203056335,
      "reward_std_all": 0.22898352146148682,
      "rewards-max/forensics_sinkhorn_reward": 0.950078159570694,
      "rewards-max/forensics_strict_edge_reward": 0.25429365411400795,
      "rewards-max/forensics_tiou_reward": 0.8983752727508545,
      "rewards/forensics_sinkhorn_reward": 0.7075035572052002,
      "rewards/forensics_strict_edge_reward": 0.028379095252603292,
      "rewards/forensics_tiou_reward": 0.4881715625524521,
      "step": 580
    },
    {
      "completion_length": 20.625,
      "epoch": 1.7834101382488479,
      "grad_norm": 11.628203392028809,
      "kl": 0.3515625,
      "learning_rate": 5.538461538461539e-07,
      "loss": -0.0273,
      "reward": 1.3802127242088318,
      "reward_std": 0.3237713426351547,
      "reward_std_all": 0.28803347796201706,
      "rewards-max/forensics_sinkhorn_reward": 0.9730145037174225,
      "rewards-max/forensics_strict_edge_reward": 0.28830963373184204,
      "rewards-max/forensics_tiou_reward": 0.9485572874546051,
      "rewards/forensics_sinkhorn_reward": 0.7767052948474884,
      "rewards/forensics_strict_edge_reward": 0.02988983318209648,
      "rewards/forensics_tiou_reward": 0.5736176073551178,
      "step": 581
    },
    {
      "completion_length": 20.828125,
      "epoch": 1.7864823348694316,
      "grad_norm": 9.95501708984375,
      "kl": 0.37890625,
      "learning_rate": 5.53076923076923e-07,
      "loss": -0.0294,
      "reward": 1.1148595213890076,
      "reward_std": 0.20488881319761276,
      "reward_std_all": 0.18187428265810013,
      "rewards-max/forensics_sinkhorn_reward": 0.9351584017276764,
      "rewards-max/forensics_strict_edge_reward": 0.15495971590280533,
      "rewards-max/forensics_tiou_reward": 0.8740280270576477,
      "rewards/forensics_sinkhorn_reward": 0.6767995059490204,
      "rewards/forensics_strict_edge_reward": 0.013087232946418226,
      "rewards/forensics_tiou_reward": 0.42497269809246063,
      "step": 582
    },
    {
      "completion_length": 21.0625,
      "epoch": 1.7895545314900154,
      "grad_norm": 7.268742561340332,
      "kl": 0.2177734375,
      "learning_rate": 5.523076923076923e-07,
      "loss": 0.1398,
      "reward": 1.1043174266815186,
      "reward_std": 0.16327619552612305,
      "reward_std_all": 0.15745843946933746,
      "rewards-max/forensics_sinkhorn_reward": 0.9464184939861298,
      "rewards-max/forensics_strict_edge_reward": 0.19990129210054874,
      "rewards-max/forensics_tiou_reward": 0.8983381986618042,
      "rewards/forensics_sinkhorn_reward": 0.6644578576087952,
      "rewards/forensics_strict_edge_reward": 0.0145031253923662,
      "rewards/forensics_tiou_reward": 0.42535650730133057,
      "step": 583
    },
    {
      "completion_length": 23.53125,
      "epoch": 1.7926267281105992,
      "grad_norm": 10.32297134399414,
      "kl": 0.298828125,
      "learning_rate": 5.515384615384615e-07,
      "loss": 0.0541,
      "reward": 1.389327883720398,
      "reward_std": 0.24242401868104935,
      "reward_std_all": 0.22090376168489456,
      "rewards-max/forensics_sinkhorn_reward": 0.9732339084148407,
      "rewards-max/forensics_strict_edge_reward": 0.2816828489303589,
      "rewards-max/forensics_tiou_reward": 0.9600657820701599,
      "rewards/forensics_sinkhorn_reward": 0.758955717086792,
      "rewards/forensics_strict_edge_reward": 0.021168767474591732,
      "rewards/forensics_tiou_reward": 0.6092033684253693,
      "step": 584
    },
    {
      "completion_length": 20.375,
      "epoch": 1.7956989247311828,
      "grad_norm": 14.743428230285645,
      "kl": 0.2119140625,
      "learning_rate": 5.507692307692308e-07,
      "loss": 0.0311,
      "reward": 1.0965837836265564,
      "reward_std": 0.17074988782405853,
      "reward_std_all": 0.15710734575986862,
      "rewards-max/forensics_sinkhorn_reward": 0.9504329860210419,
      "rewards-max/forensics_strict_edge_reward": 0.15872163511812687,
      "rewards-max/forensics_tiou_reward": 0.9011190831661224,
      "rewards/forensics_sinkhorn_reward": 0.6795504689216614,
      "rewards/forensics_strict_edge_reward": 0.015057868789881468,
      "rewards/forensics_tiou_reward": 0.4019753932952881,
      "step": 585
    },
    {
      "completion_length": 20.21875,
      "epoch": 1.7987711213517665,
      "grad_norm": 10.340853691101074,
      "kl": 0.447265625,
      "learning_rate": 5.5e-07,
      "loss": 0.0796,
      "reward": 1.1346852779388428,
      "reward_std": 0.17744626104831696,
      "reward_std_all": 0.20109965652227402,
      "rewards-max/forensics_sinkhorn_reward": 0.981714278459549,
      "rewards-max/forensics_strict_edge_reward": 0.37331727892160416,
      "rewards-max/forensics_tiou_reward": 0.9789836406707764,
      "rewards/forensics_sinkhorn_reward": 0.6725663840770721,
      "rewards/forensics_strict_edge_reward": 0.021582020446658134,
      "rewards/forensics_tiou_reward": 0.44053687155246735,
      "step": 586
    },
    {
      "completion_length": 20.84375,
      "epoch": 1.80184331797235,
      "grad_norm": 8.416585922241211,
      "kl": 0.38818359375,
      "learning_rate": 5.492307692307691e-07,
      "loss": 0.0135,
      "reward": 1.374420940876007,
      "reward_std": 0.28154607117176056,
      "reward_std_all": 0.24637316912412643,
      "rewards-max/forensics_sinkhorn_reward": 0.9843258857727051,
      "rewards-max/forensics_strict_edge_reward": 0.6593321859836578,
      "rewards-max/forensics_tiou_reward": 0.9873324632644653,
      "rewards/forensics_sinkhorn_reward": 0.7619682252407074,
      "rewards/forensics_strict_edge_reward": 0.05670931935310364,
      "rewards/forensics_tiou_reward": 0.5557433366775513,
      "step": 587
    },
    {
      "completion_length": 22.078125,
      "epoch": 1.8049155145929339,
      "grad_norm": 8.208017349243164,
      "kl": 0.21875,
      "learning_rate": 5.484615384615385e-07,
      "loss": 0.0722,
      "reward": 1.3758521676063538,
      "reward_std": 0.20937920361757278,
      "reward_std_all": 0.18653763085603714,
      "rewards-max/forensics_sinkhorn_reward": 0.9413858950138092,
      "rewards-max/forensics_strict_edge_reward": 0.2053181678056717,
      "rewards-max/forensics_tiou_reward": 0.8912339806556702,
      "rewards/forensics_sinkhorn_reward": 0.7764920592308044,
      "rewards/forensics_strict_edge_reward": 0.01983135100454092,
      "rewards/forensics_tiou_reward": 0.5795288383960724,
      "step": 588
    },
    {
      "completion_length": 21.609375,
      "epoch": 1.8079877112135176,
      "grad_norm": 13.585017204284668,
      "kl": 0.4326171875,
      "learning_rate": 5.476923076923077e-07,
      "loss": 0.0287,
      "reward": 1.2836231589317322,
      "reward_std": 0.25001993030309677,
      "reward_std_all": 0.25609874725341797,
      "rewards-max/forensics_sinkhorn_reward": 0.9753088653087616,
      "rewards-max/forensics_strict_edge_reward": 0.6154949963092804,
      "rewards-max/forensics_tiou_reward": 0.9652403891086578,
      "rewards/forensics_sinkhorn_reward": 0.7155478000640869,
      "rewards/forensics_strict_edge_reward": 0.059908926486968994,
      "rewards/forensics_tiou_reward": 0.5081663876771927,
      "step": 589
    },
    {
      "completion_length": 22.34375,
      "epoch": 1.8110599078341014,
      "grad_norm": 9.957647323608398,
      "kl": 0.30126953125,
      "learning_rate": 5.469230769230769e-07,
      "loss": -0.0764,
      "reward": 1.3962848782539368,
      "reward_std": 0.1817353516817093,
      "reward_std_all": 0.17186003178358078,
      "rewards-max/forensics_sinkhorn_reward": 0.9755579829216003,
      "rewards-max/forensics_strict_edge_reward": 0.42959781736135483,
      "rewards-max/forensics_tiou_reward": 0.9652377068996429,
      "rewards/forensics_sinkhorn_reward": 0.7921611964702606,
      "rewards/forensics_strict_edge_reward": 0.034789396449923515,
      "rewards/forensics_tiou_reward": 0.5693342089653015,
      "step": 590
    },
    {
      "completion_length": 23.390625,
      "epoch": 1.8141321044546852,
      "grad_norm": 7.657434463500977,
      "kl": 0.4345703125,
      "learning_rate": 5.461538461538461e-07,
      "loss": -0.14,
      "reward": 1.323668897151947,
      "reward_std": 0.2445986568927765,
      "reward_std_all": 0.22890855371952057,
      "rewards-max/forensics_sinkhorn_reward": 0.9673137962818146,
      "rewards-max/forensics_strict_edge_reward": 0.3208470232784748,
      "rewards-max/forensics_tiou_reward": 0.948638379573822,
      "rewards/forensics_sinkhorn_reward": 0.7464938759803772,
      "rewards/forensics_strict_edge_reward": 0.022855413844808936,
      "rewards/forensics_tiou_reward": 0.5543195605278015,
      "step": 591
    },
    {
      "completion_length": 22.359375,
      "epoch": 1.817204301075269,
      "grad_norm": 7.454367637634277,
      "kl": 0.3134765625,
      "learning_rate": 5.453846153846153e-07,
      "loss": 0.0985,
      "reward": 1.320658028125763,
      "reward_std": 0.23357345908880234,
      "reward_std_all": 0.23123463988304138,
      "rewards-max/forensics_sinkhorn_reward": 0.9803635478019714,
      "rewards-max/forensics_strict_edge_reward": 0.36904490552842617,
      "rewards-max/forensics_tiou_reward": 0.9773853719234467,
      "rewards/forensics_sinkhorn_reward": 0.73263880610466,
      "rewards/forensics_strict_edge_reward": 0.046765764243900776,
      "rewards/forensics_tiou_reward": 0.5412534475326538,
      "step": 592
    },
    {
      "completion_length": 21.375,
      "epoch": 1.8202764976958525,
      "grad_norm": 8.280555725097656,
      "kl": 0.40234375,
      "learning_rate": 5.446153846153846e-07,
      "loss": 0.0514,
      "reward": 1.22150456905365,
      "reward_std": 0.30889666080474854,
      "reward_std_all": 0.28157398104667664,
      "rewards-max/forensics_sinkhorn_reward": 0.9760498404502869,
      "rewards-max/forensics_strict_edge_reward": 0.5256026089191437,
      "rewards-max/forensics_tiou_reward": 0.9666042625904083,
      "rewards/forensics_sinkhorn_reward": 0.6827673614025116,
      "rewards/forensics_strict_edge_reward": 0.045563782565295696,
      "rewards/forensics_tiou_reward": 0.4931735396385193,
      "step": 593
    },
    {
      "completion_length": 20.3125,
      "epoch": 1.823348694316436,
      "grad_norm": 12.47342300415039,
      "kl": 0.24951171875,
      "learning_rate": 5.438461538461538e-07,
      "loss": 0.0507,
      "reward": 1.192061424255371,
      "reward_std": 0.17229463160037994,
      "reward_std_all": 0.1589062176644802,
      "rewards-max/forensics_sinkhorn_reward": 0.9634520411491394,
      "rewards-max/forensics_strict_edge_reward": 0.23905399814248085,
      "rewards-max/forensics_tiou_reward": 0.9143776297569275,
      "rewards/forensics_sinkhorn_reward": 0.6910008192062378,
      "rewards/forensics_strict_edge_reward": 0.02971740672364831,
      "rewards/forensics_tiou_reward": 0.4713432490825653,
      "step": 594
    },
    {
      "completion_length": 20.6875,
      "epoch": 1.8264208909370199,
      "grad_norm": 12.20170783996582,
      "kl": 0.40234375,
      "learning_rate": 5.430769230769231e-07,
      "loss": 0.0859,
      "reward": 1.3338899612426758,
      "reward_std": 0.23805273324251175,
      "reward_std_all": 0.21330277621746063,
      "rewards-max/forensics_sinkhorn_reward": 0.9617568850517273,
      "rewards-max/forensics_strict_edge_reward": 0.5172730684280396,
      "rewards-max/forensics_tiou_reward": 0.9347112476825714,
      "rewards/forensics_sinkhorn_reward": 0.7499642074108124,
      "rewards/forensics_strict_edge_reward": 0.05757793691009283,
      "rewards/forensics_tiou_reward": 0.5263477712869644,
      "step": 595
    },
    {
      "completion_length": 21.5,
      "epoch": 1.8294930875576036,
      "grad_norm": 11.059929847717285,
      "kl": 0.3056640625,
      "learning_rate": 5.423076923076923e-07,
      "loss": 0.1,
      "reward": 1.1825376749038696,
      "reward_std": 0.31615032255649567,
      "reward_std_all": 0.3240385353565216,
      "rewards-max/forensics_sinkhorn_reward": 0.9621250629425049,
      "rewards-max/forensics_strict_edge_reward": 0.4937888830900192,
      "rewards-max/forensics_tiou_reward": 0.9381977915763855,
      "rewards/forensics_sinkhorn_reward": 0.6693640649318695,
      "rewards/forensics_strict_edge_reward": 0.024223054759204388,
      "rewards/forensics_tiou_reward": 0.48895055055618286,
      "step": 596
    },
    {
      "completion_length": 21.171875,
      "epoch": 1.8325652841781874,
      "grad_norm": 12.06038761138916,
      "kl": 0.20849609375,
      "learning_rate": 5.415384615384615e-07,
      "loss": -0.0668,
      "reward": 1.3338430523872375,
      "reward_std": 0.23908354341983795,
      "reward_std_all": 0.21554476022720337,
      "rewards-max/forensics_sinkhorn_reward": 0.9460733532905579,
      "rewards-max/forensics_strict_edge_reward": 0.19487615209072828,
      "rewards-max/forensics_tiou_reward": 0.8816153109073639,
      "rewards/forensics_sinkhorn_reward": 0.7522657513618469,
      "rewards/forensics_strict_edge_reward": 0.019402782432734966,
      "rewards/forensics_tiou_reward": 0.5621744990348816,
      "step": 597
    },
    {
      "completion_length": 19.90625,
      "epoch": 1.8356374807987712,
      "grad_norm": 12.198859214782715,
      "kl": 0.2919921875,
      "learning_rate": 5.407692307692308e-07,
      "loss": 0.07,
      "reward": 1.3486315608024597,
      "reward_std": 0.3715952932834625,
      "reward_std_all": 0.3112022280693054,
      "rewards-max/forensics_sinkhorn_reward": 0.9816962778568268,
      "rewards-max/forensics_strict_edge_reward": 0.5702531039714813,
      "rewards-max/forensics_tiou_reward": 0.9799841344356537,
      "rewards/forensics_sinkhorn_reward": 0.7535237967967987,
      "rewards/forensics_strict_edge_reward": 0.05693329870700836,
      "rewards/forensics_tiou_reward": 0.5381744652986526,
      "step": 598
    },
    {
      "completion_length": 21.515625,
      "epoch": 1.838709677419355,
      "grad_norm": 10.316727638244629,
      "kl": 0.26513671875,
      "learning_rate": 5.4e-07,
      "loss": -0.025,
      "reward": 1.197447657585144,
      "reward_std": 0.24539951980113983,
      "reward_std_all": 0.25312136113643646,
      "rewards-max/forensics_sinkhorn_reward": 0.9826948940753937,
      "rewards-max/forensics_strict_edge_reward": 0.7158875465393066,
      "rewards-max/forensics_tiou_reward": 0.9796369969844818,
      "rewards/forensics_sinkhorn_reward": 0.6953191161155701,
      "rewards/forensics_strict_edge_reward": 0.04593879170715809,
      "rewards/forensics_tiou_reward": 0.4561897814273834,
      "step": 599
    },
    {
      "completion_length": 21.875,
      "epoch": 1.8417818740399385,
      "grad_norm": 11.739864349365234,
      "kl": 0.626953125,
      "learning_rate": 5.392307692307691e-07,
      "loss": 0.0885,
      "reward": 1.3893853425979614,
      "reward_std": 0.16503887623548508,
      "reward_std_all": 0.19849274307489395,
      "rewards-max/forensics_sinkhorn_reward": 0.952094167470932,
      "rewards-max/forensics_strict_edge_reward": 0.3187349736690521,
      "rewards-max/forensics_tiou_reward": 0.9078098237514496,
      "rewards/forensics_sinkhorn_reward": 0.7811141908168793,
      "rewards/forensics_strict_edge_reward": 0.0217605447396636,
      "rewards/forensics_tiou_reward": 0.5865105390548706,
      "step": 600
    },
    {
      "completion_length": 20.015625,
      "epoch": 1.8448540706605223,
      "grad_norm": 15.660881996154785,
      "kl": 0.392578125,
      "learning_rate": 5.384615384615384e-07,
      "loss": -0.0646,
      "reward": 1.3009360432624817,
      "reward_std": 0.21033403277397156,
      "reward_std_all": 0.18352236598730087,
      "rewards-max/forensics_sinkhorn_reward": 0.9769956171512604,
      "rewards-max/forensics_strict_edge_reward": 0.5949355661869049,
      "rewards-max/forensics_tiou_reward": 0.963822603225708,
      "rewards/forensics_sinkhorn_reward": 0.7244624495506287,
      "rewards/forensics_strict_edge_reward": 0.06101006828248501,
      "rewards/forensics_tiou_reward": 0.5154634863138199,
      "step": 601
    },
    {
      "completion_length": 21.8125,
      "epoch": 1.8479262672811059,
      "grad_norm": 6.934401512145996,
      "kl": 0.3525390625,
      "learning_rate": 5.376923076923077e-07,
      "loss": 0.011,
      "reward": 1.4157457947731018,
      "reward_std": 0.2123393937945366,
      "reward_std_all": 0.2109597772359848,
      "rewards-max/forensics_sinkhorn_reward": 0.9729798138141632,
      "rewards-max/forensics_strict_edge_reward": 0.38103729486465454,
      "rewards-max/forensics_tiou_reward": 0.9601144790649414,
      "rewards/forensics_sinkhorn_reward": 0.7781843543052673,
      "rewards/forensics_strict_edge_reward": 0.03026569075882435,
      "rewards/forensics_tiou_reward": 0.6072957813739777,
      "step": 602
    },
    {
      "completion_length": 20.609375,
      "epoch": 1.8509984639016897,
      "grad_norm": 16.937576293945312,
      "kl": 0.17333984375,
      "learning_rate": 5.369230769230769e-07,
      "loss": 0.0113,
      "reward": 1.3106263875961304,
      "reward_std": 0.22595088183879852,
      "reward_std_all": 0.21190118044614792,
      "rewards-max/forensics_sinkhorn_reward": 0.9594627916812897,
      "rewards-max/forensics_strict_edge_reward": 0.36884360760450363,
      "rewards-max/forensics_tiou_reward": 0.9235469698905945,
      "rewards/forensics_sinkhorn_reward": 0.7346547544002533,
      "rewards/forensics_strict_edge_reward": 0.05654738796874881,
      "rewards/forensics_tiou_reward": 0.5194242000579834,
      "step": 603
    },
    {
      "completion_length": 21.6875,
      "epoch": 1.8540706605222734,
      "grad_norm": 9.634864807128906,
      "kl": 0.2236328125,
      "learning_rate": 5.361538461538462e-07,
      "loss": 0.0678,
      "reward": 1.3186594247817993,
      "reward_std": 0.20488352328538895,
      "reward_std_all": 0.18612724542617798,
      "rewards-max/forensics_sinkhorn_reward": 0.9464290738105774,
      "rewards-max/forensics_strict_edge_reward": 0.4100942313671112,
      "rewards-max/forensics_tiou_reward": 0.9001017212867737,
      "rewards/forensics_sinkhorn_reward": 0.7381412386894226,
      "rewards/forensics_strict_edge_reward": 0.031451697228476405,
      "rewards/forensics_tiou_reward": 0.5490664541721344,
      "step": 604
    },
    {
      "completion_length": 20.96875,
      "epoch": 1.8571428571428572,
      "grad_norm": 8.141611099243164,
      "kl": 0.375,
      "learning_rate": 5.353846153846153e-07,
      "loss": -0.0417,
      "reward": 1.5637266039848328,
      "reward_std": 0.2995733618736267,
      "reward_std_all": 0.3034374862909317,
      "rewards-max/forensics_sinkhorn_reward": 0.9727319180965424,
      "rewards-max/forensics_strict_edge_reward": 0.5670222043991089,
      "rewards-max/forensics_tiou_reward": 0.9580017030239105,
      "rewards/forensics_sinkhorn_reward": 0.8253739178180695,
      "rewards/forensics_strict_edge_reward": 0.06971243768930435,
      "rewards/forensics_tiou_reward": 0.6686403155326843,
      "step": 605
    },
    {
      "completion_length": 21.9375,
      "epoch": 1.860215053763441,
      "grad_norm": 9.204939842224121,
      "kl": 0.22314453125,
      "learning_rate": 5.346153846153846e-07,
      "loss": 0.032,
      "reward": 1.1089822053909302,
      "reward_std": 0.18606743961572647,
      "reward_std_all": 0.16339552402496338,
      "rewards-max/forensics_sinkhorn_reward": 0.9497268795967102,
      "rewards-max/forensics_strict_edge_reward": 0.31268224120140076,
      "rewards-max/forensics_tiou_reward": 0.8987493515014648,
      "rewards/forensics_sinkhorn_reward": 0.6655628979206085,
      "rewards/forensics_strict_edge_reward": 0.025168707594275475,
      "rewards/forensics_tiou_reward": 0.4182506203651428,
      "step": 606
    },
    {
      "completion_length": 21.8125,
      "epoch": 1.8632872503840245,
      "grad_norm": 7.752347469329834,
      "kl": 0.4013671875,
      "learning_rate": 5.338461538461538e-07,
      "loss": 0.1782,
      "reward": 1.2692795991897583,
      "reward_std": 0.16889793425798416,
      "reward_std_all": 0.16308119893074036,
      "rewards-max/forensics_sinkhorn_reward": 0.9692001938819885,
      "rewards-max/forensics_strict_edge_reward": 0.12260743603110313,
      "rewards-max/forensics_tiou_reward": 0.9408377110958099,
      "rewards/forensics_sinkhorn_reward": 0.730655699968338,
      "rewards/forensics_strict_edge_reward": 0.011267495341598988,
      "rewards/forensics_tiou_reward": 0.5273563861846924,
      "step": 607
    },
    {
      "completion_length": 22.828125,
      "epoch": 1.8663594470046083,
      "grad_norm": 15.759008407592773,
      "kl": 0.2685546875,
      "learning_rate": 5.33076923076923e-07,
      "loss": 0.0982,
      "reward": 1.4118006229400635,
      "reward_std": 0.25264672935009,
      "reward_std_all": 0.22072292119264603,
      "rewards-max/forensics_sinkhorn_reward": 0.9720374643802643,
      "rewards-max/forensics_strict_edge_reward": 0.4043552428483963,
      "rewards-max/forensics_tiou_reward": 0.9491460025310516,
      "rewards/forensics_sinkhorn_reward": 0.7491552531719208,
      "rewards/forensics_strict_edge_reward": 0.06294615939259529,
      "rewards/forensics_tiou_reward": 0.5996992588043213,
      "step": 608
    },
    {
      "completion_length": 20.9375,
      "epoch": 1.8694316436251919,
      "grad_norm": 10.748035430908203,
      "kl": 0.2646484375,
      "learning_rate": 5.323076923076923e-07,
      "loss": -0.0752,
      "reward": 1.3044283390045166,
      "reward_std": 0.2864631116390228,
      "reward_std_all": 0.24984322488307953,
      "rewards-max/forensics_sinkhorn_reward": 0.9303701519966125,
      "rewards-max/forensics_strict_edge_reward": 0.7426987290382385,
      "rewards-max/forensics_tiou_reward": 0.8584035336971283,
      "rewards/forensics_sinkhorn_reward": 0.7420361340045929,
      "rewards/forensics_strict_edge_reward": 0.044507795944809914,
      "rewards/forensics_tiou_reward": 0.5178843885660172,
      "step": 609
    },
    {
      "completion_length": 21.5,
      "epoch": 1.8725038402457757,
      "grad_norm": 12.274259567260742,
      "kl": 0.1962890625,
      "learning_rate": 5.315384615384615e-07,
      "loss": 0.1188,
      "reward": 1.2498924732208252,
      "reward_std": 0.15823542699217796,
      "reward_std_all": 0.16744664311408997,
      "rewards-max/forensics_sinkhorn_reward": 0.927992582321167,
      "rewards-max/forensics_strict_edge_reward": 0.5080146789550781,
      "rewards-max/forensics_tiou_reward": 0.8576132953166962,
      "rewards/forensics_sinkhorn_reward": 0.738844633102417,
      "rewards/forensics_strict_edge_reward": 0.024083936354145408,
      "rewards/forensics_tiou_reward": 0.4869639575481415,
      "step": 610
    },
    {
      "completion_length": 19.578125,
      "epoch": 1.8755760368663594,
      "grad_norm": 13.499613761901855,
      "kl": 0.24267578125,
      "learning_rate": 5.307692307692308e-07,
      "loss": -0.0308,
      "reward": 1.4693825244903564,
      "reward_std": 0.2592737078666687,
      "reward_std_all": 0.24169541150331497,
      "rewards-max/forensics_sinkhorn_reward": 0.9700595438480377,
      "rewards-max/forensics_strict_edge_reward": 0.5090932846069336,
      "rewards-max/forensics_tiou_reward": 0.947388231754303,
      "rewards/forensics_sinkhorn_reward": 0.8121213614940643,
      "rewards/forensics_strict_edge_reward": 0.04804079420864582,
      "rewards/forensics_tiou_reward": 0.6092204451560974,
      "step": 611
    },
    {
      "completion_length": 19.765625,
      "epoch": 1.8786482334869432,
      "grad_norm": 15.24291706085205,
      "kl": 0.265625,
      "learning_rate": 5.3e-07,
      "loss": 0.0399,
      "reward": 1.1774267554283142,
      "reward_std": 0.23215070366859436,
      "reward_std_all": 0.20271605253219604,
      "rewards-max/forensics_sinkhorn_reward": 0.9652163088321686,
      "rewards-max/forensics_strict_edge_reward": 0.10971419885754585,
      "rewards-max/forensics_tiou_reward": 0.9389596581459045,
      "rewards/forensics_sinkhorn_reward": 0.7040739357471466,
      "rewards/forensics_strict_edge_reward": 0.0071748704649508,
      "rewards/forensics_tiou_reward": 0.46617794036865234,
      "step": 612
    },
    {
      "completion_length": 21.21875,
      "epoch": 1.881720430107527,
      "grad_norm": 12.046586990356445,
      "kl": 0.439453125,
      "learning_rate": 5.292307692307692e-07,
      "loss": 0.0686,
      "reward": 1.3444924354553223,
      "reward_std": 0.3163357675075531,
      "reward_std_all": 0.29361990094184875,
      "rewards-max/forensics_sinkhorn_reward": 0.9779788553714752,
      "rewards-max/forensics_strict_edge_reward": 0.22943734377622604,
      "rewards-max/forensics_tiou_reward": 0.9697596728801727,
      "rewards/forensics_sinkhorn_reward": 0.7615792453289032,
      "rewards/forensics_strict_edge_reward": 0.024310882203280926,
      "rewards/forensics_tiou_reward": 0.5586023330688477,
      "step": 613
    },
    {
      "completion_length": 22.078125,
      "epoch": 1.8847926267281108,
      "grad_norm": 14.701188087463379,
      "kl": 0.3095703125,
      "learning_rate": 5.284615384615384e-07,
      "loss": 0.0781,
      "reward": 1.3014962077140808,
      "reward_std": 0.2080528810620308,
      "reward_std_all": 0.20498375594615936,
      "rewards-max/forensics_sinkhorn_reward": 0.964371919631958,
      "rewards-max/forensics_strict_edge_reward": 0.365729495882988,
      "rewards-max/forensics_tiou_reward": 0.9370917975902557,
      "rewards/forensics_sinkhorn_reward": 0.7520342469215393,
      "rewards/forensics_strict_edge_reward": 0.04494043719023466,
      "rewards/forensics_tiou_reward": 0.5045215487480164,
      "step": 614
    },
    {
      "completion_length": 21.953125,
      "epoch": 1.8878648233486943,
      "grad_norm": 9.408236503601074,
      "kl": 0.25830078125,
      "learning_rate": 5.276923076923076e-07,
      "loss": -0.1701,
      "reward": 1.153806984424591,
      "reward_std": 0.1458933725953102,
      "reward_std_all": 0.1327618882060051,
      "rewards-max/forensics_sinkhorn_reward": 0.9572736322879791,
      "rewards-max/forensics_strict_edge_reward": 0.10483340546488762,
      "rewards-max/forensics_tiou_reward": 0.9231176376342773,
      "rewards/forensics_sinkhorn_reward": 0.6873913407325745,
      "rewards/forensics_strict_edge_reward": 0.012848662678152323,
      "rewards/forensics_tiou_reward": 0.45356693863868713,
      "step": 615
    },
    {
      "completion_length": 23.28125,
      "epoch": 1.8909370199692779,
      "grad_norm": 7.836155414581299,
      "kl": 0.20361328125,
      "learning_rate": 5.269230769230769e-07,
      "loss": 0.065,
      "reward": 1.3708621263504028,
      "reward_std": 0.22363808751106262,
      "reward_std_all": 0.2092556208372116,
      "rewards-max/forensics_sinkhorn_reward": 0.9678404033184052,
      "rewards-max/forensics_strict_edge_reward": 0.18767917901277542,
      "rewards-max/forensics_tiou_reward": 0.9380279183387756,
      "rewards/forensics_sinkhorn_reward": 0.764401763677597,
      "rewards/forensics_strict_edge_reward": 0.03114133421331644,
      "rewards/forensics_tiou_reward": 0.5753189921379089,
      "step": 616
    },
    {
      "completion_length": 20.296875,
      "epoch": 1.8940092165898617,
      "grad_norm": 8.759498596191406,
      "kl": 0.31689453125,
      "learning_rate": 5.261538461538462e-07,
      "loss": -0.0027,
      "reward": 1.3181822896003723,
      "reward_std": 0.28506723046302795,
      "reward_std_all": 0.25617431849241257,
      "rewards-max/forensics_sinkhorn_reward": 0.9690549075603485,
      "rewards-max/forensics_strict_edge_reward": 0.46688374876976013,
      "rewards-max/forensics_tiou_reward": 0.9469451010227203,
      "rewards/forensics_sinkhorn_reward": 0.740474283695221,
      "rewards/forensics_strict_edge_reward": 0.04068034142255783,
      "rewards/forensics_tiou_reward": 0.5370276868343353,
      "step": 617
    },
    {
      "completion_length": 22.09375,
      "epoch": 1.8970814132104454,
      "grad_norm": 11.374716758728027,
      "kl": 0.2666015625,
      "learning_rate": 5.253846153846153e-07,
      "loss": 0.0501,
      "reward": 1.292131781578064,
      "reward_std": 0.2595715671777725,
      "reward_std_all": 0.2240264192223549,
      "rewards-max/forensics_sinkhorn_reward": 0.9798021018505096,
      "rewards-max/forensics_strict_edge_reward": 0.4945315718650818,
      "rewards-max/forensics_tiou_reward": 0.9730681478977203,
      "rewards/forensics_sinkhorn_reward": 0.7369178235530853,
      "rewards/forensics_strict_edge_reward": 0.047722794115543365,
      "rewards/forensics_tiou_reward": 0.5074911415576935,
      "step": 618
    },
    {
      "completion_length": 20.921875,
      "epoch": 1.9001536098310292,
      "grad_norm": 10.768056869506836,
      "kl": 0.2578125,
      "learning_rate": 5.246153846153846e-07,
      "loss": 0.0616,
      "reward": 1.2845641374588013,
      "reward_std": 0.18971945345401764,
      "reward_std_all": 0.18372821062803268,
      "rewards-max/forensics_sinkhorn_reward": 0.9812347292900085,
      "rewards-max/forensics_strict_edge_reward": 0.4900742173194885,
      "rewards-max/forensics_tiou_reward": 0.9746680557727814,
      "rewards/forensics_sinkhorn_reward": 0.7309221029281616,
      "rewards/forensics_strict_edge_reward": 0.03674038127064705,
      "rewards/forensics_tiou_reward": 0.5169016122817993,
      "step": 619
    },
    {
      "completion_length": 21.1875,
      "epoch": 1.903225806451613,
      "grad_norm": 9.495413780212402,
      "kl": 0.3369140625,
      "learning_rate": 5.238461538461538e-07,
      "loss": -0.0827,
      "reward": 1.296891212463379,
      "reward_std": 0.2548677772283554,
      "reward_std_all": 0.27214179933071136,
      "rewards-max/forensics_sinkhorn_reward": 0.9750035405158997,
      "rewards-max/forensics_strict_edge_reward": 0.628528892993927,
      "rewards-max/forensics_tiou_reward": 0.9532657563686371,
      "rewards/forensics_sinkhorn_reward": 0.7248192131519318,
      "rewards/forensics_strict_edge_reward": 0.034089150838553905,
      "rewards/forensics_tiou_reward": 0.5379829108715057,
      "step": 620
    },
    {
      "completion_length": 22.390625,
      "epoch": 1.9062980030721968,
      "grad_norm": 11.641399383544922,
      "kl": 0.3017578125,
      "learning_rate": 5.23076923076923e-07,
      "loss": 0.0368,
      "reward": 1.3749194741249084,
      "reward_std": 0.15977132879197598,
      "reward_std_all": 0.17916014045476913,
      "rewards-max/forensics_sinkhorn_reward": 0.9808241724967957,
      "rewards-max/forensics_strict_edge_reward": 0.3283592611551285,
      "rewards-max/forensics_tiou_reward": 0.9775629341602325,
      "rewards/forensics_sinkhorn_reward": 0.7594283223152161,
      "rewards/forensics_strict_edge_reward": 0.04231885075569153,
      "rewards/forensics_tiou_reward": 0.5731722116470337,
      "step": 621
    },
    {
      "completion_length": 22.296875,
      "epoch": 1.9093701996927803,
      "grad_norm": 13.040586471557617,
      "kl": 0.2578125,
      "learning_rate": 5.223076923076924e-07,
      "loss": 0.0546,
      "reward": 1.2083094716072083,
      "reward_std": 0.17633531987667084,
      "reward_std_all": 0.16839420795440674,
      "rewards-max/forensics_sinkhorn_reward": 0.9511890411376953,
      "rewards-max/forensics_strict_edge_reward": 0.09887783229351044,
      "rewards-max/forensics_tiou_reward": 0.9037539064884186,
      "rewards/forensics_sinkhorn_reward": 0.7094506621360779,
      "rewards/forensics_strict_edge_reward": 0.014028645819053054,
      "rewards/forensics_tiou_reward": 0.4848301410675049,
      "step": 622
    },
    {
      "completion_length": 22.90625,
      "epoch": 1.912442396313364,
      "grad_norm": 9.976078987121582,
      "kl": 0.69921875,
      "learning_rate": 5.215384615384615e-07,
      "loss": 0.1093,
      "reward": 1.378210425376892,
      "reward_std": 0.32203029096126556,
      "reward_std_all": 0.29252369701862335,
      "rewards-max/forensics_sinkhorn_reward": 0.9876706898212433,
      "rewards-max/forensics_strict_edge_reward": 0.8187307715415955,
      "rewards-max/forensics_tiou_reward": 0.9967082440853119,
      "rewards/forensics_sinkhorn_reward": 0.7395089268684387,
      "rewards/forensics_strict_edge_reward": 0.06755492277443409,
      "rewards/forensics_tiou_reward": 0.571146696805954,
      "step": 623
    },
    {
      "completion_length": 23.890625,
      "epoch": 1.9155145929339477,
      "grad_norm": 8.111252784729004,
      "kl": 0.3427734375,
      "learning_rate": 5.207692307692308e-07,
      "loss": 0.0009,
      "reward": 1.267719328403473,
      "reward_std": 0.24669665843248367,
      "reward_std_all": 0.2220066487789154,
      "rewards-max/forensics_sinkhorn_reward": 0.9607784748077393,
      "rewards-max/forensics_strict_edge_reward": 0.24571846798062325,
      "rewards-max/forensics_tiou_reward": 0.9257514178752899,
      "rewards/forensics_sinkhorn_reward": 0.7202514708042145,
      "rewards/forensics_strict_edge_reward": 0.03478598361834884,
      "rewards/forensics_tiou_reward": 0.5126818716526031,
      "step": 624
    },
    {
      "completion_length": 23.8125,
      "epoch": 1.9185867895545314,
      "grad_norm": 9.82305908203125,
      "kl": 0.3125,
      "learning_rate": 5.2e-07,
      "loss": 0.0747,
      "reward": 1.1445437669754028,
      "reward_std": 0.2229153886437416,
      "reward_std_all": 0.2098945453763008,
      "rewards-max/forensics_sinkhorn_reward": 0.9762793481349945,
      "rewards-max/forensics_strict_edge_reward": 0.756708562374115,
      "rewards-max/forensics_tiou_reward": 0.9703385531902313,
      "rewards/forensics_sinkhorn_reward": 0.6454297006130219,
      "rewards/forensics_strict_edge_reward": 0.04761110991239548,
      "rewards/forensics_tiou_reward": 0.4515029489994049,
      "step": 625
    },
    {
      "completion_length": 24.078125,
      "epoch": 1.9216589861751152,
      "grad_norm": 10.225571632385254,
      "kl": 0.6337890625,
      "learning_rate": 5.192307692307692e-07,
      "loss": -0.0281,
      "reward": 1.4460718631744385,
      "reward_std": 0.14684680849313736,
      "reward_std_all": 0.17025341093540192,
      "rewards-max/forensics_sinkhorn_reward": 0.9684487283229828,
      "rewards-max/forensics_strict_edge_reward": 0.12067382782697678,
      "rewards-max/forensics_tiou_reward": 0.9491912424564362,
      "rewards/forensics_sinkhorn_reward": 0.8147998750209808,
      "rewards/forensics_strict_edge_reward": 0.011871017049998045,
      "rewards/forensics_tiou_reward": 0.6194008886814117,
      "step": 626
    },
    {
      "completion_length": 20.546875,
      "epoch": 1.924731182795699,
      "grad_norm": 20.014318466186523,
      "kl": 0.2333984375,
      "learning_rate": 5.184615384615384e-07,
      "loss": 0.0481,
      "reward": 1.3816941380500793,
      "reward_std": 0.1987130045890808,
      "reward_std_all": 0.18925072252750397,
      "rewards-max/forensics_sinkhorn_reward": 0.9717962443828583,
      "rewards-max/forensics_strict_edge_reward": 0.2680644690990448,
      "rewards-max/forensics_tiou_reward": 0.9573172628879547,
      "rewards/forensics_sinkhorn_reward": 0.7637619078159332,
      "rewards/forensics_strict_edge_reward": 0.03065147064626217,
      "rewards/forensics_tiou_reward": 0.5872807800769806,
      "step": 627
    },
    {
      "completion_length": 23.734375,
      "epoch": 1.9278033794162828,
      "grad_norm": 9.657833099365234,
      "kl": 0.19140625,
      "learning_rate": 5.176923076923076e-07,
      "loss": 0.0543,
      "reward": 1.3476236462593079,
      "reward_std": 0.245992973446846,
      "reward_std_all": 0.2221876084804535,
      "rewards-max/forensics_sinkhorn_reward": 0.9751774370670319,
      "rewards-max/forensics_strict_edge_reward": 0.46617764234542847,
      "rewards-max/forensics_tiou_reward": 0.9640930891036987,
      "rewards/forensics_sinkhorn_reward": 0.7567021548748016,
      "rewards/forensics_strict_edge_reward": 0.04333875887095928,
      "rewards/forensics_tiou_reward": 0.5475826561450958,
      "step": 628
    },
    {
      "completion_length": 21.96875,
      "epoch": 1.9308755760368663,
      "grad_norm": 10.59346866607666,
      "kl": 0.29833984375,
      "learning_rate": 5.16923076923077e-07,
      "loss": 0.0652,
      "reward": 1.2126662731170654,
      "reward_std": 0.1706179901957512,
      "reward_std_all": 0.17293208837509155,
      "rewards-max/forensics_sinkhorn_reward": 0.9430263638496399,
      "rewards-max/forensics_strict_edge_reward": 0.21826475113630295,
      "rewards-max/forensics_tiou_reward": 0.8880061209201813,
      "rewards/forensics_sinkhorn_reward": 0.7170166075229645,
      "rewards/forensics_strict_edge_reward": 0.014414407778531313,
      "rewards/forensics_tiou_reward": 0.4812352806329727,
      "step": 629
    },
    {
      "completion_length": 21.671875,
      "epoch": 1.93394777265745,
      "grad_norm": 13.640284538269043,
      "kl": 0.3134765625,
      "learning_rate": 5.161538461538462e-07,
      "loss": 0.0276,
      "reward": 1.2121272087097168,
      "reward_std": 0.25355929136276245,
      "reward_std_all": 0.2321174368262291,
      "rewards-max/forensics_sinkhorn_reward": 0.9531787037849426,
      "rewards-max/forensics_strict_edge_reward": 0.16782542318105698,
      "rewards-max/forensics_tiou_reward": 0.9129371345043182,
      "rewards/forensics_sinkhorn_reward": 0.6907870769500732,
      "rewards/forensics_strict_edge_reward": 0.01960540097206831,
      "rewards/forensics_tiou_reward": 0.5017347931861877,
      "step": 630
    },
    {
      "completion_length": 23.921875,
      "epoch": 1.9370199692780337,
      "grad_norm": 21.76150894165039,
      "kl": 1.58251953125,
      "learning_rate": 5.153846153846153e-07,
      "loss": 0.0232,
      "reward": 1.2944289445877075,
      "reward_std": 0.2832333743572235,
      "reward_std_all": 0.2544914558529854,
      "rewards-max/forensics_sinkhorn_reward": 0.9732629060745239,
      "rewards-max/forensics_strict_edge_reward": 0.43388937413692474,
      "rewards-max/forensics_tiou_reward": 0.9530937969684601,
      "rewards/forensics_sinkhorn_reward": 0.7041077017784119,
      "rewards/forensics_strict_edge_reward": 0.04645836725831032,
      "rewards/forensics_tiou_reward": 0.5438629388809204,
      "step": 631
    },
    {
      "completion_length": 20.9375,
      "epoch": 1.9400921658986174,
      "grad_norm": 11.369351387023926,
      "kl": 0.29296875,
      "learning_rate": 5.146153846153846e-07,
      "loss": -0.0162,
      "reward": 1.283233880996704,
      "reward_std": 0.2713906839489937,
      "reward_std_all": 0.23457077890634537,
      "rewards-max/forensics_sinkhorn_reward": 0.9390754401683807,
      "rewards-max/forensics_strict_edge_reward": 0.3360635004937649,
      "rewards-max/forensics_tiou_reward": 0.8699448704719543,
      "rewards/forensics_sinkhorn_reward": 0.7321602404117584,
      "rewards/forensics_strict_edge_reward": 0.05777016910724342,
      "rewards/forensics_tiou_reward": 0.4933035373687744,
      "step": 632
    },
    {
      "completion_length": 20.71875,
      "epoch": 1.9431643625192012,
      "grad_norm": 10.127374649047852,
      "kl": 0.330078125,
      "learning_rate": 5.138461538461538e-07,
      "loss": 0.183,
      "reward": 1.0519858598709106,
      "reward_std": 0.1964949667453766,
      "reward_std_all": 0.20148327946662903,
      "rewards-max/forensics_sinkhorn_reward": 0.9506230056285858,
      "rewards-max/forensics_strict_edge_reward": 0.1782933058857452,
      "rewards-max/forensics_tiou_reward": 0.8865641355514526,
      "rewards/forensics_sinkhorn_reward": 0.6450144350528717,
      "rewards/forensics_strict_edge_reward": 0.009357400849694386,
      "rewards/forensics_tiou_reward": 0.39761410653591156,
      "step": 633
    },
    {
      "completion_length": 22.6875,
      "epoch": 1.946236559139785,
      "grad_norm": 10.787442207336426,
      "kl": 0.34765625,
      "learning_rate": 5.13076923076923e-07,
      "loss": 0.0093,
      "reward": 1.33671635389328,
      "reward_std": 0.24979077279567719,
      "reward_std_all": 0.2596271485090256,
      "rewards-max/forensics_sinkhorn_reward": 0.939630001783371,
      "rewards-max/forensics_strict_edge_reward": 0.32734835892915726,
      "rewards-max/forensics_tiou_reward": 0.878074049949646,
      "rewards/forensics_sinkhorn_reward": 0.7448658347129822,
      "rewards/forensics_strict_edge_reward": 0.05842962022870779,
      "rewards/forensics_tiou_reward": 0.5334208905696869,
      "step": 634
    },
    {
      "completion_length": 23.921875,
      "epoch": 1.9493087557603688,
      "grad_norm": 8.504302024841309,
      "kl": 0.3427734375,
      "learning_rate": 5.123076923076923e-07,
      "loss": 0.08,
      "reward": 1.0319383442401886,
      "reward_std": 0.19423162192106247,
      "reward_std_all": 0.17275188118219376,
      "rewards-max/forensics_sinkhorn_reward": 0.8905341029167175,
      "rewards-max/forensics_strict_edge_reward": 0.289111465215683,
      "rewards-max/forensics_tiou_reward": 0.806753396987915,
      "rewards/forensics_sinkhorn_reward": 0.625797837972641,
      "rewards/forensics_strict_edge_reward": 0.016888640820980072,
      "rewards/forensics_tiou_reward": 0.38925187289714813,
      "step": 635
    },
    {
      "completion_length": 22.546875,
      "epoch": 1.9523809523809523,
      "grad_norm": 13.399771690368652,
      "kl": 0.4462890625,
      "learning_rate": 5.115384615384615e-07,
      "loss": 0.0839,
      "reward": 1.2833455801010132,
      "reward_std": 0.3021388575434685,
      "reward_std_all": 0.27703188359737396,
      "rewards-max/forensics_sinkhorn_reward": 0.9700398147106171,
      "rewards-max/forensics_strict_edge_reward": 0.27988626807928085,
      "rewards-max/forensics_tiou_reward": 0.941869705915451,
      "rewards/forensics_sinkhorn_reward": 0.707344263792038,
      "rewards/forensics_strict_edge_reward": 0.018360130488872528,
      "rewards/forensics_tiou_reward": 0.5576411187648773,
      "step": 636
    },
    {
      "completion_length": 23.609375,
      "epoch": 1.955453149001536,
      "grad_norm": 9.34805965423584,
      "kl": 0.2255859375,
      "learning_rate": 5.107692307692308e-07,
      "loss": 0.0119,
      "reward": 1.0615509748458862,
      "reward_std": 0.24632332473993301,
      "reward_std_all": 0.24684207141399384,
      "rewards-max/forensics_sinkhorn_reward": 0.9081614911556244,
      "rewards-max/forensics_strict_edge_reward": 0.34372806921601295,
      "rewards-max/forensics_tiou_reward": 0.8204611241817474,
      "rewards/forensics_sinkhorn_reward": 0.6433594226837158,
      "rewards/forensics_strict_edge_reward": 0.025693156057968736,
      "rewards/forensics_tiou_reward": 0.3924984931945801,
      "step": 637
    },
    {
      "completion_length": 23.640625,
      "epoch": 1.9585253456221197,
      "grad_norm": 8.666234970092773,
      "kl": 0.2578125,
      "learning_rate": 5.1e-07,
      "loss": 0.0346,
      "reward": 1.236330807209015,
      "reward_std": 0.18538758158683777,
      "reward_std_all": 0.1699579954147339,
      "rewards-max/forensics_sinkhorn_reward": 0.9225712418556213,
      "rewards-max/forensics_strict_edge_reward": 0.16316426545381546,
      "rewards-max/forensics_tiou_reward": 0.8441585302352905,
      "rewards/forensics_sinkhorn_reward": 0.7152060568332672,
      "rewards/forensics_strict_edge_reward": 0.015254631638526917,
      "rewards/forensics_tiou_reward": 0.5058700293302536,
      "step": 638
    },
    {
      "completion_length": 23.3125,
      "epoch": 1.9615975422427034,
      "grad_norm": 8.6369047164917,
      "kl": 0.2041015625,
      "learning_rate": 5.092307692307692e-07,
      "loss": 0.0174,
      "reward": 1.3717202544212341,
      "reward_std": 0.18063325434923172,
      "reward_std_all": 0.1632295697927475,
      "rewards-max/forensics_sinkhorn_reward": 0.9690464735031128,
      "rewards-max/forensics_strict_edge_reward": 0.3618253022432327,
      "rewards-max/forensics_tiou_reward": 0.9507711827754974,
      "rewards/forensics_sinkhorn_reward": 0.7774396538734436,
      "rewards/forensics_strict_edge_reward": 0.04159550555050373,
      "rewards/forensics_tiou_reward": 0.5526851415634155,
      "step": 639
    },
    {
      "completion_length": 19.9375,
      "epoch": 1.9646697388632872,
      "grad_norm": 11.23218059539795,
      "kl": 0.384765625,
      "learning_rate": 5.084615384615384e-07,
      "loss": 0.0327,
      "reward": 1.4011093974113464,
      "reward_std": 0.27491725981235504,
      "reward_std_all": 0.25883039087057114,
      "rewards-max/forensics_sinkhorn_reward": 0.9720661640167236,
      "rewards-max/forensics_strict_edge_reward": 0.4086042046546936,
      "rewards-max/forensics_tiou_reward": 0.9536961019039154,
      "rewards/forensics_sinkhorn_reward": 0.7656470537185669,
      "rewards/forensics_strict_edge_reward": 0.06946913711726665,
      "rewards/forensics_tiou_reward": 0.5659931004047394,
      "step": 640
    },
    {
      "completion_length": 19.671875,
      "epoch": 1.967741935483871,
      "grad_norm": 11.559683799743652,
      "kl": 0.17236328125,
      "learning_rate": 5.076923076923076e-07,
      "loss": 0.044,
      "reward": 1.2710363864898682,
      "reward_std": 0.2143954560160637,
      "reward_std_all": 0.21970205008983612,
      "rewards-max/forensics_sinkhorn_reward": 0.9624341130256653,
      "rewards-max/forensics_strict_edge_reward": 0.4505066052079201,
      "rewards-max/forensics_tiou_reward": 0.9368736147880554,
      "rewards/forensics_sinkhorn_reward": 0.7152146697044373,
      "rewards/forensics_strict_edge_reward": 0.05591512005776167,
      "rewards/forensics_tiou_reward": 0.49990662932395935,
      "step": 641
    },
    {
      "completion_length": 23.4375,
      "epoch": 1.9708141321044548,
      "grad_norm": 10.661343574523926,
      "kl": 0.3857421875,
      "learning_rate": 5.06923076923077e-07,
      "loss": 0.0446,
      "reward": 1.4962142705917358,
      "reward_std": 0.2614794671535492,
      "reward_std_all": 0.23301365971565247,
      "rewards-max/forensics_sinkhorn_reward": 0.9826452732086182,
      "rewards-max/forensics_strict_edge_reward": 0.6326714754104614,
      "rewards-max/forensics_tiou_reward": 0.9814728498458862,
      "rewards/forensics_sinkhorn_reward": 0.7954895496368408,
      "rewards/forensics_strict_edge_reward": 0.0719109233468771,
      "rewards/forensics_tiou_reward": 0.6288138031959534,
      "step": 642
    },
    {
      "completion_length": 20.890625,
      "epoch": 1.9738863287250386,
      "grad_norm": 10.913958549499512,
      "kl": 0.291015625,
      "learning_rate": 5.061538461538462e-07,
      "loss": 0.0375,
      "reward": 1.0978615283966064,
      "reward_std": 0.27188441157341003,
      "reward_std_all": 0.24397476017475128,
      "rewards-max/forensics_sinkhorn_reward": 0.9438847601413727,
      "rewards-max/forensics_strict_edge_reward": 0.13493308634497225,
      "rewards-max/forensics_tiou_reward": 0.8853196501731873,
      "rewards/forensics_sinkhorn_reward": 0.6667202115058899,
      "rewards/forensics_strict_edge_reward": 0.013072193483822048,
      "rewards/forensics_tiou_reward": 0.41806910932064056,
      "step": 643
    },
    {
      "completion_length": 20.46875,
      "epoch": 1.976958525345622,
      "grad_norm": 16.327037811279297,
      "kl": 0.2353515625,
      "learning_rate": 5.053846153846154e-07,
      "loss": 0.1142,
      "reward": 1.2986792922019958,
      "reward_std": 0.23914161324501038,
      "reward_std_all": 0.21113570779561996,
      "rewards-max/forensics_sinkhorn_reward": 0.9661409854888916,
      "rewards-max/forensics_strict_edge_reward": 0.18292110040783882,
      "rewards-max/forensics_tiou_reward": 0.9316317439079285,
      "rewards/forensics_sinkhorn_reward": 0.7506353259086609,
      "rewards/forensics_strict_edge_reward": 0.014127135276794434,
      "rewards/forensics_tiou_reward": 0.5339167416095734,
      "step": 644
    },
    {
      "completion_length": 22.765625,
      "epoch": 1.9800307219662059,
      "grad_norm": 11.135578155517578,
      "kl": 0.4599609375,
      "learning_rate": 5.046153846153846e-07,
      "loss": 0.0582,
      "reward": 1.2307060360908508,
      "reward_std": 0.30716800689697266,
      "reward_std_all": 0.31290389597415924,
      "rewards-max/forensics_sinkhorn_reward": 0.9663506746292114,
      "rewards-max/forensics_strict_edge_reward": 0.6720768958330154,
      "rewards-max/forensics_tiou_reward": 0.9482958316802979,
      "rewards/forensics_sinkhorn_reward": 0.6931510865688324,
      "rewards/forensics_strict_edge_reward": 0.05783556308597326,
      "rewards/forensics_tiou_reward": 0.4797193706035614,
      "step": 645
    },
    {
      "completion_length": 21.140625,
      "epoch": 1.9831029185867894,
      "grad_norm": 7.556153774261475,
      "kl": 0.30859375,
      "learning_rate": 5.038461538461538e-07,
      "loss": -0.0583,
      "reward": 1.3469000458717346,
      "reward_std": 0.23901376873254776,
      "reward_std_all": 0.22737212479114532,
      "rewards-max/forensics_sinkhorn_reward": 0.9707264602184296,
      "rewards-max/forensics_strict_edge_reward": 0.41129170358181,
      "rewards-max/forensics_tiou_reward": 0.9498281180858612,
      "rewards/forensics_sinkhorn_reward": 0.7700301706790924,
      "rewards/forensics_strict_edge_reward": 0.03619994316250086,
      "rewards/forensics_tiou_reward": 0.5406699776649475,
      "step": 646
    },
    {
      "completion_length": 21.265625,
      "epoch": 1.9861751152073732,
      "grad_norm": 13.882674217224121,
      "kl": 0.2333984375,
      "learning_rate": 5.03076923076923e-07,
      "loss": -0.0316,
      "reward": 1.2798756957054138,
      "reward_std": 0.266462579369545,
      "reward_std_all": 0.2525346875190735,
      "rewards-max/forensics_sinkhorn_reward": 0.9573278427124023,
      "rewards-max/forensics_strict_edge_reward": 0.25612417608499527,
      "rewards-max/forensics_tiou_reward": 0.921784520149231,
      "rewards/forensics_sinkhorn_reward": 0.7312763333320618,
      "rewards/forensics_strict_edge_reward": 0.03795302612707019,
      "rewards/forensics_tiou_reward": 0.5106463730335236,
      "step": 647
    },
    {
      "completion_length": 22.03125,
      "epoch": 1.989247311827957,
      "grad_norm": 11.335634231567383,
      "kl": 0.83984375,
      "learning_rate": 5.023076923076923e-07,
      "loss": 0.0948,
      "reward": 1.1316941976547241,
      "reward_std": 0.31287357956171036,
      "reward_std_all": 0.2816288024187088,
      "rewards-max/forensics_sinkhorn_reward": 0.9361710250377655,
      "rewards-max/forensics_strict_edge_reward": 0.17691032588481903,
      "rewards-max/forensics_tiou_reward": 0.876641184091568,
      "rewards/forensics_sinkhorn_reward": 0.6635347008705139,
      "rewards/forensics_strict_edge_reward": 0.014891460305079818,
      "rewards/forensics_tiou_reward": 0.4532681107521057,
      "step": 648
    },
    {
      "completion_length": 22.671875,
      "epoch": 1.9923195084485408,
      "grad_norm": 10.82787799835205,
      "kl": 0.2841796875,
      "learning_rate": 5.015384615384614e-07,
      "loss": 0.0673,
      "reward": 1.3516386151313782,
      "reward_std": 0.22014014422893524,
      "reward_std_all": 0.23912344872951508,
      "rewards-max/forensics_sinkhorn_reward": 0.9628037810325623,
      "rewards-max/forensics_strict_edge_reward": 0.3020675405859947,
      "rewards-max/forensics_tiou_reward": 0.9304096102714539,
      "rewards/forensics_sinkhorn_reward": 0.7500154078006744,
      "rewards/forensics_strict_edge_reward": 0.0453141825273633,
      "rewards/forensics_tiou_reward": 0.5563089847564697,
      "step": 649
    },
    {
      "completion_length": 20.515625,
      "epoch": 1.9953917050691246,
      "grad_norm": 11.264650344848633,
      "kl": 0.29296875,
      "learning_rate": 5.007692307692308e-07,
      "loss": 0.0429,
      "reward": 1.198793351650238,
      "reward_std": 0.35764622688293457,
      "reward_std_all": 0.32725051045417786,
      "rewards-max/forensics_sinkhorn_reward": 0.9678898453712463,
      "rewards-max/forensics_strict_edge_reward": 0.19672049209475517,
      "rewards-max/forensics_tiou_reward": 0.9411135315895081,
      "rewards/forensics_sinkhorn_reward": 0.6951592564582825,
      "rewards/forensics_strict_edge_reward": 0.019206988101359457,
      "rewards/forensics_tiou_reward": 0.4844270199537277,
      "step": 650
    },
    {
      "completion_length": 21.65625,
      "epoch": 1.998463901689708,
      "grad_norm": 8.913248062133789,
      "kl": 0.22021484375,
      "learning_rate": 5e-07,
      "loss": 0.0627,
      "reward": 1.4491048455238342,
      "reward_std": 0.19303718954324722,
      "reward_std_all": 0.18441380560398102,
      "rewards-max/forensics_sinkhorn_reward": 0.9810202419757843,
      "rewards-max/forensics_strict_edge_reward": 0.5254523754119873,
      "rewards-max/forensics_tiou_reward": 0.9713445007801056,
      "rewards/forensics_sinkhorn_reward": 0.791386067867279,
      "rewards/forensics_strict_edge_reward": 0.04389688326045871,
      "rewards/forensics_tiou_reward": 0.6138218939304352,
      "step": 651
    },
    {
      "completion_length": 20.0,
      "epoch": 2.0,
      "grad_norm": 6.179769515991211,
      "kl": 0.515625,
      "learning_rate": 4.992307692307692e-07,
      "loss": 0.0578,
      "reward": 0.9276540875434875,
      "reward_std": 0.41307589411735535,
      "reward_std_all": 0.4715435206890106,
      "rewards-max/forensics_sinkhorn_reward": 0.7534134387969971,
      "rewards-max/forensics_strict_edge_reward": 0.014304968528449535,
      "rewards-max/forensics_tiou_reward": 0.5860474705696106,
      "rewards/forensics_sinkhorn_reward": 0.6078674793243408,
      "rewards/forensics_strict_edge_reward": 0.0029609226621687412,
      "rewards/forensics_tiou_reward": 0.3168257176876068,
      "step": 652
    },
    {
      "completion_length": 19.4375,
      "epoch": 2.003072196620584,
      "grad_norm": 10.419689178466797,
      "kl": 0.22265625,
      "learning_rate": 4.984615384615384e-07,
      "loss": 0.0307,
      "reward": 1.3822983503341675,
      "reward_std": 0.267670676112175,
      "reward_std_all": 0.25770460814237595,
      "rewards-max/forensics_sinkhorn_reward": 0.9729809463024139,
      "rewards-max/forensics_strict_edge_reward": 0.5306462496519089,
      "rewards-max/forensics_tiou_reward": 0.9536267220973969,
      "rewards/forensics_sinkhorn_reward": 0.7508310973644257,
      "rewards/forensics_strict_edge_reward": 0.05484568327665329,
      "rewards/forensics_tiou_reward": 0.5766216367483139,
      "step": 653
    },
    {
      "completion_length": 21.9375,
      "epoch": 2.0061443932411676,
      "grad_norm": 12.048803329467773,
      "kl": 0.359375,
      "learning_rate": 4.976923076923077e-07,
      "loss": -0.0,
      "reward": 1.2619937658309937,
      "reward_std": 0.38836178183555603,
      "reward_std_all": 0.3483675569295883,
      "rewards-max/forensics_sinkhorn_reward": 0.9780583083629608,
      "rewards-max/forensics_strict_edge_reward": 0.5066750049591064,
      "rewards-max/forensics_tiou_reward": 0.9583874940872192,
      "rewards/forensics_sinkhorn_reward": 0.7206994891166687,
      "rewards/forensics_strict_edge_reward": 0.06768058985471725,
      "rewards/forensics_tiou_reward": 0.4736136198043823,
      "step": 654
    },
    {
      "completion_length": 21.328125,
      "epoch": 2.0092165898617513,
      "grad_norm": 10.492072105407715,
      "kl": 0.361328125,
      "learning_rate": 4.969230769230769e-07,
      "loss": -0.0355,
      "reward": 1.1828187108039856,
      "reward_std": 0.23537477850914001,
      "reward_std_all": 0.22306810319423676,
      "rewards-max/forensics_sinkhorn_reward": 0.9678520858287811,
      "rewards-max/forensics_strict_edge_reward": 0.3134443461894989,
      "rewards-max/forensics_tiou_reward": 0.9440284371376038,
      "rewards/forensics_sinkhorn_reward": 0.7045721113681793,
      "rewards/forensics_strict_edge_reward": 0.02403939049690962,
      "rewards/forensics_tiou_reward": 0.45420724153518677,
      "step": 655
    },
    {
      "completion_length": 21.84375,
      "epoch": 2.0122887864823347,
      "grad_norm": 12.23831844329834,
      "kl": 0.6806640625,
      "learning_rate": 4.961538461538462e-07,
      "loss": -0.0236,
      "reward": 1.4827749133110046,
      "reward_std": 0.2847675532102585,
      "reward_std_all": 0.26488588750362396,
      "rewards-max/forensics_sinkhorn_reward": 0.9800847470760345,
      "rewards-max/forensics_strict_edge_reward": 0.6408846378326416,
      "rewards-max/forensics_tiou_reward": 0.9733217656612396,
      "rewards/forensics_sinkhorn_reward": 0.7840623259544373,
      "rewards/forensics_strict_edge_reward": 0.08441182225942612,
      "rewards/forensics_tiou_reward": 0.6143007576465607,
      "step": 656
    },
    {
      "completion_length": 22.4375,
      "epoch": 2.0153609831029184,
      "grad_norm": 9.075726509094238,
      "kl": 0.32421875,
      "learning_rate": 4.953846153846154e-07,
      "loss": -0.0563,
      "reward": 1.3222858905792236,
      "reward_std": 0.3493541553616524,
      "reward_std_all": 0.30930837988853455,
      "rewards-max/forensics_sinkhorn_reward": 0.9600304961204529,
      "rewards-max/forensics_strict_edge_reward": 0.5583210289478302,
      "rewards-max/forensics_tiou_reward": 0.9222032427787781,
      "rewards/forensics_sinkhorn_reward": 0.733660876750946,
      "rewards/forensics_strict_edge_reward": 0.05793050676584244,
      "rewards/forensics_tiou_reward": 0.530694454908371,
      "step": 657
    },
    {
      "completion_length": 20.671875,
      "epoch": 2.0184331797235022,
      "grad_norm": 7.816912651062012,
      "kl": 0.29345703125,
      "learning_rate": 4.946153846153846e-07,
      "loss": 0.0878,
      "reward": 1.1766808032989502,
      "reward_std": 0.19099336117506027,
      "reward_std_all": 0.20708982646465302,
      "rewards-max/forensics_sinkhorn_reward": 0.9603576958179474,
      "rewards-max/forensics_strict_edge_reward": 0.5185378938913345,
      "rewards-max/forensics_tiou_reward": 0.9307584762573242,
      "rewards/forensics_sinkhorn_reward": 0.6670965552330017,
      "rewards/forensics_strict_edge_reward": 0.052243251353502274,
      "rewards/forensics_tiou_reward": 0.4573409855365753,
      "step": 658
    },
    {
      "completion_length": 20.171875,
      "epoch": 2.021505376344086,
      "grad_norm": 16.130807876586914,
      "kl": 0.39501953125,
      "learning_rate": 4.938461538461538e-07,
      "loss": 0.0615,
      "reward": 1.2037447690963745,
      "reward_std": 0.20671642571687698,
      "reward_std_all": 0.19666186720132828,
      "rewards-max/forensics_sinkhorn_reward": 0.9699050784111023,
      "rewards-max/forensics_strict_edge_reward": 0.22266601026058197,
      "rewards-max/forensics_tiou_reward": 0.9393965899944305,
      "rewards/forensics_sinkhorn_reward": 0.6877716481685638,
      "rewards/forensics_strict_edge_reward": 0.023814686574041843,
      "rewards/forensics_tiou_reward": 0.4921584129333496,
      "step": 659
    },
    {
      "completion_length": 20.234375,
      "epoch": 2.02457757296467,
      "grad_norm": 9.08634090423584,
      "kl": 0.2265625,
      "learning_rate": 4.93076923076923e-07,
      "loss": 0.0903,
      "reward": 1.2483885884284973,
      "reward_std": 0.25651078671216965,
      "reward_std_all": 0.24036794900894165,
      "rewards-max/forensics_sinkhorn_reward": 0.9703538417816162,
      "rewards-max/forensics_strict_edge_reward": 0.6254801601171494,
      "rewards-max/forensics_tiou_reward": 0.9550673961639404,
      "rewards/forensics_sinkhorn_reward": 0.7230103313922882,
      "rewards/forensics_strict_edge_reward": 0.04382913559675217,
      "rewards/forensics_tiou_reward": 0.4815490245819092,
      "step": 660
    },
    {
      "completion_length": 24.78125,
      "epoch": 2.0276497695852536,
      "grad_norm": 8.12386417388916,
      "kl": 0.4345703125,
      "learning_rate": 4.923076923076923e-07,
      "loss": 0.1085,
      "reward": 1.2259779572486877,
      "reward_std": 0.20717286318540573,
      "reward_std_all": 0.21457050740718842,
      "rewards-max/forensics_sinkhorn_reward": 0.9314771890640259,
      "rewards-max/forensics_strict_edge_reward": 0.14635462500154972,
      "rewards-max/forensics_tiou_reward": 0.8674584031105042,
      "rewards/forensics_sinkhorn_reward": 0.727020800113678,
      "rewards/forensics_strict_edge_reward": 0.011853282107040286,
      "rewards/forensics_tiou_reward": 0.4871038347482681,
      "step": 661
    },
    {
      "completion_length": 21.359375,
      "epoch": 2.0307219662058373,
      "grad_norm": 8.528206825256348,
      "kl": 0.25,
      "learning_rate": 4.915384615384615e-07,
      "loss": 0.0453,
      "reward": 1.4433095455169678,
      "reward_std": 0.245540589094162,
      "reward_std_all": 0.29157398641109467,
      "rewards-max/forensics_sinkhorn_reward": 0.9723297953605652,
      "rewards-max/forensics_strict_edge_reward": 0.49006423354148865,
      "rewards-max/forensics_tiou_reward": 0.958185464143753,
      "rewards/forensics_sinkhorn_reward": 0.7818073630332947,
      "rewards/forensics_strict_edge_reward": 0.06762848608195782,
      "rewards/forensics_tiou_reward": 0.5938736796379089,
      "step": 662
    },
    {
      "completion_length": 22.171875,
      "epoch": 2.033794162826421,
      "grad_norm": 8.617594718933105,
      "kl": 0.701171875,
      "learning_rate": 4.907692307692308e-07,
      "loss": 0.1744,
      "reward": 1.5263752937316895,
      "reward_std": 0.318246066570282,
      "reward_std_all": 0.28399576246738434,
      "rewards-max/forensics_sinkhorn_reward": 0.975097268819809,
      "rewards-max/forensics_strict_edge_reward": 0.7426987290382385,
      "rewards-max/forensics_tiou_reward": 0.9652537107467651,
      "rewards/forensics_sinkhorn_reward": 0.8147575557231903,
      "rewards/forensics_strict_edge_reward": 0.0747770220041275,
      "rewards/forensics_tiou_reward": 0.6368406713008881,
      "step": 663
    },
    {
      "completion_length": 20.890625,
      "epoch": 2.0368663594470044,
      "grad_norm": 13.04763126373291,
      "kl": 0.28173828125,
      "learning_rate": 4.9e-07,
      "loss": 0.0362,
      "reward": 1.2314855754375458,
      "reward_std": 0.15409980714321136,
      "reward_std_all": 0.15568266063928604,
      "rewards-max/forensics_sinkhorn_reward": 0.9266935586929321,
      "rewards-max/forensics_strict_edge_reward": 0.3428696943446994,
      "rewards-max/forensics_tiou_reward": 0.8386528491973877,
      "rewards/forensics_sinkhorn_reward": 0.7207825481891632,
      "rewards/forensics_strict_edge_reward": 0.018848414765670896,
      "rewards/forensics_tiou_reward": 0.49185462296009064,
      "step": 664
    },
    {
      "completion_length": 21.78125,
      "epoch": 2.0399385560675882,
      "grad_norm": 10.24178409576416,
      "kl": 0.28662109375,
      "learning_rate": 4.892307692307692e-07,
      "loss": 0.0139,
      "reward": 1.4440789222717285,
      "reward_std": 0.1847495287656784,
      "reward_std_all": 0.18673518300056458,
      "rewards-max/forensics_sinkhorn_reward": 0.9544202983379364,
      "rewards-max/forensics_strict_edge_reward": 0.48662105202674866,
      "rewards-max/forensics_tiou_reward": 0.9187863767147064,
      "rewards/forensics_sinkhorn_reward": 0.8062006831169128,
      "rewards/forensics_strict_edge_reward": 0.034241050481796265,
      "rewards/forensics_tiou_reward": 0.603637307882309,
      "step": 665
    },
    {
      "completion_length": 20.375,
      "epoch": 2.043010752688172,
      "grad_norm": 13.839912414550781,
      "kl": 0.21728515625,
      "learning_rate": 4.884615384615384e-07,
      "loss": 0.0581,
      "reward": 1.3406955003738403,
      "reward_std": 0.2901540845632553,
      "reward_std_all": 0.2508106306195259,
      "rewards-max/forensics_sinkhorn_reward": 0.9592928886413574,
      "rewards-max/forensics_strict_edge_reward": 0.27228695154190063,
      "rewards-max/forensics_tiou_reward": 0.9234538674354553,
      "rewards/forensics_sinkhorn_reward": 0.7713226974010468,
      "rewards/forensics_strict_edge_reward": 0.032814797945320606,
      "rewards/forensics_tiou_reward": 0.5365579724311829,
      "step": 666
    },
    {
      "completion_length": 20.640625,
      "epoch": 2.046082949308756,
      "grad_norm": 10.931879043579102,
      "kl": 0.19482421875,
      "learning_rate": 4.876923076923077e-07,
      "loss": 0.0848,
      "reward": 1.1790240406990051,
      "reward_std": 0.18732669949531555,
      "reward_std_all": 0.16516737639904022,
      "rewards-max/forensics_sinkhorn_reward": 0.961209774017334,
      "rewards-max/forensics_strict_edge_reward": 0.5124467667192221,
      "rewards-max/forensics_tiou_reward": 0.9180887639522552,
      "rewards/forensics_sinkhorn_reward": 0.6942808330059052,
      "rewards/forensics_strict_edge_reward": 0.03725081030279398,
      "rewards/forensics_tiou_reward": 0.4474923312664032,
      "step": 667
    },
    {
      "completion_length": 21.578125,
      "epoch": 2.0491551459293396,
      "grad_norm": 233.27175903320312,
      "kl": 6.1162109375,
      "learning_rate": 4.869230769230769e-07,
      "loss": 0.2488,
      "reward": 1.4271222949028015,
      "reward_std": 0.16846998780965805,
      "reward_std_all": 0.19205661863088608,
      "rewards-max/forensics_sinkhorn_reward": 0.9838642179965973,
      "rewards-max/forensics_strict_edge_reward": 0.4259992577135563,
      "rewards-max/forensics_tiou_reward": 0.9804616570472717,
      "rewards/forensics_sinkhorn_reward": 0.7744006514549255,
      "rewards/forensics_strict_edge_reward": 0.04964219429530203,
      "rewards/forensics_tiou_reward": 0.6030793935060501,
      "step": 668
    },
    {
      "completion_length": 20.609375,
      "epoch": 2.0522273425499233,
      "grad_norm": 12.379220962524414,
      "kl": 0.294921875,
      "learning_rate": 4.861538461538461e-07,
      "loss": -0.0642,
      "reward": 1.3452625274658203,
      "reward_std": 0.20675967633724213,
      "reward_std_all": 0.19967617094516754,
      "rewards-max/forensics_sinkhorn_reward": 0.9731328785419464,
      "rewards-max/forensics_strict_edge_reward": 0.40798234194517136,
      "rewards-max/forensics_tiou_reward": 0.9560041725635529,
      "rewards/forensics_sinkhorn_reward": 0.7493141889572144,
      "rewards/forensics_strict_edge_reward": 0.055604007095098495,
      "rewards/forensics_tiou_reward": 0.5403443574905396,
      "step": 669
    },
    {
      "completion_length": 21.609375,
      "epoch": 2.055299539170507,
      "grad_norm": 13.649415016174316,
      "kl": 0.2197265625,
      "learning_rate": 4.853846153846154e-07,
      "loss": 0.0355,
      "reward": 1.277322769165039,
      "reward_std": 0.20535815507173538,
      "reward_std_all": 0.2016584277153015,
      "rewards-max/forensics_sinkhorn_reward": 0.9572178721427917,
      "rewards-max/forensics_strict_edge_reward": 0.403377428650856,
      "rewards-max/forensics_tiou_reward": 0.9199503660202026,
      "rewards/forensics_sinkhorn_reward": 0.7238505780696869,
      "rewards/forensics_strict_edge_reward": 0.03390225302428007,
      "rewards/forensics_tiou_reward": 0.5195700228214264,
      "step": 670
    },
    {
      "completion_length": 20.828125,
      "epoch": 2.0583717357910905,
      "grad_norm": 8.30440902709961,
      "kl": 0.30615234375,
      "learning_rate": 4.846153846153846e-07,
      "loss": -0.0272,
      "reward": 1.3831011652946472,
      "reward_std": 0.2677600383758545,
      "reward_std_all": 0.2567658945918083,
      "rewards-max/forensics_sinkhorn_reward": 0.9803620278835297,
      "rewards-max/forensics_strict_edge_reward": 0.51341712474823,
      "rewards-max/forensics_tiou_reward": 0.9761644899845123,
      "rewards/forensics_sinkhorn_reward": 0.7660529017448425,
      "rewards/forensics_strict_edge_reward": 0.045878686010837555,
      "rewards/forensics_tiou_reward": 0.5711696147918701,
      "step": 671
    },
    {
      "completion_length": 22.75,
      "epoch": 2.0614439324116742,
      "grad_norm": 7.5090789794921875,
      "kl": 0.29345703125,
      "learning_rate": 4.838461538461539e-07,
      "loss": -0.0621,
      "reward": 1.400919795036316,
      "reward_std": 0.18582656234502792,
      "reward_std_all": 0.172136090695858,
      "rewards-max/forensics_sinkhorn_reward": 0.9492270648479462,
      "rewards-max/forensics_strict_edge_reward": 0.4628256903961301,
      "rewards-max/forensics_tiou_reward": 0.890861302614212,
      "rewards/forensics_sinkhorn_reward": 0.7629934251308441,
      "rewards/forensics_strict_edge_reward": 0.07537848339416087,
      "rewards/forensics_tiou_reward": 0.562547892332077,
      "step": 672
    },
    {
      "completion_length": 21.9375,
      "epoch": 2.064516129032258,
      "grad_norm": 6.766315460205078,
      "kl": 0.416015625,
      "learning_rate": 4.83076923076923e-07,
      "loss": 0.0423,
      "reward": 1.342289686203003,
      "reward_std": 0.19506311416625977,
      "reward_std_all": 0.17635612934827805,
      "rewards-max/forensics_sinkhorn_reward": 0.9568563997745514,
      "rewards-max/forensics_strict_edge_reward": 0.2221851944923401,
      "rewards-max/forensics_tiou_reward": 0.9189504086971283,
      "rewards/forensics_sinkhorn_reward": 0.763554036617279,
      "rewards/forensics_strict_edge_reward": 0.02492811670526862,
      "rewards/forensics_tiou_reward": 0.5538074374198914,
      "step": 673
    },
    {
      "completion_length": 19.734375,
      "epoch": 2.067588325652842,
      "grad_norm": 10.239224433898926,
      "kl": 0.1640625,
      "learning_rate": 4.823076923076923e-07,
      "loss": -0.1116,
      "reward": 1.4241540431976318,
      "reward_std": 0.22723129391670227,
      "reward_std_all": 0.1936441734433174,
      "rewards-max/forensics_sinkhorn_reward": 0.9771170616149902,
      "rewards-max/forensics_strict_edge_reward": 0.4538624435663223,
      "rewards-max/forensics_tiou_reward": 0.9658126533031464,
      "rewards/forensics_sinkhorn_reward": 0.7765375971794128,
      "rewards/forensics_strict_edge_reward": 0.05954858846962452,
      "rewards/forensics_tiou_reward": 0.5880677402019501,
      "step": 674
    },
    {
      "completion_length": 21.484375,
      "epoch": 2.0706605222734256,
      "grad_norm": 7.984202861785889,
      "kl": 0.4599609375,
      "learning_rate": 4.815384615384615e-07,
      "loss": 0.0118,
      "reward": 1.2783018350601196,
      "reward_std": 0.220899298787117,
      "reward_std_all": 0.20294400304555893,
      "rewards-max/forensics_sinkhorn_reward": 0.9685199856758118,
      "rewards-max/forensics_strict_edge_reward": 0.3856436163187027,
      "rewards-max/forensics_tiou_reward": 0.943985104560852,
      "rewards/forensics_sinkhorn_reward": 0.7248869240283966,
      "rewards/forensics_strict_edge_reward": 0.03529500588774681,
      "rewards/forensics_tiou_reward": 0.5181199461221695,
      "step": 675
    },
    {
      "completion_length": 21.8125,
      "epoch": 2.0737327188940093,
      "grad_norm": 14.597387313842773,
      "kl": 0.875,
      "learning_rate": 4.807692307692307e-07,
      "loss": 0.0216,
      "reward": 1.4431641101837158,
      "reward_std": 0.3286864906549454,
      "reward_std_all": 0.3221573531627655,
      "rewards-max/forensics_sinkhorn_reward": 0.9621735215187073,
      "rewards-max/forensics_strict_edge_reward": 0.31592947617173195,
      "rewards-max/forensics_tiou_reward": 0.9347793459892273,
      "rewards/forensics_sinkhorn_reward": 0.7759254276752472,
      "rewards/forensics_strict_edge_reward": 0.04341453593224287,
      "rewards/forensics_tiou_reward": 0.6238241195678711,
      "step": 676
    },
    {
      "completion_length": 20.65625,
      "epoch": 2.076804915514593,
      "grad_norm": 8.012065887451172,
      "kl": 0.3994140625,
      "learning_rate": 4.8e-07,
      "loss": 0.0052,
      "reward": 1.2905308604240417,
      "reward_std": 0.25649598240852356,
      "reward_std_all": 0.2260332629084587,
      "rewards-max/forensics_sinkhorn_reward": 0.920527458190918,
      "rewards-max/forensics_strict_edge_reward": 0.17052182741463184,
      "rewards-max/forensics_tiou_reward": 0.8345467150211334,
      "rewards/forensics_sinkhorn_reward": 0.7376585006713867,
      "rewards/forensics_strict_edge_reward": 0.017785617616027594,
      "rewards/forensics_tiou_reward": 0.5350868105888367,
      "step": 677
    },
    {
      "completion_length": 23.546875,
      "epoch": 2.0798771121351765,
      "grad_norm": 9.367371559143066,
      "kl": 0.4541015625,
      "learning_rate": 4.792307692307692e-07,
      "loss": -0.0489,
      "reward": 1.56876939535141,
      "reward_std": 0.24625280499458313,
      "reward_std_all": 0.25177671760320663,
      "rewards-max/forensics_sinkhorn_reward": 0.9682217240333557,
      "rewards-max/forensics_strict_edge_reward": 0.49075204133987427,
      "rewards-max/forensics_tiou_reward": 0.9442335069179535,
      "rewards/forensics_sinkhorn_reward": 0.836572527885437,
      "rewards/forensics_strict_edge_reward": 0.05843675881624222,
      "rewards/forensics_tiou_reward": 0.6737600564956665,
      "step": 678
    },
    {
      "completion_length": 20.765625,
      "epoch": 2.0829493087557602,
      "grad_norm": 12.891510009765625,
      "kl": 0.369140625,
      "learning_rate": 4.784615384615384e-07,
      "loss": -0.0302,
      "reward": 1.1442530155181885,
      "reward_std": 0.2467004656791687,
      "reward_std_all": 0.2207518219947815,
      "rewards-max/forensics_sinkhorn_reward": 0.9626556038856506,
      "rewards-max/forensics_strict_edge_reward": 0.2824089303612709,
      "rewards-max/forensics_tiou_reward": 0.9342567324638367,
      "rewards/forensics_sinkhorn_reward": 0.6839489042758942,
      "rewards/forensics_strict_edge_reward": 0.02915822621434927,
      "rewards/forensics_tiou_reward": 0.43114587664604187,
      "step": 679
    },
    {
      "completion_length": 21.8125,
      "epoch": 2.086021505376344,
      "grad_norm": 13.45989990234375,
      "kl": 0.6435546875,
      "learning_rate": 4.776923076923077e-07,
      "loss": -0.0212,
      "reward": 1.237926959991455,
      "reward_std": 0.1830049604177475,
      "reward_std_all": 0.23374482244253159,
      "rewards-max/forensics_sinkhorn_reward": 0.9573370814323425,
      "rewards-max/forensics_strict_edge_reward": 0.7753324806690216,
      "rewards-max/forensics_tiou_reward": 0.919208824634552,
      "rewards/forensics_sinkhorn_reward": 0.6955784559249878,
      "rewards/forensics_strict_edge_reward": 0.05191292054951191,
      "rewards/forensics_tiou_reward": 0.49043555557727814,
      "step": 680
    },
    {
      "completion_length": 21.625,
      "epoch": 2.089093701996928,
      "grad_norm": 13.125267028808594,
      "kl": 0.359375,
      "learning_rate": 4.769230769230769e-07,
      "loss": 0.0099,
      "reward": 1.4470187425613403,
      "reward_std": 0.29369376599788666,
      "reward_std_all": 0.27710896730422974,
      "rewards-max/forensics_sinkhorn_reward": 0.972739964723587,
      "rewards-max/forensics_strict_edge_reward": 0.6495180428028107,
      "rewards-max/forensics_tiou_reward": 0.9611165225505829,
      "rewards/forensics_sinkhorn_reward": 0.7690845429897308,
      "rewards/forensics_strict_edge_reward": 0.08212563022971153,
      "rewards/forensics_tiou_reward": 0.59580859541893,
      "step": 681
    },
    {
      "completion_length": 20.234375,
      "epoch": 2.0921658986175116,
      "grad_norm": 10.251887321472168,
      "kl": 0.2734375,
      "learning_rate": 4.7615384615384613e-07,
      "loss": 0.1095,
      "reward": 1.24835604429245,
      "reward_std": 0.22379887849092484,
      "reward_std_all": 0.2237081229686737,
      "rewards-max/forensics_sinkhorn_reward": 0.9542647302150726,
      "rewards-max/forensics_strict_edge_reward": 0.6382447481155396,
      "rewards-max/forensics_tiou_reward": 0.9075275659561157,
      "rewards/forensics_sinkhorn_reward": 0.7257578074932098,
      "rewards/forensics_strict_edge_reward": 0.04440947808325291,
      "rewards/forensics_tiou_reward": 0.47818873822689056,
      "step": 682
    },
    {
      "completion_length": 21.25,
      "epoch": 2.0952380952380953,
      "grad_norm": 9.898710250854492,
      "kl": 0.32421875,
      "learning_rate": 4.753846153846154e-07,
      "loss": 0.01,
      "reward": 1.4213398098945618,
      "reward_std": 0.31859631836414337,
      "reward_std_all": 0.34195801615715027,
      "rewards-max/forensics_sinkhorn_reward": 0.9801869988441467,
      "rewards-max/forensics_strict_edge_reward": 0.8135445415973663,
      "rewards-max/forensics_tiou_reward": 0.9793037176132202,
      "rewards/forensics_sinkhorn_reward": 0.7452457845211029,
      "rewards/forensics_strict_edge_reward": 0.09249339252710342,
      "rewards/forensics_tiou_reward": 0.583600640296936,
      "step": 683
    },
    {
      "completion_length": 21.0625,
      "epoch": 2.098310291858679,
      "grad_norm": 10.48886489868164,
      "kl": 0.25634765625,
      "learning_rate": 4.7461538461538456e-07,
      "loss": -0.0506,
      "reward": 1.330422818660736,
      "reward_std": 0.20722801983356476,
      "reward_std_all": 0.18787085264921188,
      "rewards-max/forensics_sinkhorn_reward": 0.975653737783432,
      "rewards-max/forensics_strict_edge_reward": 0.545335203409195,
      "rewards-max/forensics_tiou_reward": 0.9578148722648621,
      "rewards/forensics_sinkhorn_reward": 0.7447150647640228,
      "rewards/forensics_strict_edge_reward": 0.0548967570066452,
      "rewards/forensics_tiou_reward": 0.5308109670877457,
      "step": 684
    },
    {
      "completion_length": 20.109375,
      "epoch": 2.1013824884792625,
      "grad_norm": 10.99982738494873,
      "kl": 0.3232421875,
      "learning_rate": 4.738461538461538e-07,
      "loss": -0.0046,
      "reward": 1.2261282801628113,
      "reward_std": 0.2851061224937439,
      "reward_std_all": 0.244568333029747,
      "rewards-max/forensics_sinkhorn_reward": 0.9824151694774628,
      "rewards-max/forensics_strict_edge_reward": 0.7533732652664185,
      "rewards-max/forensics_tiou_reward": 0.9844291806221008,
      "rewards/forensics_sinkhorn_reward": 0.7136977910995483,
      "rewards/forensics_strict_edge_reward": 0.05513334088027477,
      "rewards/forensics_tiou_reward": 0.4572971761226654,
      "step": 685
    },
    {
      "completion_length": 21.625,
      "epoch": 2.1044546850998462,
      "grad_norm": 9.236431121826172,
      "kl": 0.43359375,
      "learning_rate": 4.7307692307692304e-07,
      "loss": 0.0634,
      "reward": 1.491131067276001,
      "reward_std": 0.18141089752316475,
      "reward_std_all": 0.16866928339004517,
      "rewards-max/forensics_sinkhorn_reward": 0.9790413677692413,
      "rewards-max/forensics_strict_edge_reward": 0.47400766611099243,
      "rewards-max/forensics_tiou_reward": 0.973025918006897,
      "rewards/forensics_sinkhorn_reward": 0.7933251559734344,
      "rewards/forensics_strict_edge_reward": 0.056504152715206146,
      "rewards/forensics_tiou_reward": 0.641301691532135,
      "step": 686
    },
    {
      "completion_length": 20.59375,
      "epoch": 2.10752688172043,
      "grad_norm": 11.131351470947266,
      "kl": 0.35888671875,
      "learning_rate": 4.723076923076923e-07,
      "loss": 0.0976,
      "reward": 1.2809004187583923,
      "reward_std": 0.28368543833494186,
      "reward_std_all": 0.30567291378974915,
      "rewards-max/forensics_sinkhorn_reward": 0.9730513989925385,
      "rewards-max/forensics_strict_edge_reward": 0.31347285211086273,
      "rewards-max/forensics_tiou_reward": 0.9584936797618866,
      "rewards/forensics_sinkhorn_reward": 0.7479846477508545,
      "rewards/forensics_strict_edge_reward": 0.01805230602622032,
      "rewards/forensics_tiou_reward": 0.5148634910583496,
      "step": 687
    },
    {
      "completion_length": 21.578125,
      "epoch": 2.110599078341014,
      "grad_norm": 6.753602027893066,
      "kl": 0.3232421875,
      "learning_rate": 4.715384615384615e-07,
      "loss": 0.085,
      "reward": 1.1770704984664917,
      "reward_std": 0.27078527212142944,
      "reward_std_all": 0.23721538484096527,
      "rewards-max/forensics_sinkhorn_reward": 0.9309150874614716,
      "rewards-max/forensics_strict_edge_reward": 0.11806592717766762,
      "rewards-max/forensics_tiou_reward": 0.8506103456020355,
      "rewards/forensics_sinkhorn_reward": 0.6965679824352264,
      "rewards/forensics_strict_edge_reward": 0.011605048086494207,
      "rewards/forensics_tiou_reward": 0.4688974916934967,
      "step": 688
    },
    {
      "completion_length": 20.484375,
      "epoch": 2.1136712749615976,
      "grad_norm": 9.494672775268555,
      "kl": 0.2578125,
      "learning_rate": 4.7076923076923073e-07,
      "loss": 0.026,
      "reward": 1.2594918608665466,
      "reward_std": 0.2331695482134819,
      "reward_std_all": 0.21244722604751587,
      "rewards-max/forensics_sinkhorn_reward": 0.9710036218166351,
      "rewards-max/forensics_strict_edge_reward": 0.23392322286963463,
      "rewards-max/forensics_tiou_reward": 0.9502088129520416,
      "rewards/forensics_sinkhorn_reward": 0.6914331912994385,
      "rewards/forensics_strict_edge_reward": 0.028258089907467365,
      "rewards/forensics_tiou_reward": 0.5398004949092865,
      "step": 689
    },
    {
      "completion_length": 20.90625,
      "epoch": 2.1167434715821813,
      "grad_norm": 9.491619110107422,
      "kl": 0.24658203125,
      "learning_rate": 4.6999999999999995e-07,
      "loss": -0.0114,
      "reward": 1.3341821432113647,
      "reward_std": 0.19685590267181396,
      "reward_std_all": 0.19933105260133743,
      "rewards-max/forensics_sinkhorn_reward": 0.9787154793739319,
      "rewards-max/forensics_strict_edge_reward": 0.6198739409446716,
      "rewards-max/forensics_tiou_reward": 0.9704194962978363,
      "rewards/forensics_sinkhorn_reward": 0.7527748942375183,
      "rewards/forensics_strict_edge_reward": 0.05824661534279585,
      "rewards/forensics_tiou_reward": 0.5231605768203735,
      "step": 690
    },
    {
      "completion_length": 20.25,
      "epoch": 2.119815668202765,
      "grad_norm": 11.69605827331543,
      "kl": 0.24658203125,
      "learning_rate": 4.692307692307692e-07,
      "loss": -0.0177,
      "reward": 1.5576800107955933,
      "reward_std": 0.23514331877231598,
      "reward_std_all": 0.2218088135123253,
      "rewards-max/forensics_sinkhorn_reward": 0.9795884191989899,
      "rewards-max/forensics_strict_edge_reward": 0.5500316619873047,
      "rewards-max/forensics_tiou_reward": 0.9743935763835907,
      "rewards/forensics_sinkhorn_reward": 0.8279009759426117,
      "rewards/forensics_strict_edge_reward": 0.08425088971853256,
      "rewards/forensics_tiou_reward": 0.6455282270908356,
      "step": 691
    },
    {
      "completion_length": 22.078125,
      "epoch": 2.122887864823349,
      "grad_norm": 8.529268264770508,
      "kl": 0.3251953125,
      "learning_rate": 4.684615384615385e-07,
      "loss": -0.0594,
      "reward": 1.2661495208740234,
      "reward_std": 0.2531014159321785,
      "reward_std_all": 0.2392696738243103,
      "rewards-max/forensics_sinkhorn_reward": 0.9762553572654724,
      "rewards-max/forensics_strict_edge_reward": 0.5321668684482574,
      "rewards-max/forensics_tiou_reward": 0.9715015292167664,
      "rewards/forensics_sinkhorn_reward": 0.7268747687339783,
      "rewards/forensics_strict_edge_reward": 0.03982880339026451,
      "rewards/forensics_tiou_reward": 0.49944598972797394,
      "step": 692
    },
    {
      "completion_length": 20.65625,
      "epoch": 2.1259600614439322,
      "grad_norm": 8.611410140991211,
      "kl": 0.2763671875,
      "learning_rate": 4.6769230769230764e-07,
      "loss": -0.0322,
      "reward": 1.3960794806480408,
      "reward_std": 0.19578669965267181,
      "reward_std_all": 0.18874210119247437,
      "rewards-max/forensics_sinkhorn_reward": 0.9553231298923492,
      "rewards-max/forensics_strict_edge_reward": 0.4330120086669922,
      "rewards-max/forensics_tiou_reward": 0.9198773801326752,
      "rewards/forensics_sinkhorn_reward": 0.7681742012500763,
      "rewards/forensics_strict_edge_reward": 0.056218283250927925,
      "rewards/forensics_tiou_reward": 0.5716871023178101,
      "step": 693
    },
    {
      "completion_length": 20.71875,
      "epoch": 2.129032258064516,
      "grad_norm": 10.336377143859863,
      "kl": 0.30126953125,
      "learning_rate": 4.669230769230769e-07,
      "loss": -0.1044,
      "reward": 1.2907297611236572,
      "reward_std": 0.2362707257270813,
      "reward_std_all": 0.21297304332256317,
      "rewards-max/forensics_sinkhorn_reward": 0.9608558118343353,
      "rewards-max/forensics_strict_edge_reward": 0.4506318047642708,
      "rewards-max/forensics_tiou_reward": 0.931404709815979,
      "rewards/forensics_sinkhorn_reward": 0.7246183454990387,
      "rewards/forensics_strict_edge_reward": 0.05914651555940509,
      "rewards/forensics_tiou_reward": 0.5069650113582611,
      "step": 694
    },
    {
      "completion_length": 21.0625,
      "epoch": 2.1321044546851,
      "grad_norm": 7.960299491882324,
      "kl": 0.20068359375,
      "learning_rate": 4.661538461538461e-07,
      "loss": -0.0632,
      "reward": 1.355800449848175,
      "reward_std": 0.22699221223592758,
      "reward_std_all": 0.2530505284667015,
      "rewards-max/forensics_sinkhorn_reward": 0.9676617085933685,
      "rewards-max/forensics_strict_edge_reward": 0.4982926547527313,
      "rewards-max/forensics_tiou_reward": 0.9483262598514557,
      "rewards/forensics_sinkhorn_reward": 0.7443553507328033,
      "rewards/forensics_strict_edge_reward": 0.07108859904110432,
      "rewards/forensics_tiou_reward": 0.5403565764427185,
      "step": 695
    },
    {
      "completion_length": 20.84375,
      "epoch": 2.1351766513056836,
      "grad_norm": 10.853008270263672,
      "kl": 0.3154296875,
      "learning_rate": 4.653846153846154e-07,
      "loss": -0.0601,
      "reward": 1.1515990495681763,
      "reward_std": 0.2097637951374054,
      "reward_std_all": 0.18725918978452682,
      "rewards-max/forensics_sinkhorn_reward": 0.9469800889492035,
      "rewards-max/forensics_strict_edge_reward": 0.21537704020738602,
      "rewards-max/forensics_tiou_reward": 0.8829550445079803,
      "rewards/forensics_sinkhorn_reward": 0.6866351366043091,
      "rewards/forensics_strict_edge_reward": 0.015173074323683977,
      "rewards/forensics_tiou_reward": 0.4497908800840378,
      "step": 696
    },
    {
      "completion_length": 22.390625,
      "epoch": 2.1382488479262673,
      "grad_norm": 11.55024242401123,
      "kl": 0.23681640625,
      "learning_rate": 4.646153846153846e-07,
      "loss": 0.1121,
      "reward": 1.4256747364997864,
      "reward_std": 0.23848296701908112,
      "reward_std_all": 0.2227957919239998,
      "rewards-max/forensics_sinkhorn_reward": 0.965951144695282,
      "rewards-max/forensics_strict_edge_reward": 0.4361710250377655,
      "rewards-max/forensics_tiou_reward": 0.9386634230613708,
      "rewards/forensics_sinkhorn_reward": 0.7771891355514526,
      "rewards/forensics_strict_edge_reward": 0.06382765993475914,
      "rewards/forensics_tiou_reward": 0.5846579074859619,
      "step": 697
    },
    {
      "completion_length": 19.5625,
      "epoch": 2.141321044546851,
      "grad_norm": 13.088051795959473,
      "kl": 0.2607421875,
      "learning_rate": 4.638461538461538e-07,
      "loss": -0.0959,
      "reward": 1.3577320575714111,
      "reward_std": 0.22823115438222885,
      "reward_std_all": 0.21338441967964172,
      "rewards-max/forensics_sinkhorn_reward": 0.9727123081684113,
      "rewards-max/forensics_strict_edge_reward": 0.4235072433948517,
      "rewards-max/forensics_tiou_reward": 0.9549775123596191,
      "rewards/forensics_sinkhorn_reward": 0.7442861497402191,
      "rewards/forensics_strict_edge_reward": 0.058692128863185644,
      "rewards/forensics_tiou_reward": 0.5547538101673126,
      "step": 698
    },
    {
      "completion_length": 22.78125,
      "epoch": 2.144393241167435,
      "grad_norm": 6.459594249725342,
      "kl": 0.3173828125,
      "learning_rate": 4.6307692307692303e-07,
      "loss": -0.0589,
      "reward": 1.4628621935844421,
      "reward_std": 0.2622726112604141,
      "reward_std_all": 0.22572068870067596,
      "rewards-max/forensics_sinkhorn_reward": 0.9677946865558624,
      "rewards-max/forensics_strict_edge_reward": 0.5321605801582336,
      "rewards-max/forensics_tiou_reward": 0.9377835392951965,
      "rewards/forensics_sinkhorn_reward": 0.7945825457572937,
      "rewards/forensics_strict_edge_reward": 0.049943823367357254,
      "rewards/forensics_tiou_reward": 0.6183357238769531,
      "step": 699
    },
    {
      "completion_length": 19.625,
      "epoch": 2.1474654377880182,
      "grad_norm": 11.437468528747559,
      "kl": 0.3525390625,
      "learning_rate": 4.623076923076923e-07,
      "loss": -0.0148,
      "reward": 1.3352082967758179,
      "reward_std": 0.2052048221230507,
      "reward_std_all": 0.1818777695298195,
      "rewards-max/forensics_sinkhorn_reward": 0.9691068232059479,
      "rewards-max/forensics_strict_edge_reward": 0.45430223643779755,
      "rewards-max/forensics_tiou_reward": 0.9488049149513245,
      "rewards/forensics_sinkhorn_reward": 0.7559658885002136,
      "rewards/forensics_strict_edge_reward": 0.0559240598231554,
      "rewards/forensics_tiou_reward": 0.523318350315094,
      "step": 700
    },
    {
      "completion_length": 20.671875,
      "epoch": 2.150537634408602,
      "grad_norm": 15.595083236694336,
      "kl": 0.33251953125,
      "learning_rate": 4.6153846153846156e-07,
      "loss": 0.0853,
      "reward": 1.3221425414085388,
      "reward_std": 0.2805793732404709,
      "reward_std_all": 0.2484591007232666,
      "rewards-max/forensics_sinkhorn_reward": 0.970489889383316,
      "rewards-max/forensics_strict_edge_reward": 0.4641416519880295,
      "rewards-max/forensics_tiou_reward": 0.9524203836917877,
      "rewards/forensics_sinkhorn_reward": 0.758839339017868,
      "rewards/forensics_strict_edge_reward": 0.04276365414261818,
      "rewards/forensics_tiou_reward": 0.5205394923686981,
      "step": 701
    },
    {
      "completion_length": 22.0,
      "epoch": 2.153609831029186,
      "grad_norm": 22.309036254882812,
      "kl": 0.23876953125,
      "learning_rate": 4.607692307692307e-07,
      "loss": 0.0453,
      "reward": 1.3153014183044434,
      "reward_std": 0.19658008962869644,
      "reward_std_all": 0.2136518731713295,
      "rewards-max/forensics_sinkhorn_reward": 0.9721085727214813,
      "rewards-max/forensics_strict_edge_reward": 0.26344073191285133,
      "rewards-max/forensics_tiou_reward": 0.9587971866130829,
      "rewards/forensics_sinkhorn_reward": 0.7388766407966614,
      "rewards/forensics_strict_edge_reward": 0.037350501865148544,
      "rewards/forensics_tiou_reward": 0.5390742719173431,
      "step": 702
    },
    {
      "completion_length": 21.6875,
      "epoch": 2.1566820276497696,
      "grad_norm": 11.48072624206543,
      "kl": 0.373046875,
      "learning_rate": 4.6e-07,
      "loss": -0.0984,
      "reward": 1.121176540851593,
      "reward_std": 0.18957220762968063,
      "reward_std_all": 0.1644003763794899,
      "rewards-max/forensics_sinkhorn_reward": 0.9642283022403717,
      "rewards-max/forensics_strict_edge_reward": 0.1289210058748722,
      "rewards-max/forensics_tiou_reward": 0.931624561548233,
      "rewards/forensics_sinkhorn_reward": 0.6611246168613434,
      "rewards/forensics_strict_edge_reward": 0.012058103922754526,
      "rewards/forensics_tiou_reward": 0.4479936957359314,
      "step": 703
    },
    {
      "completion_length": 21.84375,
      "epoch": 2.1597542242703534,
      "grad_norm": 8.483817100524902,
      "kl": 0.2861328125,
      "learning_rate": 4.592307692307692e-07,
      "loss": -0.1007,
      "reward": 1.4516103863716125,
      "reward_std": 0.1524648442864418,
      "reward_std_all": 0.150277741253376,
      "rewards-max/forensics_sinkhorn_reward": 0.9648514091968536,
      "rewards-max/forensics_strict_edge_reward": 0.3513035327196121,
      "rewards-max/forensics_tiou_reward": 0.9253180623054504,
      "rewards/forensics_sinkhorn_reward": 0.804649829864502,
      "rewards/forensics_strict_edge_reward": 0.03000751230865717,
      "rewards/forensics_tiou_reward": 0.6169531047344208,
      "step": 704
    },
    {
      "completion_length": 21.921875,
      "epoch": 2.162826420890937,
      "grad_norm": 12.06200885772705,
      "kl": 0.630859375,
      "learning_rate": 4.5846153846153847e-07,
      "loss": 0.0265,
      "reward": 1.4476107358932495,
      "reward_std": 0.25794145464897156,
      "reward_std_all": 0.26569584012031555,
      "rewards-max/forensics_sinkhorn_reward": 0.9734825491905212,
      "rewards-max/forensics_strict_edge_reward": 0.40798234194517136,
      "rewards-max/forensics_tiou_reward": 0.9611148238182068,
      "rewards/forensics_sinkhorn_reward": 0.7997897267341614,
      "rewards/forensics_strict_edge_reward": 0.046231191605329514,
      "rewards/forensics_tiou_reward": 0.6015898287296295,
      "step": 705
    },
    {
      "completion_length": 19.953125,
      "epoch": 2.165898617511521,
      "grad_norm": 12.782917022705078,
      "kl": 0.3388671875,
      "learning_rate": 4.5769230769230763e-07,
      "loss": -0.0685,
      "reward": 1.2891181111335754,
      "reward_std": 0.27443544566631317,
      "reward_std_all": 0.2403155267238617,
      "rewards-max/forensics_sinkhorn_reward": 0.9623187780380249,
      "rewards-max/forensics_strict_edge_reward": 0.34273045510053635,
      "rewards-max/forensics_tiou_reward": 0.9302962422370911,
      "rewards/forensics_sinkhorn_reward": 0.7432150840759277,
      "rewards/forensics_strict_edge_reward": 0.04105719830840826,
      "rewards/forensics_tiou_reward": 0.5048459023237228,
      "step": 706
    },
    {
      "completion_length": 20.734375,
      "epoch": 2.1689708141321047,
      "grad_norm": 8.654350280761719,
      "kl": 0.2255859375,
      "learning_rate": 4.569230769230769e-07,
      "loss": -0.0696,
      "reward": 1.515337586402893,
      "reward_std": 0.23219755291938782,
      "reward_std_all": 0.21709486097097397,
      "rewards-max/forensics_sinkhorn_reward": 0.9615204036235809,
      "rewards-max/forensics_strict_edge_reward": 0.49555233120918274,
      "rewards-max/forensics_tiou_reward": 0.9359354376792908,
      "rewards/forensics_sinkhorn_reward": 0.8046700358390808,
      "rewards/forensics_strict_edge_reward": 0.07380995526909828,
      "rewards/forensics_tiou_reward": 0.636857658624649,
      "step": 707
    },
    {
      "completion_length": 22.0,
      "epoch": 2.172043010752688,
      "grad_norm": 11.219345092773438,
      "kl": 0.3779296875,
      "learning_rate": 4.561538461538461e-07,
      "loss": 0.0578,
      "reward": 1.3381458520889282,
      "reward_std": 0.21601669490337372,
      "reward_std_all": 0.19623523205518723,
      "rewards-max/forensics_sinkhorn_reward": 0.9800419807434082,
      "rewards-max/forensics_strict_edge_reward": 0.47400766611099243,
      "rewards-max/forensics_tiou_reward": 0.9753933846950531,
      "rewards/forensics_sinkhorn_reward": 0.7343713641166687,
      "rewards/forensics_strict_edge_reward": 0.049149940721690655,
      "rewards/forensics_tiou_reward": 0.554624542593956,
      "step": 708
    },
    {
      "completion_length": 21.453125,
      "epoch": 2.175115207373272,
      "grad_norm": 17.42472267150879,
      "kl": 0.26806640625,
      "learning_rate": 4.553846153846154e-07,
      "loss": -0.0472,
      "reward": 1.263942301273346,
      "reward_std": 0.24560096114873886,
      "reward_std_all": 0.26415004581213,
      "rewards-max/forensics_sinkhorn_reward": 0.9772967398166656,
      "rewards-max/forensics_strict_edge_reward": 0.44648270308971405,
      "rewards-max/forensics_tiou_reward": 0.9648316502571106,
      "rewards/forensics_sinkhorn_reward": 0.7071040570735931,
      "rewards/forensics_strict_edge_reward": 0.047247666865587234,
      "rewards/forensics_tiou_reward": 0.5095905661582947,
      "step": 709
    },
    {
      "completion_length": 21.21875,
      "epoch": 2.1781874039938556,
      "grad_norm": 9.648982048034668,
      "kl": 0.341796875,
      "learning_rate": 4.546153846153846e-07,
      "loss": -0.0198,
      "reward": 1.361824870109558,
      "reward_std": 0.34737439453601837,
      "reward_std_all": 0.3347952961921692,
      "rewards-max/forensics_sinkhorn_reward": 0.982454240322113,
      "rewards-max/forensics_strict_edge_reward": 0.7837068438529968,
      "rewards-max/forensics_tiou_reward": 0.9820577502250671,
      "rewards/forensics_sinkhorn_reward": 0.7198574244976044,
      "rewards/forensics_strict_edge_reward": 0.0788383036851883,
      "rewards/forensics_tiou_reward": 0.5631291568279266,
      "step": 710
    },
    {
      "completion_length": 22.15625,
      "epoch": 2.1812596006144394,
      "grad_norm": 11.646665573120117,
      "kl": 0.3740234375,
      "learning_rate": 4.538461538461538e-07,
      "loss": -0.0155,
      "reward": 1.422175645828247,
      "reward_std": 0.33276990056037903,
      "reward_std_all": 0.3337786793708801,
      "rewards-max/forensics_sinkhorn_reward": 0.9769936203956604,
      "rewards-max/forensics_strict_edge_reward": 0.5404154062271118,
      "rewards-max/forensics_tiou_reward": 0.971167653799057,
      "rewards/forensics_sinkhorn_reward": 0.7540155947208405,
      "rewards/forensics_strict_edge_reward": 0.0713046919554472,
      "rewards/forensics_tiou_reward": 0.5968553721904755,
      "step": 711
    },
    {
      "completion_length": 21.875,
      "epoch": 2.184331797235023,
      "grad_norm": 10.00827693939209,
      "kl": 0.4140625,
      "learning_rate": 4.5307692307692307e-07,
      "loss": 0.0089,
      "reward": 1.2361518144607544,
      "reward_std": 0.24148204922676086,
      "reward_std_all": 0.21234389394521713,
      "rewards-max/forensics_sinkhorn_reward": 0.9670754373073578,
      "rewards-max/forensics_strict_edge_reward": 0.3975932002067566,
      "rewards-max/forensics_tiou_reward": 0.9467686116695404,
      "rewards/forensics_sinkhorn_reward": 0.7119346261024475,
      "rewards/forensics_strict_edge_reward": 0.03520648181438446,
      "rewards/forensics_tiou_reward": 0.4890108108520508,
      "step": 712
    },
    {
      "completion_length": 19.234375,
      "epoch": 2.187403993855607,
      "grad_norm": 13.000314712524414,
      "kl": 0.46484375,
      "learning_rate": 4.523076923076923e-07,
      "loss": -0.0073,
      "reward": 1.4392178654670715,
      "reward_std": 0.2671937942504883,
      "reward_std_all": 0.24569661915302277,
      "rewards-max/forensics_sinkhorn_reward": 0.9779225587844849,
      "rewards-max/forensics_strict_edge_reward": 0.7403165996074677,
      "rewards-max/forensics_tiou_reward": 0.9734298288822174,
      "rewards/forensics_sinkhorn_reward": 0.7965807318687439,
      "rewards/forensics_strict_edge_reward": 0.07510275393724442,
      "rewards/forensics_tiou_reward": 0.5675344467163086,
      "step": 713
    },
    {
      "completion_length": 21.953125,
      "epoch": 2.1904761904761907,
      "grad_norm": 10.910383224487305,
      "kl": 0.3779296875,
      "learning_rate": 4.5153846153846155e-07,
      "loss": 0.1185,
      "reward": 1.4843064546585083,
      "reward_std": 0.27418413758277893,
      "reward_std_all": 0.25715016573667526,
      "rewards-max/forensics_sinkhorn_reward": 0.9645135402679443,
      "rewards-max/forensics_strict_edge_reward": 0.7010898292064667,
      "rewards-max/forensics_tiou_reward": 0.9434104263782501,
      "rewards/forensics_sinkhorn_reward": 0.8017411231994629,
      "rewards/forensics_strict_edge_reward": 0.07239437103271484,
      "rewards/forensics_tiou_reward": 0.6101710796356201,
      "step": 714
    },
    {
      "completion_length": 21.953125,
      "epoch": 2.193548387096774,
      "grad_norm": 11.191259384155273,
      "kl": 0.701171875,
      "learning_rate": 4.507692307692307e-07,
      "loss": 0.2101,
      "reward": 1.2161570191383362,
      "reward_std": 0.2733670249581337,
      "reward_std_all": 0.2565016895532608,
      "rewards-max/forensics_sinkhorn_reward": 0.9237059950828552,
      "rewards-max/forensics_strict_edge_reward": 0.2670212350785732,
      "rewards-max/forensics_tiou_reward": 0.8511207699775696,
      "rewards/forensics_sinkhorn_reward": 0.6932043731212616,
      "rewards/forensics_strict_edge_reward": 0.03662133775651455,
      "rewards/forensics_tiou_reward": 0.4863312840461731,
      "step": 715
    },
    {
      "completion_length": 21.6875,
      "epoch": 2.196620583717358,
      "grad_norm": 8.36439037322998,
      "kl": 0.271484375,
      "learning_rate": 4.5e-07,
      "loss": -0.0668,
      "reward": 1.3713954091072083,
      "reward_std": 0.16890062391757965,
      "reward_std_all": 0.16727619618177414,
      "rewards-max/forensics_sinkhorn_reward": 0.9651155471801758,
      "rewards-max/forensics_strict_edge_reward": 0.22279208153486252,
      "rewards-max/forensics_tiou_reward": 0.9375087916851044,
      "rewards/forensics_sinkhorn_reward": 0.7647013068199158,
      "rewards/forensics_strict_edge_reward": 0.02797537576407194,
      "rewards/forensics_tiou_reward": 0.57871875166893,
      "step": 716
    },
    {
      "completion_length": 21.609375,
      "epoch": 2.1996927803379416,
      "grad_norm": 10.300077438354492,
      "kl": 0.25146484375,
      "learning_rate": 4.492307692307692e-07,
      "loss": -0.0518,
      "reward": 1.475437879562378,
      "reward_std": 0.1832326203584671,
      "reward_std_all": 0.16718190163373947,
      "rewards-max/forensics_sinkhorn_reward": 0.9820025563240051,
      "rewards-max/forensics_strict_edge_reward": 0.6377167105674744,
      "rewards-max/forensics_tiou_reward": 0.9823332130908966,
      "rewards/forensics_sinkhorn_reward": 0.7906276285648346,
      "rewards/forensics_strict_edge_reward": 0.0801125168800354,
      "rewards/forensics_tiou_reward": 0.6046976745128632,
      "step": 717
    },
    {
      "completion_length": 22.125,
      "epoch": 2.2027649769585254,
      "grad_norm": 9.953198432922363,
      "kl": 0.3271484375,
      "learning_rate": 4.4846153846153846e-07,
      "loss": 0.0946,
      "reward": 1.1605981588363647,
      "reward_std": 0.3086789846420288,
      "reward_std_all": 0.2838786691427231,
      "rewards-max/forensics_sinkhorn_reward": 0.9638788104057312,
      "rewards-max/forensics_strict_edge_reward": 0.45961035788059235,
      "rewards-max/forensics_tiou_reward": 0.9381281435489655,
      "rewards/forensics_sinkhorn_reward": 0.6637387871742249,
      "rewards/forensics_strict_edge_reward": 0.04971974343061447,
      "rewards/forensics_tiou_reward": 0.44713956117630005,
      "step": 718
    },
    {
      "completion_length": 20.515625,
      "epoch": 2.205837173579109,
      "grad_norm": 10.3635835647583,
      "kl": 0.388671875,
      "learning_rate": 4.476923076923076e-07,
      "loss": 0.0005,
      "reward": 1.5365540385246277,
      "reward_std": 0.27817563712596893,
      "reward_std_all": 0.2720798999071121,
      "rewards-max/forensics_sinkhorn_reward": 0.9412882328033447,
      "rewards-max/forensics_strict_edge_reward": 0.7037424445152283,
      "rewards-max/forensics_tiou_reward": 0.870152473449707,
      "rewards/forensics_sinkhorn_reward": 0.81577929854393,
      "rewards/forensics_strict_edge_reward": 0.08805413916707039,
      "rewards/forensics_tiou_reward": 0.6327205896377563,
      "step": 719
    },
    {
      "completion_length": 22.40625,
      "epoch": 2.208909370199693,
      "grad_norm": 9.74746322631836,
      "kl": 0.3876953125,
      "learning_rate": 4.469230769230769e-07,
      "loss": -0.0052,
      "reward": 1.5304589867591858,
      "reward_std": 0.25829870998859406,
      "reward_std_all": 0.24662068486213684,
      "rewards-max/forensics_sinkhorn_reward": 0.979648619890213,
      "rewards-max/forensics_strict_edge_reward": 0.48898807168006897,
      "rewards-max/forensics_tiou_reward": 0.9679623544216156,
      "rewards/forensics_sinkhorn_reward": 0.8078097701072693,
      "rewards/forensics_strict_edge_reward": 0.0623945165425539,
      "rewards/forensics_tiou_reward": 0.660254716873169,
      "step": 720
    },
    {
      "completion_length": 22.59375,
      "epoch": 2.2119815668202767,
      "grad_norm": 10.007206916809082,
      "kl": 0.24072265625,
      "learning_rate": 4.4615384615384615e-07,
      "loss": -0.007,
      "reward": 1.626905918121338,
      "reward_std": 0.26432283222675323,
      "reward_std_all": 0.2370743229985237,
      "rewards-max/forensics_sinkhorn_reward": 0.9847652018070221,
      "rewards-max/forensics_strict_edge_reward": 0.7958523333072662,
      "rewards-max/forensics_tiou_reward": 0.9877971112728119,
      "rewards/forensics_sinkhorn_reward": 0.838290274143219,
      "rewards/forensics_strict_edge_reward": 0.08665209263563156,
      "rewards/forensics_tiou_reward": 0.7019636929035187,
      "step": 721
    },
    {
      "completion_length": 23.40625,
      "epoch": 2.21505376344086,
      "grad_norm": 10.084322929382324,
      "kl": 0.25830078125,
      "learning_rate": 4.4538461538461537e-07,
      "loss": -0.0229,
      "reward": 1.5619785785675049,
      "reward_std": 0.3010032922029495,
      "reward_std_all": 0.2889935225248337,
      "rewards-max/forensics_sinkhorn_reward": 0.9799292385578156,
      "rewards-max/forensics_strict_edge_reward": 0.575991302728653,
      "rewards-max/forensics_tiou_reward": 0.9763544797897339,
      "rewards/forensics_sinkhorn_reward": 0.7830242216587067,
      "rewards/forensics_strict_edge_reward": 0.11150642484426498,
      "rewards/forensics_tiou_reward": 0.6674479842185974,
      "step": 722
    },
    {
      "completion_length": 22.359375,
      "epoch": 2.218125960061444,
      "grad_norm": 11.262130737304688,
      "kl": 0.298828125,
      "learning_rate": 4.4461538461538463e-07,
      "loss": -0.0109,
      "reward": 1.3179817199707031,
      "reward_std": 0.22304177284240723,
      "reward_std_all": 0.22268333286046982,
      "rewards-max/forensics_sinkhorn_reward": 0.9555537700653076,
      "rewards-max/forensics_strict_edge_reward": 0.3887161910533905,
      "rewards-max/forensics_tiou_reward": 0.9149158596992493,
      "rewards/forensics_sinkhorn_reward": 0.7657421827316284,
      "rewards/forensics_strict_edge_reward": 0.030557976104319096,
      "rewards/forensics_tiou_reward": 0.5216816663742065,
      "step": 723
    },
    {
      "completion_length": 21.90625,
      "epoch": 2.2211981566820276,
      "grad_norm": 16.34966278076172,
      "kl": 0.24560546875,
      "learning_rate": 4.438461538461538e-07,
      "loss": -0.0296,
      "reward": 1.3728585243225098,
      "reward_std": 0.1921650692820549,
      "reward_std_all": 0.1885600984096527,
      "rewards-max/forensics_sinkhorn_reward": 0.9319854974746704,
      "rewards-max/forensics_strict_edge_reward": 0.2223440185189247,
      "rewards-max/forensics_tiou_reward": 0.8634262382984161,
      "rewards/forensics_sinkhorn_reward": 0.7747371792793274,
      "rewards/forensics_strict_edge_reward": 0.025263018906116486,
      "rewards/forensics_tiou_reward": 0.5728584229946136,
      "step": 724
    },
    {
      "completion_length": 24.453125,
      "epoch": 2.2242703533026114,
      "grad_norm": 13.500412940979004,
      "kl": 0.25390625,
      "learning_rate": 4.4307692307692306e-07,
      "loss": -0.0898,
      "reward": 1.3813928365707397,
      "reward_std": 0.21359767764806747,
      "reward_std_all": 0.21477162837982178,
      "rewards-max/forensics_sinkhorn_reward": 0.9791212975978851,
      "rewards-max/forensics_strict_edge_reward": 0.3445742726325989,
      "rewards-max/forensics_tiou_reward": 0.9743356704711914,
      "rewards/forensics_sinkhorn_reward": 0.7687372863292694,
      "rewards/forensics_strict_edge_reward": 0.029463255777955055,
      "rewards/forensics_tiou_reward": 0.5831924080848694,
      "step": 725
    },
    {
      "completion_length": 21.90625,
      "epoch": 2.227342549923195,
      "grad_norm": 11.822771072387695,
      "kl": 0.478515625,
      "learning_rate": 4.423076923076923e-07,
      "loss": 0.0204,
      "reward": 1.5224640369415283,
      "reward_std": 0.20150822401046753,
      "reward_std_all": 0.19714518636465073,
      "rewards-max/forensics_sinkhorn_reward": 0.9493581354618073,
      "rewards-max/forensics_strict_edge_reward": 0.30745653808116913,
      "rewards-max/forensics_tiou_reward": 0.8960050046443939,
      "rewards/forensics_sinkhorn_reward": 0.8198699951171875,
      "rewards/forensics_strict_edge_reward": 0.04026193078607321,
      "rewards/forensics_tiou_reward": 0.662332147359848,
      "step": 726
    },
    {
      "completion_length": 23.1875,
      "epoch": 2.230414746543779,
      "grad_norm": 8.564302444458008,
      "kl": 0.26318359375,
      "learning_rate": 4.4153846153846154e-07,
      "loss": 0.0239,
      "reward": 1.3990784883499146,
      "reward_std": 0.26309624314308167,
      "reward_std_all": 0.23047670722007751,
      "rewards-max/forensics_sinkhorn_reward": 0.9739688336849213,
      "rewards-max/forensics_strict_edge_reward": 0.3966260254383087,
      "rewards-max/forensics_tiou_reward": 0.9634871184825897,
      "rewards/forensics_sinkhorn_reward": 0.7786312401294708,
      "rewards/forensics_strict_edge_reward": 0.03882203157991171,
      "rewards/forensics_tiou_reward": 0.5816251933574677,
      "step": 727
    },
    {
      "completion_length": 22.078125,
      "epoch": 2.2334869431643627,
      "grad_norm": 9.660177230834961,
      "kl": 0.36328125,
      "learning_rate": 4.407692307692307e-07,
      "loss": 0.0581,
      "reward": 1.1119601726531982,
      "reward_std": 0.24001862108707428,
      "reward_std_all": 0.21247507631778717,
      "rewards-max/forensics_sinkhorn_reward": 0.9672731459140778,
      "rewards-max/forensics_strict_edge_reward": 0.1318835316778859,
      "rewards-max/forensics_tiou_reward": 0.933944433927536,
      "rewards/forensics_sinkhorn_reward": 0.6778896152973175,
      "rewards/forensics_strict_edge_reward": 0.01311583441838593,
      "rewards/forensics_tiou_reward": 0.42095471918582916,
      "step": 728
    },
    {
      "completion_length": 22.515625,
      "epoch": 2.236559139784946,
      "grad_norm": 10.161126136779785,
      "kl": 0.34912109375,
      "learning_rate": 4.3999999999999997e-07,
      "loss": -0.2306,
      "reward": 1.1370119452476501,
      "reward_std": 0.28330712020397186,
      "reward_std_all": 0.30774733424186707,
      "rewards-max/forensics_sinkhorn_reward": 0.92772576212883,
      "rewards-max/forensics_strict_edge_reward": 0.4476890116930008,
      "rewards-max/forensics_tiou_reward": 0.8596951961517334,
      "rewards/forensics_sinkhorn_reward": 0.6306265592575073,
      "rewards/forensics_strict_edge_reward": 0.0592255387455225,
      "rewards/forensics_tiou_reward": 0.4471598118543625,
      "step": 729
    },
    {
      "completion_length": 24.046875,
      "epoch": 2.23963133640553,
      "grad_norm": 10.855547904968262,
      "kl": 0.2685546875,
      "learning_rate": 4.3923076923076924e-07,
      "loss": 0.0077,
      "reward": 1.324657380580902,
      "reward_std": 0.16356263309717178,
      "reward_std_all": 0.15955912321805954,
      "rewards-max/forensics_sinkhorn_reward": 0.925651878118515,
      "rewards-max/forensics_strict_edge_reward": 0.2865639915689826,
      "rewards-max/forensics_tiou_reward": 0.8422591388225555,
      "rewards/forensics_sinkhorn_reward": 0.7538972795009613,
      "rewards/forensics_strict_edge_reward": 0.028716945205815136,
      "rewards/forensics_tiou_reward": 0.5420430898666382,
      "step": 730
    },
    {
      "completion_length": 20.125,
      "epoch": 2.2427035330261136,
      "grad_norm": 14.005341529846191,
      "kl": 0.205078125,
      "learning_rate": 4.3846153846153845e-07,
      "loss": -0.0288,
      "reward": 1.2268341779708862,
      "reward_std": 0.21204491704702377,
      "reward_std_all": 0.22542542219161987,
      "rewards-max/forensics_sinkhorn_reward": 0.9461596310138702,
      "rewards-max/forensics_strict_edge_reward": 0.19741079211235046,
      "rewards-max/forensics_tiou_reward": 0.8957471251487732,
      "rewards/forensics_sinkhorn_reward": 0.7163525223731995,
      "rewards/forensics_strict_edge_reward": 0.022036297246813774,
      "rewards/forensics_tiou_reward": 0.4884454160928726,
      "step": 731
    },
    {
      "completion_length": 21.90625,
      "epoch": 2.2457757296466974,
      "grad_norm": 11.6885347366333,
      "kl": 0.4091796875,
      "learning_rate": 4.376923076923077e-07,
      "loss": 0.1365,
      "reward": 1.2558321356773376,
      "reward_std": 0.20920977741479874,
      "reward_std_all": 0.2018069326877594,
      "rewards-max/forensics_sinkhorn_reward": 0.9483367800712585,
      "rewards-max/forensics_strict_edge_reward": 0.19785186648368835,
      "rewards-max/forensics_tiou_reward": 0.9032813906669617,
      "rewards/forensics_sinkhorn_reward": 0.740571916103363,
      "rewards/forensics_strict_edge_reward": 0.017348979599773884,
      "rewards/forensics_tiou_reward": 0.49791133403778076,
      "step": 732
    },
    {
      "completion_length": 22.46875,
      "epoch": 2.248847926267281,
      "grad_norm": 14.92142105102539,
      "kl": 0.283203125,
      "learning_rate": 4.369230769230769e-07,
      "loss": 0.0221,
      "reward": 1.324789047241211,
      "reward_std": 0.20472301542758942,
      "reward_std_all": 0.1885865554213524,
      "rewards-max/forensics_sinkhorn_reward": 0.9796895384788513,
      "rewards-max/forensics_strict_edge_reward": 0.6065583229064941,
      "rewards-max/forensics_tiou_reward": 0.9743199646472931,
      "rewards/forensics_sinkhorn_reward": 0.7478969395160675,
      "rewards/forensics_strict_edge_reward": 0.04333548992872238,
      "rewards/forensics_tiou_reward": 0.5335565656423569,
      "step": 733
    },
    {
      "completion_length": 21.828125,
      "epoch": 2.251920122887865,
      "grad_norm": 11.624807357788086,
      "kl": 0.3740234375,
      "learning_rate": 4.3615384615384614e-07,
      "loss": 0.1533,
      "reward": 1.3018800020217896,
      "reward_std": 0.31861570477485657,
      "reward_std_all": 0.30134670436382294,
      "rewards-max/forensics_sinkhorn_reward": 0.9690133333206177,
      "rewards-max/forensics_strict_edge_reward": 0.4208007901906967,
      "rewards-max/forensics_tiou_reward": 0.9424042701721191,
      "rewards/forensics_sinkhorn_reward": 0.7115757465362549,
      "rewards/forensics_strict_edge_reward": 0.05826290976256132,
      "rewards/forensics_tiou_reward": 0.5320413410663605,
      "step": 734
    },
    {
      "completion_length": 22.296875,
      "epoch": 2.2549923195084487,
      "grad_norm": 8.052652359008789,
      "kl": 0.2646484375,
      "learning_rate": 4.3538461538461536e-07,
      "loss": 0.0255,
      "reward": 1.3431405425071716,
      "reward_std": 0.148933507502079,
      "reward_std_all": 0.13603219389915466,
      "rewards-max/forensics_sinkhorn_reward": 0.9715666770935059,
      "rewards-max/forensics_strict_edge_reward": 0.29351864755153656,
      "rewards-max/forensics_tiou_reward": 0.9565431773662567,
      "rewards/forensics_sinkhorn_reward": 0.7517254948616028,
      "rewards/forensics_strict_edge_reward": 0.02468063496053219,
      "rewards/forensics_tiou_reward": 0.5667343139648438,
      "step": 735
    },
    {
      "completion_length": 26.71875,
      "epoch": 2.258064516129032,
      "grad_norm": 7.662326335906982,
      "kl": 0.26611328125,
      "learning_rate": 4.346153846153846e-07,
      "loss": -0.1458,
      "reward": 1.495607614517212,
      "reward_std": 0.18138045072555542,
      "reward_std_all": 0.1672164425253868,
      "rewards-max/forensics_sinkhorn_reward": 0.9675647616386414,
      "rewards-max/forensics_strict_edge_reward": 0.41797660291194916,
      "rewards-max/forensics_tiou_reward": 0.9432624280452728,
      "rewards/forensics_sinkhorn_reward": 0.7865244746208191,
      "rewards/forensics_strict_edge_reward": 0.06510131433606148,
      "rewards/forensics_tiou_reward": 0.6439820230007172,
      "step": 736
    },
    {
      "completion_length": 24.765625,
      "epoch": 2.261136712749616,
      "grad_norm": 8.644054412841797,
      "kl": 0.458984375,
      "learning_rate": 4.338461538461538e-07,
      "loss": -0.0131,
      "reward": 1.4115614295005798,
      "reward_std": 0.3260042890906334,
      "reward_std_all": 0.2897116467356682,
      "rewards-max/forensics_sinkhorn_reward": 0.9702663421630859,
      "rewards-max/forensics_strict_edge_reward": 0.36730754375457764,
      "rewards-max/forensics_tiou_reward": 0.9449813365936279,
      "rewards/forensics_sinkhorn_reward": 0.7492784857749939,
      "rewards/forensics_strict_edge_reward": 0.05759519152343273,
      "rewards/forensics_tiou_reward": 0.6046876907348633,
      "step": 737
    },
    {
      "completion_length": 20.484375,
      "epoch": 2.2642089093701996,
      "grad_norm": 10.604193687438965,
      "kl": 0.35791015625,
      "learning_rate": 4.3307692307692305e-07,
      "loss": -0.068,
      "reward": 1.336747944355011,
      "reward_std": 0.2944851964712143,
      "reward_std_all": 0.27974945306777954,
      "rewards-max/forensics_sinkhorn_reward": 0.9720652401447296,
      "rewards-max/forensics_strict_edge_reward": 0.4991503059864044,
      "rewards-max/forensics_tiou_reward": 0.9530445635318756,
      "rewards/forensics_sinkhorn_reward": 0.7534739971160889,
      "rewards/forensics_strict_edge_reward": 0.03857223689556122,
      "rewards/forensics_tiou_reward": 0.5447016954421997,
      "step": 738
    },
    {
      "completion_length": 21.078125,
      "epoch": 2.2672811059907834,
      "grad_norm": 11.704586029052734,
      "kl": 0.37548828125,
      "learning_rate": 4.3230769230769227e-07,
      "loss": 0.0277,
      "reward": 1.275401771068573,
      "reward_std": 0.3575722873210907,
      "reward_std_all": 0.3477369695901871,
      "rewards-max/forensics_sinkhorn_reward": 0.9870880544185638,
      "rewards-max/forensics_strict_edge_reward": 0.7009071409702301,
      "rewards-max/forensics_tiou_reward": 0.9942867755889893,
      "rewards/forensics_sinkhorn_reward": 0.6946324110031128,
      "rewards/forensics_strict_edge_reward": 0.09340810403227806,
      "rewards/forensics_tiou_reward": 0.48736128211021423,
      "step": 739
    },
    {
      "completion_length": 22.25,
      "epoch": 2.270353302611367,
      "grad_norm": 7.553930759429932,
      "kl": 0.3115234375,
      "learning_rate": 4.3153846153846153e-07,
      "loss": 0.1238,
      "reward": 1.1210686564445496,
      "reward_std": 0.22238174080848694,
      "reward_std_all": 0.19638194888830185,
      "rewards-max/forensics_sinkhorn_reward": 0.9396398365497589,
      "rewards-max/forensics_strict_edge_reward": 0.1441498138010502,
      "rewards-max/forensics_tiou_reward": 0.8812834024429321,
      "rewards/forensics_sinkhorn_reward": 0.6923031806945801,
      "rewards/forensics_strict_edge_reward": 0.011136237648315728,
      "rewards/forensics_tiou_reward": 0.417629212141037,
      "step": 740
    },
    {
      "completion_length": 21.625,
      "epoch": 2.273425499231951,
      "grad_norm": 12.309929847717285,
      "kl": 0.259765625,
      "learning_rate": 4.307692307692308e-07,
      "loss": -0.0131,
      "reward": 1.396692156791687,
      "reward_std": 0.31103575229644775,
      "reward_std_all": 0.2729092389345169,
      "rewards-max/forensics_sinkhorn_reward": 0.9592139422893524,
      "rewards-max/forensics_strict_edge_reward": 0.40362464636564255,
      "rewards-max/forensics_tiou_reward": 0.9180813431739807,
      "rewards/forensics_sinkhorn_reward": 0.7735587656497955,
      "rewards/forensics_strict_edge_reward": 0.05210092244669795,
      "rewards/forensics_tiou_reward": 0.5710324943065643,
      "step": 741
    },
    {
      "completion_length": 20.46875,
      "epoch": 2.2764976958525347,
      "grad_norm": 13.317874908447266,
      "kl": 0.24365234375,
      "learning_rate": 4.2999999999999996e-07,
      "loss": 0.0205,
      "reward": 1.3524386286735535,
      "reward_std": 0.2757514417171478,
      "reward_std_all": 0.25540369749069214,
      "rewards-max/forensics_sinkhorn_reward": 0.9700523614883423,
      "rewards-max/forensics_strict_edge_reward": 0.29772263020277023,
      "rewards-max/forensics_tiou_reward": 0.9532538056373596,
      "rewards/forensics_sinkhorn_reward": 0.7709943652153015,
      "rewards/forensics_strict_edge_reward": 0.0261903814971447,
      "rewards/forensics_tiou_reward": 0.5552537441253662,
      "step": 742
    },
    {
      "completion_length": 24.1875,
      "epoch": 2.279569892473118,
      "grad_norm": 13.875226974487305,
      "kl": 0.21630859375,
      "learning_rate": 4.292307692307692e-07,
      "loss": 0.0983,
      "reward": 1.349799633026123,
      "reward_std": 0.2210988849401474,
      "reward_std_all": 0.22216610610485077,
      "rewards-max/forensics_sinkhorn_reward": 0.9833870828151703,
      "rewards-max/forensics_strict_edge_reward": 0.8062050342559814,
      "rewards-max/forensics_tiou_reward": 0.9860038459300995,
      "rewards/forensics_sinkhorn_reward": 0.7210002541542053,
      "rewards/forensics_strict_edge_reward": 0.06421911716461182,
      "rewards/forensics_tiou_reward": 0.5645802170038223,
      "step": 743
    },
    {
      "completion_length": 22.75,
      "epoch": 2.282642089093702,
      "grad_norm": 10.489765167236328,
      "kl": 0.4521484375,
      "learning_rate": 4.2846153846153844e-07,
      "loss": -0.0076,
      "reward": 1.2203658521175385,
      "reward_std": 0.24217602610588074,
      "reward_std_all": 0.2417335882782936,
      "rewards-max/forensics_sinkhorn_reward": 0.9591322839260101,
      "rewards-max/forensics_strict_edge_reward": 0.3243762031197548,
      "rewards-max/forensics_tiou_reward": 0.9245466589927673,
      "rewards/forensics_sinkhorn_reward": 0.6895740926265717,
      "rewards/forensics_strict_edge_reward": 0.06318969652056694,
      "rewards/forensics_tiou_reward": 0.467602014541626,
      "step": 744
    },
    {
      "completion_length": 22.140625,
      "epoch": 2.2857142857142856,
      "grad_norm": 12.495147705078125,
      "kl": 0.3232421875,
      "learning_rate": 4.276923076923077e-07,
      "loss": 0.0078,
      "reward": 1.3295302391052246,
      "reward_std": 0.23708367347717285,
      "reward_std_all": 0.2349260449409485,
      "rewards-max/forensics_sinkhorn_reward": 0.9692468643188477,
      "rewards-max/forensics_strict_edge_reward": 0.39925330877304077,
      "rewards-max/forensics_tiou_reward": 0.9547601938247681,
      "rewards/forensics_sinkhorn_reward": 0.737119048833847,
      "rewards/forensics_strict_edge_reward": 0.040471216663718224,
      "rewards/forensics_tiou_reward": 0.5519398748874664,
      "step": 745
    },
    {
      "completion_length": 23.90625,
      "epoch": 2.2887864823348694,
      "grad_norm": 10.845864295959473,
      "kl": 0.4013671875,
      "learning_rate": 4.2692307692307687e-07,
      "loss": 0.0169,
      "reward": 1.453328013420105,
      "reward_std": 0.284435398876667,
      "reward_std_all": 0.27358561009168625,
      "rewards-max/forensics_sinkhorn_reward": 0.9671390056610107,
      "rewards-max/forensics_strict_edge_reward": 0.3715486377477646,
      "rewards-max/forensics_tiou_reward": 0.9396080672740936,
      "rewards/forensics_sinkhorn_reward": 0.7787862122058868,
      "rewards/forensics_strict_edge_reward": 0.029561336152255535,
      "rewards/forensics_tiou_reward": 0.6449804902076721,
      "step": 746
    },
    {
      "completion_length": 20.265625,
      "epoch": 2.291858678955453,
      "grad_norm": 9.884746551513672,
      "kl": 0.4267578125,
      "learning_rate": 4.2615384615384613e-07,
      "loss": 0.0398,
      "reward": 1.214787483215332,
      "reward_std": 0.29527566581964493,
      "reward_std_all": 0.28654882311820984,
      "rewards-max/forensics_sinkhorn_reward": 0.946279376745224,
      "rewards-max/forensics_strict_edge_reward": 0.2516818791627884,
      "rewards-max/forensics_tiou_reward": 0.8862074315547943,
      "rewards/forensics_sinkhorn_reward": 0.7038319110870361,
      "rewards/forensics_strict_edge_reward": 0.022423611022531986,
      "rewards/forensics_tiou_reward": 0.4885319769382477,
      "step": 747
    },
    {
      "completion_length": 21.28125,
      "epoch": 2.294930875576037,
      "grad_norm": 13.6007080078125,
      "kl": 0.27392578125,
      "learning_rate": 4.2538461538461535e-07,
      "loss": -0.0053,
      "reward": 1.4038035869598389,
      "reward_std": 0.1984306499361992,
      "reward_std_all": 0.1916390210390091,
      "rewards-max/forensics_sinkhorn_reward": 0.963518351316452,
      "rewards-max/forensics_strict_edge_reward": 0.4606466591358185,
      "rewards-max/forensics_tiou_reward": 0.9327622056007385,
      "rewards/forensics_sinkhorn_reward": 0.7587123811244965,
      "rewards/forensics_strict_edge_reward": 0.062148647382855415,
      "rewards/forensics_tiou_reward": 0.5829426050186157,
      "step": 748
    },
    {
      "completion_length": 21.578125,
      "epoch": 2.2980030721966207,
      "grad_norm": 10.195428848266602,
      "kl": 0.3369140625,
      "learning_rate": 4.246153846153846e-07,
      "loss": 0.0482,
      "reward": 1.3742907047271729,
      "reward_std": 0.3163410574197769,
      "reward_std_all": 0.2758568376302719,
      "rewards-max/forensics_sinkhorn_reward": 0.9783694446086884,
      "rewards-max/forensics_strict_edge_reward": 0.49907028675079346,
      "rewards-max/forensics_tiou_reward": 0.9688303768634796,
      "rewards/forensics_sinkhorn_reward": 0.7673153877258301,
      "rewards/forensics_strict_edge_reward": 0.03234481252729893,
      "rewards/forensics_tiou_reward": 0.5746304988861084,
      "step": 749
    },
    {
      "completion_length": 22.21875,
      "epoch": 2.3010752688172045,
      "grad_norm": 9.622030258178711,
      "kl": 0.43896484375,
      "learning_rate": 4.2384615384615383e-07,
      "loss": -0.01,
      "reward": 1.327721655368805,
      "reward_std": 0.19281350821256638,
      "reward_std_all": 0.18491213023662567,
      "rewards-max/forensics_sinkhorn_reward": 0.9732184410095215,
      "rewards-max/forensics_strict_edge_reward": 0.5713528990745544,
      "rewards-max/forensics_tiou_reward": 0.9612651169300079,
      "rewards/forensics_sinkhorn_reward": 0.7236663103103638,
      "rewards/forensics_strict_edge_reward": 0.08374086022377014,
      "rewards/forensics_tiou_reward": 0.5203145295381546,
      "step": 750
    },
    {
      "completion_length": 22.96875,
      "epoch": 2.3041474654377883,
      "grad_norm": 12.39954662322998,
      "kl": 0.396484375,
      "learning_rate": 4.2307692307692304e-07,
      "loss": -0.1136,
      "reward": 1.294877827167511,
      "reward_std": 0.19173681735992432,
      "reward_std_all": 0.1952090635895729,
      "rewards-max/forensics_sinkhorn_reward": 0.9770511388778687,
      "rewards-max/forensics_strict_edge_reward": 0.5112937390804291,
      "rewards-max/forensics_tiou_reward": 0.96414515376091,
      "rewards/forensics_sinkhorn_reward": 0.7455402612686157,
      "rewards/forensics_strict_edge_reward": 0.03083231206983328,
      "rewards/forensics_tiou_reward": 0.518505185842514,
      "step": 751
    },
    {
      "completion_length": 20.765625,
      "epoch": 2.3072196620583716,
      "grad_norm": 11.730884552001953,
      "kl": 0.4189453125,
      "learning_rate": 4.223076923076923e-07,
      "loss": -0.0002,
      "reward": 1.2531389594078064,
      "reward_std": 0.2926245480775833,
      "reward_std_all": 0.27505116164684296,
      "rewards-max/forensics_sinkhorn_reward": 0.9644801616668701,
      "rewards-max/forensics_strict_edge_reward": 0.35702988505363464,
      "rewards-max/forensics_tiou_reward": 0.9392202794551849,
      "rewards/forensics_sinkhorn_reward": 0.7354250550270081,
      "rewards/forensics_strict_edge_reward": 0.0388605697080493,
      "rewards/forensics_tiou_reward": 0.478853315114975,
      "step": 752
    },
    {
      "completion_length": 21.765625,
      "epoch": 2.3102918586789554,
      "grad_norm": 11.224684715270996,
      "kl": 0.35205078125,
      "learning_rate": 4.215384615384615e-07,
      "loss": -0.0247,
      "reward": 1.39872145652771,
      "reward_std": 0.2007618024945259,
      "reward_std_all": 0.18426711857318878,
      "rewards-max/forensics_sinkhorn_reward": 0.956025093793869,
      "rewards-max/forensics_strict_edge_reward": 0.27939875423908234,
      "rewards-max/forensics_tiou_reward": 0.909830778837204,
      "rewards/forensics_sinkhorn_reward": 0.7781442999839783,
      "rewards/forensics_strict_edge_reward": 0.028734908439218998,
      "rewards/forensics_tiou_reward": 0.5918421745300293,
      "step": 753
    },
    {
      "completion_length": 24.0625,
      "epoch": 2.313364055299539,
      "grad_norm": 9.164552688598633,
      "kl": 0.23583984375,
      "learning_rate": 4.207692307692308e-07,
      "loss": -0.0363,
      "reward": 1.4512322545051575,
      "reward_std": 0.19206750392913818,
      "reward_std_all": 0.17886072397232056,
      "rewards-max/forensics_sinkhorn_reward": 0.9725695550441742,
      "rewards-max/forensics_strict_edge_reward": 0.2657751739025116,
      "rewards-max/forensics_tiou_reward": 0.9564333558082581,
      "rewards/forensics_sinkhorn_reward": 0.796686589717865,
      "rewards/forensics_strict_edge_reward": 0.04204991739243269,
      "rewards/forensics_tiou_reward": 0.6124956607818604,
      "step": 754
    },
    {
      "completion_length": 21.25,
      "epoch": 2.316436251920123,
      "grad_norm": 10.725934028625488,
      "kl": 0.2890625,
      "learning_rate": 4.1999999999999995e-07,
      "loss": -0.0222,
      "reward": 1.2700047492980957,
      "reward_std": 0.27646180987358093,
      "reward_std_all": 0.24161545187234879,
      "rewards-max/forensics_sinkhorn_reward": 0.9780028164386749,
      "rewards-max/forensics_strict_edge_reward": 0.578471302986145,
      "rewards-max/forensics_tiou_reward": 0.968295693397522,
      "rewards/forensics_sinkhorn_reward": 0.7312234342098236,
      "rewards/forensics_strict_edge_reward": 0.057602958753705025,
      "rewards/forensics_tiou_reward": 0.4811783730983734,
      "step": 755
    },
    {
      "completion_length": 23.046875,
      "epoch": 2.3195084485407067,
      "grad_norm": 11.347561836242676,
      "kl": 0.37109375,
      "learning_rate": 4.192307692307692e-07,
      "loss": -0.1196,
      "reward": 1.6332468390464783,
      "reward_std": 0.3116030693054199,
      "reward_std_all": 0.2849421054124832,
      "rewards-max/forensics_sinkhorn_reward": 0.9660328924655914,
      "rewards-max/forensics_strict_edge_reward": 0.5670294612646103,
      "rewards-max/forensics_tiou_reward": 0.9444810152053833,
      "rewards/forensics_sinkhorn_reward": 0.8358586728572845,
      "rewards/forensics_strict_edge_reward": 0.1013268232345581,
      "rewards/forensics_tiou_reward": 0.696061372756958,
      "step": 756
    },
    {
      "completion_length": 20.765625,
      "epoch": 2.3225806451612905,
      "grad_norm": 15.727842330932617,
      "kl": 0.26953125,
      "learning_rate": 4.1846153846153843e-07,
      "loss": 0.0535,
      "reward": 1.026478260755539,
      "reward_std": 0.2765759378671646,
      "reward_std_all": 0.24432802200317383,
      "rewards-max/forensics_sinkhorn_reward": 0.9572172462940216,
      "rewards-max/forensics_strict_edge_reward": 0.4051655661314726,
      "rewards-max/forensics_tiou_reward": 0.9186146259307861,
      "rewards/forensics_sinkhorn_reward": 0.6246815919876099,
      "rewards/forensics_strict_edge_reward": 0.03136992361396551,
      "rewards/forensics_tiou_reward": 0.37042680382728577,
      "step": 757
    },
    {
      "completion_length": 21.0625,
      "epoch": 2.3256528417818743,
      "grad_norm": 11.782613754272461,
      "kl": 0.27587890625,
      "learning_rate": 4.176923076923077e-07,
      "loss": -0.05,
      "reward": 1.3714932203292847,
      "reward_std": 0.32203513383865356,
      "reward_std_all": 0.29782266914844513,
      "rewards-max/forensics_sinkhorn_reward": 0.9804673492908478,
      "rewards-max/forensics_strict_edge_reward": 0.3782782405614853,
      "rewards-max/forensics_tiou_reward": 0.9738859534263611,
      "rewards/forensics_sinkhorn_reward": 0.7522217929363251,
      "rewards/forensics_strict_edge_reward": 0.06730222702026367,
      "rewards/forensics_tiou_reward": 0.5519692152738571,
      "step": 758
    },
    {
      "completion_length": 24.3125,
      "epoch": 2.3287250384024576,
      "grad_norm": 9.517403602600098,
      "kl": 0.25537109375,
      "learning_rate": 4.1692307692307686e-07,
      "loss": 0.0302,
      "reward": 1.2813096046447754,
      "reward_std": 0.24926328659057617,
      "reward_std_all": 0.28568488359451294,
      "rewards-max/forensics_sinkhorn_reward": 0.9604698717594147,
      "rewards-max/forensics_strict_edge_reward": 0.5579978227615356,
      "rewards-max/forensics_tiou_reward": 0.9200118780136108,
      "rewards/forensics_sinkhorn_reward": 0.7220068573951721,
      "rewards/forensics_strict_edge_reward": 0.04649675264954567,
      "rewards/forensics_tiou_reward": 0.5128060132265091,
      "step": 759
    },
    {
      "completion_length": 23.640625,
      "epoch": 2.3317972350230414,
      "grad_norm": 10.939369201660156,
      "kl": 0.3525390625,
      "learning_rate": 4.161538461538461e-07,
      "loss": 0.0332,
      "reward": 1.243879497051239,
      "reward_std": 0.3182436227798462,
      "reward_std_all": 0.28602173179388046,
      "rewards-max/forensics_sinkhorn_reward": 0.9857254922389984,
      "rewards-max/forensics_strict_edge_reward": 0.7610765993595123,
      "rewards-max/forensics_tiou_reward": 0.9916837811470032,
      "rewards/forensics_sinkhorn_reward": 0.687411904335022,
      "rewards/forensics_strict_edge_reward": 0.057482123374938965,
      "rewards/forensics_tiou_reward": 0.4989854544401169,
      "step": 760
    },
    {
      "completion_length": 23.65625,
      "epoch": 2.334869431643625,
      "grad_norm": 9.888252258300781,
      "kl": 0.302734375,
      "learning_rate": 4.153846153846154e-07,
      "loss": 0.088,
      "reward": 1.3105416297912598,
      "reward_std": 0.25601087510585785,
      "reward_std_all": 0.22053851932287216,
      "rewards-max/forensics_sinkhorn_reward": 0.9881342351436615,
      "rewards-max/forensics_strict_edge_reward": 0.6388925462961197,
      "rewards-max/forensics_tiou_reward": 0.9967740774154663,
      "rewards/forensics_sinkhorn_reward": 0.7054074704647064,
      "rewards/forensics_strict_edge_reward": 0.04916177969425917,
      "rewards/forensics_tiou_reward": 0.5559723079204559,
      "step": 761
    },
    {
      "completion_length": 26.140625,
      "epoch": 2.337941628264209,
      "grad_norm": 9.412310600280762,
      "kl": 0.24267578125,
      "learning_rate": 4.146153846153846e-07,
      "loss": -0.0321,
      "reward": 1.5419813394546509,
      "reward_std": 0.2363906055688858,
      "reward_std_all": 0.25844020396471024,
      "rewards-max/forensics_sinkhorn_reward": 0.9686270654201508,
      "rewards-max/forensics_strict_edge_reward": 0.3095613196492195,
      "rewards-max/forensics_tiou_reward": 0.9467291533946991,
      "rewards/forensics_sinkhorn_reward": 0.8163830935955048,
      "rewards/forensics_strict_edge_reward": 0.04819411691278219,
      "rewards/forensics_tiou_reward": 0.6774041056632996,
      "step": 762
    },
    {
      "completion_length": 22.9375,
      "epoch": 2.3410138248847927,
      "grad_norm": 10.986971855163574,
      "kl": 0.22412109375,
      "learning_rate": 4.1384615384615387e-07,
      "loss": 0.1148,
      "reward": 1.463632583618164,
      "reward_std": 0.2017507180571556,
      "reward_std_all": 0.19584538787603378,
      "rewards-max/forensics_sinkhorn_reward": 0.9720035493373871,
      "rewards-max/forensics_strict_edge_reward": 0.3928476870059967,
      "rewards-max/forensics_tiou_reward": 0.9562292397022247,
      "rewards/forensics_sinkhorn_reward": 0.7943265438079834,
      "rewards/forensics_strict_edge_reward": 0.06581865902990103,
      "rewards/forensics_tiou_reward": 0.6034873723983765,
      "step": 763
    },
    {
      "completion_length": 20.25,
      "epoch": 2.3440860215053765,
      "grad_norm": 11.390172004699707,
      "kl": 0.26025390625,
      "learning_rate": 4.1307692307692303e-07,
      "loss": 0.1039,
      "reward": 1.3612224459648132,
      "reward_std": 0.23423030972480774,
      "reward_std_all": 0.2065541297197342,
      "rewards-max/forensics_sinkhorn_reward": 0.9598428308963776,
      "rewards-max/forensics_strict_edge_reward": 0.18221481144428253,
      "rewards-max/forensics_tiou_reward": 0.9173426628112793,
      "rewards/forensics_sinkhorn_reward": 0.7639724612236023,
      "rewards/forensics_strict_edge_reward": 0.02181901689618826,
      "rewards/forensics_tiou_reward": 0.5754309594631195,
      "step": 764
    },
    {
      "completion_length": 19.796875,
      "epoch": 2.3471582181259603,
      "grad_norm": 11.755607604980469,
      "kl": 0.3134765625,
      "learning_rate": 4.123076923076923e-07,
      "loss": -0.0187,
      "reward": 1.3631086349487305,
      "reward_std": 0.2584025710821152,
      "reward_std_all": 0.24230988323688507,
      "rewards-max/forensics_sinkhorn_reward": 0.9637530446052551,
      "rewards-max/forensics_strict_edge_reward": 0.35939208418130875,
      "rewards-max/forensics_tiou_reward": 0.9293697476387024,
      "rewards/forensics_sinkhorn_reward": 0.7694149911403656,
      "rewards/forensics_strict_edge_reward": 0.042337330523878336,
      "rewards/forensics_tiou_reward": 0.5513563305139542,
      "step": 765
    },
    {
      "completion_length": 23.28125,
      "epoch": 2.3502304147465436,
      "grad_norm": 11.662776947021484,
      "kl": 0.244140625,
      "learning_rate": 4.115384615384615e-07,
      "loss": 0.1385,
      "reward": 1.3223245739936829,
      "reward_std": 0.1989414095878601,
      "reward_std_all": 0.17655076831579208,
      "rewards-max/forensics_sinkhorn_reward": 0.9661204218864441,
      "rewards-max/forensics_strict_edge_reward": 0.16919660102576017,
      "rewards-max/forensics_tiou_reward": 0.9323014616966248,
      "rewards/forensics_sinkhorn_reward": 0.7477664053440094,
      "rewards/forensics_strict_edge_reward": 0.013812702149152756,
      "rewards/forensics_tiou_reward": 0.5607453882694244,
      "step": 766
    },
    {
      "completion_length": 20.890625,
      "epoch": 2.3533026113671274,
      "grad_norm": 12.805598258972168,
      "kl": 0.4521484375,
      "learning_rate": 4.107692307692308e-07,
      "loss": -0.0947,
      "reward": 1.3474024534225464,
      "reward_std": 0.23038172721862793,
      "reward_std_all": 0.2153305858373642,
      "rewards-max/forensics_sinkhorn_reward": 0.9757857620716095,
      "rewards-max/forensics_strict_edge_reward": 0.5598244965076447,
      "rewards-max/forensics_tiou_reward": 0.9652746915817261,
      "rewards/forensics_sinkhorn_reward": 0.7381727695465088,
      "rewards/forensics_strict_edge_reward": 0.06623298674821854,
      "rewards/forensics_tiou_reward": 0.542996734380722,
      "step": 767
    },
    {
      "completion_length": 20.265625,
      "epoch": 2.356374807987711,
      "grad_norm": 10.086044311523438,
      "kl": 0.2294921875,
      "learning_rate": 4.0999999999999994e-07,
      "loss": -0.0912,
      "reward": 1.2709108591079712,
      "reward_std": 0.13778051361441612,
      "reward_std_all": 0.1344408243894577,
      "rewards-max/forensics_sinkhorn_reward": 0.9586798846721649,
      "rewards-max/forensics_strict_edge_reward": 0.19715256989002228,
      "rewards-max/forensics_tiou_reward": 0.925335556268692,
      "rewards/forensics_sinkhorn_reward": 0.7385047078132629,
      "rewards/forensics_strict_edge_reward": 0.013027136912569404,
      "rewards/forensics_tiou_reward": 0.5193790197372437,
      "step": 768
    },
    {
      "completion_length": 22.09375,
      "epoch": 2.359447004608295,
      "grad_norm": 10.869244575500488,
      "kl": 0.283203125,
      "learning_rate": 4.092307692307692e-07,
      "loss": -0.1422,
      "reward": 1.4435021877288818,
      "reward_std": 0.28070826828479767,
      "reward_std_all": 0.23902779817581177,
      "rewards-max/forensics_sinkhorn_reward": 0.971327394247055,
      "rewards-max/forensics_strict_edge_reward": 0.5054457187652588,
      "rewards-max/forensics_tiou_reward": 0.954770416021347,
      "rewards/forensics_sinkhorn_reward": 0.7697545289993286,
      "rewards/forensics_strict_edge_reward": 0.0650859922170639,
      "rewards/forensics_tiou_reward": 0.608661562204361,
      "step": 769
    },
    {
      "completion_length": 22.6875,
      "epoch": 2.3625192012288787,
      "grad_norm": 8.81008529663086,
      "kl": 0.22119140625,
      "learning_rate": 4.0846153846153847e-07,
      "loss": -0.0487,
      "reward": 1.3520228266716003,
      "reward_std": 0.2242378145456314,
      "reward_std_all": 0.1977166086435318,
      "rewards-max/forensics_sinkhorn_reward": 0.9828217625617981,
      "rewards-max/forensics_strict_edge_reward": 0.3909902833402157,
      "rewards-max/forensics_tiou_reward": 0.9830019474029541,
      "rewards/forensics_sinkhorn_reward": 0.7342125773429871,
      "rewards/forensics_strict_edge_reward": 0.060633168555796146,
      "rewards/forensics_tiou_reward": 0.5571771264076233,
      "step": 770
    },
    {
      "completion_length": 22.671875,
      "epoch": 2.3655913978494625,
      "grad_norm": 9.60224723815918,
      "kl": 0.38037109375,
      "learning_rate": 4.076923076923077e-07,
      "loss": -0.0833,
      "reward": 1.3658120036125183,
      "reward_std": 0.2287544384598732,
      "reward_std_all": 0.2103642150759697,
      "rewards-max/forensics_sinkhorn_reward": 0.9699285626411438,
      "rewards-max/forensics_strict_edge_reward": 0.3607759475708008,
      "rewards-max/forensics_tiou_reward": 0.9482632577419281,
      "rewards/forensics_sinkhorn_reward": 0.7397760152816772,
      "rewards/forensics_strict_edge_reward": 0.06020559091120958,
      "rewards/forensics_tiou_reward": 0.565830335021019,
      "step": 771
    },
    {
      "completion_length": 20.09375,
      "epoch": 2.3686635944700463,
      "grad_norm": 10.554018020629883,
      "kl": 0.2763671875,
      "learning_rate": 4.069230769230769e-07,
      "loss": -0.0561,
      "reward": 1.3868802785873413,
      "reward_std": 0.24731146544218063,
      "reward_std_all": 0.2449600100517273,
      "rewards-max/forensics_sinkhorn_reward": 0.9684901833534241,
      "rewards-max/forensics_strict_edge_reward": 0.38785459846258163,
      "rewards-max/forensics_tiou_reward": 0.9357369840145111,
      "rewards/forensics_sinkhorn_reward": 0.7515300810337067,
      "rewards/forensics_strict_edge_reward": 0.04466840624809265,
      "rewards/forensics_tiou_reward": 0.5906818211078644,
      "step": 772
    },
    {
      "completion_length": 22.484375,
      "epoch": 2.3717357910906296,
      "grad_norm": 10.654083251953125,
      "kl": 0.2666015625,
      "learning_rate": 4.061538461538461e-07,
      "loss": -0.0256,
      "reward": 1.2953341007232666,
      "reward_std": 0.2820679396390915,
      "reward_std_all": 0.25822117924690247,
      "rewards-max/forensics_sinkhorn_reward": 0.9595139622688293,
      "rewards-max/forensics_strict_edge_reward": 0.5629585310816765,
      "rewards-max/forensics_tiou_reward": 0.9277314841747284,
      "rewards/forensics_sinkhorn_reward": 0.7364250719547272,
      "rewards/forensics_strict_edge_reward": 0.0491037592291832,
      "rewards/forensics_tiou_reward": 0.5098052322864532,
      "step": 773
    },
    {
      "completion_length": 20.6875,
      "epoch": 2.3748079877112134,
      "grad_norm": 11.596759796142578,
      "kl": 0.2216796875,
      "learning_rate": 4.053846153846154e-07,
      "loss": -0.0064,
      "reward": 1.3237289190292358,
      "reward_std": 0.13332410901784897,
      "reward_std_all": 0.11775187402963638,
      "rewards-max/forensics_sinkhorn_reward": 0.9438167810440063,
      "rewards-max/forensics_strict_edge_reward": 0.19952420890331268,
      "rewards-max/forensics_tiou_reward": 0.856501042842865,
      "rewards/forensics_sinkhorn_reward": 0.7659505307674408,
      "rewards/forensics_strict_edge_reward": 0.02041999064385891,
      "rewards/forensics_tiou_reward": 0.5373583734035492,
      "step": 774
    },
    {
      "completion_length": 21.65625,
      "epoch": 2.377880184331797,
      "grad_norm": 10.174251556396484,
      "kl": 0.29931640625,
      "learning_rate": 4.046153846153846e-07,
      "loss": -0.05,
      "reward": 1.4338592290878296,
      "reward_std": 0.20418281853199005,
      "reward_std_all": 0.2038515955209732,
      "rewards-max/forensics_sinkhorn_reward": 0.975926548242569,
      "rewards-max/forensics_strict_edge_reward": 0.43341680616140366,
      "rewards-max/forensics_tiou_reward": 0.9634720683097839,
      "rewards/forensics_sinkhorn_reward": 0.7816879153251648,
      "rewards/forensics_strict_edge_reward": 0.047806454822421074,
      "rewards/forensics_tiou_reward": 0.6043649315834045,
      "step": 775
    },
    {
      "completion_length": 21.5,
      "epoch": 2.380952380952381,
      "grad_norm": 10.895867347717285,
      "kl": 0.291015625,
      "learning_rate": 4.0384615384615386e-07,
      "loss": -0.0717,
      "reward": 1.3456632494926453,
      "reward_std": 0.2364686131477356,
      "reward_std_all": 0.22236812114715576,
      "rewards-max/forensics_sinkhorn_reward": 0.9539422690868378,
      "rewards-max/forensics_strict_edge_reward": 0.2724028080701828,
      "rewards-max/forensics_tiou_reward": 0.9087848961353302,
      "rewards/forensics_sinkhorn_reward": 0.7522060573101044,
      "rewards/forensics_strict_edge_reward": 0.03931000642478466,
      "rewards/forensics_tiou_reward": 0.5541472733020782,
      "step": 776
    },
    {
      "completion_length": 22.203125,
      "epoch": 2.3840245775729647,
      "grad_norm": 13.154898643493652,
      "kl": 0.201171875,
      "learning_rate": 4.03076923076923e-07,
      "loss": -0.1414,
      "reward": 1.6368860602378845,
      "reward_std": 0.20124142616987228,
      "reward_std_all": 0.20315495133399963,
      "rewards-max/forensics_sinkhorn_reward": 0.9814525544643402,
      "rewards-max/forensics_strict_edge_reward": 0.5488366484642029,
      "rewards-max/forensics_tiou_reward": 0.9784179329872131,
      "rewards/forensics_sinkhorn_reward": 0.8313721120357513,
      "rewards/forensics_strict_edge_reward": 0.08481402695178986,
      "rewards/forensics_tiou_reward": 0.7206999361515045,
      "step": 777
    },
    {
      "completion_length": 24.109375,
      "epoch": 2.3870967741935485,
      "grad_norm": 14.374053955078125,
      "kl": 0.328125,
      "learning_rate": 4.023076923076923e-07,
      "loss": 0.03,
      "reward": 1.4306070804595947,
      "reward_std": 0.25353682041168213,
      "reward_std_all": 0.22391550987958908,
      "rewards-max/forensics_sinkhorn_reward": 0.9761478006839752,
      "rewards-max/forensics_strict_edge_reward": 0.26856348663568497,
      "rewards-max/forensics_tiou_reward": 0.9663487672805786,
      "rewards/forensics_sinkhorn_reward": 0.7820566594600677,
      "rewards/forensics_strict_edge_reward": 0.043684765696525574,
      "rewards/forensics_tiou_reward": 0.6048655807971954,
      "step": 778
    },
    {
      "completion_length": 23.171875,
      "epoch": 2.3901689708141323,
      "grad_norm": 10.27751350402832,
      "kl": 0.23291015625,
      "learning_rate": 4.015384615384615e-07,
      "loss": 0.0185,
      "reward": 1.20009046792984,
      "reward_std": 0.21904271841049194,
      "reward_std_all": 0.22290873527526855,
      "rewards-max/forensics_sinkhorn_reward": 0.9439882338047028,
      "rewards-max/forensics_strict_edge_reward": 0.12693892791867256,
      "rewards-max/forensics_tiou_reward": 0.8845783174037933,
      "rewards/forensics_sinkhorn_reward": 0.6988508999347687,
      "rewards/forensics_strict_edge_reward": 0.014504353050142527,
      "rewards/forensics_tiou_reward": 0.48673516511917114,
      "step": 779
    },
    {
      "completion_length": 22.96875,
      "epoch": 2.3932411674347156,
      "grad_norm": 10.01811408996582,
      "kl": 0.3037109375,
      "learning_rate": 4.0076923076923077e-07,
      "loss": -0.0243,
      "reward": 1.473280906677246,
      "reward_std": 0.27837902307510376,
      "reward_std_all": 0.24183350056409836,
      "rewards-max/forensics_sinkhorn_reward": 0.9700564444065094,
      "rewards-max/forensics_strict_edge_reward": 0.5770299434661865,
      "rewards-max/forensics_tiou_reward": 0.9501800537109375,
      "rewards/forensics_sinkhorn_reward": 0.800848126411438,
      "rewards/forensics_strict_edge_reward": 0.05308285355567932,
      "rewards/forensics_tiou_reward": 0.6193499267101288,
      "step": 780
    },
    {
      "completion_length": 21.921875,
      "epoch": 2.3963133640552994,
      "grad_norm": 10.939592361450195,
      "kl": 0.2861328125,
      "learning_rate": 4e-07,
      "loss": 0.015,
      "reward": 1.4780138731002808,
      "reward_std": 0.25723959505558014,
      "reward_std_all": 0.2628229707479477,
      "rewards-max/forensics_sinkhorn_reward": 0.9787856936454773,
      "rewards-max/forensics_strict_edge_reward": 0.39828625321388245,
      "rewards-max/forensics_tiou_reward": 0.9689232409000397,
      "rewards/forensics_sinkhorn_reward": 0.7943938076496124,
      "rewards/forensics_strict_edge_reward": 0.07094770297408104,
      "rewards/forensics_tiou_reward": 0.6126723885536194,
      "step": 781
    },
    {
      "completion_length": 21.359375,
      "epoch": 2.399385560675883,
      "grad_norm": 10.768976211547852,
      "kl": 0.2822265625,
      "learning_rate": 3.992307692307692e-07,
      "loss": 0.0717,
      "reward": 1.271946132183075,
      "reward_std": 0.17726386711001396,
      "reward_std_all": 0.2086327038705349,
      "rewards-max/forensics_sinkhorn_reward": 0.9796218276023865,
      "rewards-max/forensics_strict_edge_reward": 0.44064828753471375,
      "rewards-max/forensics_tiou_reward": 0.9751729965209961,
      "rewards/forensics_sinkhorn_reward": 0.7199679911136627,
      "rewards/forensics_strict_edge_reward": 0.04237673711031675,
      "rewards/forensics_tiou_reward": 0.5096014142036438,
      "step": 782
    },
    {
      "completion_length": 22.5625,
      "epoch": 2.402457757296467,
      "grad_norm": 10.943729400634766,
      "kl": 0.4951171875,
      "learning_rate": 3.9846153846153846e-07,
      "loss": 0.0726,
      "reward": 1.3694891333580017,
      "reward_std": 0.33554697036743164,
      "reward_std_all": 0.2915696054697037,
      "rewards-max/forensics_sinkhorn_reward": 0.9806016683578491,
      "rewards-max/forensics_strict_edge_reward": 0.6799494326114655,
      "rewards-max/forensics_tiou_reward": 0.9796177446842194,
      "rewards/forensics_sinkhorn_reward": 0.7348682284355164,
      "rewards/forensics_strict_edge_reward": 0.07928145676851273,
      "rewards/forensics_tiou_reward": 0.555339440703392,
      "step": 783
    },
    {
      "completion_length": 23.921875,
      "epoch": 2.4055299539170507,
      "grad_norm": 11.679808616638184,
      "kl": 0.455078125,
      "learning_rate": 3.976923076923077e-07,
      "loss": 0.104,
      "reward": 1.3858295679092407,
      "reward_std": 0.3572094142436981,
      "reward_std_all": 0.37737976014614105,
      "rewards-max/forensics_sinkhorn_reward": 0.9715498089790344,
      "rewards-max/forensics_strict_edge_reward": 0.38069719076156616,
      "rewards-max/forensics_tiou_reward": 0.957266241312027,
      "rewards/forensics_sinkhorn_reward": 0.7426102757453918,
      "rewards/forensics_strict_edge_reward": 0.0726019199937582,
      "rewards/forensics_tiou_reward": 0.5706174373626709,
      "step": 784
    },
    {
      "completion_length": 20.671875,
      "epoch": 2.4086021505376345,
      "grad_norm": 13.971087455749512,
      "kl": 0.43798828125,
      "learning_rate": 3.969230769230769e-07,
      "loss": -0.006,
      "reward": 1.3823739886283875,
      "reward_std": 0.28692860901355743,
      "reward_std_all": 0.26587916910648346,
      "rewards-max/forensics_sinkhorn_reward": 0.9799542427062988,
      "rewards-max/forensics_strict_edge_reward": 0.6720215678215027,
      "rewards-max/forensics_tiou_reward": 0.9702102839946747,
      "rewards/forensics_sinkhorn_reward": 0.7493628561496735,
      "rewards/forensics_strict_edge_reward": 0.08463025838136673,
      "rewards/forensics_tiou_reward": 0.5483809411525726,
      "step": 785
    },
    {
      "completion_length": 21.515625,
      "epoch": 2.4116743471582183,
      "grad_norm": 10.354826927185059,
      "kl": 0.2734375,
      "learning_rate": 3.961538461538461e-07,
      "loss": -0.0386,
      "reward": 1.3949527144432068,
      "reward_std": 0.21504557877779007,
      "reward_std_all": 0.2210068553686142,
      "rewards-max/forensics_sinkhorn_reward": 0.9790656268596649,
      "rewards-max/forensics_strict_edge_reward": 0.4874049425125122,
      "rewards-max/forensics_tiou_reward": 0.9681521356105804,
      "rewards/forensics_sinkhorn_reward": 0.7549998164176941,
      "rewards/forensics_strict_edge_reward": 0.05850530415773392,
      "rewards/forensics_tiou_reward": 0.5814476013183594,
      "step": 786
    },
    {
      "completion_length": 19.609375,
      "epoch": 2.4147465437788016,
      "grad_norm": 8.67326831817627,
      "kl": 0.6083984375,
      "learning_rate": 3.9538461538461537e-07,
      "loss": -0.0345,
      "reward": 1.4648275971412659,
      "reward_std": 0.2264089584350586,
      "reward_std_all": 0.20283548533916473,
      "rewards-max/forensics_sinkhorn_reward": 0.977522224187851,
      "rewards-max/forensics_strict_edge_reward": 0.4745236784219742,
      "rewards-max/forensics_tiou_reward": 0.9655039608478546,
      "rewards/forensics_sinkhorn_reward": 0.7988663911819458,
      "rewards/forensics_strict_edge_reward": 0.061660733073949814,
      "rewards/forensics_tiou_reward": 0.6043003499507904,
      "step": 787
    },
    {
      "completion_length": 21.28125,
      "epoch": 2.4178187403993854,
      "grad_norm": 13.170387268066406,
      "kl": 0.4033203125,
      "learning_rate": 3.946153846153846e-07,
      "loss": 0.019,
      "reward": 1.4443080425262451,
      "reward_std": 0.31746093928813934,
      "reward_std_all": 0.2883835583925247,
      "rewards-max/forensics_sinkhorn_reward": 0.9769907593727112,
      "rewards-max/forensics_strict_edge_reward": 0.5305682122707367,
      "rewards-max/forensics_tiou_reward": 0.963786244392395,
      "rewards/forensics_sinkhorn_reward": 0.7650296688079834,
      "rewards/forensics_strict_edge_reward": 0.07452582381665707,
      "rewards/forensics_tiou_reward": 0.6047524809837341,
      "step": 788
    },
    {
      "completion_length": 20.28125,
      "epoch": 2.420890937019969,
      "grad_norm": 13.159246444702148,
      "kl": 0.185546875,
      "learning_rate": 3.9384615384615385e-07,
      "loss": 0.0459,
      "reward": 1.29258394241333,
      "reward_std": 0.15053492039442062,
      "reward_std_all": 0.16194980964064598,
      "rewards-max/forensics_sinkhorn_reward": 0.958722323179245,
      "rewards-max/forensics_strict_edge_reward": 0.2186184898018837,
      "rewards-max/forensics_tiou_reward": 0.9189734756946564,
      "rewards/forensics_sinkhorn_reward": 0.7371228039264679,
      "rewards/forensics_strict_edge_reward": 0.024296922609210014,
      "rewards/forensics_tiou_reward": 0.5311641991138458,
      "step": 789
    },
    {
      "completion_length": 21.8125,
      "epoch": 2.423963133640553,
      "grad_norm": 13.845876693725586,
      "kl": 0.19140625,
      "learning_rate": 3.9307692307692306e-07,
      "loss": -0.0036,
      "reward": 1.456020474433899,
      "reward_std": 0.16335858404636383,
      "reward_std_all": 0.14579328149557114,
      "rewards-max/forensics_sinkhorn_reward": 0.9635195732116699,
      "rewards-max/forensics_strict_edge_reward": 0.3416936919093132,
      "rewards-max/forensics_tiou_reward": 0.925542801618576,
      "rewards/forensics_sinkhorn_reward": 0.8031153976917267,
      "rewards/forensics_strict_edge_reward": 0.050154754891991615,
      "rewards/forensics_tiou_reward": 0.602750301361084,
      "step": 790
    },
    {
      "completion_length": 20.921875,
      "epoch": 2.4270353302611367,
      "grad_norm": 9.931966781616211,
      "kl": 0.3271484375,
      "learning_rate": 3.923076923076923e-07,
      "loss": -0.014,
      "reward": 1.3769100904464722,
      "reward_std": 0.27725084125995636,
      "reward_std_all": 0.23835259675979614,
      "rewards-max/forensics_sinkhorn_reward": 0.984178364276886,
      "rewards-max/forensics_strict_edge_reward": 0.6236754059791565,
      "rewards-max/forensics_tiou_reward": 0.9861176311969757,
      "rewards/forensics_sinkhorn_reward": 0.7582228481769562,
      "rewards/forensics_strict_edge_reward": 0.0884432066231966,
      "rewards/forensics_tiou_reward": 0.5302440226078033,
      "step": 791
    },
    {
      "completion_length": 22.734375,
      "epoch": 2.4301075268817205,
      "grad_norm": 10.616737365722656,
      "kl": 0.5322265625,
      "learning_rate": 3.9153846153846154e-07,
      "loss": 0.013,
      "reward": 1.3398504257202148,
      "reward_std": 0.1377052739262581,
      "reward_std_all": 0.1270044669508934,
      "rewards-max/forensics_sinkhorn_reward": 0.9762083292007446,
      "rewards-max/forensics_strict_edge_reward": 0.27389559149742126,
      "rewards-max/forensics_tiou_reward": 0.9678812623023987,
      "rewards/forensics_sinkhorn_reward": 0.7331571280956268,
      "rewards/forensics_strict_edge_reward": 0.04024490714073181,
      "rewards/forensics_tiou_reward": 0.5664485096931458,
      "step": 792
    },
    {
      "completion_length": 21.1875,
      "epoch": 2.4331797235023043,
      "grad_norm": 12.227115631103516,
      "kl": 0.3037109375,
      "learning_rate": 3.9076923076923076e-07,
      "loss": -0.0879,
      "reward": 1.3973493576049805,
      "reward_std": 0.2091999426484108,
      "reward_std_all": 0.1875329092144966,
      "rewards-max/forensics_sinkhorn_reward": 0.9872865080833435,
      "rewards-max/forensics_strict_edge_reward": 0.6862294971942902,
      "rewards-max/forensics_tiou_reward": 0.9944769740104675,
      "rewards/forensics_sinkhorn_reward": 0.7644168734550476,
      "rewards/forensics_strict_edge_reward": 0.05680612288415432,
      "rewards/forensics_tiou_reward": 0.5761262625455856,
      "step": 793
    },
    {
      "completion_length": 24.015625,
      "epoch": 2.436251920122888,
      "grad_norm": 9.725693702697754,
      "kl": 0.26904296875,
      "learning_rate": 3.8999999999999997e-07,
      "loss": -0.0281,
      "reward": 1.4644604921340942,
      "reward_std": 0.2481151968240738,
      "reward_std_all": 0.2138824164867401,
      "rewards-max/forensics_sinkhorn_reward": 0.9710765480995178,
      "rewards-max/forensics_strict_edge_reward": 0.27180543541908264,
      "rewards-max/forensics_tiou_reward": 0.9562309384346008,
      "rewards/forensics_sinkhorn_reward": 0.7995815873146057,
      "rewards/forensics_strict_edge_reward": 0.04008668940514326,
      "rewards/forensics_tiou_reward": 0.6247922778129578,
      "step": 794
    },
    {
      "completion_length": 23.0,
      "epoch": 2.439324116743472,
      "grad_norm": 12.042327880859375,
      "kl": 0.25244140625,
      "learning_rate": 3.892307692307692e-07,
      "loss": 0.0583,
      "reward": 1.2589755654335022,
      "reward_std": 0.14855455607175827,
      "reward_std_all": 0.14315586537122726,
      "rewards-max/forensics_sinkhorn_reward": 0.9563076794147491,
      "rewards-max/forensics_strict_edge_reward": 0.21962502971291542,
      "rewards-max/forensics_tiou_reward": 0.9171238839626312,
      "rewards/forensics_sinkhorn_reward": 0.7236918807029724,
      "rewards/forensics_strict_edge_reward": 0.026951636653393507,
      "rewards/forensics_tiou_reward": 0.5083320438861847,
      "step": 795
    },
    {
      "completion_length": 23.015625,
      "epoch": 2.442396313364055,
      "grad_norm": 15.708654403686523,
      "kl": 0.8388671875,
      "learning_rate": 3.8846153846153845e-07,
      "loss": 0.0058,
      "reward": 1.5332238674163818,
      "reward_std": 0.3167601078748703,
      "reward_std_all": 0.29440926015377045,
      "rewards-max/forensics_sinkhorn_reward": 0.9873577058315277,
      "rewards-max/forensics_strict_edge_reward": 0.8929734230041504,
      "rewards-max/forensics_tiou_reward": 0.9947774112224579,
      "rewards/forensics_sinkhorn_reward": 0.7896906733512878,
      "rewards/forensics_strict_edge_reward": 0.11177186667919159,
      "rewards/forensics_tiou_reward": 0.631761223077774,
      "step": 796
    },
    {
      "completion_length": 19.921875,
      "epoch": 2.445468509984639,
      "grad_norm": 11.226948738098145,
      "kl": 0.20849609375,
      "learning_rate": 3.8769230769230767e-07,
      "loss": 0.0391,
      "reward": 1.448264241218567,
      "reward_std": 0.1421080157160759,
      "reward_std_all": 0.13557954132556915,
      "rewards-max/forensics_sinkhorn_reward": 0.9620386064052582,
      "rewards-max/forensics_strict_edge_reward": 0.2365882769227028,
      "rewards-max/forensics_tiou_reward": 0.9290476441383362,
      "rewards/forensics_sinkhorn_reward": 0.8027100563049316,
      "rewards/forensics_strict_edge_reward": 0.022408622317016125,
      "rewards/forensics_tiou_reward": 0.6231456995010376,
      "step": 797
    },
    {
      "completion_length": 21.265625,
      "epoch": 2.4485407066052227,
      "grad_norm": 10.725789070129395,
      "kl": 0.3994140625,
      "learning_rate": 3.8692307692307693e-07,
      "loss": 0.1203,
      "reward": 1.3880788683891296,
      "reward_std": 0.18379812687635422,
      "reward_std_all": 0.17541685700416565,
      "rewards-max/forensics_sinkhorn_reward": 0.9687990844249725,
      "rewards-max/forensics_strict_edge_reward": 0.4264484792947769,
      "rewards-max/forensics_tiou_reward": 0.9430351853370667,
      "rewards/forensics_sinkhorn_reward": 0.7669564187526703,
      "rewards/forensics_strict_edge_reward": 0.056791868060827255,
      "rewards/forensics_tiou_reward": 0.5643305629491806,
      "step": 798
    },
    {
      "completion_length": 21.109375,
      "epoch": 2.4516129032258065,
      "grad_norm": 11.75031566619873,
      "kl": 0.365234375,
      "learning_rate": 3.8615384615384615e-07,
      "loss": -0.0238,
      "reward": 1.529003620147705,
      "reward_std": 0.26075195521116257,
      "reward_std_all": 0.24238275736570358,
      "rewards-max/forensics_sinkhorn_reward": 0.9875153601169586,
      "rewards-max/forensics_strict_edge_reward": 0.6317985653877258,
      "rewards-max/forensics_tiou_reward": 0.9950161874294281,
      "rewards/forensics_sinkhorn_reward": 0.7866809368133545,
      "rewards/forensics_strict_edge_reward": 0.09184180852025747,
      "rewards/forensics_tiou_reward": 0.6504807472229004,
      "step": 799
    },
    {
      "completion_length": 22.625,
      "epoch": 2.4546850998463903,
      "grad_norm": 13.372367858886719,
      "kl": 0.318359375,
      "learning_rate": 3.8538461538461536e-07,
      "loss": -0.0421,
      "reward": 1.472043216228485,
      "reward_std": 0.406621515750885,
      "reward_std_all": 0.3607543408870697,
      "rewards-max/forensics_sinkhorn_reward": 0.9742529988288879,
      "rewards-max/forensics_strict_edge_reward": 0.5702531039714813,
      "rewards-max/forensics_tiou_reward": 0.9632005989551544,
      "rewards/forensics_sinkhorn_reward": 0.7656646966934204,
      "rewards/forensics_strict_edge_reward": 0.08745726570487022,
      "rewards/forensics_tiou_reward": 0.6189212799072266,
      "step": 800
    },
    {
      "completion_length": 22.3125,
      "epoch": 2.457757296466974,
      "grad_norm": 14.652676582336426,
      "kl": 0.671875,
      "learning_rate": 3.8461538461538463e-07,
      "loss": 0.0299,
      "reward": 1.4071400165557861,
      "reward_std": 0.13959667831659317,
      "reward_std_all": 0.16489717364311218,
      "rewards-max/forensics_sinkhorn_reward": 0.9812693893909454,
      "rewards-max/forensics_strict_edge_reward": 0.5548860281705856,
      "rewards-max/forensics_tiou_reward": 0.976308673620224,
      "rewards/forensics_sinkhorn_reward": 0.7790223062038422,
      "rewards/forensics_strict_edge_reward": 0.052538828924298286,
      "rewards/forensics_tiou_reward": 0.5755788683891296,
      "step": 801
    },
    {
      "completion_length": 20.390625,
      "epoch": 2.460829493087558,
      "grad_norm": 21.051029205322266,
      "kl": 0.3173828125,
      "learning_rate": 3.8384615384615384e-07,
      "loss": 0.0186,
      "reward": 1.1862305402755737,
      "reward_std": 0.2542012557387352,
      "reward_std_all": 0.22983192652463913,
      "rewards-max/forensics_sinkhorn_reward": 0.9485830068588257,
      "rewards-max/forensics_strict_edge_reward": 0.402368500828743,
      "rewards-max/forensics_tiou_reward": 0.907780796289444,
      "rewards/forensics_sinkhorn_reward": 0.7020200490951538,
      "rewards/forensics_strict_edge_reward": 0.0288679301738739,
      "rewards/forensics_tiou_reward": 0.4553425908088684,
      "step": 802
    },
    {
      "completion_length": 21.578125,
      "epoch": 2.463901689708141,
      "grad_norm": 12.476534843444824,
      "kl": 0.20849609375,
      "learning_rate": 3.8307692307692305e-07,
      "loss": 0.0617,
      "reward": 1.3158809542655945,
      "reward_std": 0.258402481675148,
      "reward_std_all": 0.22601058334112167,
      "rewards-max/forensics_sinkhorn_reward": 0.9845887422561646,
      "rewards-max/forensics_strict_edge_reward": 0.6736744344234467,
      "rewards-max/forensics_tiou_reward": 0.988658219575882,
      "rewards/forensics_sinkhorn_reward": 0.7118820250034332,
      "rewards/forensics_strict_edge_reward": 0.0638783760368824,
      "rewards/forensics_tiou_reward": 0.5401205718517303,
      "step": 803
    },
    {
      "completion_length": 20.65625,
      "epoch": 2.466973886328725,
      "grad_norm": 14.823670387268066,
      "kl": 0.3056640625,
      "learning_rate": 3.8230769230769227e-07,
      "loss": 0.1264,
      "reward": 1.3236182928085327,
      "reward_std": 0.26941124349832535,
      "reward_std_all": 0.2384197860956192,
      "rewards-max/forensics_sinkhorn_reward": 0.9822399616241455,
      "rewards-max/forensics_strict_edge_reward": 0.4639832302927971,
      "rewards-max/forensics_tiou_reward": 0.9818016290664673,
      "rewards/forensics_sinkhorn_reward": 0.7650987505912781,
      "rewards/forensics_strict_edge_reward": 0.05469747312599793,
      "rewards/forensics_tiou_reward": 0.503822073340416,
      "step": 804
    },
    {
      "completion_length": 22.546875,
      "epoch": 2.4700460829493087,
      "grad_norm": 9.264262199401855,
      "kl": 0.2998046875,
      "learning_rate": 3.8153846153846153e-07,
      "loss": 0.0793,
      "reward": 1.2861084342002869,
      "reward_std": 0.23551979660987854,
      "reward_std_all": 0.21287335455417633,
      "rewards-max/forensics_sinkhorn_reward": 0.9743353128433228,
      "rewards-max/forensics_strict_edge_reward": 0.12306453473865986,
      "rewards-max/forensics_tiou_reward": 0.9613737165927887,
      "rewards/forensics_sinkhorn_reward": 0.7484031319618225,
      "rewards/forensics_strict_edge_reward": 0.011118072085082531,
      "rewards/forensics_tiou_reward": 0.5265872180461884,
      "step": 805
    },
    {
      "completion_length": 20.53125,
      "epoch": 2.4731182795698925,
      "grad_norm": 13.676507949829102,
      "kl": 0.36767578125,
      "learning_rate": 3.8076923076923075e-07,
      "loss": 0.0335,
      "reward": 1.3152554035186768,
      "reward_std": 0.18082130700349808,
      "reward_std_all": 0.17470596730709076,
      "rewards-max/forensics_sinkhorn_reward": 0.9578921794891357,
      "rewards-max/forensics_strict_edge_reward": 0.36807506904006004,
      "rewards-max/forensics_tiou_reward": 0.9057590961456299,
      "rewards/forensics_sinkhorn_reward": 0.7558682858943939,
      "rewards/forensics_strict_edge_reward": 0.033938915468752384,
      "rewards/forensics_tiou_reward": 0.5254481732845306,
      "step": 806
    },
    {
      "completion_length": 22.578125,
      "epoch": 2.4761904761904763,
      "grad_norm": 15.679752349853516,
      "kl": 0.2314453125,
      "learning_rate": 3.7999999999999996e-07,
      "loss": -0.0138,
      "reward": 1.2911050915718079,
      "reward_std": 0.25962021946907043,
      "reward_std_all": 0.24997369945049286,
      "rewards-max/forensics_sinkhorn_reward": 0.9304059743881226,
      "rewards-max/forensics_strict_edge_reward": 0.44296814128756523,
      "rewards-max/forensics_tiou_reward": 0.8667373955249786,
      "rewards/forensics_sinkhorn_reward": 0.7323543131351471,
      "rewards/forensics_strict_edge_reward": 0.03929785266518593,
      "rewards/forensics_tiou_reward": 0.5194529592990875,
      "step": 807
    },
    {
      "completion_length": 22.859375,
      "epoch": 2.47926267281106,
      "grad_norm": 14.478668212890625,
      "kl": 0.43994140625,
      "learning_rate": 3.792307692307692e-07,
      "loss": -0.0024,
      "reward": 1.3508346676826477,
      "reward_std": 0.2645101696252823,
      "reward_std_all": 0.24727731943130493,
      "rewards-max/forensics_sinkhorn_reward": 0.9690653681755066,
      "rewards-max/forensics_strict_edge_reward": 0.3243762031197548,
      "rewards-max/forensics_tiou_reward": 0.9485154747962952,
      "rewards/forensics_sinkhorn_reward": 0.7532071769237518,
      "rewards/forensics_strict_edge_reward": 0.0312563506886363,
      "rewards/forensics_tiou_reward": 0.5663711130619049,
      "step": 808
    },
    {
      "completion_length": 22.59375,
      "epoch": 2.482334869431644,
      "grad_norm": 10.866449356079102,
      "kl": 0.27392578125,
      "learning_rate": 3.7846153846153844e-07,
      "loss": 0.1118,
      "reward": 1.4502019882202148,
      "reward_std": 0.2275846153497696,
      "reward_std_all": 0.20357206463813782,
      "rewards-max/forensics_sinkhorn_reward": 0.9798027276992798,
      "rewards-max/forensics_strict_edge_reward": 0.5396527349948883,
      "rewards-max/forensics_tiou_reward": 0.9759854972362518,
      "rewards/forensics_sinkhorn_reward": 0.7622422575950623,
      "rewards/forensics_strict_edge_reward": 0.06122559681534767,
      "rewards/forensics_tiou_reward": 0.62673419713974,
      "step": 809
    },
    {
      "completion_length": 24.96875,
      "epoch": 2.485407066052227,
      "grad_norm": 15.854036331176758,
      "kl": 0.24609375,
      "learning_rate": 3.776923076923077e-07,
      "loss": 0.0834,
      "reward": 1.3326559662818909,
      "reward_std": 0.3129993826150894,
      "reward_std_all": 0.2709457725286484,
      "rewards-max/forensics_sinkhorn_reward": 0.9631122946739197,
      "rewards-max/forensics_strict_edge_reward": 0.4724728614091873,
      "rewards-max/forensics_tiou_reward": 0.9351387917995453,
      "rewards/forensics_sinkhorn_reward": 0.7555347681045532,
      "rewards/forensics_strict_edge_reward": 0.05364038422703743,
      "rewards/forensics_tiou_reward": 0.5234808474779129,
      "step": 810
    },
    {
      "completion_length": 23.046875,
      "epoch": 2.488479262672811,
      "grad_norm": 12.467412948608398,
      "kl": 0.25,
      "learning_rate": 3.769230769230769e-07,
      "loss": 0.0487,
      "reward": 1.3859661221504211,
      "reward_std": 0.2511444464325905,
      "reward_std_all": 0.23491643369197845,
      "rewards-max/forensics_sinkhorn_reward": 0.9621122777462006,
      "rewards-max/forensics_strict_edge_reward": 0.29376552999019623,
      "rewards-max/forensics_tiou_reward": 0.9224263727664948,
      "rewards/forensics_sinkhorn_reward": 0.7750163972377777,
      "rewards/forensics_strict_edge_reward": 0.033214143477380276,
      "rewards/forensics_tiou_reward": 0.57773557305336,
      "step": 811
    },
    {
      "completion_length": 20.921875,
      "epoch": 2.4915514592933947,
      "grad_norm": 10.557191848754883,
      "kl": 0.29296875,
      "learning_rate": 3.7615384615384614e-07,
      "loss": 0.0393,
      "reward": 1.2858757972717285,
      "reward_std": 0.234201580286026,
      "reward_std_all": 0.26550451666116714,
      "rewards-max/forensics_sinkhorn_reward": 0.9643381536006927,
      "rewards-max/forensics_strict_edge_reward": 0.3480503261089325,
      "rewards-max/forensics_tiou_reward": 0.9364834427833557,
      "rewards/forensics_sinkhorn_reward": 0.739255279302597,
      "rewards/forensics_strict_edge_reward": 0.029222752898931503,
      "rewards/forensics_tiou_reward": 0.5173977315425873,
      "step": 812
    },
    {
      "completion_length": 20.96875,
      "epoch": 2.4946236559139785,
      "grad_norm": 8.223738670349121,
      "kl": 0.40185546875,
      "learning_rate": 3.7538461538461535e-07,
      "loss": -0.0886,
      "reward": 1.611822247505188,
      "reward_std": 0.3103214502334595,
      "reward_std_all": 0.2884303033351898,
      "rewards-max/forensics_sinkhorn_reward": 0.9807911217212677,
      "rewards-max/forensics_strict_edge_reward": 0.6001806855201721,
      "rewards-max/forensics_tiou_reward": 0.9806228280067444,
      "rewards/forensics_sinkhorn_reward": 0.8182318806648254,
      "rewards/forensics_strict_edge_reward": 0.10918726027011871,
      "rewards/forensics_tiou_reward": 0.6844031810760498,
      "step": 813
    },
    {
      "completion_length": 25.015625,
      "epoch": 2.4976958525345623,
      "grad_norm": 10.320680618286133,
      "kl": 0.34765625,
      "learning_rate": 3.746153846153846e-07,
      "loss": -0.0023,
      "reward": 1.3967410922050476,
      "reward_std": 0.1990077644586563,
      "reward_std_all": 0.18129755556583405,
      "rewards-max/forensics_sinkhorn_reward": 0.9283011257648468,
      "rewards-max/forensics_strict_edge_reward": 0.21962502971291542,
      "rewards-max/forensics_tiou_reward": 0.8523160815238953,
      "rewards/forensics_sinkhorn_reward": 0.7780065834522247,
      "rewards/forensics_strict_edge_reward": 0.019492370076477528,
      "rewards/forensics_tiou_reward": 0.5992421507835388,
      "step": 814
    },
    {
      "completion_length": 20.015625,
      "epoch": 2.500768049155146,
      "grad_norm": 14.740644454956055,
      "kl": 0.365234375,
      "learning_rate": 3.7384615384615383e-07,
      "loss": 0.1609,
      "reward": 1.2297052145004272,
      "reward_std": 0.21139171719551086,
      "reward_std_all": 0.22270585596561432,
      "rewards-max/forensics_sinkhorn_reward": 0.9665407538414001,
      "rewards-max/forensics_strict_edge_reward": 0.6232984811067581,
      "rewards-max/forensics_tiou_reward": 0.9440245628356934,
      "rewards/forensics_sinkhorn_reward": 0.6923803687095642,
      "rewards/forensics_strict_edge_reward": 0.06597676686942577,
      "rewards/forensics_tiou_reward": 0.4713481515645981,
      "step": 815
    },
    {
      "completion_length": 19.390625,
      "epoch": 2.50384024577573,
      "grad_norm": 14.344622611999512,
      "kl": 0.72607421875,
      "learning_rate": 3.7307692307692304e-07,
      "loss": 0.0221,
      "reward": 1.4876002669334412,
      "reward_std": 0.23439017683267593,
      "reward_std_all": 0.20889262855052948,
      "rewards-max/forensics_sinkhorn_reward": 0.9836641550064087,
      "rewards-max/forensics_strict_edge_reward": 0.6839397251605988,
      "rewards-max/forensics_tiou_reward": 0.9844760894775391,
      "rewards/forensics_sinkhorn_reward": 0.7998906672000885,
      "rewards/forensics_strict_edge_reward": 0.06738024204969406,
      "rewards/forensics_tiou_reward": 0.620329350233078,
      "step": 816
    },
    {
      "completion_length": 21.046875,
      "epoch": 2.506912442396313,
      "grad_norm": 16.63906478881836,
      "kl": 0.26953125,
      "learning_rate": 3.7230769230769226e-07,
      "loss": 0.0723,
      "reward": 1.2865065336227417,
      "reward_std": 0.2080957069993019,
      "reward_std_all": 0.1814602091908455,
      "rewards-max/forensics_sinkhorn_reward": 0.9760933220386505,
      "rewards-max/forensics_strict_edge_reward": 0.34197375178337097,
      "rewards-max/forensics_tiou_reward": 0.9650320708751678,
      "rewards/forensics_sinkhorn_reward": 0.7455405592918396,
      "rewards/forensics_strict_edge_reward": 0.02412697672843933,
      "rewards/forensics_tiou_reward": 0.5168389528989792,
      "step": 817
    },
    {
      "completion_length": 22.421875,
      "epoch": 2.509984639016897,
      "grad_norm": 10.821161270141602,
      "kl": 0.263671875,
      "learning_rate": 3.715384615384615e-07,
      "loss": -0.0151,
      "reward": 1.3000353574752808,
      "reward_std": 0.2123006284236908,
      "reward_std_all": 0.20709021389484406,
      "rewards-max/forensics_sinkhorn_reward": 0.9616250693798065,
      "rewards-max/forensics_strict_edge_reward": 0.38873492181301117,
      "rewards-max/forensics_tiou_reward": 0.9224903881549835,
      "rewards/forensics_sinkhorn_reward": 0.7539881765842438,
      "rewards/forensics_strict_edge_reward": 0.041340577183291316,
      "rewards/forensics_tiou_reward": 0.5047066807746887,
      "step": 818
    },
    {
      "completion_length": 23.015625,
      "epoch": 2.5130568356374807,
      "grad_norm": 11.039938926696777,
      "kl": 0.3203125,
      "learning_rate": 3.707692307692308e-07,
      "loss": 0.0797,
      "reward": 1.4968701601028442,
      "reward_std": 0.20248237997293472,
      "reward_std_all": 0.2070152834057808,
      "rewards-max/forensics_sinkhorn_reward": 0.9749345481395721,
      "rewards-max/forensics_strict_edge_reward": 0.3216277211904526,
      "rewards-max/forensics_tiou_reward": 0.9652225971221924,
      "rewards/forensics_sinkhorn_reward": 0.814423680305481,
      "rewards/forensics_strict_edge_reward": 0.037852345034480095,
      "rewards/forensics_tiou_reward": 0.6445941627025604,
      "step": 819
    },
    {
      "completion_length": 23.484375,
      "epoch": 2.5161290322580645,
      "grad_norm": 9.278578758239746,
      "kl": 0.27001953125,
      "learning_rate": 3.7e-07,
      "loss": 0.0343,
      "reward": 1.485130488872528,
      "reward_std": 0.2021770477294922,
      "reward_std_all": 0.19562505185604095,
      "rewards-max/forensics_sinkhorn_reward": 0.9719536900520325,
      "rewards-max/forensics_strict_edge_reward": 0.5755647718906403,
      "rewards-max/forensics_tiou_reward": 0.9589256942272186,
      "rewards/forensics_sinkhorn_reward": 0.7867908775806427,
      "rewards/forensics_strict_edge_reward": 0.05248953960835934,
      "rewards/forensics_tiou_reward": 0.6458501815795898,
      "step": 820
    },
    {
      "completion_length": 21.21875,
      "epoch": 2.5192012288786483,
      "grad_norm": 11.795641899108887,
      "kl": 0.4306640625,
      "learning_rate": 3.692307692307692e-07,
      "loss": -0.0136,
      "reward": 1.446698248386383,
      "reward_std": 0.2698117196559906,
      "reward_std_all": 0.23776060342788696,
      "rewards-max/forensics_sinkhorn_reward": 0.958881288766861,
      "rewards-max/forensics_strict_edge_reward": 0.44404928386211395,
      "rewards-max/forensics_tiou_reward": 0.9263617396354675,
      "rewards/forensics_sinkhorn_reward": 0.788431316614151,
      "rewards/forensics_strict_edge_reward": 0.0504170311614871,
      "rewards/forensics_tiou_reward": 0.6078499257564545,
      "step": 821
    },
    {
      "completion_length": 22.5625,
      "epoch": 2.522273425499232,
      "grad_norm": 16.700092315673828,
      "kl": 0.4326171875,
      "learning_rate": 3.6846153846153843e-07,
      "loss": 0.1135,
      "reward": 1.4477497935295105,
      "reward_std": 0.21870720386505127,
      "reward_std_all": 0.18187980353832245,
      "rewards-max/forensics_sinkhorn_reward": 0.9796695709228516,
      "rewards-max/forensics_strict_edge_reward": 0.7412298321723938,
      "rewards-max/forensics_tiou_reward": 0.9772332310676575,
      "rewards/forensics_sinkhorn_reward": 0.7955661118030548,
      "rewards/forensics_strict_edge_reward": 0.050220007076859474,
      "rewards/forensics_tiou_reward": 0.6019637286663055,
      "step": 822
    },
    {
      "completion_length": 21.765625,
      "epoch": 2.525345622119816,
      "grad_norm": 11.136881828308105,
      "kl": 0.24267578125,
      "learning_rate": 3.676923076923077e-07,
      "loss": 0.0282,
      "reward": 1.375968337059021,
      "reward_std": 0.20072059333324432,
      "reward_std_all": 0.17514332383871078,
      "rewards-max/forensics_sinkhorn_reward": 0.9499469697475433,
      "rewards-max/forensics_strict_edge_reward": 0.3532397598028183,
      "rewards-max/forensics_tiou_reward": 0.8996310532093048,
      "rewards/forensics_sinkhorn_reward": 0.7774007618427277,
      "rewards/forensics_strict_edge_reward": 0.034256720915436745,
      "rewards/forensics_tiou_reward": 0.5643109083175659,
      "step": 823
    },
    {
      "completion_length": 22.203125,
      "epoch": 2.528417818740399,
      "grad_norm": 10.39769458770752,
      "kl": 0.4287109375,
      "learning_rate": 3.669230769230769e-07,
      "loss": 0.0932,
      "reward": 1.1625877022743225,
      "reward_std": 0.2137155905365944,
      "reward_std_all": 0.2149164080619812,
      "rewards-max/forensics_sinkhorn_reward": 0.9385727941989899,
      "rewards-max/forensics_strict_edge_reward": 0.5251649767160416,
      "rewards-max/forensics_tiou_reward": 0.8841567635536194,
      "rewards/forensics_sinkhorn_reward": 0.6677074134349823,
      "rewards/forensics_strict_edge_reward": 0.04172064736485481,
      "rewards/forensics_tiou_reward": 0.4531596004962921,
      "step": 824
    },
    {
      "completion_length": 23.3125,
      "epoch": 2.531490015360983,
      "grad_norm": 11.011792182922363,
      "kl": 0.322265625,
      "learning_rate": 3.661538461538461e-07,
      "loss": 0.0256,
      "reward": 1.2700690627098083,
      "reward_std": 0.23857365548610687,
      "reward_std_all": 0.23253512382507324,
      "rewards-max/forensics_sinkhorn_reward": 0.9410479366779327,
      "rewards-max/forensics_strict_edge_reward": 0.2059265561401844,
      "rewards-max/forensics_tiou_reward": 0.8810407519340515,
      "rewards/forensics_sinkhorn_reward": 0.7282306253910065,
      "rewards/forensics_strict_edge_reward": 0.02308773878030479,
      "rewards/forensics_tiou_reward": 0.5187506377696991,
      "step": 825
    },
    {
      "completion_length": 22.5625,
      "epoch": 2.5345622119815667,
      "grad_norm": 10.743854522705078,
      "kl": 0.26953125,
      "learning_rate": 3.6538461538461534e-07,
      "loss": 0.1296,
      "reward": 1.2542181611061096,
      "reward_std": 0.20564201474189758,
      "reward_std_all": 0.19719305634498596,
      "rewards-max/forensics_sinkhorn_reward": 0.98228919506073,
      "rewards-max/forensics_strict_edge_reward": 0.6593321859836578,
      "rewards-max/forensics_tiou_reward": 0.9830977022647858,
      "rewards/forensics_sinkhorn_reward": 0.7025550305843353,
      "rewards/forensics_strict_edge_reward": 0.05926799401640892,
      "rewards/forensics_tiou_reward": 0.49239514768123627,
      "step": 826
    },
    {
      "completion_length": 21.703125,
      "epoch": 2.5376344086021505,
      "grad_norm": 8.330087661743164,
      "kl": 0.5185546875,
      "learning_rate": 3.646153846153846e-07,
      "loss": -0.0294,
      "reward": 1.3499549627304077,
      "reward_std": 0.2213415950536728,
      "reward_std_all": 0.2291705086827278,
      "rewards-max/forensics_sinkhorn_reward": 0.9574387669563293,
      "rewards-max/forensics_strict_edge_reward": 0.320647157728672,
      "rewards-max/forensics_tiou_reward": 0.91363325715065,
      "rewards/forensics_sinkhorn_reward": 0.7732534408569336,
      "rewards/forensics_strict_edge_reward": 0.028573296032845974,
      "rewards/forensics_tiou_reward": 0.5481282472610474,
      "step": 827
    },
    {
      "completion_length": 21.5,
      "epoch": 2.5407066052227343,
      "grad_norm": 17.580320358276367,
      "kl": 0.615234375,
      "learning_rate": 3.638461538461538e-07,
      "loss": 0.0858,
      "reward": 1.156257688999176,
      "reward_std": 0.27816492319107056,
      "reward_std_all": 0.23649699240922928,
      "rewards-max/forensics_sinkhorn_reward": 0.9486404061317444,
      "rewards-max/forensics_strict_edge_reward": 0.2731504626572132,
      "rewards-max/forensics_tiou_reward": 0.903345912694931,
      "rewards/forensics_sinkhorn_reward": 0.7053904831409454,
      "rewards/forensics_strict_edge_reward": 0.01894389558583498,
      "rewards/forensics_tiou_reward": 0.43192335963249207,
      "step": 828
    },
    {
      "completion_length": 21.796875,
      "epoch": 2.543778801843318,
      "grad_norm": 10.349560737609863,
      "kl": 0.3974609375,
      "learning_rate": 3.6307692307692303e-07,
      "loss": 0.0356,
      "reward": 1.3110806941986084,
      "reward_std": 0.23966259509325027,
      "reward_std_all": 0.22206756472587585,
      "rewards-max/forensics_sinkhorn_reward": 0.9642205238342285,
      "rewards-max/forensics_strict_edge_reward": 0.474128395318985,
      "rewards-max/forensics_tiou_reward": 0.9365043342113495,
      "rewards/forensics_sinkhorn_reward": 0.7402773201465607,
      "rewards/forensics_strict_edge_reward": 0.03701033256947994,
      "rewards/forensics_tiou_reward": 0.5337931513786316,
      "step": 829
    },
    {
      "completion_length": 22.609375,
      "epoch": 2.546850998463902,
      "grad_norm": 12.331088066101074,
      "kl": 0.25146484375,
      "learning_rate": 3.623076923076923e-07,
      "loss": 0.0392,
      "reward": 1.286211609840393,
      "reward_std": 0.20416603982448578,
      "reward_std_all": 0.20520223677158356,
      "rewards-max/forensics_sinkhorn_reward": 0.9264039099216461,
      "rewards-max/forensics_strict_edge_reward": 0.2951847454532981,
      "rewards-max/forensics_tiou_reward": 0.8478600382804871,
      "rewards/forensics_sinkhorn_reward": 0.7316369414329529,
      "rewards/forensics_strict_edge_reward": 0.03624230483546853,
      "rewards/forensics_tiou_reward": 0.5183323472738266,
      "step": 830
    },
    {
      "completion_length": 22.609375,
      "epoch": 2.549923195084485,
      "grad_norm": 11.664569854736328,
      "kl": 0.23046875,
      "learning_rate": 3.615384615384615e-07,
      "loss": -0.043,
      "reward": 1.2105081379413605,
      "reward_std": 0.26829321682453156,
      "reward_std_all": 0.24004900455474854,
      "rewards-max/forensics_sinkhorn_reward": 0.934448629617691,
      "rewards-max/forensics_strict_edge_reward": 0.52165737003088,
      "rewards-max/forensics_tiou_reward": 0.8738157451152802,
      "rewards/forensics_sinkhorn_reward": 0.6807996034622192,
      "rewards/forensics_strict_edge_reward": 0.05214686878025532,
      "rewards/forensics_tiou_reward": 0.47756174206733704,
      "step": 831
    },
    {
      "completion_length": 22.125,
      "epoch": 2.5529953917050694,
      "grad_norm": 12.992718696594238,
      "kl": 0.43017578125,
      "learning_rate": 3.607692307692308e-07,
      "loss": 0.0233,
      "reward": 1.5961801409721375,
      "reward_std": 0.2926499545574188,
      "reward_std_all": 0.25432708859443665,
      "rewards-max/forensics_sinkhorn_reward": 0.9767538607120514,
      "rewards-max/forensics_strict_edge_reward": 0.5422053933143616,
      "rewards-max/forensics_tiou_reward": 0.9676424264907837,
      "rewards/forensics_sinkhorn_reward": 0.8431607186794281,
      "rewards/forensics_strict_edge_reward": 0.053657191805541515,
      "rewards/forensics_tiou_reward": 0.6993623375892639,
      "step": 832
    },
    {
      "completion_length": 21.921875,
      "epoch": 2.5560675883256527,
      "grad_norm": 15.806733131408691,
      "kl": 0.7275390625,
      "learning_rate": 3.6e-07,
      "loss": 0.011,
      "reward": 1.2826501727104187,
      "reward_std": 0.1670038402080536,
      "reward_std_all": 0.1568947583436966,
      "rewards-max/forensics_sinkhorn_reward": 0.9646613001823425,
      "rewards-max/forensics_strict_edge_reward": 0.3178005963563919,
      "rewards-max/forensics_tiou_reward": 0.9292654991149902,
      "rewards/forensics_sinkhorn_reward": 0.7435906529426575,
      "rewards/forensics_strict_edge_reward": 0.02727564610540867,
      "rewards/forensics_tiou_reward": 0.5117838084697723,
      "step": 833
    },
    {
      "completion_length": 21.015625,
      "epoch": 2.5591397849462365,
      "grad_norm": 12.399951934814453,
      "kl": 0.345703125,
      "learning_rate": 3.592307692307692e-07,
      "loss": 0.0139,
      "reward": 1.3591012954711914,
      "reward_std": 0.19016408920288086,
      "reward_std_all": 0.18678878992795944,
      "rewards-max/forensics_sinkhorn_reward": 0.981543242931366,
      "rewards-max/forensics_strict_edge_reward": 0.27528148889541626,
      "rewards-max/forensics_tiou_reward": 0.9765272438526154,
      "rewards/forensics_sinkhorn_reward": 0.7779442965984344,
      "rewards/forensics_strict_edge_reward": 0.03140035644173622,
      "rewards/forensics_tiou_reward": 0.549756646156311,
      "step": 834
    },
    {
      "completion_length": 20.9375,
      "epoch": 2.5622119815668203,
      "grad_norm": 13.32717227935791,
      "kl": 0.5595703125,
      "learning_rate": 3.584615384615384e-07,
      "loss": 0.1453,
      "reward": 1.2113847136497498,
      "reward_std": 0.1851339116692543,
      "reward_std_all": 0.18757043778896332,
      "rewards-max/forensics_sinkhorn_reward": 0.9578103423118591,
      "rewards-max/forensics_strict_edge_reward": 0.5099192559719086,
      "rewards-max/forensics_tiou_reward": 0.925025999546051,
      "rewards/forensics_sinkhorn_reward": 0.7100747227668762,
      "rewards/forensics_strict_edge_reward": 0.046502978540956974,
      "rewards/forensics_tiou_reward": 0.45480701327323914,
      "step": 835
    },
    {
      "completion_length": 20.828125,
      "epoch": 2.565284178187404,
      "grad_norm": 13.672258377075195,
      "kl": 0.2587890625,
      "learning_rate": 3.576923076923077e-07,
      "loss": 0.0166,
      "reward": 1.120669662952423,
      "reward_std": 0.18298590183258057,
      "reward_std_all": 0.16282694786787033,
      "rewards-max/forensics_sinkhorn_reward": 0.9152350127696991,
      "rewards-max/forensics_strict_edge_reward": 0.2711677588522434,
      "rewards-max/forensics_tiou_reward": 0.7990691363811493,
      "rewards/forensics_sinkhorn_reward": 0.6900840103626251,
      "rewards/forensics_strict_edge_reward": 0.022569107823073864,
      "rewards/forensics_tiou_reward": 0.40801648795604706,
      "step": 836
    },
    {
      "completion_length": 21.125,
      "epoch": 2.568356374807988,
      "grad_norm": 16.901611328125,
      "kl": 0.955078125,
      "learning_rate": 3.569230769230769e-07,
      "loss": 0.1386,
      "reward": 1.3853377103805542,
      "reward_std": 0.353484645485878,
      "reward_std_all": 0.30356447398662567,
      "rewards-max/forensics_sinkhorn_reward": 0.9720926880836487,
      "rewards-max/forensics_strict_edge_reward": 0.46688374876976013,
      "rewards-max/forensics_tiou_reward": 0.9578152298927307,
      "rewards/forensics_sinkhorn_reward": 0.7419376075267792,
      "rewards/forensics_strict_edge_reward": 0.044030994176864624,
      "rewards/forensics_tiou_reward": 0.599369078874588,
      "step": 837
    },
    {
      "completion_length": 21.765625,
      "epoch": 2.571428571428571,
      "grad_norm": 10.609417915344238,
      "kl": 0.44140625,
      "learning_rate": 3.561538461538461e-07,
      "loss": 0.0922,
      "reward": 1.3916453123092651,
      "reward_std": 0.22075695544481277,
      "reward_std_all": 0.19723086059093475,
      "rewards-max/forensics_sinkhorn_reward": 0.9698594212532043,
      "rewards-max/forensics_strict_edge_reward": 0.29105816036462784,
      "rewards-max/forensics_tiou_reward": 0.9471550583839417,
      "rewards/forensics_sinkhorn_reward": 0.779748260974884,
      "rewards/forensics_strict_edge_reward": 0.0329593401402235,
      "rewards/forensics_tiou_reward": 0.5789376497268677,
      "step": 838
    },
    {
      "completion_length": 23.453125,
      "epoch": 2.5745007680491554,
      "grad_norm": 17.301008224487305,
      "kl": 0.2744140625,
      "learning_rate": 3.553846153846154e-07,
      "loss": -0.0281,
      "reward": 0.9773033857345581,
      "reward_std": 0.18371398746967316,
      "reward_std_all": 0.18691638857126236,
      "rewards-max/forensics_sinkhorn_reward": 0.9474270045757294,
      "rewards-max/forensics_strict_edge_reward": 0.07913222908973694,
      "rewards-max/forensics_tiou_reward": 0.8872522711753845,
      "rewards/forensics_sinkhorn_reward": 0.6133217513561249,
      "rewards/forensics_strict_edge_reward": 0.0056845201179385185,
      "rewards/forensics_tiou_reward": 0.3582971394062042,
      "step": 839
    },
    {
      "completion_length": 21.3125,
      "epoch": 2.5775729646697387,
      "grad_norm": 15.965240478515625,
      "kl": 0.1953125,
      "learning_rate": 3.546153846153846e-07,
      "loss": 0.2141,
      "reward": 1.468152642250061,
      "reward_std": 0.21799743175506592,
      "reward_std_all": 0.2060217633843422,
      "rewards-max/forensics_sinkhorn_reward": 0.9836636483669281,
      "rewards-max/forensics_strict_edge_reward": 0.7875787317752838,
      "rewards-max/forensics_tiou_reward": 0.9860922992229462,
      "rewards/forensics_sinkhorn_reward": 0.8000130355358124,
      "rewards/forensics_strict_edge_reward": 0.047912100329995155,
      "rewards/forensics_tiou_reward": 0.6202275454998016,
      "step": 840
    },
    {
      "completion_length": 21.890625,
      "epoch": 2.5806451612903225,
      "grad_norm": 11.584981918334961,
      "kl": 0.611328125,
      "learning_rate": 3.5384615384615386e-07,
      "loss": 0.0588,
      "reward": 1.3946582674980164,
      "reward_std": 0.26013362407684326,
      "reward_std_all": 0.22845450043678284,
      "rewards-max/forensics_sinkhorn_reward": 0.9621565043926239,
      "rewards-max/forensics_strict_edge_reward": 0.3137419670820236,
      "rewards-max/forensics_tiou_reward": 0.9295792281627655,
      "rewards/forensics_sinkhorn_reward": 0.762425422668457,
      "rewards/forensics_strict_edge_reward": 0.04980576038360596,
      "rewards/forensics_tiou_reward": 0.5824271142482758,
      "step": 841
    },
    {
      "completion_length": 20.15625,
      "epoch": 2.5837173579109063,
      "grad_norm": 12.017770767211914,
      "kl": 0.595703125,
      "learning_rate": 3.530769230769231e-07,
      "loss": 0.1241,
      "reward": 1.3017261624336243,
      "reward_std": 0.25045353174209595,
      "reward_std_all": 0.23477132618427277,
      "rewards-max/forensics_sinkhorn_reward": 0.9755734801292419,
      "rewards-max/forensics_strict_edge_reward": 0.29823410511016846,
      "rewards-max/forensics_tiou_reward": 0.9628115296363831,
      "rewards/forensics_sinkhorn_reward": 0.7567379474639893,
      "rewards/forensics_strict_edge_reward": 0.0268370034173131,
      "rewards/forensics_tiou_reward": 0.5181512236595154,
      "step": 842
    },
    {
      "completion_length": 23.265625,
      "epoch": 2.58678955453149,
      "grad_norm": 9.639327049255371,
      "kl": 0.453125,
      "learning_rate": 3.523076923076923e-07,
      "loss": 0.0383,
      "reward": 1.30471932888031,
      "reward_std": 0.3060033619403839,
      "reward_std_all": 0.28344595432281494,
      "rewards-max/forensics_sinkhorn_reward": 0.9409059584140778,
      "rewards-max/forensics_strict_edge_reward": 0.615896612405777,
      "rewards-max/forensics_tiou_reward": 0.8883415758609772,
      "rewards/forensics_sinkhorn_reward": 0.7241007685661316,
      "rewards/forensics_strict_edge_reward": 0.06044421344995499,
      "rewards/forensics_tiou_reward": 0.5201742649078369,
      "step": 843
    },
    {
      "completion_length": 22.03125,
      "epoch": 2.589861751152074,
      "grad_norm": 11.366477966308594,
      "kl": 0.3369140625,
      "learning_rate": 3.515384615384615e-07,
      "loss": 0.1591,
      "reward": 1.4095312356948853,
      "reward_std": 0.28268104791641235,
      "reward_std_all": 0.2917833924293518,
      "rewards-max/forensics_sinkhorn_reward": 0.9847189486026764,
      "rewards-max/forensics_strict_edge_reward": 0.4506318047642708,
      "rewards-max/forensics_tiou_reward": 0.9872523546218872,
      "rewards/forensics_sinkhorn_reward": 0.7764540910720825,
      "rewards/forensics_strict_edge_reward": 0.05288800969719887,
      "rewards/forensics_tiou_reward": 0.5801891088485718,
      "step": 844
    },
    {
      "completion_length": 20.625,
      "epoch": 2.592933947772657,
      "grad_norm": 16.619211196899414,
      "kl": 0.3984375,
      "learning_rate": 3.5076923076923077e-07,
      "loss": -0.0493,
      "reward": 1.1421822905540466,
      "reward_std": 0.21726644784212112,
      "reward_std_all": 0.22649198025465012,
      "rewards-max/forensics_sinkhorn_reward": 0.9748388230800629,
      "rewards-max/forensics_strict_edge_reward": 0.5548860281705856,
      "rewards-max/forensics_tiou_reward": 0.9638355672359467,
      "rewards/forensics_sinkhorn_reward": 0.6661737263202667,
      "rewards/forensics_strict_edge_reward": 0.05740141309797764,
      "rewards/forensics_tiou_reward": 0.4186071306467056,
      "step": 845
    },
    {
      "completion_length": 24.40625,
      "epoch": 2.5960061443932414,
      "grad_norm": 9.427268028259277,
      "kl": 0.3525390625,
      "learning_rate": 3.5e-07,
      "loss": 0.0047,
      "reward": 1.5063959956169128,
      "reward_std": 0.20686906576156616,
      "reward_std_all": 0.1960577592253685,
      "rewards-max/forensics_sinkhorn_reward": 0.9679119288921356,
      "rewards-max/forensics_strict_edge_reward": 0.4150864779949188,
      "rewards-max/forensics_tiou_reward": 0.948811799287796,
      "rewards/forensics_sinkhorn_reward": 0.8020090460777283,
      "rewards/forensics_strict_edge_reward": 0.06926831230521202,
      "rewards/forensics_tiou_reward": 0.6351186037063599,
      "step": 846
    },
    {
      "completion_length": 21.0,
      "epoch": 2.5990783410138247,
      "grad_norm": 10.939075469970703,
      "kl": 0.3837890625,
      "learning_rate": 3.492307692307692e-07,
      "loss": 0.0448,
      "reward": 1.4501755833625793,
      "reward_std": 0.24802161008119583,
      "reward_std_all": 0.2144923061132431,
      "rewards-max/forensics_sinkhorn_reward": 0.979484885931015,
      "rewards-max/forensics_strict_edge_reward": 0.5749746412038803,
      "rewards-max/forensics_tiou_reward": 0.9732734858989716,
      "rewards/forensics_sinkhorn_reward": 0.7723811268806458,
      "rewards/forensics_strict_edge_reward": 0.08060220628976822,
      "rewards/forensics_tiou_reward": 0.5971923172473907,
      "step": 847
    },
    {
      "completion_length": 22.1875,
      "epoch": 2.6021505376344085,
      "grad_norm": 11.025545120239258,
      "kl": 0.2353515625,
      "learning_rate": 3.4846153846153846e-07,
      "loss": -0.0537,
      "reward": 1.2288850545883179,
      "reward_std": 0.20371581614017487,
      "reward_std_all": 0.18446564674377441,
      "rewards-max/forensics_sinkhorn_reward": 0.9301648139953613,
      "rewards-max/forensics_strict_edge_reward": 0.5054102391004562,
      "rewards-max/forensics_tiou_reward": 0.8606403470039368,
      "rewards/forensics_sinkhorn_reward": 0.7146922647953033,
      "rewards/forensics_strict_edge_reward": 0.05305819120258093,
      "rewards/forensics_tiou_reward": 0.46113458275794983,
      "step": 848
    },
    {
      "completion_length": 21.203125,
      "epoch": 2.6052227342549923,
      "grad_norm": 13.372859954833984,
      "kl": 0.4541015625,
      "learning_rate": 3.476923076923077e-07,
      "loss": 0.0661,
      "reward": 1.500036895275116,
      "reward_std": 0.19847969710826874,
      "reward_std_all": 0.178087517619133,
      "rewards-max/forensics_sinkhorn_reward": 0.9788085222244263,
      "rewards-max/forensics_strict_edge_reward": 0.20148923993110657,
      "rewards-max/forensics_tiou_reward": 0.9737923741340637,
      "rewards/forensics_sinkhorn_reward": 0.8214394152164459,
      "rewards/forensics_strict_edge_reward": 0.023014957085251808,
      "rewards/forensics_tiou_reward": 0.6555824875831604,
      "step": 849
    },
    {
      "completion_length": 20.28125,
      "epoch": 2.608294930875576,
      "grad_norm": 14.06412410736084,
      "kl": 0.326171875,
      "learning_rate": 3.4692307692307694e-07,
      "loss": -0.031,
      "reward": 1.6857468485832214,
      "reward_std": 0.38471001386642456,
      "reward_std_all": 0.3333579748868942,
      "rewards-max/forensics_sinkhorn_reward": 0.983651876449585,
      "rewards-max/forensics_strict_edge_reward": 0.5564709603786469,
      "rewards-max/forensics_tiou_reward": 0.9846810400485992,
      "rewards/forensics_sinkhorn_reward": 0.8541283905506134,
      "rewards/forensics_strict_edge_reward": 0.12445631623268127,
      "rewards/forensics_tiou_reward": 0.7071621119976044,
      "step": 850
    },
    {
      "completion_length": 20.140625,
      "epoch": 2.61136712749616,
      "grad_norm": 10.277031898498535,
      "kl": 0.41015625,
      "learning_rate": 3.461538461538461e-07,
      "loss": -0.0441,
      "reward": 1.4503998756408691,
      "reward_std": 0.21202902495861053,
      "reward_std_all": 0.19330227375030518,
      "rewards-max/forensics_sinkhorn_reward": 0.9764211475849152,
      "rewards-max/forensics_strict_edge_reward": 0.5702531039714813,
      "rewards-max/forensics_tiou_reward": 0.9665713012218475,
      "rewards/forensics_sinkhorn_reward": 0.7761455178260803,
      "rewards/forensics_strict_edge_reward": 0.06349664181470871,
      "rewards/forensics_tiou_reward": 0.61075758934021,
      "step": 851
    },
    {
      "completion_length": 22.53125,
      "epoch": 2.6144393241167436,
      "grad_norm": 10.662809371948242,
      "kl": 0.3759765625,
      "learning_rate": 3.4538461538461537e-07,
      "loss": 0.0102,
      "reward": 1.349230945110321,
      "reward_std": 0.26080890744924545,
      "reward_std_all": 0.24250617623329163,
      "rewards-max/forensics_sinkhorn_reward": 0.9796046018600464,
      "rewards-max/forensics_strict_edge_reward": 0.5339391529560089,
      "rewards-max/forensics_tiou_reward": 0.9770475625991821,
      "rewards/forensics_sinkhorn_reward": 0.7391420006752014,
      "rewards/forensics_strict_edge_reward": 0.060120828449726105,
      "rewards/forensics_tiou_reward": 0.5499680042266846,
      "step": 852
    },
    {
      "completion_length": 21.078125,
      "epoch": 2.6175115207373274,
      "grad_norm": 11.076810836791992,
      "kl": 0.302734375,
      "learning_rate": 3.446153846153846e-07,
      "loss": -0.0162,
      "reward": 1.315802812576294,
      "reward_std": 0.3010356277227402,
      "reward_std_all": 0.2803306132555008,
      "rewards-max/forensics_sinkhorn_reward": 0.9748817980289459,
      "rewards-max/forensics_strict_edge_reward": 0.37403713166713715,
      "rewards-max/forensics_tiou_reward": 0.9621632695198059,
      "rewards/forensics_sinkhorn_reward": 0.7493383288383484,
      "rewards/forensics_strict_edge_reward": 0.027453158050775528,
      "rewards/forensics_tiou_reward": 0.5390112698078156,
      "step": 853
    },
    {
      "completion_length": 23.984375,
      "epoch": 2.6205837173579107,
      "grad_norm": 12.7742338180542,
      "kl": 0.4912109375,
      "learning_rate": 3.4384615384615385e-07,
      "loss": 0.0547,
      "reward": 1.464029610157013,
      "reward_std": 0.3103485554456711,
      "reward_std_all": 0.33020298182964325,
      "rewards-max/forensics_sinkhorn_reward": 0.9724812805652618,
      "rewards-max/forensics_strict_edge_reward": 0.36426396667957306,
      "rewards-max/forensics_tiou_reward": 0.956336259841919,
      "rewards/forensics_sinkhorn_reward": 0.7841375470161438,
      "rewards/forensics_strict_edge_reward": 0.07104887440800667,
      "rewards/forensics_tiou_reward": 0.6088431477546692,
      "step": 854
    },
    {
      "completion_length": 19.296875,
      "epoch": 2.6236559139784945,
      "grad_norm": 15.828043937683105,
      "kl": 0.46875,
      "learning_rate": 3.4307692307692307e-07,
      "loss": 0.0262,
      "reward": 1.4367947578430176,
      "reward_std": 0.31287163496017456,
      "reward_std_all": 0.2802603989839554,
      "rewards-max/forensics_sinkhorn_reward": 0.9766258597373962,
      "rewards-max/forensics_strict_edge_reward": 0.6837711930274963,
      "rewards-max/forensics_tiou_reward": 0.9700659215450287,
      "rewards/forensics_sinkhorn_reward": 0.7826957106590271,
      "rewards/forensics_strict_edge_reward": 0.07025955989956856,
      "rewards/forensics_tiou_reward": 0.5838395059108734,
      "step": 855
    },
    {
      "completion_length": 21.75,
      "epoch": 2.6267281105990783,
      "grad_norm": 10.943902969360352,
      "kl": 0.2705078125,
      "learning_rate": 3.423076923076923e-07,
      "loss": 0.1094,
      "reward": 1.2276135087013245,
      "reward_std": 0.23020268976688385,
      "reward_std_all": 0.2021089568734169,
      "rewards-max/forensics_sinkhorn_reward": 0.9204473495483398,
      "rewards-max/forensics_strict_edge_reward": 0.11628350056707859,
      "rewards-max/forensics_tiou_reward": 0.8243480324745178,
      "rewards/forensics_sinkhorn_reward": 0.7185506522655487,
      "rewards/forensics_strict_edge_reward": 0.010965477442368865,
      "rewards/forensics_tiou_reward": 0.49809738993644714,
      "step": 856
    },
    {
      "completion_length": 22.3125,
      "epoch": 2.629800307219662,
      "grad_norm": 10.88845443725586,
      "kl": 0.4140625,
      "learning_rate": 3.415384615384615e-07,
      "loss": 0.0671,
      "reward": 1.419040322303772,
      "reward_std": 0.2817063182592392,
      "reward_std_all": 0.2634151428937912,
      "rewards-max/forensics_sinkhorn_reward": 0.9688800275325775,
      "rewards-max/forensics_strict_edge_reward": 0.4866199493408203,
      "rewards-max/forensics_tiou_reward": 0.952681839466095,
      "rewards/forensics_sinkhorn_reward": 0.784953385591507,
      "rewards/forensics_strict_edge_reward": 0.057142920792102814,
      "rewards/forensics_tiou_reward": 0.5769439935684204,
      "step": 857
    },
    {
      "completion_length": 22.03125,
      "epoch": 2.632872503840246,
      "grad_norm": 14.52580451965332,
      "kl": 0.3115234375,
      "learning_rate": 3.4076923076923076e-07,
      "loss": 0.0215,
      "reward": 1.1467392444610596,
      "reward_std": 0.24687347561120987,
      "reward_std_all": 0.2255857065320015,
      "rewards-max/forensics_sinkhorn_reward": 0.9288860559463501,
      "rewards-max/forensics_strict_edge_reward": 0.4602060317993164,
      "rewards-max/forensics_tiou_reward": 0.8626261055469513,
      "rewards/forensics_sinkhorn_reward": 0.6699997186660767,
      "rewards/forensics_strict_edge_reward": 0.034092748537659645,
      "rewards/forensics_tiou_reward": 0.442646786570549,
      "step": 858
    },
    {
      "completion_length": 21.859375,
      "epoch": 2.6359447004608296,
      "grad_norm": 41.57416534423828,
      "kl": 2.359375,
      "learning_rate": 3.4000000000000003e-07,
      "loss": 0.1303,
      "reward": 1.2940523028373718,
      "reward_std": 0.3752932846546173,
      "reward_std_all": 0.33738239109516144,
      "rewards-max/forensics_sinkhorn_reward": 0.9794795215129852,
      "rewards-max/forensics_strict_edge_reward": 0.7412298321723938,
      "rewards-max/forensics_tiou_reward": 0.9772390425205231,
      "rewards/forensics_sinkhorn_reward": 0.7203270494937897,
      "rewards/forensics_strict_edge_reward": 0.0733078233897686,
      "rewards/forensics_tiou_reward": 0.5004174113273621,
      "step": 859
    },
    {
      "completion_length": 20.46875,
      "epoch": 2.6390168970814134,
      "grad_norm": 10.524837493896484,
      "kl": 0.28271484375,
      "learning_rate": 3.392307692307692e-07,
      "loss": -0.0121,
      "reward": 1.4093321561813354,
      "reward_std": 0.2078278362751007,
      "reward_std_all": 0.19637132436037064,
      "rewards-max/forensics_sinkhorn_reward": 0.9693898856639862,
      "rewards-max/forensics_strict_edge_reward": 0.43902117013931274,
      "rewards-max/forensics_tiou_reward": 0.9410974383354187,
      "rewards/forensics_sinkhorn_reward": 0.7753261923789978,
      "rewards/forensics_strict_edge_reward": 0.034978076815605164,
      "rewards/forensics_tiou_reward": 0.5990278422832489,
      "step": 860
    },
    {
      "completion_length": 21.5625,
      "epoch": 2.6420890937019967,
      "grad_norm": 10.61242389678955,
      "kl": 0.341796875,
      "learning_rate": 3.3846153846153845e-07,
      "loss": 0.1388,
      "reward": 1.4140268564224243,
      "reward_std": 0.2683066576719284,
      "reward_std_all": 0.2741682082414627,
      "rewards-max/forensics_sinkhorn_reward": 0.9526523947715759,
      "rewards-max/forensics_strict_edge_reward": 0.12306921929121017,
      "rewards-max/forensics_tiou_reward": 0.911179929971695,
      "rewards/forensics_sinkhorn_reward": 0.7861507833003998,
      "rewards/forensics_strict_edge_reward": 0.01922709960490465,
      "rewards/forensics_tiou_reward": 0.6086490154266357,
      "step": 861
    },
    {
      "completion_length": 20.296875,
      "epoch": 2.6451612903225805,
      "grad_norm": 13.599178314208984,
      "kl": 0.267578125,
      "learning_rate": 3.3769230769230767e-07,
      "loss": -0.0153,
      "reward": 1.2485336065292358,
      "reward_std": 0.2800626754760742,
      "reward_std_all": 0.2572662979364395,
      "rewards-max/forensics_sinkhorn_reward": 0.9442660510540009,
      "rewards-max/forensics_strict_edge_reward": 0.6492071002721786,
      "rewards-max/forensics_tiou_reward": 0.8939968645572662,
      "rewards/forensics_sinkhorn_reward": 0.6946078836917877,
      "rewards/forensics_strict_edge_reward": 0.07367649301886559,
      "rewards/forensics_tiou_reward": 0.48024922609329224,
      "step": 862
    },
    {
      "completion_length": 22.03125,
      "epoch": 2.6482334869431643,
      "grad_norm": 8.343388557434082,
      "kl": 0.28125,
      "learning_rate": 3.3692307692307693e-07,
      "loss": 0.0948,
      "reward": 1.4192545413970947,
      "reward_std": 0.27863937616348267,
      "reward_std_all": 0.25950857996940613,
      "rewards-max/forensics_sinkhorn_reward": 0.9845030307769775,
      "rewards-max/forensics_strict_edge_reward": 0.7797744870185852,
      "rewards-max/forensics_tiou_reward": 0.989090085029602,
      "rewards/forensics_sinkhorn_reward": 0.7514547109603882,
      "rewards/forensics_strict_edge_reward": 0.06101346202194691,
      "rewards/forensics_tiou_reward": 0.6067863404750824,
      "step": 863
    },
    {
      "completion_length": 23.421875,
      "epoch": 2.651305683563748,
      "grad_norm": 13.843122482299805,
      "kl": 0.30517578125,
      "learning_rate": 3.361538461538461e-07,
      "loss": -0.0187,
      "reward": 1.5596026182174683,
      "reward_std": 0.2187962755560875,
      "reward_std_all": 0.1925378292798996,
      "rewards-max/forensics_sinkhorn_reward": 0.9854446053504944,
      "rewards-max/forensics_strict_edge_reward": 0.5415481477975845,
      "rewards-max/forensics_tiou_reward": 0.9908658862113953,
      "rewards/forensics_sinkhorn_reward": 0.8269389271736145,
      "rewards/forensics_strict_edge_reward": 0.059267327189445496,
      "rewards/forensics_tiou_reward": 0.6733963787555695,
      "step": 864
    },
    {
      "completion_length": 22.125,
      "epoch": 2.654377880184332,
      "grad_norm": 15.386033058166504,
      "kl": 0.3505859375,
      "learning_rate": 3.3538461538461536e-07,
      "loss": 0.0298,
      "reward": 1.2626963257789612,
      "reward_std": 0.1987934149801731,
      "reward_std_all": 0.208004679530859,
      "rewards-max/forensics_sinkhorn_reward": 0.9807830452919006,
      "rewards-max/forensics_strict_edge_reward": 0.3993081673979759,
      "rewards-max/forensics_tiou_reward": 0.9782377183437347,
      "rewards/forensics_sinkhorn_reward": 0.7236402332782745,
      "rewards/forensics_strict_edge_reward": 0.03460444835945964,
      "rewards/forensics_tiou_reward": 0.5044516324996948,
      "step": 865
    },
    {
      "completion_length": 23.515625,
      "epoch": 2.6574500768049156,
      "grad_norm": 15.144036293029785,
      "kl": 0.47265625,
      "learning_rate": 3.346153846153846e-07,
      "loss": -0.0125,
      "reward": 1.5024925470352173,
      "reward_std": 0.31134211272001266,
      "reward_std_all": 0.28543321788311005,
      "rewards-max/forensics_sinkhorn_reward": 0.9881668090820312,
      "rewards-max/forensics_strict_edge_reward": 0.824172168970108,
      "rewards-max/forensics_tiou_reward": 0.996953696012497,
      "rewards/forensics_sinkhorn_reward": 0.7931247353553772,
      "rewards/forensics_strict_edge_reward": 0.06951707229018211,
      "rewards/forensics_tiou_reward": 0.6398507952690125,
      "step": 866
    },
    {
      "completion_length": 21.21875,
      "epoch": 2.6605222734254994,
      "grad_norm": 13.622079849243164,
      "kl": 0.20556640625,
      "learning_rate": 3.3384615384615384e-07,
      "loss": 0.0446,
      "reward": 1.314159631729126,
      "reward_std": 0.2371494099497795,
      "reward_std_all": 0.2169865220785141,
      "rewards-max/forensics_sinkhorn_reward": 0.9441811740398407,
      "rewards-max/forensics_strict_edge_reward": 0.6049744635820389,
      "rewards-max/forensics_tiou_reward": 0.8791965842247009,
      "rewards/forensics_sinkhorn_reward": 0.7158629894256592,
      "rewards/forensics_strict_edge_reward": 0.0729371327906847,
      "rewards/forensics_tiou_reward": 0.5253595411777496,
      "step": 867
    },
    {
      "completion_length": 22.171875,
      "epoch": 2.6635944700460827,
      "grad_norm": 11.000904083251953,
      "kl": 0.513671875,
      "learning_rate": 3.330769230769231e-07,
      "loss": 0.0364,
      "reward": 1.3397390246391296,
      "reward_std": 0.1770365759730339,
      "reward_std_all": 0.17678267508745193,
      "rewards-max/forensics_sinkhorn_reward": 0.9777423143386841,
      "rewards-max/forensics_strict_edge_reward": 0.5044723898172379,
      "rewards-max/forensics_tiou_reward": 0.9634112417697906,
      "rewards/forensics_sinkhorn_reward": 0.7173604369163513,
      "rewards/forensics_strict_edge_reward": 0.061872225254774094,
      "rewards/forensics_tiou_reward": 0.5605064034461975,
      "step": 868
    },
    {
      "completion_length": 22.28125,
      "epoch": 2.6666666666666665,
      "grad_norm": 10.37425422668457,
      "kl": 0.3525390625,
      "learning_rate": 3.3230769230769227e-07,
      "loss": -0.0206,
      "reward": 1.4777119159698486,
      "reward_std": 0.17274516075849533,
      "reward_std_all": 0.1941073164343834,
      "rewards-max/forensics_sinkhorn_reward": 0.9761733710765839,
      "rewards-max/forensics_strict_edge_reward": 0.2727290540933609,
      "rewards-max/forensics_tiou_reward": 0.9630368053913116,
      "rewards/forensics_sinkhorn_reward": 0.8026564121246338,
      "rewards/forensics_strict_edge_reward": 0.03736655507236719,
      "rewards/forensics_tiou_reward": 0.6376890242099762,
      "step": 869
    },
    {
      "completion_length": 21.96875,
      "epoch": 2.6697388632872503,
      "grad_norm": 10.752123832702637,
      "kl": 0.22998046875,
      "learning_rate": 3.3153846153846154e-07,
      "loss": -0.0721,
      "reward": 1.5402384996414185,
      "reward_std": 0.18975521624088287,
      "reward_std_all": 0.17994992434978485,
      "rewards-max/forensics_sinkhorn_reward": 0.9742807149887085,
      "rewards-max/forensics_strict_edge_reward": 0.347290962934494,
      "rewards-max/forensics_tiou_reward": 0.955068975687027,
      "rewards/forensics_sinkhorn_reward": 0.8291219472885132,
      "rewards/forensics_strict_edge_reward": 0.04495329037308693,
      "rewards/forensics_tiou_reward": 0.6661633253097534,
      "step": 870
    },
    {
      "completion_length": 24.078125,
      "epoch": 2.672811059907834,
      "grad_norm": 11.990580558776855,
      "kl": 0.349609375,
      "learning_rate": 3.3076923076923075e-07,
      "loss": -0.1009,
      "reward": 1.198646605014801,
      "reward_std": 0.3131175711750984,
      "reward_std_all": 0.27392031997442245,
      "rewards-max/forensics_sinkhorn_reward": 0.9814333021640778,
      "rewards-max/forensics_strict_edge_reward": 0.732279509305954,
      "rewards-max/forensics_tiou_reward": 0.9795618951320648,
      "rewards/forensics_sinkhorn_reward": 0.6580050587654114,
      "rewards/forensics_strict_edge_reward": 0.07456820458173752,
      "rewards/forensics_tiou_reward": 0.46607333421707153,
      "step": 871
    },
    {
      "completion_length": 19.578125,
      "epoch": 2.675883256528418,
      "grad_norm": 11.066929817199707,
      "kl": 0.2294921875,
      "learning_rate": 3.3e-07,
      "loss": -0.0165,
      "reward": 1.3213805556297302,
      "reward_std": 0.16651419177651405,
      "reward_std_all": 0.14559945464134216,
      "rewards-max/forensics_sinkhorn_reward": 0.9838558733463287,
      "rewards-max/forensics_strict_edge_reward": 0.31986102275550365,
      "rewards-max/forensics_tiou_reward": 0.9848543405532837,
      "rewards/forensics_sinkhorn_reward": 0.7278532683849335,
      "rewards/forensics_strict_edge_reward": 0.07192021305672824,
      "rewards/forensics_tiou_reward": 0.521607056260109,
      "step": 872
    },
    {
      "completion_length": 22.15625,
      "epoch": 2.6789554531490016,
      "grad_norm": 9.800715446472168,
      "kl": 0.314453125,
      "learning_rate": 3.292307692307692e-07,
      "loss": 0.0587,
      "reward": 1.5730538964271545,
      "reward_std": 0.29240454733371735,
      "reward_std_all": 0.25276947021484375,
      "rewards-max/forensics_sinkhorn_reward": 0.9830659925937653,
      "rewards-max/forensics_strict_edge_reward": 0.4856214374303818,
      "rewards-max/forensics_tiou_reward": 0.986111581325531,
      "rewards/forensics_sinkhorn_reward": 0.8020954728126526,
      "rewards/forensics_strict_edge_reward": 0.08302382379770279,
      "rewards/forensics_tiou_reward": 0.6879346072673798,
      "step": 873
    },
    {
      "completion_length": 22.421875,
      "epoch": 2.6820276497695854,
      "grad_norm": 14.609977722167969,
      "kl": 0.3271484375,
      "learning_rate": 3.2846153846153844e-07,
      "loss": 0.0652,
      "reward": 1.3015085458755493,
      "reward_std": 0.19701005518436432,
      "reward_std_all": 0.23386436700820923,
      "rewards-max/forensics_sinkhorn_reward": 0.9442574083805084,
      "rewards-max/forensics_strict_edge_reward": 0.2995662987232208,
      "rewards-max/forensics_tiou_reward": 0.8932165205478668,
      "rewards/forensics_sinkhorn_reward": 0.7396965324878693,
      "rewards/forensics_strict_edge_reward": 0.03176676109433174,
      "rewards/forensics_tiou_reward": 0.5300453007221222,
      "step": 874
    },
    {
      "completion_length": 21.3125,
      "epoch": 2.6850998463901687,
      "grad_norm": 13.38276481628418,
      "kl": 0.2392578125,
      "learning_rate": 3.2769230769230766e-07,
      "loss": -0.0656,
      "reward": 1.4164384603500366,
      "reward_std": 0.2004793956875801,
      "reward_std_all": 0.17725419253110886,
      "rewards-max/forensics_sinkhorn_reward": 0.973438173532486,
      "rewards-max/forensics_strict_edge_reward": 0.556451678276062,
      "rewards-max/forensics_tiou_reward": 0.959226131439209,
      "rewards/forensics_sinkhorn_reward": 0.769395262002945,
      "rewards/forensics_strict_edge_reward": 0.0659263152629137,
      "rewards/forensics_tiou_reward": 0.5811168253421783,
      "step": 875
    },
    {
      "completion_length": 21.578125,
      "epoch": 2.688172043010753,
      "grad_norm": 11.545065879821777,
      "kl": 0.2685546875,
      "learning_rate": 3.269230769230769e-07,
      "loss": 0.1446,
      "reward": 1.3093574047088623,
      "reward_std": 0.19741570949554443,
      "reward_std_all": 0.1916269287467003,
      "rewards-max/forensics_sinkhorn_reward": 0.9711076617240906,
      "rewards-max/forensics_strict_edge_reward": 0.6482091546058655,
      "rewards-max/forensics_tiou_reward": 0.941326230764389,
      "rewards/forensics_sinkhorn_reward": 0.7344608306884766,
      "rewards/forensics_strict_edge_reward": 0.03149254713207483,
      "rewards/forensics_tiou_reward": 0.5434040725231171,
      "step": 876
    },
    {
      "completion_length": 22.015625,
      "epoch": 2.6912442396313363,
      "grad_norm": 12.918631553649902,
      "kl": 0.23876953125,
      "learning_rate": 3.2615384615384614e-07,
      "loss": -0.0538,
      "reward": 1.4596696496009827,
      "reward_std": 0.23278721421957016,
      "reward_std_all": 0.21412087976932526,
      "rewards-max/forensics_sinkhorn_reward": 0.9731228351593018,
      "rewards-max/forensics_strict_edge_reward": 0.3072382062673569,
      "rewards-max/forensics_tiou_reward": 0.9513894617557526,
      "rewards/forensics_sinkhorn_reward": 0.7921600043773651,
      "rewards/forensics_strict_edge_reward": 0.039640747010707855,
      "rewards/forensics_tiou_reward": 0.6278688907623291,
      "step": 877
    },
    {
      "completion_length": 21.171875,
      "epoch": 2.69431643625192,
      "grad_norm": 16.218324661254883,
      "kl": 0.337890625,
      "learning_rate": 3.2538461538461535e-07,
      "loss": 0.0691,
      "reward": 1.4147975444793701,
      "reward_std": 0.18953366577625275,
      "reward_std_all": 0.17510897666215897,
      "rewards-max/forensics_sinkhorn_reward": 0.9829678237438202,
      "rewards-max/forensics_strict_edge_reward": 0.5343473553657532,
      "rewards-max/forensics_tiou_reward": 0.9812397658824921,
      "rewards/forensics_sinkhorn_reward": 0.7755277156829834,
      "rewards/forensics_strict_edge_reward": 0.05030125193297863,
      "rewards/forensics_tiou_reward": 0.5889685600996017,
      "step": 878
    },
    {
      "completion_length": 21.765625,
      "epoch": 2.697388632872504,
      "grad_norm": 12.649942398071289,
      "kl": 0.345703125,
      "learning_rate": 3.246153846153846e-07,
      "loss": 0.0588,
      "reward": 1.4406041502952576,
      "reward_std": 0.35512423515319824,
      "reward_std_all": 0.3198421448469162,
      "rewards-max/forensics_sinkhorn_reward": 0.9796486496925354,
      "rewards-max/forensics_strict_edge_reward": 0.6009648144245148,
      "rewards-max/forensics_tiou_reward": 0.9763875603675842,
      "rewards/forensics_sinkhorn_reward": 0.7654176950454712,
      "rewards/forensics_strict_edge_reward": 0.06339138559997082,
      "rewards/forensics_tiou_reward": 0.6117951571941376,
      "step": 879
    },
    {
      "completion_length": 24.375,
      "epoch": 2.7004608294930876,
      "grad_norm": 15.717339515686035,
      "kl": 0.3994140625,
      "learning_rate": 3.2384615384615383e-07,
      "loss": 0.0053,
      "reward": 1.3710676431655884,
      "reward_std": 0.19740483909845352,
      "reward_std_all": 0.18357788771390915,
      "rewards-max/forensics_sinkhorn_reward": 0.9748099744319916,
      "rewards-max/forensics_strict_edge_reward": 0.7388437688350677,
      "rewards-max/forensics_tiou_reward": 0.9633382856845856,
      "rewards/forensics_sinkhorn_reward": 0.76514932513237,
      "rewards/forensics_strict_edge_reward": 0.05693063512444496,
      "rewards/forensics_tiou_reward": 0.5489876866340637,
      "step": 880
    },
    {
      "completion_length": 21.8125,
      "epoch": 2.7035330261136714,
      "grad_norm": 12.382796287536621,
      "kl": 0.3388671875,
      "learning_rate": 3.230769230769231e-07,
      "loss": 0.0225,
      "reward": 1.2833625078201294,
      "reward_std": 0.2696625739336014,
      "reward_std_all": 0.23033790290355682,
      "rewards-max/forensics_sinkhorn_reward": 0.9747643768787384,
      "rewards-max/forensics_strict_edge_reward": 0.6934256851673126,
      "rewards-max/forensics_tiou_reward": 0.96733158826828,
      "rewards/forensics_sinkhorn_reward": 0.7391396164894104,
      "rewards/forensics_strict_edge_reward": 0.06162170507013798,
      "rewards/forensics_tiou_reward": 0.4826011657714844,
      "step": 881
    },
    {
      "completion_length": 22.171875,
      "epoch": 2.7066052227342547,
      "grad_norm": 11.034876823425293,
      "kl": 0.263671875,
      "learning_rate": 3.2230769230769226e-07,
      "loss": -0.0766,
      "reward": 1.3834820985794067,
      "reward_std": 0.23827607929706573,
      "reward_std_all": 0.21096622943878174,
      "rewards-max/forensics_sinkhorn_reward": 0.979178786277771,
      "rewards-max/forensics_strict_edge_reward": 0.6901743710041046,
      "rewards-max/forensics_tiou_reward": 0.9737413227558136,
      "rewards/forensics_sinkhorn_reward": 0.7433564364910126,
      "rewards/forensics_strict_edge_reward": 0.07499738410115242,
      "rewards/forensics_tiou_reward": 0.5651282966136932,
      "step": 882
    },
    {
      "completion_length": 23.03125,
      "epoch": 2.709677419354839,
      "grad_norm": 8.136452674865723,
      "kl": 0.33544921875,
      "learning_rate": 3.215384615384615e-07,
      "loss": 0.0257,
      "reward": 1.4411154985427856,
      "reward_std": 0.3142326921224594,
      "reward_std_all": 0.35189466178417206,
      "rewards-max/forensics_sinkhorn_reward": 0.9753483235836029,
      "rewards-max/forensics_strict_edge_reward": 0.6149742305278778,
      "rewards-max/forensics_tiou_reward": 0.9666672945022583,
      "rewards/forensics_sinkhorn_reward": 0.7850185334682465,
      "rewards/forensics_strict_edge_reward": 0.05488318018615246,
      "rewards/forensics_tiou_reward": 0.6012137532234192,
      "step": 883
    },
    {
      "completion_length": 27.421875,
      "epoch": 2.7127496159754223,
      "grad_norm": 9.812796592712402,
      "kl": 0.45703125,
      "learning_rate": 3.2076923076923074e-07,
      "loss": 0.1501,
      "reward": 1.3699390888214111,
      "reward_std": 0.23244848102331161,
      "reward_std_all": 0.2545342892408371,
      "rewards-max/forensics_sinkhorn_reward": 0.9678540825843811,
      "rewards-max/forensics_strict_edge_reward": 0.5900419056415558,
      "rewards-max/forensics_tiou_reward": 0.9446823596954346,
      "rewards/forensics_sinkhorn_reward": 0.7664946615695953,
      "rewards/forensics_strict_edge_reward": 0.049354616552591324,
      "rewards/forensics_tiou_reward": 0.5540898442268372,
      "step": 884
    },
    {
      "completion_length": 25.140625,
      "epoch": 2.715821812596006,
      "grad_norm": 9.32098388671875,
      "kl": 0.275390625,
      "learning_rate": 3.2e-07,
      "loss": 0.0747,
      "reward": 1.3386062979698181,
      "reward_std": 0.29555609077215195,
      "reward_std_all": 0.27118852734565735,
      "rewards-max/forensics_sinkhorn_reward": 0.9726379811763763,
      "rewards-max/forensics_strict_edge_reward": 0.5524591207504272,
      "rewards-max/forensics_tiou_reward": 0.9588903188705444,
      "rewards/forensics_sinkhorn_reward": 0.71843421459198,
      "rewards/forensics_strict_edge_reward": 0.08225243911147118,
      "rewards/forensics_tiou_reward": 0.5379197001457214,
      "step": 885
    },
    {
      "completion_length": 22.0,
      "epoch": 2.71889400921659,
      "grad_norm": 9.300785064697266,
      "kl": 0.42578125,
      "learning_rate": 3.1923076923076917e-07,
      "loss": 0.0958,
      "reward": 1.4010407328605652,
      "reward_std": 0.2631375193595886,
      "reward_std_all": 0.25088070333004,
      "rewards-max/forensics_sinkhorn_reward": 0.9774422347545624,
      "rewards-max/forensics_strict_edge_reward": 0.44343748688697815,
      "rewards-max/forensics_tiou_reward": 0.9685327112674713,
      "rewards/forensics_sinkhorn_reward": 0.7647432684898376,
      "rewards/forensics_strict_edge_reward": 0.046469708904623985,
      "rewards/forensics_tiou_reward": 0.5898277461528778,
      "step": 886
    },
    {
      "completion_length": 22.0,
      "epoch": 2.7219662058371736,
      "grad_norm": 11.6956787109375,
      "kl": 0.28173828125,
      "learning_rate": 3.1846153846153843e-07,
      "loss": 0.0495,
      "reward": 1.325099766254425,
      "reward_std": 0.23116962611675262,
      "reward_std_all": 0.1958823725581169,
      "rewards-max/forensics_sinkhorn_reward": 0.9693664014339447,
      "rewards-max/forensics_strict_edge_reward": 0.3065662086009979,
      "rewards-max/forensics_tiou_reward": 0.9491823017597198,
      "rewards/forensics_sinkhorn_reward": 0.7439035773277283,
      "rewards/forensics_strict_edge_reward": 0.036456664092838764,
      "rewards/forensics_tiou_reward": 0.5447395443916321,
      "step": 887
    },
    {
      "completion_length": 21.5,
      "epoch": 2.7250384024577574,
      "grad_norm": 13.425134658813477,
      "kl": 0.37548828125,
      "learning_rate": 3.176923076923077e-07,
      "loss": 0.082,
      "reward": 1.3901885747909546,
      "reward_std": 0.1890997737646103,
      "reward_std_all": 0.16844823956489563,
      "rewards-max/forensics_sinkhorn_reward": 0.9613555669784546,
      "rewards-max/forensics_strict_edge_reward": 0.2693905830383301,
      "rewards-max/forensics_tiou_reward": 0.9263650476932526,
      "rewards/forensics_sinkhorn_reward": 0.780620813369751,
      "rewards/forensics_strict_edge_reward": 0.036223096773028374,
      "rewards/forensics_tiou_reward": 0.5733446478843689,
      "step": 888
    },
    {
      "completion_length": 22.125,
      "epoch": 2.7281105990783407,
      "grad_norm": 11.192734718322754,
      "kl": 0.4541015625,
      "learning_rate": 3.169230769230769e-07,
      "loss": 0.0293,
      "reward": 1.4233810305595398,
      "reward_std": 0.23394303768873215,
      "reward_std_all": 0.22887777537107468,
      "rewards-max/forensics_sinkhorn_reward": 0.9780988395214081,
      "rewards-max/forensics_strict_edge_reward": 0.8333333432674408,
      "rewards-max/forensics_tiou_reward": 0.9742840230464935,
      "rewards/forensics_sinkhorn_reward": 0.771208643913269,
      "rewards/forensics_strict_edge_reward": 0.07819973304867744,
      "rewards/forensics_tiou_reward": 0.5739726424217224,
      "step": 889
    },
    {
      "completion_length": 20.859375,
      "epoch": 2.731182795698925,
      "grad_norm": 14.453432083129883,
      "kl": 0.220703125,
      "learning_rate": 3.161538461538462e-07,
      "loss": -0.0366,
      "reward": 1.2709794640541077,
      "reward_std": 0.17128360271453857,
      "reward_std_all": 0.14961637556552887,
      "rewards-max/forensics_sinkhorn_reward": 0.9056549072265625,
      "rewards-max/forensics_strict_edge_reward": 0.4200858473777771,
      "rewards-max/forensics_tiou_reward": 0.7994451224803925,
      "rewards/forensics_sinkhorn_reward": 0.7357407510280609,
      "rewards/forensics_strict_edge_reward": 0.03669359814375639,
      "rewards/forensics_tiou_reward": 0.4985451102256775,
      "step": 890
    },
    {
      "completion_length": 23.03125,
      "epoch": 2.7342549923195083,
      "grad_norm": 11.708611488342285,
      "kl": 0.3828125,
      "learning_rate": 3.1538461538461534e-07,
      "loss": 0.0154,
      "reward": 1.3979794383049011,
      "reward_std": 0.29344383627176285,
      "reward_std_all": 0.25689711421728134,
      "rewards-max/forensics_sinkhorn_reward": 0.9859542846679688,
      "rewards-max/forensics_strict_edge_reward": 0.7727466821670532,
      "rewards-max/forensics_tiou_reward": 0.9918849170207977,
      "rewards/forensics_sinkhorn_reward": 0.765309602022171,
      "rewards/forensics_strict_edge_reward": 0.06784110888838768,
      "rewards/forensics_tiou_reward": 0.5648287534713745,
      "step": 891
    },
    {
      "completion_length": 22.765625,
      "epoch": 2.737327188940092,
      "grad_norm": 12.70610523223877,
      "kl": 0.3212890625,
      "learning_rate": 3.146153846153846e-07,
      "loss": 0.1319,
      "reward": 1.2732486724853516,
      "reward_std": 0.29891733825206757,
      "reward_std_all": 0.3064529299736023,
      "rewards-max/forensics_sinkhorn_reward": 0.9708744287490845,
      "rewards-max/forensics_strict_edge_reward": 0.32904115319252014,
      "rewards-max/forensics_tiou_reward": 0.953662782907486,
      "rewards/forensics_sinkhorn_reward": 0.6968410909175873,
      "rewards/forensics_strict_edge_reward": 0.042481884360313416,
      "rewards/forensics_tiou_reward": 0.5339256525039673,
      "step": 892
    },
    {
      "completion_length": 21.5625,
      "epoch": 2.740399385560676,
      "grad_norm": 13.565186500549316,
      "kl": 0.23291015625,
      "learning_rate": 3.138461538461538e-07,
      "loss": 0.1392,
      "reward": 1.5123777389526367,
      "reward_std": 0.2990330308675766,
      "reward_std_all": 0.2731620818376541,
      "rewards-max/forensics_sinkhorn_reward": 0.9828027486801147,
      "rewards-max/forensics_strict_edge_reward": 0.6032288670539856,
      "rewards-max/forensics_tiou_reward": 0.9804246425628662,
      "rewards/forensics_sinkhorn_reward": 0.8175157308578491,
      "rewards/forensics_strict_edge_reward": 0.06394875049591064,
      "rewards/forensics_tiou_reward": 0.6309134066104889,
      "step": 893
    },
    {
      "completion_length": 25.734375,
      "epoch": 2.7434715821812596,
      "grad_norm": 10.843696594238281,
      "kl": 0.27783203125,
      "learning_rate": 3.130769230769231e-07,
      "loss": 0.103,
      "reward": 1.341556966304779,
      "reward_std": 0.216501846909523,
      "reward_std_all": 0.1922774761915207,
      "rewards-max/forensics_sinkhorn_reward": 0.9774905443191528,
      "rewards-max/forensics_strict_edge_reward": 0.2345871850848198,
      "rewards-max/forensics_tiou_reward": 0.9700287282466888,
      "rewards/forensics_sinkhorn_reward": 0.7477476000785828,
      "rewards/forensics_strict_edge_reward": 0.02961694449186325,
      "rewards/forensics_tiou_reward": 0.5641923993825912,
      "step": 894
    },
    {
      "completion_length": 23.984375,
      "epoch": 2.7465437788018434,
      "grad_norm": 10.070709228515625,
      "kl": 0.3515625,
      "learning_rate": 3.1230769230769225e-07,
      "loss": 0.1382,
      "reward": 1.3766353130340576,
      "reward_std": 0.24976520985364914,
      "reward_std_all": 0.21437698602676392,
      "rewards-max/forensics_sinkhorn_reward": 0.977675050497055,
      "rewards-max/forensics_strict_edge_reward": 0.3810797333717346,
      "rewards-max/forensics_tiou_reward": 0.9676638841629028,
      "rewards/forensics_sinkhorn_reward": 0.7477382123470306,
      "rewards/forensics_strict_edge_reward": 0.05290362052619457,
      "rewards/forensics_tiou_reward": 0.5759935081005096,
      "step": 895
    },
    {
      "completion_length": 22.328125,
      "epoch": 2.749615975422427,
      "grad_norm": 11.166543006896973,
      "kl": 0.34716796875,
      "learning_rate": 3.115384615384615e-07,
      "loss": -0.018,
      "reward": 1.4532004594802856,
      "reward_std": 0.18994402885437012,
      "reward_std_all": 0.17787813395261765,
      "rewards-max/forensics_sinkhorn_reward": 0.9435377418994904,
      "rewards-max/forensics_strict_edge_reward": 0.3813142329454422,
      "rewards-max/forensics_tiou_reward": 0.8892029821872711,
      "rewards/forensics_sinkhorn_reward": 0.7958644330501556,
      "rewards/forensics_strict_edge_reward": 0.04321678355336189,
      "rewards/forensics_tiou_reward": 0.6141193211078644,
      "step": 896
    },
    {
      "completion_length": 22.375,
      "epoch": 2.752688172043011,
      "grad_norm": 10.311776161193848,
      "kl": 0.24853515625,
      "learning_rate": 3.1076923076923073e-07,
      "loss": 0.0621,
      "reward": 1.4965898990631104,
      "reward_std": 0.33858777582645416,
      "reward_std_all": 0.30347439646720886,
      "rewards-max/forensics_sinkhorn_reward": 0.982373833656311,
      "rewards-max/forensics_strict_edge_reward": 0.49006423354148865,
      "rewards-max/forensics_tiou_reward": 0.9822108149528503,
      "rewards/forensics_sinkhorn_reward": 0.778502494096756,
      "rewards/forensics_strict_edge_reward": 0.0924306120723486,
      "rewards/forensics_tiou_reward": 0.6256568431854248,
      "step": 897
    },
    {
      "completion_length": 21.296875,
      "epoch": 2.7557603686635943,
      "grad_norm": 11.712973594665527,
      "kl": 0.2705078125,
      "learning_rate": 3.1e-07,
      "loss": 0.1988,
      "reward": 1.4209023714065552,
      "reward_std": 0.37219177186489105,
      "reward_std_all": 0.3250475898385048,
      "rewards-max/forensics_sinkhorn_reward": 0.9641098380088806,
      "rewards-max/forensics_strict_edge_reward": 0.5531879812479019,
      "rewards-max/forensics_tiou_reward": 0.9382501542568207,
      "rewards/forensics_sinkhorn_reward": 0.7744826674461365,
      "rewards/forensics_strict_edge_reward": 0.06233395170420408,
      "rewards/forensics_tiou_reward": 0.5840857028961182,
      "step": 898
    },
    {
      "completion_length": 21.1875,
      "epoch": 2.758832565284178,
      "grad_norm": 14.035881996154785,
      "kl": 0.31201171875,
      "learning_rate": 3.0923076923076926e-07,
      "loss": 0.0357,
      "reward": 1.387680172920227,
      "reward_std": 0.23543398827314377,
      "reward_std_all": 0.2262447252869606,
      "rewards-max/forensics_sinkhorn_reward": 0.9742448031902313,
      "rewards-max/forensics_strict_edge_reward": 0.31910227984189987,
      "rewards-max/forensics_tiou_reward": 0.9577888548374176,
      "rewards/forensics_sinkhorn_reward": 0.7654189467430115,
      "rewards/forensics_strict_edge_reward": 0.041299653239548206,
      "rewards/forensics_tiou_reward": 0.5809616446495056,
      "step": 899
    },
    {
      "completion_length": 23.6875,
      "epoch": 2.761904761904762,
      "grad_norm": 12.377790451049805,
      "kl": 0.5078125,
      "learning_rate": 3.084615384615384e-07,
      "loss": 0.133,
      "reward": 1.3989129662513733,
      "reward_std": 0.3101435899734497,
      "reward_std_all": 0.2756780982017517,
      "rewards-max/forensics_sinkhorn_reward": 0.9741982221603394,
      "rewards-max/forensics_strict_edge_reward": 0.6188668608665466,
      "rewards-max/forensics_tiou_reward": 0.9609953463077545,
      "rewards/forensics_sinkhorn_reward": 0.7708919942378998,
      "rewards/forensics_strict_edge_reward": 0.050131943076848984,
      "rewards/forensics_tiou_reward": 0.5778890252113342,
      "step": 900
    },
    {
      "completion_length": 22.46875,
      "epoch": 2.7649769585253456,
      "grad_norm": 11.378509521484375,
      "kl": 0.224609375,
      "learning_rate": 3.076923076923077e-07,
      "loss": 0.0029,
      "reward": 1.2491450905799866,
      "reward_std": 0.18217771500349045,
      "reward_std_all": 0.1967444121837616,
      "rewards-max/forensics_sinkhorn_reward": 0.9666292667388916,
      "rewards-max/forensics_strict_edge_reward": 0.5299537032842636,
      "rewards-max/forensics_tiou_reward": 0.9421573877334595,
      "rewards/forensics_sinkhorn_reward": 0.6932992339134216,
      "rewards/forensics_strict_edge_reward": 0.0361232440918684,
      "rewards/forensics_tiou_reward": 0.5197227001190186,
      "step": 901
    },
    {
      "completion_length": 21.890625,
      "epoch": 2.7680491551459294,
      "grad_norm": 13.312487602233887,
      "kl": 0.33203125,
      "learning_rate": 3.069230769230769e-07,
      "loss": -0.0893,
      "reward": 1.2476996183395386,
      "reward_std": 0.20800361782312393,
      "reward_std_all": 0.20090456306934357,
      "rewards-max/forensics_sinkhorn_reward": 0.9601556658744812,
      "rewards-max/forensics_strict_edge_reward": 0.37814464420080185,
      "rewards-max/forensics_tiou_reward": 0.9277006387710571,
      "rewards/forensics_sinkhorn_reward": 0.7122526466846466,
      "rewards/forensics_strict_edge_reward": 0.049268027767539024,
      "rewards/forensics_tiou_reward": 0.4861789792776108,
      "step": 902
    },
    {
      "completion_length": 24.375,
      "epoch": 2.771121351766513,
      "grad_norm": 19.673057556152344,
      "kl": 1.2216796875,
      "learning_rate": 3.0615384615384617e-07,
      "loss": 0.0572,
      "reward": 1.389203131198883,
      "reward_std": 0.26563749462366104,
      "reward_std_all": 0.2485550045967102,
      "rewards-max/forensics_sinkhorn_reward": 0.9659497737884521,
      "rewards-max/forensics_strict_edge_reward": 0.3831244856119156,
      "rewards-max/forensics_tiou_reward": 0.9297664165496826,
      "rewards/forensics_sinkhorn_reward": 0.7675102055072784,
      "rewards/forensics_strict_edge_reward": 0.023509250953793526,
      "rewards/forensics_tiou_reward": 0.5981836915016174,
      "step": 903
    },
    {
      "completion_length": 23.265625,
      "epoch": 2.774193548387097,
      "grad_norm": 9.849416732788086,
      "kl": 0.48046875,
      "learning_rate": 3.0538461538461533e-07,
      "loss": 0.0749,
      "reward": 1.355811059474945,
      "reward_std": 0.26976779103279114,
      "reward_std_all": 0.2358371913433075,
      "rewards-max/forensics_sinkhorn_reward": 0.9767940938472748,
      "rewards-max/forensics_strict_edge_reward": 0.7654097378253937,
      "rewards-max/forensics_tiou_reward": 0.9714982211589813,
      "rewards/forensics_sinkhorn_reward": 0.7496925890445709,
      "rewards/forensics_strict_edge_reward": 0.06843614391982555,
      "rewards/forensics_tiou_reward": 0.5376822352409363,
      "step": 904
    },
    {
      "completion_length": 20.71875,
      "epoch": 2.7772657450076803,
      "grad_norm": 15.073267936706543,
      "kl": 0.2841796875,
      "learning_rate": 3.046153846153846e-07,
      "loss": 0.06,
      "reward": 1.2703156471252441,
      "reward_std": 0.18574869632720947,
      "reward_std_all": 0.18885228037834167,
      "rewards-max/forensics_sinkhorn_reward": 0.9535554349422455,
      "rewards-max/forensics_strict_edge_reward": 0.07828757725656033,
      "rewards-max/forensics_tiou_reward": 0.8976298272609711,
      "rewards/forensics_sinkhorn_reward": 0.7212902009487152,
      "rewards/forensics_strict_edge_reward": 0.012891562073491514,
      "rewards/forensics_tiou_reward": 0.536133885383606,
      "step": 905
    },
    {
      "completion_length": 22.09375,
      "epoch": 2.780337941628264,
      "grad_norm": 10.465709686279297,
      "kl": 0.33740234375,
      "learning_rate": 3.038461538461538e-07,
      "loss": 0.0525,
      "reward": 1.56813246011734,
      "reward_std": 0.17532455176115036,
      "reward_std_all": 0.19054170697927475,
      "rewards-max/forensics_sinkhorn_reward": 0.9767318069934845,
      "rewards-max/forensics_strict_edge_reward": 0.4025607258081436,
      "rewards-max/forensics_tiou_reward": 0.9622006714344025,
      "rewards/forensics_sinkhorn_reward": 0.8359674215316772,
      "rewards/forensics_strict_edge_reward": 0.049410492181777954,
      "rewards/forensics_tiou_reward": 0.6827545464038849,
      "step": 906
    },
    {
      "completion_length": 21.265625,
      "epoch": 2.783410138248848,
      "grad_norm": 12.003198623657227,
      "kl": 0.53125,
      "learning_rate": 3.030769230769231e-07,
      "loss": -0.0296,
      "reward": 1.303417682647705,
      "reward_std": 0.2833789736032486,
      "reward_std_all": 0.26462437212467194,
      "rewards-max/forensics_sinkhorn_reward": 0.9684489667415619,
      "rewards-max/forensics_strict_edge_reward": 0.3972083032131195,
      "rewards-max/forensics_tiou_reward": 0.9492048025131226,
      "rewards/forensics_sinkhorn_reward": 0.7091173827648163,
      "rewards/forensics_strict_edge_reward": 0.06098382826894522,
      "rewards/forensics_tiou_reward": 0.5333164930343628,
      "step": 907
    },
    {
      "completion_length": 22.75,
      "epoch": 2.7864823348694316,
      "grad_norm": 14.66158676147461,
      "kl": 0.4697265625,
      "learning_rate": 3.0230769230769235e-07,
      "loss": 0.0487,
      "reward": 1.3569221496582031,
      "reward_std": 0.24376510083675385,
      "reward_std_all": 0.2082449346780777,
      "rewards-max/forensics_sinkhorn_reward": 0.9850437939167023,
      "rewards-max/forensics_strict_edge_reward": 0.40149129927158356,
      "rewards-max/forensics_tiou_reward": 0.9870900809764862,
      "rewards/forensics_sinkhorn_reward": 0.7604441046714783,
      "rewards/forensics_strict_edge_reward": 0.0430546966381371,
      "rewards/forensics_tiou_reward": 0.553423285484314,
      "step": 908
    },
    {
      "completion_length": 22.484375,
      "epoch": 2.7895545314900154,
      "grad_norm": 21.459064483642578,
      "kl": 0.28271484375,
      "learning_rate": 3.015384615384615e-07,
      "loss": 0.0695,
      "reward": 1.433622658252716,
      "reward_std": 0.26119011640548706,
      "reward_std_all": 0.2355065792798996,
      "rewards-max/forensics_sinkhorn_reward": 0.9559248089790344,
      "rewards-max/forensics_strict_edge_reward": 0.5754836797714233,
      "rewards-max/forensics_tiou_reward": 0.9140046536922455,
      "rewards/forensics_sinkhorn_reward": 0.7678196132183075,
      "rewards/forensics_strict_edge_reward": 0.08114159107208252,
      "rewards/forensics_tiou_reward": 0.5846615135669708,
      "step": 909
    },
    {
      "completion_length": 23.34375,
      "epoch": 2.792626728110599,
      "grad_norm": 13.376349449157715,
      "kl": 0.2802734375,
      "learning_rate": 3.0076923076923077e-07,
      "loss": 0.109,
      "reward": 1.3297265768051147,
      "reward_std": 0.18252381682395935,
      "reward_std_all": 0.1891215592622757,
      "rewards-max/forensics_sinkhorn_reward": 0.9569790065288544,
      "rewards-max/forensics_strict_edge_reward": 0.2186601385474205,
      "rewards-max/forensics_tiou_reward": 0.91624316573143,
      "rewards/forensics_sinkhorn_reward": 0.7426655292510986,
      "rewards/forensics_strict_edge_reward": 0.022602021228522062,
      "rewards/forensics_tiou_reward": 0.564458891749382,
      "step": 910
    },
    {
      "completion_length": 21.71875,
      "epoch": 2.795698924731183,
      "grad_norm": 15.732534408569336,
      "kl": 0.25244140625,
      "learning_rate": 3e-07,
      "loss": 0.0141,
      "reward": 1.3276798129081726,
      "reward_std": 0.27917955815792084,
      "reward_std_all": 0.24084362387657166,
      "rewards-max/forensics_sinkhorn_reward": 0.9853854477405548,
      "rewards-max/forensics_strict_edge_reward": 0.6147179305553436,
      "rewards-max/forensics_tiou_reward": 0.9862470626831055,
      "rewards/forensics_sinkhorn_reward": 0.7246844172477722,
      "rewards/forensics_strict_edge_reward": 0.08041836880147457,
      "rewards/forensics_tiou_reward": 0.5225770771503448,
      "step": 911
    },
    {
      "completion_length": 21.90625,
      "epoch": 2.7987711213517663,
      "grad_norm": 13.107465744018555,
      "kl": 0.4423828125,
      "learning_rate": 2.9923076923076925e-07,
      "loss": -0.0502,
      "reward": 1.3738975524902344,
      "reward_std": 0.19359386712312698,
      "reward_std_all": 0.19575800001621246,
      "rewards-max/forensics_sinkhorn_reward": 0.9711708724498749,
      "rewards-max/forensics_strict_edge_reward": 0.4508197009563446,
      "rewards-max/forensics_tiou_reward": 0.9511889219284058,
      "rewards/forensics_sinkhorn_reward": 0.7520975470542908,
      "rewards/forensics_strict_edge_reward": 0.05010915733873844,
      "rewards/forensics_tiou_reward": 0.5716909170150757,
      "step": 912
    },
    {
      "completion_length": 25.515625,
      "epoch": 2.80184331797235,
      "grad_norm": 9.509676933288574,
      "kl": 0.38232421875,
      "learning_rate": 2.984615384615384e-07,
      "loss": -0.0202,
      "reward": 1.337321698665619,
      "reward_std": 0.19922374188899994,
      "reward_std_all": 0.1961427479982376,
      "rewards-max/forensics_sinkhorn_reward": 0.945536345243454,
      "rewards-max/forensics_strict_edge_reward": 0.16478049382567406,
      "rewards-max/forensics_tiou_reward": 0.890963077545166,
      "rewards/forensics_sinkhorn_reward": 0.7466284036636353,
      "rewards/forensics_strict_edge_reward": 0.035042010713368654,
      "rewards/forensics_tiou_reward": 0.5556513071060181,
      "step": 913
    },
    {
      "completion_length": 23.09375,
      "epoch": 2.804915514592934,
      "grad_norm": 24.737411499023438,
      "kl": 0.4228515625,
      "learning_rate": 2.976923076923077e-07,
      "loss": -0.0909,
      "reward": 1.45927232503891,
      "reward_std": 0.19940506666898727,
      "reward_std_all": 0.22989434003829956,
      "rewards-max/forensics_sinkhorn_reward": 0.975060224533081,
      "rewards-max/forensics_strict_edge_reward": 0.57895427942276,
      "rewards-max/forensics_tiou_reward": 0.9645665884017944,
      "rewards/forensics_sinkhorn_reward": 0.7772410213947296,
      "rewards/forensics_strict_edge_reward": 0.07412723638117313,
      "rewards/forensics_tiou_reward": 0.6079041361808777,
      "step": 914
    },
    {
      "completion_length": 22.4375,
      "epoch": 2.8079877112135176,
      "grad_norm": 9.68719482421875,
      "kl": 0.3232421875,
      "learning_rate": 2.969230769230769e-07,
      "loss": 0.0155,
      "reward": 1.4464058876037598,
      "reward_std": 0.22377817332744598,
      "reward_std_all": 0.20190656930208206,
      "rewards-max/forensics_sinkhorn_reward": 0.9763282239437103,
      "rewards-max/forensics_strict_edge_reward": 0.5221183001995087,
      "rewards-max/forensics_tiou_reward": 0.9655804336071014,
      "rewards/forensics_sinkhorn_reward": 0.7907002866268158,
      "rewards/forensics_strict_edge_reward": 0.06878525018692017,
      "rewards/forensics_tiou_reward": 0.5869203805923462,
      "step": 915
    },
    {
      "completion_length": 22.546875,
      "epoch": 2.8110599078341014,
      "grad_norm": 12.189176559448242,
      "kl": 0.21875,
      "learning_rate": 2.9615384615384616e-07,
      "loss": -0.0362,
      "reward": 1.3043816685676575,
      "reward_std": 0.21413271874189377,
      "reward_std_all": 0.21678397059440613,
      "rewards-max/forensics_sinkhorn_reward": 0.9626278877258301,
      "rewards-max/forensics_strict_edge_reward": 0.3264661692082882,
      "rewards-max/forensics_tiou_reward": 0.9296941757202148,
      "rewards/forensics_sinkhorn_reward": 0.718603640794754,
      "rewards/forensics_strict_edge_reward": 0.043392484076321125,
      "rewards/forensics_tiou_reward": 0.5423855483531952,
      "step": 916
    },
    {
      "completion_length": 25.34375,
      "epoch": 2.814132104454685,
      "grad_norm": 9.808574676513672,
      "kl": 0.3984375,
      "learning_rate": 2.953846153846154e-07,
      "loss": 0.0213,
      "reward": 1.3319623470306396,
      "reward_std": 0.2824878841638565,
      "reward_std_all": 0.27506884187459946,
      "rewards-max/forensics_sinkhorn_reward": 0.9676377773284912,
      "rewards-max/forensics_strict_edge_reward": 0.4329897463321686,
      "rewards-max/forensics_tiou_reward": 0.9435532987117767,
      "rewards/forensics_sinkhorn_reward": 0.7281379997730255,
      "rewards/forensics_strict_edge_reward": 0.057081046514213085,
      "rewards/forensics_tiou_reward": 0.5467432737350464,
      "step": 917
    },
    {
      "completion_length": 21.671875,
      "epoch": 2.817204301075269,
      "grad_norm": 12.673433303833008,
      "kl": 0.388671875,
      "learning_rate": 2.946153846153846e-07,
      "loss": -0.0458,
      "reward": 1.55899578332901,
      "reward_std": 0.233500674366951,
      "reward_std_all": 0.22814316302537918,
      "rewards-max/forensics_sinkhorn_reward": 0.9651082456111908,
      "rewards-max/forensics_strict_edge_reward": 0.4149532616138458,
      "rewards-max/forensics_tiou_reward": 0.9415490031242371,
      "rewards/forensics_sinkhorn_reward": 0.8188828229904175,
      "rewards/forensics_strict_edge_reward": 0.06365206092596054,
      "rewards/forensics_tiou_reward": 0.6764608919620514,
      "step": 918
    },
    {
      "completion_length": 21.96875,
      "epoch": 2.8202764976958523,
      "grad_norm": 13.359613418579102,
      "kl": 0.2900390625,
      "learning_rate": 2.9384615384615385e-07,
      "loss": -0.0163,
      "reward": 1.356988787651062,
      "reward_std": 0.27949807047843933,
      "reward_std_all": 0.2503616437315941,
      "rewards-max/forensics_sinkhorn_reward": 0.9797488451004028,
      "rewards-max/forensics_strict_edge_reward": 0.3332939688116312,
      "rewards-max/forensics_tiou_reward": 0.9749878644943237,
      "rewards/forensics_sinkhorn_reward": 0.7456057369709015,
      "rewards/forensics_strict_edge_reward": 0.044378308462910354,
      "rewards/forensics_tiou_reward": 0.5670047104358673,
      "step": 919
    },
    {
      "completion_length": 24.015625,
      "epoch": 2.823348694316436,
      "grad_norm": 10.63493537902832,
      "kl": 0.3076171875,
      "learning_rate": 2.9307692307692307e-07,
      "loss": 0.0473,
      "reward": 1.3850961923599243,
      "reward_std": 0.2837272435426712,
      "reward_std_all": 0.26641470193862915,
      "rewards-max/forensics_sinkhorn_reward": 0.986704409122467,
      "rewards-max/forensics_strict_edge_reward": 0.8582656681537628,
      "rewards-max/forensics_tiou_reward": 0.9938460290431976,
      "rewards/forensics_sinkhorn_reward": 0.7590766847133636,
      "rewards/forensics_strict_edge_reward": 0.06424316763877869,
      "rewards/forensics_tiou_reward": 0.5617762506008148,
      "step": 920
    },
    {
      "completion_length": 24.65625,
      "epoch": 2.82642089093702,
      "grad_norm": 12.690006256103516,
      "kl": 0.240234375,
      "learning_rate": 2.9230769230769234e-07,
      "loss": 0.1632,
      "reward": 1.270141839981079,
      "reward_std": 0.19195937365293503,
      "reward_std_all": 0.23123158514499664,
      "rewards-max/forensics_sinkhorn_reward": 0.9469112753868103,
      "rewards-max/forensics_strict_edge_reward": 0.35020002722740173,
      "rewards-max/forensics_tiou_reward": 0.8898530304431915,
      "rewards/forensics_sinkhorn_reward": 0.7006348967552185,
      "rewards/forensics_strict_edge_reward": 0.04002309963107109,
      "rewards/forensics_tiou_reward": 0.52948397397995,
      "step": 921
    },
    {
      "completion_length": 21.5625,
      "epoch": 2.8294930875576036,
      "grad_norm": 15.498927116394043,
      "kl": 0.2763671875,
      "learning_rate": 2.915384615384615e-07,
      "loss": 0.0008,
      "reward": 1.3852139115333557,
      "reward_std": 0.29746805131435394,
      "reward_std_all": 0.266168974339962,
      "rewards-max/forensics_sinkhorn_reward": 0.9748750329017639,
      "rewards-max/forensics_strict_edge_reward": 0.5210531502962112,
      "rewards-max/forensics_tiou_reward": 0.9626083672046661,
      "rewards/forensics_sinkhorn_reward": 0.7481576204299927,
      "rewards/forensics_strict_edge_reward": 0.05865752883255482,
      "rewards/forensics_tiou_reward": 0.578398734331131,
      "step": 922
    },
    {
      "completion_length": 22.859375,
      "epoch": 2.8325652841781874,
      "grad_norm": 13.725482940673828,
      "kl": 0.2978515625,
      "learning_rate": 2.9076923076923076e-07,
      "loss": 0.0598,
      "reward": 1.3190218806266785,
      "reward_std": 0.2493104413151741,
      "reward_std_all": 0.23520517349243164,
      "rewards-max/forensics_sinkhorn_reward": 0.9641157388687134,
      "rewards-max/forensics_strict_edge_reward": 0.3821936398744583,
      "rewards-max/forensics_tiou_reward": 0.936599999666214,
      "rewards/forensics_sinkhorn_reward": 0.7175838649272919,
      "rewards/forensics_strict_edge_reward": 0.0398457832634449,
      "rewards/forensics_tiou_reward": 0.5615922808647156,
      "step": 923
    },
    {
      "completion_length": 22.890625,
      "epoch": 2.835637480798771,
      "grad_norm": 70.32025909423828,
      "kl": 0.2802734375,
      "learning_rate": 2.9e-07,
      "loss": 0.0174,
      "reward": 1.296116828918457,
      "reward_std": 0.3051757365465164,
      "reward_std_all": 0.2712337225675583,
      "rewards-max/forensics_sinkhorn_reward": 0.9688863158226013,
      "rewards-max/forensics_strict_edge_reward": 0.3444886729121208,
      "rewards-max/forensics_tiou_reward": 0.9450859427452087,
      "rewards/forensics_sinkhorn_reward": 0.723068505525589,
      "rewards/forensics_strict_edge_reward": 0.030428696423768997,
      "rewards/forensics_tiou_reward": 0.5426196604967117,
      "step": 924
    },
    {
      "completion_length": 20.609375,
      "epoch": 2.838709677419355,
      "grad_norm": 13.29162311553955,
      "kl": 0.33056640625,
      "learning_rate": 2.8923076923076924e-07,
      "loss": -0.0261,
      "reward": 1.4574323296546936,
      "reward_std": 0.29006239026784897,
      "reward_std_all": 0.2885734736919403,
      "rewards-max/forensics_sinkhorn_reward": 0.9745753407478333,
      "rewards-max/forensics_strict_edge_reward": 0.5910519808530807,
      "rewards-max/forensics_tiou_reward": 0.961227297782898,
      "rewards/forensics_sinkhorn_reward": 0.7829619646072388,
      "rewards/forensics_strict_edge_reward": 0.07878262177109718,
      "rewards/forensics_tiou_reward": 0.5956877768039703,
      "step": 925
    },
    {
      "completion_length": 21.640625,
      "epoch": 2.8417818740399383,
      "grad_norm": 18.288259506225586,
      "kl": 0.55078125,
      "learning_rate": 2.884615384615384e-07,
      "loss": -0.0015,
      "reward": 1.341488540172577,
      "reward_std": 0.22427164018154144,
      "reward_std_all": 0.213381826877594,
      "rewards-max/forensics_sinkhorn_reward": 0.9657860398292542,
      "rewards-max/forensics_strict_edge_reward": 0.31910227984189987,
      "rewards-max/forensics_tiou_reward": 0.9466930329799652,
      "rewards/forensics_sinkhorn_reward": 0.7647296190261841,
      "rewards/forensics_strict_edge_reward": 0.025124669075012207,
      "rewards/forensics_tiou_reward": 0.551634281873703,
      "step": 926
    },
    {
      "completion_length": 23.96875,
      "epoch": 2.8448540706605225,
      "grad_norm": 12.802728652954102,
      "kl": 0.58251953125,
      "learning_rate": 2.8769230769230767e-07,
      "loss": 0.0684,
      "reward": 1.3439885377883911,
      "reward_std": 0.26135361194610596,
      "reward_std_all": 0.26786812394857407,
      "rewards-max/forensics_sinkhorn_reward": 0.9778899848461151,
      "rewards-max/forensics_strict_edge_reward": 0.5495419502258301,
      "rewards-max/forensics_tiou_reward": 0.9708270728588104,
      "rewards/forensics_sinkhorn_reward": 0.6983522772789001,
      "rewards/forensics_strict_edge_reward": 0.0676611177623272,
      "rewards/forensics_tiou_reward": 0.5779750645160675,
      "step": 927
    },
    {
      "completion_length": 22.390625,
      "epoch": 2.847926267281106,
      "grad_norm": 14.16667652130127,
      "kl": 0.5078125,
      "learning_rate": 2.8692307692307694e-07,
      "loss": -0.0141,
      "reward": 1.4161638617515564,
      "reward_std": 0.29623064398765564,
      "reward_std_all": 0.2813160866498947,
      "rewards-max/forensics_sinkhorn_reward": 0.9736617803573608,
      "rewards-max/forensics_strict_edge_reward": 0.49895569682121277,
      "rewards-max/forensics_tiou_reward": 0.9602905511856079,
      "rewards/forensics_sinkhorn_reward": 0.7858405709266663,
      "rewards/forensics_strict_edge_reward": 0.059198809787631035,
      "rewards/forensics_tiou_reward": 0.5711245685815811,
      "step": 928
    },
    {
      "completion_length": 23.109375,
      "epoch": 2.8509984639016897,
      "grad_norm": 14.027689933776855,
      "kl": 0.3408203125,
      "learning_rate": 2.8615384615384615e-07,
      "loss": -0.0117,
      "reward": 1.3719660639762878,
      "reward_std": 0.268836110830307,
      "reward_std_all": 0.23668333888053894,
      "rewards-max/forensics_sinkhorn_reward": 0.9366475045681,
      "rewards-max/forensics_strict_edge_reward": 0.21809175424277782,
      "rewards-max/forensics_tiou_reward": 0.8600861430168152,
      "rewards/forensics_sinkhorn_reward": 0.769381582736969,
      "rewards/forensics_strict_edge_reward": 0.029446294531226158,
      "rewards/forensics_tiou_reward": 0.5731381773948669,
      "step": 929
    },
    {
      "completion_length": 21.515625,
      "epoch": 2.8540706605222734,
      "grad_norm": 14.500003814697266,
      "kl": 0.20947265625,
      "learning_rate": 2.8538461538461536e-07,
      "loss": -0.0433,
      "reward": 1.3466894626617432,
      "reward_std": 0.2214297652244568,
      "reward_std_all": 0.19137071818113327,
      "rewards-max/forensics_sinkhorn_reward": 0.9713792204856873,
      "rewards-max/forensics_strict_edge_reward": 0.5602131485939026,
      "rewards-max/forensics_tiou_reward": 0.9544388949871063,
      "rewards/forensics_sinkhorn_reward": 0.7439550757408142,
      "rewards/forensics_strict_edge_reward": 0.06268332432955503,
      "rewards/forensics_tiou_reward": 0.5400510728359222,
      "step": 930
    },
    {
      "completion_length": 20.421875,
      "epoch": 2.857142857142857,
      "grad_norm": 12.078885078430176,
      "kl": 0.21484375,
      "learning_rate": 2.846153846153846e-07,
      "loss": -0.1123,
      "reward": 1.4801506996154785,
      "reward_std": 0.2893797606229782,
      "reward_std_all": 0.2650364488363266,
      "rewards-max/forensics_sinkhorn_reward": 0.9855067729949951,
      "rewards-max/forensics_strict_edge_reward": 0.45133768022060394,
      "rewards-max/forensics_tiou_reward": 0.9907363951206207,
      "rewards/forensics_sinkhorn_reward": 0.79292231798172,
      "rewards/forensics_strict_edge_reward": 0.08999672532081604,
      "rewards/forensics_tiou_reward": 0.5972315967082977,
      "step": 931
    },
    {
      "completion_length": 22.359375,
      "epoch": 2.860215053763441,
      "grad_norm": 11.812883377075195,
      "kl": 0.25244140625,
      "learning_rate": 2.8384615384615384e-07,
      "loss": 0.0478,
      "reward": 1.100515604019165,
      "reward_std": 0.2371804416179657,
      "reward_std_all": 0.2154531553387642,
      "rewards-max/forensics_sinkhorn_reward": 0.8883009552955627,
      "rewards-max/forensics_strict_edge_reward": 0.21475112903863192,
      "rewards-max/forensics_tiou_reward": 0.7833648025989532,
      "rewards/forensics_sinkhorn_reward": 0.6747693717479706,
      "rewards/forensics_strict_edge_reward": 0.013415610417723656,
      "rewards/forensics_tiou_reward": 0.41233062744140625,
      "step": 932
    },
    {
      "completion_length": 19.96875,
      "epoch": 2.8632872503840243,
      "grad_norm": 13.25533676147461,
      "kl": 0.328125,
      "learning_rate": 2.8307692307692306e-07,
      "loss": 0.0103,
      "reward": 1.3389836549758911,
      "reward_std": 0.2766493484377861,
      "reward_std_all": 0.24934181571006775,
      "rewards-max/forensics_sinkhorn_reward": 0.9736685156822205,
      "rewards-max/forensics_strict_edge_reward": 0.6226264834403992,
      "rewards-max/forensics_tiou_reward": 0.9622591435909271,
      "rewards/forensics_sinkhorn_reward": 0.7448316812515259,
      "rewards/forensics_strict_edge_reward": 0.0482111144810915,
      "rewards/forensics_tiou_reward": 0.5459408164024353,
      "step": 933
    },
    {
      "completion_length": 21.796875,
      "epoch": 2.8663594470046085,
      "grad_norm": 9.564692497253418,
      "kl": 0.26904296875,
      "learning_rate": 2.823076923076923e-07,
      "loss": 0.0647,
      "reward": 1.542129933834076,
      "reward_std": 0.34498655796051025,
      "reward_std_all": 0.3165447264909744,
      "rewards-max/forensics_sinkhorn_reward": 0.9733303189277649,
      "rewards-max/forensics_strict_edge_reward": 0.46283967792987823,
      "rewards-max/forensics_tiou_reward": 0.9596344232559204,
      "rewards/forensics_sinkhorn_reward": 0.8073596358299255,
      "rewards/forensics_strict_edge_reward": 0.07366380095481873,
      "rewards/forensics_tiou_reward": 0.6611064672470093,
      "step": 934
    },
    {
      "completion_length": 22.875,
      "epoch": 2.869431643625192,
      "grad_norm": 9.177580833435059,
      "kl": 0.345703125,
      "learning_rate": 2.815384615384615e-07,
      "loss": 0.06,
      "reward": 1.3061496019363403,
      "reward_std": 0.22541388869285583,
      "reward_std_all": 0.22974424809217453,
      "rewards-max/forensics_sinkhorn_reward": 0.9545987546443939,
      "rewards-max/forensics_strict_edge_reward": 0.10807029157876968,
      "rewards-max/forensics_tiou_reward": 0.8946615159511566,
      "rewards/forensics_sinkhorn_reward": 0.7400180697441101,
      "rewards/forensics_strict_edge_reward": 0.011298978235572577,
      "rewards/forensics_tiou_reward": 0.5548325181007385,
      "step": 935
    },
    {
      "completion_length": 22.796875,
      "epoch": 2.8725038402457757,
      "grad_norm": 14.417346000671387,
      "kl": 0.373046875,
      "learning_rate": 2.8076923076923075e-07,
      "loss": 0.1052,
      "reward": 1.2300866842269897,
      "reward_std": 0.2255825400352478,
      "reward_std_all": 0.20658227056264877,
      "rewards-max/forensics_sinkhorn_reward": 0.9453489482402802,
      "rewards-max/forensics_strict_edge_reward": 0.3694510543718934,
      "rewards-max/forensics_tiou_reward": 0.8949925601482391,
      "rewards/forensics_sinkhorn_reward": 0.694749653339386,
      "rewards/forensics_strict_edge_reward": 0.03701569413533434,
      "rewards/forensics_tiou_reward": 0.49832141399383545,
      "step": 936
    },
    {
      "completion_length": 22.71875,
      "epoch": 2.8755760368663594,
      "grad_norm": 8.647092819213867,
      "kl": 0.353515625,
      "learning_rate": 2.8e-07,
      "loss": -0.036,
      "reward": 1.4558262825012207,
      "reward_std": 0.21933585405349731,
      "reward_std_all": 0.20451223850250244,
      "rewards-max/forensics_sinkhorn_reward": 0.9771409034729004,
      "rewards-max/forensics_strict_edge_reward": 0.5737849175930023,
      "rewards-max/forensics_tiou_reward": 0.9684003591537476,
      "rewards/forensics_sinkhorn_reward": 0.782550722360611,
      "rewards/forensics_strict_edge_reward": 0.04704934172332287,
      "rewards/forensics_tiou_reward": 0.6262262165546417,
      "step": 937
    },
    {
      "completion_length": 20.265625,
      "epoch": 2.878648233486943,
      "grad_norm": 16.910295486450195,
      "kl": 0.208984375,
      "learning_rate": 2.7923076923076923e-07,
      "loss": 0.0142,
      "reward": 1.5371640920639038,
      "reward_std": 0.19822001457214355,
      "reward_std_all": 0.20576217770576477,
      "rewards-max/forensics_sinkhorn_reward": 0.9767889380455017,
      "rewards-max/forensics_strict_edge_reward": 0.750005841255188,
      "rewards-max/forensics_tiou_reward": 0.9702880084514618,
      "rewards/forensics_sinkhorn_reward": 0.8115541338920593,
      "rewards/forensics_strict_edge_reward": 0.09184172376990318,
      "rewards/forensics_tiou_reward": 0.6337683200836182,
      "step": 938
    },
    {
      "completion_length": 22.53125,
      "epoch": 2.881720430107527,
      "grad_norm": 10.865653038024902,
      "kl": 0.2724609375,
      "learning_rate": 2.7846153846153845e-07,
      "loss": 0.0501,
      "reward": 1.4352381229400635,
      "reward_std": 0.3503062278032303,
      "reward_std_all": 0.31623509526252747,
      "rewards-max/forensics_sinkhorn_reward": 0.9752661287784576,
      "rewards-max/forensics_strict_edge_reward": 0.384186252951622,
      "rewards-max/forensics_tiou_reward": 0.9596192836761475,
      "rewards/forensics_sinkhorn_reward": 0.7845797538757324,
      "rewards/forensics_strict_edge_reward": 0.06615013256669044,
      "rewards/forensics_tiou_reward": 0.5845082402229309,
      "step": 939
    },
    {
      "completion_length": 23.421875,
      "epoch": 2.8847926267281108,
      "grad_norm": 9.610053062438965,
      "kl": 0.310546875,
      "learning_rate": 2.7769230769230766e-07,
      "loss": 0.0206,
      "reward": 1.4552603960037231,
      "reward_std": 0.3506586253643036,
      "reward_std_all": 0.2989492565393448,
      "rewards-max/forensics_sinkhorn_reward": 0.9802624583244324,
      "rewards-max/forensics_strict_edge_reward": 0.6363584399223328,
      "rewards-max/forensics_tiou_reward": 0.9757070541381836,
      "rewards/forensics_sinkhorn_reward": 0.7448646426200867,
      "rewards/forensics_strict_edge_reward": 0.09826121479272842,
      "rewards/forensics_tiou_reward": 0.6121345162391663,
      "step": 940
    },
    {
      "completion_length": 21.59375,
      "epoch": 2.8878648233486945,
      "grad_norm": 13.059893608093262,
      "kl": 0.2880859375,
      "learning_rate": 2.7692307692307693e-07,
      "loss": -0.013,
      "reward": 1.3228400349617004,
      "reward_std": 0.28219713270664215,
      "reward_std_all": 0.2531043216586113,
      "rewards-max/forensics_sinkhorn_reward": 0.9550345242023468,
      "rewards-max/forensics_strict_edge_reward": 0.32078617066144943,
      "rewards-max/forensics_tiou_reward": 0.9194949567317963,
      "rewards/forensics_sinkhorn_reward": 0.7418705821037292,
      "rewards/forensics_strict_edge_reward": 0.0409513246268034,
      "rewards/forensics_tiou_reward": 0.5400181412696838,
      "step": 941
    },
    {
      "completion_length": 22.53125,
      "epoch": 2.890937019969278,
      "grad_norm": 11.919137954711914,
      "kl": 0.28125,
      "learning_rate": 2.7615384615384614e-07,
      "loss": 0.0793,
      "reward": 1.4472553730010986,
      "reward_std": 0.205590158700943,
      "reward_std_all": 0.2857162207365036,
      "rewards-max/forensics_sinkhorn_reward": 0.9557322561740875,
      "rewards-max/forensics_strict_edge_reward": 0.5254621207714081,
      "rewards-max/forensics_tiou_reward": 0.9155553579330444,
      "rewards/forensics_sinkhorn_reward": 0.779782235622406,
      "rewards/forensics_strict_edge_reward": 0.08020131848752499,
      "rewards/forensics_tiou_reward": 0.5872716903686523,
      "step": 942
    },
    {
      "completion_length": 22.203125,
      "epoch": 2.8940092165898617,
      "grad_norm": 14.242620468139648,
      "kl": 0.431640625,
      "learning_rate": 2.753846153846154e-07,
      "loss": 0.013,
      "reward": 1.5413551330566406,
      "reward_std": 0.20373231172561646,
      "reward_std_all": 0.20083393901586533,
      "rewards-max/forensics_sinkhorn_reward": 0.9751162528991699,
      "rewards-max/forensics_strict_edge_reward": 0.4259721636772156,
      "rewards-max/forensics_tiou_reward": 0.9596825540065765,
      "rewards/forensics_sinkhorn_reward": 0.8131902515888214,
      "rewards/forensics_strict_edge_reward": 0.06948006059974432,
      "rewards/forensics_tiou_reward": 0.6586848199367523,
      "step": 943
    },
    {
      "completion_length": 20.765625,
      "epoch": 2.8970814132104454,
      "grad_norm": 11.995780944824219,
      "kl": 0.29296875,
      "learning_rate": 2.7461538461538457e-07,
      "loss": 0.0715,
      "reward": 1.300951600074768,
      "reward_std": 0.21324950456619263,
      "reward_std_all": 0.198837049305439,
      "rewards-max/forensics_sinkhorn_reward": 0.9712100923061371,
      "rewards-max/forensics_strict_edge_reward": 0.5320244655013084,
      "rewards-max/forensics_tiou_reward": 0.9593895375728607,
      "rewards/forensics_sinkhorn_reward": 0.7312858700752258,
      "rewards/forensics_strict_edge_reward": 0.04858618974685669,
      "rewards/forensics_tiou_reward": 0.5210796296596527,
      "step": 944
    },
    {
      "completion_length": 19.671875,
      "epoch": 2.900153609831029,
      "grad_norm": 15.474132537841797,
      "kl": 0.26416015625,
      "learning_rate": 2.7384615384615383e-07,
      "loss": 0.0729,
      "reward": 1.2174378037452698,
      "reward_std": 0.186996191740036,
      "reward_std_all": 0.19012810289859772,
      "rewards-max/forensics_sinkhorn_reward": 0.9430162012577057,
      "rewards-max/forensics_strict_edge_reward": 0.34791333973407745,
      "rewards-max/forensics_tiou_reward": 0.8845400810241699,
      "rewards/forensics_sinkhorn_reward": 0.7098247706890106,
      "rewards/forensics_strict_edge_reward": 0.02169762272387743,
      "rewards/forensics_tiou_reward": 0.485915407538414,
      "step": 945
    },
    {
      "completion_length": 24.0,
      "epoch": 2.903225806451613,
      "grad_norm": 13.68740177154541,
      "kl": 0.25537109375,
      "learning_rate": 2.7307692307692305e-07,
      "loss": 0.173,
      "reward": 1.331258773803711,
      "reward_std": 0.22456437349319458,
      "reward_std_all": 0.20543326437473297,
      "rewards-max/forensics_sinkhorn_reward": 0.9625864624977112,
      "rewards-max/forensics_strict_edge_reward": 0.2911505401134491,
      "rewards-max/forensics_tiou_reward": 0.9299532473087311,
      "rewards/forensics_sinkhorn_reward": 0.7267318367958069,
      "rewards/forensics_strict_edge_reward": 0.04203628934919834,
      "rewards/forensics_tiou_reward": 0.5624906420707703,
      "step": 946
    },
    {
      "completion_length": 23.375,
      "epoch": 2.9062980030721968,
      "grad_norm": 18.944067001342773,
      "kl": 0.25732421875,
      "learning_rate": 2.723076923076923e-07,
      "loss": 0.014,
      "reward": 1.5141109228134155,
      "reward_std": 0.29388774931430817,
      "reward_std_all": 0.2720661759376526,
      "rewards-max/forensics_sinkhorn_reward": 0.9813455939292908,
      "rewards-max/forensics_strict_edge_reward": 0.7381869554519653,
      "rewards-max/forensics_tiou_reward": 0.9793850779533386,
      "rewards/forensics_sinkhorn_reward": 0.7789008617401123,
      "rewards/forensics_strict_edge_reward": 0.12293345108628273,
      "rewards/forensics_tiou_reward": 0.6122766137123108,
      "step": 947
    },
    {
      "completion_length": 20.3125,
      "epoch": 2.9093701996927805,
      "grad_norm": 12.834728240966797,
      "kl": 0.443359375,
      "learning_rate": 2.7153846153846153e-07,
      "loss": 0.0222,
      "reward": 1.4756608605384827,
      "reward_std": 0.18326688557863235,
      "reward_std_all": 0.2356945425271988,
      "rewards-max/forensics_sinkhorn_reward": 0.9720912277698517,
      "rewards-max/forensics_strict_edge_reward": 0.4974842667579651,
      "rewards-max/forensics_tiou_reward": 0.9563914835453033,
      "rewards/forensics_sinkhorn_reward": 0.8099819421768188,
      "rewards/forensics_strict_edge_reward": 0.04573217686265707,
      "rewards/forensics_tiou_reward": 0.6199466288089752,
      "step": 948
    },
    {
      "completion_length": 22.484375,
      "epoch": 2.912442396313364,
      "grad_norm": 11.561810493469238,
      "kl": 0.23193359375,
      "learning_rate": 2.7076923076923074e-07,
      "loss": 0.0374,
      "reward": 1.422132968902588,
      "reward_std": 0.20773330330848694,
      "reward_std_all": 0.21970577538013458,
      "rewards-max/forensics_sinkhorn_reward": 0.9406644999980927,
      "rewards-max/forensics_strict_edge_reward": 0.26513541489839554,
      "rewards-max/forensics_tiou_reward": 0.8854773044586182,
      "rewards/forensics_sinkhorn_reward": 0.7660071551799774,
      "rewards/forensics_strict_edge_reward": 0.059336043894290924,
      "rewards/forensics_tiou_reward": 0.5967898666858673,
      "step": 949
    },
    {
      "completion_length": 22.0,
      "epoch": 2.9155145929339477,
      "grad_norm": 12.505314826965332,
      "kl": 0.94140625,
      "learning_rate": 2.7e-07,
      "loss": 0.0563,
      "reward": 1.2265591621398926,
      "reward_std": 0.25331375002861023,
      "reward_std_all": 0.26734408736228943,
      "rewards-max/forensics_sinkhorn_reward": 0.9648584723472595,
      "rewards-max/forensics_strict_edge_reward": 0.32841892540454865,
      "rewards-max/forensics_tiou_reward": 0.9366106986999512,
      "rewards/forensics_sinkhorn_reward": 0.6953570544719696,
      "rewards/forensics_strict_edge_reward": 0.042275868356227875,
      "rewards/forensics_tiou_reward": 0.4889262318611145,
      "step": 950
    },
    {
      "completion_length": 22.609375,
      "epoch": 2.9185867895545314,
      "grad_norm": 12.046134948730469,
      "kl": 0.3125,
      "learning_rate": 2.692307692307692e-07,
      "loss": 0.0394,
      "reward": 1.379410445690155,
      "reward_std": 0.19464049488306046,
      "reward_std_all": 0.20967605710029602,
      "rewards-max/forensics_sinkhorn_reward": 0.9616053402423859,
      "rewards-max/forensics_strict_edge_reward": 0.3850628435611725,
      "rewards-max/forensics_tiou_reward": 0.9300369322299957,
      "rewards/forensics_sinkhorn_reward": 0.7634441256523132,
      "rewards/forensics_strict_edge_reward": 0.044575247913599014,
      "rewards/forensics_tiou_reward": 0.5713910460472107,
      "step": 951
    },
    {
      "completion_length": 21.125,
      "epoch": 2.921658986175115,
      "grad_norm": 33.20677185058594,
      "kl": 1.748046875,
      "learning_rate": 2.6846153846153844e-07,
      "loss": 0.0018,
      "reward": 1.393178403377533,
      "reward_std": 0.26700131595134735,
      "reward_std_all": 0.24051332473754883,
      "rewards-max/forensics_sinkhorn_reward": 0.9511905610561371,
      "rewards-max/forensics_strict_edge_reward": 0.21653109788894653,
      "rewards-max/forensics_tiou_reward": 0.8995859622955322,
      "rewards/forensics_sinkhorn_reward": 0.7807802259922028,
      "rewards/forensics_strict_edge_reward": 0.02872518263757229,
      "rewards/forensics_tiou_reward": 0.5836730301380157,
      "step": 952
    },
    {
      "completion_length": 24.21875,
      "epoch": 2.924731182795699,
      "grad_norm": 11.749404907226562,
      "kl": 0.2890625,
      "learning_rate": 2.6769230769230765e-07,
      "loss": 0.0192,
      "reward": 1.4871370792388916,
      "reward_std": 0.3107675313949585,
      "reward_std_all": 0.2673511430621147,
      "rewards-max/forensics_sinkhorn_reward": 0.9827407002449036,
      "rewards-max/forensics_strict_edge_reward": 0.6824378371238708,
      "rewards-max/forensics_tiou_reward": 0.984342485666275,
      "rewards/forensics_sinkhorn_reward": 0.8029029667377472,
      "rewards/forensics_strict_edge_reward": 0.06282591819763184,
      "rewards/forensics_tiou_reward": 0.6214083731174469,
      "step": 953
    },
    {
      "completion_length": 20.90625,
      "epoch": 2.9278033794162828,
      "grad_norm": 14.158266067504883,
      "kl": 0.2421875,
      "learning_rate": 2.669230769230769e-07,
      "loss": 0.0384,
      "reward": 1.200877845287323,
      "reward_std": 0.2548849582672119,
      "reward_std_all": 0.23770643025636673,
      "rewards-max/forensics_sinkhorn_reward": 0.9865117967128754,
      "rewards-max/forensics_strict_edge_reward": 0.5439433008432388,
      "rewards-max/forensics_tiou_reward": 0.99191614985466,
      "rewards/forensics_sinkhorn_reward": 0.6717851459980011,
      "rewards/forensics_strict_edge_reward": 0.047994617372751236,
      "rewards/forensics_tiou_reward": 0.4810980409383774,
      "step": 954
    },
    {
      "completion_length": 21.6875,
      "epoch": 2.9308755760368665,
      "grad_norm": 13.71070671081543,
      "kl": 0.380859375,
      "learning_rate": 2.6615384615384613e-07,
      "loss": 0.0672,
      "reward": 1.444670855998993,
      "reward_std": 0.2194429561495781,
      "reward_std_all": 0.20100633054971695,
      "rewards-max/forensics_sinkhorn_reward": 0.9662971496582031,
      "rewards-max/forensics_strict_edge_reward": 0.4216773808002472,
      "rewards-max/forensics_tiou_reward": 0.9309473037719727,
      "rewards/forensics_sinkhorn_reward": 0.7923955321311951,
      "rewards/forensics_strict_edge_reward": 0.041067502461373806,
      "rewards/forensics_tiou_reward": 0.6112078428268433,
      "step": 955
    },
    {
      "completion_length": 24.203125,
      "epoch": 2.93394777265745,
      "grad_norm": 12.315363883972168,
      "kl": 0.3408203125,
      "learning_rate": 2.653846153846154e-07,
      "loss": -0.07,
      "reward": 1.3805056810379028,
      "reward_std": 0.23658134043216705,
      "reward_std_all": 0.2201951965689659,
      "rewards-max/forensics_sinkhorn_reward": 0.9724555611610413,
      "rewards-max/forensics_strict_edge_reward": 0.49686121940612793,
      "rewards-max/forensics_tiou_reward": 0.958019345998764,
      "rewards/forensics_sinkhorn_reward": 0.7269156277179718,
      "rewards/forensics_strict_edge_reward": 0.08209886960685253,
      "rewards/forensics_tiou_reward": 0.5714912414550781,
      "step": 956
    },
    {
      "completion_length": 22.5625,
      "epoch": 2.9370199692780337,
      "grad_norm": 21.579452514648438,
      "kl": 0.376953125,
      "learning_rate": 2.646153846153846e-07,
      "loss": -0.0232,
      "reward": 1.392899751663208,
      "reward_std": 0.1769150346517563,
      "reward_std_all": 0.1796232908964157,
      "rewards-max/forensics_sinkhorn_reward": 0.9627043604850769,
      "rewards-max/forensics_strict_edge_reward": 0.604914128780365,
      "rewards-max/forensics_tiou_reward": 0.9379510283470154,
      "rewards/forensics_sinkhorn_reward": 0.7689982354640961,
      "rewards/forensics_strict_edge_reward": 0.046571023762226105,
      "rewards/forensics_tiou_reward": 0.5773303508758545,
      "step": 957
    },
    {
      "completion_length": 22.859375,
      "epoch": 2.9400921658986174,
      "grad_norm": 9.950685501098633,
      "kl": 0.4208984375,
      "learning_rate": 2.638461538461538e-07,
      "loss": 0.0494,
      "reward": 1.3743836879730225,
      "reward_std": 0.27949658036231995,
      "reward_std_all": 0.25601381063461304,
      "rewards-max/forensics_sinkhorn_reward": 0.9665851593017578,
      "rewards-max/forensics_strict_edge_reward": 0.4945315718650818,
      "rewards-max/forensics_tiou_reward": 0.94670769572258,
      "rewards/forensics_sinkhorn_reward": 0.7782714664936066,
      "rewards/forensics_strict_edge_reward": 0.03781159408390522,
      "rewards/forensics_tiou_reward": 0.5583007037639618,
      "step": 958
    },
    {
      "completion_length": 21.46875,
      "epoch": 2.943164362519201,
      "grad_norm": 10.814718246459961,
      "kl": 0.5263671875,
      "learning_rate": 2.630769230769231e-07,
      "loss": 0.0528,
      "reward": 1.4226229786872864,
      "reward_std": 0.2867204248905182,
      "reward_std_all": 0.27032075822353363,
      "rewards-max/forensics_sinkhorn_reward": 0.9614943861961365,
      "rewards-max/forensics_strict_edge_reward": 0.5190257281064987,
      "rewards-max/forensics_tiou_reward": 0.9345571100711823,
      "rewards/forensics_sinkhorn_reward": 0.7864639759063721,
      "rewards/forensics_strict_edge_reward": 0.05643320083618164,
      "rewards/forensics_tiou_reward": 0.5797258019447327,
      "step": 959
    },
    {
      "completion_length": 23.71875,
      "epoch": 2.946236559139785,
      "grad_norm": 11.175861358642578,
      "kl": 0.265625,
      "learning_rate": 2.623076923076923e-07,
      "loss": 0.1297,
      "reward": 1.290695309638977,
      "reward_std": 0.17253120243549347,
      "reward_std_all": 0.2019556164741516,
      "rewards-max/forensics_sinkhorn_reward": 0.9629998803138733,
      "rewards-max/forensics_strict_edge_reward": 0.37797266244888306,
      "rewards-max/forensics_tiou_reward": 0.9305156767368317,
      "rewards/forensics_sinkhorn_reward": 0.7254425287246704,
      "rewards/forensics_strict_edge_reward": 0.034875527024269104,
      "rewards/forensics_tiou_reward": 0.5303773283958435,
      "step": 960
    },
    {
      "completion_length": 21.8125,
      "epoch": 2.9493087557603688,
      "grad_norm": 14.90368366241455,
      "kl": 0.25439453125,
      "learning_rate": 2.615384615384615e-07,
      "loss": -0.0451,
      "reward": 1.2320603132247925,
      "reward_std": 0.3262484669685364,
      "reward_std_all": 0.2822238430380821,
      "rewards-max/forensics_sinkhorn_reward": 0.9703212082386017,
      "rewards-max/forensics_strict_edge_reward": 0.5261107087135315,
      "rewards-max/forensics_tiou_reward": 0.9550401270389557,
      "rewards/forensics_sinkhorn_reward": 0.6879628300666809,
      "rewards/forensics_strict_edge_reward": 0.06753231398761272,
      "rewards/forensics_tiou_reward": 0.4765651822090149,
      "step": 961
    },
    {
      "completion_length": 21.5625,
      "epoch": 2.9523809523809526,
      "grad_norm": 9.508164405822754,
      "kl": 0.263671875,
      "learning_rate": 2.6076923076923073e-07,
      "loss": 0.0216,
      "reward": 1.3942239880561829,
      "reward_std": 0.2843650281429291,
      "reward_std_all": 0.2518014758825302,
      "rewards-max/forensics_sinkhorn_reward": 0.9727663397789001,
      "rewards-max/forensics_strict_edge_reward": 0.5559406280517578,
      "rewards-max/forensics_tiou_reward": 0.9534910619258881,
      "rewards/forensics_sinkhorn_reward": 0.7614523470401764,
      "rewards/forensics_strict_edge_reward": 0.07200594432651997,
      "rewards/forensics_tiou_reward": 0.5607656836509705,
      "step": 962
    },
    {
      "completion_length": 21.21875,
      "epoch": 2.955453149001536,
      "grad_norm": 8.798666000366211,
      "kl": 0.2841796875,
      "learning_rate": 2.6e-07,
      "loss": -0.0601,
      "reward": 1.3314338326454163,
      "reward_std": 0.22920186072587967,
      "reward_std_all": 0.22841620072722435,
      "rewards-max/forensics_sinkhorn_reward": 0.9850738346576691,
      "rewards-max/forensics_strict_edge_reward": 0.4268360957503319,
      "rewards-max/forensics_tiou_reward": 0.9884650707244873,
      "rewards/forensics_sinkhorn_reward": 0.7360198497772217,
      "rewards/forensics_strict_edge_reward": 0.024954780703410506,
      "rewards/forensics_tiou_reward": 0.570459246635437,
      "step": 963
    },
    {
      "completion_length": 21.90625,
      "epoch": 2.9585253456221197,
      "grad_norm": 15.421111106872559,
      "kl": 0.4375,
      "learning_rate": 2.592307692307692e-07,
      "loss": 0.0022,
      "reward": 1.466229498386383,
      "reward_std": 0.2589130252599716,
      "reward_std_all": 0.23786013573408127,
      "rewards-max/forensics_sinkhorn_reward": 0.9570586383342743,
      "rewards-max/forensics_strict_edge_reward": 0.6497763246297836,
      "rewards-max/forensics_tiou_reward": 0.9209339618682861,
      "rewards/forensics_sinkhorn_reward": 0.7830567955970764,
      "rewards/forensics_strict_edge_reward": 0.07489523850381374,
      "rewards/forensics_tiou_reward": 0.608277440071106,
      "step": 964
    },
    {
      "completion_length": 24.75,
      "epoch": 2.9615975422427034,
      "grad_norm": 12.039477348327637,
      "kl": 0.22900390625,
      "learning_rate": 2.584615384615385e-07,
      "loss": 0.0612,
      "reward": 1.5752986669540405,
      "reward_std": 0.21299729496240616,
      "reward_std_all": 0.1943359375,
      "rewards-max/forensics_sinkhorn_reward": 0.967024952173233,
      "rewards-max/forensics_strict_edge_reward": 0.5539533495903015,
      "rewards-max/forensics_tiou_reward": 0.9479478895664215,
      "rewards/forensics_sinkhorn_reward": 0.8329424262046814,
      "rewards/forensics_strict_edge_reward": 0.05979350209236145,
      "rewards/forensics_tiou_reward": 0.6825628876686096,
      "step": 965
    },
    {
      "completion_length": 22.625,
      "epoch": 2.964669738863287,
      "grad_norm": 12.286970138549805,
      "kl": 0.333984375,
      "learning_rate": 2.5769230769230764e-07,
      "loss": 0.033,
      "reward": 1.4706887006759644,
      "reward_std": 0.275016188621521,
      "reward_std_all": 0.2701643779873848,
      "rewards-max/forensics_sinkhorn_reward": 0.9776108264923096,
      "rewards-max/forensics_strict_edge_reward": 0.503647968173027,
      "rewards-max/forensics_tiou_reward": 0.9668504595756531,
      "rewards/forensics_sinkhorn_reward": 0.7829302847385406,
      "rewards/forensics_strict_edge_reward": 0.07269743084907532,
      "rewards/forensics_tiou_reward": 0.6150609254837036,
      "step": 966
    },
    {
      "completion_length": 23.859375,
      "epoch": 2.967741935483871,
      "grad_norm": 10.449833869934082,
      "kl": 0.4033203125,
      "learning_rate": 2.569230769230769e-07,
      "loss": 0.0701,
      "reward": 1.2132362127304077,
      "reward_std": 0.20648907870054245,
      "reward_std_all": 0.19159181416034698,
      "rewards-max/forensics_sinkhorn_reward": 0.9755417108535767,
      "rewards-max/forensics_strict_edge_reward": 0.3728483244776726,
      "rewards-max/forensics_tiou_reward": 0.9662280380725861,
      "rewards/forensics_sinkhorn_reward": 0.7027043104171753,
      "rewards/forensics_strict_edge_reward": 0.01836958108469844,
      "rewards/forensics_tiou_reward": 0.4921623170375824,
      "step": 967
    },
    {
      "completion_length": 20.140625,
      "epoch": 2.9708141321044548,
      "grad_norm": 22.65896224975586,
      "kl": 0.3037109375,
      "learning_rate": 2.5615384615384617e-07,
      "loss": -0.0683,
      "reward": 1.4556320905685425,
      "reward_std": 0.18246348202228546,
      "reward_std_all": 0.16165746748447418,
      "rewards-max/forensics_sinkhorn_reward": 0.9393466114997864,
      "rewards-max/forensics_strict_edge_reward": 0.2798697128891945,
      "rewards-max/forensics_tiou_reward": 0.8806538879871368,
      "rewards/forensics_sinkhorn_reward": 0.7987735867500305,
      "rewards/forensics_strict_edge_reward": 0.03731168480589986,
      "rewards/forensics_tiou_reward": 0.6195468306541443,
      "step": 968
    },
    {
      "completion_length": 23.59375,
      "epoch": 2.9738863287250386,
      "grad_norm": 9.606636047363281,
      "kl": 0.3173828125,
      "learning_rate": 2.553846153846154e-07,
      "loss": -0.0295,
      "reward": 1.3002055287361145,
      "reward_std": 0.15835846215486526,
      "reward_std_all": 0.18669076263904572,
      "rewards-max/forensics_sinkhorn_reward": 0.941860020160675,
      "rewards-max/forensics_strict_edge_reward": 0.4890154078602791,
      "rewards-max/forensics_tiou_reward": 0.8857113420963287,
      "rewards/forensics_sinkhorn_reward": 0.6932017803192139,
      "rewards/forensics_strict_edge_reward": 0.06266859360039234,
      "rewards/forensics_tiou_reward": 0.544335126876831,
      "step": 969
    },
    {
      "completion_length": 24.46875,
      "epoch": 2.976958525345622,
      "grad_norm": 13.639596939086914,
      "kl": 0.3984375,
      "learning_rate": 2.546153846153846e-07,
      "loss": 0.0174,
      "reward": 1.3846960663795471,
      "reward_std": 0.2375044822692871,
      "reward_std_all": 0.2614033818244934,
      "rewards-max/forensics_sinkhorn_reward": 0.9693346917629242,
      "rewards-max/forensics_strict_edge_reward": 0.5579978227615356,
      "rewards-max/forensics_tiou_reward": 0.9496556222438812,
      "rewards/forensics_sinkhorn_reward": 0.7497566342353821,
      "rewards/forensics_strict_edge_reward": 0.06387260928750038,
      "rewards/forensics_tiou_reward": 0.5710667669773102,
      "step": 970
    },
    {
      "completion_length": 22.625,
      "epoch": 2.980030721966206,
      "grad_norm": 13.861035346984863,
      "kl": 0.29833984375,
      "learning_rate": 2.538461538461538e-07,
      "loss": 0.0077,
      "reward": 1.2585646510124207,
      "reward_std": 0.2535795271396637,
      "reward_std_all": 0.23528771102428436,
      "rewards-max/forensics_sinkhorn_reward": 0.9444066882133484,
      "rewards-max/forensics_strict_edge_reward": 0.3003734350204468,
      "rewards-max/forensics_tiou_reward": 0.8952149748802185,
      "rewards/forensics_sinkhorn_reward": 0.7146672606468201,
      "rewards/forensics_strict_edge_reward": 0.03520877659320831,
      "rewards/forensics_tiou_reward": 0.5086885988712311,
      "step": 971
    },
    {
      "completion_length": 22.65625,
      "epoch": 2.9831029185867894,
      "grad_norm": 9.941447257995605,
      "kl": 0.24951171875,
      "learning_rate": 2.530769230769231e-07,
      "loss": -0.0077,
      "reward": 1.2879900932312012,
      "reward_std": 0.2914106547832489,
      "reward_std_all": 0.2669985890388489,
      "rewards-max/forensics_sinkhorn_reward": 0.9631634056568146,
      "rewards-max/forensics_strict_edge_reward": 0.4109206646680832,
      "rewards-max/forensics_tiou_reward": 0.9261579215526581,
      "rewards/forensics_sinkhorn_reward": 0.7221721112728119,
      "rewards/forensics_strict_edge_reward": 0.05612512305378914,
      "rewards/forensics_tiou_reward": 0.5096929371356964,
      "step": 972
    },
    {
      "completion_length": 22.4375,
      "epoch": 2.986175115207373,
      "grad_norm": 20.883371353149414,
      "kl": 1.39453125,
      "learning_rate": 2.523076923076923e-07,
      "loss": -0.0429,
      "reward": 1.474467933177948,
      "reward_std": 0.22297364473342896,
      "reward_std_all": 0.1997951790690422,
      "rewards-max/forensics_sinkhorn_reward": 0.9707134366035461,
      "rewards-max/forensics_strict_edge_reward": 0.3640780746936798,
      "rewards-max/forensics_tiou_reward": 0.9558257460594177,
      "rewards/forensics_sinkhorn_reward": 0.8017080724239349,
      "rewards/forensics_strict_edge_reward": 0.04608146473765373,
      "rewards/forensics_tiou_reward": 0.6266784965991974,
      "step": 973
    },
    {
      "completion_length": 21.828125,
      "epoch": 2.989247311827957,
      "grad_norm": 11.230242729187012,
      "kl": 0.6044921875,
      "learning_rate": 2.515384615384615e-07,
      "loss": 0.0677,
      "reward": 1.318827748298645,
      "reward_std": 0.23529879748821259,
      "reward_std_all": 0.2593298703432083,
      "rewards-max/forensics_sinkhorn_reward": 0.9811611175537109,
      "rewards-max/forensics_strict_edge_reward": 0.4648469015955925,
      "rewards-max/forensics_tiou_reward": 0.9752172529697418,
      "rewards/forensics_sinkhorn_reward": 0.7260473966598511,
      "rewards/forensics_strict_edge_reward": 0.04528442211449146,
      "rewards/forensics_tiou_reward": 0.5474958568811417,
      "step": 974
    },
    {
      "completion_length": 22.515625,
      "epoch": 2.9923195084485408,
      "grad_norm": 11.721611022949219,
      "kl": 0.27197265625,
      "learning_rate": 2.507692307692307e-07,
      "loss": -0.0048,
      "reward": 1.4363465309143066,
      "reward_std": 0.20948120951652527,
      "reward_std_all": 0.20370257645845413,
      "rewards-max/forensics_sinkhorn_reward": 0.9803700745105743,
      "rewards-max/forensics_strict_edge_reward": 0.6733285784721375,
      "rewards-max/forensics_tiou_reward": 0.9732729196548462,
      "rewards/forensics_sinkhorn_reward": 0.7725706994533539,
      "rewards/forensics_strict_edge_reward": 0.07874000817537308,
      "rewards/forensics_tiou_reward": 0.5850358307361603,
      "step": 975
    },
    {
      "completion_length": 21.34375,
      "epoch": 2.9953917050691246,
      "grad_norm": 8.981985092163086,
      "kl": 0.44921875,
      "learning_rate": 2.5e-07,
      "loss": 0.1032,
      "reward": 1.3729363679885864,
      "reward_std": 0.2609715238213539,
      "reward_std_all": 0.2648051232099533,
      "rewards-max/forensics_sinkhorn_reward": 0.9729902148246765,
      "rewards-max/forensics_strict_edge_reward": 0.3391084223985672,
      "rewards-max/forensics_tiou_reward": 0.9599611461162567,
      "rewards/forensics_sinkhorn_reward": 0.7558833062648773,
      "rewards/forensics_strict_edge_reward": 0.056129617150872946,
      "rewards/forensics_tiou_reward": 0.5609235167503357,
      "step": 976
    },
    {
      "completion_length": 21.890625,
      "epoch": 2.998463901689708,
      "grad_norm": 20.82079315185547,
      "kl": 0.39453125,
      "learning_rate": 2.492307692307692e-07,
      "loss": 0.0726,
      "reward": 1.350846529006958,
      "reward_std": 0.2455911561846733,
      "reward_std_all": 0.2146880030632019,
      "rewards-max/forensics_sinkhorn_reward": 0.9660845100879669,
      "rewards-max/forensics_strict_edge_reward": 0.1778687834739685,
      "rewards-max/forensics_tiou_reward": 0.9272667169570923,
      "rewards/forensics_sinkhorn_reward": 0.7554458677768707,
      "rewards/forensics_strict_edge_reward": 0.020012879744172096,
      "rewards/forensics_tiou_reward": 0.5753878057003021,
      "step": 977
    },
    {
      "completion_length": 35.60000228881836,
      "epoch": 3.0,
      "grad_norm": 7.250450611114502,
      "kl": 0.482421875,
      "learning_rate": 2.4846153846153847e-07,
      "loss": -0.0593,
      "reward": 1.5352014303207397,
      "reward_std": 0.4152957499027252,
      "reward_std_all": 0.432708740234375,
      "rewards-max/forensics_sinkhorn_reward": 0.9443325996398926,
      "rewards-max/forensics_strict_edge_reward": 0.04243364557623863,
      "rewards-max/forensics_tiou_reward": 0.8955219388008118,
      "rewards/forensics_sinkhorn_reward": 0.8442444801330566,
      "rewards/forensics_strict_edge_reward": 0.010878035798668861,
      "rewards/forensics_tiou_reward": 0.68007892370224,
      "step": 978
    },
    {
      "completion_length": 22.015625,
      "epoch": 3.003072196620584,
      "grad_norm": 12.828496932983398,
      "kl": 0.2734375,
      "learning_rate": 2.476923076923077e-07,
      "loss": -0.0443,
      "reward": 1.5689812302589417,
      "reward_std": 0.17952080070972443,
      "reward_std_all": 0.18888023495674133,
      "rewards-max/forensics_sinkhorn_reward": 0.9594714343547821,
      "rewards-max/forensics_strict_edge_reward": 0.5299537032842636,
      "rewards-max/forensics_tiou_reward": 0.9203267395496368,
      "rewards/forensics_sinkhorn_reward": 0.8296765983104706,
      "rewards/forensics_strict_edge_reward": 0.0767427608370781,
      "rewards/forensics_tiou_reward": 0.6625618636608124,
      "step": 979
    },
    {
      "completion_length": 20.140625,
      "epoch": 3.0061443932411676,
      "grad_norm": 12.863929748535156,
      "kl": 0.4443359375,
      "learning_rate": 2.469230769230769e-07,
      "loss": 0.0415,
      "reward": 1.2534247636795044,
      "reward_std": 0.24524198472499847,
      "reward_std_all": 0.22389957308769226,
      "rewards-max/forensics_sinkhorn_reward": 0.9520056545734406,
      "rewards-max/forensics_strict_edge_reward": 0.24642999470233917,
      "rewards-max/forensics_tiou_reward": 0.906699001789093,
      "rewards/forensics_sinkhorn_reward": 0.7061751186847687,
      "rewards/forensics_strict_edge_reward": 0.02762369718402624,
      "rewards/forensics_tiou_reward": 0.5196260213851929,
      "step": 980
    },
    {
      "completion_length": 20.484375,
      "epoch": 3.0092165898617513,
      "grad_norm": 11.842999458312988,
      "kl": 0.640625,
      "learning_rate": 2.4615384615384616e-07,
      "loss": 0.1168,
      "reward": 1.2072536945343018,
      "reward_std": 0.30900293588638306,
      "reward_std_all": 0.2944592088460922,
      "rewards-max/forensics_sinkhorn_reward": 0.9726584553718567,
      "rewards-max/forensics_strict_edge_reward": 0.5159863531589508,
      "rewards-max/forensics_tiou_reward": 0.9556099474430084,
      "rewards/forensics_sinkhorn_reward": 0.7012749016284943,
      "rewards/forensics_strict_edge_reward": 0.05272845830768347,
      "rewards/forensics_tiou_reward": 0.45325033366680145,
      "step": 981
    },
    {
      "completion_length": 22.421875,
      "epoch": 3.0122887864823347,
      "grad_norm": 10.927701950073242,
      "kl": 0.2880859375,
      "learning_rate": 2.453846153846154e-07,
      "loss": -0.0563,
      "reward": 1.5864855647087097,
      "reward_std": 0.23704780638217926,
      "reward_std_all": 0.21472900360822678,
      "rewards-max/forensics_sinkhorn_reward": 0.9612418115139008,
      "rewards-max/forensics_strict_edge_reward": 0.19554343819618225,
      "rewards-max/forensics_tiou_reward": 0.9303869903087616,
      "rewards/forensics_sinkhorn_reward": 0.8431476056575775,
      "rewards/forensics_strict_edge_reward": 0.02760407468304038,
      "rewards/forensics_tiou_reward": 0.7157338857650757,
      "step": 982
    },
    {
      "completion_length": 21.328125,
      "epoch": 3.0153609831029184,
      "grad_norm": 17.131135940551758,
      "kl": 0.94140625,
      "learning_rate": 2.446153846153846e-07,
      "loss": 0.0736,
      "reward": 1.4096214175224304,
      "reward_std": 0.3358810544013977,
      "reward_std_all": 0.3052142783999443,
      "rewards-max/forensics_sinkhorn_reward": 0.9541407823562622,
      "rewards-max/forensics_strict_edge_reward": 0.5062990635633469,
      "rewards-max/forensics_tiou_reward": 0.91196009516716,
      "rewards/forensics_sinkhorn_reward": 0.7724066376686096,
      "rewards/forensics_strict_edge_reward": 0.05356409214437008,
      "rewards/forensics_tiou_reward": 0.5836507081985474,
      "step": 983
    },
    {
      "completion_length": 20.0,
      "epoch": 3.0184331797235022,
      "grad_norm": 13.261197090148926,
      "kl": 0.38671875,
      "learning_rate": 2.4384615384615386e-07,
      "loss": 0.0722,
      "reward": 1.2924976348876953,
      "reward_std": 0.21568097174167633,
      "reward_std_all": 0.22742947936058044,
      "rewards-max/forensics_sinkhorn_reward": 0.9707913100719452,
      "rewards-max/forensics_strict_edge_reward": 0.25334038585424423,
      "rewards-max/forensics_tiou_reward": 0.9484304785728455,
      "rewards/forensics_sinkhorn_reward": 0.7399002313613892,
      "rewards/forensics_strict_edge_reward": 0.026688099838793278,
      "rewards/forensics_tiou_reward": 0.5259093046188354,
      "step": 984
    },
    {
      "completion_length": 22.265625,
      "epoch": 3.021505376344086,
      "grad_norm": 11.424779891967773,
      "kl": 0.2880859375,
      "learning_rate": 2.4307692307692307e-07,
      "loss": -0.0487,
      "reward": 1.3132058382034302,
      "reward_std": 0.2792353108525276,
      "reward_std_all": 0.2835703194141388,
      "rewards-max/forensics_sinkhorn_reward": 0.9604232609272003,
      "rewards-max/forensics_strict_edge_reward": 0.2484866762533784,
      "rewards-max/forensics_tiou_reward": 0.9330622851848602,
      "rewards/forensics_sinkhorn_reward": 0.7369918823242188,
      "rewards/forensics_strict_edge_reward": 0.02835703524760902,
      "rewards/forensics_tiou_reward": 0.5478569716215134,
      "step": 985
    },
    {
      "completion_length": 22.234375,
      "epoch": 3.02457757296467,
      "grad_norm": 10.483904838562012,
      "kl": 0.337890625,
      "learning_rate": 2.423076923076923e-07,
      "loss": -0.0425,
      "reward": 1.5889541506767273,
      "reward_std": 0.30983859300613403,
      "reward_std_all": 0.27080781757831573,
      "rewards-max/forensics_sinkhorn_reward": 0.987285315990448,
      "rewards-max/forensics_strict_edge_reward": 0.6770831942558289,
      "rewards-max/forensics_tiou_reward": 0.9950769245624542,
      "rewards/forensics_sinkhorn_reward": 0.8313614130020142,
      "rewards/forensics_strict_edge_reward": 0.08092242106795311,
      "rewards/forensics_tiou_reward": 0.6766704022884369,
      "step": 986
    },
    {
      "completion_length": 22.65625,
      "epoch": 3.0276497695852536,
      "grad_norm": 12.21804428100586,
      "kl": 0.2822265625,
      "learning_rate": 2.415384615384615e-07,
      "loss": -0.0033,
      "reward": 1.4903764128684998,
      "reward_std": 0.2620956301689148,
      "reward_std_all": 0.23617111146450043,
      "rewards-max/forensics_sinkhorn_reward": 0.9607810378074646,
      "rewards-max/forensics_strict_edge_reward": 0.5853640735149384,
      "rewards-max/forensics_tiou_reward": 0.9315247535705566,
      "rewards/forensics_sinkhorn_reward": 0.7864903807640076,
      "rewards/forensics_strict_edge_reward": 0.08491591364145279,
      "rewards/forensics_tiou_reward": 0.6189702153205872,
      "step": 987
    },
    {
      "completion_length": 20.25,
      "epoch": 3.0307219662058373,
      "grad_norm": 16.012100219726562,
      "kl": 0.23486328125,
      "learning_rate": 2.4076923076923076e-07,
      "loss": -0.0055,
      "reward": 1.3274773359298706,
      "reward_std": 0.21428745985031128,
      "reward_std_all": 0.18954279273748398,
      "rewards-max/forensics_sinkhorn_reward": 0.9784795343875885,
      "rewards-max/forensics_strict_edge_reward": 0.4852591007947922,
      "rewards-max/forensics_tiou_reward": 0.9751404523849487,
      "rewards/forensics_sinkhorn_reward": 0.7416786849498749,
      "rewards/forensics_strict_edge_reward": 0.06020732503384352,
      "rewards/forensics_tiou_reward": 0.5255913138389587,
      "step": 988
    },
    {
      "completion_length": 21.609375,
      "epoch": 3.033794162826421,
      "grad_norm": 14.869386672973633,
      "kl": 0.4736328125,
      "learning_rate": 2.4e-07,
      "loss": 0.0064,
      "reward": 1.225094974040985,
      "reward_std": 0.1546926237642765,
      "reward_std_all": 0.13007987290620804,
      "rewards-max/forensics_sinkhorn_reward": 0.9642766416072845,
      "rewards-max/forensics_strict_edge_reward": 0.37659838050603867,
      "rewards-max/forensics_tiou_reward": 0.9401244521141052,
      "rewards/forensics_sinkhorn_reward": 0.6996138989925385,
      "rewards/forensics_strict_edge_reward": 0.05649790447205305,
      "rewards/forensics_tiou_reward": 0.46898317337036133,
      "step": 989
    },
    {
      "completion_length": 19.96875,
      "epoch": 3.0368663594470044,
      "grad_norm": 13.017195701599121,
      "kl": 0.71875,
      "learning_rate": 2.392307692307692e-07,
      "loss": 0.0026,
      "reward": 1.3733193278312683,
      "reward_std": 0.23112548142671585,
      "reward_std_all": 0.20230722427368164,
      "rewards-max/forensics_sinkhorn_reward": 0.9271093308925629,
      "rewards-max/forensics_strict_edge_reward": 0.3572895973920822,
      "rewards-max/forensics_tiou_reward": 0.8537293076515198,
      "rewards/forensics_sinkhorn_reward": 0.7736510336399078,
      "rewards/forensics_strict_edge_reward": 0.04277743212878704,
      "rewards/forensics_tiou_reward": 0.5568909347057343,
      "step": 990
    },
    {
      "completion_length": 21.421875,
      "epoch": 3.0399385560675882,
      "grad_norm": 9.978898048400879,
      "kl": 0.470703125,
      "learning_rate": 2.3846153846153846e-07,
      "loss": -0.0356,
      "reward": 1.4862638711929321,
      "reward_std": 0.18230139464139938,
      "reward_std_all": 0.18075818568468094,
      "rewards-max/forensics_sinkhorn_reward": 0.9571696221828461,
      "rewards-max/forensics_strict_edge_reward": 0.38450370728969574,
      "rewards-max/forensics_tiou_reward": 0.9201252460479736,
      "rewards/forensics_sinkhorn_reward": 0.8102364540100098,
      "rewards/forensics_strict_edge_reward": 0.04742486774921417,
      "rewards/forensics_tiou_reward": 0.6286025643348694,
      "step": 991
    },
    {
      "completion_length": 23.75,
      "epoch": 3.043010752688172,
      "grad_norm": 14.311792373657227,
      "kl": 0.4052734375,
      "learning_rate": 2.376923076923077e-07,
      "loss": 0.0097,
      "reward": 1.551591157913208,
      "reward_std": 0.2963717132806778,
      "reward_std_all": 0.2796526253223419,
      "rewards-max/forensics_sinkhorn_reward": 0.9771620631217957,
      "rewards-max/forensics_strict_edge_reward": 0.3399255871772766,
      "rewards-max/forensics_tiou_reward": 0.9620621502399445,
      "rewards/forensics_sinkhorn_reward": 0.8089400231838226,
      "rewards/forensics_strict_edge_reward": 0.07237237133085728,
      "rewards/forensics_tiou_reward": 0.6702787578105927,
      "step": 992
    },
    {
      "completion_length": 23.03125,
      "epoch": 3.046082949308756,
      "grad_norm": 9.200197219848633,
      "kl": 0.3046875,
      "learning_rate": 2.369230769230769e-07,
      "loss": -0.0508,
      "reward": 1.4727312922477722,
      "reward_std": 0.18045996129512787,
      "reward_std_all": 0.19287830591201782,
      "rewards-max/forensics_sinkhorn_reward": 0.9673849940299988,
      "rewards-max/forensics_strict_edge_reward": 0.5134025514125824,
      "rewards-max/forensics_tiou_reward": 0.9398858845233917,
      "rewards/forensics_sinkhorn_reward": 0.762938529253006,
      "rewards/forensics_strict_edge_reward": 0.07148137874901295,
      "rewards/forensics_tiou_reward": 0.6383114755153656,
      "step": 993
    },
    {
      "completion_length": 22.34375,
      "epoch": 3.0491551459293396,
      "grad_norm": 14.138434410095215,
      "kl": 0.294921875,
      "learning_rate": 2.3615384615384615e-07,
      "loss": 0.0578,
      "reward": 1.3811296224594116,
      "reward_std": 0.21996186673641205,
      "reward_std_all": 0.19474390149116516,
      "rewards-max/forensics_sinkhorn_reward": 0.9601520895957947,
      "rewards-max/forensics_strict_edge_reward": 0.5321668684482574,
      "rewards-max/forensics_tiou_reward": 0.9296753108501434,
      "rewards/forensics_sinkhorn_reward": 0.7589812278747559,
      "rewards/forensics_strict_edge_reward": 0.04866848140954971,
      "rewards/forensics_tiou_reward": 0.573479950428009,
      "step": 994
    },
    {
      "completion_length": 20.625,
      "epoch": 3.0522273425499233,
      "grad_norm": 7.924882888793945,
      "kl": 0.30859375,
      "learning_rate": 2.3538461538461537e-07,
      "loss": -0.0167,
      "reward": 1.39179128408432,
      "reward_std": 0.24578729271888733,
      "reward_std_all": 0.22334877401590347,
      "rewards-max/forensics_sinkhorn_reward": 0.954557329416275,
      "rewards-max/forensics_strict_edge_reward": 0.373710498213768,
      "rewards-max/forensics_tiou_reward": 0.9140744507312775,
      "rewards/forensics_sinkhorn_reward": 0.7681556940078735,
      "rewards/forensics_strict_edge_reward": 0.05599244497716427,
      "rewards/forensics_tiou_reward": 0.5676431357860565,
      "step": 995
    },
    {
      "completion_length": 22.15625,
      "epoch": 3.055299539170507,
      "grad_norm": 10.588197708129883,
      "kl": 0.25048828125,
      "learning_rate": 2.346153846153846e-07,
      "loss": 0.0301,
      "reward": 1.4689775705337524,
      "reward_std": 0.35105355083942413,
      "reward_std_all": 0.3136649429798126,
      "rewards-max/forensics_sinkhorn_reward": 0.9810528457164764,
      "rewards-max/forensics_strict_edge_reward": 0.8582656681537628,
      "rewards-max/forensics_tiou_reward": 0.9819550514221191,
      "rewards/forensics_sinkhorn_reward": 0.7679394483566284,
      "rewards/forensics_strict_edge_reward": 0.0812872163951397,
      "rewards/forensics_tiou_reward": 0.6197509169578552,
      "step": 996
    },
    {
      "completion_length": 22.4375,
      "epoch": 3.0583717357910905,
      "grad_norm": 11.762428283691406,
      "kl": 0.306640625,
      "learning_rate": 2.3384615384615382e-07,
      "loss": 0.0207,
      "reward": 1.417456030845642,
      "reward_std": 0.28349485993385315,
      "reward_std_all": 0.2528010457754135,
      "rewards-max/forensics_sinkhorn_reward": 0.9651115536689758,
      "rewards-max/forensics_strict_edge_reward": 0.487403929233551,
      "rewards-max/forensics_tiou_reward": 0.9412116408348083,
      "rewards/forensics_sinkhorn_reward": 0.7695762515068054,
      "rewards/forensics_strict_edge_reward": 0.0610632486641407,
      "rewards/forensics_tiou_reward": 0.5868164896965027,
      "step": 997
    },
    {
      "completion_length": 21.96875,
      "epoch": 3.0614439324116742,
      "grad_norm": 11.682266235351562,
      "kl": 0.28515625,
      "learning_rate": 2.3307692307692306e-07,
      "loss": 0.0675,
      "reward": 1.2811124324798584,
      "reward_std": 0.2838660478591919,
      "reward_std_all": 0.24910957366228104,
      "rewards-max/forensics_sinkhorn_reward": 0.9525575637817383,
      "rewards-max/forensics_strict_edge_reward": 0.5063450932502747,
      "rewards-max/forensics_tiou_reward": 0.9074669778347015,
      "rewards/forensics_sinkhorn_reward": 0.7153915464878082,
      "rewards/forensics_strict_edge_reward": 0.06685873493552208,
      "rewards/forensics_tiou_reward": 0.4988621920347214,
      "step": 998
    },
    {
      "completion_length": 20.53125,
      "epoch": 3.064516129032258,
      "grad_norm": 15.35761833190918,
      "kl": 0.58349609375,
      "learning_rate": 2.323076923076923e-07,
      "loss": 0.0541,
      "reward": 1.4893994331359863,
      "reward_std": 0.21162202954292297,
      "reward_std_all": 0.1842338666319847,
      "rewards-max/forensics_sinkhorn_reward": 0.9725658297538757,
      "rewards-max/forensics_strict_edge_reward": 0.23573141545057297,
      "rewards-max/forensics_tiou_reward": 0.9558137655258179,
      "rewards/forensics_sinkhorn_reward": 0.8108330070972443,
      "rewards/forensics_strict_edge_reward": 0.03673736751079559,
      "rewards/forensics_tiou_reward": 0.6418288946151733,
      "step": 999
    },
    {
      "completion_length": 23.0625,
      "epoch": 3.067588325652842,
      "grad_norm": 11.526473045349121,
      "kl": 0.3828125,
      "learning_rate": 2.3153846153846151e-07,
      "loss": 0.0198,
      "reward": 1.5516103506088257,
      "reward_std": 0.26722755283117294,
      "reward_std_all": 0.2500587999820709,
      "rewards-max/forensics_sinkhorn_reward": 0.9780772626399994,
      "rewards-max/forensics_strict_edge_reward": 0.7445254027843475,
      "rewards-max/forensics_tiou_reward": 0.9741252362728119,
      "rewards/forensics_sinkhorn_reward": 0.8267060816287994,
      "rewards/forensics_strict_edge_reward": 0.06517668440937996,
      "rewards/forensics_tiou_reward": 0.6597275733947754,
      "step": 1000
    },
    {
      "completion_length": 21.484375,
      "epoch": 3.0706605222734256,
      "grad_norm": 9.279770851135254,
      "kl": 0.361328125,
      "learning_rate": 2.3076923076923078e-07,
      "loss": 0.0887,
      "reward": 1.2620832920074463,
      "reward_std": 0.257281132042408,
      "reward_std_all": 0.24288180470466614,
      "rewards-max/forensics_sinkhorn_reward": 0.9613000154495239,
      "rewards-max/forensics_strict_edge_reward": 0.22357474267482758,
      "rewards-max/forensics_tiou_reward": 0.9317683279514313,
      "rewards/forensics_sinkhorn_reward": 0.6950906813144684,
      "rewards/forensics_strict_edge_reward": 0.03599105216562748,
      "rewards/forensics_tiou_reward": 0.5310014933347702,
      "step": 1001
    },
    {
      "completion_length": 19.9375,
      "epoch": 3.0737327188940093,
      "grad_norm": 20.974868774414062,
      "kl": 0.36767578125,
      "learning_rate": 2.3e-07,
      "loss": 0.0754,
      "reward": 1.3521226644515991,
      "reward_std": 0.21751180291175842,
      "reward_std_all": 0.21068711578845978,
      "rewards-max/forensics_sinkhorn_reward": 0.9636455774307251,
      "rewards-max/forensics_strict_edge_reward": 0.48458099365234375,
      "rewards-max/forensics_tiou_reward": 0.940078467130661,
      "rewards/forensics_sinkhorn_reward": 0.7643064856529236,
      "rewards/forensics_strict_edge_reward": 0.04306875541806221,
      "rewards/forensics_tiou_reward": 0.5447474420070648,
      "step": 1002
    },
    {
      "completion_length": 21.890625,
      "epoch": 3.076804915514593,
      "grad_norm": 9.444611549377441,
      "kl": 0.48046875,
      "learning_rate": 2.2923076923076924e-07,
      "loss": -0.0163,
      "reward": 1.5092962384223938,
      "reward_std": 0.2166134938597679,
      "reward_std_all": 0.19996461272239685,
      "rewards-max/forensics_sinkhorn_reward": 0.9759024083614349,
      "rewards-max/forensics_strict_edge_reward": 0.4728233367204666,
      "rewards-max/forensics_tiou_reward": 0.9678993225097656,
      "rewards/forensics_sinkhorn_reward": 0.8073884844779968,
      "rewards/forensics_strict_edge_reward": 0.06836395896971226,
      "rewards/forensics_tiou_reward": 0.6335437893867493,
      "step": 1003
    },
    {
      "completion_length": 21.65625,
      "epoch": 3.0798771121351765,
      "grad_norm": 12.364532470703125,
      "kl": 0.291015625,
      "learning_rate": 2.2846153846153845e-07,
      "loss": -0.0089,
      "reward": 1.3075904250144958,
      "reward_std": 0.1963944286108017,
      "reward_std_all": 0.17534375935792923,
      "rewards-max/forensics_sinkhorn_reward": 0.9544074535369873,
      "rewards-max/forensics_strict_edge_reward": 0.3277985006570816,
      "rewards-max/forensics_tiou_reward": 0.9058031141757965,
      "rewards/forensics_sinkhorn_reward": 0.7352415323257446,
      "rewards/forensics_strict_edge_reward": 0.04182937927544117,
      "rewards/forensics_tiou_reward": 0.5305195301771164,
      "step": 1004
    },
    {
      "completion_length": 21.84375,
      "epoch": 3.0829493087557602,
      "grad_norm": 16.70408058166504,
      "kl": 0.30859375,
      "learning_rate": 2.276923076923077e-07,
      "loss": 0.0401,
      "reward": 1.573967695236206,
      "reward_std": 0.2953896224498749,
      "reward_std_all": 0.2926088869571686,
      "rewards-max/forensics_sinkhorn_reward": 0.9814671576023102,
      "rewards-max/forensics_strict_edge_reward": 0.5677289068698883,
      "rewards-max/forensics_tiou_reward": 0.9787665605545044,
      "rewards/forensics_sinkhorn_reward": 0.8183388113975525,
      "rewards/forensics_strict_edge_reward": 0.10082105174660683,
      "rewards/forensics_tiou_reward": 0.6548079252243042,
      "step": 1005
    },
    {
      "completion_length": 20.234375,
      "epoch": 3.086021505376344,
      "grad_norm": 13.138063430786133,
      "kl": 0.296875,
      "learning_rate": 2.269230769230769e-07,
      "loss": -0.045,
      "reward": 1.53023761510849,
      "reward_std": 0.23449838906526566,
      "reward_std_all": 0.20658448338508606,
      "rewards-max/forensics_sinkhorn_reward": 0.9814994037151337,
      "rewards-max/forensics_strict_edge_reward": 0.606867641210556,
      "rewards-max/forensics_tiou_reward": 0.978020966053009,
      "rewards/forensics_sinkhorn_reward": 0.8056087493896484,
      "rewards/forensics_strict_edge_reward": 0.08701402693986893,
      "rewards/forensics_tiou_reward": 0.6376148462295532,
      "step": 1006
    },
    {
      "completion_length": 24.421875,
      "epoch": 3.089093701996928,
      "grad_norm": 11.384961128234863,
      "kl": 0.2880859375,
      "learning_rate": 2.2615384615384614e-07,
      "loss": -0.0634,
      "reward": 1.5353730916976929,
      "reward_std": 0.21050862967967987,
      "reward_std_all": 0.20847298949956894,
      "rewards-max/forensics_sinkhorn_reward": 0.9751096665859222,
      "rewards-max/forensics_strict_edge_reward": 0.47466468811035156,
      "rewards-max/forensics_tiou_reward": 0.960218220949173,
      "rewards/forensics_sinkhorn_reward": 0.8126778900623322,
      "rewards/forensics_strict_edge_reward": 0.0760817602276802,
      "rewards/forensics_tiou_reward": 0.646613359451294,
      "step": 1007
    },
    {
      "completion_length": 20.828125,
      "epoch": 3.0921658986175116,
      "grad_norm": 14.859304428100586,
      "kl": 0.50390625,
      "learning_rate": 2.2538461538461536e-07,
      "loss": 0.0753,
      "reward": 1.3385884165763855,
      "reward_std": 0.2641366049647331,
      "reward_std_all": 0.2636273577809334,
      "rewards-max/forensics_sinkhorn_reward": 0.9738365411758423,
      "rewards-max/forensics_strict_edge_reward": 0.43268609046936035,
      "rewards-max/forensics_tiou_reward": 0.9575579166412354,
      "rewards/forensics_sinkhorn_reward": 0.7447781264781952,
      "rewards/forensics_strict_edge_reward": 0.05307728797197342,
      "rewards/forensics_tiou_reward": 0.5407329201698303,
      "step": 1008
    },
    {
      "completion_length": 23.453125,
      "epoch": 3.0952380952380953,
      "grad_norm": 14.427529335021973,
      "kl": 0.35546875,
      "learning_rate": 2.246153846153846e-07,
      "loss": 0.0501,
      "reward": 1.2756794691085815,
      "reward_std": 0.36738942563533783,
      "reward_std_all": 0.32982011139392853,
      "rewards-max/forensics_sinkhorn_reward": 0.9733599126338959,
      "rewards-max/forensics_strict_edge_reward": 0.3240565210580826,
      "rewards-max/forensics_tiou_reward": 0.9558500647544861,
      "rewards/forensics_sinkhorn_reward": 0.6940663158893585,
      "rewards/forensics_strict_edge_reward": 0.04534011520445347,
      "rewards/forensics_tiou_reward": 0.5362730026245117,
      "step": 1009
    },
    {
      "completion_length": 23.40625,
      "epoch": 3.098310291858679,
      "grad_norm": 13.271814346313477,
      "kl": 0.47998046875,
      "learning_rate": 2.238461538461538e-07,
      "loss": -0.0434,
      "reward": 1.6581478118896484,
      "reward_std": 0.3206551969051361,
      "reward_std_all": 0.2837671488523483,
      "rewards-max/forensics_sinkhorn_reward": 0.9843640327453613,
      "rewards-max/forensics_strict_edge_reward": 0.6779094636440277,
      "rewards-max/forensics_tiou_reward": 0.9829245805740356,
      "rewards/forensics_sinkhorn_reward": 0.8527721762657166,
      "rewards/forensics_strict_edge_reward": 0.08960932865738869,
      "rewards/forensics_tiou_reward": 0.7157663106918335,
      "step": 1010
    },
    {
      "completion_length": 24.25,
      "epoch": 3.1013824884792625,
      "grad_norm": 11.175384521484375,
      "kl": 0.4404296875,
      "learning_rate": 2.2307692307692308e-07,
      "loss": 0.0193,
      "reward": 1.5622703433036804,
      "reward_std": 0.3765622228384018,
      "reward_std_all": 0.3485147953033447,
      "rewards-max/forensics_sinkhorn_reward": 0.9779841601848602,
      "rewards-max/forensics_strict_edge_reward": 0.7709200084209442,
      "rewards-max/forensics_tiou_reward": 0.9696208536624908,
      "rewards/forensics_sinkhorn_reward": 0.7754855751991272,
      "rewards/forensics_strict_edge_reward": 0.1626761108636856,
      "rewards/forensics_tiou_reward": 0.624108761548996,
      "step": 1011
    },
    {
      "completion_length": 20.53125,
      "epoch": 3.1044546850998462,
      "grad_norm": 13.179275512695312,
      "kl": 0.5322265625,
      "learning_rate": 2.2230769230769232e-07,
      "loss": 0.0059,
      "reward": 1.424791693687439,
      "reward_std": 0.27637097239494324,
      "reward_std_all": 0.24621835350990295,
      "rewards-max/forensics_sinkhorn_reward": 0.9780674874782562,
      "rewards-max/forensics_strict_edge_reward": 0.5093767195940018,
      "rewards-max/forensics_tiou_reward": 0.9731379449367523,
      "rewards/forensics_sinkhorn_reward": 0.794201672077179,
      "rewards/forensics_strict_edge_reward": 0.056245969608426094,
      "rewards/forensics_tiou_reward": 0.5743440687656403,
      "step": 1012
    },
    {
      "completion_length": 20.140625,
      "epoch": 3.10752688172043,
      "grad_norm": 11.238017082214355,
      "kl": 0.4765625,
      "learning_rate": 2.2153846153846153e-07,
      "loss": 0.0376,
      "reward": 1.452987551689148,
      "reward_std": 0.20386452227830887,
      "reward_std_all": 0.19609671086072922,
      "rewards-max/forensics_sinkhorn_reward": 0.9781023859977722,
      "rewards-max/forensics_strict_edge_reward": 0.5294047147035599,
      "rewards-max/forensics_tiou_reward": 0.9700512588024139,
      "rewards/forensics_sinkhorn_reward": 0.7883738875389099,
      "rewards/forensics_strict_edge_reward": 0.05350521206855774,
      "rewards/forensics_tiou_reward": 0.6111083626747131,
      "step": 1013
    },
    {
      "completion_length": 26.53125,
      "epoch": 3.110599078341014,
      "grad_norm": 6.227747440338135,
      "kl": 0.3095703125,
      "learning_rate": 2.2076923076923077e-07,
      "loss": 0.0491,
      "reward": 1.640174388885498,
      "reward_std": 0.26112572103738785,
      "reward_std_all": 0.2723454535007477,
      "rewards-max/forensics_sinkhorn_reward": 0.98528853058815,
      "rewards-max/forensics_strict_edge_reward": 0.4811747968196869,
      "rewards-max/forensics_tiou_reward": 0.9897549450397491,
      "rewards/forensics_sinkhorn_reward": 0.8392227590084076,
      "rewards/forensics_strict_edge_reward": 0.10149627551436424,
      "rewards/forensics_tiou_reward": 0.6994554996490479,
      "step": 1014
    },
    {
      "completion_length": 24.140625,
      "epoch": 3.1136712749615976,
      "grad_norm": 14.88934326171875,
      "kl": 0.482421875,
      "learning_rate": 2.1999999999999998e-07,
      "loss": 0.1836,
      "reward": 1.450058937072754,
      "reward_std": 0.2798813506960869,
      "reward_std_all": 0.2695840448141098,
      "rewards-max/forensics_sinkhorn_reward": 0.9738147556781769,
      "rewards-max/forensics_strict_edge_reward": 0.25453535467386246,
      "rewards-max/forensics_tiou_reward": 0.9572103023529053,
      "rewards/forensics_sinkhorn_reward": 0.793156772851944,
      "rewards/forensics_strict_edge_reward": 0.025309564545750618,
      "rewards/forensics_tiou_reward": 0.6315926313400269,
      "step": 1015
    },
    {
      "completion_length": 21.484375,
      "epoch": 3.1167434715821813,
      "grad_norm": 13.5647554397583,
      "kl": 0.32421875,
      "learning_rate": 2.1923076923076922e-07,
      "loss": -0.0254,
      "reward": 1.5281705260276794,
      "reward_std": 0.3123071938753128,
      "reward_std_all": 0.2835005670785904,
      "rewards-max/forensics_sinkhorn_reward": 0.9738141596317291,
      "rewards-max/forensics_strict_edge_reward": 0.5145584642887115,
      "rewards-max/forensics_tiou_reward": 0.9620360732078552,
      "rewards/forensics_sinkhorn_reward": 0.7973181903362274,
      "rewards/forensics_strict_edge_reward": 0.07889657095074654,
      "rewards/forensics_tiou_reward": 0.6519557535648346,
      "step": 1016
    },
    {
      "completion_length": 22.078125,
      "epoch": 3.119815668202765,
      "grad_norm": 7.5157389640808105,
      "kl": 0.4814453125,
      "learning_rate": 2.1846153846153844e-07,
      "loss": 0.0857,
      "reward": 1.3843470215797424,
      "reward_std": 0.28486186265945435,
      "reward_std_all": 0.26933513581752777,
      "rewards-max/forensics_sinkhorn_reward": 0.9681433141231537,
      "rewards-max/forensics_strict_edge_reward": 0.3420734480023384,
      "rewards-max/forensics_tiou_reward": 0.9454829692840576,
      "rewards/forensics_sinkhorn_reward": 0.7683299481868744,
      "rewards/forensics_strict_edge_reward": 0.055179400369524956,
      "rewards/forensics_tiou_reward": 0.5608377158641815,
      "step": 1017
    },
    {
      "completion_length": 23.734375,
      "epoch": 3.122887864823349,
      "grad_norm": 8.915888786315918,
      "kl": 0.396484375,
      "learning_rate": 2.1769230769230768e-07,
      "loss": 0.0866,
      "reward": 1.5488582253456116,
      "reward_std": 0.28348416090011597,
      "reward_std_all": 0.2564370259642601,
      "rewards-max/forensics_sinkhorn_reward": 0.9762686789035797,
      "rewards-max/forensics_strict_edge_reward": 0.6032288670539856,
      "rewards-max/forensics_tiou_reward": 0.9695180058479309,
      "rewards/forensics_sinkhorn_reward": 0.8136704862117767,
      "rewards/forensics_strict_edge_reward": 0.07215401157736778,
      "rewards/forensics_tiou_reward": 0.6630335748195648,
      "step": 1018
    },
    {
      "completion_length": 21.625,
      "epoch": 3.1259600614439322,
      "grad_norm": 11.832427978515625,
      "kl": 0.3642578125,
      "learning_rate": 2.169230769230769e-07,
      "loss": 0.0802,
      "reward": 1.3095468878746033,
      "reward_std": 0.2999027222394943,
      "reward_std_all": 0.26093870401382446,
      "rewards-max/forensics_sinkhorn_reward": 0.9358002543449402,
      "rewards-max/forensics_strict_edge_reward": 0.3342469185590744,
      "rewards-max/forensics_tiou_reward": 0.864954024553299,
      "rewards/forensics_sinkhorn_reward": 0.7191725075244904,
      "rewards/forensics_strict_edge_reward": 0.06273196451365948,
      "rewards/forensics_tiou_reward": 0.5276424586772919,
      "step": 1019
    },
    {
      "completion_length": 23.171875,
      "epoch": 3.129032258064516,
      "grad_norm": 12.71010971069336,
      "kl": 0.3154296875,
      "learning_rate": 2.1615384615384613e-07,
      "loss": -0.0735,
      "reward": 1.4956538677215576,
      "reward_std": 0.20072419941425323,
      "reward_std_all": 0.19325830042362213,
      "rewards-max/forensics_sinkhorn_reward": 0.9824314713478088,
      "rewards-max/forensics_strict_edge_reward": 0.42918433248996735,
      "rewards-max/forensics_tiou_reward": 0.9790819585323334,
      "rewards/forensics_sinkhorn_reward": 0.7936272323131561,
      "rewards/forensics_strict_edge_reward": 0.05988572910428047,
      "rewards/forensics_tiou_reward": 0.6421408653259277,
      "step": 1020
    },
    {
      "completion_length": 24.9375,
      "epoch": 3.1321044546851,
      "grad_norm": 10.56097412109375,
      "kl": 0.25927734375,
      "learning_rate": 2.153846153846154e-07,
      "loss": -0.0498,
      "reward": 1.3481075763702393,
      "reward_std": 0.23778808861970901,
      "reward_std_all": 0.22725866734981537,
      "rewards-max/forensics_sinkhorn_reward": 0.9756174087524414,
      "rewards-max/forensics_strict_edge_reward": 0.5500316619873047,
      "rewards-max/forensics_tiou_reward": 0.9671681225299835,
      "rewards/forensics_sinkhorn_reward": 0.751875102519989,
      "rewards/forensics_strict_edge_reward": 0.046663036569952965,
      "rewards/forensics_tiou_reward": 0.5495694577693939,
      "step": 1021
    },
    {
      "completion_length": 21.28125,
      "epoch": 3.1351766513056836,
      "grad_norm": 9.962472915649414,
      "kl": 0.29833984375,
      "learning_rate": 2.146153846153846e-07,
      "loss": 0.0414,
      "reward": 1.2016160488128662,
      "reward_std": 0.2542303055524826,
      "reward_std_all": 0.2272879183292389,
      "rewards-max/forensics_sinkhorn_reward": 0.966572105884552,
      "rewards-max/forensics_strict_edge_reward": 0.27899545431137085,
      "rewards-max/forensics_tiou_reward": 0.9375788867473602,
      "rewards/forensics_sinkhorn_reward": 0.6772758662700653,
      "rewards/forensics_strict_edge_reward": 0.031916745007038116,
      "rewards/forensics_tiou_reward": 0.49242353439331055,
      "step": 1022
    },
    {
      "completion_length": 23.671875,
      "epoch": 3.1382488479262673,
      "grad_norm": 13.049381256103516,
      "kl": 0.23681640625,
      "learning_rate": 2.1384615384615385e-07,
      "loss": 0.0026,
      "reward": 1.4942212104797363,
      "reward_std": 0.22200937569141388,
      "reward_std_all": 0.20518619567155838,
      "rewards-max/forensics_sinkhorn_reward": 0.9674708545207977,
      "rewards-max/forensics_strict_edge_reward": 0.75,
      "rewards-max/forensics_tiou_reward": 0.9436276257038116,
      "rewards/forensics_sinkhorn_reward": 0.7767207324504852,
      "rewards/forensics_strict_edge_reward": 0.12709021382033825,
      "rewards/forensics_tiou_reward": 0.5904103219509125,
      "step": 1023
    },
    {
      "completion_length": 20.984375,
      "epoch": 3.141321044546851,
      "grad_norm": 11.689162254333496,
      "kl": 0.50341796875,
      "learning_rate": 2.1307692307692307e-07,
      "loss": -0.0231,
      "reward": 1.6026581525802612,
      "reward_std": 0.27631908655166626,
      "reward_std_all": 0.2755056470632553,
      "rewards-max/forensics_sinkhorn_reward": 0.9826642274856567,
      "rewards-max/forensics_strict_edge_reward": 0.5955080687999725,
      "rewards-max/forensics_tiou_reward": 0.9817019104957581,
      "rewards/forensics_sinkhorn_reward": 0.8228223025798798,
      "rewards/forensics_strict_edge_reward": 0.10655912011861801,
      "rewards/forensics_tiou_reward": 0.6732766628265381,
      "step": 1024
    },
    {
      "completion_length": 19.375,
      "epoch": 3.144393241167435,
      "grad_norm": 17.01479721069336,
      "kl": 0.3720703125,
      "learning_rate": 2.123076923076923e-07,
      "loss": 0.1165,
      "reward": 1.4595385789871216,
      "reward_std": 0.3016910254955292,
      "reward_std_all": 0.30003052949905396,
      "rewards-max/forensics_sinkhorn_reward": 0.9777629971504211,
      "rewards-max/forensics_strict_edge_reward": 0.5984183251857758,
      "rewards-max/forensics_tiou_reward": 0.9663439691066742,
      "rewards/forensics_sinkhorn_reward": 0.7788327634334564,
      "rewards/forensics_strict_edge_reward": 0.08661065436899662,
      "rewards/forensics_tiou_reward": 0.5940951406955719,
      "step": 1025
    },
    {
      "completion_length": 24.515625,
      "epoch": 3.1474654377880182,
      "grad_norm": 13.511968612670898,
      "kl": 0.66015625,
      "learning_rate": 2.1153846153846152e-07,
      "loss": 0.0599,
      "reward": 1.3976624608039856,
      "reward_std": 0.19531627744436264,
      "reward_std_all": 0.17330075800418854,
      "rewards-max/forensics_sinkhorn_reward": 0.9764833748340607,
      "rewards-max/forensics_strict_edge_reward": 0.43131066765636206,
      "rewards-max/forensics_tiou_reward": 0.968149870634079,
      "rewards/forensics_sinkhorn_reward": 0.7477349638938904,
      "rewards/forensics_strict_edge_reward": 0.0611686990596354,
      "rewards/forensics_tiou_reward": 0.5887588262557983,
      "step": 1026
    },
    {
      "completion_length": 23.65625,
      "epoch": 3.150537634408602,
      "grad_norm": 13.35755729675293,
      "kl": 0.578125,
      "learning_rate": 2.1076923076923076e-07,
      "loss": 0.0535,
      "reward": 1.4297797083854675,
      "reward_std": 0.2745879292488098,
      "reward_std_all": 0.26615093648433685,
      "rewards-max/forensics_sinkhorn_reward": 0.9634766578674316,
      "rewards-max/forensics_strict_edge_reward": 0.37659838050603867,
      "rewards-max/forensics_tiou_reward": 0.940322607755661,
      "rewards/forensics_sinkhorn_reward": 0.7722096741199493,
      "rewards/forensics_strict_edge_reward": 0.05855295620858669,
      "rewards/forensics_tiou_reward": 0.5990170538425446,
      "step": 1027
    },
    {
      "completion_length": 19.65625,
      "epoch": 3.153609831029186,
      "grad_norm": 12.714494705200195,
      "kl": 0.3759765625,
      "learning_rate": 2.0999999999999997e-07,
      "loss": -0.0016,
      "reward": 1.4511354565620422,
      "reward_std": 0.2979995161294937,
      "reward_std_all": 0.2616409957408905,
      "rewards-max/forensics_sinkhorn_reward": 0.9617739319801331,
      "rewards-max/forensics_strict_edge_reward": 0.3958035260438919,
      "rewards-max/forensics_tiou_reward": 0.9172552525997162,
      "rewards/forensics_sinkhorn_reward": 0.7807036936283112,
      "rewards/forensics_strict_edge_reward": 0.06127633713185787,
      "rewards/forensics_tiou_reward": 0.6091553568840027,
      "step": 1028
    },
    {
      "completion_length": 23.4375,
      "epoch": 3.1566820276497696,
      "grad_norm": 13.386408805847168,
      "kl": 0.396484375,
      "learning_rate": 2.0923076923076921e-07,
      "loss": 0.0092,
      "reward": 1.2727947235107422,
      "reward_std": 0.1587974652647972,
      "reward_std_all": 0.19168463349342346,
      "rewards-max/forensics_sinkhorn_reward": 0.9681945145130157,
      "rewards-max/forensics_strict_edge_reward": 0.2649521976709366,
      "rewards-max/forensics_tiou_reward": 0.944582611322403,
      "rewards/forensics_sinkhorn_reward": 0.7169187366962433,
      "rewards/forensics_strict_edge_reward": 0.034249489195644855,
      "rewards/forensics_tiou_reward": 0.5216265916824341,
      "step": 1029
    },
    {
      "completion_length": 21.546875,
      "epoch": 3.1597542242703534,
      "grad_norm": 12.388603210449219,
      "kl": 0.38037109375,
      "learning_rate": 2.0846153846153843e-07,
      "loss": 0.184,
      "reward": 1.4570367932319641,
      "reward_std": 0.32051512598991394,
      "reward_std_all": 0.28683262318372726,
      "rewards-max/forensics_sinkhorn_reward": 0.9779479205608368,
      "rewards-max/forensics_strict_edge_reward": 0.5686750710010529,
      "rewards-max/forensics_tiou_reward": 0.9699316024780273,
      "rewards/forensics_sinkhorn_reward": 0.7816829979419708,
      "rewards/forensics_strict_edge_reward": 0.08044182509183884,
      "rewards/forensics_tiou_reward": 0.5949119627475739,
      "step": 1030
    },
    {
      "completion_length": 22.765625,
      "epoch": 3.162826420890937,
      "grad_norm": 11.474058151245117,
      "kl": 0.27490234375,
      "learning_rate": 2.076923076923077e-07,
      "loss": -0.0124,
      "reward": 1.5793802738189697,
      "reward_std": 0.23655399680137634,
      "reward_std_all": 0.21220355480909348,
      "rewards-max/forensics_sinkhorn_reward": 0.973344624042511,
      "rewards-max/forensics_strict_edge_reward": 0.5965884327888489,
      "rewards-max/forensics_tiou_reward": 0.9616615474224091,
      "rewards/forensics_sinkhorn_reward": 0.8231483697891235,
      "rewards/forensics_strict_edge_reward": 0.09736024215817451,
      "rewards/forensics_tiou_reward": 0.6588715016841888,
      "step": 1031
    },
    {
      "completion_length": 24.296875,
      "epoch": 3.165898617511521,
      "grad_norm": 11.565089225769043,
      "kl": 0.2626953125,
      "learning_rate": 2.0692307692307694e-07,
      "loss": -0.0532,
      "reward": 1.4878785014152527,
      "reward_std": 0.21963536739349365,
      "reward_std_all": 0.1974879577755928,
      "rewards-max/forensics_sinkhorn_reward": 0.969296783208847,
      "rewards-max/forensics_strict_edge_reward": 0.3239518702030182,
      "rewards-max/forensics_tiou_reward": 0.9502600431442261,
      "rewards/forensics_sinkhorn_reward": 0.7932643890380859,
      "rewards/forensics_strict_edge_reward": 0.06663582753390074,
      "rewards/forensics_tiou_reward": 0.6279783546924591,
      "step": 1032
    },
    {
      "completion_length": 22.15625,
      "epoch": 3.1689708141321047,
      "grad_norm": 8.992938041687012,
      "kl": 0.462890625,
      "learning_rate": 2.0615384615384615e-07,
      "loss": 0.0993,
      "reward": 1.324907660484314,
      "reward_std": 0.2904350161552429,
      "reward_std_all": 0.262290894985199,
      "rewards-max/forensics_sinkhorn_reward": 0.9716341495513916,
      "rewards-max/forensics_strict_edge_reward": 0.6087257117033005,
      "rewards-max/forensics_tiou_reward": 0.9546021819114685,
      "rewards/forensics_sinkhorn_reward": 0.7256454825401306,
      "rewards/forensics_strict_edge_reward": 0.0820884583517909,
      "rewards/forensics_tiou_reward": 0.5171736776828766,
      "step": 1033
    },
    {
      "completion_length": 23.546875,
      "epoch": 3.172043010752688,
      "grad_norm": 12.160707473754883,
      "kl": 0.4931640625,
      "learning_rate": 2.053846153846154e-07,
      "loss": -0.0383,
      "reward": 1.4588632583618164,
      "reward_std": 0.21960683166980743,
      "reward_std_all": 0.2391514629125595,
      "rewards-max/forensics_sinkhorn_reward": 0.9742502868175507,
      "rewards-max/forensics_strict_edge_reward": 0.5750451982021332,
      "rewards-max/forensics_tiou_reward": 0.9633205831050873,
      "rewards/forensics_sinkhorn_reward": 0.7809323966503143,
      "rewards/forensics_strict_edge_reward": 0.06747937388718128,
      "rewards/forensics_tiou_reward": 0.6104516088962555,
      "step": 1034
    },
    {
      "completion_length": 21.25,
      "epoch": 3.175115207373272,
      "grad_norm": 11.347728729248047,
      "kl": 0.3759765625,
      "learning_rate": 2.046153846153846e-07,
      "loss": 0.0509,
      "reward": 1.4459822177886963,
      "reward_std": 0.2545575276017189,
      "reward_std_all": 0.24009990692138672,
      "rewards-max/forensics_sinkhorn_reward": 0.9808336794376373,
      "rewards-max/forensics_strict_edge_reward": 0.48555219173431396,
      "rewards-max/forensics_tiou_reward": 0.9788329899311066,
      "rewards/forensics_sinkhorn_reward": 0.7836979627609253,
      "rewards/forensics_strict_edge_reward": 0.0706475842744112,
      "rewards/forensics_tiou_reward": 0.591636598110199,
      "step": 1035
    },
    {
      "completion_length": 20.171875,
      "epoch": 3.1781874039938556,
      "grad_norm": 12.011959075927734,
      "kl": 0.255859375,
      "learning_rate": 2.0384615384615384e-07,
      "loss": 0.0345,
      "reward": 1.3036642670631409,
      "reward_std": 0.1882207840681076,
      "reward_std_all": 0.17774666845798492,
      "rewards-max/forensics_sinkhorn_reward": 0.9447462260723114,
      "rewards-max/forensics_strict_edge_reward": 0.3413814939558506,
      "rewards-max/forensics_tiou_reward": 0.8952619731426239,
      "rewards/forensics_sinkhorn_reward": 0.7305355966091156,
      "rewards/forensics_strict_edge_reward": 0.0507693849503994,
      "rewards/forensics_tiou_reward": 0.522359311580658,
      "step": 1036
    },
    {
      "completion_length": 22.796875,
      "epoch": 3.1812596006144394,
      "grad_norm": 9.747697830200195,
      "kl": 0.3623046875,
      "learning_rate": 2.0307692307692306e-07,
      "loss": 0.0883,
      "reward": 1.3349088430404663,
      "reward_std": 0.19421860575675964,
      "reward_std_all": 0.19123149663209915,
      "rewards-max/forensics_sinkhorn_reward": 0.9673157632350922,
      "rewards-max/forensics_strict_edge_reward": 0.26833562925457954,
      "rewards-max/forensics_tiou_reward": 0.9447431266307831,
      "rewards/forensics_sinkhorn_reward": 0.7519942820072174,
      "rewards/forensics_strict_edge_reward": 0.03509973827749491,
      "rewards/forensics_tiou_reward": 0.5478148758411407,
      "step": 1037
    },
    {
      "completion_length": 25.296875,
      "epoch": 3.184331797235023,
      "grad_norm": 17.073223114013672,
      "kl": 0.251953125,
      "learning_rate": 2.023076923076923e-07,
      "loss": 0.0381,
      "reward": 1.3336769938468933,
      "reward_std": 0.1440938636660576,
      "reward_std_all": 0.1694328635931015,
      "rewards-max/forensics_sinkhorn_reward": 0.9641712605953217,
      "rewards-max/forensics_strict_edge_reward": 0.27244652435183525,
      "rewards-max/forensics_tiou_reward": 0.9383763372898102,
      "rewards/forensics_sinkhorn_reward": 0.7354166507720947,
      "rewards/forensics_strict_edge_reward": 0.04435076471418142,
      "rewards/forensics_tiou_reward": 0.553909569978714,
      "step": 1038
    },
    {
      "completion_length": 23.359375,
      "epoch": 3.187403993855607,
      "grad_norm": 10.061944961547852,
      "kl": 0.4853515625,
      "learning_rate": 2.015384615384615e-07,
      "loss": 0.0859,
      "reward": 1.4248017072677612,
      "reward_std": 0.2283472716808319,
      "reward_std_all": 0.23745792359113693,
      "rewards-max/forensics_sinkhorn_reward": 0.9571244418621063,
      "rewards-max/forensics_strict_edge_reward": 0.4200603812932968,
      "rewards-max/forensics_tiou_reward": 0.9191235601902008,
      "rewards/forensics_sinkhorn_reward": 0.7907922565937042,
      "rewards/forensics_strict_edge_reward": 0.044195229187607765,
      "rewards/forensics_tiou_reward": 0.5898141860961914,
      "step": 1039
    },
    {
      "completion_length": 23.515625,
      "epoch": 3.1904761904761907,
      "grad_norm": 13.63623332977295,
      "kl": 0.21875,
      "learning_rate": 2.0076923076923075e-07,
      "loss": -0.0958,
      "reward": 1.425912857055664,
      "reward_std": 0.19099175930023193,
      "reward_std_all": 0.20603343471884727,
      "rewards-max/forensics_sinkhorn_reward": 0.9801930785179138,
      "rewards-max/forensics_strict_edge_reward": 0.44576262682676315,
      "rewards-max/forensics_tiou_reward": 0.9769158661365509,
      "rewards/forensics_sinkhorn_reward": 0.7609174847602844,
      "rewards/forensics_strict_edge_reward": 0.1040600836277008,
      "rewards/forensics_tiou_reward": 0.5609352886676788,
      "step": 1040
    },
    {
      "completion_length": 22.921875,
      "epoch": 3.193548387096774,
      "grad_norm": 11.401023864746094,
      "kl": 0.283203125,
      "learning_rate": 2e-07,
      "loss": 0.0078,
      "reward": 1.4800165891647339,
      "reward_std": 0.22549059987068176,
      "reward_std_all": 0.20107775181531906,
      "rewards-max/forensics_sinkhorn_reward": 0.9771788716316223,
      "rewards-max/forensics_strict_edge_reward": 0.3144318461418152,
      "rewards-max/forensics_tiou_reward": 0.9673121869564056,
      "rewards/forensics_sinkhorn_reward": 0.813060462474823,
      "rewards/forensics_strict_edge_reward": 0.042396364733576775,
      "rewards/forensics_tiou_reward": 0.6245597898960114,
      "step": 1041
    },
    {
      "completion_length": 22.515625,
      "epoch": 3.196620583717358,
      "grad_norm": 12.043800354003906,
      "kl": 0.23828125,
      "learning_rate": 1.9923076923076923e-07,
      "loss": -0.0653,
      "reward": 1.49391770362854,
      "reward_std": 0.19325324147939682,
      "reward_std_all": 0.16670487821102142,
      "rewards-max/forensics_sinkhorn_reward": 0.965975433588028,
      "rewards-max/forensics_strict_edge_reward": 0.48601071536540985,
      "rewards-max/forensics_tiou_reward": 0.9462716281414032,
      "rewards/forensics_sinkhorn_reward": 0.7954668998718262,
      "rewards/forensics_strict_edge_reward": 0.07981464453041553,
      "rewards/forensics_tiou_reward": 0.6186362653970718,
      "step": 1042
    },
    {
      "completion_length": 22.03125,
      "epoch": 3.1996927803379416,
      "grad_norm": 12.659132957458496,
      "kl": 0.2451171875,
      "learning_rate": 1.9846153846153844e-07,
      "loss": 0.0178,
      "reward": 1.5503466129302979,
      "reward_std": 0.2466682642698288,
      "reward_std_all": 0.2478724718093872,
      "rewards-max/forensics_sinkhorn_reward": 0.9866662621498108,
      "rewards-max/forensics_strict_edge_reward": 0.9211947023868561,
      "rewards-max/forensics_tiou_reward": 0.9934326708316803,
      "rewards/forensics_sinkhorn_reward": 0.7918815016746521,
      "rewards/forensics_strict_edge_reward": 0.12720512598752975,
      "rewards/forensics_tiou_reward": 0.631259948015213,
      "step": 1043
    },
    {
      "completion_length": 25.015625,
      "epoch": 3.2027649769585254,
      "grad_norm": 11.18786334991455,
      "kl": 0.318359375,
      "learning_rate": 1.9769230769230768e-07,
      "loss": -0.148,
      "reward": 1.50338476896286,
      "reward_std": 0.24165025353431702,
      "reward_std_all": 0.2278166562318802,
      "rewards-max/forensics_sinkhorn_reward": 0.9783163666725159,
      "rewards-max/forensics_strict_edge_reward": 0.7197895646095276,
      "rewards-max/forensics_tiou_reward": 0.9711841940879822,
      "rewards/forensics_sinkhorn_reward": 0.77406245470047,
      "rewards/forensics_strict_edge_reward": 0.11153554171323776,
      "rewards/forensics_tiou_reward": 0.6177868247032166,
      "step": 1044
    },
    {
      "completion_length": 22.1875,
      "epoch": 3.205837173579109,
      "grad_norm": 14.630146980285645,
      "kl": 0.353515625,
      "learning_rate": 1.9692307692307693e-07,
      "loss": -0.0249,
      "reward": 1.26413232088089,
      "reward_std": 0.3115810602903366,
      "reward_std_all": 0.2722310423851013,
      "rewards-max/forensics_sinkhorn_reward": 0.9754873812198639,
      "rewards-max/forensics_strict_edge_reward": 0.29988496378064156,
      "rewards-max/forensics_tiou_reward": 0.956038773059845,
      "rewards/forensics_sinkhorn_reward": 0.6972275376319885,
      "rewards/forensics_strict_edge_reward": 0.04139811417553574,
      "rewards/forensics_tiou_reward": 0.5255066603422165,
      "step": 1045
    },
    {
      "completion_length": 22.921875,
      "epoch": 3.208909370199693,
      "grad_norm": 18.065109252929688,
      "kl": 0.376953125,
      "learning_rate": 1.9615384615384614e-07,
      "loss": 0.0566,
      "reward": 1.393079698085785,
      "reward_std": 0.2408355176448822,
      "reward_std_all": 0.24841118603944778,
      "rewards-max/forensics_sinkhorn_reward": 0.9820845127105713,
      "rewards-max/forensics_strict_edge_reward": 0.7659632563591003,
      "rewards-max/forensics_tiou_reward": 0.9822339117527008,
      "rewards/forensics_sinkhorn_reward": 0.7401426434516907,
      "rewards/forensics_strict_edge_reward": 0.06331642158329487,
      "rewards/forensics_tiou_reward": 0.5896207094192505,
      "step": 1046
    },
    {
      "completion_length": 20.984375,
      "epoch": 3.2119815668202767,
      "grad_norm": 15.436614036560059,
      "kl": 0.44580078125,
      "learning_rate": 1.9538461538461538e-07,
      "loss": 0.1106,
      "reward": 1.462855339050293,
      "reward_std": 0.2773725613951683,
      "reward_std_all": 0.24964144825935364,
      "rewards-max/forensics_sinkhorn_reward": 0.9799505770206451,
      "rewards-max/forensics_strict_edge_reward": 0.5706980228424072,
      "rewards-max/forensics_tiou_reward": 0.9763006567955017,
      "rewards/forensics_sinkhorn_reward": 0.7870705723762512,
      "rewards/forensics_strict_edge_reward": 0.07623794302344322,
      "rewards/forensics_tiou_reward": 0.5995468497276306,
      "step": 1047
    },
    {
      "completion_length": 20.46875,
      "epoch": 3.21505376344086,
      "grad_norm": 13.847407341003418,
      "kl": 0.26123046875,
      "learning_rate": 1.946153846153846e-07,
      "loss": -0.1164,
      "reward": 1.5278599858283997,
      "reward_std": 0.23998793959617615,
      "reward_std_all": 0.21719762682914734,
      "rewards-max/forensics_sinkhorn_reward": 0.9762793481349945,
      "rewards-max/forensics_strict_edge_reward": 0.7388437688350677,
      "rewards-max/forensics_tiou_reward": 0.9643381237983704,
      "rewards/forensics_sinkhorn_reward": 0.7915265560150146,
      "rewards/forensics_strict_edge_reward": 0.11206614971160889,
      "rewards/forensics_tiou_reward": 0.6242672204971313,
      "step": 1048
    },
    {
      "completion_length": 20.96875,
      "epoch": 3.218125960061444,
      "grad_norm": 9.430074691772461,
      "kl": 0.421875,
      "learning_rate": 1.9384615384615383e-07,
      "loss": 0.1099,
      "reward": 1.2906277775764465,
      "reward_std": 0.2953309863805771,
      "reward_std_all": 0.25661203265190125,
      "rewards-max/forensics_sinkhorn_reward": 0.9714698195457458,
      "rewards-max/forensics_strict_edge_reward": 0.6342410445213318,
      "rewards-max/forensics_tiou_reward": 0.9550639092922211,
      "rewards/forensics_sinkhorn_reward": 0.7114287912845612,
      "rewards/forensics_strict_edge_reward": 0.08191436901688576,
      "rewards/forensics_tiou_reward": 0.49728457629680634,
      "step": 1049
    },
    {
      "completion_length": 24.203125,
      "epoch": 3.2211981566820276,
      "grad_norm": 11.76928997039795,
      "kl": 0.2548828125,
      "learning_rate": 1.9307692307692307e-07,
      "loss": 0.0104,
      "reward": 1.6811891198158264,
      "reward_std": 0.25165707617998123,
      "reward_std_all": 0.2730900049209595,
      "rewards-max/forensics_sinkhorn_reward": 0.9788590967655182,
      "rewards-max/forensics_strict_edge_reward": 0.6515887677669525,
      "rewards-max/forensics_tiou_reward": 0.9741096794605255,
      "rewards/forensics_sinkhorn_reward": 0.8392674028873444,
      "rewards/forensics_strict_edge_reward": 0.13141454011201859,
      "rewards/forensics_tiou_reward": 0.7105071842670441,
      "step": 1050
    },
    {
      "completion_length": 20.65625,
      "epoch": 3.2242703533026114,
      "grad_norm": 14.217641830444336,
      "kl": 0.5556640625,
      "learning_rate": 1.9230769230769231e-07,
      "loss": 0.0892,
      "reward": 1.4528328776359558,
      "reward_std": 0.23146717995405197,
      "reward_std_all": 0.2802664339542389,
      "rewards-max/forensics_sinkhorn_reward": 0.982891172170639,
      "rewards-max/forensics_strict_edge_reward": 0.5829301476478577,
      "rewards-max/forensics_tiou_reward": 0.9826687872409821,
      "rewards/forensics_sinkhorn_reward": 0.7629619538784027,
      "rewards/forensics_strict_edge_reward": 0.0798509307205677,
      "rewards/forensics_tiou_reward": 0.6100200414657593,
      "step": 1051
    },
    {
      "completion_length": 22.15625,
      "epoch": 3.227342549923195,
      "grad_norm": 10.263608932495117,
      "kl": 0.23681640625,
      "learning_rate": 1.9153846153846153e-07,
      "loss": -0.0269,
      "reward": 1.4583678245544434,
      "reward_std": 0.2468814253807068,
      "reward_std_all": 0.22670665383338928,
      "rewards-max/forensics_sinkhorn_reward": 0.9717090129852295,
      "rewards-max/forensics_strict_edge_reward": 0.4838104844093323,
      "rewards-max/forensics_tiou_reward": 0.9607217013835907,
      "rewards/forensics_sinkhorn_reward": 0.7832086682319641,
      "rewards/forensics_strict_edge_reward": 0.05726510751992464,
      "rewards/forensics_tiou_reward": 0.6178940832614899,
      "step": 1052
    },
    {
      "completion_length": 23.09375,
      "epoch": 3.230414746543779,
      "grad_norm": 11.498668670654297,
      "kl": 0.4765625,
      "learning_rate": 1.9076923076923077e-07,
      "loss": -0.0455,
      "reward": 1.3641985058784485,
      "reward_std": 0.21414832770824432,
      "reward_std_all": 0.2006305530667305,
      "rewards-max/forensics_sinkhorn_reward": 0.9689917266368866,
      "rewards-max/forensics_strict_edge_reward": 0.2893172800540924,
      "rewards-max/forensics_tiou_reward": 0.9492744505405426,
      "rewards/forensics_sinkhorn_reward": 0.7586384415626526,
      "rewards/forensics_strict_edge_reward": 0.052266634767875075,
      "rewards/forensics_tiou_reward": 0.5532935112714767,
      "step": 1053
    },
    {
      "completion_length": 22.140625,
      "epoch": 3.2334869431643627,
      "grad_norm": 12.065277099609375,
      "kl": 0.36328125,
      "learning_rate": 1.8999999999999998e-07,
      "loss": -0.0054,
      "reward": 1.5006370544433594,
      "reward_std": 0.2961782366037369,
      "reward_std_all": 0.2620653137564659,
      "rewards-max/forensics_sinkhorn_reward": 0.9659278988838196,
      "rewards-max/forensics_strict_edge_reward": 0.5844750255346298,
      "rewards-max/forensics_tiou_reward": 0.9391155242919922,
      "rewards/forensics_sinkhorn_reward": 0.7882046401500702,
      "rewards/forensics_strict_edge_reward": 0.07372118160128593,
      "rewards/forensics_tiou_reward": 0.6387111842632294,
      "step": 1054
    },
    {
      "completion_length": 22.15625,
      "epoch": 3.236559139784946,
      "grad_norm": 10.060911178588867,
      "kl": 0.29296875,
      "learning_rate": 1.8923076923076922e-07,
      "loss": 0.093,
      "reward": 1.34318345785141,
      "reward_std": 0.15314709395170212,
      "reward_std_all": 0.19104410707950592,
      "rewards-max/forensics_sinkhorn_reward": 0.9825005829334259,
      "rewards-max/forensics_strict_edge_reward": 0.5599738955497742,
      "rewards-max/forensics_tiou_reward": 0.9834116101264954,
      "rewards/forensics_sinkhorn_reward": 0.7387011647224426,
      "rewards/forensics_strict_edge_reward": 0.06815196573734283,
      "rewards/forensics_tiou_reward": 0.5363303422927856,
      "step": 1055
    },
    {
      "completion_length": 22.453125,
      "epoch": 3.23963133640553,
      "grad_norm": 14.317773818969727,
      "kl": 0.4404296875,
      "learning_rate": 1.8846153846153846e-07,
      "loss": 0.0936,
      "reward": 1.2380120158195496,
      "reward_std": 0.19610323011875153,
      "reward_std_all": 0.23057082295417786,
      "rewards-max/forensics_sinkhorn_reward": 0.9353845715522766,
      "rewards-max/forensics_strict_edge_reward": 0.2644799053668976,
      "rewards-max/forensics_tiou_reward": 0.8630131185054779,
      "rewards/forensics_sinkhorn_reward": 0.6854995787143707,
      "rewards/forensics_strict_edge_reward": 0.03189989924430847,
      "rewards/forensics_tiou_reward": 0.5206125676631927,
      "step": 1056
    },
    {
      "completion_length": 23.1875,
      "epoch": 3.2427035330261136,
      "grad_norm": 15.642374038696289,
      "kl": 0.26025390625,
      "learning_rate": 1.8769230769230767e-07,
      "loss": -0.0208,
      "reward": 1.5081447958946228,
      "reward_std": 0.2984677404165268,
      "reward_std_all": 0.26124534010887146,
      "rewards-max/forensics_sinkhorn_reward": 0.9744012653827667,
      "rewards-max/forensics_strict_edge_reward": 0.6428841799497604,
      "rewards-max/forensics_tiou_reward": 0.9601799845695496,
      "rewards/forensics_sinkhorn_reward": 0.7562470734119415,
      "rewards/forensics_strict_edge_reward": 0.12412146851420403,
      "rewards/forensics_tiou_reward": 0.6277762949466705,
      "step": 1057
    },
    {
      "completion_length": 20.09375,
      "epoch": 3.2457757296466974,
      "grad_norm": 12.643536567687988,
      "kl": 0.3779296875,
      "learning_rate": 1.8692307692307692e-07,
      "loss": -0.0223,
      "reward": 1.4423682689666748,
      "reward_std": 0.2516558766365051,
      "reward_std_all": 0.21850769966840744,
      "rewards-max/forensics_sinkhorn_reward": 0.9679457247257233,
      "rewards-max/forensics_strict_edge_reward": 0.42784760892391205,
      "rewards-max/forensics_tiou_reward": 0.9484648406505585,
      "rewards/forensics_sinkhorn_reward": 0.783907949924469,
      "rewards/forensics_strict_edge_reward": 0.05540154315531254,
      "rewards/forensics_tiou_reward": 0.6030588150024414,
      "step": 1058
    },
    {
      "completion_length": 21.40625,
      "epoch": 3.248847926267281,
      "grad_norm": 15.648175239562988,
      "kl": 0.4765625,
      "learning_rate": 1.8615384615384613e-07,
      "loss": 0.1461,
      "reward": 1.4294722080230713,
      "reward_std": 0.24471653997898102,
      "reward_std_all": 0.27160659432411194,
      "rewards-max/forensics_sinkhorn_reward": 0.970976322889328,
      "rewards-max/forensics_strict_edge_reward": 0.5500316619873047,
      "rewards-max/forensics_tiou_reward": 0.9575011730194092,
      "rewards/forensics_sinkhorn_reward": 0.766831785440445,
      "rewards/forensics_strict_edge_reward": 0.07848497480154037,
      "rewards/forensics_tiou_reward": 0.5841554403305054,
      "step": 1059
    },
    {
      "completion_length": 22.640625,
      "epoch": 3.251920122887865,
      "grad_norm": 12.690512657165527,
      "kl": 0.3544921875,
      "learning_rate": 1.853846153846154e-07,
      "loss": -0.1598,
      "reward": 1.5267288088798523,
      "reward_std": 0.22339874505996704,
      "reward_std_all": 0.24907849729061127,
      "rewards-max/forensics_sinkhorn_reward": 0.9806263744831085,
      "rewards-max/forensics_strict_edge_reward": 0.5905451774597168,
      "rewards-max/forensics_tiou_reward": 0.9820835590362549,
      "rewards/forensics_sinkhorn_reward": 0.8033087849617004,
      "rewards/forensics_strict_edge_reward": 0.06935422867536545,
      "rewards/forensics_tiou_reward": 0.6540656983852386,
      "step": 1060
    },
    {
      "completion_length": 22.0,
      "epoch": 3.2549923195084487,
      "grad_norm": 9.785438537597656,
      "kl": 0.384765625,
      "learning_rate": 1.846153846153846e-07,
      "loss": 0.0133,
      "reward": 1.4366998672485352,
      "reward_std": 0.31149449944496155,
      "reward_std_all": 0.27360180020332336,
      "rewards-max/forensics_sinkhorn_reward": 0.9676340222358704,
      "rewards-max/forensics_strict_edge_reward": 0.7101751863956451,
      "rewards-max/forensics_tiou_reward": 0.9521320760250092,
      "rewards/forensics_sinkhorn_reward": 0.7702412009239197,
      "rewards/forensics_strict_edge_reward": 0.07482536695897579,
      "rewards/forensics_tiou_reward": 0.5916332006454468,
      "step": 1061
    },
    {
      "completion_length": 23.015625,
      "epoch": 3.258064516129032,
      "grad_norm": 14.599536895751953,
      "kl": 0.333984375,
      "learning_rate": 1.8384615384615385e-07,
      "loss": -0.0327,
      "reward": 1.304967999458313,
      "reward_std": 0.24110674858093262,
      "reward_std_all": 0.20948732644319534,
      "rewards-max/forensics_sinkhorn_reward": 0.9694620668888092,
      "rewards-max/forensics_strict_edge_reward": 0.5083713233470917,
      "rewards-max/forensics_tiou_reward": 0.951349139213562,
      "rewards/forensics_sinkhorn_reward": 0.7168587744235992,
      "rewards/forensics_strict_edge_reward": 0.05146513320505619,
      "rewards/forensics_tiou_reward": 0.5366440713405609,
      "step": 1062
    },
    {
      "completion_length": 20.34375,
      "epoch": 3.261136712749616,
      "grad_norm": 14.88631534576416,
      "kl": 0.2939453125,
      "learning_rate": 1.8307692307692306e-07,
      "loss": 0.2302,
      "reward": 1.3418598771095276,
      "reward_std": 0.2022145763039589,
      "reward_std_all": 0.19817745685577393,
      "rewards-max/forensics_sinkhorn_reward": 0.9676994979381561,
      "rewards-max/forensics_strict_edge_reward": 0.23576828837394714,
      "rewards-max/forensics_tiou_reward": 0.9361318051815033,
      "rewards/forensics_sinkhorn_reward": 0.7560109496116638,
      "rewards/forensics_strict_edge_reward": 0.02327996026724577,
      "rewards/forensics_tiou_reward": 0.5625689923763275,
      "step": 1063
    },
    {
      "completion_length": 23.421875,
      "epoch": 3.2642089093701996,
      "grad_norm": 10.91881275177002,
      "kl": 0.2353515625,
      "learning_rate": 1.823076923076923e-07,
      "loss": -0.2213,
      "reward": 1.340228259563446,
      "reward_std": 0.1431305669248104,
      "reward_std_all": 0.16850555688142776,
      "rewards-max/forensics_sinkhorn_reward": 0.9652936160564423,
      "rewards-max/forensics_strict_edge_reward": 0.2968953996896744,
      "rewards-max/forensics_tiou_reward": 0.9343673586845398,
      "rewards/forensics_sinkhorn_reward": 0.7496009171009064,
      "rewards/forensics_strict_edge_reward": 0.05509163998067379,
      "rewards/forensics_tiou_reward": 0.535535603761673,
      "step": 1064
    },
    {
      "completion_length": 23.296875,
      "epoch": 3.2672811059907834,
      "grad_norm": 9.921048164367676,
      "kl": 0.24609375,
      "learning_rate": 1.8153846153846152e-07,
      "loss": 0.0693,
      "reward": 1.504786491394043,
      "reward_std": 0.2633357048034668,
      "reward_std_all": 0.2448146864771843,
      "rewards-max/forensics_sinkhorn_reward": 0.9808904528617859,
      "rewards-max/forensics_strict_edge_reward": 0.7026884853839874,
      "rewards-max/forensics_tiou_reward": 0.9750023782253265,
      "rewards/forensics_sinkhorn_reward": 0.7912941575050354,
      "rewards/forensics_strict_edge_reward": 0.07907166704535484,
      "rewards/forensics_tiou_reward": 0.63442063331604,
      "step": 1065
    },
    {
      "completion_length": 20.875,
      "epoch": 3.270353302611367,
      "grad_norm": 12.420321464538574,
      "kl": 0.3955078125,
      "learning_rate": 1.8076923076923076e-07,
      "loss": -0.1485,
      "reward": 1.4998034834861755,
      "reward_std": 0.22720690816640854,
      "reward_std_all": 0.21779506653547287,
      "rewards-max/forensics_sinkhorn_reward": 0.9806002378463745,
      "rewards-max/forensics_strict_edge_reward": 0.4768719747662544,
      "rewards-max/forensics_tiou_reward": 0.9741199016571045,
      "rewards/forensics_sinkhorn_reward": 0.786754697561264,
      "rewards/forensics_strict_edge_reward": 0.06589088030159473,
      "rewards/forensics_tiou_reward": 0.6471579372882843,
      "step": 1066
    },
    {
      "completion_length": 22.25,
      "epoch": 3.273425499231951,
      "grad_norm": 18.046077728271484,
      "kl": 0.2685546875,
      "learning_rate": 1.8e-07,
      "loss": -0.006,
      "reward": 1.4884611368179321,
      "reward_std": 0.20369791984558105,
      "reward_std_all": 0.17996658384799957,
      "rewards-max/forensics_sinkhorn_reward": 0.9564855098724365,
      "rewards-max/forensics_strict_edge_reward": 0.7821605801582336,
      "rewards-max/forensics_tiou_reward": 0.9158651828765869,
      "rewards/forensics_sinkhorn_reward": 0.7860012650489807,
      "rewards/forensics_strict_edge_reward": 0.07143562845885754,
      "rewards/forensics_tiou_reward": 0.6310242116451263,
      "step": 1067
    },
    {
      "completion_length": 20.90625,
      "epoch": 3.2764976958525347,
      "grad_norm": 10.568256378173828,
      "kl": 0.318359375,
      "learning_rate": 1.792307692307692e-07,
      "loss": -0.0086,
      "reward": 1.5867146253585815,
      "reward_std": 0.331555612385273,
      "reward_std_all": 0.30269723385572433,
      "rewards-max/forensics_sinkhorn_reward": 0.9726839363574982,
      "rewards-max/forensics_strict_edge_reward": 0.514585480093956,
      "rewards-max/forensics_tiou_reward": 0.9540906250476837,
      "rewards/forensics_sinkhorn_reward": 0.8123798966407776,
      "rewards/forensics_strict_edge_reward": 0.10887400805950165,
      "rewards/forensics_tiou_reward": 0.6654607057571411,
      "step": 1068
    },
    {
      "completion_length": 21.34375,
      "epoch": 3.279569892473118,
      "grad_norm": 10.656712532043457,
      "kl": 0.27294921875,
      "learning_rate": 1.7846153846153845e-07,
      "loss": -0.058,
      "reward": 1.3030312657356262,
      "reward_std": 0.23106713593006134,
      "reward_std_all": 0.20557387173175812,
      "rewards-max/forensics_sinkhorn_reward": 0.9584026038646698,
      "rewards-max/forensics_strict_edge_reward": 0.45888544619083405,
      "rewards-max/forensics_tiou_reward": 0.9249774217605591,
      "rewards/forensics_sinkhorn_reward": 0.7213341593742371,
      "rewards/forensics_strict_edge_reward": 0.04124578274786472,
      "rewards/forensics_tiou_reward": 0.5404513478279114,
      "step": 1069
    },
    {
      "completion_length": 23.359375,
      "epoch": 3.282642089093702,
      "grad_norm": 25.996076583862305,
      "kl": 0.251953125,
      "learning_rate": 1.776923076923077e-07,
      "loss": 0.0348,
      "reward": 1.4582282304763794,
      "reward_std": 0.2969907522201538,
      "reward_std_all": 0.26540273427963257,
      "rewards-max/forensics_sinkhorn_reward": 0.9695649743080139,
      "rewards-max/forensics_strict_edge_reward": 0.634945809841156,
      "rewards-max/forensics_tiou_reward": 0.9471922218799591,
      "rewards/forensics_sinkhorn_reward": 0.7734373211860657,
      "rewards/forensics_strict_edge_reward": 0.09066174924373627,
      "rewards/forensics_tiou_reward": 0.5941291749477386,
      "step": 1070
    },
    {
      "completion_length": 23.9375,
      "epoch": 3.2857142857142856,
      "grad_norm": 11.53810977935791,
      "kl": 0.419921875,
      "learning_rate": 1.7692307692307693e-07,
      "loss": 0.0186,
      "reward": 1.2908666133880615,
      "reward_std": 0.3635491728782654,
      "reward_std_all": 0.3220093697309494,
      "rewards-max/forensics_sinkhorn_reward": 0.9599923193454742,
      "rewards-max/forensics_strict_edge_reward": 0.4132732152938843,
      "rewards-max/forensics_tiou_reward": 0.9259867072105408,
      "rewards/forensics_sinkhorn_reward": 0.6949779689311981,
      "rewards/forensics_strict_edge_reward": 0.05864991154521704,
      "rewards/forensics_tiou_reward": 0.537238746881485,
      "step": 1071
    },
    {
      "completion_length": 21.859375,
      "epoch": 3.2887864823348694,
      "grad_norm": 14.321222305297852,
      "kl": 0.3876953125,
      "learning_rate": 1.7615384615384615e-07,
      "loss": 0.0864,
      "reward": 1.465566635131836,
      "reward_std": 0.19388631731271744,
      "reward_std_all": 0.2038596272468567,
      "rewards-max/forensics_sinkhorn_reward": 0.956888735294342,
      "rewards-max/forensics_strict_edge_reward": 0.3672138899564743,
      "rewards-max/forensics_tiou_reward": 0.9220135509967804,
      "rewards/forensics_sinkhorn_reward": 0.7844909131526947,
      "rewards/forensics_strict_edge_reward": 0.06196390464901924,
      "rewards/forensics_tiou_reward": 0.6191118657588959,
      "step": 1072
    },
    {
      "completion_length": 22.859375,
      "epoch": 3.291858678955453,
      "grad_norm": 13.523300170898438,
      "kl": 0.4638671875,
      "learning_rate": 1.7538461538461539e-07,
      "loss": 0.0235,
      "reward": 1.6262053847312927,
      "reward_std": 0.23777300119400024,
      "reward_std_all": 0.22652606666088104,
      "rewards-max/forensics_sinkhorn_reward": 0.981645256280899,
      "rewards-max/forensics_strict_edge_reward": 0.7482926547527313,
      "rewards-max/forensics_tiou_reward": 0.9800990223884583,
      "rewards/forensics_sinkhorn_reward": 0.8408433198928833,
      "rewards/forensics_strict_edge_reward": 0.07715468481183052,
      "rewards/forensics_tiou_reward": 0.708207368850708,
      "step": 1073
    },
    {
      "completion_length": 25.9375,
      "epoch": 3.294930875576037,
      "grad_norm": 11.755936622619629,
      "kl": 0.4541015625,
      "learning_rate": 1.746153846153846e-07,
      "loss": 0.0488,
      "reward": 1.5202381014823914,
      "reward_std": 0.25287503004074097,
      "reward_std_all": 0.21687674522399902,
      "rewards-max/forensics_sinkhorn_reward": 0.9701557457447052,
      "rewards-max/forensics_strict_edge_reward": 0.42324090003967285,
      "rewards-max/forensics_tiou_reward": 0.9495117664337158,
      "rewards/forensics_sinkhorn_reward": 0.8178026080131531,
      "rewards/forensics_strict_edge_reward": 0.04770323261618614,
      "rewards/forensics_tiou_reward": 0.6547322273254395,
      "step": 1074
    },
    {
      "completion_length": 23.859375,
      "epoch": 3.2980030721966207,
      "grad_norm": 16.107812881469727,
      "kl": 0.890625,
      "learning_rate": 1.7384615384615384e-07,
      "loss": 0.0084,
      "reward": 1.288074553012848,
      "reward_std": 0.31386642158031464,
      "reward_std_all": 0.27409540116786957,
      "rewards-max/forensics_sinkhorn_reward": 0.9665192663669586,
      "rewards-max/forensics_strict_edge_reward": 0.5183674395084381,
      "rewards-max/forensics_tiou_reward": 0.9458030164241791,
      "rewards/forensics_sinkhorn_reward": 0.6913349628448486,
      "rewards/forensics_strict_edge_reward": 0.054556699469685555,
      "rewards/forensics_tiou_reward": 0.5421828627586365,
      "step": 1075
    },
    {
      "completion_length": 22.28125,
      "epoch": 3.3010752688172045,
      "grad_norm": 10.955864906311035,
      "kl": 0.2822265625,
      "learning_rate": 1.7307692307692305e-07,
      "loss": 0.0179,
      "reward": 1.4991626143455505,
      "reward_std": 0.3208422362804413,
      "reward_std_all": 0.29733987152576447,
      "rewards-max/forensics_sinkhorn_reward": 0.9835395812988281,
      "rewards-max/forensics_strict_edge_reward": 0.7229099273681641,
      "rewards-max/forensics_tiou_reward": 0.9854608178138733,
      "rewards/forensics_sinkhorn_reward": 0.7901351749897003,
      "rewards/forensics_strict_edge_reward": 0.1015717051923275,
      "rewards/forensics_tiou_reward": 0.6074557900428772,
      "step": 1076
    },
    {
      "completion_length": 21.515625,
      "epoch": 3.3041474654377883,
      "grad_norm": 10.474072456359863,
      "kl": 0.22216796875,
      "learning_rate": 1.723076923076923e-07,
      "loss": 0.0219,
      "reward": 1.5133126974105835,
      "reward_std": 0.2767488658428192,
      "reward_std_all": 0.2539442926645279,
      "rewards-max/forensics_sinkhorn_reward": 0.9769833385944366,
      "rewards-max/forensics_strict_edge_reward": 0.6363584399223328,
      "rewards-max/forensics_tiou_reward": 0.9683437943458557,
      "rewards/forensics_sinkhorn_reward": 0.7958947420120239,
      "rewards/forensics_strict_edge_reward": 0.10399673134088516,
      "rewards/forensics_tiou_reward": 0.6134212017059326,
      "step": 1077
    },
    {
      "completion_length": 20.75,
      "epoch": 3.3072196620583716,
      "grad_norm": 11.747723579406738,
      "kl": 0.27734375,
      "learning_rate": 1.7153846153846153e-07,
      "loss": 0.0016,
      "reward": 1.30191969871521,
      "reward_std": 0.18594791740179062,
      "reward_std_all": 0.16681332141160965,
      "rewards-max/forensics_sinkhorn_reward": 0.9768950343132019,
      "rewards-max/forensics_strict_edge_reward": 0.5063450932502747,
      "rewards-max/forensics_tiou_reward": 0.9673628211021423,
      "rewards/forensics_sinkhorn_reward": 0.7322023510932922,
      "rewards/forensics_strict_edge_reward": 0.03576026111841202,
      "rewards/forensics_tiou_reward": 0.5339571684598923,
      "step": 1078
    },
    {
      "completion_length": 24.28125,
      "epoch": 3.3102918586789554,
      "grad_norm": 12.509533882141113,
      "kl": 0.3203125,
      "learning_rate": 1.7076923076923075e-07,
      "loss": -0.0914,
      "reward": 1.3988662362098694,
      "reward_std": 0.21965571492910385,
      "reward_std_all": 0.2579801231622696,
      "rewards-max/forensics_sinkhorn_reward": 0.9763431549072266,
      "rewards-max/forensics_strict_edge_reward": 0.40144599974155426,
      "rewards-max/forensics_tiou_reward": 0.965495765209198,
      "rewards/forensics_sinkhorn_reward": 0.7627125084400177,
      "rewards/forensics_strict_edge_reward": 0.05257090926170349,
      "rewards/forensics_tiou_reward": 0.5835827589035034,
      "step": 1079
    },
    {
      "completion_length": 20.34375,
      "epoch": 3.313364055299539,
      "grad_norm": 12.164851188659668,
      "kl": 0.22509765625,
      "learning_rate": 1.7000000000000001e-07,
      "loss": -0.0,
      "reward": 1.3118932247161865,
      "reward_std": 0.22797538340091705,
      "reward_std_all": 0.20827794075012207,
      "rewards-max/forensics_sinkhorn_reward": 0.9841763973236084,
      "rewards-max/forensics_strict_edge_reward": 0.5842172801494598,
      "rewards-max/forensics_tiou_reward": 0.9830988645553589,
      "rewards/forensics_sinkhorn_reward": 0.7207518219947815,
      "rewards/forensics_strict_edge_reward": 0.06353229284286499,
      "rewards/forensics_tiou_reward": 0.5276090502738953,
      "step": 1080
    },
    {
      "completion_length": 22.03125,
      "epoch": 3.316436251920123,
      "grad_norm": 19.16971206665039,
      "kl": 0.2470703125,
      "learning_rate": 1.6923076923076923e-07,
      "loss": 0.0538,
      "reward": 1.3020423650741577,
      "reward_std": 0.3095851391553879,
      "reward_std_all": 0.27832742035388947,
      "rewards-max/forensics_sinkhorn_reward": 0.9695412814617157,
      "rewards-max/forensics_strict_edge_reward": 0.6558744013309479,
      "rewards-max/forensics_tiou_reward": 0.9534798860549927,
      "rewards/forensics_sinkhorn_reward": 0.7210636138916016,
      "rewards/forensics_strict_edge_reward": 0.07472650706768036,
      "rewards/forensics_tiou_reward": 0.5062523186206818,
      "step": 1081
    },
    {
      "completion_length": 23.890625,
      "epoch": 3.3195084485407067,
      "grad_norm": 12.470845222473145,
      "kl": 0.3212890625,
      "learning_rate": 1.6846153846153847e-07,
      "loss": 0.0942,
      "reward": 1.3831883668899536,
      "reward_std": 0.20930856466293335,
      "reward_std_all": 0.19391324371099472,
      "rewards-max/forensics_sinkhorn_reward": 0.9605508148670197,
      "rewards-max/forensics_strict_edge_reward": 0.31976478546857834,
      "rewards-max/forensics_tiou_reward": 0.9245806038379669,
      "rewards/forensics_sinkhorn_reward": 0.7589278519153595,
      "rewards/forensics_strict_edge_reward": 0.04187904950231314,
      "rewards/forensics_tiou_reward": 0.5823814570903778,
      "step": 1082
    },
    {
      "completion_length": 21.1875,
      "epoch": 3.3225806451612905,
      "grad_norm": 12.681854248046875,
      "kl": 0.36328125,
      "learning_rate": 1.6769230769230768e-07,
      "loss": 0.0919,
      "reward": 1.252071499824524,
      "reward_std": 0.2973792627453804,
      "reward_std_all": 0.2591434717178345,
      "rewards-max/forensics_sinkhorn_reward": 0.9720395803451538,
      "rewards-max/forensics_strict_edge_reward": 0.5237197279930115,
      "rewards-max/forensics_tiou_reward": 0.9592142105102539,
      "rewards/forensics_sinkhorn_reward": 0.690636157989502,
      "rewards/forensics_strict_edge_reward": 0.06329125538468361,
      "rewards/forensics_tiou_reward": 0.49814414978027344,
      "step": 1083
    },
    {
      "completion_length": 23.96875,
      "epoch": 3.3256528417818743,
      "grad_norm": 16.904682159423828,
      "kl": 0.2685546875,
      "learning_rate": 1.6692307692307692e-07,
      "loss": 0.0582,
      "reward": 1.3066465258598328,
      "reward_std": 0.14761142060160637,
      "reward_std_all": 0.13202998414635658,
      "rewards-max/forensics_sinkhorn_reward": 0.9435759782791138,
      "rewards-max/forensics_strict_edge_reward": 0.44787345826625824,
      "rewards-max/forensics_tiou_reward": 0.8872294127941132,
      "rewards/forensics_sinkhorn_reward": 0.7262400090694427,
      "rewards/forensics_strict_edge_reward": 0.03498053178191185,
      "rewards/forensics_tiou_reward": 0.5454259365797043,
      "step": 1084
    },
    {
      "completion_length": 21.234375,
      "epoch": 3.3287250384024576,
      "grad_norm": 14.448725700378418,
      "kl": 0.4150390625,
      "learning_rate": 1.6615384615384613e-07,
      "loss": -0.1166,
      "reward": 1.5030133724212646,
      "reward_std": 0.25893935561180115,
      "reward_std_all": 0.24263033270835876,
      "rewards-max/forensics_sinkhorn_reward": 0.9710709452629089,
      "rewards-max/forensics_strict_edge_reward": 0.569886103272438,
      "rewards-max/forensics_tiou_reward": 0.9450939297676086,
      "rewards/forensics_sinkhorn_reward": 0.7664448320865631,
      "rewards/forensics_strict_edge_reward": 0.11675849370658398,
      "rewards/forensics_tiou_reward": 0.6198100447654724,
      "step": 1085
    },
    {
      "completion_length": 25.75,
      "epoch": 3.3317972350230414,
      "grad_norm": 12.849519729614258,
      "kl": 0.529296875,
      "learning_rate": 1.6538461538461538e-07,
      "loss": 0.0515,
      "reward": 1.4366373419761658,
      "reward_std": 0.23597048968076706,
      "reward_std_all": 0.2360316812992096,
      "rewards-max/forensics_sinkhorn_reward": 0.9762147068977356,
      "rewards-max/forensics_strict_edge_reward": 0.39423424005508423,
      "rewards-max/forensics_tiou_reward": 0.96659716963768,
      "rewards/forensics_sinkhorn_reward": 0.764733761548996,
      "rewards/forensics_strict_edge_reward": 0.06429655477404594,
      "rewards/forensics_tiou_reward": 0.6076070666313171,
      "step": 1086
    },
    {
      "completion_length": 25.5625,
      "epoch": 3.334869431643625,
      "grad_norm": 9.808213233947754,
      "kl": 0.328125,
      "learning_rate": 1.646153846153846e-07,
      "loss": -0.0356,
      "reward": 1.3588446974754333,
      "reward_std": 0.22823285311460495,
      "reward_std_all": 0.20874421298503876,
      "rewards-max/forensics_sinkhorn_reward": 0.967778742313385,
      "rewards-max/forensics_strict_edge_reward": 0.3727487623691559,
      "rewards-max/forensics_tiou_reward": 0.9461327791213989,
      "rewards/forensics_sinkhorn_reward": 0.7445952296257019,
      "rewards/forensics_strict_edge_reward": 0.04954778216779232,
      "rewards/forensics_tiou_reward": 0.564701646566391,
      "step": 1087
    },
    {
      "completion_length": 21.34375,
      "epoch": 3.337941628264209,
      "grad_norm": 12.813779830932617,
      "kl": 0.4052734375,
      "learning_rate": 1.6384615384615383e-07,
      "loss": -0.084,
      "reward": 1.5489044785499573,
      "reward_std": 0.28632882237434387,
      "reward_std_all": 0.26052697002887726,
      "rewards-max/forensics_sinkhorn_reward": 0.9818606376647949,
      "rewards-max/forensics_strict_edge_reward": 0.5834503769874573,
      "rewards-max/forensics_tiou_reward": 0.9806950092315674,
      "rewards/forensics_sinkhorn_reward": 0.786020964384079,
      "rewards/forensics_strict_edge_reward": 0.13235627114772797,
      "rewards/forensics_tiou_reward": 0.6305272579193115,
      "step": 1088
    },
    {
      "completion_length": 22.171875,
      "epoch": 3.3410138248847927,
      "grad_norm": 21.73383140563965,
      "kl": 0.2890625,
      "learning_rate": 1.6307692307692307e-07,
      "loss": -0.0303,
      "reward": 1.6281749606132507,
      "reward_std": 0.19958390295505524,
      "reward_std_all": 0.18764303624629974,
      "rewards-max/forensics_sinkhorn_reward": 0.9814956188201904,
      "rewards-max/forensics_strict_edge_reward": 0.7165313363075256,
      "rewards-max/forensics_tiou_reward": 0.9811436533927917,
      "rewards/forensics_sinkhorn_reward": 0.8507389724254608,
      "rewards/forensics_strict_edge_reward": 0.08080195263028145,
      "rewards/forensics_tiou_reward": 0.69663405418396,
      "step": 1089
    },
    {
      "completion_length": 23.921875,
      "epoch": 3.3440860215053765,
      "grad_norm": 11.822328567504883,
      "kl": 0.2294921875,
      "learning_rate": 1.623076923076923e-07,
      "loss": 0.0062,
      "reward": 1.6103686690330505,
      "reward_std": 0.19097314774990082,
      "reward_std_all": 0.16599520295858383,
      "rewards-max/forensics_sinkhorn_reward": 0.9721175730228424,
      "rewards-max/forensics_strict_edge_reward": 0.4341297447681427,
      "rewards-max/forensics_tiou_reward": 0.9528308808803558,
      "rewards/forensics_sinkhorn_reward": 0.8297304809093475,
      "rewards/forensics_strict_edge_reward": 0.08081144839525223,
      "rewards/forensics_tiou_reward": 0.699826717376709,
      "step": 1090
    },
    {
      "completion_length": 20.34375,
      "epoch": 3.3471582181259603,
      "grad_norm": 10.437202453613281,
      "kl": 0.26025390625,
      "learning_rate": 1.6153846153846155e-07,
      "loss": -0.046,
      "reward": 1.4932451248168945,
      "reward_std": 0.2612839564681053,
      "reward_std_all": 0.2508236765861511,
      "rewards-max/forensics_sinkhorn_reward": 0.977163553237915,
      "rewards-max/forensics_strict_edge_reward": 0.40215258300304413,
      "rewards-max/forensics_tiou_reward": 0.9653252065181732,
      "rewards/forensics_sinkhorn_reward": 0.8039386868476868,
      "rewards/forensics_strict_edge_reward": 0.06720611080527306,
      "rewards/forensics_tiou_reward": 0.6221003830432892,
      "step": 1091
    },
    {
      "completion_length": 20.65625,
      "epoch": 3.3502304147465436,
      "grad_norm": 14.536355972290039,
      "kl": 0.2685546875,
      "learning_rate": 1.6076923076923076e-07,
      "loss": 0.1083,
      "reward": 1.4393959045410156,
      "reward_std": 0.35860252380371094,
      "reward_std_all": 0.32717031240463257,
      "rewards-max/forensics_sinkhorn_reward": 0.9768789708614349,
      "rewards-max/forensics_strict_edge_reward": 0.5918685793876648,
      "rewards-max/forensics_tiou_reward": 0.9701731204986572,
      "rewards/forensics_sinkhorn_reward": 0.7308928072452545,
      "rewards/forensics_strict_edge_reward": 0.1154155321419239,
      "rewards/forensics_tiou_reward": 0.5930876433849335,
      "step": 1092
    },
    {
      "completion_length": 23.984375,
      "epoch": 3.3533026113671274,
      "grad_norm": 17.572250366210938,
      "kl": 0.3935546875,
      "learning_rate": 1.6e-07,
      "loss": -0.045,
      "reward": 1.2908586859703064,
      "reward_std": 0.30180947482585907,
      "reward_std_all": 0.27831658720970154,
      "rewards-max/forensics_sinkhorn_reward": 0.9676767587661743,
      "rewards-max/forensics_strict_edge_reward": 0.38241781294345856,
      "rewards-max/forensics_tiou_reward": 0.9439014792442322,
      "rewards/forensics_sinkhorn_reward": 0.6930291950702667,
      "rewards/forensics_strict_edge_reward": 0.04820537194609642,
      "rewards/forensics_tiou_reward": 0.5496241450309753,
      "step": 1093
    },
    {
      "completion_length": 20.03125,
      "epoch": 3.356374807987711,
      "grad_norm": 13.2024507522583,
      "kl": 0.271484375,
      "learning_rate": 1.5923076923076922e-07,
      "loss": -0.1265,
      "reward": 1.436416506767273,
      "reward_std": 0.19661813229322433,
      "reward_std_all": 0.170725017786026,
      "rewards-max/forensics_sinkhorn_reward": 0.9561077356338501,
      "rewards-max/forensics_strict_edge_reward": 0.26892485469579697,
      "rewards-max/forensics_tiou_reward": 0.9203843474388123,
      "rewards/forensics_sinkhorn_reward": 0.7836796939373016,
      "rewards/forensics_strict_edge_reward": 0.038183096796274185,
      "rewards/forensics_tiou_reward": 0.6145537495613098,
      "step": 1094
    },
    {
      "completion_length": 20.921875,
      "epoch": 3.359447004608295,
      "grad_norm": 12.888647079467773,
      "kl": 0.388671875,
      "learning_rate": 1.5846153846153846e-07,
      "loss": 0.149,
      "reward": 1.3630508184432983,
      "reward_std": 0.2665565013885498,
      "reward_std_all": 0.23182518780231476,
      "rewards-max/forensics_sinkhorn_reward": 0.9861897230148315,
      "rewards-max/forensics_strict_edge_reward": 0.5630159378051758,
      "rewards-max/forensics_tiou_reward": 0.9904930591583252,
      "rewards/forensics_sinkhorn_reward": 0.7470660507678986,
      "rewards/forensics_strict_edge_reward": 0.051702793687582016,
      "rewards/forensics_tiou_reward": 0.5642819702625275,
      "step": 1095
    },
    {
      "completion_length": 22.65625,
      "epoch": 3.3625192012288787,
      "grad_norm": 10.905396461486816,
      "kl": 0.29931640625,
      "learning_rate": 1.5769230769230767e-07,
      "loss": 0.0699,
      "reward": 1.4416583776474,
      "reward_std": 0.2385159507393837,
      "reward_std_all": 0.21306262910366058,
      "rewards-max/forensics_sinkhorn_reward": 0.9776637554168701,
      "rewards-max/forensics_strict_edge_reward": 0.6250577867031097,
      "rewards-max/forensics_tiou_reward": 0.9692197144031525,
      "rewards/forensics_sinkhorn_reward": 0.7372628152370453,
      "rewards/forensics_strict_edge_reward": 0.10887188464403152,
      "rewards/forensics_tiou_reward": 0.5955237150192261,
      "step": 1096
    },
    {
      "completion_length": 20.5,
      "epoch": 3.3655913978494625,
      "grad_norm": 12.426338195800781,
      "kl": 0.22998046875,
      "learning_rate": 1.569230769230769e-07,
      "loss": -0.0119,
      "reward": 1.550164520740509,
      "reward_std": 0.2838141918182373,
      "reward_std_all": 0.2541898936033249,
      "rewards-max/forensics_sinkhorn_reward": 0.9851534962654114,
      "rewards-max/forensics_strict_edge_reward": 0.499885693192482,
      "rewards-max/forensics_tiou_reward": 0.9895072281360626,
      "rewards/forensics_sinkhorn_reward": 0.7977953255176544,
      "rewards/forensics_strict_edge_reward": 0.1064319871366024,
      "rewards/forensics_tiou_reward": 0.6459372341632843,
      "step": 1097
    },
    {
      "completion_length": 22.40625,
      "epoch": 3.3686635944700463,
      "grad_norm": 11.971382141113281,
      "kl": 0.43408203125,
      "learning_rate": 1.5615384615384612e-07,
      "loss": 0.0095,
      "reward": 1.399364948272705,
      "reward_std": 0.29002247750759125,
      "reward_std_all": 0.2646748721599579,
      "rewards-max/forensics_sinkhorn_reward": 0.9721597135066986,
      "rewards-max/forensics_strict_edge_reward": 0.49303558468818665,
      "rewards-max/forensics_tiou_reward": 0.9583017230033875,
      "rewards/forensics_sinkhorn_reward": 0.7417188286781311,
      "rewards/forensics_strict_edge_reward": 0.06271819584071636,
      "rewards/forensics_tiou_reward": 0.5949280560016632,
      "step": 1098
    },
    {
      "completion_length": 20.171875,
      "epoch": 3.3717357910906296,
      "grad_norm": 14.36856746673584,
      "kl": 0.228515625,
      "learning_rate": 1.5538461538461536e-07,
      "loss": -0.0822,
      "reward": 1.6008285880088806,
      "reward_std": 0.1796261966228485,
      "reward_std_all": 0.16732604056596756,
      "rewards-max/forensics_sinkhorn_reward": 0.9720757901668549,
      "rewards-max/forensics_strict_edge_reward": 0.4277711436152458,
      "rewards-max/forensics_tiou_reward": 0.9583760201931,
      "rewards/forensics_sinkhorn_reward": 0.8430952429771423,
      "rewards/forensics_strict_edge_reward": 0.06363246496766806,
      "rewards/forensics_tiou_reward": 0.6941008567810059,
      "step": 1099
    },
    {
      "completion_length": 21.109375,
      "epoch": 3.3748079877112134,
      "grad_norm": 11.00801944732666,
      "kl": 0.4365234375,
      "learning_rate": 1.5461538461538463e-07,
      "loss": 0.0156,
      "reward": 1.3590204119682312,
      "reward_std": 0.26156269758939743,
      "reward_std_all": 0.22987311333417892,
      "rewards-max/forensics_sinkhorn_reward": 0.9697428941726685,
      "rewards-max/forensics_strict_edge_reward": 0.2680644690990448,
      "rewards-max/forensics_tiou_reward": 0.9489727914333344,
      "rewards/forensics_sinkhorn_reward": 0.7388618290424347,
      "rewards/forensics_strict_edge_reward": 0.03677869774401188,
      "rewards/forensics_tiou_reward": 0.583379864692688,
      "step": 1100
    },
    {
      "completion_length": 20.78125,
      "epoch": 3.377880184331797,
      "grad_norm": 25.06653594970703,
      "kl": 0.71875,
      "learning_rate": 1.5384615384615385e-07,
      "loss": -0.0057,
      "reward": 1.3590950965881348,
      "reward_std": 0.29596225917339325,
      "reward_std_all": 0.2579689472913742,
      "rewards-max/forensics_sinkhorn_reward": 0.9736950695514679,
      "rewards-max/forensics_strict_edge_reward": 0.4489254802465439,
      "rewards-max/forensics_tiou_reward": 0.9577851295471191,
      "rewards/forensics_sinkhorn_reward": 0.7583009302616119,
      "rewards/forensics_strict_edge_reward": 0.05908758565783501,
      "rewards/forensics_tiou_reward": 0.5417065918445587,
      "step": 1101
    },
    {
      "completion_length": 21.84375,
      "epoch": 3.380952380952381,
      "grad_norm": 18.970170974731445,
      "kl": 0.36376953125,
      "learning_rate": 1.5307692307692309e-07,
      "loss": 0.0603,
      "reward": 1.4037643671035767,
      "reward_std": 0.25448672473430634,
      "reward_std_all": 0.2214456871151924,
      "rewards-max/forensics_sinkhorn_reward": 0.9736932814121246,
      "rewards-max/forensics_strict_edge_reward": 0.36184918880462646,
      "rewards-max/forensics_tiou_reward": 0.9587360322475433,
      "rewards/forensics_sinkhorn_reward": 0.7725018858909607,
      "rewards/forensics_strict_edge_reward": 0.04876359738409519,
      "rewards/forensics_tiou_reward": 0.5824988484382629,
      "step": 1102
    },
    {
      "completion_length": 23.1875,
      "epoch": 3.3840245775729647,
      "grad_norm": 12.952630996704102,
      "kl": 0.283203125,
      "learning_rate": 1.523076923076923e-07,
      "loss": 0.0433,
      "reward": 1.630056083202362,
      "reward_std": 0.1965995952486992,
      "reward_std_all": 0.18922316282987595,
      "rewards-max/forensics_sinkhorn_reward": 0.9745569229125977,
      "rewards-max/forensics_strict_edge_reward": 0.541615754365921,
      "rewards-max/forensics_tiou_reward": 0.9626327157020569,
      "rewards/forensics_sinkhorn_reward": 0.8288698792457581,
      "rewards/forensics_strict_edge_reward": 0.10001150518655777,
      "rewards/forensics_tiou_reward": 0.7011747658252716,
      "step": 1103
    },
    {
      "completion_length": 23.09375,
      "epoch": 3.3870967741935485,
      "grad_norm": 10.93875789642334,
      "kl": 0.544921875,
      "learning_rate": 1.5153846153846154e-07,
      "loss": 0.0282,
      "reward": 1.4381037950515747,
      "reward_std": 0.25902096927165985,
      "reward_std_all": 0.23610376566648483,
      "rewards-max/forensics_sinkhorn_reward": 0.9718727171421051,
      "rewards-max/forensics_strict_edge_reward": 0.5121164321899414,
      "rewards-max/forensics_tiou_reward": 0.9539735019207001,
      "rewards/forensics_sinkhorn_reward": 0.7664092481136322,
      "rewards/forensics_strict_edge_reward": 0.06716747209429741,
      "rewards/forensics_tiou_reward": 0.6045270562171936,
      "step": 1104
    },
    {
      "completion_length": 23.734375,
      "epoch": 3.3901689708141323,
      "grad_norm": 9.505158424377441,
      "kl": 0.6748046875,
      "learning_rate": 1.5076923076923075e-07,
      "loss": -0.0115,
      "reward": 1.3484841585159302,
      "reward_std": 0.2613275796175003,
      "reward_std_all": 0.26589149236679077,
      "rewards-max/forensics_sinkhorn_reward": 0.959143191576004,
      "rewards-max/forensics_strict_edge_reward": 0.3663562908768654,
      "rewards-max/forensics_tiou_reward": 0.9252069592475891,
      "rewards/forensics_sinkhorn_reward": 0.7492127418518066,
      "rewards/forensics_strict_edge_reward": 0.029472002759575844,
      "rewards/forensics_tiou_reward": 0.5697994232177734,
      "step": 1105
    },
    {
      "completion_length": 22.34375,
      "epoch": 3.3932411674347156,
      "grad_norm": 13.212321281433105,
      "kl": 0.408203125,
      "learning_rate": 1.5e-07,
      "loss": 0.0075,
      "reward": 1.4975318908691406,
      "reward_std": 0.19883470982313156,
      "reward_std_all": 0.18666033446788788,
      "rewards-max/forensics_sinkhorn_reward": 0.98540398478508,
      "rewards-max/forensics_strict_edge_reward": 0.7162975072860718,
      "rewards-max/forensics_tiou_reward": 0.9893608391284943,
      "rewards/forensics_sinkhorn_reward": 0.7875466346740723,
      "rewards/forensics_strict_edge_reward": 0.10200406610965729,
      "rewards/forensics_tiou_reward": 0.6079811453819275,
      "step": 1106
    },
    {
      "completion_length": 20.8125,
      "epoch": 3.3963133640552994,
      "grad_norm": 14.613962173461914,
      "kl": 0.3349609375,
      "learning_rate": 1.492307692307692e-07,
      "loss": -0.0271,
      "reward": 1.6180258393287659,
      "reward_std": 0.18588116765022278,
      "reward_std_all": 0.16515365988016129,
      "rewards-max/forensics_sinkhorn_reward": 0.9736851751804352,
      "rewards-max/forensics_strict_edge_reward": 0.4835272878408432,
      "rewards-max/forensics_tiou_reward": 0.9605657756328583,
      "rewards/forensics_sinkhorn_reward": 0.8415535688400269,
      "rewards/forensics_strict_edge_reward": 0.08126138150691986,
      "rewards/forensics_tiou_reward": 0.6952109634876251,
      "step": 1107
    },
    {
      "completion_length": 21.96875,
      "epoch": 3.399385560675883,
      "grad_norm": 12.941728591918945,
      "kl": 0.3818359375,
      "learning_rate": 1.4846153846153845e-07,
      "loss": 0.0916,
      "reward": 1.5169730186462402,
      "reward_std": 0.21851886808872223,
      "reward_std_all": 0.2539747729897499,
      "rewards-max/forensics_sinkhorn_reward": 0.9626506865024567,
      "rewards-max/forensics_strict_edge_reward": 0.5761755704879761,
      "rewards-max/forensics_tiou_reward": 0.9369329512119293,
      "rewards/forensics_sinkhorn_reward": 0.8041931986808777,
      "rewards/forensics_strict_edge_reward": 0.06484504230320454,
      "rewards/forensics_tiou_reward": 0.6479347348213196,
      "step": 1108
    },
    {
      "completion_length": 25.78125,
      "epoch": 3.402457757296467,
      "grad_norm": 18.132465362548828,
      "kl": 0.3994140625,
      "learning_rate": 1.476923076923077e-07,
      "loss": 0.1041,
      "reward": 1.488734781742096,
      "reward_std": 0.19066385179758072,
      "reward_std_all": 0.18084069341421127,
      "rewards-max/forensics_sinkhorn_reward": 0.9738697111606598,
      "rewards-max/forensics_strict_edge_reward": 0.2080069910734892,
      "rewards-max/forensics_tiou_reward": 0.9572309851646423,
      "rewards/forensics_sinkhorn_reward": 0.8121668696403503,
      "rewards/forensics_strict_edge_reward": 0.029640314634889364,
      "rewards/forensics_tiou_reward": 0.6469275951385498,
      "step": 1109
    },
    {
      "completion_length": 19.40625,
      "epoch": 3.4055299539170507,
      "grad_norm": 11.590115547180176,
      "kl": 0.7255859375,
      "learning_rate": 1.4692307692307693e-07,
      "loss": 0.0969,
      "reward": 1.3221303820610046,
      "reward_std": 0.20732884109020233,
      "reward_std_all": 0.21761919558048248,
      "rewards-max/forensics_sinkhorn_reward": 0.9476373791694641,
      "rewards-max/forensics_strict_edge_reward": 0.307451494038105,
      "rewards-max/forensics_tiou_reward": 0.8987419307231903,
      "rewards/forensics_sinkhorn_reward": 0.7380953133106232,
      "rewards/forensics_strict_edge_reward": 0.04858798533678055,
      "rewards/forensics_tiou_reward": 0.5354470908641815,
      "step": 1110
    },
    {
      "completion_length": 21.421875,
      "epoch": 3.4086021505376345,
      "grad_norm": 14.763511657714844,
      "kl": 0.3662109375,
      "learning_rate": 1.4615384615384617e-07,
      "loss": -0.0863,
      "reward": 1.4606234431266785,
      "reward_std": 0.28007933497428894,
      "reward_std_all": 0.2836093455553055,
      "rewards-max/forensics_sinkhorn_reward": 0.9774592518806458,
      "rewards-max/forensics_strict_edge_reward": 0.5587328374385834,
      "rewards-max/forensics_tiou_reward": 0.9691548645496368,
      "rewards/forensics_sinkhorn_reward": 0.7685824334621429,
      "rewards/forensics_strict_edge_reward": 0.08642861247062683,
      "rewards/forensics_tiou_reward": 0.6056123971939087,
      "step": 1111
    },
    {
      "completion_length": 22.171875,
      "epoch": 3.4116743471582183,
      "grad_norm": 10.806276321411133,
      "kl": 0.68359375,
      "learning_rate": 1.4538461538461538e-07,
      "loss": 0.095,
      "reward": 1.5052822828292847,
      "reward_std": 0.29553908109664917,
      "reward_std_all": 0.2577502280473709,
      "rewards-max/forensics_sinkhorn_reward": 0.948124349117279,
      "rewards-max/forensics_strict_edge_reward": 0.5374585837125778,
      "rewards-max/forensics_tiou_reward": 0.898341953754425,
      "rewards/forensics_sinkhorn_reward": 0.7947448492050171,
      "rewards/forensics_strict_edge_reward": 0.07049761340022087,
      "rewards/forensics_tiou_reward": 0.640039712190628,
      "step": 1112
    },
    {
      "completion_length": 21.921875,
      "epoch": 3.4147465437788016,
      "grad_norm": 21.3290958404541,
      "kl": 1.0791015625,
      "learning_rate": 1.4461538461538462e-07,
      "loss": 0.1138,
      "reward": 1.524783730506897,
      "reward_std": 0.2798696756362915,
      "reward_std_all": 0.24850532412528992,
      "rewards-max/forensics_sinkhorn_reward": 0.9735784530639648,
      "rewards-max/forensics_strict_edge_reward": 0.5572862923145294,
      "rewards-max/forensics_tiou_reward": 0.9545636475086212,
      "rewards/forensics_sinkhorn_reward": 0.7941503524780273,
      "rewards/forensics_strict_edge_reward": 0.07996099814772606,
      "rewards/forensics_tiou_reward": 0.650672435760498,
      "step": 1113
    },
    {
      "completion_length": 22.15625,
      "epoch": 3.4178187403993854,
      "grad_norm": 13.051504135131836,
      "kl": 0.318359375,
      "learning_rate": 1.4384615384615384e-07,
      "loss": -0.0202,
      "reward": 1.4500973224639893,
      "reward_std": 0.28462083637714386,
      "reward_std_all": 0.251738965511322,
      "rewards-max/forensics_sinkhorn_reward": 0.9765410423278809,
      "rewards-max/forensics_strict_edge_reward": 0.44489842653274536,
      "rewards-max/forensics_tiou_reward": 0.9639612138271332,
      "rewards/forensics_sinkhorn_reward": 0.7699966728687286,
      "rewards/forensics_strict_edge_reward": 0.07283201441168785,
      "rewards/forensics_tiou_reward": 0.6072686016559601,
      "step": 1114
    },
    {
      "completion_length": 23.46875,
      "epoch": 3.420890937019969,
      "grad_norm": 16.808700561523438,
      "kl": 0.2529296875,
      "learning_rate": 1.4307692307692308e-07,
      "loss": -0.0525,
      "reward": 1.6320761442184448,
      "reward_std": 0.2043992057442665,
      "reward_std_all": 0.19454824924468994,
      "rewards-max/forensics_sinkhorn_reward": 0.9795708060264587,
      "rewards-max/forensics_strict_edge_reward": 0.57895427942276,
      "rewards-max/forensics_tiou_reward": 0.9726435244083405,
      "rewards/forensics_sinkhorn_reward": 0.8367257416248322,
      "rewards/forensics_strict_edge_reward": 0.09758356586098671,
      "rewards/forensics_tiou_reward": 0.6977668404579163,
      "step": 1115
    },
    {
      "completion_length": 21.671875,
      "epoch": 3.423963133640553,
      "grad_norm": 11.879801750183105,
      "kl": 0.5810546875,
      "learning_rate": 1.423076923076923e-07,
      "loss": 0.1863,
      "reward": 1.242835521697998,
      "reward_std": 0.29400642216205597,
      "reward_std_all": 0.2825765311717987,
      "rewards-max/forensics_sinkhorn_reward": 0.9621035754680634,
      "rewards-max/forensics_strict_edge_reward": 0.5665471255779266,
      "rewards-max/forensics_tiou_reward": 0.9356460273265839,
      "rewards/forensics_sinkhorn_reward": 0.692142903804779,
      "rewards/forensics_strict_edge_reward": 0.06259659118950367,
      "rewards/forensics_tiou_reward": 0.48809608817100525,
      "step": 1116
    },
    {
      "completion_length": 21.609375,
      "epoch": 3.4270353302611367,
      "grad_norm": 14.949662208557129,
      "kl": 0.29296875,
      "learning_rate": 1.4153846153846153e-07,
      "loss": 0.1206,
      "reward": 1.2560486197471619,
      "reward_std": 0.2228328287601471,
      "reward_std_all": 0.23352084308862686,
      "rewards-max/forensics_sinkhorn_reward": 0.9530891478061676,
      "rewards-max/forensics_strict_edge_reward": 0.6701013445854187,
      "rewards-max/forensics_tiou_reward": 0.9205408096313477,
      "rewards/forensics_sinkhorn_reward": 0.7164799273014069,
      "rewards/forensics_strict_edge_reward": 0.058886148035526276,
      "rewards/forensics_tiou_reward": 0.48068249225616455,
      "step": 1117
    },
    {
      "completion_length": 23.46875,
      "epoch": 3.4301075268817205,
      "grad_norm": 16.571678161621094,
      "kl": 0.31787109375,
      "learning_rate": 1.4076923076923074e-07,
      "loss": -0.0065,
      "reward": 1.514144241809845,
      "reward_std": 0.2556391954421997,
      "reward_std_all": 0.25316284596920013,
      "rewards-max/forensics_sinkhorn_reward": 0.9766733646392822,
      "rewards-max/forensics_strict_edge_reward": 0.6289739608764648,
      "rewards-max/forensics_tiou_reward": 0.9691748023033142,
      "rewards/forensics_sinkhorn_reward": 0.7685454487800598,
      "rewards/forensics_strict_edge_reward": 0.13040917925536633,
      "rewards/forensics_tiou_reward": 0.6151895523071289,
      "step": 1118
    },
    {
      "completion_length": 19.828125,
      "epoch": 3.4331797235023043,
      "grad_norm": 12.68332576751709,
      "kl": 0.3564453125,
      "learning_rate": 1.4e-07,
      "loss": -0.0432,
      "reward": 1.4475704431533813,
      "reward_std": 0.26598262786865234,
      "reward_std_all": 0.2338794320821762,
      "rewards-max/forensics_sinkhorn_reward": 0.9802520573139191,
      "rewards-max/forensics_strict_edge_reward": 0.7514603435993195,
      "rewards-max/forensics_tiou_reward": 0.9784935712814331,
      "rewards/forensics_sinkhorn_reward": 0.7774041593074799,
      "rewards/forensics_strict_edge_reward": 0.09068343043327332,
      "rewards/forensics_tiou_reward": 0.5794828832149506,
      "step": 1119
    },
    {
      "completion_length": 20.171875,
      "epoch": 3.436251920122888,
      "grad_norm": 13.46518325805664,
      "kl": 0.2314453125,
      "learning_rate": 1.3923076923076922e-07,
      "loss": 0.0298,
      "reward": 1.436627447605133,
      "reward_std": 0.14427268505096436,
      "reward_std_all": 0.1371747851371765,
      "rewards-max/forensics_sinkhorn_reward": 0.986997663974762,
      "rewards-max/forensics_strict_edge_reward": 0.608120322227478,
      "rewards-max/forensics_tiou_reward": 0.9934546649456024,
      "rewards/forensics_sinkhorn_reward": 0.7773682177066803,
      "rewards/forensics_strict_edge_reward": 0.07029002159833908,
      "rewards/forensics_tiou_reward": 0.5889692008495331,
      "step": 1120
    },
    {
      "completion_length": 21.609375,
      "epoch": 3.439324116743472,
      "grad_norm": 16.33791160583496,
      "kl": 0.5263671875,
      "learning_rate": 1.3846153846153846e-07,
      "loss": 0.0405,
      "reward": 1.4233968257904053,
      "reward_std": 0.20600482821464539,
      "reward_std_all": 0.19338425248861313,
      "rewards-max/forensics_sinkhorn_reward": 0.9649673104286194,
      "rewards-max/forensics_strict_edge_reward": 0.23880668729543686,
      "rewards-max/forensics_tiou_reward": 0.9369333386421204,
      "rewards/forensics_sinkhorn_reward": 0.7871536910533905,
      "rewards/forensics_strict_edge_reward": 0.03439422883093357,
      "rewards/forensics_tiou_reward": 0.6018487513065338,
      "step": 1121
    },
    {
      "completion_length": 20.5,
      "epoch": 3.442396313364055,
      "grad_norm": 13.755521774291992,
      "kl": 0.31884765625,
      "learning_rate": 1.376923076923077e-07,
      "loss": 0.0788,
      "reward": 1.4918586611747742,
      "reward_std": 0.1828513741493225,
      "reward_std_all": 0.20664937794208527,
      "rewards-max/forensics_sinkhorn_reward": 0.9776639044284821,
      "rewards-max/forensics_strict_edge_reward": 0.3885071277618408,
      "rewards-max/forensics_tiou_reward": 0.967894583940506,
      "rewards/forensics_sinkhorn_reward": 0.8131013810634613,
      "rewards/forensics_strict_edge_reward": 0.045886049047112465,
      "rewards/forensics_tiou_reward": 0.6328712701797485,
      "step": 1122
    },
    {
      "completion_length": 20.328125,
      "epoch": 3.445468509984639,
      "grad_norm": 13.972482681274414,
      "kl": 0.40625,
      "learning_rate": 1.3692307692307692e-07,
      "loss": 0.051,
      "reward": 1.298470675945282,
      "reward_std": 0.1701442450284958,
      "reward_std_all": 0.15250609442591667,
      "rewards-max/forensics_sinkhorn_reward": 0.9514071345329285,
      "rewards-max/forensics_strict_edge_reward": 0.17770091071724892,
      "rewards-max/forensics_tiou_reward": 0.908210813999176,
      "rewards/forensics_sinkhorn_reward": 0.7221675515174866,
      "rewards/forensics_strict_edge_reward": 0.026273942436091602,
      "rewards/forensics_tiou_reward": 0.5500291287899017,
      "step": 1123
    },
    {
      "completion_length": 23.84375,
      "epoch": 3.4485407066052227,
      "grad_norm": 14.65048599243164,
      "kl": 0.2861328125,
      "learning_rate": 1.3615384615384616e-07,
      "loss": 0.0045,
      "reward": 1.5748623609542847,
      "reward_std": 0.2147921472787857,
      "reward_std_all": 0.18712297827005386,
      "rewards-max/forensics_sinkhorn_reward": 0.9785230457782745,
      "rewards-max/forensics_strict_edge_reward": 0.5373985022306442,
      "rewards-max/forensics_tiou_reward": 0.972679615020752,
      "rewards/forensics_sinkhorn_reward": 0.803823709487915,
      "rewards/forensics_strict_edge_reward": 0.09376318752765656,
      "rewards/forensics_tiou_reward": 0.6772754788398743,
      "step": 1124
    },
    {
      "completion_length": 22.21875,
      "epoch": 3.4516129032258065,
      "grad_norm": 18.077503204345703,
      "kl": 0.29833984375,
      "learning_rate": 1.3538461538461537e-07,
      "loss": 0.0165,
      "reward": 1.5059105157852173,
      "reward_std": 0.18790636211633682,
      "reward_std_all": 0.17649798095226288,
      "rewards-max/forensics_sinkhorn_reward": 0.9728735089302063,
      "rewards-max/forensics_strict_edge_reward": 0.293737955391407,
      "rewards-max/forensics_tiou_reward": 0.9597029089927673,
      "rewards/forensics_sinkhorn_reward": 0.8229002356529236,
      "rewards/forensics_strict_edge_reward": 0.043189527466893196,
      "rewards/forensics_tiou_reward": 0.6398208439350128,
      "step": 1125
    },
    {
      "completion_length": 24.453125,
      "epoch": 3.4546850998463903,
      "grad_norm": 16.170461654663086,
      "kl": 0.27490234375,
      "learning_rate": 1.346153846153846e-07,
      "loss": -0.0071,
      "reward": 1.658773958683014,
      "reward_std": 0.2526192143559456,
      "reward_std_all": 0.23074062168598175,
      "rewards-max/forensics_sinkhorn_reward": 0.9791596829891205,
      "rewards-max/forensics_strict_edge_reward": 0.5829301476478577,
      "rewards-max/forensics_tiou_reward": 0.9731951057910919,
      "rewards/forensics_sinkhorn_reward": 0.8519887328147888,
      "rewards/forensics_strict_edge_reward": 0.08912266418337822,
      "rewards/forensics_tiou_reward": 0.7176624536514282,
      "step": 1126
    },
    {
      "completion_length": 22.65625,
      "epoch": 3.457757296466974,
      "grad_norm": 12.219635009765625,
      "kl": 0.52587890625,
      "learning_rate": 1.3384615384615383e-07,
      "loss": -0.0211,
      "reward": 1.387851595878601,
      "reward_std": 0.23613981157541275,
      "reward_std_all": 0.2106512039899826,
      "rewards-max/forensics_sinkhorn_reward": 0.9831650853157043,
      "rewards-max/forensics_strict_edge_reward": 0.6515887677669525,
      "rewards-max/forensics_tiou_reward": 0.9838131368160248,
      "rewards/forensics_sinkhorn_reward": 0.7520989775657654,
      "rewards/forensics_strict_edge_reward": 0.07045968994498253,
      "rewards/forensics_tiou_reward": 0.5652929246425629,
      "step": 1127
    },
    {
      "completion_length": 23.84375,
      "epoch": 3.460829493087558,
      "grad_norm": 13.834656715393066,
      "kl": 0.31494140625,
      "learning_rate": 1.3307692307692307e-07,
      "loss": -0.0674,
      "reward": 1.5222322344779968,
      "reward_std": 0.2087024375796318,
      "reward_std_all": 0.193263228982687,
      "rewards-max/forensics_sinkhorn_reward": 0.9636173844337463,
      "rewards-max/forensics_strict_edge_reward": 0.36554911732673645,
      "rewards-max/forensics_tiou_reward": 0.937388002872467,
      "rewards/forensics_sinkhorn_reward": 0.8228119313716888,
      "rewards/forensics_strict_edge_reward": 0.0421520434319973,
      "rewards/forensics_tiou_reward": 0.6572682857513428,
      "step": 1128
    },
    {
      "completion_length": 21.25,
      "epoch": 3.463901689708141,
      "grad_norm": 14.220627784729004,
      "kl": 0.25732421875,
      "learning_rate": 1.323076923076923e-07,
      "loss": 0.028,
      "reward": 1.6520122289657593,
      "reward_std": 0.3070145845413208,
      "reward_std_all": 0.30298957228660583,
      "rewards-max/forensics_sinkhorn_reward": 0.9858180582523346,
      "rewards-max/forensics_strict_edge_reward": 0.9232408702373505,
      "rewards-max/forensics_tiou_reward": 0.992517352104187,
      "rewards/forensics_sinkhorn_reward": 0.8303830623626709,
      "rewards/forensics_strict_edge_reward": 0.12316404283046722,
      "rewards/forensics_tiou_reward": 0.6984652280807495,
      "step": 1129
    },
    {
      "completion_length": 24.046875,
      "epoch": 3.466973886328725,
      "grad_norm": 10.391096115112305,
      "kl": 0.4072265625,
      "learning_rate": 1.3153846153846155e-07,
      "loss": 0.0198,
      "reward": 1.5185996890068054,
      "reward_std": 0.19345282018184662,
      "reward_std_all": 0.18013522773981094,
      "rewards-max/forensics_sinkhorn_reward": 0.9848478436470032,
      "rewards-max/forensics_strict_edge_reward": 0.480892151594162,
      "rewards-max/forensics_tiou_reward": 0.9878726303577423,
      "rewards/forensics_sinkhorn_reward": 0.8186473250389099,
      "rewards/forensics_strict_edge_reward": 0.060752193443477154,
      "rewards/forensics_tiou_reward": 0.6392001807689667,
      "step": 1130
    },
    {
      "completion_length": 21.5,
      "epoch": 3.4700460829493087,
      "grad_norm": 32.03148651123047,
      "kl": 0.240234375,
      "learning_rate": 1.3076923076923076e-07,
      "loss": -0.0414,
      "reward": 1.4551693797111511,
      "reward_std": 0.1470373421907425,
      "reward_std_all": 0.1947208195924759,
      "rewards-max/forensics_sinkhorn_reward": 0.980452835559845,
      "rewards-max/forensics_strict_edge_reward": 0.4635082483291626,
      "rewards-max/forensics_tiou_reward": 0.9727477729320526,
      "rewards/forensics_sinkhorn_reward": 0.8021484911441803,
      "rewards/forensics_strict_edge_reward": 0.06018087640404701,
      "rewards/forensics_tiou_reward": 0.5928401947021484,
      "step": 1131
    },
    {
      "completion_length": 22.8125,
      "epoch": 3.4731182795698925,
      "grad_norm": 19.26459312438965,
      "kl": 0.423828125,
      "learning_rate": 1.3e-07,
      "loss": 0.0965,
      "reward": 1.3281316757202148,
      "reward_std": 0.257322296500206,
      "reward_std_all": 0.23264236748218536,
      "rewards-max/forensics_sinkhorn_reward": 0.9830484688282013,
      "rewards-max/forensics_strict_edge_reward": 0.4866199493408203,
      "rewards-max/forensics_tiou_reward": 0.9845185279846191,
      "rewards/forensics_sinkhorn_reward": 0.7274004220962524,
      "rewards/forensics_strict_edge_reward": 0.0568777397274971,
      "rewards/forensics_tiou_reward": 0.5438534915447235,
      "step": 1132
    },
    {
      "completion_length": 24.453125,
      "epoch": 3.4761904761904763,
      "grad_norm": 8.452847480773926,
      "kl": 0.23095703125,
      "learning_rate": 1.2923076923076924e-07,
      "loss": 0.0556,
      "reward": 1.4443687796592712,
      "reward_std": 0.2497027963399887,
      "reward_std_all": 0.22624950855970383,
      "rewards-max/forensics_sinkhorn_reward": 0.9800090789794922,
      "rewards-max/forensics_strict_edge_reward": 0.43600474298000336,
      "rewards-max/forensics_tiou_reward": 0.9703216254711151,
      "rewards/forensics_sinkhorn_reward": 0.7799939513206482,
      "rewards/forensics_strict_edge_reward": 0.058567481115460396,
      "rewards/forensics_tiou_reward": 0.605807363986969,
      "step": 1133
    },
    {
      "completion_length": 22.25,
      "epoch": 3.47926267281106,
      "grad_norm": 11.527697563171387,
      "kl": 0.599609375,
      "learning_rate": 1.2846153846153845e-07,
      "loss": -0.0442,
      "reward": 1.5904603600502014,
      "reward_std": 0.3518364429473877,
      "reward_std_all": 0.32352229952812195,
      "rewards-max/forensics_sinkhorn_reward": 0.9861449897289276,
      "rewards-max/forensics_strict_edge_reward": 0.7212935090065002,
      "rewards-max/forensics_tiou_reward": 0.992126852273941,
      "rewards/forensics_sinkhorn_reward": 0.8107967674732208,
      "rewards/forensics_strict_edge_reward": 0.12121394649147987,
      "rewards/forensics_tiou_reward": 0.6584495902061462,
      "step": 1134
    },
    {
      "completion_length": 20.8125,
      "epoch": 3.482334869431644,
      "grad_norm": 14.87739372253418,
      "kl": 0.419921875,
      "learning_rate": 1.276923076923077e-07,
      "loss": 0.0259,
      "reward": 1.3330163359642029,
      "reward_std": 0.12539618834853172,
      "reward_std_all": 0.1439792849123478,
      "rewards-max/forensics_sinkhorn_reward": 0.9738632142543793,
      "rewards-max/forensics_strict_edge_reward": 0.2284184694290161,
      "rewards-max/forensics_tiou_reward": 0.9546278119087219,
      "rewards/forensics_sinkhorn_reward": 0.7590895295143127,
      "rewards/forensics_strict_edge_reward": 0.015235533472150564,
      "rewards/forensics_tiou_reward": 0.5586913079023361,
      "step": 1135
    },
    {
      "completion_length": 20.703125,
      "epoch": 3.485407066052227,
      "grad_norm": 11.960951805114746,
      "kl": 0.361328125,
      "learning_rate": 1.269230769230769e-07,
      "loss": -0.0313,
      "reward": 1.4171680212020874,
      "reward_std": 0.28351810574531555,
      "reward_std_all": 0.2557908520102501,
      "rewards-max/forensics_sinkhorn_reward": 0.974370926618576,
      "rewards-max/forensics_strict_edge_reward": 0.49006423354148865,
      "rewards-max/forensics_tiou_reward": 0.9612709283828735,
      "rewards/forensics_sinkhorn_reward": 0.7532077133655548,
      "rewards/forensics_strict_edge_reward": 0.08796868100762367,
      "rewards/forensics_tiou_reward": 0.575991690158844,
      "step": 1136
    },
    {
      "completion_length": 19.5,
      "epoch": 3.488479262672811,
      "grad_norm": 15.505084991455078,
      "kl": 0.3349609375,
      "learning_rate": 1.2615384615384615e-07,
      "loss": 0.0327,
      "reward": 1.3955009579658508,
      "reward_std": 0.26428931951522827,
      "reward_std_all": 0.2783960700035095,
      "rewards-max/forensics_sinkhorn_reward": 0.9598658680915833,
      "rewards-max/forensics_strict_edge_reward": 0.6167194992303848,
      "rewards-max/forensics_tiou_reward": 0.9240719377994537,
      "rewards/forensics_sinkhorn_reward": 0.7517480850219727,
      "rewards/forensics_strict_edge_reward": 0.08844322990626097,
      "rewards/forensics_tiou_reward": 0.5553096383810043,
      "step": 1137
    },
    {
      "completion_length": 23.75,
      "epoch": 3.4915514592933947,
      "grad_norm": 12.58053207397461,
      "kl": 0.2919921875,
      "learning_rate": 1.2538461538461536e-07,
      "loss": -0.0081,
      "reward": 1.2697352766990662,
      "reward_std": 0.1417706273496151,
      "reward_std_all": 0.13262946903705597,
      "rewards-max/forensics_sinkhorn_reward": 0.9652430713176727,
      "rewards-max/forensics_strict_edge_reward": 0.1183152983430773,
      "rewards-max/forensics_tiou_reward": 0.9345998764038086,
      "rewards/forensics_sinkhorn_reward": 0.7469092011451721,
      "rewards/forensics_strict_edge_reward": 0.006420842342777178,
      "rewards/forensics_tiou_reward": 0.5164051949977875,
      "step": 1138
    },
    {
      "completion_length": 21.84375,
      "epoch": 3.4946236559139785,
      "grad_norm": 15.38195514678955,
      "kl": 0.255859375,
      "learning_rate": 1.246153846153846e-07,
      "loss": -0.1344,
      "reward": 1.4955807328224182,
      "reward_std": 0.29616883397102356,
      "reward_std_all": 0.2549894079566002,
      "rewards-max/forensics_sinkhorn_reward": 0.9770427644252777,
      "rewards-max/forensics_strict_edge_reward": 0.5820994228124619,
      "rewards-max/forensics_tiou_reward": 0.9659421443939209,
      "rewards/forensics_sinkhorn_reward": 0.7991633713245392,
      "rewards/forensics_strict_edge_reward": 0.0699932910501957,
      "rewards/forensics_tiou_reward": 0.6264240443706512,
      "step": 1139
    },
    {
      "completion_length": 22.640625,
      "epoch": 3.4976958525345623,
      "grad_norm": 13.359721183776855,
      "kl": 0.5498046875,
      "learning_rate": 1.2384615384615384e-07,
      "loss": -0.0325,
      "reward": 1.4054362773895264,
      "reward_std": 0.2236236408352852,
      "reward_std_all": 0.1961451694369316,
      "rewards-max/forensics_sinkhorn_reward": 0.9699001908302307,
      "rewards-max/forensics_strict_edge_reward": 0.46139125525951385,
      "rewards-max/forensics_tiou_reward": 0.9452098309993744,
      "rewards/forensics_sinkhorn_reward": 0.7589430212974548,
      "rewards/forensics_strict_edge_reward": 0.06474996171891689,
      "rewards/forensics_tiou_reward": 0.5817432403564453,
      "step": 1140
    },
    {
      "completion_length": 20.671875,
      "epoch": 3.500768049155146,
      "grad_norm": 11.217842102050781,
      "kl": 0.26171875,
      "learning_rate": 1.2307692307692308e-07,
      "loss": 0.0154,
      "reward": 1.6009863018989563,
      "reward_std": 0.21246357262134552,
      "reward_std_all": 0.1915287747979164,
      "rewards-max/forensics_sinkhorn_reward": 0.9827211499214172,
      "rewards-max/forensics_strict_edge_reward": 0.7511312067508698,
      "rewards-max/forensics_tiou_reward": 0.9820871651172638,
      "rewards/forensics_sinkhorn_reward": 0.8245699107646942,
      "rewards/forensics_strict_edge_reward": 0.10333362966775894,
      "rewards/forensics_tiou_reward": 0.6730827689170837,
      "step": 1141
    },
    {
      "completion_length": 23.515625,
      "epoch": 3.50384024577573,
      "grad_norm": 13.546146392822266,
      "kl": 0.3134765625,
      "learning_rate": 1.223076923076923e-07,
      "loss": 0.0453,
      "reward": 1.7127419710159302,
      "reward_std": 0.22910666465759277,
      "reward_std_all": 0.23798874020576477,
      "rewards-max/forensics_sinkhorn_reward": 0.9761262536048889,
      "rewards-max/forensics_strict_edge_reward": 0.5179875791072845,
      "rewards-max/forensics_tiou_reward": 0.9700073301792145,
      "rewards/forensics_sinkhorn_reward": 0.8652490973472595,
      "rewards/forensics_strict_edge_reward": 0.10975663736462593,
      "rewards/forensics_tiou_reward": 0.7377362549304962,
      "step": 1142
    },
    {
      "completion_length": 22.40625,
      "epoch": 3.506912442396313,
      "grad_norm": 17.38442611694336,
      "kl": 0.453125,
      "learning_rate": 1.2153846153846154e-07,
      "loss": 0.0542,
      "reward": 1.4706982970237732,
      "reward_std": 0.2954019606113434,
      "reward_std_all": 0.29622291028499603,
      "rewards-max/forensics_sinkhorn_reward": 0.9742194414138794,
      "rewards-max/forensics_strict_edge_reward": 0.39296941459178925,
      "rewards-max/forensics_tiou_reward": 0.9556620419025421,
      "rewards/forensics_sinkhorn_reward": 0.7854408025741577,
      "rewards/forensics_strict_edge_reward": 0.055704984813928604,
      "rewards/forensics_tiou_reward": 0.62955242395401,
      "step": 1143
    },
    {
      "completion_length": 22.859375,
      "epoch": 3.509984639016897,
      "grad_norm": 14.566142082214355,
      "kl": 0.3115234375,
      "learning_rate": 1.2076923076923075e-07,
      "loss": 0.0662,
      "reward": 1.5371080040931702,
      "reward_std": 0.26201532036066055,
      "reward_std_all": 0.23140932619571686,
      "rewards-max/forensics_sinkhorn_reward": 0.9687182307243347,
      "rewards-max/forensics_strict_edge_reward": 0.5284441187977791,
      "rewards-max/forensics_tiou_reward": 0.9490217566490173,
      "rewards/forensics_sinkhorn_reward": 0.8161498606204987,
      "rewards/forensics_strict_edge_reward": 0.07354173110798001,
      "rewards/forensics_tiou_reward": 0.6474163234233856,
      "step": 1144
    },
    {
      "completion_length": 23.546875,
      "epoch": 3.5130568356374807,
      "grad_norm": 11.792001724243164,
      "kl": 0.5634765625,
      "learning_rate": 1.2e-07,
      "loss": 0.0185,
      "reward": 1.510349452495575,
      "reward_std": 0.35430386662483215,
      "reward_std_all": 0.3124210089445114,
      "rewards-max/forensics_sinkhorn_reward": 0.970463901758194,
      "rewards-max/forensics_strict_edge_reward": 0.5483183413743973,
      "rewards-max/forensics_tiou_reward": 0.9492641985416412,
      "rewards/forensics_sinkhorn_reward": 0.794170081615448,
      "rewards/forensics_strict_edge_reward": 0.09223421663045883,
      "rewards/forensics_tiou_reward": 0.6239451467990875,
      "step": 1145
    },
    {
      "completion_length": 20.359375,
      "epoch": 3.5161290322580645,
      "grad_norm": 12.969362258911133,
      "kl": 0.24560546875,
      "learning_rate": 1.1923076923076923e-07,
      "loss": 0.0809,
      "reward": 1.2618558406829834,
      "reward_std": 0.17603865265846252,
      "reward_std_all": 0.16120994091033936,
      "rewards-max/forensics_sinkhorn_reward": 0.9558223485946655,
      "rewards-max/forensics_strict_edge_reward": 0.4109206646680832,
      "rewards-max/forensics_tiou_reward": 0.9182797074317932,
      "rewards/forensics_sinkhorn_reward": 0.7090527415275574,
      "rewards/forensics_strict_edge_reward": 0.038210026919841766,
      "rewards/forensics_tiou_reward": 0.5145930349826813,
      "step": 1146
    },
    {
      "completion_length": 25.015625,
      "epoch": 3.5192012288786483,
      "grad_norm": 15.72164535522461,
      "kl": 0.537109375,
      "learning_rate": 1.1846153846153846e-07,
      "loss": 0.0493,
      "reward": 1.419912040233612,
      "reward_std": 0.24787290394306183,
      "reward_std_all": 0.22611111402511597,
      "rewards-max/forensics_sinkhorn_reward": 0.9828190803527832,
      "rewards-max/forensics_strict_edge_reward": 0.6327048689126968,
      "rewards-max/forensics_tiou_reward": 0.9807020127773285,
      "rewards/forensics_sinkhorn_reward": 0.7548223733901978,
      "rewards/forensics_strict_edge_reward": 0.055574157275259495,
      "rewards/forensics_tiou_reward": 0.6095155477523804,
      "step": 1147
    },
    {
      "completion_length": 22.015625,
      "epoch": 3.522273425499232,
      "grad_norm": 10.1067533493042,
      "kl": 0.3056640625,
      "learning_rate": 1.1769230769230768e-07,
      "loss": 0.1236,
      "reward": 1.3559532165527344,
      "reward_std": 0.24038316309452057,
      "reward_std_all": 0.23402085900306702,
      "rewards-max/forensics_sinkhorn_reward": 0.9704903066158295,
      "rewards-max/forensics_strict_edge_reward": 0.5071093738079071,
      "rewards-max/forensics_tiou_reward": 0.9534367620944977,
      "rewards/forensics_sinkhorn_reward": 0.7609948217868805,
      "rewards/forensics_strict_edge_reward": 0.054206643253564835,
      "rewards/forensics_tiou_reward": 0.5407518148422241,
      "step": 1148
    },
    {
      "completion_length": 24.515625,
      "epoch": 3.525345622119816,
      "grad_norm": 15.390432357788086,
      "kl": 0.4130859375,
      "learning_rate": 1.1692307692307691e-07,
      "loss": 0.0744,
      "reward": 1.3461554646492004,
      "reward_std": 0.27261531352996826,
      "reward_std_all": 0.24080395698547363,
      "rewards-max/forensics_sinkhorn_reward": 0.9717639088630676,
      "rewards-max/forensics_strict_edge_reward": 0.5693852305412292,
      "rewards-max/forensics_tiou_reward": 0.9591191709041595,
      "rewards/forensics_sinkhorn_reward": 0.7207381725311279,
      "rewards/forensics_strict_edge_reward": 0.060793742537498474,
      "rewards/forensics_tiou_reward": 0.5646235346794128,
      "step": 1149
    },
    {
      "completion_length": 21.359375,
      "epoch": 3.528417818740399,
      "grad_norm": 13.056419372558594,
      "kl": 0.28125,
      "learning_rate": 1.1615384615384615e-07,
      "loss": 0.0574,
      "reward": 1.356688916683197,
      "reward_std": 0.22332699596881866,
      "reward_std_all": 0.2151997685432434,
      "rewards-max/forensics_sinkhorn_reward": 0.9570513963699341,
      "rewards-max/forensics_strict_edge_reward": 0.23880668729543686,
      "rewards-max/forensics_tiou_reward": 0.922418475151062,
      "rewards/forensics_sinkhorn_reward": 0.7560504674911499,
      "rewards/forensics_strict_edge_reward": 0.025700521655380726,
      "rewards/forensics_tiou_reward": 0.5749379396438599,
      "step": 1150
    },
    {
      "completion_length": 21.0,
      "epoch": 3.531490015360983,
      "grad_norm": 10.063502311706543,
      "kl": 0.2841796875,
      "learning_rate": 1.1538461538461539e-07,
      "loss": 0.0294,
      "reward": 1.4689524173736572,
      "reward_std": 0.1629912257194519,
      "reward_std_all": 0.1808258518576622,
      "rewards-max/forensics_sinkhorn_reward": 0.984484851360321,
      "rewards-max/forensics_strict_edge_reward": 0.5422053933143616,
      "rewards-max/forensics_tiou_reward": 0.9875247776508331,
      "rewards/forensics_sinkhorn_reward": 0.783484160900116,
      "rewards/forensics_strict_edge_reward": 0.07878406904637814,
      "rewards/forensics_tiou_reward": 0.606684148311615,
      "step": 1151
    },
    {
      "completion_length": 21.34375,
      "epoch": 3.5345622119815667,
      "grad_norm": 13.516379356384277,
      "kl": 0.4326171875,
      "learning_rate": 1.1461538461538462e-07,
      "loss": 0.0801,
      "reward": 1.432556390762329,
      "reward_std": 0.22694851458072662,
      "reward_std_all": 0.2112201601266861,
      "rewards-max/forensics_sinkhorn_reward": 0.9861655831336975,
      "rewards-max/forensics_strict_edge_reward": 0.7331632375717163,
      "rewards-max/forensics_tiou_reward": 0.9893508851528168,
      "rewards/forensics_sinkhorn_reward": 0.7953074276447296,
      "rewards/forensics_strict_edge_reward": 0.06556032225489616,
      "rewards/forensics_tiou_reward": 0.5716886222362518,
      "step": 1152
    },
    {
      "completion_length": 21.28125,
      "epoch": 3.5376344086021505,
      "grad_norm": 13.021549224853516,
      "kl": 0.27783203125,
      "learning_rate": 1.1384615384615384e-07,
      "loss": 0.1174,
      "reward": 1.4697426557540894,
      "reward_std": 0.22350820153951645,
      "reward_std_all": 0.19348671287298203,
      "rewards-max/forensics_sinkhorn_reward": 0.9558830261230469,
      "rewards-max/forensics_strict_edge_reward": 0.42340075969696045,
      "rewards-max/forensics_tiou_reward": 0.9079833328723907,
      "rewards/forensics_sinkhorn_reward": 0.7924492359161377,
      "rewards/forensics_strict_edge_reward": 0.062446387484669685,
      "rewards/forensics_tiou_reward": 0.6148470938205719,
      "step": 1153
    },
    {
      "completion_length": 21.859375,
      "epoch": 3.5407066052227343,
      "grad_norm": 23.433889389038086,
      "kl": 0.3525390625,
      "learning_rate": 1.1307692307692307e-07,
      "loss": -0.0958,
      "reward": 1.5573630332946777,
      "reward_std": 0.26516734063625336,
      "reward_std_all": 0.23374737054109573,
      "rewards-max/forensics_sinkhorn_reward": 0.9720644354820251,
      "rewards-max/forensics_strict_edge_reward": 0.5474212914705276,
      "rewards-max/forensics_tiou_reward": 0.9556550979614258,
      "rewards/forensics_sinkhorn_reward": 0.8153549432754517,
      "rewards/forensics_strict_edge_reward": 0.09574661403894424,
      "rewards/forensics_tiou_reward": 0.6462615132331848,
      "step": 1154
    },
    {
      "completion_length": 24.140625,
      "epoch": 3.543778801843318,
      "grad_norm": 15.84994888305664,
      "kl": 0.32568359375,
      "learning_rate": 1.123076923076923e-07,
      "loss": -0.038,
      "reward": 1.510966956615448,
      "reward_std": 0.30691133439540863,
      "reward_std_all": 0.2973692715167999,
      "rewards-max/forensics_sinkhorn_reward": 0.9771021902561188,
      "rewards-max/forensics_strict_edge_reward": 0.6032288670539856,
      "rewards-max/forensics_tiou_reward": 0.970399022102356,
      "rewards/forensics_sinkhorn_reward": 0.7842106223106384,
      "rewards/forensics_strict_edge_reward": 0.08631323650479317,
      "rewards/forensics_tiou_reward": 0.6404431462287903,
      "step": 1155
    },
    {
      "completion_length": 22.125,
      "epoch": 3.546850998463902,
      "grad_norm": 13.82271957397461,
      "kl": 0.31689453125,
      "learning_rate": 1.1153846153846154e-07,
      "loss": -0.0164,
      "reward": 1.5276620388031006,
      "reward_std": 0.14155763015151024,
      "reward_std_all": 0.12325824052095413,
      "rewards-max/forensics_sinkhorn_reward": 0.9395292103290558,
      "rewards-max/forensics_strict_edge_reward": 0.27180543541908264,
      "rewards-max/forensics_tiou_reward": 0.879317045211792,
      "rewards/forensics_sinkhorn_reward": 0.8250281512737274,
      "rewards/forensics_strict_edge_reward": 0.04615373536944389,
      "rewards/forensics_tiou_reward": 0.6564801633358002,
      "step": 1156
    },
    {
      "completion_length": 26.109375,
      "epoch": 3.549923195084485,
      "grad_norm": 15.627351760864258,
      "kl": 0.3798828125,
      "learning_rate": 1.1076923076923077e-07,
      "loss": -0.0267,
      "reward": 1.7531954050064087,
      "reward_std": 0.17773603647947311,
      "reward_std_all": 0.19032613933086395,
      "rewards-max/forensics_sinkhorn_reward": 0.9834182262420654,
      "rewards-max/forensics_strict_edge_reward": 0.4788461923599243,
      "rewards-max/forensics_tiou_reward": 0.9850775003433228,
      "rewards/forensics_sinkhorn_reward": 0.8780244290828705,
      "rewards/forensics_strict_edge_reward": 0.10464124381542206,
      "rewards/forensics_tiou_reward": 0.7705296874046326,
      "step": 1157
    },
    {
      "completion_length": 22.03125,
      "epoch": 3.5529953917050694,
      "grad_norm": 13.692770004272461,
      "kl": 0.29296875,
      "learning_rate": 1.0999999999999999e-07,
      "loss": -0.0308,
      "reward": 1.529724359512329,
      "reward_std": 0.34701403975486755,
      "reward_std_all": 0.329643115401268,
      "rewards-max/forensics_sinkhorn_reward": 0.9841702580451965,
      "rewards-max/forensics_strict_edge_reward": 0.9093653857707977,
      "rewards-max/forensics_tiou_reward": 0.9890348017215729,
      "rewards/forensics_sinkhorn_reward": 0.7707529067993164,
      "rewards/forensics_strict_edge_reward": 0.13555877283215523,
      "rewards/forensics_tiou_reward": 0.6234126687049866,
      "step": 1158
    },
    {
      "completion_length": 21.5,
      "epoch": 3.5560675883256527,
      "grad_norm": 10.917449951171875,
      "kl": 0.275390625,
      "learning_rate": 1.0923076923076922e-07,
      "loss": 0.0109,
      "reward": 1.441208302974701,
      "reward_std": 0.24401167035102844,
      "reward_std_all": 0.24269971996545792,
      "rewards-max/forensics_sinkhorn_reward": 0.9726881384849548,
      "rewards-max/forensics_strict_edge_reward": 0.55436572432518,
      "rewards-max/forensics_tiou_reward": 0.9585534036159515,
      "rewards/forensics_sinkhorn_reward": 0.7928178906440735,
      "rewards/forensics_strict_edge_reward": 0.06972663104534149,
      "rewards/forensics_tiou_reward": 0.5786637663841248,
      "step": 1159
    },
    {
      "completion_length": 22.265625,
      "epoch": 3.5591397849462365,
      "grad_norm": 12.926118850708008,
      "kl": 0.4189453125,
      "learning_rate": 1.0846153846153845e-07,
      "loss": -0.0918,
      "reward": 1.3878355622291565,
      "reward_std": 0.20420721173286438,
      "reward_std_all": 0.20476646721363068,
      "rewards-max/forensics_sinkhorn_reward": 0.9598614573478699,
      "rewards-max/forensics_strict_edge_reward": 0.4208383075892925,
      "rewards-max/forensics_tiou_reward": 0.930579423904419,
      "rewards/forensics_sinkhorn_reward": 0.7565744519233704,
      "rewards/forensics_strict_edge_reward": 0.05213419906795025,
      "rewards/forensics_tiou_reward": 0.5791269242763519,
      "step": 1160
    },
    {
      "completion_length": 21.03125,
      "epoch": 3.5622119815668203,
      "grad_norm": 14.408655166625977,
      "kl": 0.3857421875,
      "learning_rate": 1.076923076923077e-07,
      "loss": 0.0508,
      "reward": 1.5714796781539917,
      "reward_std": 0.2365589439868927,
      "reward_std_all": 0.22039882838726044,
      "rewards-max/forensics_sinkhorn_reward": 0.9802533686161041,
      "rewards-max/forensics_strict_edge_reward": 0.4974842667579651,
      "rewards-max/forensics_tiou_reward": 0.9765088260173798,
      "rewards/forensics_sinkhorn_reward": 0.8277361690998077,
      "rewards/forensics_strict_edge_reward": 0.07304956391453743,
      "rewards/forensics_tiou_reward": 0.6706938743591309,
      "step": 1161
    },
    {
      "completion_length": 24.0625,
      "epoch": 3.565284178187404,
      "grad_norm": 11.836222648620605,
      "kl": 0.41845703125,
      "learning_rate": 1.0692307692307693e-07,
      "loss": -0.0608,
      "reward": 1.7464916110038757,
      "reward_std": 0.3958979547023773,
      "reward_std_all": 0.3471621870994568,
      "rewards-max/forensics_sinkhorn_reward": 0.9830393493175507,
      "rewards-max/forensics_strict_edge_reward": 0.8351600170135498,
      "rewards-max/forensics_tiou_reward": 0.9853540062904358,
      "rewards/forensics_sinkhorn_reward": 0.8467775583267212,
      "rewards/forensics_strict_edge_reward": 0.17099174112081528,
      "rewards/forensics_tiou_reward": 0.7287222743034363,
      "step": 1162
    },
    {
      "completion_length": 20.234375,
      "epoch": 3.568356374807988,
      "grad_norm": 11.470514297485352,
      "kl": 0.28369140625,
      "learning_rate": 1.0615384615384615e-07,
      "loss": 0.0531,
      "reward": 1.4656199216842651,
      "reward_std": 0.3091510385274887,
      "reward_std_all": 0.25956444442272186,
      "rewards-max/forensics_sinkhorn_reward": 0.9807169735431671,
      "rewards-max/forensics_strict_edge_reward": 0.5987430810928345,
      "rewards-max/forensics_tiou_reward": 0.9765800535678864,
      "rewards/forensics_sinkhorn_reward": 0.78180792927742,
      "rewards/forensics_strict_edge_reward": 0.09889054670929909,
      "rewards/forensics_tiou_reward": 0.5849214196205139,
      "step": 1163
    },
    {
      "completion_length": 19.765625,
      "epoch": 3.571428571428571,
      "grad_norm": 16.20220184326172,
      "kl": 0.26123046875,
      "learning_rate": 1.0538461538461538e-07,
      "loss": 0.0305,
      "reward": 1.327883005142212,
      "reward_std": 0.2907251864671707,
      "reward_std_all": 0.25446461141109467,
      "rewards-max/forensics_sinkhorn_reward": 0.9729939699172974,
      "rewards-max/forensics_strict_edge_reward": 0.5933051109313965,
      "rewards-max/forensics_tiou_reward": 0.962435245513916,
      "rewards/forensics_sinkhorn_reward": 0.7213523983955383,
      "rewards/forensics_strict_edge_reward": 0.07034305483102798,
      "rewards/forensics_tiou_reward": 0.5361875295639038,
      "step": 1164
    },
    {
      "completion_length": 21.625,
      "epoch": 3.5745007680491554,
      "grad_norm": 14.890814781188965,
      "kl": 0.30029296875,
      "learning_rate": 1.0461538461538461e-07,
      "loss": 0.1752,
      "reward": 1.25127512216568,
      "reward_std": 0.2048385590314865,
      "reward_std_all": 0.2275685966014862,
      "rewards-max/forensics_sinkhorn_reward": 0.9350411593914032,
      "rewards-max/forensics_strict_edge_reward": 0.202406607568264,
      "rewards-max/forensics_tiou_reward": 0.8783154189586639,
      "rewards/forensics_sinkhorn_reward": 0.7354398965835571,
      "rewards/forensics_strict_edge_reward": 0.022767492569983006,
      "rewards/forensics_tiou_reward": 0.4930677264928818,
      "step": 1165
    },
    {
      "completion_length": 21.203125,
      "epoch": 3.5775729646697387,
      "grad_norm": 13.584766387939453,
      "kl": 0.21923828125,
      "learning_rate": 1.0384615384615385e-07,
      "loss": 0.0134,
      "reward": 1.3698809146881104,
      "reward_std": 0.19919279217720032,
      "reward_std_all": 0.1800437942147255,
      "rewards-max/forensics_sinkhorn_reward": 0.9753601551055908,
      "rewards-max/forensics_strict_edge_reward": 0.33765287697315216,
      "rewards-max/forensics_tiou_reward": 0.9569361209869385,
      "rewards/forensics_sinkhorn_reward": 0.7607525587081909,
      "rewards/forensics_strict_edge_reward": 0.05460223741829395,
      "rewards/forensics_tiou_reward": 0.5545260310173035,
      "step": 1166
    },
    {
      "completion_length": 19.71875,
      "epoch": 3.5806451612903225,
      "grad_norm": 20.93928337097168,
      "kl": 0.8671875,
      "learning_rate": 1.0307692307692307e-07,
      "loss": 0.1339,
      "reward": 1.3327445983886719,
      "reward_std": 0.2436562217772007,
      "reward_std_all": 0.26784825325012207,
      "rewards-max/forensics_sinkhorn_reward": 0.9613244533538818,
      "rewards-max/forensics_strict_edge_reward": 0.3955255150794983,
      "rewards-max/forensics_tiou_reward": 0.9273821711540222,
      "rewards/forensics_sinkhorn_reward": 0.7606383860111237,
      "rewards/forensics_strict_edge_reward": 0.04276546649634838,
      "rewards/forensics_tiou_reward": 0.5293407440185547,
      "step": 1167
    },
    {
      "completion_length": 24.765625,
      "epoch": 3.5837173579109063,
      "grad_norm": 10.915739059448242,
      "kl": 0.3271484375,
      "learning_rate": 1.023076923076923e-07,
      "loss": -0.0631,
      "reward": 1.5971670746803284,
      "reward_std": 0.2726321145892143,
      "reward_std_all": 0.28534573316574097,
      "rewards-max/forensics_sinkhorn_reward": 0.9726710021495819,
      "rewards-max/forensics_strict_edge_reward": 0.6609791368246078,
      "rewards-max/forensics_tiou_reward": 0.9600017368793488,
      "rewards/forensics_sinkhorn_reward": 0.80381840467453,
      "rewards/forensics_strict_edge_reward": 0.11805551499128342,
      "rewards/forensics_tiou_reward": 0.6752931773662567,
      "step": 1168
    },
    {
      "completion_length": 26.03125,
      "epoch": 3.58678955453149,
      "grad_norm": 10.487375259399414,
      "kl": 0.26025390625,
      "learning_rate": 1.0153846153846153e-07,
      "loss": -0.013,
      "reward": 1.6030444502830505,
      "reward_std": 0.1903563067317009,
      "reward_std_all": 0.19676777720451355,
      "rewards-max/forensics_sinkhorn_reward": 0.9851884841918945,
      "rewards-max/forensics_strict_edge_reward": 0.5254523754119873,
      "rewards-max/forensics_tiou_reward": 0.9870169758796692,
      "rewards/forensics_sinkhorn_reward": 0.8227128982543945,
      "rewards/forensics_strict_edge_reward": 0.08128845132887363,
      "rewards/forensics_tiou_reward": 0.6990431249141693,
      "step": 1169
    },
    {
      "completion_length": 22.203125,
      "epoch": 3.589861751152074,
      "grad_norm": 11.224981307983398,
      "kl": 0.337890625,
      "learning_rate": 1.0076923076923076e-07,
      "loss": 0.0508,
      "reward": 1.5561108589172363,
      "reward_std": 0.21743444353342056,
      "reward_std_all": 0.20834416896104813,
      "rewards-max/forensics_sinkhorn_reward": 0.971277266740799,
      "rewards-max/forensics_strict_edge_reward": 0.5136008560657501,
      "rewards-max/forensics_tiou_reward": 0.9550816118717194,
      "rewards/forensics_sinkhorn_reward": 0.8129557073116302,
      "rewards/forensics_strict_edge_reward": 0.08764985576272011,
      "rewards/forensics_tiou_reward": 0.6555052399635315,
      "step": 1170
    },
    {
      "completion_length": 20.09375,
      "epoch": 3.592933947772657,
      "grad_norm": 13.63050365447998,
      "kl": 0.24853515625,
      "learning_rate": 1e-07,
      "loss": 0.0733,
      "reward": 1.5040867328643799,
      "reward_std": 0.2060464695096016,
      "reward_std_all": 0.1890576332807541,
      "rewards-max/forensics_sinkhorn_reward": 0.9808063507080078,
      "rewards-max/forensics_strict_edge_reward": 0.6858793199062347,
      "rewards-max/forensics_tiou_reward": 0.9783102571964264,
      "rewards/forensics_sinkhorn_reward": 0.7996759712696075,
      "rewards/forensics_strict_edge_reward": 0.09017866104841232,
      "rewards/forensics_tiou_reward": 0.6142320036888123,
      "step": 1171
    },
    {
      "completion_length": 22.515625,
      "epoch": 3.5960061443932414,
      "grad_norm": 12.935108184814453,
      "kl": 0.31640625,
      "learning_rate": 9.923076923076922e-08,
      "loss": 0.0163,
      "reward": 1.419227659702301,
      "reward_std": 0.2603874132037163,
      "reward_std_all": 0.22549262642860413,
      "rewards-max/forensics_sinkhorn_reward": 0.9778027534484863,
      "rewards-max/forensics_strict_edge_reward": 0.6724361926317215,
      "rewards-max/forensics_tiou_reward": 0.9678871333599091,
      "rewards/forensics_sinkhorn_reward": 0.7614935040473938,
      "rewards/forensics_strict_edge_reward": 0.07284235209226608,
      "rewards/forensics_tiou_reward": 0.5848917663097382,
      "step": 1172
    },
    {
      "completion_length": 22.765625,
      "epoch": 3.5990783410138247,
      "grad_norm": 13.23903751373291,
      "kl": 0.33642578125,
      "learning_rate": 9.846153846153846e-08,
      "loss": 0.0941,
      "reward": 1.2525516748428345,
      "reward_std": 0.16657166928052902,
      "reward_std_all": 0.15812034904956818,
      "rewards-max/forensics_sinkhorn_reward": 0.96562659740448,
      "rewards-max/forensics_strict_edge_reward": 0.12828864343464375,
      "rewards-max/forensics_tiou_reward": 0.9401966035366058,
      "rewards/forensics_sinkhorn_reward": 0.7192671597003937,
      "rewards/forensics_strict_edge_reward": 0.01596419094130397,
      "rewards/forensics_tiou_reward": 0.5173202753067017,
      "step": 1173
    },
    {
      "completion_length": 20.0,
      "epoch": 3.6021505376344085,
      "grad_norm": 12.12562370300293,
      "kl": 0.330078125,
      "learning_rate": 9.769230769230769e-08,
      "loss": -0.0335,
      "reward": 1.5618323683738708,
      "reward_std": 0.23277872055768967,
      "reward_std_all": 0.20208293199539185,
      "rewards-max/forensics_sinkhorn_reward": 0.9785160422325134,
      "rewards-max/forensics_strict_edge_reward": 0.6115650832653046,
      "rewards-max/forensics_tiou_reward": 0.9684135913848877,
      "rewards/forensics_sinkhorn_reward": 0.8114695847034454,
      "rewards/forensics_strict_edge_reward": 0.10964228957891464,
      "rewards/forensics_tiou_reward": 0.6407204866409302,
      "step": 1174
    },
    {
      "completion_length": 20.3125,
      "epoch": 3.6052227342549923,
      "grad_norm": 16.280776977539062,
      "kl": 0.283203125,
      "learning_rate": 9.692307692307692e-08,
      "loss": -0.0776,
      "reward": 1.4832438826560974,
      "reward_std": 0.24518316984176636,
      "reward_std_all": 0.23479555547237396,
      "rewards-max/forensics_sinkhorn_reward": 0.984294593334198,
      "rewards-max/forensics_strict_edge_reward": 0.5971094369888306,
      "rewards-max/forensics_tiou_reward": 0.9876165688037872,
      "rewards/forensics_sinkhorn_reward": 0.7835817039012909,
      "rewards/forensics_strict_edge_reward": 0.06774390302598476,
      "rewards/forensics_tiou_reward": 0.6319183707237244,
      "step": 1175
    },
    {
      "completion_length": 21.203125,
      "epoch": 3.608294930875576,
      "grad_norm": 10.133692741394043,
      "kl": 0.32275390625,
      "learning_rate": 9.615384615384616e-08,
      "loss": 0.0245,
      "reward": 1.4396224617958069,
      "reward_std": 0.22358384728431702,
      "reward_std_all": 0.19898468255996704,
      "rewards-max/forensics_sinkhorn_reward": 0.9648990035057068,
      "rewards-max/forensics_strict_edge_reward": 0.37411409616470337,
      "rewards-max/forensics_tiou_reward": 0.9385961890220642,
      "rewards/forensics_sinkhorn_reward": 0.7851454615592957,
      "rewards/forensics_strict_edge_reward": 0.04245593957602978,
      "rewards/forensics_tiou_reward": 0.6120211184024811,
      "step": 1176
    },
    {
      "completion_length": 20.796875,
      "epoch": 3.61136712749616,
      "grad_norm": 11.24094295501709,
      "kl": 0.3251953125,
      "learning_rate": 9.538461538461538e-08,
      "loss": 0.0734,
      "reward": 1.3753355741500854,
      "reward_std": 0.3194015771150589,
      "reward_std_all": 0.279158353805542,
      "rewards-max/forensics_sinkhorn_reward": 0.9688495993614197,
      "rewards-max/forensics_strict_edge_reward": 0.4695305973291397,
      "rewards-max/forensics_tiou_reward": 0.9472824931144714,
      "rewards/forensics_sinkhorn_reward": 0.7380185723304749,
      "rewards/forensics_strict_edge_reward": 0.07006832398474216,
      "rewards/forensics_tiou_reward": 0.567248672246933,
      "step": 1177
    },
    {
      "completion_length": 20.421875,
      "epoch": 3.6144393241167436,
      "grad_norm": 11.311583518981934,
      "kl": 0.26416015625,
      "learning_rate": 9.461538461538461e-08,
      "loss": 0.0527,
      "reward": 1.4201536178588867,
      "reward_std": 0.2136230170726776,
      "reward_std_all": 0.20540816336870193,
      "rewards-max/forensics_sinkhorn_reward": 0.9696572422981262,
      "rewards-max/forensics_strict_edge_reward": 0.43208183348178864,
      "rewards-max/forensics_tiou_reward": 0.950196772813797,
      "rewards/forensics_sinkhorn_reward": 0.7688784301280975,
      "rewards/forensics_strict_edge_reward": 0.05966071039438248,
      "rewards/forensics_tiou_reward": 0.5916144251823425,
      "step": 1178
    },
    {
      "completion_length": 22.828125,
      "epoch": 3.6175115207373274,
      "grad_norm": 17.815351486206055,
      "kl": 0.25830078125,
      "learning_rate": 9.384615384615384e-08,
      "loss": 0.0107,
      "reward": 1.5635475516319275,
      "reward_std": 0.20540890842676163,
      "reward_std_all": 0.1988227218389511,
      "rewards-max/forensics_sinkhorn_reward": 0.9721279740333557,
      "rewards-max/forensics_strict_edge_reward": 0.43394339084625244,
      "rewards-max/forensics_tiou_reward": 0.959175705909729,
      "rewards/forensics_sinkhorn_reward": 0.821024477481842,
      "rewards/forensics_strict_edge_reward": 0.08398620039224625,
      "rewards/forensics_tiou_reward": 0.658536821603775,
      "step": 1179
    },
    {
      "completion_length": 22.375,
      "epoch": 3.6205837173579107,
      "grad_norm": 11.889981269836426,
      "kl": 0.2958984375,
      "learning_rate": 9.307692307692306e-08,
      "loss": -0.0945,
      "reward": 1.6144185066223145,
      "reward_std": 0.16237092018127441,
      "reward_std_all": 0.17748308926820755,
      "rewards-max/forensics_sinkhorn_reward": 0.981394499540329,
      "rewards-max/forensics_strict_edge_reward": 0.5067686438560486,
      "rewards-max/forensics_tiou_reward": 0.9795129597187042,
      "rewards/forensics_sinkhorn_reward": 0.843727171421051,
      "rewards/forensics_strict_edge_reward": 0.07147565111517906,
      "rewards/forensics_tiou_reward": 0.6992157995700836,
      "step": 1180
    },
    {
      "completion_length": 21.921875,
      "epoch": 3.6236559139784945,
      "grad_norm": 14.102791786193848,
      "kl": 0.326171875,
      "learning_rate": 9.23076923076923e-08,
      "loss": 0.1162,
      "reward": 1.5098265409469604,
      "reward_std": 0.29138006269931793,
      "reward_std_all": 0.27431609481573105,
      "rewards-max/forensics_sinkhorn_reward": 0.9819625616073608,
      "rewards-max/forensics_strict_edge_reward": 0.6312568187713623,
      "rewards-max/forensics_tiou_reward": 0.9797770082950592,
      "rewards/forensics_sinkhorn_reward": 0.8196048140525818,
      "rewards/forensics_strict_edge_reward": 0.058300042524933815,
      "rewards/forensics_tiou_reward": 0.6319216191768646,
      "step": 1181
    },
    {
      "completion_length": 21.640625,
      "epoch": 3.6267281105990783,
      "grad_norm": 10.258770942687988,
      "kl": 0.5732421875,
      "learning_rate": 9.153846153846153e-08,
      "loss": -0.0514,
      "reward": 1.568163514137268,
      "reward_std": 0.22140101343393326,
      "reward_std_all": 0.21582740545272827,
      "rewards-max/forensics_sinkhorn_reward": 0.9776959419250488,
      "rewards-max/forensics_strict_edge_reward": 0.6770831942558289,
      "rewards-max/forensics_tiou_reward": 0.9716013073921204,
      "rewards/forensics_sinkhorn_reward": 0.8304807841777802,
      "rewards/forensics_strict_edge_reward": 0.0731959156692028,
      "rewards/forensics_tiou_reward": 0.664486825466156,
      "step": 1182
    },
    {
      "completion_length": 21.359375,
      "epoch": 3.629800307219662,
      "grad_norm": 11.375157356262207,
      "kl": 0.32177734375,
      "learning_rate": 9.076923076923076e-08,
      "loss": -0.0984,
      "reward": 1.536864161491394,
      "reward_std": 0.2274901196360588,
      "reward_std_all": 0.203691266477108,
      "rewards-max/forensics_sinkhorn_reward": 0.9695218801498413,
      "rewards-max/forensics_strict_edge_reward": 0.5185378938913345,
      "rewards-max/forensics_tiou_reward": 0.9494919180870056,
      "rewards/forensics_sinkhorn_reward": 0.8134899139404297,
      "rewards/forensics_strict_edge_reward": 0.06483201682567596,
      "rewards/forensics_tiou_reward": 0.6585423052310944,
      "step": 1183
    },
    {
      "completion_length": 26.078125,
      "epoch": 3.632872503840246,
      "grad_norm": 16.756771087646484,
      "kl": 0.4912109375,
      "learning_rate": 9e-08,
      "loss": 0.0521,
      "reward": 1.4533289074897766,
      "reward_std": 0.37311841547489166,
      "reward_std_all": 0.32396432757377625,
      "rewards-max/forensics_sinkhorn_reward": 0.9716624021530151,
      "rewards-max/forensics_strict_edge_reward": 0.42382363975048065,
      "rewards-max/forensics_tiou_reward": 0.9535903930664062,
      "rewards/forensics_sinkhorn_reward": 0.7815190255641937,
      "rewards/forensics_strict_edge_reward": 0.05700407549738884,
      "rewards/forensics_tiou_reward": 0.6148057579994202,
      "step": 1184
    },
    {
      "completion_length": 22.640625,
      "epoch": 3.6359447004608296,
      "grad_norm": 14.879897117614746,
      "kl": 0.431640625,
      "learning_rate": 8.923076923076923e-08,
      "loss": 0.0067,
      "reward": 1.3700454235076904,
      "reward_std": 0.23327042162418365,
      "reward_std_all": 0.20520158857107162,
      "rewards-max/forensics_sinkhorn_reward": 0.9752586483955383,
      "rewards-max/forensics_strict_edge_reward": 0.5548860281705856,
      "rewards-max/forensics_tiou_reward": 0.9647233188152313,
      "rewards/forensics_sinkhorn_reward": 0.7308155000209808,
      "rewards/forensics_strict_edge_reward": 0.07376411370933056,
      "rewards/forensics_tiou_reward": 0.5654657781124115,
      "step": 1185
    },
    {
      "completion_length": 22.0625,
      "epoch": 3.6390168970814134,
      "grad_norm": 12.722293853759766,
      "kl": 0.54736328125,
      "learning_rate": 8.846153846153847e-08,
      "loss": 0.0857,
      "reward": 1.4254460334777832,
      "reward_std": 0.2149241492152214,
      "reward_std_all": 0.1965138539671898,
      "rewards-max/forensics_sinkhorn_reward": 0.9877383410930634,
      "rewards-max/forensics_strict_edge_reward": 0.5805044323205948,
      "rewards-max/forensics_tiou_reward": 0.9963415563106537,
      "rewards/forensics_sinkhorn_reward": 0.7605707049369812,
      "rewards/forensics_strict_edge_reward": 0.07678288966417313,
      "rewards/forensics_tiou_reward": 0.5880922973155975,
      "step": 1186
    },
    {
      "completion_length": 23.03125,
      "epoch": 3.6420890937019967,
      "grad_norm": 11.351012229919434,
      "kl": 0.51953125,
      "learning_rate": 8.769230769230769e-08,
      "loss": 0.0102,
      "reward": 1.4967743754386902,
      "reward_std": 0.20284519344568253,
      "reward_std_all": 0.19204551726579666,
      "rewards-max/forensics_sinkhorn_reward": 0.9602234065532684,
      "rewards-max/forensics_strict_edge_reward": 0.4209369271993637,
      "rewards-max/forensics_tiou_reward": 0.9220173954963684,
      "rewards/forensics_sinkhorn_reward": 0.8047350943088531,
      "rewards/forensics_strict_edge_reward": 0.0655650906264782,
      "rewards/forensics_tiou_reward": 0.6264741718769073,
      "step": 1187
    },
    {
      "completion_length": 23.609375,
      "epoch": 3.6451612903225805,
      "grad_norm": 9.527201652526855,
      "kl": 0.3544921875,
      "learning_rate": 8.692307692307692e-08,
      "loss": 0.0062,
      "reward": 1.4589137434959412,
      "reward_std": 0.15847064927220345,
      "reward_std_all": 0.18793019652366638,
      "rewards-max/forensics_sinkhorn_reward": 0.9818570017814636,
      "rewards-max/forensics_strict_edge_reward": 0.40982385724782944,
      "rewards-max/forensics_tiou_reward": 0.9780555069446564,
      "rewards/forensics_sinkhorn_reward": 0.7707939147949219,
      "rewards/forensics_strict_edge_reward": 0.05514703318476677,
      "rewards/forensics_tiou_reward": 0.6329728662967682,
      "step": 1188
    },
    {
      "completion_length": 21.5625,
      "epoch": 3.6482334869431643,
      "grad_norm": 13.334453582763672,
      "kl": 0.767578125,
      "learning_rate": 8.615384615384615e-08,
      "loss": 0.1483,
      "reward": 1.5785249471664429,
      "reward_std": 0.2800447642803192,
      "reward_std_all": 0.24246473610401154,
      "rewards-max/forensics_sinkhorn_reward": 0.9781610071659088,
      "rewards-max/forensics_strict_edge_reward": 0.5303425639867783,
      "rewards-max/forensics_tiou_reward": 0.9715803861618042,
      "rewards/forensics_sinkhorn_reward": 0.8348929286003113,
      "rewards/forensics_strict_edge_reward": 0.06097199767827988,
      "rewards/forensics_tiou_reward": 0.6826600730419159,
      "step": 1189
    },
    {
      "completion_length": 20.15625,
      "epoch": 3.651305683563748,
      "grad_norm": 16.148069381713867,
      "kl": 0.578125,
      "learning_rate": 8.538461538461537e-08,
      "loss": -0.0492,
      "reward": 1.4709137678146362,
      "reward_std": 0.21138925105333328,
      "reward_std_all": 0.19389421492815018,
      "rewards-max/forensics_sinkhorn_reward": 0.9765156209468842,
      "rewards-max/forensics_strict_edge_reward": 0.6358188390731812,
      "rewards-max/forensics_tiou_reward": 0.9692667126655579,
      "rewards/forensics_sinkhorn_reward": 0.7981022894382477,
      "rewards/forensics_strict_edge_reward": 0.06717919372022152,
      "rewards/forensics_tiou_reward": 0.6056323051452637,
      "step": 1190
    },
    {
      "completion_length": 20.21875,
      "epoch": 3.654377880184332,
      "grad_norm": 14.075907707214355,
      "kl": 0.564453125,
      "learning_rate": 8.461538461538461e-08,
      "loss": 0.0119,
      "reward": 1.7185649275779724,
      "reward_std": 0.19971036165952682,
      "reward_std_all": 0.21665740758180618,
      "rewards-max/forensics_sinkhorn_reward": 0.9850217998027802,
      "rewards-max/forensics_strict_edge_reward": 0.7212606072425842,
      "rewards-max/forensics_tiou_reward": 0.9903178811073303,
      "rewards/forensics_sinkhorn_reward": 0.873337060213089,
      "rewards/forensics_strict_edge_reward": 0.09662967920303345,
      "rewards/forensics_tiou_reward": 0.7485981583595276,
      "step": 1191
    },
    {
      "completion_length": 20.609375,
      "epoch": 3.6574500768049156,
      "grad_norm": 11.256272315979004,
      "kl": 0.583984375,
      "learning_rate": 8.384615384615384e-08,
      "loss": 0.0314,
      "reward": 1.6069529056549072,
      "reward_std": 0.3918232321739197,
      "reward_std_all": 0.37519319355487823,
      "rewards-max/forensics_sinkhorn_reward": 0.9815933108329773,
      "rewards-max/forensics_strict_edge_reward": 0.8187307715415955,
      "rewards-max/forensics_tiou_reward": 0.9822294116020203,
      "rewards/forensics_sinkhorn_reward": 0.8150915503501892,
      "rewards/forensics_strict_edge_reward": 0.13301394134759903,
      "rewards/forensics_tiou_reward": 0.6588473618030548,
      "step": 1192
    },
    {
      "completion_length": 23.40625,
      "epoch": 3.6605222734254994,
      "grad_norm": 15.54460620880127,
      "kl": 0.5517578125,
      "learning_rate": 8.307692307692307e-08,
      "loss": -0.0262,
      "reward": 1.5620314478874207,
      "reward_std": 0.2733374089002609,
      "reward_std_all": 0.2575625032186508,
      "rewards-max/forensics_sinkhorn_reward": 0.9801641404628754,
      "rewards-max/forensics_strict_edge_reward": 0.5834526717662811,
      "rewards-max/forensics_tiou_reward": 0.9791636168956757,
      "rewards/forensics_sinkhorn_reward": 0.7874554991722107,
      "rewards/forensics_strict_edge_reward": 0.10357092693448067,
      "rewards/forensics_tiou_reward": 0.6710050404071808,
      "step": 1193
    },
    {
      "completion_length": 20.75,
      "epoch": 3.6635944700460827,
      "grad_norm": 16.20049476623535,
      "kl": 0.3271484375,
      "learning_rate": 8.23076923076923e-08,
      "loss": -0.0071,
      "reward": 1.473436415195465,
      "reward_std": 0.2970905378460884,
      "reward_std_all": 0.30386433005332947,
      "rewards-max/forensics_sinkhorn_reward": 0.9821824729442596,
      "rewards-max/forensics_strict_edge_reward": 0.7181241810321808,
      "rewards-max/forensics_tiou_reward": 0.9801999628543854,
      "rewards/forensics_sinkhorn_reward": 0.7827156186103821,
      "rewards/forensics_strict_edge_reward": 0.06449860706925392,
      "rewards/forensics_tiou_reward": 0.626222163438797,
      "step": 1194
    },
    {
      "completion_length": 19.625,
      "epoch": 3.6666666666666665,
      "grad_norm": 13.460834503173828,
      "kl": 0.5107421875,
      "learning_rate": 8.153846153846153e-08,
      "loss": 0.022,
      "reward": 1.4403313994407654,
      "reward_std": 0.28549133986234665,
      "reward_std_all": 0.30970923602581024,
      "rewards-max/forensics_sinkhorn_reward": 0.9781637787818909,
      "rewards-max/forensics_strict_edge_reward": 0.4856214374303818,
      "rewards-max/forensics_tiou_reward": 0.9659876823425293,
      "rewards/forensics_sinkhorn_reward": 0.7542831897735596,
      "rewards/forensics_strict_edge_reward": 0.10521227866411209,
      "rewards/forensics_tiou_reward": 0.5808359682559967,
      "step": 1195
    },
    {
      "completion_length": 20.640625,
      "epoch": 3.6697388632872503,
      "grad_norm": 17.508256912231445,
      "kl": 0.283203125,
      "learning_rate": 8.076923076923077e-08,
      "loss": 0.1455,
      "reward": 1.2338240146636963,
      "reward_std": 0.23015432059764862,
      "reward_std_all": 0.20864955335855484,
      "rewards-max/forensics_sinkhorn_reward": 0.958286702632904,
      "rewards-max/forensics_strict_edge_reward": 0.26825200393795967,
      "rewards-max/forensics_tiou_reward": 0.9286763668060303,
      "rewards/forensics_sinkhorn_reward": 0.6907230019569397,
      "rewards/forensics_strict_edge_reward": 0.03555054683238268,
      "rewards/forensics_tiou_reward": 0.5075504779815674,
      "step": 1196
    },
    {
      "completion_length": 20.90625,
      "epoch": 3.672811059907834,
      "grad_norm": 22.738563537597656,
      "kl": 0.283203125,
      "learning_rate": 8e-08,
      "loss": -0.0618,
      "reward": 1.3089839220046997,
      "reward_std": 0.1638568714261055,
      "reward_std_all": 0.17700199782848358,
      "rewards-max/forensics_sinkhorn_reward": 0.8616932332515717,
      "rewards-max/forensics_strict_edge_reward": 0.24884484708309174,
      "rewards-max/forensics_tiou_reward": 0.7894763350486755,
      "rewards/forensics_sinkhorn_reward": 0.7321159243583679,
      "rewards/forensics_strict_edge_reward": 0.03184989653527737,
      "rewards/forensics_tiou_reward": 0.5450181365013123,
      "step": 1197
    },
    {
      "completion_length": 24.171875,
      "epoch": 3.675883256528418,
      "grad_norm": 12.604308128356934,
      "kl": 0.43359375,
      "learning_rate": 7.923076923076923e-08,
      "loss": 0.0058,
      "reward": 1.2219173908233643,
      "reward_std": 0.3036974221467972,
      "reward_std_all": 0.30255092680454254,
      "rewards-max/forensics_sinkhorn_reward": 0.9457089304924011,
      "rewards-max/forensics_strict_edge_reward": 0.3758818507194519,
      "rewards-max/forensics_tiou_reward": 0.886040985584259,
      "rewards/forensics_sinkhorn_reward": 0.6823537647724152,
      "rewards/forensics_strict_edge_reward": 0.04668642394244671,
      "rewards/forensics_tiou_reward": 0.49287712574005127,
      "step": 1198
    },
    {
      "completion_length": 22.328125,
      "epoch": 3.6789554531490016,
      "grad_norm": 15.36886978149414,
      "kl": 0.376953125,
      "learning_rate": 7.846153846153846e-08,
      "loss": 0.0935,
      "reward": 1.5050826668739319,
      "reward_std": 0.22819752246141434,
      "reward_std_all": 0.20942479372024536,
      "rewards-max/forensics_sinkhorn_reward": 0.9723995625972748,
      "rewards-max/forensics_strict_edge_reward": 0.538663849234581,
      "rewards-max/forensics_tiou_reward": 0.9543881416320801,
      "rewards/forensics_sinkhorn_reward": 0.8005509972572327,
      "rewards/forensics_strict_edge_reward": 0.08277385495603085,
      "rewards/forensics_tiou_reward": 0.6217577457427979,
      "step": 1199
    },
    {
      "completion_length": 21.796875,
      "epoch": 3.6820276497695854,
      "grad_norm": 10.835665702819824,
      "kl": 0.31103515625,
      "learning_rate": 7.769230769230768e-08,
      "loss": -0.0062,
      "reward": 1.5540485978126526,
      "reward_std": 0.25672122836112976,
      "reward_std_all": 0.23621375858783722,
      "rewards-max/forensics_sinkhorn_reward": 0.9669260382652283,
      "rewards-max/forensics_strict_edge_reward": 0.4490247070789337,
      "rewards-max/forensics_tiou_reward": 0.9300172030925751,
      "rewards/forensics_sinkhorn_reward": 0.7947546243667603,
      "rewards/forensics_strict_edge_reward": 0.09724600985646248,
      "rewards/forensics_tiou_reward": 0.6620479077100754,
      "step": 1200
    },
    {
      "completion_length": 21.203125,
      "epoch": 3.6850998463901687,
      "grad_norm": 13.567333221435547,
      "kl": 0.3017578125,
      "learning_rate": 7.692307692307692e-08,
      "loss": -0.0719,
      "reward": 1.650819718837738,
      "reward_std": 0.22496619075536728,
      "reward_std_all": 0.22539778798818588,
      "rewards-max/forensics_sinkhorn_reward": 0.965095043182373,
      "rewards-max/forensics_strict_edge_reward": 0.3640780746936798,
      "rewards-max/forensics_tiou_reward": 0.9392834305763245,
      "rewards/forensics_sinkhorn_reward": 0.84896320104599,
      "rewards/forensics_strict_edge_reward": 0.07320381328463554,
      "rewards/forensics_tiou_reward": 0.728652685880661,
      "step": 1201
    },
    {
      "completion_length": 21.375,
      "epoch": 3.688172043010753,
      "grad_norm": 14.994315147399902,
      "kl": 0.576171875,
      "learning_rate": 7.615384615384615e-08,
      "loss": -0.0217,
      "reward": 1.4440522193908691,
      "reward_std": 0.16586291044950485,
      "reward_std_all": 0.15491227805614471,
      "rewards-max/forensics_sinkhorn_reward": 0.9757440984249115,
      "rewards-max/forensics_strict_edge_reward": 0.5956533849239349,
      "rewards-max/forensics_tiou_reward": 0.965694010257721,
      "rewards/forensics_sinkhorn_reward": 0.7942918241024017,
      "rewards/forensics_strict_edge_reward": 0.061898622661828995,
      "rewards/forensics_tiou_reward": 0.5878616869449615,
      "step": 1202
    },
    {
      "completion_length": 25.203125,
      "epoch": 3.6912442396313363,
      "grad_norm": 11.526715278625488,
      "kl": 0.283203125,
      "learning_rate": 7.538461538461538e-08,
      "loss": -0.0399,
      "reward": 1.4867497682571411,
      "reward_std": 0.22495771199464798,
      "reward_std_all": 0.21410098671913147,
      "rewards-max/forensics_sinkhorn_reward": 0.9784602522850037,
      "rewards-max/forensics_strict_edge_reward": 0.5984183251857758,
      "rewards-max/forensics_tiou_reward": 0.9723435342311859,
      "rewards/forensics_sinkhorn_reward": 0.7652847766876221,
      "rewards/forensics_strict_edge_reward": 0.11635401844978333,
      "rewards/forensics_tiou_reward": 0.6051109731197357,
      "step": 1203
    },
    {
      "completion_length": 21.953125,
      "epoch": 3.69431643625192,
      "grad_norm": 14.47504997253418,
      "kl": 0.3203125,
      "learning_rate": 7.46153846153846e-08,
      "loss": 0.0735,
      "reward": 1.3878160119056702,
      "reward_std": 0.16784800589084625,
      "reward_std_all": 0.15191654488444328,
      "rewards-max/forensics_sinkhorn_reward": 0.9721055924892426,
      "rewards-max/forensics_strict_edge_reward": 0.31884680688381195,
      "rewards-max/forensics_tiou_reward": 0.9519280791282654,
      "rewards/forensics_sinkhorn_reward": 0.7935814261436462,
      "rewards/forensics_strict_edge_reward": 0.02613675221800804,
      "rewards/forensics_tiou_reward": 0.5680978298187256,
      "step": 1204
    },
    {
      "completion_length": 20.5625,
      "epoch": 3.697388632872504,
      "grad_norm": 12.093963623046875,
      "kl": 0.2021484375,
      "learning_rate": 7.384615384615384e-08,
      "loss": 0.0692,
      "reward": 1.676020622253418,
      "reward_std": 0.23036856949329376,
      "reward_std_all": 0.26628577709198,
      "rewards-max/forensics_sinkhorn_reward": 0.9819158315658569,
      "rewards-max/forensics_strict_edge_reward": 0.6382447481155396,
      "rewards-max/forensics_tiou_reward": 0.9773622453212738,
      "rewards/forensics_sinkhorn_reward": 0.8644300103187561,
      "rewards/forensics_strict_edge_reward": 0.08964655920863152,
      "rewards/forensics_tiou_reward": 0.7219440340995789,
      "step": 1205
    },
    {
      "completion_length": 23.15625,
      "epoch": 3.7004608294930876,
      "grad_norm": 10.42260456085205,
      "kl": 0.318359375,
      "learning_rate": 7.307692307692308e-08,
      "loss": -0.0694,
      "reward": 1.484461784362793,
      "reward_std": 0.17720182985067368,
      "reward_std_all": 0.18471022695302963,
      "rewards-max/forensics_sinkhorn_reward": 0.9715356230735779,
      "rewards-max/forensics_strict_edge_reward": 0.2783227041363716,
      "rewards-max/forensics_tiou_reward": 0.9433808624744415,
      "rewards/forensics_sinkhorn_reward": 0.8178682923316956,
      "rewards/forensics_strict_edge_reward": 0.030692055821418762,
      "rewards/forensics_tiou_reward": 0.6359013319015503,
      "step": 1206
    },
    {
      "completion_length": 24.9375,
      "epoch": 3.7035330261136714,
      "grad_norm": 12.014887809753418,
      "kl": 0.2958984375,
      "learning_rate": 7.230769230769231e-08,
      "loss": -0.1051,
      "reward": 1.640864074230194,
      "reward_std": 0.2761099264025688,
      "reward_std_all": 0.2449059933423996,
      "rewards-max/forensics_sinkhorn_reward": 0.9791883230209351,
      "rewards-max/forensics_strict_edge_reward": 0.5851600170135498,
      "rewards-max/forensics_tiou_reward": 0.9716394245624542,
      "rewards/forensics_sinkhorn_reward": 0.8092182278633118,
      "rewards/forensics_strict_edge_reward": 0.136390782892704,
      "rewards/forensics_tiou_reward": 0.6952550113201141,
      "step": 1207
    },
    {
      "completion_length": 21.890625,
      "epoch": 3.7066052227342547,
      "grad_norm": 14.137670516967773,
      "kl": 0.21630859375,
      "learning_rate": 7.153846153846154e-08,
      "loss": 0.0767,
      "reward": 1.4949983358383179,
      "reward_std": 0.2089412957429886,
      "reward_std_all": 0.1897849217057228,
      "rewards-max/forensics_sinkhorn_reward": 0.9652807414531708,
      "rewards-max/forensics_strict_edge_reward": 0.5649383366107941,
      "rewards-max/forensics_tiou_reward": 0.9342481195926666,
      "rewards/forensics_sinkhorn_reward": 0.791344553232193,
      "rewards/forensics_strict_edge_reward": 0.0832426305860281,
      "rewards/forensics_tiou_reward": 0.620411217212677,
      "step": 1208
    },
    {
      "completion_length": 22.59375,
      "epoch": 3.709677419354839,
      "grad_norm": 11.435212135314941,
      "kl": 0.43896484375,
      "learning_rate": 7.076923076923076e-08,
      "loss": 0.027,
      "reward": 1.5776425004005432,
      "reward_std": 0.22875171899795532,
      "reward_std_all": 0.22095011919736862,
      "rewards-max/forensics_sinkhorn_reward": 0.9783132672309875,
      "rewards-max/forensics_strict_edge_reward": 0.5755647718906403,
      "rewards-max/forensics_tiou_reward": 0.9736563861370087,
      "rewards/forensics_sinkhorn_reward": 0.8043866753578186,
      "rewards/forensics_strict_edge_reward": 0.12123198434710503,
      "rewards/forensics_tiou_reward": 0.6520238816738129,
      "step": 1209
    },
    {
      "completion_length": 23.234375,
      "epoch": 3.7127496159754223,
      "grad_norm": 11.000686645507812,
      "kl": 0.6591796875,
      "learning_rate": 7e-08,
      "loss": 0.105,
      "reward": 1.4344196915626526,
      "reward_std": 0.3049621134996414,
      "reward_std_all": 0.26484689116477966,
      "rewards-max/forensics_sinkhorn_reward": 0.9873187839984894,
      "rewards-max/forensics_strict_edge_reward": 0.8704091012477875,
      "rewards-max/forensics_tiou_reward": 0.9957875907421112,
      "rewards/forensics_sinkhorn_reward": 0.7650985419750214,
      "rewards/forensics_strict_edge_reward": 0.07853168621659279,
      "rewards/forensics_tiou_reward": 0.5907894670963287,
      "step": 1210
    },
    {
      "completion_length": 21.609375,
      "epoch": 3.715821812596006,
      "grad_norm": 13.426464080810547,
      "kl": 0.2470703125,
      "learning_rate": 6.923076923076923e-08,
      "loss": 0.0214,
      "reward": 1.556739330291748,
      "reward_std": 0.21172255277633667,
      "reward_std_all": 0.1903703734278679,
      "rewards-max/forensics_sinkhorn_reward": 0.9756771624088287,
      "rewards-max/forensics_strict_edge_reward": 0.4871364235877991,
      "rewards-max/forensics_tiou_reward": 0.963463693857193,
      "rewards/forensics_sinkhorn_reward": 0.8348744511604309,
      "rewards/forensics_strict_edge_reward": 0.05023595876991749,
      "rewards/forensics_tiou_reward": 0.6716288626194,
      "step": 1211
    },
    {
      "completion_length": 23.96875,
      "epoch": 3.71889400921659,
      "grad_norm": 25.69651985168457,
      "kl": 0.3095703125,
      "learning_rate": 6.846153846153846e-08,
      "loss": 0.0602,
      "reward": 1.3145359754562378,
      "reward_std": 0.20944354683160782,
      "reward_std_all": 0.1970359832048416,
      "rewards-max/forensics_sinkhorn_reward": 0.9346991181373596,
      "rewards-max/forensics_strict_edge_reward": 0.2691841498017311,
      "rewards-max/forensics_tiou_reward": 0.8817006051540375,
      "rewards/forensics_sinkhorn_reward": 0.7433163523674011,
      "rewards/forensics_strict_edge_reward": 0.02408407162874937,
      "rewards/forensics_tiou_reward": 0.5471355766057968,
      "step": 1212
    },
    {
      "completion_length": 25.15625,
      "epoch": 3.7219662058371736,
      "grad_norm": 10.944231986999512,
      "kl": 0.29736328125,
      "learning_rate": 6.769230769230769e-08,
      "loss": 0.0143,
      "reward": 1.4380473494529724,
      "reward_std": 0.21734821051359177,
      "reward_std_all": 0.25023437291383743,
      "rewards-max/forensics_sinkhorn_reward": 0.9715414643287659,
      "rewards-max/forensics_strict_edge_reward": 0.4410206526517868,
      "rewards-max/forensics_tiou_reward": 0.9512232840061188,
      "rewards/forensics_sinkhorn_reward": 0.7749650776386261,
      "rewards/forensics_strict_edge_reward": 0.06505529209971428,
      "rewards/forensics_tiou_reward": 0.5980269610881805,
      "step": 1213
    },
    {
      "completion_length": 21.640625,
      "epoch": 3.7250384024577574,
      "grad_norm": 11.301273345947266,
      "kl": 0.591796875,
      "learning_rate": 6.692307692307691e-08,
      "loss": 0.1099,
      "reward": 1.236173003911972,
      "reward_std": 0.18610695004463196,
      "reward_std_all": 0.1863366663455963,
      "rewards-max/forensics_sinkhorn_reward": 0.9711289405822754,
      "rewards-max/forensics_strict_edge_reward": 0.4643923193216324,
      "rewards-max/forensics_tiou_reward": 0.9533521831035614,
      "rewards/forensics_sinkhorn_reward": 0.7092556953430176,
      "rewards/forensics_strict_edge_reward": 0.045259881764650345,
      "rewards/forensics_tiou_reward": 0.4816575348377228,
      "step": 1214
    },
    {
      "completion_length": 22.46875,
      "epoch": 3.7281105990783407,
      "grad_norm": 12.812104225158691,
      "kl": 0.2919921875,
      "learning_rate": 6.615384615384615e-08,
      "loss": 0.0891,
      "reward": 1.6149566173553467,
      "reward_std": 0.2656511887907982,
      "reward_std_all": 0.2480163425207138,
      "rewards-max/forensics_sinkhorn_reward": 0.9761077761650085,
      "rewards-max/forensics_strict_edge_reward": 0.4120378792285919,
      "rewards-max/forensics_tiou_reward": 0.9607149362564087,
      "rewards/forensics_sinkhorn_reward": 0.8216458857059479,
      "rewards/forensics_strict_edge_reward": 0.100181526504457,
      "rewards/forensics_tiou_reward": 0.693129301071167,
      "step": 1215
    },
    {
      "completion_length": 21.0625,
      "epoch": 3.731182795698925,
      "grad_norm": 11.86478042602539,
      "kl": 0.484375,
      "learning_rate": 6.538461538461538e-08,
      "loss": 0.0223,
      "reward": 1.3011432886123657,
      "reward_std": 0.3083798289299011,
      "reward_std_all": 0.31078697741031647,
      "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": 0.7042204737663269,
      "rewards/forensics_strict_edge_reward": 0.07464052923023701,
      "rewards/forensics_tiou_reward": 0.5222822427749634,
      "step": 1216
    },
    {
      "completion_length": 24.875,
      "epoch": 3.7342549923195083,
      "grad_norm": 14.990229606628418,
      "kl": 0.2470703125,
      "learning_rate": 6.461538461538462e-08,
      "loss": 0.0085,
      "reward": 1.4435514211654663,
      "reward_std": 0.2577839195728302,
      "reward_std_all": 0.2453710213303566,
      "rewards-max/forensics_sinkhorn_reward": 0.9835713505744934,
      "rewards-max/forensics_strict_edge_reward": 0.7165313363075256,
      "rewards-max/forensics_tiou_reward": 0.984746664762497,
      "rewards/forensics_sinkhorn_reward": 0.7791679799556732,
      "rewards/forensics_strict_edge_reward": 0.06686607003211975,
      "rewards/forensics_tiou_reward": 0.5975173115730286,
      "step": 1217
    },
    {
      "completion_length": 22.046875,
      "epoch": 3.737327188940092,
      "grad_norm": 12.111122131347656,
      "kl": 0.5947265625,
      "learning_rate": 6.384615384615385e-08,
      "loss": 0.0763,
      "reward": 1.6672340035438538,
      "reward_std": 0.29002923518419266,
      "reward_std_all": 0.29649996757507324,
      "rewards-max/forensics_sinkhorn_reward": 0.9768354892730713,
      "rewards-max/forensics_strict_edge_reward": 0.6706924736499786,
      "rewards-max/forensics_tiou_reward": 0.9713174998760223,
      "rewards/forensics_sinkhorn_reward": 0.84199458360672,
      "rewards/forensics_strict_edge_reward": 0.12216108292341232,
      "rewards/forensics_tiou_reward": 0.703078418970108,
      "step": 1218
    },
    {
      "completion_length": 20.640625,
      "epoch": 3.740399385560676,
      "grad_norm": 14.708993911743164,
      "kl": 0.732421875,
      "learning_rate": 6.307692307692307e-08,
      "loss": -0.1258,
      "reward": 1.4514049291610718,
      "reward_std": 0.20505010336637497,
      "reward_std_all": 0.1824052706360817,
      "rewards-max/forensics_sinkhorn_reward": 0.9641295969486237,
      "rewards-max/forensics_strict_edge_reward": 0.3696350008249283,
      "rewards-max/forensics_tiou_reward": 0.9330345690250397,
      "rewards/forensics_sinkhorn_reward": 0.790814608335495,
      "rewards/forensics_strict_edge_reward": 0.059358443133533,
      "rewards/forensics_tiou_reward": 0.6012318432331085,
      "step": 1219
    },
    {
      "completion_length": 20.765625,
      "epoch": 3.7434715821812596,
      "grad_norm": 15.118330955505371,
      "kl": 0.380859375,
      "learning_rate": 6.23076923076923e-08,
      "loss": 0.0476,
      "reward": 1.3992759585380554,
      "reward_std": 0.2327343076467514,
      "reward_std_all": 0.21080531924962997,
      "rewards-max/forensics_sinkhorn_reward": 0.9516964852809906,
      "rewards-max/forensics_strict_edge_reward": 0.49555233120918274,
      "rewards-max/forensics_tiou_reward": 0.9097053706645966,
      "rewards/forensics_sinkhorn_reward": 0.7788634896278381,
      "rewards/forensics_strict_edge_reward": 0.052761977538466454,
      "rewards/forensics_tiou_reward": 0.5676504969596863,
      "step": 1220
    },
    {
      "completion_length": 22.046875,
      "epoch": 3.7465437788018434,
      "grad_norm": 12.83151912689209,
      "kl": 0.31640625,
      "learning_rate": 6.153846153846154e-08,
      "loss": 0.0388,
      "reward": 1.3605149388313293,
      "reward_std": 0.12137703597545624,
      "reward_std_all": 0.14221592992544174,
      "rewards-max/forensics_sinkhorn_reward": 0.9310883581638336,
      "rewards-max/forensics_strict_edge_reward": 0.08925481140613556,
      "rewards-max/forensics_tiou_reward": 0.8558656573295593,
      "rewards/forensics_sinkhorn_reward": 0.78046053647995,
      "rewards/forensics_strict_edge_reward": 0.011933713220059872,
      "rewards/forensics_tiou_reward": 0.568120688199997,
      "step": 1221
    },
    {
      "completion_length": 20.453125,
      "epoch": 3.749615975422427,
      "grad_norm": 13.328353881835938,
      "kl": 0.310546875,
      "learning_rate": 6.076923076923077e-08,
      "loss": 0.0666,
      "reward": 1.4832103252410889,
      "reward_std": 0.2535398304462433,
      "reward_std_all": 0.22788186371326447,
      "rewards-max/forensics_sinkhorn_reward": 0.9886462390422821,
      "rewards-max/forensics_strict_edge_reward": 0.7654097378253937,
      "rewards-max/forensics_tiou_reward": 0.9988633692264557,
      "rewards/forensics_sinkhorn_reward": 0.771951287984848,
      "rewards/forensics_strict_edge_reward": 0.09420815855264664,
      "rewards/forensics_tiou_reward": 0.6170508563518524,
      "step": 1222
    },
    {
      "completion_length": 22.359375,
      "epoch": 3.752688172043011,
      "grad_norm": 13.9771146774292,
      "kl": 0.62109375,
      "learning_rate": 6e-08,
      "loss": 0.0646,
      "reward": 1.3089892864227295,
      "reward_std": 0.30090777575969696,
      "reward_std_all": 0.292149156332016,
      "rewards-max/forensics_sinkhorn_reward": 0.9386450946331024,
      "rewards-max/forensics_strict_edge_reward": 0.3996507991105318,
      "rewards-max/forensics_tiou_reward": 0.8824295103549957,
      "rewards/forensics_sinkhorn_reward": 0.7421984076499939,
      "rewards/forensics_strict_edge_reward": 0.048770498018711805,
      "rewards/forensics_tiou_reward": 0.518020361661911,
      "step": 1223
    },
    {
      "completion_length": 20.5,
      "epoch": 3.7557603686635943,
      "grad_norm": 14.715097427368164,
      "kl": 0.2314453125,
      "learning_rate": 5.923076923076923e-08,
      "loss": 0.1574,
      "reward": 1.5984033346176147,
      "reward_std": 0.29451896250247955,
      "reward_std_all": 0.2790762782096863,
      "rewards-max/forensics_sinkhorn_reward": 0.9852658808231354,
      "rewards-max/forensics_strict_edge_reward": 0.8540171384811401,
      "rewards-max/forensics_tiou_reward": 0.9892303645610809,
      "rewards/forensics_sinkhorn_reward": 0.8161270022392273,
      "rewards/forensics_strict_edge_reward": 0.10092667490243912,
      "rewards/forensics_tiou_reward": 0.6813495755195618,
      "step": 1224
    },
    {
      "completion_length": 19.75,
      "epoch": 3.758832565284178,
      "grad_norm": 19.592273712158203,
      "kl": 0.39453125,
      "learning_rate": 5.8461538461538455e-08,
      "loss": -0.0034,
      "reward": 1.670938491821289,
      "reward_std": 0.3007744550704956,
      "reward_std_all": 0.2712128534913063,
      "rewards-max/forensics_sinkhorn_reward": 0.9847595691680908,
      "rewards-max/forensics_strict_edge_reward": 0.7026884853839874,
      "rewards-max/forensics_tiou_reward": 0.9872500896453857,
      "rewards/forensics_sinkhorn_reward": 0.843900740146637,
      "rewards/forensics_strict_edge_reward": 0.12095064669847488,
      "rewards/forensics_tiou_reward": 0.7060869932174683,
      "step": 1225
    },
    {
      "completion_length": 21.734375,
      "epoch": 3.761904761904762,
      "grad_norm": 15.05764102935791,
      "kl": 0.53173828125,
      "learning_rate": 5.7692307692307695e-08,
      "loss": -0.0426,
      "reward": 1.3924859166145325,
      "reward_std": 0.21433386206626892,
      "reward_std_all": 0.21061839163303375,
      "rewards-max/forensics_sinkhorn_reward": 0.9577133059501648,
      "rewards-max/forensics_strict_edge_reward": 0.5210043489933014,
      "rewards-max/forensics_tiou_reward": 0.9223025441169739,
      "rewards/forensics_sinkhorn_reward": 0.765135794878006,
      "rewards/forensics_strict_edge_reward": 0.06841129064559937,
      "rewards/forensics_tiou_reward": 0.5589388012886047,
      "step": 1226
    },
    {
      "completion_length": 23.015625,
      "epoch": 3.7649769585253456,
      "grad_norm": 12.605094909667969,
      "kl": 0.53515625,
      "learning_rate": 5.692307692307692e-08,
      "loss": 0.0312,
      "reward": 1.4253953695297241,
      "reward_std": 0.24225742369890213,
      "reward_std_all": 0.22597669064998627,
      "rewards-max/forensics_sinkhorn_reward": 0.9413165748119354,
      "rewards-max/forensics_strict_edge_reward": 0.41594958305358887,
      "rewards-max/forensics_tiou_reward": 0.8860142529010773,
      "rewards/forensics_sinkhorn_reward": 0.7657046020030975,
      "rewards/forensics_strict_edge_reward": 0.09630518965423107,
      "rewards/forensics_tiou_reward": 0.5633854866027832,
      "step": 1227
    },
    {
      "completion_length": 22.1875,
      "epoch": 3.7680491551459294,
      "grad_norm": 19.68528175354004,
      "kl": 0.294921875,
      "learning_rate": 5.615384615384615e-08,
      "loss": -0.0542,
      "reward": 1.4674967527389526,
      "reward_std": 0.2281968891620636,
      "reward_std_all": 0.23870818316936493,
      "rewards-max/forensics_sinkhorn_reward": 0.9733947515487671,
      "rewards-max/forensics_strict_edge_reward": 0.349030464887619,
      "rewards-max/forensics_tiou_reward": 0.962107926607132,
      "rewards/forensics_sinkhorn_reward": 0.8016782104969025,
      "rewards/forensics_strict_edge_reward": 0.039323147386312485,
      "rewards/forensics_tiou_reward": 0.6264954805374146,
      "step": 1228
    },
    {
      "completion_length": 20.375,
      "epoch": 3.771121351766513,
      "grad_norm": 32.516876220703125,
      "kl": 0.869140625,
      "learning_rate": 5.538461538461538e-08,
      "loss": 0.1584,
      "reward": 1.2163947224617004,
      "reward_std": 0.19375228881835938,
      "reward_std_all": 0.18091165274381638,
      "rewards-max/forensics_sinkhorn_reward": 0.9290765821933746,
      "rewards-max/forensics_strict_edge_reward": 0.2556847035884857,
      "rewards-max/forensics_tiou_reward": 0.8529095351696014,
      "rewards/forensics_sinkhorn_reward": 0.7180753350257874,
      "rewards/forensics_strict_edge_reward": 0.027695599477738142,
      "rewards/forensics_tiou_reward": 0.47062385082244873,
      "step": 1229
    },
    {
      "completion_length": 20.234375,
      "epoch": 3.774193548387097,
      "grad_norm": 15.817482948303223,
      "kl": 0.3662109375,
      "learning_rate": 5.461538461538461e-08,
      "loss": -0.0515,
      "reward": 1.4789803624153137,
      "reward_std": 0.26343825459480286,
      "reward_std_all": 0.2452303171157837,
      "rewards-max/forensics_sinkhorn_reward": 0.9791397452354431,
      "rewards-max/forensics_strict_edge_reward": 0.5918685793876648,
      "rewards-max/forensics_tiou_reward": 0.976677268743515,
      "rewards/forensics_sinkhorn_reward": 0.775905042886734,
      "rewards/forensics_strict_edge_reward": 0.0824689157307148,
      "rewards/forensics_tiou_reward": 0.6206064820289612,
      "step": 1230
    },
    {
      "completion_length": 24.640625,
      "epoch": 3.7772657450076803,
      "grad_norm": 9.405116081237793,
      "kl": 0.2646484375,
      "learning_rate": 5.384615384615385e-08,
      "loss": 0.0494,
      "reward": 1.5738797783851624,
      "reward_std": 0.25213032960891724,
      "reward_std_all": 0.2574779838323593,
      "rewards-max/forensics_sinkhorn_reward": 0.9804307222366333,
      "rewards-max/forensics_strict_edge_reward": 0.5659761875867844,
      "rewards-max/forensics_tiou_reward": 0.9788453578948975,
      "rewards/forensics_sinkhorn_reward": 0.8224056661128998,
      "rewards/forensics_strict_edge_reward": 0.07611648738384247,
      "rewards/forensics_tiou_reward": 0.6753575205802917,
      "step": 1231
    },
    {
      "completion_length": 19.71875,
      "epoch": 3.780337941628264,
      "grad_norm": 14.340348243713379,
      "kl": 0.3603515625,
      "learning_rate": 5.307692307692308e-08,
      "loss": 0.0398,
      "reward": 1.4756828546524048,
      "reward_std": 0.26375989615917206,
      "reward_std_all": 0.24853572249412537,
      "rewards-max/forensics_sinkhorn_reward": 0.9586518108844757,
      "rewards-max/forensics_strict_edge_reward": 0.3846660181879997,
      "rewards-max/forensics_tiou_reward": 0.9238814413547516,
      "rewards/forensics_sinkhorn_reward": 0.7930733263492584,
      "rewards/forensics_strict_edge_reward": 0.07107982225716114,
      "rewards/forensics_tiou_reward": 0.6115297675132751,
      "step": 1232
    },
    {
      "completion_length": 23.765625,
      "epoch": 3.783410138248848,
      "grad_norm": 11.149214744567871,
      "kl": 0.29296875,
      "learning_rate": 5.2307692307692304e-08,
      "loss": -0.0496,
      "reward": 1.5309777855873108,
      "reward_std": 0.2256820648908615,
      "reward_std_all": 0.1991971656680107,
      "rewards-max/forensics_sinkhorn_reward": 0.9731602668762207,
      "rewards-max/forensics_strict_edge_reward": 0.4789964258670807,
      "rewards-max/forensics_tiou_reward": 0.9573434591293335,
      "rewards/forensics_sinkhorn_reward": 0.8123128116130829,
      "rewards/forensics_strict_edge_reward": 0.08596867695450783,
      "rewards/forensics_tiou_reward": 0.632696270942688,
      "step": 1233
    },
    {
      "completion_length": 21.640625,
      "epoch": 3.7864823348694316,
      "grad_norm": 12.644694328308105,
      "kl": 0.30078125,
      "learning_rate": 5.153846153846154e-08,
      "loss": 0.1122,
      "reward": 1.4744101762771606,
      "reward_std": 0.24876131117343903,
      "reward_std_all": 0.21815980225801468,
      "rewards-max/forensics_sinkhorn_reward": 0.9678620398044586,
      "rewards-max/forensics_strict_edge_reward": 0.5062990635633469,
      "rewards-max/forensics_tiou_reward": 0.9467733800411224,
      "rewards/forensics_sinkhorn_reward": 0.7922838032245636,
      "rewards/forensics_strict_edge_reward": 0.06390683352947235,
      "rewards/forensics_tiou_reward": 0.6182195544242859,
      "step": 1234
    },
    {
      "completion_length": 22.0625,
      "epoch": 3.7895545314900154,
      "grad_norm": 10.554624557495117,
      "kl": 0.259765625,
      "learning_rate": 5.0769230769230764e-08,
      "loss": 0.1175,
      "reward": 1.416016936302185,
      "reward_std": 0.3123209923505783,
      "reward_std_all": 0.28224220871925354,
      "rewards-max/forensics_sinkhorn_reward": 0.9763124883174896,
      "rewards-max/forensics_strict_edge_reward": 0.43349720537662506,
      "rewards-max/forensics_tiou_reward": 0.9664057493209839,
      "rewards/forensics_sinkhorn_reward": 0.757125198841095,
      "rewards/forensics_strict_edge_reward": 0.04744712635874748,
      "rewards/forensics_tiou_reward": 0.6114446222782135,
      "step": 1235
    },
    {
      "completion_length": 22.0,
      "epoch": 3.792626728110599,
      "grad_norm": 14.213752746582031,
      "kl": 0.35546875,
      "learning_rate": 5e-08,
      "loss": 0.0358,
      "reward": 1.3783977627754211,
      "reward_std": 0.2728726491332054,
      "reward_std_all": 0.23350553214550018,
      "rewards-max/forensics_sinkhorn_reward": 0.95973339676857,
      "rewards-max/forensics_strict_edge_reward": 0.3298932909965515,
      "rewards-max/forensics_tiou_reward": 0.9253755211830139,
      "rewards/forensics_sinkhorn_reward": 0.772398054599762,
      "rewards/forensics_strict_edge_reward": 0.04851457756012678,
      "rewards/forensics_tiou_reward": 0.5574851334095001,
      "step": 1236
    },
    {
      "completion_length": 20.4375,
      "epoch": 3.795698924731183,
      "grad_norm": 13.26047420501709,
      "kl": 0.4384765625,
      "learning_rate": 4.923076923076923e-08,
      "loss": -0.0711,
      "reward": 1.5513738989830017,
      "reward_std": 0.23174114897847176,
      "reward_std_all": 0.22377651184797287,
      "rewards-max/forensics_sinkhorn_reward": 0.9816528558731079,
      "rewards-max/forensics_strict_edge_reward": 0.568440854549408,
      "rewards-max/forensics_tiou_reward": 0.9802664518356323,
      "rewards/forensics_sinkhorn_reward": 0.8112049698829651,
      "rewards/forensics_strict_edge_reward": 0.0809822604060173,
      "rewards/forensics_tiou_reward": 0.659186840057373,
      "step": 1237
    },
    {
      "completion_length": 24.765625,
      "epoch": 3.7987711213517663,
      "grad_norm": 13.388447761535645,
      "kl": 0.28857421875,
      "learning_rate": 4.846153846153846e-08,
      "loss": 0.1234,
      "reward": 1.5554443001747131,
      "reward_std": 0.24972278624773026,
      "reward_std_all": 0.22092247009277344,
      "rewards-max/forensics_sinkhorn_reward": 0.9804331362247467,
      "rewards-max/forensics_strict_edge_reward": 0.507948026061058,
      "rewards-max/forensics_tiou_reward": 0.9759716689586639,
      "rewards/forensics_sinkhorn_reward": 0.8086155354976654,
      "rewards/forensics_strict_edge_reward": 0.09296504408121109,
      "rewards/forensics_tiou_reward": 0.6538637280464172,
      "step": 1238
    },
    {
      "completion_length": 22.125,
      "epoch": 3.80184331797235,
      "grad_norm": 10.996404647827148,
      "kl": 0.2802734375,
      "learning_rate": 4.769230769230769e-08,
      "loss": 0.0669,
      "reward": 1.341672658920288,
      "reward_std": 0.17890659719705582,
      "reward_std_all": 0.16357451677322388,
      "rewards-max/forensics_sinkhorn_reward": 0.9564136564731598,
      "rewards-max/forensics_strict_edge_reward": 0.48458099365234375,
      "rewards-max/forensics_tiou_reward": 0.9258379936218262,
      "rewards/forensics_sinkhorn_reward": 0.7584189772605896,
      "rewards/forensics_strict_edge_reward": 0.04852818697690964,
      "rewards/forensics_tiou_reward": 0.5347254872322083,
      "step": 1239
    },
    {
      "completion_length": 19.84375,
      "epoch": 3.804915514592934,
      "grad_norm": 15.774970054626465,
      "kl": 0.30859375,
      "learning_rate": 4.692307692307692e-08,
      "loss": 0.0633,
      "reward": 1.4813156723976135,
      "reward_std": 0.2637479603290558,
      "reward_std_all": 0.22778358310461044,
      "rewards-max/forensics_sinkhorn_reward": 0.9709904193878174,
      "rewards-max/forensics_strict_edge_reward": 0.554103210568428,
      "rewards-max/forensics_tiou_reward": 0.9577713310718536,
      "rewards/forensics_sinkhorn_reward": 0.7988819479942322,
      "rewards/forensics_strict_edge_reward": 0.06635311618447304,
      "rewards/forensics_tiou_reward": 0.6160805821418762,
      "step": 1240
    },
    {
      "completion_length": 20.625,
      "epoch": 3.8079877112135176,
      "grad_norm": 21.86056137084961,
      "kl": 0.38134765625,
      "learning_rate": 4.615384615384615e-08,
      "loss": -0.0142,
      "reward": 1.4682486057281494,
      "reward_std": 0.23137734830379486,
      "reward_std_all": 0.19931073486804962,
      "rewards-max/forensics_sinkhorn_reward": 0.9719466269016266,
      "rewards-max/forensics_strict_edge_reward": 0.5139672756195068,
      "rewards-max/forensics_tiou_reward": 0.9558691382408142,
      "rewards/forensics_sinkhorn_reward": 0.8013017177581787,
      "rewards/forensics_strict_edge_reward": 0.06718263402581215,
      "rewards/forensics_tiou_reward": 0.5997642278671265,
      "step": 1241
    },
    {
      "completion_length": 20.96875,
      "epoch": 3.8110599078341014,
      "grad_norm": 16.15167808532715,
      "kl": 0.69140625,
      "learning_rate": 4.538461538461538e-08,
      "loss": 0.0445,
      "reward": 1.4328638315200806,
      "reward_std": 0.18775150179862976,
      "reward_std_all": 0.25443561375141144,
      "rewards-max/forensics_sinkhorn_reward": 0.9693759679794312,
      "rewards-max/forensics_strict_edge_reward": 0.6032288670539856,
      "rewards-max/forensics_tiou_reward": 0.9551281332969666,
      "rewards/forensics_sinkhorn_reward": 0.7653552889823914,
      "rewards/forensics_strict_edge_reward": 0.0871620699763298,
      "rewards/forensics_tiou_reward": 0.580346405506134,
      "step": 1242
    },
    {
      "completion_length": 23.984375,
      "epoch": 3.814132104454685,
      "grad_norm": 14.590988159179688,
      "kl": 0.3125,
      "learning_rate": 4.461538461538461e-08,
      "loss": 0.0775,
      "reward": 1.4506820440292358,
      "reward_std": 0.20403430610895157,
      "reward_std_all": 0.22745077311992645,
      "rewards-max/forensics_sinkhorn_reward": 0.9771041572093964,
      "rewards-max/forensics_strict_edge_reward": 0.7388437688350677,
      "rewards-max/forensics_tiou_reward": 0.9704351723194122,
      "rewards/forensics_sinkhorn_reward": 0.7523634135723114,
      "rewards/forensics_strict_edge_reward": 0.08032025396823883,
      "rewards/forensics_tiou_reward": 0.617998331785202,
      "step": 1243
    },
    {
      "completion_length": 22.453125,
      "epoch": 3.817204301075269,
      "grad_norm": 16.185144424438477,
      "kl": 0.3046875,
      "learning_rate": 4.3846153846153846e-08,
      "loss": -0.0201,
      "reward": 1.5358060598373413,
      "reward_std": 0.22705164551734924,
      "reward_std_all": 0.20022828876972198,
      "rewards-max/forensics_sinkhorn_reward": 0.9842562973499298,
      "rewards-max/forensics_strict_edge_reward": 0.48297472298145294,
      "rewards-max/forensics_tiou_reward": 0.9823758602142334,
      "rewards/forensics_sinkhorn_reward": 0.8117113411426544,
      "rewards/forensics_strict_edge_reward": 0.05856453627347946,
      "rewards/forensics_tiou_reward": 0.665530264377594,
      "step": 1244
    },
    {
      "completion_length": 25.5,
      "epoch": 3.8202764976958523,
      "grad_norm": 10.531407356262207,
      "kl": 0.3408203125,
      "learning_rate": 4.307692307692307e-08,
      "loss": -0.028,
      "reward": 1.3976856470108032,
      "reward_std": 0.16510628908872604,
      "reward_std_all": 0.22382274270057678,
      "rewards-max/forensics_sinkhorn_reward": 0.9721398055553436,
      "rewards-max/forensics_strict_edge_reward": 0.5473489910364151,
      "rewards-max/forensics_tiou_reward": 0.9500750005245209,
      "rewards/forensics_sinkhorn_reward": 0.7573430240154266,
      "rewards/forensics_strict_edge_reward": 0.06002316018566489,
      "rewards/forensics_tiou_reward": 0.5803193300962448,
      "step": 1245
    },
    {
      "completion_length": 21.921875,
      "epoch": 3.823348694316436,
      "grad_norm": 13.68431282043457,
      "kl": 0.24560546875,
      "learning_rate": 4.230769230769231e-08,
      "loss": 0.0092,
      "reward": 1.622489333152771,
      "reward_std": 0.2600446045398712,
      "reward_std_all": 0.25824519991874695,
      "rewards-max/forensics_sinkhorn_reward": 0.9781674444675446,
      "rewards-max/forensics_strict_edge_reward": 0.6799494326114655,
      "rewards-max/forensics_tiou_reward": 0.9714112877845764,
      "rewards/forensics_sinkhorn_reward": 0.8144244253635406,
      "rewards/forensics_strict_edge_reward": 0.1297810822725296,
      "rewards/forensics_tiou_reward": 0.6782838404178619,
      "step": 1246
    },
    {
      "completion_length": 22.25,
      "epoch": 3.82642089093702,
      "grad_norm": 25.11623191833496,
      "kl": 0.90625,
      "learning_rate": 4.1538461538461534e-08,
      "loss": 0.0013,
      "reward": 1.3736899495124817,
      "reward_std": 0.2069050893187523,
      "reward_std_all": 0.19530199468135834,
      "rewards-max/forensics_sinkhorn_reward": 0.9385403692722321,
      "rewards-max/forensics_strict_edge_reward": 0.34904851019382477,
      "rewards-max/forensics_tiou_reward": 0.8784964680671692,
      "rewards/forensics_sinkhorn_reward": 0.7474434673786163,
      "rewards/forensics_strict_edge_reward": 0.03858568146824837,
      "rewards/forensics_tiou_reward": 0.5876607596874237,
      "step": 1247
    },
    {
      "completion_length": 23.078125,
      "epoch": 3.8294930875576036,
      "grad_norm": 12.388428688049316,
      "kl": 0.5126953125,
      "learning_rate": 4.076923076923077e-08,
      "loss": 0.0391,
      "reward": 1.470203161239624,
      "reward_std": 0.22575999796390533,
      "reward_std_all": 0.2554832547903061,
      "rewards-max/forensics_sinkhorn_reward": 0.9680795967578888,
      "rewards-max/forensics_strict_edge_reward": 0.4392491579055786,
      "rewards-max/forensics_tiou_reward": 0.9336774945259094,
      "rewards/forensics_sinkhorn_reward": 0.7933366894721985,
      "rewards/forensics_strict_edge_reward": 0.07077745348215103,
      "rewards/forensics_tiou_reward": 0.6060889363288879,
      "step": 1248
    },
    {
      "completion_length": 20.0,
      "epoch": 3.8325652841781874,
      "grad_norm": 9.927780151367188,
      "kl": 0.4580078125,
      "learning_rate": 4e-08,
      "loss": -0.0557,
      "reward": 1.506454885005951,
      "reward_std": 0.23908057808876038,
      "reward_std_all": 0.2466941997408867,
      "rewards-max/forensics_sinkhorn_reward": 0.9800406396389008,
      "rewards-max/forensics_strict_edge_reward": 0.6965087056159973,
      "rewards-max/forensics_tiou_reward": 0.9769651889801025,
      "rewards/forensics_sinkhorn_reward": 0.7747161984443665,
      "rewards/forensics_strict_edge_reward": 0.12108508870005608,
      "rewards/forensics_tiou_reward": 0.6106535792350769,
      "step": 1249
    },
    {
      "completion_length": 21.328125,
      "epoch": 3.835637480798771,
      "grad_norm": 14.296932220458984,
      "kl": 0.298828125,
      "learning_rate": 3.923076923076923e-08,
      "loss": 0.0477,
      "reward": 1.5677350759506226,
      "reward_std": 0.24142905324697495,
      "reward_std_all": 0.24296852946281433,
      "rewards-max/forensics_sinkhorn_reward": 0.9838833808898926,
      "rewards-max/forensics_strict_edge_reward": 0.6408714950084686,
      "rewards-max/forensics_tiou_reward": 0.9860771000385284,
      "rewards/forensics_sinkhorn_reward": 0.8067797422409058,
      "rewards/forensics_strict_edge_reward": 0.08962669596076012,
      "rewards/forensics_tiou_reward": 0.6713286638259888,
      "step": 1250
    },
    {
      "completion_length": 20.796875,
      "epoch": 3.838709677419355,
      "grad_norm": 16.818710327148438,
      "kl": 0.46484375,
      "learning_rate": 3.846153846153846e-08,
      "loss": -0.0371,
      "reward": 1.5392972230911255,
      "reward_std": 0.20396798849105835,
      "reward_std_all": 0.19409432262182236,
      "rewards-max/forensics_sinkhorn_reward": 0.9804418683052063,
      "rewards-max/forensics_strict_edge_reward": 0.6720768958330154,
      "rewards-max/forensics_tiou_reward": 0.977034866809845,
      "rewards/forensics_sinkhorn_reward": 0.8039015233516693,
      "rewards/forensics_strict_edge_reward": 0.10266419500112534,
      "rewards/forensics_tiou_reward": 0.6327315866947174,
      "step": 1251
    },
    {
      "completion_length": 20.609375,
      "epoch": 3.8417818740399383,
      "grad_norm": 14.598250389099121,
      "kl": 0.580078125,
      "learning_rate": 3.769230769230769e-08,
      "loss": 0.0303,
      "reward": 1.5398970246315002,
      "reward_std": 0.15472520887851715,
      "reward_std_all": 0.18331336975097656,
      "rewards-max/forensics_sinkhorn_reward": 0.9781686961650848,
      "rewards-max/forensics_strict_edge_reward": 0.6516932249069214,
      "rewards-max/forensics_tiou_reward": 0.9735040068626404,
      "rewards/forensics_sinkhorn_reward": 0.8107281625270844,
      "rewards/forensics_strict_edge_reward": 0.07726612687110901,
      "rewards/forensics_tiou_reward": 0.6519027054309845,
      "step": 1252
    },
    {
      "completion_length": 23.015625,
      "epoch": 3.8448540706605225,
      "grad_norm": 15.6972074508667,
      "kl": 0.509765625,
      "learning_rate": 3.692307692307692e-08,
      "loss": 0.1195,
      "reward": 1.5612676739692688,
      "reward_std": 0.2190246805548668,
      "reward_std_all": 0.22415414452552795,
      "rewards-max/forensics_sinkhorn_reward": 0.9743139445781708,
      "rewards-max/forensics_strict_edge_reward": 0.480892151594162,
      "rewards-max/forensics_tiou_reward": 0.9556133151054382,
      "rewards/forensics_sinkhorn_reward": 0.8170600235462189,
      "rewards/forensics_strict_edge_reward": 0.07043610792607069,
      "rewards/forensics_tiou_reward": 0.6737715601921082,
      "step": 1253
    },
    {
      "completion_length": 19.078125,
      "epoch": 3.847926267281106,
      "grad_norm": 20.251523971557617,
      "kl": 0.27197265625,
      "learning_rate": 3.6153846153846155e-08,
      "loss": -0.0576,
      "reward": 1.4069954752922058,
      "reward_std": 0.2744007259607315,
      "reward_std_all": 0.23562051355838776,
      "rewards-max/forensics_sinkhorn_reward": 0.9799361228942871,
      "rewards-max/forensics_strict_edge_reward": 0.756708562374115,
      "rewards-max/forensics_tiou_reward": 0.9740203022956848,
      "rewards/forensics_sinkhorn_reward": 0.7694313228130341,
      "rewards/forensics_strict_edge_reward": 0.08340968936681747,
      "rewards/forensics_tiou_reward": 0.5541543662548065,
      "step": 1254
    },
    {
      "completion_length": 23.109375,
      "epoch": 3.8509984639016897,
      "grad_norm": 20.564809799194336,
      "kl": 0.4384765625,
      "learning_rate": 3.538461538461538e-08,
      "loss": 0.0231,
      "reward": 1.386605679988861,
      "reward_std": 0.20660168677568436,
      "reward_std_all": 0.21363622695207596,
      "rewards-max/forensics_sinkhorn_reward": 0.982330709695816,
      "rewards-max/forensics_strict_edge_reward": 0.7388437688350677,
      "rewards-max/forensics_tiou_reward": 0.9819130003452301,
      "rewards/forensics_sinkhorn_reward": 0.7286490201950073,
      "rewards/forensics_strict_edge_reward": 0.0778132975101471,
      "rewards/forensics_tiou_reward": 0.5801434814929962,
      "step": 1255
    },
    {
      "completion_length": 22.765625,
      "epoch": 3.8540706605222734,
      "grad_norm": 18.57024383544922,
      "kl": 1.09375,
      "learning_rate": 3.4615384615384616e-08,
      "loss": 0.1762,
      "reward": 1.3297542333602905,
      "reward_std": 0.24047646671533585,
      "reward_std_all": 0.24448296427726746,
      "rewards-max/forensics_sinkhorn_reward": 0.9699028134346008,
      "rewards-max/forensics_strict_edge_reward": 0.30461298674345016,
      "rewards-max/forensics_tiou_reward": 0.9529444575309753,
      "rewards/forensics_sinkhorn_reward": 0.7373635768890381,
      "rewards/forensics_strict_edge_reward": 0.038001193664968014,
      "rewards/forensics_tiou_reward": 0.5543894171714783,
      "step": 1256
    },
    {
      "completion_length": 22.15625,
      "epoch": 3.857142857142857,
      "grad_norm": 10.176319122314453,
      "kl": 0.685546875,
      "learning_rate": 3.384615384615384e-08,
      "loss": 0.1287,
      "reward": 1.4242749214172363,
      "reward_std": 0.2798374891281128,
      "reward_std_all": 0.2805977612733841,
      "rewards-max/forensics_sinkhorn_reward": 0.9827971458435059,
      "rewards-max/forensics_strict_edge_reward": 0.756708562374115,
      "rewards-max/forensics_tiou_reward": 0.984728217124939,
      "rewards/forensics_sinkhorn_reward": 0.7609052360057831,
      "rewards/forensics_strict_edge_reward": 0.097510676831007,
      "rewards/forensics_tiou_reward": 0.5658589005470276,
      "step": 1257
    },
    {
      "completion_length": 26.40625,
      "epoch": 3.860215053763441,
      "grad_norm": 8.623017311096191,
      "kl": 0.447265625,
      "learning_rate": 3.3076923076923076e-08,
      "loss": 0.104,
      "reward": 1.4502997994422913,
      "reward_std": 0.2472587153315544,
      "reward_std_all": 0.22598858922719955,
      "rewards-max/forensics_sinkhorn_reward": 0.9693280756473541,
      "rewards-max/forensics_strict_edge_reward": 0.38454997539520264,
      "rewards-max/forensics_tiou_reward": 0.9513300955295563,
      "rewards/forensics_sinkhorn_reward": 0.7817583680152893,
      "rewards/forensics_strict_edge_reward": 0.06447987072169781,
      "rewards/forensics_tiou_reward": 0.6040615737438202,
      "step": 1258
    },
    {
      "completion_length": 20.84375,
      "epoch": 3.8632872503840243,
      "grad_norm": 10.680822372436523,
      "kl": 0.22900390625,
      "learning_rate": 3.230769230769231e-08,
      "loss": -0.0539,
      "reward": 1.533252775669098,
      "reward_std": 0.1892911121249199,
      "reward_std_all": 0.18990376591682434,
      "rewards-max/forensics_sinkhorn_reward": 0.9736879467964172,
      "rewards-max/forensics_strict_edge_reward": 0.3185875788331032,
      "rewards-max/forensics_tiou_reward": 0.9558651149272919,
      "rewards/forensics_sinkhorn_reward": 0.8149867355823517,
      "rewards/forensics_strict_edge_reward": 0.06326074153184891,
      "rewards/forensics_tiou_reward": 0.6550052762031555,
      "step": 1259
    },
    {
      "completion_length": 19.609375,
      "epoch": 3.8663594470046085,
      "grad_norm": 15.157567024230957,
      "kl": 0.2724609375,
      "learning_rate": 3.153846153846154e-08,
      "loss": -0.0296,
      "reward": 1.3751916289329529,
      "reward_std": 0.2040107697248459,
      "reward_std_all": 0.1827736422419548,
      "rewards-max/forensics_sinkhorn_reward": 0.9734403789043427,
      "rewards-max/forensics_strict_edge_reward": 0.30294420570135117,
      "rewards-max/forensics_tiou_reward": 0.9548213183879852,
      "rewards/forensics_sinkhorn_reward": 0.7579797804355621,
      "rewards/forensics_strict_edge_reward": 0.0466513205319643,
      "rewards/forensics_tiou_reward": 0.5705605745315552,
      "step": 1260
    },
    {
      "completion_length": 23.234375,
      "epoch": 3.869431643625192,
      "grad_norm": 13.030008316040039,
      "kl": 0.48046875,
      "learning_rate": 3.076923076923077e-08,
      "loss": 0.0219,
      "reward": 1.3546053171157837,
      "reward_std": 0.31013719737529755,
      "reward_std_all": 0.27473045140504837,
      "rewards-max/forensics_sinkhorn_reward": 0.9781580567359924,
      "rewards-max/forensics_strict_edge_reward": 0.46446214616298676,
      "rewards-max/forensics_tiou_reward": 0.963972419500351,
      "rewards/forensics_sinkhorn_reward": 0.7108963429927826,
      "rewards/forensics_strict_edge_reward": 0.08753836154937744,
      "rewards/forensics_tiou_reward": 0.5561705529689789,
      "step": 1261
    },
    {
      "completion_length": 22.703125,
      "epoch": 3.8725038402457757,
      "grad_norm": 13.271458625793457,
      "kl": 0.2802734375,
      "learning_rate": 3e-08,
      "loss": -0.0298,
      "reward": 1.5782638788223267,
      "reward_std": 0.2207413613796234,
      "reward_std_all": 0.19810143113136292,
      "rewards-max/forensics_sinkhorn_reward": 0.9765545129776001,
      "rewards-max/forensics_strict_edge_reward": 0.6408846378326416,
      "rewards-max/forensics_tiou_reward": 0.968513548374176,
      "rewards/forensics_sinkhorn_reward": 0.8307933211326599,
      "rewards/forensics_strict_edge_reward": 0.07396256551146507,
      "rewards/forensics_tiou_reward": 0.6735080778598785,
      "step": 1262
    },
    {
      "completion_length": 20.9375,
      "epoch": 3.8755760368663594,
      "grad_norm": 11.368239402770996,
      "kl": 0.3896484375,
      "learning_rate": 2.9230769230769228e-08,
      "loss": 0.0954,
      "reward": 1.5736841559410095,
      "reward_std": 0.2117890566587448,
      "reward_std_all": 0.247587651014328,
      "rewards-max/forensics_sinkhorn_reward": 0.9828279316425323,
      "rewards-max/forensics_strict_edge_reward": 0.7709200084209442,
      "rewards-max/forensics_tiou_reward": 0.9802365005016327,
      "rewards/forensics_sinkhorn_reward": 0.8186528086662292,
      "rewards/forensics_strict_edge_reward": 0.09612752497196198,
      "rewards/forensics_tiou_reward": 0.6589037775993347,
      "step": 1263
    },
    {
      "completion_length": 23.9375,
      "epoch": 3.878648233486943,
      "grad_norm": 11.633086204528809,
      "kl": 0.3681640625,
      "learning_rate": 2.846153846153846e-08,
      "loss": 0.0539,
      "reward": 1.601948857307434,
      "reward_std": 0.2884613275527954,
      "reward_std_all": 0.25349923968315125,
      "rewards-max/forensics_sinkhorn_reward": 0.9659242331981659,
      "rewards-max/forensics_strict_edge_reward": 0.49500927329063416,
      "rewards-max/forensics_tiou_reward": 0.944848358631134,
      "rewards/forensics_sinkhorn_reward": 0.8263892531394958,
      "rewards/forensics_strict_edge_reward": 0.08427267894148827,
      "rewards/forensics_tiou_reward": 0.6912869513034821,
      "step": 1264
    },
    {
      "completion_length": 21.875,
      "epoch": 3.881720430107527,
      "grad_norm": 12.324650764465332,
      "kl": 0.88671875,
      "learning_rate": 2.769230769230769e-08,
      "loss": 0.1022,
      "reward": 1.5078725218772888,
      "reward_std": 0.22755693644285202,
      "reward_std_all": 0.2004573792219162,
      "rewards-max/forensics_sinkhorn_reward": 0.9794729053974152,
      "rewards-max/forensics_strict_edge_reward": 0.47998008131980896,
      "rewards-max/forensics_tiou_reward": 0.9766450226306915,
      "rewards/forensics_sinkhorn_reward": 0.8084350228309631,
      "rewards/forensics_strict_edge_reward": 0.06087705120444298,
      "rewards/forensics_tiou_reward": 0.63856041431427,
      "step": 1265
    },
    {
      "completion_length": 20.375,
      "epoch": 3.8847926267281108,
      "grad_norm": 21.625492095947266,
      "kl": 0.80859375,
      "learning_rate": 2.6923076923076925e-08,
      "loss": 0.117,
      "reward": 1.3617688417434692,
      "reward_std": 0.20258213579654694,
      "reward_std_all": 0.18631302565336227,
      "rewards-max/forensics_sinkhorn_reward": 0.9648595154285431,
      "rewards-max/forensics_strict_edge_reward": 0.3105753306299448,
      "rewards-max/forensics_tiou_reward": 0.9253288805484772,
      "rewards/forensics_sinkhorn_reward": 0.7549325227737427,
      "rewards/forensics_strict_edge_reward": 0.045383086428046227,
      "rewards/forensics_tiou_reward": 0.5614531934261322,
      "step": 1266
    },
    {
      "completion_length": 20.5,
      "epoch": 3.8878648233486945,
      "grad_norm": 16.778053283691406,
      "kl": 0.568359375,
      "learning_rate": 2.6153846153846152e-08,
      "loss": 0.0757,
      "reward": 1.4356753826141357,
      "reward_std": 0.30212724208831787,
      "reward_std_all": 0.2878076434135437,
      "rewards-max/forensics_sinkhorn_reward": 0.9691683053970337,
      "rewards-max/forensics_strict_edge_reward": 0.4714259207248688,
      "rewards-max/forensics_tiou_reward": 0.9488361775875092,
      "rewards/forensics_sinkhorn_reward": 0.7776508033275604,
      "rewards/forensics_strict_edge_reward": 0.0651310496032238,
      "rewards/forensics_tiou_reward": 0.5928934514522552,
      "step": 1267
    },
    {
      "completion_length": 20.8125,
      "epoch": 3.890937019969278,
      "grad_norm": 10.932205200195312,
      "kl": 0.4345703125,
      "learning_rate": 2.5384615384615382e-08,
      "loss": 0.1497,
      "reward": 1.4494584798812866,
      "reward_std": 0.28174157440662384,
      "reward_std_all": 0.30498212575912476,
      "rewards-max/forensics_sinkhorn_reward": 0.9734452366828918,
      "rewards-max/forensics_strict_edge_reward": 0.5238488018512726,
      "rewards-max/forensics_tiou_reward": 0.9584243595600128,
      "rewards/forensics_sinkhorn_reward": 0.7686093747615814,
      "rewards/forensics_strict_edge_reward": 0.08952074870467186,
      "rewards/forensics_tiou_reward": 0.5913283228874207,
      "step": 1268
    },
    {
      "completion_length": 23.875,
      "epoch": 3.8940092165898617,
      "grad_norm": 13.670315742492676,
      "kl": 0.37890625,
      "learning_rate": 2.4615384615384616e-08,
      "loss": 0.1647,
      "reward": 1.4030143022537231,
      "reward_std": 0.2904176265001297,
      "reward_std_all": 0.2749422639608383,
      "rewards-max/forensics_sinkhorn_reward": 0.9761814773082733,
      "rewards-max/forensics_strict_edge_reward": 0.49768924713134766,
      "rewards-max/forensics_tiou_reward": 0.9678610563278198,
      "rewards/forensics_sinkhorn_reward": 0.7599989473819733,
      "rewards/forensics_strict_edge_reward": 0.07353643141686916,
      "rewards/forensics_tiou_reward": 0.5694788992404938,
      "step": 1269
    },
    {
      "completion_length": 22.109375,
      "epoch": 3.8970814132104454,
      "grad_norm": 18.965587615966797,
      "kl": 0.3447265625,
      "learning_rate": 2.3846153846153846e-08,
      "loss": -0.0692,
      "reward": 1.5958915948867798,
      "reward_std": 0.37650204449892044,
      "reward_std_all": 0.32805366814136505,
      "rewards-max/forensics_sinkhorn_reward": 0.9792797863483429,
      "rewards-max/forensics_strict_edge_reward": 0.6231168210506439,
      "rewards-max/forensics_tiou_reward": 0.9736958146095276,
      "rewards/forensics_sinkhorn_reward": 0.8159209489822388,
      "rewards/forensics_strict_edge_reward": 0.12543000653386116,
      "rewards/forensics_tiou_reward": 0.6545405685901642,
      "step": 1270
    },
    {
      "completion_length": 24.203125,
      "epoch": 3.900153609831029,
      "grad_norm": 11.028079986572266,
      "kl": 0.35546875,
      "learning_rate": 2.3076923076923076e-08,
      "loss": -0.0229,
      "reward": 1.4147267937660217,
      "reward_std": 0.24852438271045685,
      "reward_std_all": 0.2269551232457161,
      "rewards-max/forensics_sinkhorn_reward": 0.9739267230033875,
      "rewards-max/forensics_strict_edge_reward": 0.5408188104629517,
      "rewards-max/forensics_tiou_reward": 0.9503246545791626,
      "rewards/forensics_sinkhorn_reward": 0.7616488337516785,
      "rewards/forensics_strict_edge_reward": 0.07892881846055388,
      "rewards/forensics_tiou_reward": 0.5741492509841919,
      "step": 1271
    },
    {
      "completion_length": 23.6875,
      "epoch": 3.903225806451613,
      "grad_norm": 26.75109100341797,
      "kl": 1.01171875,
      "learning_rate": 2.2307692307692306e-08,
      "loss": 0.1876,
      "reward": 1.3590798377990723,
      "reward_std": 0.29390138387680054,
      "reward_std_all": 0.2762724310159683,
      "rewards-max/forensics_sinkhorn_reward": 0.9573158919811249,
      "rewards-max/forensics_strict_edge_reward": 0.4029480069875717,
      "rewards-max/forensics_tiou_reward": 0.9215469062328339,
      "rewards/forensics_sinkhorn_reward": 0.7349328696727753,
      "rewards/forensics_strict_edge_reward": 0.05577663704752922,
      "rewards/forensics_tiou_reward": 0.5683703422546387,
      "step": 1272
    },
    {
      "completion_length": 21.390625,
      "epoch": 3.9062980030721968,
      "grad_norm": 14.050885200500488,
      "kl": 0.4453125,
      "learning_rate": 2.1538461538461537e-08,
      "loss": 0.09,
      "reward": 1.559920072555542,
      "reward_std": 0.2369047850370407,
      "reward_std_all": 0.2132304459810257,
      "rewards-max/forensics_sinkhorn_reward": 0.9691860675811768,
      "rewards-max/forensics_strict_edge_reward": 0.5957051217556,
      "rewards-max/forensics_tiou_reward": 0.9499012231826782,
      "rewards/forensics_sinkhorn_reward": 0.8129291832447052,
      "rewards/forensics_strict_edge_reward": 0.0928255170583725,
      "rewards/forensics_tiou_reward": 0.6541652977466583,
      "step": 1273
    },
    {
      "completion_length": 22.609375,
      "epoch": 3.9093701996927805,
      "grad_norm": 9.85362720489502,
      "kl": 0.5361328125,
      "learning_rate": 2.0769230769230767e-08,
      "loss": -0.0571,
      "reward": 1.4071131944656372,
      "reward_std": 0.2308553010225296,
      "reward_std_all": 0.22006218880414963,
      "rewards-max/forensics_sinkhorn_reward": 0.984169214963913,
      "rewards-max/forensics_strict_edge_reward": 0.45225730538368225,
      "rewards-max/forensics_tiou_reward": 0.9885453581809998,
      "rewards/forensics_sinkhorn_reward": 0.7509986460208893,
      "rewards/forensics_strict_edge_reward": 0.07112741097807884,
      "rewards/forensics_tiou_reward": 0.5849871337413788,
      "step": 1274
    },
    {
      "completion_length": 22.109375,
      "epoch": 3.912442396313364,
      "grad_norm": 11.988099098205566,
      "kl": 0.3984375,
      "learning_rate": 2e-08,
      "loss": 0.0327,
      "reward": 1.6036184430122375,
      "reward_std": 0.31323903799057007,
      "reward_std_all": 0.28264328837394714,
      "rewards-max/forensics_sinkhorn_reward": 0.9831361472606659,
      "rewards-max/forensics_strict_edge_reward": 0.6930722296237946,
      "rewards-max/forensics_tiou_reward": 0.9853795170783997,
      "rewards/forensics_sinkhorn_reward": 0.8079582452774048,
      "rewards/forensics_strict_edge_reward": 0.10818715021014214,
      "rewards/forensics_tiou_reward": 0.6874729990959167,
      "step": 1275
    },
    {
      "completion_length": 22.765625,
      "epoch": 3.9155145929339477,
      "grad_norm": 11.600899696350098,
      "kl": 0.2998046875,
      "learning_rate": 1.923076923076923e-08,
      "loss": -0.0657,
      "reward": 1.423899531364441,
      "reward_std": 0.28710395097732544,
      "reward_std_all": 0.2599906027317047,
      "rewards-max/forensics_sinkhorn_reward": 0.969594419002533,
      "rewards-max/forensics_strict_edge_reward": 0.7181241810321808,
      "rewards-max/forensics_tiou_reward": 0.9530568718910217,
      "rewards/forensics_sinkhorn_reward": 0.7764461040496826,
      "rewards/forensics_strict_edge_reward": 0.07251620851457119,
      "rewards/forensics_tiou_reward": 0.5749371647834778,
      "step": 1276
    },
    {
      "completion_length": 19.796875,
      "epoch": 3.9185867895545314,
      "grad_norm": 13.072968482971191,
      "kl": 0.3671875,
      "learning_rate": 1.846153846153846e-08,
      "loss": -0.0163,
      "reward": 1.583605706691742,
      "reward_std": 0.2703028842806816,
      "reward_std_all": 0.24049126356840134,
      "rewards-max/forensics_sinkhorn_reward": 0.9811829030513763,
      "rewards-max/forensics_strict_edge_reward": 0.8507176637649536,
      "rewards-max/forensics_tiou_reward": 0.9821801781654358,
      "rewards/forensics_sinkhorn_reward": 0.8251717686653137,
      "rewards/forensics_strict_edge_reward": 0.09333059191703796,
      "rewards/forensics_tiou_reward": 0.6651033163070679,
      "step": 1277
    },
    {
      "completion_length": 23.8125,
      "epoch": 3.921658986175115,
      "grad_norm": 14.055949211120605,
      "kl": 0.4501953125,
      "learning_rate": 1.769230769230769e-08,
      "loss": -0.0625,
      "reward": 1.3215749263763428,
      "reward_std": 0.27984718978405,
      "reward_std_all": 0.24165478348731995,
      "rewards-max/forensics_sinkhorn_reward": 0.9642853438854218,
      "rewards-max/forensics_strict_edge_reward": 0.3243762031197548,
      "rewards-max/forensics_tiou_reward": 0.938495397567749,
      "rewards/forensics_sinkhorn_reward": 0.7143318355083466,
      "rewards/forensics_strict_edge_reward": 0.036426068283617496,
      "rewards/forensics_tiou_reward": 0.5708169937133789,
      "step": 1278
    },
    {
      "completion_length": 20.140625,
      "epoch": 3.924731182795699,
      "grad_norm": 14.134719848632812,
      "kl": 0.29248046875,
      "learning_rate": 1.692307692307692e-08,
      "loss": 0.1211,
      "reward": 1.4108757972717285,
      "reward_std": 0.2017485797405243,
      "reward_std_all": 0.17982128262519836,
      "rewards-max/forensics_sinkhorn_reward": 0.9723352491855621,
      "rewards-max/forensics_strict_edge_reward": 0.5240039527416229,
      "rewards-max/forensics_tiou_reward": 0.9589346051216125,
      "rewards/forensics_sinkhorn_reward": 0.7761141657829285,
      "rewards/forensics_strict_edge_reward": 0.06800772249698639,
      "rewards/forensics_tiou_reward": 0.5667539834976196,
      "step": 1279
    },
    {
      "completion_length": 23.171875,
      "epoch": 3.9278033794162828,
      "grad_norm": 10.034931182861328,
      "kl": 0.251953125,
      "learning_rate": 1.6153846153846155e-08,
      "loss": 0.0595,
      "reward": 1.5658334493637085,
      "reward_std": 0.22678110748529434,
      "reward_std_all": 0.25641168653964996,
      "rewards-max/forensics_sinkhorn_reward": 0.9714924693107605,
      "rewards-max/forensics_strict_edge_reward": 0.21038316190242767,
      "rewards-max/forensics_tiou_reward": 0.9557759761810303,
      "rewards/forensics_sinkhorn_reward": 0.8226637244224548,
      "rewards/forensics_strict_edge_reward": 0.05043875798583031,
      "rewards/forensics_tiou_reward": 0.6927309930324554,
      "step": 1280
    },
    {
      "completion_length": 25.609375,
      "epoch": 3.9308755760368665,
      "grad_norm": 14.301567077636719,
      "kl": 0.3857421875,
      "learning_rate": 1.5384615384615385e-08,
      "loss": 0.005,
      "reward": 1.489460527896881,
      "reward_std": 0.2944066971540451,
      "reward_std_all": 0.2877180129289627,
      "rewards-max/forensics_sinkhorn_reward": 0.9594024121761322,
      "rewards-max/forensics_strict_edge_reward": 0.49021559953689575,
      "rewards-max/forensics_tiou_reward": 0.9346702396869659,
      "rewards/forensics_sinkhorn_reward": 0.7526581287384033,
      "rewards/forensics_strict_edge_reward": 0.08943658322095871,
      "rewards/forensics_tiou_reward": 0.6473659574985504,
      "step": 1281
    },
    {
      "completion_length": 24.75,
      "epoch": 3.93394777265745,
      "grad_norm": 12.697685241699219,
      "kl": 0.482421875,
      "learning_rate": 1.4615384615384614e-08,
      "loss": 0.0052,
      "reward": 1.4747043251991272,
      "reward_std": 0.19827628880739212,
      "reward_std_all": 0.1846652552485466,
      "rewards-max/forensics_sinkhorn_reward": 0.983586996793747,
      "rewards-max/forensics_strict_edge_reward": 0.522201269865036,
      "rewards-max/forensics_tiou_reward": 0.9875593781471252,
      "rewards/forensics_sinkhorn_reward": 0.7672742307186127,
      "rewards/forensics_strict_edge_reward": 0.07978934422135353,
      "rewards/forensics_tiou_reward": 0.6276407837867737,
      "step": 1282
    },
    {
      "completion_length": 22.25,
      "epoch": 3.9370199692780337,
      "grad_norm": 19.137575149536133,
      "kl": 0.291015625,
      "learning_rate": 1.3846153846153846e-08,
      "loss": 0.0529,
      "reward": 1.5553804636001587,
      "reward_std": 0.3397684693336487,
      "reward_std_all": 0.3073476701974869,
      "rewards-max/forensics_sinkhorn_reward": 0.9760790169239044,
      "rewards-max/forensics_strict_edge_reward": 0.6149742305278778,
      "rewards-max/forensics_tiou_reward": 0.9660452902317047,
      "rewards/forensics_sinkhorn_reward": 0.8052221238613129,
      "rewards/forensics_strict_edge_reward": 0.09243160113692284,
      "rewards/forensics_tiou_reward": 0.6577268838882446,
      "step": 1283
    },
    {
      "completion_length": 21.25,
      "epoch": 3.9400921658986174,
      "grad_norm": 16.876136779785156,
      "kl": 0.30517578125,
      "learning_rate": 1.3076923076923076e-08,
      "loss": 0.0397,
      "reward": 1.3117429614067078,
      "reward_std": 0.18475782871246338,
      "reward_std_all": 0.17477213591337204,
      "rewards-max/forensics_sinkhorn_reward": 0.9620340168476105,
      "rewards-max/forensics_strict_edge_reward": 0.5470698475837708,
      "rewards-max/forensics_tiou_reward": 0.9361189305782318,
      "rewards/forensics_sinkhorn_reward": 0.7229754328727722,
      "rewards/forensics_strict_edge_reward": 0.07201121374964714,
      "rewards/forensics_tiou_reward": 0.5167562812566757,
      "step": 1284
    },
    {
      "completion_length": 20.546875,
      "epoch": 3.943164362519201,
      "grad_norm": 14.854860305786133,
      "kl": 0.3779296875,
      "learning_rate": 1.2307692307692308e-08,
      "loss": 0.0269,
      "reward": 1.410122811794281,
      "reward_std": 0.20593001693487167,
      "reward_std_all": 0.19301926344633102,
      "rewards-max/forensics_sinkhorn_reward": 0.9669139385223389,
      "rewards-max/forensics_strict_edge_reward": 0.5934597104787827,
      "rewards-max/forensics_tiou_reward": 0.9436655044555664,
      "rewards/forensics_sinkhorn_reward": 0.7735224366188049,
      "rewards/forensics_strict_edge_reward": 0.06724836491048336,
      "rewards/forensics_tiou_reward": 0.569352000951767,
      "step": 1285
    },
    {
      "completion_length": 21.484375,
      "epoch": 3.946236559139785,
      "grad_norm": 10.536168098449707,
      "kl": 0.296875,
      "learning_rate": 1.1538461538461538e-08,
      "loss": 0.0712,
      "reward": 1.5346530079841614,
      "reward_std": 0.19626886397600174,
      "reward_std_all": 0.1700897067785263,
      "rewards-max/forensics_sinkhorn_reward": 0.9841594099998474,
      "rewards-max/forensics_strict_edge_reward": 0.6377167105674744,
      "rewards-max/forensics_tiou_reward": 0.9871557652950287,
      "rewards/forensics_sinkhorn_reward": 0.7970166802406311,
      "rewards/forensics_strict_edge_reward": 0.0868614912033081,
      "rewards/forensics_tiou_reward": 0.6507748663425446,
      "step": 1286
    },
    {
      "completion_length": 21.9375,
      "epoch": 3.9493087557603688,
      "grad_norm": 14.286109924316406,
      "kl": 0.3212890625,
      "learning_rate": 1.0769230769230768e-08,
      "loss": -0.018,
      "reward": 1.7305435538291931,
      "reward_std": 0.38549724221229553,
      "reward_std_all": 0.3447847366333008,
      "rewards-max/forensics_sinkhorn_reward": 0.982771635055542,
      "rewards-max/forensics_strict_edge_reward": 0.6942952275276184,
      "rewards-max/forensics_tiou_reward": 0.9814805388450623,
      "rewards/forensics_sinkhorn_reward": 0.8616800904273987,
      "rewards/forensics_strict_edge_reward": 0.12825299054384232,
      "rewards/forensics_tiou_reward": 0.7406105697154999,
      "step": 1287
    },
    {
      "completion_length": 22.8125,
      "epoch": 3.9523809523809526,
      "grad_norm": 11.553106307983398,
      "kl": 0.591796875,
      "learning_rate": 1e-08,
      "loss": -0.0043,
      "reward": 1.4143970012664795,
      "reward_std": 0.19870023429393768,
      "reward_std_all": 0.18077731877565384,
      "rewards-max/forensics_sinkhorn_reward": 0.9755703508853912,
      "rewards-max/forensics_strict_edge_reward": 0.48620253801345825,
      "rewards-max/forensics_tiou_reward": 0.9644341766834259,
      "rewards/forensics_sinkhorn_reward": 0.7680864334106445,
      "rewards/forensics_strict_edge_reward": 0.07128168642520905,
      "rewards/forensics_tiou_reward": 0.5750288665294647,
      "step": 1288
    },
    {
      "completion_length": 20.703125,
      "epoch": 3.955453149001536,
      "grad_norm": 19.036714553833008,
      "kl": 0.2373046875,
      "learning_rate": 9.23076923076923e-09,
      "loss": 0.0592,
      "reward": 1.4840625524520874,
      "reward_std": 0.2981954962015152,
      "reward_std_all": 0.27298980951309204,
      "rewards-max/forensics_sinkhorn_reward": 0.9786189198493958,
      "rewards-max/forensics_strict_edge_reward": 0.6934256851673126,
      "rewards-max/forensics_tiou_reward": 0.9750387668609619,
      "rewards/forensics_sinkhorn_reward": 0.7953131198883057,
      "rewards/forensics_strict_edge_reward": 0.09611684456467628,
      "rewards/forensics_tiou_reward": 0.5926325619220734,
      "step": 1289
    },
    {
      "completion_length": 21.6875,
      "epoch": 3.9585253456221197,
      "grad_norm": 12.633459091186523,
      "kl": 0.4013671875,
      "learning_rate": 8.46153846153846e-09,
      "loss": 0.182,
      "reward": 1.3866632580757141,
      "reward_std": 0.30140525102615356,
      "reward_std_all": 0.2629035711288452,
      "rewards-max/forensics_sinkhorn_reward": 0.9724671542644501,
      "rewards-max/forensics_strict_edge_reward": 0.5692775547504425,
      "rewards-max/forensics_tiou_reward": 0.9578287899494171,
      "rewards/forensics_sinkhorn_reward": 0.7463182210922241,
      "rewards/forensics_strict_edge_reward": 0.08446403592824936,
      "rewards/forensics_tiou_reward": 0.5558810830116272,
      "step": 1290
    },
    {
      "completion_length": 21.015625,
      "epoch": 3.9615975422427034,
      "grad_norm": 13.33427619934082,
      "kl": 0.4306640625,
      "learning_rate": 7.692307692307693e-09,
      "loss": 0.0016,
      "reward": 1.5776130557060242,
      "reward_std": 0.3071470409631729,
      "reward_std_all": 0.28236013650894165,
      "rewards-max/forensics_sinkhorn_reward": 0.9678298830986023,
      "rewards-max/forensics_strict_edge_reward": 0.31924980878829956,
      "rewards-max/forensics_tiou_reward": 0.9427829384803772,
      "rewards/forensics_sinkhorn_reward": 0.8232938647270203,
      "rewards/forensics_strict_edge_reward": 0.0645814910531044,
      "rewards/forensics_tiou_reward": 0.6897378265857697,
      "step": 1291
    },
    {
      "completion_length": 21.5625,
      "epoch": 3.964669738863287,
      "grad_norm": 11.490555763244629,
      "kl": 0.484375,
      "learning_rate": 6.923076923076923e-09,
      "loss": 0.0194,
      "reward": 1.3700802326202393,
      "reward_std": 0.22059743851423264,
      "reward_std_all": 0.19268299639225006,
      "rewards-max/forensics_sinkhorn_reward": 0.9755571782588959,
      "rewards-max/forensics_strict_edge_reward": 0.3623322397470474,
      "rewards-max/forensics_tiou_reward": 0.960362583398819,
      "rewards/forensics_sinkhorn_reward": 0.7554986774921417,
      "rewards/forensics_strict_edge_reward": 0.05699879489839077,
      "rewards/forensics_tiou_reward": 0.5575827062129974,
      "step": 1292
    },
    {
      "completion_length": 23.078125,
      "epoch": 3.967741935483871,
      "grad_norm": 11.867653846740723,
      "kl": 0.2939453125,
      "learning_rate": 6.153846153846154e-09,
      "loss": -0.0242,
      "reward": 1.213417410850525,
      "reward_std": 0.2376270294189453,
      "reward_std_all": 0.2365291714668274,
      "rewards-max/forensics_sinkhorn_reward": 0.9783742725849152,
      "rewards-max/forensics_strict_edge_reward": 0.40857692062854767,
      "rewards-max/forensics_tiou_reward": 0.9720948934555054,
      "rewards/forensics_sinkhorn_reward": 0.6774598658084869,
      "rewards/forensics_strict_edge_reward": 0.05926985293626785,
      "rewards/forensics_tiou_reward": 0.4766877293586731,
      "step": 1293
    },
    {
      "completion_length": 21.171875,
      "epoch": 3.9708141321044548,
      "grad_norm": 12.813315391540527,
      "kl": 0.4072265625,
      "learning_rate": 5.384615384615384e-09,
      "loss": 0.1539,
      "reward": 1.3430229425430298,
      "reward_std": 0.19023188948631287,
      "reward_std_all": 0.16874831914901733,
      "rewards-max/forensics_sinkhorn_reward": 0.9796105325222015,
      "rewards-max/forensics_strict_edge_reward": 0.37697482109069824,
      "rewards-max/forensics_tiou_reward": 0.9740490615367889,
      "rewards/forensics_sinkhorn_reward": 0.7552652955055237,
      "rewards/forensics_strict_edge_reward": 0.04354419931769371,
      "rewards/forensics_tiou_reward": 0.5442134737968445,
      "step": 1294
    },
    {
      "completion_length": 23.0625,
      "epoch": 3.9738863287250386,
      "grad_norm": 11.046984672546387,
      "kl": 0.3251953125,
      "learning_rate": 4.615384615384615e-09,
      "loss": -0.1389,
      "reward": 1.4893075227737427,
      "reward_std": 0.2356838956475258,
      "reward_std_all": 0.2063385397195816,
      "rewards-max/forensics_sinkhorn_reward": 0.9783909320831299,
      "rewards-max/forensics_strict_edge_reward": 0.5080419927835464,
      "rewards-max/forensics_tiou_reward": 0.9757157862186432,
      "rewards/forensics_sinkhorn_reward": 0.7953185439109802,
      "rewards/forensics_strict_edge_reward": 0.08175621554255486,
      "rewards/forensics_tiou_reward": 0.6122328341007233,
      "step": 1295
    },
    {
      "completion_length": 22.421875,
      "epoch": 3.976958525345622,
      "grad_norm": 16.063583374023438,
      "kl": 0.306640625,
      "learning_rate": 3.846153846153846e-09,
      "loss": 0.0388,
      "reward": 1.576362133026123,
      "reward_std": 0.21536172181367874,
      "reward_std_all": 0.21245530992746353,
      "rewards-max/forensics_sinkhorn_reward": 0.9776090681552887,
      "rewards-max/forensics_strict_edge_reward": 0.5101649016141891,
      "rewards-max/forensics_tiou_reward": 0.9658872783184052,
      "rewards/forensics_sinkhorn_reward": 0.8109565079212189,
      "rewards/forensics_strict_edge_reward": 0.08878512494266033,
      "rewards/forensics_tiou_reward": 0.6766204535961151,
      "step": 1296
    },
    {
      "completion_length": 21.796875,
      "epoch": 3.980030721966206,
      "grad_norm": 14.770612716674805,
      "kl": 0.3955078125,
      "learning_rate": 3.076923076923077e-09,
      "loss": -0.116,
      "reward": 1.3934175372123718,
      "reward_std": 0.20842088758945465,
      "reward_std_all": 0.19703101366758347,
      "rewards-max/forensics_sinkhorn_reward": 0.9840694665908813,
      "rewards-max/forensics_strict_edge_reward": 0.6942952275276184,
      "rewards-max/forensics_tiou_reward": 0.9851465225219727,
      "rewards/forensics_sinkhorn_reward": 0.7419648170471191,
      "rewards/forensics_strict_edge_reward": 0.0875637773424387,
      "rewards/forensics_tiou_reward": 0.5638889372348785,
      "step": 1297
    },
    {
      "completion_length": 20.40625,
      "epoch": 3.9831029185867894,
      "grad_norm": 16.8264102935791,
      "kl": 0.37890625,
      "learning_rate": 2.3076923076923076e-09,
      "loss": -0.0603,
      "reward": 1.5728030800819397,
      "reward_std": 0.2108752503991127,
      "reward_std_all": 0.2268161177635193,
      "rewards-max/forensics_sinkhorn_reward": 0.9723771512508392,
      "rewards-max/forensics_strict_edge_reward": 0.5853640735149384,
      "rewards-max/forensics_tiou_reward": 0.9577998518943787,
      "rewards/forensics_sinkhorn_reward": 0.8135458827018738,
      "rewards/forensics_strict_edge_reward": 0.1154710054397583,
      "rewards/forensics_tiou_reward": 0.6437862515449524,
      "step": 1298
    },
    {
      "completion_length": 19.984375,
      "epoch": 3.986175115207373,
      "grad_norm": 14.466212272644043,
      "kl": 0.318359375,
      "learning_rate": 1.5384615384615385e-09,
      "loss": -0.0134,
      "reward": 1.446820318698883,
      "reward_std": 0.2841941639780998,
      "reward_std_all": 0.2597198411822319,
      "rewards-max/forensics_sinkhorn_reward": 0.9706117510795593,
      "rewards-max/forensics_strict_edge_reward": 0.4962427020072937,
      "rewards-max/forensics_tiou_reward": 0.9476171433925629,
      "rewards/forensics_sinkhorn_reward": 0.7663558721542358,
      "rewards/forensics_strict_edge_reward": 0.08243591710925102,
      "rewards/forensics_tiou_reward": 0.5980285704135895,
      "step": 1299
    },
    {
      "completion_length": 20.796875,
      "epoch": 3.989247311827957,
      "grad_norm": 16.107498168945312,
      "kl": 0.3251953125,
      "learning_rate": 7.692307692307692e-10,
      "loss": 0.0175,
      "reward": 1.4553614854812622,
      "reward_std": 0.21480435132980347,
      "reward_std_all": 0.2250678837299347,
      "rewards-max/forensics_sinkhorn_reward": 0.976652979850769,
      "rewards-max/forensics_strict_edge_reward": 0.5734860002994537,
      "rewards-max/forensics_tiou_reward": 0.9697609543800354,
      "rewards/forensics_sinkhorn_reward": 0.7856123745441437,
      "rewards/forensics_strict_edge_reward": 0.07500530034303665,
      "rewards/forensics_tiou_reward": 0.5947438180446625,
      "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
}