{
  "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": 26.5625,
      "epoch": 0.0030721966205837174,
      "grad_norm": 11.1958646774292,
      "kl": 0.0,
      "learning_rate": 1e-06,
      "loss": -0.1132,
      "reward": 0.3940541446208954,
      "reward_std": 0.29250316321849823,
      "reward_std_all": 1.3422992825508118,
      "rewards-max/forensics_sinkhorn_reward": 0.82574662566185,
      "rewards-max/forensics_strict_edge_reward": 0.01118538625742991,
      "rewards-max/forensics_tiou_reward": 0.6565207988023758,
      "rewards/forensics_sinkhorn_reward": 0.3149970918893814,
      "rewards/forensics_strict_edge_reward": 0.0008476336387306647,
      "rewards/forensics_tiou_reward": 0.07820940017700195,
      "step": 1
    },
    {
      "completion_length": 25.15625,
      "epoch": 0.006144393241167435,
      "grad_norm": 7.8773016929626465,
      "kl": 0.0005636215209960938,
      "learning_rate": 9.992307692307693e-07,
      "loss": -0.1706,
      "reward": 0.4983289837837219,
      "reward_std": 0.26724541932344437,
      "reward_std_all": 1.2877849340438843,
      "rewards-max/forensics_sinkhorn_reward": 0.9156321585178375,
      "rewards-max/forensics_strict_edge_reward": 0.03598110191524029,
      "rewards-max/forensics_tiou_reward": 0.8120884895324707,
      "rewards/forensics_sinkhorn_reward": 0.3680003732442856,
      "rewards/forensics_strict_edge_reward": 0.001504895422840491,
      "rewards/forensics_tiou_reward": 0.12882371991872787,
      "step": 2
    },
    {
      "completion_length": 24.984375,
      "epoch": 0.009216589861751152,
      "grad_norm": 8.275260925292969,
      "kl": 0.00333404541015625,
      "learning_rate": 9.984615384615384e-07,
      "loss": -0.1408,
      "reward": 0.3818422555923462,
      "reward_std": 0.14291005209088326,
      "reward_std_all": 1.3187931776046753,
      "rewards-max/forensics_sinkhorn_reward": 0.8053024709224701,
      "rewards-max/forensics_strict_edge_reward": 2.4281384851221333e-05,
      "rewards-max/forensics_tiou_reward": 0.5773855894804001,
      "rewards/forensics_sinkhorn_reward": 0.3163292706012726,
      "rewards/forensics_strict_edge_reward": 1.0117528826958377e-06,
      "rewards/forensics_tiou_reward": 0.06551197543740273,
      "step": 3
    },
    {
      "completion_length": 25.28125,
      "epoch": 0.01228878648233487,
      "grad_norm": 9.345783233642578,
      "kl": 0.005218505859375,
      "learning_rate": 9.976923076923077e-07,
      "loss": -0.107,
      "reward": 0.3876323103904724,
      "reward_std": 0.12363637983798981,
      "reward_std_all": 1.3060373067855835,
      "rewards-max/forensics_sinkhorn_reward": 0.6409639120101929,
      "rewards-max/forensics_strict_edge_reward": 1.4675101939554472e-06,
      "rewards-max/forensics_tiou_reward": 0.28499235957860947,
      "rewards/forensics_sinkhorn_reward": 0.34712690114974976,
      "rewards/forensics_strict_edge_reward": 6.550994205404442e-08,
      "rewards/forensics_tiou_reward": 0.04050532355904579,
      "step": 4
    },
    {
      "completion_length": 23.703125,
      "epoch": 0.015360983102918587,
      "grad_norm": 7.408661842346191,
      "kl": 0.024658203125,
      "learning_rate": 9.969230769230768e-07,
      "loss": -0.146,
      "reward": 0.38043294847011566,
      "reward_std": 0.2388673648238182,
      "reward_std_all": 1.3380218148231506,
      "rewards-max/forensics_sinkhorn_reward": 0.8545466363430023,
      "rewards-max/forensics_strict_edge_reward": 0.013661876251652494,
      "rewards-max/forensics_tiou_reward": 0.6963701248168945,
      "rewards/forensics_sinkhorn_reward": 0.30355488508939743,
      "rewards/forensics_strict_edge_reward": 0.0005745445410275662,
      "rewards/forensics_tiou_reward": 0.07630351558327675,
      "step": 5
    },
    {
      "completion_length": 25.546875,
      "epoch": 0.018433179723502304,
      "grad_norm": 5.6102705001831055,
      "kl": 0.04052734375,
      "learning_rate": 9.961538461538461e-07,
      "loss": -0.1738,
      "reward": 0.4615098088979721,
      "reward_std": 0.1868392825126648,
      "reward_std_all": 1.2886309623718262,
      "rewards-max/forensics_sinkhorn_reward": 0.8067167699337006,
      "rewards-max/forensics_strict_edge_reward": 0.0010224556608591229,
      "rewards-max/forensics_tiou_reward": 0.6495614796876907,
      "rewards/forensics_sinkhorn_reward": 0.3476274013519287,
      "rewards/forensics_strict_edge_reward": 4.624676785169868e-05,
      "rewards/forensics_tiou_reward": 0.11383618786931038,
      "step": 6
    },
    {
      "completion_length": 26.921875,
      "epoch": 0.021505376344086023,
      "grad_norm": 4.636892318725586,
      "kl": 0.0743408203125,
      "learning_rate": 9.953846153846154e-07,
      "loss": -0.1512,
      "reward": 0.34780192375183105,
      "reward_std": 0.09131139144301414,
      "reward_std_all": 1.3243870735168457,
      "rewards-max/forensics_sinkhorn_reward": 0.593824028968811,
      "rewards-max/forensics_strict_edge_reward": 1.1517023949636496e-05,
      "rewards-max/forensics_tiou_reward": 0.25181838870048523,
      "rewards/forensics_sinkhorn_reward": 0.2892417758703232,
      "rewards/forensics_strict_edge_reward": 4.830671400668507e-07,
      "rewards/forensics_tiou_reward": 0.058559663593769073,
      "step": 7
    },
    {
      "completion_length": 24.703125,
      "epoch": 0.02457757296466974,
      "grad_norm": 6.846715927124023,
      "kl": 0.071044921875,
      "learning_rate": 9.946153846153846e-07,
      "loss": -0.1718,
      "reward": 0.4003940224647522,
      "reward_std": 0.18433703109622002,
      "reward_std_all": 1.3220415115356445,
      "rewards-max/forensics_sinkhorn_reward": 0.7260840237140656,
      "rewards-max/forensics_strict_edge_reward": 0.0878657112173411,
      "rewards-max/forensics_tiou_reward": 0.5592242032289505,
      "rewards/forensics_sinkhorn_reward": 0.3168761432170868,
      "rewards/forensics_strict_edge_reward": 0.0037839226824219682,
      "rewards/forensics_tiou_reward": 0.0797339640557766,
      "step": 8
    },
    {
      "completion_length": 25.0625,
      "epoch": 0.027649769585253458,
      "grad_norm": 4.187779426574707,
      "kl": 0.108154296875,
      "learning_rate": 9.938461538461539e-07,
      "loss": -0.1505,
      "reward": 0.5438766479492188,
      "reward_std": 0.29911768436431885,
      "reward_std_all": 1.2675523161888123,
      "rewards-max/forensics_sinkhorn_reward": 0.8865947127342224,
      "rewards-max/forensics_strict_edge_reward": 0.16462573781609535,
      "rewards-max/forensics_tiou_reward": 0.7575401663780212,
      "rewards/forensics_sinkhorn_reward": 0.40203604102134705,
      "rewards/forensics_strict_edge_reward": 0.00704411044716835,
      "rewards/forensics_tiou_reward": 0.13479650765657425,
      "step": 9
    },
    {
      "completion_length": 30.40625,
      "epoch": 0.030721966205837174,
      "grad_norm": 7.461478233337402,
      "kl": 0.19287109375,
      "learning_rate": 9.93076923076923e-07,
      "loss": -0.1106,
      "reward": 0.4144246131181717,
      "reward_std": 0.09398166835308075,
      "reward_std_all": 1.2912858724594116,
      "rewards-max/forensics_sinkhorn_reward": 0.6317728459835052,
      "rewards-max/forensics_strict_edge_reward": 0.017245799172087573,
      "rewards-max/forensics_tiou_reward": 0.4871901124715805,
      "rewards/forensics_sinkhorn_reward": 0.3198162019252777,
      "rewards/forensics_strict_edge_reward": 0.0007334073688980425,
      "rewards/forensics_tiou_reward": 0.09387500956654549,
      "step": 10
    },
    {
      "completion_length": 30.78125,
      "epoch": 0.03379416282642089,
      "grad_norm": 8.809865951538086,
      "kl": 0.21435546875,
      "learning_rate": 9.923076923076923e-07,
      "loss": -0.1509,
      "reward": 0.3654591590166092,
      "reward_std": 0.18569406867027283,
      "reward_std_all": 1.3395617008209229,
      "rewards-max/forensics_sinkhorn_reward": 0.7386344969272614,
      "rewards-max/forensics_strict_edge_reward": 0.02038121884957178,
      "rewards-max/forensics_tiou_reward": 0.584500640630722,
      "rewards/forensics_sinkhorn_reward": 0.259238600730896,
      "rewards/forensics_strict_edge_reward": 0.0008493322115894486,
      "rewards/forensics_tiou_reward": 0.10537124425172806,
      "step": 11
    },
    {
      "completion_length": 32.265625,
      "epoch": 0.03686635944700461,
      "grad_norm": 4.442704200744629,
      "kl": 0.2216796875,
      "learning_rate": 9.915384615384616e-07,
      "loss": -0.1592,
      "reward": 0.4362516701221466,
      "reward_std": 0.1333378069102764,
      "reward_std_all": 1.2917985320091248,
      "rewards-max/forensics_sinkhorn_reward": 0.6689943969249725,
      "rewards-max/forensics_strict_edge_reward": 0.2639648346485046,
      "rewards-max/forensics_tiou_reward": 0.49939651787281036,
      "rewards/forensics_sinkhorn_reward": 0.32279732823371887,
      "rewards/forensics_strict_edge_reward": 0.010998783106145993,
      "rewards/forensics_tiou_reward": 0.10245556384325027,
      "step": 12
    },
    {
      "completion_length": 27.125,
      "epoch": 0.039938556067588324,
      "grad_norm": 7.238396644592285,
      "kl": 0.23095703125,
      "learning_rate": 9.907692307692307e-07,
      "loss": -0.1818,
      "reward": 0.428651824593544,
      "reward_std": 0.2255651280283928,
      "reward_std_all": 1.3235796689987183,
      "rewards-max/forensics_sinkhorn_reward": 0.9508825540542603,
      "rewards-max/forensics_strict_edge_reward": 0.21017520345965401,
      "rewards-max/forensics_tiou_reward": 0.9007682800292969,
      "rewards/forensics_sinkhorn_reward": 0.321326345205307,
      "rewards/forensics_strict_edge_reward": 0.009848845907027726,
      "rewards/forensics_tiou_reward": 0.09747661650180817,
      "step": 13
    },
    {
      "completion_length": 35.421875,
      "epoch": 0.043010752688172046,
      "grad_norm": 5.918492317199707,
      "kl": 0.16748046875,
      "learning_rate": 9.9e-07,
      "loss": -0.1092,
      "reward": 0.37833331525325775,
      "reward_std": 0.1283337101340294,
      "reward_std_all": 1.3121069073677063,
      "rewards-max/forensics_sinkhorn_reward": 0.6611891984939575,
      "rewards-max/forensics_strict_edge_reward": 0.00026620406218569315,
      "rewards-max/forensics_tiou_reward": 0.4109000265598297,
      "rewards/forensics_sinkhorn_reward": 0.2574455440044403,
      "rewards/forensics_strict_edge_reward": 1.1226117065277208e-05,
      "rewards/forensics_tiou_reward": 0.12087655812501907,
      "step": 14
    },
    {
      "completion_length": 30.6875,
      "epoch": 0.04608294930875576,
      "grad_norm": 6.173614501953125,
      "kl": 0.220703125,
      "learning_rate": 9.892307692307692e-07,
      "loss": -0.1208,
      "reward": 0.436187744140625,
      "reward_std": 0.1862214133143425,
      "reward_std_all": 1.3019255995750427,
      "rewards-max/forensics_sinkhorn_reward": 0.7877554595470428,
      "rewards-max/forensics_strict_edge_reward": 0.26630911231040955,
      "rewards-max/forensics_tiou_reward": 0.7931505739688873,
      "rewards/forensics_sinkhorn_reward": 0.3108120560646057,
      "rewards/forensics_strict_edge_reward": 0.011162742739543319,
      "rewards/forensics_tiou_reward": 0.11421292647719383,
      "step": 15
    },
    {
      "completion_length": 26.15625,
      "epoch": 0.04915514592933948,
      "grad_norm": 3.9143004417419434,
      "kl": 0.21484375,
      "learning_rate": 9.884615384615385e-07,
      "loss": -0.1132,
      "reward": 0.5134466290473938,
      "reward_std": 0.30397067964076996,
      "reward_std_all": 1.2850096225738525,
      "rewards-max/forensics_sinkhorn_reward": 0.8920511305332184,
      "rewards-max/forensics_strict_edge_reward": 0.004334909885074012,
      "rewards-max/forensics_tiou_reward": 0.7555508315563202,
      "rewards/forensics_sinkhorn_reward": 0.3713267594575882,
      "rewards/forensics_strict_edge_reward": 0.00024819600093906047,
      "rewards/forensics_tiou_reward": 0.14187165349721909,
      "step": 16
    },
    {
      "completion_length": 27.28125,
      "epoch": 0.05222734254992319,
      "grad_norm": 3.8146493434906006,
      "kl": 0.1962890625,
      "learning_rate": 9.876923076923076e-07,
      "loss": -0.1382,
      "reward": 0.3593156635761261,
      "reward_std": 0.1540566310286522,
      "reward_std_all": 1.3342419862747192,
      "rewards-max/forensics_sinkhorn_reward": 0.7859238386154175,
      "rewards-max/forensics_strict_edge_reward": 0.00010273499985558043,
      "rewards-max/forensics_tiou_reward": 0.5374405309557915,
      "rewards/forensics_sinkhorn_reward": 0.30130261182785034,
      "rewards/forensics_strict_edge_reward": 4.2955520673168546e-06,
      "rewards/forensics_tiou_reward": 0.05800873599946499,
      "step": 17
    },
    {
      "completion_length": 21.03125,
      "epoch": 0.055299539170506916,
      "grad_norm": 4.4787068367004395,
      "kl": 0.22607421875,
      "learning_rate": 9.86923076923077e-07,
      "loss": -0.089,
      "reward": 0.48965421319007874,
      "reward_std": 0.19427943229675293,
      "reward_std_all": 1.2746570706367493,
      "rewards-max/forensics_sinkhorn_reward": 0.8713568150997162,
      "rewards-max/forensics_strict_edge_reward": 0.0067681989166885614,
      "rewards-max/forensics_tiou_reward": 0.7298399806022644,
      "rewards/forensics_sinkhorn_reward": 0.40520182251930237,
      "rewards/forensics_strict_edge_reward": 0.0002832458121702075,
      "rewards/forensics_tiou_reward": 0.0841691754758358,
      "step": 18
    },
    {
      "completion_length": 24.0,
      "epoch": 0.05837173579109063,
      "grad_norm": 3.8896219730377197,
      "kl": 0.1845703125,
      "learning_rate": 9.86153846153846e-07,
      "loss": -0.1633,
      "reward": 0.5193594098091125,
      "reward_std": 0.25546708703041077,
      "reward_std_all": 1.282039761543274,
      "rewards-max/forensics_sinkhorn_reward": 0.9282115995883942,
      "rewards-max/forensics_strict_edge_reward": 0.17031004093587399,
      "rewards-max/forensics_tiou_reward": 0.8438183665275574,
      "rewards/forensics_sinkhorn_reward": 0.3961738795042038,
      "rewards/forensics_strict_edge_reward": 0.011042308411560953,
      "rewards/forensics_tiou_reward": 0.11214318126440048,
      "step": 19
    },
    {
      "completion_length": 25.625,
      "epoch": 0.06144393241167435,
      "grad_norm": 3.759608268737793,
      "kl": 0.18798828125,
      "learning_rate": 9.853846153846153e-07,
      "loss": -0.1428,
      "reward": 0.5166539400815964,
      "reward_std": 0.21971212327480316,
      "reward_std_all": 1.270405888557434,
      "rewards-max/forensics_sinkhorn_reward": 0.9207989871501923,
      "rewards-max/forensics_strict_edge_reward": 0.3619650974869728,
      "rewards-max/forensics_tiou_reward": 0.8079496324062347,
      "rewards/forensics_sinkhorn_reward": 0.3770024925470352,
      "rewards/forensics_strict_edge_reward": 0.015081914141774178,
      "rewards/forensics_tiou_reward": 0.12456955015659332,
      "step": 20
    },
    {
      "completion_length": 21.40625,
      "epoch": 0.06451612903225806,
      "grad_norm": 5.015868663787842,
      "kl": 0.2265625,
      "learning_rate": 9.846153846153847e-07,
      "loss": -0.1122,
      "reward": 0.49773918092250824,
      "reward_std": 0.22470834478735924,
      "reward_std_all": 1.2805359959602356,
      "rewards-max/forensics_sinkhorn_reward": 0.8084620237350464,
      "rewards-max/forensics_strict_edge_reward": 0.018446792147187807,
      "rewards-max/forensics_tiou_reward": 0.627473920583725,
      "rewards/forensics_sinkhorn_reward": 0.39962562918663025,
      "rewards/forensics_strict_edge_reward": 0.0007878240031402584,
      "rewards/forensics_tiou_reward": 0.09732570499181747,
      "step": 21
    },
    {
      "completion_length": 22.484375,
      "epoch": 0.06758832565284179,
      "grad_norm": 5.681004524230957,
      "kl": 0.1826171875,
      "learning_rate": 9.838461538461538e-07,
      "loss": -0.1236,
      "reward": 0.4609204977750778,
      "reward_std": 0.11833734065294266,
      "reward_std_all": 1.2772359251976013,
      "rewards-max/forensics_sinkhorn_reward": 0.6889024674892426,
      "rewards-max/forensics_strict_edge_reward": 0.23884377083141706,
      "rewards-max/forensics_tiou_reward": 0.5256055891513824,
      "rewards/forensics_sinkhorn_reward": 0.3839259445667267,
      "rewards/forensics_strict_edge_reward": 0.009951885872814023,
      "rewards/forensics_tiou_reward": 0.06704270839691162,
      "step": 22
    },
    {
      "completion_length": 23.296875,
      "epoch": 0.0706605222734255,
      "grad_norm": 3.7150990962982178,
      "kl": 0.24267578125,
      "learning_rate": 9.83076923076923e-07,
      "loss": -0.1046,
      "reward": 0.5954944491386414,
      "reward_std": 0.14031445235013962,
      "reward_std_all": 1.2034409046173096,
      "rewards-max/forensics_sinkhorn_reward": 0.7538225054740906,
      "rewards-max/forensics_strict_edge_reward": 0.00852439044183484,
      "rewards-max/forensics_tiou_reward": 0.5070266127586365,
      "rewards/forensics_sinkhorn_reward": 0.4709973782300949,
      "rewards/forensics_strict_edge_reward": 0.00035522342236049553,
      "rewards/forensics_tiou_reward": 0.12414181977510452,
      "step": 23
    },
    {
      "completion_length": 24.890625,
      "epoch": 0.07373271889400922,
      "grad_norm": 5.320615291595459,
      "kl": 0.177734375,
      "learning_rate": 9.823076923076922e-07,
      "loss": -0.1484,
      "reward": 0.554190993309021,
      "reward_std": 0.16289880871772766,
      "reward_std_all": 1.236474871635437,
      "rewards-max/forensics_sinkhorn_reward": 0.774649053812027,
      "rewards-max/forensics_strict_edge_reward": 0.012780872357794237,
      "rewards-max/forensics_tiou_reward": 0.5396134853363037,
      "rewards/forensics_sinkhorn_reward": 0.4354008287191391,
      "rewards/forensics_strict_edge_reward": 0.0005325413964669679,
      "rewards/forensics_tiou_reward": 0.11825758591294289,
      "step": 24
    },
    {
      "completion_length": 21.78125,
      "epoch": 0.07680491551459294,
      "grad_norm": 3.9735171794891357,
      "kl": 0.2373046875,
      "learning_rate": 9.815384615384615e-07,
      "loss": -0.1717,
      "reward": 0.4730145037174225,
      "reward_std": 0.24233540892601013,
      "reward_std_all": 1.2946405410766602,
      "rewards-max/forensics_sinkhorn_reward": 0.8904722034931183,
      "rewards-max/forensics_strict_edge_reward": 0.06301653233822435,
      "rewards-max/forensics_tiou_reward": 0.7455790340900421,
      "rewards/forensics_sinkhorn_reward": 0.38738909363746643,
      "rewards/forensics_strict_edge_reward": 0.0026257384743075818,
      "rewards/forensics_tiou_reward": 0.08299969136714935,
      "step": 25
    },
    {
      "completion_length": 24.59375,
      "epoch": 0.07987711213517665,
      "grad_norm": 4.19720983505249,
      "kl": 0.1953125,
      "learning_rate": 9.807692307692306e-07,
      "loss": -0.1924,
      "reward": 0.5794048011302948,
      "reward_std": 0.30084648728370667,
      "reward_std_all": 1.2646671533584595,
      "rewards-max/forensics_sinkhorn_reward": 0.9457493424415588,
      "rewards-max/forensics_strict_edge_reward": 0.24097890773555264,
      "rewards-max/forensics_tiou_reward": 0.8469889760017395,
      "rewards/forensics_sinkhorn_reward": 0.4338945150375366,
      "rewards/forensics_strict_edge_reward": 0.012694474062300287,
      "rewards/forensics_tiou_reward": 0.13281582295894623,
      "step": 26
    },
    {
      "completion_length": 25.59375,
      "epoch": 0.08294930875576037,
      "grad_norm": 4.522003650665283,
      "kl": 0.20751953125,
      "learning_rate": 9.8e-07,
      "loss": -0.1332,
      "reward": 0.5866266787052155,
      "reward_std": 0.19968610256910324,
      "reward_std_all": 1.2263636589050293,
      "rewards-max/forensics_sinkhorn_reward": 0.8658015429973602,
      "rewards-max/forensics_strict_edge_reward": 0.0640855427045608,
      "rewards-max/forensics_tiou_reward": 0.729079395532608,
      "rewards/forensics_sinkhorn_reward": 0.43979401886463165,
      "rewards/forensics_strict_edge_reward": 0.002670930960903206,
      "rewards/forensics_tiou_reward": 0.14416176825761795,
      "step": 27
    },
    {
      "completion_length": 24.328125,
      "epoch": 0.08602150537634409,
      "grad_norm": 3.604429006576538,
      "kl": 0.216796875,
      "learning_rate": 9.792307692307693e-07,
      "loss": -0.1195,
      "reward": 0.5131649672985077,
      "reward_std": 0.13236935064196587,
      "reward_std_all": 1.2489749193191528,
      "rewards-max/forensics_sinkhorn_reward": 0.7421555817127228,
      "rewards-max/forensics_strict_edge_reward": 0.00019951620548730453,
      "rewards-max/forensics_tiou_reward": 0.5353869497776031,
      "rewards/forensics_sinkhorn_reward": 0.404898464679718,
      "rewards/forensics_strict_edge_reward": 8.584006069650707e-06,
      "rewards/forensics_tiou_reward": 0.1082579530775547,
      "step": 28
    },
    {
      "completion_length": 23.578125,
      "epoch": 0.0890937019969278,
      "grad_norm": 4.718179702758789,
      "kl": 0.2119140625,
      "learning_rate": 9.784615384615384e-07,
      "loss": -0.1494,
      "reward": 0.5802572220563889,
      "reward_std": 0.28185104206204414,
      "reward_std_all": 1.2559031248092651,
      "rewards-max/forensics_sinkhorn_reward": 0.7904944121837616,
      "rewards-max/forensics_strict_edge_reward": 0.08264951313388025,
      "rewards-max/forensics_tiou_reward": 0.5582785084843636,
      "rewards/forensics_sinkhorn_reward": 0.43253660202026367,
      "rewards/forensics_strict_edge_reward": 0.003520771356031238,
      "rewards/forensics_tiou_reward": 0.14419987611472607,
      "step": 29
    },
    {
      "completion_length": 33.0,
      "epoch": 0.09216589861751152,
      "grad_norm": 4.893036365509033,
      "kl": 0.17919921875,
      "learning_rate": 9.776923076923077e-07,
      "loss": -0.12,
      "reward": 0.5156619846820831,
      "reward_std": 0.23033608496189117,
      "reward_std_all": 1.2752195000648499,
      "rewards-max/forensics_sinkhorn_reward": 0.9363621473312378,
      "rewards-max/forensics_strict_edge_reward": 0.07923215627670288,
      "rewards-max/forensics_tiou_reward": 0.8598074913024902,
      "rewards/forensics_sinkhorn_reward": 0.36911769211292267,
      "rewards/forensics_strict_edge_reward": 0.003755218116566539,
      "rewards/forensics_tiou_reward": 0.1427890658378601,
      "step": 30
    },
    {
      "completion_length": 27.046875,
      "epoch": 0.09523809523809523,
      "grad_norm": 4.096416473388672,
      "kl": 0.17041015625,
      "learning_rate": 9.769230769230768e-07,
      "loss": -0.1517,
      "reward": 0.5411942005157471,
      "reward_std": 0.183096781373024,
      "reward_std_all": 1.2402859330177307,
      "rewards-max/forensics_sinkhorn_reward": 0.8006228506565094,
      "rewards-max/forensics_strict_edge_reward": 0.013661861579044338,
      "rewards-max/forensics_tiou_reward": 0.5741160213947296,
      "rewards/forensics_sinkhorn_reward": 0.42355312407016754,
      "rewards/forensics_strict_edge_reward": 0.0005692450280876818,
      "rewards/forensics_tiou_reward": 0.11707179993391037,
      "step": 31
    },
    {
      "completion_length": 28.015625,
      "epoch": 0.09831029185867896,
      "grad_norm": 5.186222553253174,
      "kl": 0.20068359375,
      "learning_rate": 9.761538461538461e-07,
      "loss": -0.1517,
      "reward": 0.5879552066326141,
      "reward_std": 0.3217540383338928,
      "reward_std_all": 1.2535142302513123,
      "rewards-max/forensics_sinkhorn_reward": 0.8780783116817474,
      "rewards-max/forensics_strict_edge_reward": 0.19672049209475517,
      "rewards-max/forensics_tiou_reward": 0.7580855786800385,
      "rewards/forensics_sinkhorn_reward": 0.4074255973100662,
      "rewards/forensics_strict_edge_reward": 0.008312728954479098,
      "rewards/forensics_tiou_reward": 0.17221688479185104,
      "step": 32
    },
    {
      "completion_length": 28.53125,
      "epoch": 0.10138248847926268,
      "grad_norm": 3.1409788131713867,
      "kl": 0.1943359375,
      "learning_rate": 9.753846153846154e-07,
      "loss": -0.1697,
      "reward": 0.5431902408599854,
      "reward_std": 0.23658551275730133,
      "reward_std_all": 1.2650447487831116,
      "rewards-max/forensics_sinkhorn_reward": 0.9003156423568726,
      "rewards-max/forensics_strict_edge_reward": 0.02533146063797176,
      "rewards-max/forensics_tiou_reward": 0.7977602481842041,
      "rewards/forensics_sinkhorn_reward": 0.4108152836561203,
      "rewards/forensics_strict_edge_reward": 0.001166638350696303,
      "rewards/forensics_tiou_reward": 0.1312083601951599,
      "step": 33
    },
    {
      "completion_length": 27.890625,
      "epoch": 0.10445468509984639,
      "grad_norm": 4.488781929016113,
      "kl": 0.20947265625,
      "learning_rate": 9.746153846153845e-07,
      "loss": -0.1403,
      "reward": 0.602284699678421,
      "reward_std": 0.2726464346051216,
      "reward_std_all": 1.2352269291877747,
      "rewards-max/forensics_sinkhorn_reward": 0.9117324650287628,
      "rewards-max/forensics_strict_edge_reward": 0.016686777398930985,
      "rewards-max/forensics_tiou_reward": 0.7607530951499939,
      "rewards/forensics_sinkhorn_reward": 0.4323710799217224,
      "rewards/forensics_strict_edge_reward": 0.0011112638670578434,
      "rewards/forensics_tiou_reward": 0.16880235075950623,
      "step": 34
    },
    {
      "completion_length": 27.328125,
      "epoch": 0.10752688172043011,
      "grad_norm": 3.7323853969573975,
      "kl": 0.22216796875,
      "learning_rate": 9.738461538461539e-07,
      "loss": -0.1622,
      "reward": 0.5894497036933899,
      "reward_std": 0.23558174073696136,
      "reward_std_all": 1.2358797192573547,
      "rewards-max/forensics_sinkhorn_reward": 0.9016894102096558,
      "rewards-max/forensics_strict_edge_reward": 0.2513144686818123,
      "rewards-max/forensics_tiou_reward": 0.7997856736183167,
      "rewards/forensics_sinkhorn_reward": 0.4456307739019394,
      "rewards/forensics_strict_edge_reward": 0.013022658880800009,
      "rewards/forensics_tiou_reward": 0.13079623132944107,
      "step": 35
    },
    {
      "completion_length": 27.015625,
      "epoch": 0.11059907834101383,
      "grad_norm": 4.149996280670166,
      "kl": 0.193359375,
      "learning_rate": 9.73076923076923e-07,
      "loss": -0.1177,
      "reward": 0.5739648342132568,
      "reward_std": 0.2719877362251282,
      "reward_std_all": 1.2546907663345337,
      "rewards-max/forensics_sinkhorn_reward": 0.9355571568012238,
      "rewards-max/forensics_strict_edge_reward": 0.07479411933582014,
      "rewards-max/forensics_tiou_reward": 0.8479482233524323,
      "rewards/forensics_sinkhorn_reward": 0.39083848893642426,
      "rewards/forensics_strict_edge_reward": 0.0034724449532745894,
      "rewards/forensics_tiou_reward": 0.17965389043092728,
      "step": 36
    },
    {
      "completion_length": 29.109375,
      "epoch": 0.11367127496159754,
      "grad_norm": 4.237454891204834,
      "kl": 0.21630859375,
      "learning_rate": 9.723076923076923e-07,
      "loss": -0.1363,
      "reward": 0.41797277331352234,
      "reward_std": 0.1406511291861534,
      "reward_std_all": 1.297800898551941,
      "rewards-max/forensics_sinkhorn_reward": 0.7377596199512482,
      "rewards-max/forensics_strict_edge_reward": 0.0002976624164148234,
      "rewards-max/forensics_tiou_reward": 0.5623148828744888,
      "rewards/forensics_sinkhorn_reward": 0.30078670382499695,
      "rewards/forensics_strict_edge_reward": 1.2878904612989572e-05,
      "rewards/forensics_tiou_reward": 0.11717319115996361,
      "step": 37
    },
    {
      "completion_length": 27.5,
      "epoch": 0.11674347158218126,
      "grad_norm": 4.6025309562683105,
      "kl": 0.2060546875,
      "learning_rate": 9.715384615384616e-07,
      "loss": -0.1203,
      "reward": 0.5103887915611267,
      "reward_std": 0.19178726896643639,
      "reward_std_all": 1.2660963535308838,
      "rewards-max/forensics_sinkhorn_reward": 0.8025510013103485,
      "rewards-max/forensics_strict_edge_reward": 0.1318770222933381,
      "rewards-max/forensics_tiou_reward": 0.6678169965744019,
      "rewards/forensics_sinkhorn_reward": 0.3538770377635956,
      "rewards/forensics_strict_edge_reward": 0.006611316059206729,
      "rewards/forensics_tiou_reward": 0.1499004289507866,
      "step": 38
    },
    {
      "completion_length": 27.59375,
      "epoch": 0.11981566820276497,
      "grad_norm": 4.003723621368408,
      "kl": 0.22412109375,
      "learning_rate": 9.707692307692307e-07,
      "loss": -0.0958,
      "reward": 0.5163972675800323,
      "reward_std": 0.2119985967874527,
      "reward_std_all": 1.2660390138626099,
      "rewards-max/forensics_sinkhorn_reward": 0.8984632194042206,
      "rewards-max/forensics_strict_edge_reward": 0.0005970713384613191,
      "rewards-max/forensics_tiou_reward": 0.7865651249885559,
      "rewards/forensics_sinkhorn_reward": 0.3707522600889206,
      "rewards/forensics_strict_edge_reward": 2.588214708509895e-05,
      "rewards/forensics_tiou_reward": 0.14561911672353745,
      "step": 39
    },
    {
      "completion_length": 30.703125,
      "epoch": 0.1228878648233487,
      "grad_norm": 3.75766658782959,
      "kl": 0.18798828125,
      "learning_rate": 9.7e-07,
      "loss": -0.1193,
      "reward": 0.5329866111278534,
      "reward_std": 0.2107878401875496,
      "reward_std_all": 1.2547361254692078,
      "rewards-max/forensics_sinkhorn_reward": 0.8160233795642853,
      "rewards-max/forensics_strict_edge_reward": 0.00039714889888387006,
      "rewards-max/forensics_tiou_reward": 0.692937508225441,
      "rewards/forensics_sinkhorn_reward": 0.3780204504728317,
      "rewards/forensics_strict_edge_reward": 1.7581455040271976e-05,
      "rewards/forensics_tiou_reward": 0.15494856238365173,
      "step": 40
    },
    {
      "completion_length": 26.765625,
      "epoch": 0.1259600614439324,
      "grad_norm": 4.786641597747803,
      "kl": 0.17041015625,
      "learning_rate": 9.692307692307691e-07,
      "loss": -0.15,
      "reward": 0.5788355320692062,
      "reward_std": 0.2253374382853508,
      "reward_std_all": 1.2375217080116272,
      "rewards-max/forensics_sinkhorn_reward": 0.8274794518947601,
      "rewards-max/forensics_strict_edge_reward": 0.01375637393988427,
      "rewards-max/forensics_tiou_reward": 0.6463418900966644,
      "rewards/forensics_sinkhorn_reward": 0.44254449009895325,
      "rewards/forensics_strict_edge_reward": 0.0006024746243209478,
      "rewards/forensics_tiou_reward": 0.13568859547376633,
      "step": 41
    },
    {
      "completion_length": 24.078125,
      "epoch": 0.12903225806451613,
      "grad_norm": 3.3373165130615234,
      "kl": 0.17919921875,
      "learning_rate": 9.684615384615385e-07,
      "loss": -0.1432,
      "reward": 0.5898805260658264,
      "reward_std": 0.24092119932174683,
      "reward_std_all": 1.2324743866920471,
      "rewards-max/forensics_sinkhorn_reward": 0.7901397049427032,
      "rewards-max/forensics_strict_edge_reward": 0.005025920658694449,
      "rewards-max/forensics_tiou_reward": 0.5903613865375519,
      "rewards/forensics_sinkhorn_reward": 0.4489065855741501,
      "rewards/forensics_strict_edge_reward": 0.0003216212698391313,
      "rewards/forensics_tiou_reward": 0.1406523510813713,
      "step": 42
    },
    {
      "completion_length": 23.859375,
      "epoch": 0.13210445468509985,
      "grad_norm": 4.355539798736572,
      "kl": 0.181640625,
      "learning_rate": 9.676923076923078e-07,
      "loss": -0.0805,
      "reward": 0.5495947003364563,
      "reward_std": 0.18880809843540192,
      "reward_std_all": 1.2401345372200012,
      "rewards-max/forensics_sinkhorn_reward": 0.8446557819843292,
      "rewards-max/forensics_strict_edge_reward": 6.947309838523097e-08,
      "rewards-max/forensics_tiou_reward": 0.6171526312828064,
      "rewards/forensics_sinkhorn_reward": 0.4341787099838257,
      "rewards/forensics_strict_edge_reward": 6.723255285834284e-09,
      "rewards/forensics_tiou_reward": 0.11541596800088882,
      "step": 43
    },
    {
      "completion_length": 23.34375,
      "epoch": 0.13517665130568357,
      "grad_norm": 3.6316516399383545,
      "kl": 0.1884765625,
      "learning_rate": 9.669230769230769e-07,
      "loss": -0.0914,
      "reward": 0.5076144337654114,
      "reward_std": 0.18329399079084396,
      "reward_std_all": 1.2581992149353027,
      "rewards-max/forensics_sinkhorn_reward": 0.7392106354236603,
      "rewards-max/forensics_strict_edge_reward": 0.17113909955048,
      "rewards-max/forensics_tiou_reward": 0.4923073351383209,
      "rewards/forensics_sinkhorn_reward": 0.4192349314689636,
      "rewards/forensics_strict_edge_reward": 0.007130816425785724,
      "rewards/forensics_tiou_reward": 0.0812486968934536,
      "step": 44
    },
    {
      "completion_length": 22.546875,
      "epoch": 0.1382488479262673,
      "grad_norm": 4.413534164428711,
      "kl": 0.19384765625,
      "learning_rate": 9.66153846153846e-07,
      "loss": -0.0834,
      "reward": 0.6536144614219666,
      "reward_std": 0.275473415851593,
      "reward_std_all": 1.2030264735221863,
      "rewards-max/forensics_sinkhorn_reward": 0.9137557446956635,
      "rewards-max/forensics_strict_edge_reward": 0.01263437932357192,
      "rewards-max/forensics_tiou_reward": 0.7865800559520721,
      "rewards/forensics_sinkhorn_reward": 0.502713680267334,
      "rewards/forensics_strict_edge_reward": 0.0005321782373357564,
      "rewards/forensics_tiou_reward": 0.1503685787320137,
      "step": 45
    },
    {
      "completion_length": 23.96875,
      "epoch": 0.141321044546851,
      "grad_norm": 5.54576301574707,
      "kl": 0.21923828125,
      "learning_rate": 9.653846153846153e-07,
      "loss": -0.1734,
      "reward": 0.5740984678268433,
      "reward_std": 0.2702847719192505,
      "reward_std_all": 1.2556475400924683,
      "rewards-max/forensics_sinkhorn_reward": 0.9359227120876312,
      "rewards-max/forensics_strict_edge_reward": 0.01759448880329728,
      "rewards-max/forensics_tiou_reward": 0.8393693268299103,
      "rewards/forensics_sinkhorn_reward": 0.419009268283844,
      "rewards/forensics_strict_edge_reward": 0.0008101934799924493,
      "rewards/forensics_tiou_reward": 0.15427901223301888,
      "step": 46
    },
    {
      "completion_length": 25.703125,
      "epoch": 0.1443932411674347,
      "grad_norm": 4.055161952972412,
      "kl": 0.20654296875,
      "learning_rate": 9.646153846153846e-07,
      "loss": -0.1186,
      "reward": 0.6016701757907867,
      "reward_std": 0.24820144474506378,
      "reward_std_all": 1.234801173210144,
      "rewards-max/forensics_sinkhorn_reward": 0.9515325427055359,
      "rewards-max/forensics_strict_edge_reward": 0.2217167615890503,
      "rewards-max/forensics_tiou_reward": 0.890499085187912,
      "rewards/forensics_sinkhorn_reward": 0.43220506608486176,
      "rewards/forensics_strict_edge_reward": 0.010541165480390191,
      "rewards/forensics_tiou_reward": 0.1589239537715912,
      "step": 47
    },
    {
      "completion_length": 24.03125,
      "epoch": 0.14746543778801843,
      "grad_norm": 3.9309651851654053,
      "kl": 0.2109375,
      "learning_rate": 9.638461538461537e-07,
      "loss": -0.1513,
      "reward": 0.6104657649993896,
      "reward_std": 0.19393349066376686,
      "reward_std_all": 1.2121371626853943,
      "rewards-max/forensics_sinkhorn_reward": 0.786678671836853,
      "rewards-max/forensics_strict_edge_reward": 0.005025921085805685,
      "rewards-max/forensics_tiou_reward": 0.5667612850666046,
      "rewards/forensics_sinkhorn_reward": 0.47301529347896576,
      "rewards/forensics_strict_edge_reward": 0.00020943973092052426,
      "rewards/forensics_tiou_reward": 0.13724102824926376,
      "step": 48
    },
    {
      "completion_length": 24.171875,
      "epoch": 0.15053763440860216,
      "grad_norm": 3.684154987335205,
      "kl": 0.21875,
      "learning_rate": 9.63076923076923e-07,
      "loss": -0.083,
      "reward": 0.580735370516777,
      "reward_std": 0.1904025487601757,
      "reward_std_all": 1.225302815437317,
      "rewards-max/forensics_sinkhorn_reward": 0.6926053464412689,
      "rewards-max/forensics_strict_edge_reward": 0.08902962198772002,
      "rewards-max/forensics_tiou_reward": 0.4703674241900444,
      "rewards/forensics_sinkhorn_reward": 0.45235709846019745,
      "rewards/forensics_strict_edge_reward": 0.004364489854197018,
      "rewards/forensics_tiou_reward": 0.12401382997632027,
      "step": 49
    },
    {
      "completion_length": 26.40625,
      "epoch": 0.15360983102918588,
      "grad_norm": 4.785772323608398,
      "kl": 0.18408203125,
      "learning_rate": 9.623076923076922e-07,
      "loss": -0.1758,
      "reward": 0.5470135509967804,
      "reward_std": 0.2743440717458725,
      "reward_std_all": 1.2721489071846008,
      "rewards-max/forensics_sinkhorn_reward": 0.9258136749267578,
      "rewards-max/forensics_strict_edge_reward": 0.008502052107360214,
      "rewards-max/forensics_tiou_reward": 0.8327035307884216,
      "rewards/forensics_sinkhorn_reward": 0.3943830281496048,
      "rewards/forensics_strict_edge_reward": 0.0003715722959896084,
      "rewards/forensics_tiou_reward": 0.1522589735686779,
      "step": 50
    },
    {
      "completion_length": 25.84375,
      "epoch": 0.15668202764976957,
      "grad_norm": 3.6024656295776367,
      "kl": 0.2041015625,
      "learning_rate": 9.615384615384615e-07,
      "loss": -0.1603,
      "reward": 0.5753266215324402,
      "reward_std": 0.22966965287923813,
      "reward_std_all": 1.2341354489326477,
      "rewards-max/forensics_sinkhorn_reward": 0.8202514946460724,
      "rewards-max/forensics_strict_edge_reward": 0.1830303780734539,
      "rewards-max/forensics_tiou_reward": 0.7095365822315216,
      "rewards/forensics_sinkhorn_reward": 0.413878932595253,
      "rewards/forensics_strict_edge_reward": 0.00762635882711038,
      "rewards/forensics_tiou_reward": 0.15382131189107895,
      "step": 51
    },
    {
      "completion_length": 25.359375,
      "epoch": 0.1597542242703533,
      "grad_norm": 4.685666084289551,
      "kl": 0.21044921875,
      "learning_rate": 9.607692307692308e-07,
      "loss": -0.1317,
      "reward": 0.7260808348655701,
      "reward_std": 0.26307062804698944,
      "reward_std_all": 1.173309087753296,
      "rewards-max/forensics_sinkhorn_reward": 0.944805771112442,
      "rewards-max/forensics_strict_edge_reward": 0.35607264935970306,
      "rewards-max/forensics_tiou_reward": 0.8704377710819244,
      "rewards/forensics_sinkhorn_reward": 0.5024839639663696,
      "rewards/forensics_strict_edge_reward": 0.015860290732234716,
      "rewards/forensics_tiou_reward": 0.20773663371801376,
      "step": 52
    },
    {
      "completion_length": 30.859375,
      "epoch": 0.16282642089093702,
      "grad_norm": 4.35025691986084,
      "kl": 0.17333984375,
      "learning_rate": 9.6e-07,
      "loss": -0.1096,
      "reward": 0.6817749440670013,
      "reward_std": 0.2765054404735565,
      "reward_std_all": 1.1944644451141357,
      "rewards-max/forensics_sinkhorn_reward": 0.9660372138023376,
      "rewards-max/forensics_strict_edge_reward": 0.08201125077903271,
      "rewards-max/forensics_tiou_reward": 0.9430513978004456,
      "rewards/forensics_sinkhorn_reward": 0.4557470828294754,
      "rewards/forensics_strict_edge_reward": 0.0036175919231027365,
      "rewards/forensics_tiou_reward": 0.22241030633449554,
      "step": 53
    },
    {
      "completion_length": 25.546875,
      "epoch": 0.16589861751152074,
      "grad_norm": 4.7644877433776855,
      "kl": 0.17822265625,
      "learning_rate": 9.592307692307692e-07,
      "loss": -0.1697,
      "reward": 0.6822539567947388,
      "reward_std": 0.31797249615192413,
      "reward_std_all": 1.2113213539123535,
      "rewards-max/forensics_sinkhorn_reward": 0.9206923544406891,
      "rewards-max/forensics_strict_edge_reward": 0.12668556347489357,
      "rewards-max/forensics_tiou_reward": 0.8324877619743347,
      "rewards/forensics_sinkhorn_reward": 0.4432571828365326,
      "rewards/forensics_strict_edge_reward": 0.006106054992415011,
      "rewards/forensics_tiou_reward": 0.232890747487545,
      "step": 54
    },
    {
      "completion_length": 27.125,
      "epoch": 0.16897081413210446,
      "grad_norm": 3.7419161796569824,
      "kl": 0.2431640625,
      "learning_rate": 9.584615384615383e-07,
      "loss": -0.1755,
      "reward": 0.6843858063220978,
      "reward_std": 0.2568824514746666,
      "reward_std_all": 1.1946929097175598,
      "rewards-max/forensics_sinkhorn_reward": 0.938902884721756,
      "rewards-max/forensics_strict_edge_reward": 0.36595215648412704,
      "rewards-max/forensics_tiou_reward": 0.878780722618103,
      "rewards/forensics_sinkhorn_reward": 0.4583958685398102,
      "rewards/forensics_strict_edge_reward": 0.01713996473699808,
      "rewards/forensics_tiou_reward": 0.2088499329984188,
      "step": 55
    },
    {
      "completion_length": 27.3125,
      "epoch": 0.17204301075268819,
      "grad_norm": 3.824681043624878,
      "kl": 0.236328125,
      "learning_rate": 9.576923076923077e-07,
      "loss": -0.0874,
      "reward": 0.4549226462841034,
      "reward_std": 0.18917106464505196,
      "reward_std_all": 1.2918716073036194,
      "rewards-max/forensics_sinkhorn_reward": 0.7729831039905548,
      "rewards-max/forensics_strict_edge_reward": 0.01195649616695212,
      "rewards-max/forensics_tiou_reward": 0.574035033583641,
      "rewards/forensics_sinkhorn_reward": 0.3458065241575241,
      "rewards/forensics_strict_edge_reward": 0.0004983046093425635,
      "rewards/forensics_tiou_reward": 0.10861779749393463,
      "step": 56
    },
    {
      "completion_length": 24.71875,
      "epoch": 0.17511520737327188,
      "grad_norm": 3.5868988037109375,
      "kl": 0.21484375,
      "learning_rate": 9.569230769230768e-07,
      "loss": -0.1241,
      "reward": 0.669460117816925,
      "reward_std": 0.27882810682058334,
      "reward_std_all": 1.2063607573509216,
      "rewards-max/forensics_sinkhorn_reward": 0.9378860592842102,
      "rewards-max/forensics_strict_edge_reward": 0.2884473651647568,
      "rewards-max/forensics_tiou_reward": 0.8747000098228455,
      "rewards/forensics_sinkhorn_reward": 0.4591553509235382,
      "rewards/forensics_strict_edge_reward": 0.01565828826278448,
      "rewards/forensics_tiou_reward": 0.1946464627981186,
      "step": 57
    },
    {
      "completion_length": 24.5,
      "epoch": 0.1781874039938556,
      "grad_norm": 4.705322265625,
      "kl": 0.19384765625,
      "learning_rate": 9.56153846153846e-07,
      "loss": -0.1566,
      "reward": 0.6407491564750671,
      "reward_std": 0.32678473740816116,
      "reward_std_all": 1.2515900135040283,
      "rewards-max/forensics_sinkhorn_reward": 0.9328555464744568,
      "rewards-max/forensics_strict_edge_reward": 0.32465688584488817,
      "rewards-max/forensics_tiou_reward": 0.8086560368537903,
      "rewards/forensics_sinkhorn_reward": 0.4348820000886917,
      "rewards/forensics_strict_edge_reward": 0.02843669574940577,
      "rewards/forensics_tiou_reward": 0.1774304285645485,
      "step": 58
    },
    {
      "completion_length": 28.96875,
      "epoch": 0.18125960061443933,
      "grad_norm": 4.697961330413818,
      "kl": 0.21337890625,
      "learning_rate": 9.553846153846154e-07,
      "loss": -0.1324,
      "reward": 0.44205084443092346,
      "reward_std": 0.17482366785407066,
      "reward_std_all": 1.3005483150482178,
      "rewards-max/forensics_sinkhorn_reward": 0.7975657880306244,
      "rewards-max/forensics_strict_edge_reward": 0.03780001413542777,
      "rewards-max/forensics_tiou_reward": 0.6298219561576843,
      "rewards/forensics_sinkhorn_reward": 0.3133952021598816,
      "rewards/forensics_strict_edge_reward": 0.0015760432870592922,
      "rewards/forensics_tiou_reward": 0.12707960233092308,
      "step": 59
    },
    {
      "completion_length": 26.78125,
      "epoch": 0.18433179723502305,
      "grad_norm": 5.391278266906738,
      "kl": 0.23291015625,
      "learning_rate": 9.546153846153845e-07,
      "loss": -0.0975,
      "reward": 0.48954230546951294,
      "reward_std": 0.1123570166528225,
      "reward_std_all": 1.2554545402526855,
      "rewards-max/forensics_sinkhorn_reward": 0.6630142033100128,
      "rewards-max/forensics_strict_edge_reward": 1.7898905397828457e-06,
      "rewards-max/forensics_tiou_reward": 0.3133802115917206,
      "rewards/forensics_sinkhorn_reward": 0.3812762200832367,
      "rewards/forensics_strict_edge_reward": 7.804469603911457e-08,
      "rewards/forensics_tiou_reward": 0.10826598852872849,
      "step": 60
    },
    {
      "completion_length": 29.140625,
      "epoch": 0.18740399385560677,
      "grad_norm": 4.358666896820068,
      "kl": 0.1923828125,
      "learning_rate": 9.538461538461538e-07,
      "loss": -0.1553,
      "reward": 0.4679064452648163,
      "reward_std": 0.1624574363231659,
      "reward_std_all": 1.2792755365371704,
      "rewards-max/forensics_sinkhorn_reward": 0.8278732597827911,
      "rewards-max/forensics_strict_edge_reward": 0.1254037325270474,
      "rewards-max/forensics_tiou_reward": 0.7165287137031555,
      "rewards/forensics_sinkhorn_reward": 0.3505748063325882,
      "rewards/forensics_strict_edge_reward": 0.005233706135186367,
      "rewards/forensics_tiou_reward": 0.11209793016314507,
      "step": 61
    },
    {
      "completion_length": 26.140625,
      "epoch": 0.19047619047619047,
      "grad_norm": 4.021928787231445,
      "kl": 0.22021484375,
      "learning_rate": 9.53076923076923e-07,
      "loss": -0.1308,
      "reward": 0.5815093219280243,
      "reward_std": 0.22141142189502716,
      "reward_std_all": 1.2361329197883606,
      "rewards-max/forensics_sinkhorn_reward": 0.8870708644390106,
      "rewards-max/forensics_strict_edge_reward": 0.15060665232977044,
      "rewards-max/forensics_tiou_reward": 0.7407096326351166,
      "rewards/forensics_sinkhorn_reward": 0.4337289035320282,
      "rewards/forensics_strict_edge_reward": 0.006275278701735942,
      "rewards/forensics_tiou_reward": 0.1415051594376564,
      "step": 62
    },
    {
      "completion_length": 24.609375,
      "epoch": 0.1935483870967742,
      "grad_norm": 3.185966968536377,
      "kl": 0.201171875,
      "learning_rate": 9.523076923076923e-07,
      "loss": -0.0726,
      "reward": 0.4948492646217346,
      "reward_std": 0.13634946942329407,
      "reward_std_all": 1.2546008229255676,
      "rewards-max/forensics_sinkhorn_reward": 0.7104527950286865,
      "rewards-max/forensics_strict_edge_reward": 0.028159676119685173,
      "rewards-max/forensics_tiou_reward": 0.5741734206676483,
      "rewards/forensics_sinkhorn_reward": 0.38005152344703674,
      "rewards/forensics_strict_edge_reward": 0.0011734834697563201,
      "rewards/forensics_tiou_reward": 0.11362428218126297,
      "step": 63
    },
    {
      "completion_length": 27.046875,
      "epoch": 0.1966205837173579,
      "grad_norm": 3.8759286403656006,
      "kl": 0.1953125,
      "learning_rate": 9.515384615384615e-07,
      "loss": -0.0714,
      "reward": 0.592516154050827,
      "reward_std": 0.22964372113347054,
      "reward_std_all": 1.2284786701202393,
      "rewards-max/forensics_sinkhorn_reward": 0.7999715805053711,
      "rewards-max/forensics_strict_edge_reward": 0.0018033950971130253,
      "rewards-max/forensics_tiou_reward": 0.6124520003795624,
      "rewards/forensics_sinkhorn_reward": 0.4238215833902359,
      "rewards/forensics_strict_edge_reward": 7.525209622372131e-05,
      "rewards/forensics_tiou_reward": 0.16861934959888458,
      "step": 64
    },
    {
      "completion_length": 25.78125,
      "epoch": 0.19969278033794163,
      "grad_norm": 2.971221923828125,
      "kl": 0.171875,
      "learning_rate": 9.507692307692308e-07,
      "loss": -0.0464,
      "reward": 0.6478650569915771,
      "reward_std": 0.20228565484285355,
      "reward_std_all": 1.193384051322937,
      "rewards-max/forensics_sinkhorn_reward": 0.8295961022377014,
      "rewards-max/forensics_strict_edge_reward": 0.003615864602579677,
      "rewards-max/forensics_tiou_reward": 0.5969041734933853,
      "rewards/forensics_sinkhorn_reward": 0.4789330065250397,
      "rewards/forensics_strict_edge_reward": 0.00015081466153787915,
      "rewards/forensics_tiou_reward": 0.16878119111061096,
      "step": 65
    },
    {
      "completion_length": 25.015625,
      "epoch": 0.20276497695852536,
      "grad_norm": 3.936776876449585,
      "kl": 0.2119140625,
      "learning_rate": 9.499999999999999e-07,
      "loss": -0.1539,
      "reward": 0.599419355392456,
      "reward_std": 0.28139103949069977,
      "reward_std_all": 1.243406057357788,
      "rewards-max/forensics_sinkhorn_reward": 0.9395139217376709,
      "rewards-max/forensics_strict_edge_reward": 0.1154277044115588,
      "rewards-max/forensics_tiou_reward": 0.8829672634601593,
      "rewards/forensics_sinkhorn_reward": 0.4439091384410858,
      "rewards/forensics_strict_edge_reward": 0.004816192416001286,
      "rewards/forensics_tiou_reward": 0.15069394558668137,
      "step": 66
    },
    {
      "completion_length": 24.484375,
      "epoch": 0.20583717357910905,
      "grad_norm": 4.068624019622803,
      "kl": 0.189453125,
      "learning_rate": 9.492307692307691e-07,
      "loss": -0.116,
      "reward": 0.492275208234787,
      "reward_std": 0.16921687871217728,
      "reward_std_all": 1.2642695903778076,
      "rewards-max/forensics_sinkhorn_reward": 0.7418169379234314,
      "rewards-max/forensics_strict_edge_reward": 0.00018370342149864882,
      "rewards-max/forensics_tiou_reward": 0.5052126199007034,
      "rewards/forensics_sinkhorn_reward": 0.38373909890651703,
      "rewards/forensics_strict_edge_reward": 7.808363079675473e-06,
      "rewards/forensics_tiou_reward": 0.108528271317482,
      "step": 67
    },
    {
      "completion_length": 25.53125,
      "epoch": 0.20890937019969277,
      "grad_norm": 4.468069553375244,
      "kl": 0.1884765625,
      "learning_rate": 9.484615384615384e-07,
      "loss": -0.1202,
      "reward": 0.5475893914699554,
      "reward_std": 0.25651081651449203,
      "reward_std_all": 1.277584433555603,
      "rewards-max/forensics_sinkhorn_reward": 0.9765310883522034,
      "rewards-max/forensics_strict_edge_reward": 0.4714259207248688,
      "rewards-max/forensics_tiou_reward": 0.9701370894908905,
      "rewards/forensics_sinkhorn_reward": 0.39981304109096527,
      "rewards/forensics_strict_edge_reward": 0.01964278519153595,
      "rewards/forensics_tiou_reward": 0.12813356518745422,
      "step": 68
    },
    {
      "completion_length": 23.71875,
      "epoch": 0.2119815668202765,
      "grad_norm": 3.550971746444702,
      "kl": 0.18798828125,
      "learning_rate": 9.476923076923077e-07,
      "loss": -0.1912,
      "reward": 0.6416870951652527,
      "reward_std": 0.447289377450943,
      "reward_std_all": 1.2641912698745728,
      "rewards-max/forensics_sinkhorn_reward": 0.9412312805652618,
      "rewards-max/forensics_strict_edge_reward": 0.17284106463193893,
      "rewards-max/forensics_tiou_reward": 0.8837761878967285,
      "rewards/forensics_sinkhorn_reward": 0.4347153306007385,
      "rewards/forensics_strict_edge_reward": 0.008080908097326756,
      "rewards/forensics_tiou_reward": 0.19889087229967117,
      "step": 69
    },
    {
      "completion_length": 28.875,
      "epoch": 0.21505376344086022,
      "grad_norm": 3.3017616271972656,
      "kl": 0.17431640625,
      "learning_rate": 9.46923076923077e-07,
      "loss": -0.1683,
      "reward": 0.5807040929794312,
      "reward_std": 0.23802901059389114,
      "reward_std_all": 1.2432583570480347,
      "rewards-max/forensics_sinkhorn_reward": 0.9117431044578552,
      "rewards-max/forensics_strict_edge_reward": 0.10025409515947104,
      "rewards-max/forensics_tiou_reward": 0.8081941604614258,
      "rewards/forensics_sinkhorn_reward": 0.40119488537311554,
      "rewards/forensics_strict_edge_reward": 0.005085438606329262,
      "rewards/forensics_tiou_reward": 0.17442376911640167,
      "step": 70
    },
    {
      "completion_length": 25.046875,
      "epoch": 0.21812596006144394,
      "grad_norm": 3.207207202911377,
      "kl": 0.18896484375,
      "learning_rate": 9.461538461538461e-07,
      "loss": -0.0464,
      "reward": 0.6283261775970459,
      "reward_std": 0.20792506635189056,
      "reward_std_all": 1.207359492778778,
      "rewards-max/forensics_sinkhorn_reward": 0.8755140900611877,
      "rewards-max/forensics_strict_edge_reward": 0.015610637731455768,
      "rewards-max/forensics_tiou_reward": 0.7342118620872498,
      "rewards/forensics_sinkhorn_reward": 0.45931947231292725,
      "rewards/forensics_strict_edge_reward": 0.000650450030445171,
      "rewards/forensics_tiou_reward": 0.1683562695980072,
      "step": 71
    },
    {
      "completion_length": 23.421875,
      "epoch": 0.22119815668202766,
      "grad_norm": 3.3704683780670166,
      "kl": 0.18310546875,
      "learning_rate": 9.453846153846153e-07,
      "loss": -0.0988,
      "reward": 0.49175870418548584,
      "reward_std": 0.22166930884122849,
      "reward_std_all": 1.2843703031539917,
      "rewards-max/forensics_sinkhorn_reward": 0.9078764319419861,
      "rewards-max/forensics_strict_edge_reward": 0.05124426539987326,
      "rewards-max/forensics_tiou_reward": 0.8111063539981842,
      "rewards/forensics_sinkhorn_reward": 0.387490838766098,
      "rewards/forensics_strict_edge_reward": 0.002144387166481465,
      "rewards/forensics_tiou_reward": 0.10212351754307747,
      "step": 72
    },
    {
      "completion_length": 23.46875,
      "epoch": 0.22427035330261136,
      "grad_norm": 3.0334556102752686,
      "kl": 0.1875,
      "learning_rate": 9.446153846153846e-07,
      "loss": -0.1098,
      "reward": 0.7359307706356049,
      "reward_std": 0.27109768241643906,
      "reward_std_all": 1.174526333808899,
      "rewards-max/forensics_sinkhorn_reward": 0.9029435515403748,
      "rewards-max/forensics_strict_edge_reward": 0.4027831554412842,
      "rewards-max/forensics_tiou_reward": 0.7515157759189606,
      "rewards/forensics_sinkhorn_reward": 0.4998118281364441,
      "rewards/forensics_strict_edge_reward": 0.020029394887387753,
      "rewards/forensics_tiou_reward": 0.21608953922986984,
      "step": 73
    },
    {
      "completion_length": 24.75,
      "epoch": 0.22734254992319508,
      "grad_norm": 4.503236770629883,
      "kl": 0.220703125,
      "learning_rate": 9.438461538461538e-07,
      "loss": -0.0839,
      "reward": 0.5579365342855453,
      "reward_std": 0.173659086227417,
      "reward_std_all": 1.233432412147522,
      "rewards-max/forensics_sinkhorn_reward": 0.8355679214000702,
      "rewards-max/forensics_strict_edge_reward": 3.055438767669838e-06,
      "rewards-max/forensics_tiou_reward": 0.5782559663057327,
      "rewards/forensics_sinkhorn_reward": 0.4318079799413681,
      "rewards/forensics_strict_edge_reward": 1.4274751469322416e-07,
      "rewards/forensics_tiou_reward": 0.12612834572792053,
      "step": 74
    },
    {
      "completion_length": 22.984375,
      "epoch": 0.2304147465437788,
      "grad_norm": 4.898789882659912,
      "kl": 0.20751953125,
      "learning_rate": 9.43076923076923e-07,
      "loss": -0.1138,
      "reward": 0.6964720487594604,
      "reward_std": 0.25770825147628784,
      "reward_std_all": 1.190110206604004,
      "rewards-max/forensics_sinkhorn_reward": 0.9387616217136383,
      "rewards-max/forensics_strict_edge_reward": 0.06796641834080219,
      "rewards-max/forensics_tiou_reward": 0.8612698018550873,
      "rewards/forensics_sinkhorn_reward": 0.48090873658657074,
      "rewards/forensics_strict_edge_reward": 0.0028394300024956465,
      "rewards/forensics_tiou_reward": 0.21272391825914383,
      "step": 75
    },
    {
      "completion_length": 21.203125,
      "epoch": 0.23348694316436253,
      "grad_norm": 3.9669737815856934,
      "kl": 0.19921875,
      "learning_rate": 9.423076923076923e-07,
      "loss": -0.0977,
      "reward": 0.5824858397245407,
      "reward_std": 0.17449332028627396,
      "reward_std_all": 1.2199462056159973,
      "rewards-max/forensics_sinkhorn_reward": 0.8764417469501495,
      "rewards-max/forensics_strict_edge_reward": 0.0022790880757384002,
      "rewards-max/forensics_tiou_reward": 0.7134937644004822,
      "rewards/forensics_sinkhorn_reward": 0.461177259683609,
      "rewards/forensics_strict_edge_reward": 9.907976163958665e-05,
      "rewards/forensics_tiou_reward": 0.1212095096707344,
      "step": 76
    },
    {
      "completion_length": 22.59375,
      "epoch": 0.23655913978494625,
      "grad_norm": 4.2066216468811035,
      "kl": 0.2001953125,
      "learning_rate": 9.415384615384615e-07,
      "loss": -0.1043,
      "reward": 0.47426638007164,
      "reward_std": 0.1334007829427719,
      "reward_std_all": 1.2678257822990417,
      "rewards-max/forensics_sinkhorn_reward": 0.7338027358055115,
      "rewards-max/forensics_strict_edge_reward": 6.181163114149513e-05,
      "rewards-max/forensics_tiou_reward": 0.41388964653015137,
      "rewards/forensics_sinkhorn_reward": 0.3885818123817444,
      "rewards/forensics_strict_edge_reward": 2.8494162291892167e-06,
      "rewards/forensics_tiou_reward": 0.08568169176578522,
      "step": 77
    },
    {
      "completion_length": 21.6875,
      "epoch": 0.23963133640552994,
      "grad_norm": 3.7567520141601562,
      "kl": 0.2001953125,
      "learning_rate": 9.407692307692308e-07,
      "loss": -0.0754,
      "reward": 0.7275541424751282,
      "reward_std": 0.2918770760297775,
      "reward_std_all": 1.1832395792007446,
      "rewards-max/forensics_sinkhorn_reward": 0.9684055745601654,
      "rewards-max/forensics_strict_edge_reward": 0.2386450693011284,
      "rewards-max/forensics_tiou_reward": 0.9493469297885895,
      "rewards/forensics_sinkhorn_reward": 0.5156828314065933,
      "rewards/forensics_strict_edge_reward": 0.010439061094075441,
      "rewards/forensics_tiou_reward": 0.2014322727918625,
      "step": 78
    },
    {
      "completion_length": 24.9375,
      "epoch": 0.24270353302611367,
      "grad_norm": 3.3074846267700195,
      "kl": 0.20458984375,
      "learning_rate": 9.399999999999999e-07,
      "loss": -0.1387,
      "reward": 0.6009069085121155,
      "reward_std": 0.20747506245970726,
      "reward_std_all": 1.2214503288269043,
      "rewards-max/forensics_sinkhorn_reward": 0.802160233259201,
      "rewards-max/forensics_strict_edge_reward": 0.033729707822203636,
      "rewards-max/forensics_tiou_reward": 0.6718096435070038,
      "rewards/forensics_sinkhorn_reward": 0.426323264837265,
      "rewards/forensics_strict_edge_reward": 0.001614327891729772,
      "rewards/forensics_tiou_reward": 0.17296932637691498,
      "step": 79
    },
    {
      "completion_length": 22.09375,
      "epoch": 0.2457757296466974,
      "grad_norm": 3.5058743953704834,
      "kl": 0.23193359375,
      "learning_rate": 9.392307692307691e-07,
      "loss": -0.0678,
      "reward": 0.6737700700759888,
      "reward_std": 0.23452170193195343,
      "reward_std_all": 1.2022840976715088,
      "rewards-max/forensics_sinkhorn_reward": 0.8875068426132202,
      "rewards-max/forensics_strict_edge_reward": 0.2246644824722981,
      "rewards-max/forensics_tiou_reward": 0.7176894545555115,
      "rewards/forensics_sinkhorn_reward": 0.45074139535427094,
      "rewards/forensics_strict_edge_reward": 0.017027167780914176,
      "rewards/forensics_tiou_reward": 0.20600154995918274,
      "step": 80
    },
    {
      "completion_length": 25.828125,
      "epoch": 0.2488479262672811,
      "grad_norm": 4.695311069488525,
      "kl": 0.19970703125,
      "learning_rate": 9.384615384615384e-07,
      "loss": -0.0835,
      "reward": 0.6565687954425812,
      "reward_std": 0.25452762842178345,
      "reward_std_all": 1.2122334837913513,
      "rewards-max/forensics_sinkhorn_reward": 0.9706761240959167,
      "rewards-max/forensics_strict_edge_reward": 0.20382346049882472,
      "rewards-max/forensics_tiou_reward": 0.950086921453476,
      "rewards/forensics_sinkhorn_reward": 0.44005316495895386,
      "rewards/forensics_strict_edge_reward": 0.008523810691258404,
      "rewards/forensics_tiou_reward": 0.2079918533563614,
      "step": 81
    },
    {
      "completion_length": 27.359375,
      "epoch": 0.2519201228878648,
      "grad_norm": 4.903571128845215,
      "kl": 0.259765625,
      "learning_rate": 9.376923076923076e-07,
      "loss": -0.0923,
      "reward": 0.6323717832565308,
      "reward_std": 0.27279703319072723,
      "reward_std_all": 1.2164369821548462,
      "rewards-max/forensics_sinkhorn_reward": 0.9314071238040924,
      "rewards-max/forensics_strict_edge_reward": 0.07305609062314034,
      "rewards-max/forensics_tiou_reward": 0.8524621725082397,
      "rewards/forensics_sinkhorn_reward": 0.4369144141674042,
      "rewards/forensics_strict_edge_reward": 0.003044096170924604,
      "rewards/forensics_tiou_reward": 0.19241327047348022,
      "step": 82
    },
    {
      "completion_length": 24.609375,
      "epoch": 0.25499231950844853,
      "grad_norm": 3.183236598968506,
      "kl": 0.1767578125,
      "learning_rate": 9.36923076923077e-07,
      "loss": -0.1307,
      "reward": 0.5830034613609314,
      "reward_std": 0.22242460399866104,
      "reward_std_all": 1.2395784258842468,
      "rewards-max/forensics_sinkhorn_reward": 0.830968052148819,
      "rewards-max/forensics_strict_edge_reward": 0.12330256100722181,
      "rewards-max/forensics_tiou_reward": 0.6975933611392975,
      "rewards/forensics_sinkhorn_reward": 0.40777163207530975,
      "rewards/forensics_strict_edge_reward": 0.0053597907906208775,
      "rewards/forensics_tiou_reward": 0.169871985912323,
      "step": 83
    },
    {
      "completion_length": 22.34375,
      "epoch": 0.25806451612903225,
      "grad_norm": 4.474985599517822,
      "kl": 0.2509765625,
      "learning_rate": 9.361538461538461e-07,
      "loss": -0.1322,
      "reward": 0.6462559103965759,
      "reward_std": 0.31725814938545227,
      "reward_std_all": 1.2265020608901978,
      "rewards-max/forensics_sinkhorn_reward": 0.9525248110294342,
      "rewards-max/forensics_strict_edge_reward": 0.24168037623167038,
      "rewards-max/forensics_tiou_reward": 0.9013889729976654,
      "rewards/forensics_sinkhorn_reward": 0.4531036615371704,
      "rewards/forensics_strict_edge_reward": 0.010254994849674404,
      "rewards/forensics_tiou_reward": 0.18289728462696075,
      "step": 84
    },
    {
      "completion_length": 27.75,
      "epoch": 0.261136712749616,
      "grad_norm": 3.783344268798828,
      "kl": 0.20751953125,
      "learning_rate": 9.353846153846153e-07,
      "loss": -0.1004,
      "reward": 0.6226890087127686,
      "reward_std": 0.2395569011569023,
      "reward_std_all": 1.225530743598938,
      "rewards-max/forensics_sinkhorn_reward": 0.9025161564350128,
      "rewards-max/forensics_strict_edge_reward": 0.17565325321629643,
      "rewards-max/forensics_tiou_reward": 0.8049976229667664,
      "rewards/forensics_sinkhorn_reward": 0.4291418045759201,
      "rewards/forensics_strict_edge_reward": 0.011671666688926052,
      "rewards/forensics_tiou_reward": 0.1818755567073822,
      "step": 85
    },
    {
      "completion_length": 23.765625,
      "epoch": 0.2642089093701997,
      "grad_norm": 4.353981971740723,
      "kl": 0.23388671875,
      "learning_rate": 9.346153846153846e-07,
      "loss": -0.0579,
      "reward": 0.691448986530304,
      "reward_std": 0.23747825622558594,
      "reward_std_all": 1.177795648574829,
      "rewards-max/forensics_sinkhorn_reward": 0.8125932812690735,
      "rewards-max/forensics_strict_edge_reward": 0.03954867215361446,
      "rewards-max/forensics_tiou_reward": 0.6823189556598663,
      "rewards/forensics_sinkhorn_reward": 0.5023747086524963,
      "rewards/forensics_strict_edge_reward": 0.0021851699930266477,
      "rewards/forensics_tiou_reward": 0.18688911199569702,
      "step": 86
    },
    {
      "completion_length": 26.046875,
      "epoch": 0.2672811059907834,
      "grad_norm": 3.677826404571533,
      "kl": 0.224609375,
      "learning_rate": 9.338461538461538e-07,
      "loss": -0.1024,
      "reward": 0.6572063565254211,
      "reward_std": 0.27450714260339737,
      "reward_std_all": 1.2192556858062744,
      "rewards-max/forensics_sinkhorn_reward": 0.9585237801074982,
      "rewards-max/forensics_strict_edge_reward": 0.0003248125431127846,
      "rewards-max/forensics_tiou_reward": 0.9084151685237885,
      "rewards/forensics_sinkhorn_reward": 0.4402554780244827,
      "rewards/forensics_strict_edge_reward": 2.1926893168711103e-05,
      "rewards/forensics_tiou_reward": 0.21692897379398346,
      "step": 87
    },
    {
      "completion_length": 25.34375,
      "epoch": 0.27035330261136714,
      "grad_norm": 3.4071762561798096,
      "kl": 0.22119140625,
      "learning_rate": 9.33076923076923e-07,
      "loss": -0.0757,
      "reward": 0.7733998894691467,
      "reward_std": 0.36548617482185364,
      "reward_std_all": 1.177419126033783,
      "rewards-max/forensics_sinkhorn_reward": 0.9478439390659332,
      "rewards-max/forensics_strict_edge_reward": 0.2048705443739891,
      "rewards-max/forensics_tiou_reward": 0.9013735949993134,
      "rewards/forensics_sinkhorn_reward": 0.4911404848098755,
      "rewards/forensics_strict_edge_reward": 0.009780643973499537,
      "rewards/forensics_tiou_reward": 0.27247871458530426,
      "step": 88
    },
    {
      "completion_length": 25.53125,
      "epoch": 0.27342549923195086,
      "grad_norm": 3.937356948852539,
      "kl": 0.2060546875,
      "learning_rate": 9.323076923076922e-07,
      "loss": -0.1378,
      "reward": 0.7164096534252167,
      "reward_std": 0.23771218955516815,
      "reward_std_all": 1.1848647594451904,
      "rewards-max/forensics_sinkhorn_reward": 0.9657383859157562,
      "rewards-max/forensics_strict_edge_reward": 0.26888667792081833,
      "rewards-max/forensics_tiou_reward": 0.9319477677345276,
      "rewards/forensics_sinkhorn_reward": 0.49111002683639526,
      "rewards/forensics_strict_edge_reward": 0.014478133991360664,
      "rewards/forensics_tiou_reward": 0.2108214944601059,
      "step": 89
    },
    {
      "completion_length": 25.265625,
      "epoch": 0.2764976958525346,
      "grad_norm": 4.428747177124023,
      "kl": 0.2021484375,
      "learning_rate": 9.315384615384615e-07,
      "loss": -0.1113,
      "reward": 0.755386233329773,
      "reward_std": 0.16855482757091522,
      "reward_std_all": 1.1440384984016418,
      "rewards-max/forensics_sinkhorn_reward": 0.8289439678192139,
      "rewards-max/forensics_strict_edge_reward": 0.19662056789900362,
      "rewards-max/forensics_tiou_reward": 0.6325919926166534,
      "rewards/forensics_sinkhorn_reward": 0.5181674659252167,
      "rewards/forensics_strict_edge_reward": 0.012024023591278521,
      "rewards/forensics_tiou_reward": 0.22519472241401672,
      "step": 90
    },
    {
      "completion_length": 21.875,
      "epoch": 0.27956989247311825,
      "grad_norm": 4.456953048706055,
      "kl": 0.26416015625,
      "learning_rate": 9.307692307692308e-07,
      "loss": -0.1768,
      "reward": 0.8537881672382355,
      "reward_std": 0.4806436747312546,
      "reward_std_all": 1.1699681878089905,
      "rewards-max/forensics_sinkhorn_reward": 0.9519807398319244,
      "rewards-max/forensics_strict_edge_reward": 0.33664843440055847,
      "rewards-max/forensics_tiou_reward": 0.9064554274082184,
      "rewards/forensics_sinkhorn_reward": 0.5344991385936737,
      "rewards/forensics_strict_edge_reward": 0.016148042399436235,
      "rewards/forensics_tiou_reward": 0.30314093828201294,
      "step": 91
    },
    {
      "completion_length": 23.828125,
      "epoch": 0.282642089093702,
      "grad_norm": 4.467982292175293,
      "kl": 0.23095703125,
      "learning_rate": 9.3e-07,
      "loss": -0.1397,
      "reward": 0.6369224637746811,
      "reward_std": 0.22429580986499786,
      "reward_std_all": 1.2145153284072876,
      "rewards-max/forensics_sinkhorn_reward": 0.8570336997509003,
      "rewards-max/forensics_strict_edge_reward": 0.0883492181640122,
      "rewards-max/forensics_tiou_reward": 0.6949992775917053,
      "rewards/forensics_sinkhorn_reward": 0.4456433057785034,
      "rewards/forensics_strict_edge_reward": 0.00632011583356018,
      "rewards/forensics_tiou_reward": 0.1849590353667736,
      "step": 92
    },
    {
      "completion_length": 27.734375,
      "epoch": 0.2857142857142857,
      "grad_norm": 5.529465198516846,
      "kl": 0.25537109375,
      "learning_rate": 9.292307692307692e-07,
      "loss": -0.1802,
      "reward": 0.6046651601791382,
      "reward_std": 0.2004835605621338,
      "reward_std_all": 1.2226530313491821,
      "rewards-max/forensics_sinkhorn_reward": 0.77016881108284,
      "rewards-max/forensics_strict_edge_reward": 0.07762309096870013,
      "rewards-max/forensics_tiou_reward": 0.7024856507778168,
      "rewards/forensics_sinkhorn_reward": 0.3899301588535309,
      "rewards/forensics_strict_edge_reward": 0.0035345927462913096,
      "rewards/forensics_tiou_reward": 0.21120042726397514,
      "step": 93
    },
    {
      "completion_length": 26.046875,
      "epoch": 0.2887864823348694,
      "grad_norm": 4.845691680908203,
      "kl": 0.2236328125,
      "learning_rate": 9.284615384615384e-07,
      "loss": -0.0707,
      "reward": 0.5231927335262299,
      "reward_std": 0.19667831808328629,
      "reward_std_all": 1.258561372756958,
      "rewards-max/forensics_sinkhorn_reward": 0.881174772977829,
      "rewards-max/forensics_strict_edge_reward": 0.014301850693300366,
      "rewards-max/forensics_tiou_reward": 0.7377609312534332,
      "rewards/forensics_sinkhorn_reward": 0.38724249601364136,
      "rewards/forensics_strict_edge_reward": 0.0005970372731098905,
      "rewards/forensics_tiou_reward": 0.1353531926870346,
      "step": 94
    },
    {
      "completion_length": 27.828125,
      "epoch": 0.29185867895545314,
      "grad_norm": 5.164733409881592,
      "kl": 0.19580078125,
      "learning_rate": 9.276923076923076e-07,
      "loss": -0.1305,
      "reward": 0.6225506067276001,
      "reward_std": 0.33224262297153473,
      "reward_std_all": 1.2426900267601013,
      "rewards-max/forensics_sinkhorn_reward": 0.962208092212677,
      "rewards-max/forensics_strict_edge_reward": 0.02339547371957451,
      "rewards-max/forensics_tiou_reward": 0.902397096157074,
      "rewards/forensics_sinkhorn_reward": 0.41276805102825165,
      "rewards/forensics_strict_edge_reward": 0.0009869574569165707,
      "rewards/forensics_tiou_reward": 0.20879559963941574,
      "step": 95
    },
    {
      "completion_length": 25.03125,
      "epoch": 0.29493087557603687,
      "grad_norm": 3.912236213684082,
      "kl": 0.20751953125,
      "learning_rate": 9.26923076923077e-07,
      "loss": -0.1638,
      "reward": 0.6877931654453278,
      "reward_std": 0.21905896067619324,
      "reward_std_all": 1.1848536729812622,
      "rewards-max/forensics_sinkhorn_reward": 0.9393472671508789,
      "rewards-max/forensics_strict_edge_reward": 0.014369936077855527,
      "rewards-max/forensics_tiou_reward": 0.8511930108070374,
      "rewards/forensics_sinkhorn_reward": 0.5013483166694641,
      "rewards/forensics_strict_edge_reward": 0.0010457887074153405,
      "rewards/forensics_tiou_reward": 0.18539907783269882,
      "step": 96
    },
    {
      "completion_length": 24.109375,
      "epoch": 0.2980030721966206,
      "grad_norm": 4.628684997558594,
      "kl": 0.25244140625,
      "learning_rate": 9.261538461538461e-07,
      "loss": -0.1322,
      "reward": 0.7638233304023743,
      "reward_std": 0.31596166640520096,
      "reward_std_all": 1.186242401599884,
      "rewards-max/forensics_sinkhorn_reward": 0.9713691174983978,
      "rewards-max/forensics_strict_edge_reward": 0.41472819447517395,
      "rewards-max/forensics_tiou_reward": 0.9523387849330902,
      "rewards/forensics_sinkhorn_reward": 0.49329259991645813,
      "rewards/forensics_strict_edge_reward": 0.028671087697148323,
      "rewards/forensics_tiou_reward": 0.24185965955257416,
      "step": 97
    },
    {
      "completion_length": 25.1875,
      "epoch": 0.3010752688172043,
      "grad_norm": 4.267979145050049,
      "kl": 0.2177734375,
      "learning_rate": 9.253846153846153e-07,
      "loss": -0.1448,
      "reward": 0.6506400108337402,
      "reward_std": 0.12712985277175903,
      "reward_std_all": 1.181099772453308,
      "rewards-max/forensics_sinkhorn_reward": 0.7565285265445709,
      "rewards-max/forensics_strict_edge_reward": 0.004715474271506537,
      "rewards-max/forensics_tiou_reward": 0.6005391627550125,
      "rewards/forensics_sinkhorn_reward": 0.467810794711113,
      "rewards/forensics_strict_edge_reward": 0.000196821768895461,
      "rewards/forensics_tiou_reward": 0.18263237178325653,
      "step": 98
    },
    {
      "completion_length": 25.296875,
      "epoch": 0.30414746543778803,
      "grad_norm": 3.7212772369384766,
      "kl": 0.26416015625,
      "learning_rate": 9.246153846153846e-07,
      "loss": -0.097,
      "reward": 0.561502993106842,
      "reward_std": 0.12204620987176895,
      "reward_std_all": 1.2208786010742188,
      "rewards-max/forensics_sinkhorn_reward": 0.7639069855213165,
      "rewards-max/forensics_strict_edge_reward": 1.0219011230105934e-05,
      "rewards-max/forensics_tiou_reward": 0.5556606352329254,
      "rewards/forensics_sinkhorn_reward": 0.4204822778701782,
      "rewards/forensics_strict_edge_reward": 5.502882994834479e-07,
      "rewards/forensics_tiou_reward": 0.14102023094892502,
      "step": 99
    },
    {
      "completion_length": 23.375,
      "epoch": 0.30721966205837176,
      "grad_norm": 3.890244245529175,
      "kl": 0.236328125,
      "learning_rate": 9.238461538461538e-07,
      "loss": -0.1513,
      "reward": 0.6126362383365631,
      "reward_std": 0.1998247653245926,
      "reward_std_all": 1.2228412628173828,
      "rewards-max/forensics_sinkhorn_reward": 0.8398756682872772,
      "rewards-max/forensics_strict_edge_reward": 0.09134268903540033,
      "rewards-max/forensics_tiou_reward": 0.6507904231548309,
      "rewards/forensics_sinkhorn_reward": 0.4523557126522064,
      "rewards/forensics_strict_edge_reward": 0.006193910982808859,
      "rewards/forensics_tiou_reward": 0.1540866158902645,
      "step": 100
    },
    {
      "completion_length": 23.78125,
      "epoch": 0.3102918586789555,
      "grad_norm": 4.371968746185303,
      "kl": 0.271484375,
      "learning_rate": 9.230769230769231e-07,
      "loss": -0.1136,
      "reward": 0.6057609617710114,
      "reward_std": 0.3916286826133728,
      "reward_std_all": 1.2811661958694458,
      "rewards-max/forensics_sinkhorn_reward": 0.9615724682807922,
      "rewards-max/forensics_strict_edge_reward": 0.2798496223986149,
      "rewards-max/forensics_tiou_reward": 0.9359732568264008,
      "rewards/forensics_sinkhorn_reward": 0.43044625222682953,
      "rewards/forensics_strict_edge_reward": 0.021565535105764866,
      "rewards/forensics_tiou_reward": 0.15374916791915894,
      "step": 101
    },
    {
      "completion_length": 23.9375,
      "epoch": 0.31336405529953915,
      "grad_norm": 4.383841514587402,
      "kl": 0.29296875,
      "learning_rate": 9.223076923076922e-07,
      "loss": -0.0831,
      "reward": 0.6167464256286621,
      "reward_std": 0.23088999092578888,
      "reward_std_all": 1.232787549495697,
      "rewards-max/forensics_sinkhorn_reward": 0.8064968883991241,
      "rewards-max/forensics_strict_edge_reward": 0.42324097276357975,
      "rewards-max/forensics_tiou_reward": 0.7276858538389206,
      "rewards/forensics_sinkhorn_reward": 0.41486552357673645,
      "rewards/forensics_strict_edge_reward": 0.01995632868802577,
      "rewards/forensics_tiou_reward": 0.18192455917596817,
      "step": 102
    },
    {
      "completion_length": 26.390625,
      "epoch": 0.31643625192012287,
      "grad_norm": 4.482112884521484,
      "kl": 0.2783203125,
      "learning_rate": 9.215384615384614e-07,
      "loss": -0.1548,
      "reward": 0.6943931579589844,
      "reward_std": 0.34759193658828735,
      "reward_std_all": 1.2116678953170776,
      "rewards-max/forensics_sinkhorn_reward": 0.9358438849449158,
      "rewards-max/forensics_strict_edge_reward": 0.14346841350197792,
      "rewards-max/forensics_tiou_reward": 0.8807200789451599,
      "rewards/forensics_sinkhorn_reward": 0.4833451062440872,
      "rewards/forensics_strict_edge_reward": 0.007732338272035122,
      "rewards/forensics_tiou_reward": 0.20331566035747528,
      "step": 103
    },
    {
      "completion_length": 28.4375,
      "epoch": 0.3195084485407066,
      "grad_norm": 4.972875118255615,
      "kl": 0.23193359375,
      "learning_rate": 9.207692307692308e-07,
      "loss": -0.1582,
      "reward": 0.566039428114891,
      "reward_std": 0.2196979820728302,
      "reward_std_all": 1.249595046043396,
      "rewards-max/forensics_sinkhorn_reward": 0.940264880657196,
      "rewards-max/forensics_strict_edge_reward": 0.11641974188387394,
      "rewards-max/forensics_tiou_reward": 0.8689634799957275,
      "rewards/forensics_sinkhorn_reward": 0.3845110088586807,
      "rewards/forensics_strict_edge_reward": 0.004851298115681857,
      "rewards/forensics_tiou_reward": 0.17667712271213531,
      "step": 104
    },
    {
      "completion_length": 25.40625,
      "epoch": 0.3225806451612903,
      "grad_norm": 3.753541946411133,
      "kl": 0.2373046875,
      "learning_rate": 9.2e-07,
      "loss": -0.0989,
      "reward": 0.7214126884937286,
      "reward_std": 0.21237455308437347,
      "reward_std_all": 1.1573545932769775,
      "rewards-max/forensics_sinkhorn_reward": 0.8775275349617004,
      "rewards-max/forensics_strict_edge_reward": 0.0008594785467721522,
      "rewards-max/forensics_tiou_reward": 0.6808474659919739,
      "rewards/forensics_sinkhorn_reward": 0.5123358368873596,
      "rewards/forensics_strict_edge_reward": 9.593312097422313e-05,
      "rewards/forensics_tiou_reward": 0.20898091793060303,
      "step": 105
    },
    {
      "completion_length": 28.0,
      "epoch": 0.32565284178187404,
      "grad_norm": 4.102039337158203,
      "kl": 0.2353515625,
      "learning_rate": 9.192307692307692e-07,
      "loss": -0.1276,
      "reward": 0.5773871541023254,
      "reward_std": 0.12897293269634247,
      "reward_std_all": 1.2173874378204346,
      "rewards-max/forensics_sinkhorn_reward": 0.7433816492557526,
      "rewards-max/forensics_strict_edge_reward": 0.005711474797574834,
      "rewards-max/forensics_tiou_reward": 0.610978364944458,
      "rewards/forensics_sinkhorn_reward": 0.43469342589378357,
      "rewards/forensics_strict_edge_reward": 0.00027931609204945307,
      "rewards/forensics_tiou_reward": 0.14241443574428558,
      "step": 106
    },
    {
      "completion_length": 27.34375,
      "epoch": 0.32872503840245776,
      "grad_norm": 5.087400913238525,
      "kl": 0.21630859375,
      "learning_rate": 9.184615384615384e-07,
      "loss": -0.0218,
      "reward": 0.6141737401485443,
      "reward_std": 0.2844928652048111,
      "reward_std_all": 1.2361814379692078,
      "rewards-max/forensics_sinkhorn_reward": 0.9475991725921631,
      "rewards-max/forensics_strict_edge_reward": 0.14197011478245258,
      "rewards-max/forensics_tiou_reward": 0.8854624032974243,
      "rewards/forensics_sinkhorn_reward": 0.39839673042297363,
      "rewards/forensics_strict_edge_reward": 0.007611550739966333,
      "rewards/forensics_tiou_reward": 0.20816544443368912,
      "step": 107
    },
    {
      "completion_length": 23.34375,
      "epoch": 0.3317972350230415,
      "grad_norm": 4.671716690063477,
      "kl": 0.2392578125,
      "learning_rate": 9.176923076923076e-07,
      "loss": -0.1124,
      "reward": 0.7711265981197357,
      "reward_std": 0.28358612209558487,
      "reward_std_all": 1.1671332120895386,
      "rewards-max/forensics_sinkhorn_reward": 0.9137777388095856,
      "rewards-max/forensics_strict_edge_reward": 0.2571620183880441,
      "rewards-max/forensics_tiou_reward": 0.8154164254665375,
      "rewards/forensics_sinkhorn_reward": 0.5065692961215973,
      "rewards/forensics_strict_edge_reward": 0.019152701668645022,
      "rewards/forensics_tiou_reward": 0.24540463089942932,
      "step": 108
    },
    {
      "completion_length": 26.40625,
      "epoch": 0.3348694316436252,
      "grad_norm": 3.9532597064971924,
      "kl": 0.22021484375,
      "learning_rate": 9.169230769230769e-07,
      "loss": -0.1093,
      "reward": 0.5660233497619629,
      "reward_std": 0.23436468839645386,
      "reward_std_all": 1.2492684125900269,
      "rewards-max/forensics_sinkhorn_reward": 0.9176590740680695,
      "rewards-max/forensics_strict_edge_reward": 0.00013073396144136495,
      "rewards-max/forensics_tiou_reward": 0.8051052987575531,
      "rewards/forensics_sinkhorn_reward": 0.41290557384490967,
      "rewards/forensics_strict_edge_reward": 6.453820986962455e-06,
      "rewards/forensics_tiou_reward": 0.1531113013625145,
      "step": 109
    },
    {
      "completion_length": 24.96875,
      "epoch": 0.3379416282642089,
      "grad_norm": 3.989546298980713,
      "kl": 0.1982421875,
      "learning_rate": 9.161538461538462e-07,
      "loss": -0.0614,
      "reward": 0.6140117347240448,
      "reward_std": 0.15244388952851295,
      "reward_std_all": 1.199021339416504,
      "rewards-max/forensics_sinkhorn_reward": 0.7351812124252319,
      "rewards-max/forensics_strict_edge_reward": 0.045123110479835304,
      "rewards-max/forensics_tiou_reward": 0.5072305798530579,
      "rewards/forensics_sinkhorn_reward": 0.46241407096385956,
      "rewards/forensics_strict_edge_reward": 0.0018807978151471616,
      "rewards/forensics_tiou_reward": 0.14971685037016869,
      "step": 110
    },
    {
      "completion_length": 23.796875,
      "epoch": 0.34101382488479265,
      "grad_norm": 4.441440105438232,
      "kl": 0.203125,
      "learning_rate": 9.153846153846153e-07,
      "loss": -0.0688,
      "reward": 0.6723786294460297,
      "reward_std": 0.260866641998291,
      "reward_std_all": 1.2067184448242188,
      "rewards-max/forensics_sinkhorn_reward": 0.9740096926689148,
      "rewards-max/forensics_strict_edge_reward": 0.4358583316206932,
      "rewards-max/forensics_tiou_reward": 0.9589051306247711,
      "rewards/forensics_sinkhorn_reward": 0.44551587104797363,
      "rewards/forensics_strict_edge_reward": 0.0182620610576123,
      "rewards/forensics_tiou_reward": 0.208600752055645,
      "step": 111
    },
    {
      "completion_length": 25.875,
      "epoch": 0.34408602150537637,
      "grad_norm": 5.705254077911377,
      "kl": 0.36083984375,
      "learning_rate": 9.146153846153846e-07,
      "loss": -0.0823,
      "reward": 0.6977436244487762,
      "reward_std": 0.15953315049409866,
      "reward_std_all": 1.1643383502960205,
      "rewards-max/forensics_sinkhorn_reward": 0.8516969382762909,
      "rewards-max/forensics_strict_edge_reward": 0.020685063005657867,
      "rewards-max/forensics_tiou_reward": 0.7212970852851868,
      "rewards/forensics_sinkhorn_reward": 0.47679632902145386,
      "rewards/forensics_strict_edge_reward": 0.0008697938828845508,
      "rewards/forensics_tiou_reward": 0.2200774922966957,
      "step": 112
    },
    {
      "completion_length": 26.140625,
      "epoch": 0.34715821812596004,
      "grad_norm": 3.926574945449829,
      "kl": 0.2080078125,
      "learning_rate": 9.138461538461538e-07,
      "loss": -0.0674,
      "reward": 0.6302379965782166,
      "reward_std": 0.22718621045351028,
      "reward_std_all": 1.2207489609718323,
      "rewards-max/forensics_sinkhorn_reward": 0.8581801652908325,
      "rewards-max/forensics_strict_edge_reward": 0.4094455089289113,
      "rewards-max/forensics_tiou_reward": 0.755630224943161,
      "rewards/forensics_sinkhorn_reward": 0.4201083481311798,
      "rewards/forensics_strict_edge_reward": 0.01785081219532003,
      "rewards/forensics_tiou_reward": 0.19227878749370575,
      "step": 113
    },
    {
      "completion_length": 25.90625,
      "epoch": 0.35023041474654376,
      "grad_norm": 4.39113187789917,
      "kl": 0.24169921875,
      "learning_rate": 9.130769230769231e-07,
      "loss": -0.1516,
      "reward": 0.7138338088989258,
      "reward_std": 0.263514868915081,
      "reward_std_all": 1.1858415007591248,
      "rewards-max/forensics_sinkhorn_reward": 0.9244360327720642,
      "rewards-max/forensics_strict_edge_reward": 0.04019418265670538,
      "rewards-max/forensics_tiou_reward": 0.8358617424964905,
      "rewards/forensics_sinkhorn_reward": 0.48716744780540466,
      "rewards/forensics_strict_edge_reward": 0.002057390141999349,
      "rewards/forensics_tiou_reward": 0.22460899502038956,
      "step": 114
    },
    {
      "completion_length": 26.859375,
      "epoch": 0.3533026113671275,
      "grad_norm": 4.614264965057373,
      "kl": 0.18896484375,
      "learning_rate": 9.123076923076922e-07,
      "loss": -0.0888,
      "reward": 0.6216209828853607,
      "reward_std": 0.21419671177864075,
      "reward_std_all": 1.2090017795562744,
      "rewards-max/forensics_sinkhorn_reward": 0.7751968801021576,
      "rewards-max/forensics_strict_edge_reward": 0.02441011805422022,
      "rewards-max/forensics_tiou_reward": 0.5763607025146484,
      "rewards/forensics_sinkhorn_reward": 0.4574781358242035,
      "rewards/forensics_strict_edge_reward": 0.0010241940883446432,
      "rewards/forensics_tiou_reward": 0.1631186679005623,
      "step": 115
    },
    {
      "completion_length": 26.359375,
      "epoch": 0.3563748079877112,
      "grad_norm": 3.7256388664245605,
      "kl": 0.22021484375,
      "learning_rate": 9.115384615384614e-07,
      "loss": -0.1138,
      "reward": 0.655228316783905,
      "reward_std": 0.25672728568315506,
      "reward_std_all": 1.2094014286994934,
      "rewards-max/forensics_sinkhorn_reward": 0.9685737788677216,
      "rewards-max/forensics_strict_edge_reward": 0.07002133475180017,
      "rewards-max/forensics_tiou_reward": 0.9459256827831268,
      "rewards/forensics_sinkhorn_reward": 0.43289709091186523,
      "rewards/forensics_strict_edge_reward": 0.005424956703791395,
      "rewards/forensics_tiou_reward": 0.21690621972084045,
      "step": 116
    },
    {
      "completion_length": 24.046875,
      "epoch": 0.35944700460829493,
      "grad_norm": 5.3866472244262695,
      "kl": 0.3505859375,
      "learning_rate": 9.107692307692308e-07,
      "loss": -0.1541,
      "reward": 0.6104428470134735,
      "reward_std": 0.30164284259080887,
      "reward_std_all": 1.240394949913025,
      "rewards-max/forensics_sinkhorn_reward": 0.9516716599464417,
      "rewards-max/forensics_strict_edge_reward": 0.10545190516859293,
      "rewards-max/forensics_tiou_reward": 0.8907947242259979,
      "rewards/forensics_sinkhorn_reward": 0.4375806897878647,
      "rewards/forensics_strict_edge_reward": 0.0052837703842669725,
      "rewards/forensics_tiou_reward": 0.16757835447788239,
      "step": 117
    },
    {
      "completion_length": 28.53125,
      "epoch": 0.36251920122887865,
      "grad_norm": 3.7808761596679688,
      "kl": 0.29833984375,
      "learning_rate": 9.1e-07,
      "loss": -0.1205,
      "reward": 0.7343905568122864,
      "reward_std": 0.23047292232513428,
      "reward_std_all": 1.1617873311042786,
      "rewards-max/forensics_sinkhorn_reward": 0.9362702369689941,
      "rewards-max/forensics_strict_edge_reward": 0.1408906214783201,
      "rewards-max/forensics_tiou_reward": 0.8690129816532135,
      "rewards/forensics_sinkhorn_reward": 0.4848071038722992,
      "rewards/forensics_strict_edge_reward": 0.006177390176162589,
      "rewards/forensics_tiou_reward": 0.24340605735778809,
      "step": 118
    },
    {
      "completion_length": 23.53125,
      "epoch": 0.3655913978494624,
      "grad_norm": 4.140354156494141,
      "kl": 0.2548828125,
      "learning_rate": 9.092307692307692e-07,
      "loss": -0.0824,
      "reward": 0.7565351128578186,
      "reward_std": 0.2819647789001465,
      "reward_std_all": 1.1609025597572327,
      "rewards-max/forensics_sinkhorn_reward": 0.9137576222419739,
      "rewards-max/forensics_strict_edge_reward": 0.28474223241209984,
      "rewards-max/forensics_tiou_reward": 0.8228673040866852,
      "rewards/forensics_sinkhorn_reward": 0.5092705190181732,
      "rewards/forensics_strict_edge_reward": 0.012661856482736766,
      "rewards/forensics_tiou_reward": 0.23460276424884796,
      "step": 119
    },
    {
      "completion_length": 25.90625,
      "epoch": 0.3686635944700461,
      "grad_norm": 4.23823881149292,
      "kl": 0.20556640625,
      "learning_rate": 9.084615384615384e-07,
      "loss": -0.0978,
      "reward": 0.5323606133460999,
      "reward_std": 0.23109053075313568,
      "reward_std_all": 1.2634700536727905,
      "rewards-max/forensics_sinkhorn_reward": 0.933284342288971,
      "rewards-max/forensics_strict_edge_reward": 0.03089575603371486,
      "rewards-max/forensics_tiou_reward": 0.8526354432106018,
      "rewards/forensics_sinkhorn_reward": 0.36401642858982086,
      "rewards/forensics_strict_edge_reward": 0.0014309535381471505,
      "rewards/forensics_tiou_reward": 0.166913241147995,
      "step": 120
    },
    {
      "completion_length": 26.484375,
      "epoch": 0.3717357910906298,
      "grad_norm": 4.441474437713623,
      "kl": 0.25390625,
      "learning_rate": 9.076923076923076e-07,
      "loss": -0.0899,
      "reward": 0.5298372209072113,
      "reward_std": 0.23232832551002502,
      "reward_std_all": 1.2671997547149658,
      "rewards-max/forensics_sinkhorn_reward": 0.9251623749732971,
      "rewards-max/forensics_strict_edge_reward": 0.19539857283234596,
      "rewards-max/forensics_tiou_reward": 0.8509188890457153,
      "rewards/forensics_sinkhorn_reward": 0.3740623891353607,
      "rewards/forensics_strict_edge_reward": 0.008145933272317052,
      "rewards/forensics_tiou_reward": 0.14762889593839645,
      "step": 121
    },
    {
      "completion_length": 23.71875,
      "epoch": 0.37480798771121354,
      "grad_norm": 3.9316999912261963,
      "kl": 0.19580078125,
      "learning_rate": 9.069230769230769e-07,
      "loss": -0.1152,
      "reward": 0.808172881603241,
      "reward_std": 0.2862129211425781,
      "reward_std_all": 1.1403716206550598,
      "rewards-max/forensics_sinkhorn_reward": 0.8814508616924286,
      "rewards-max/forensics_strict_edge_reward": 0.25674792195059126,
      "rewards-max/forensics_tiou_reward": 0.7629342377185822,
      "rewards/forensics_sinkhorn_reward": 0.5247374325990677,
      "rewards/forensics_strict_edge_reward": 0.013506150037756015,
      "rewards/forensics_tiou_reward": 0.26992934197187424,
      "step": 122
    },
    {
      "completion_length": 26.609375,
      "epoch": 0.3778801843317972,
      "grad_norm": 4.229092597961426,
      "kl": 0.2607421875,
      "learning_rate": 9.061538461538461e-07,
      "loss": -0.1037,
      "reward": 0.6505668461322784,
      "reward_std": 0.23197831213474274,
      "reward_std_all": 1.2035942673683167,
      "rewards-max/forensics_sinkhorn_reward": 0.9390270709991455,
      "rewards-max/forensics_strict_edge_reward": 0.024143346287019085,
      "rewards-max/forensics_tiou_reward": 0.8823477923870087,
      "rewards/forensics_sinkhorn_reward": 0.46033112704753876,
      "rewards/forensics_strict_edge_reward": 0.001008838818052027,
      "rewards/forensics_tiou_reward": 0.18922685831785202,
      "step": 123
    },
    {
      "completion_length": 24.703125,
      "epoch": 0.38095238095238093,
      "grad_norm": 4.0178608894348145,
      "kl": 0.2216796875,
      "learning_rate": 9.053846153846153e-07,
      "loss": -0.0546,
      "reward": 0.6731729805469513,
      "reward_std": 0.28371645510196686,
      "reward_std_all": 1.2133587002754211,
      "rewards-max/forensics_sinkhorn_reward": 0.9447854459285736,
      "rewards-max/forensics_strict_edge_reward": 0.287186574190855,
      "rewards-max/forensics_tiou_reward": 0.8940654695034027,
      "rewards/forensics_sinkhorn_reward": 0.4285302758216858,
      "rewards/forensics_strict_edge_reward": 0.02279897767584771,
      "rewards/forensics_tiou_reward": 0.22184374928474426,
      "step": 124
    },
    {
      "completion_length": 29.203125,
      "epoch": 0.38402457757296465,
      "grad_norm": 3.706345796585083,
      "kl": 0.26220703125,
      "learning_rate": 9.046153846153846e-07,
      "loss": -0.1087,
      "reward": 0.6572085618972778,
      "reward_std": 0.350699320435524,
      "reward_std_all": 1.2358711957931519,
      "rewards-max/forensics_sinkhorn_reward": 0.9565605223178864,
      "rewards-max/forensics_strict_edge_reward": 0.15988703072071075,
      "rewards-max/forensics_tiou_reward": 0.9123554527759552,
      "rewards/forensics_sinkhorn_reward": 0.4336799830198288,
      "rewards/forensics_strict_edge_reward": 0.006689268629997969,
      "rewards/forensics_tiou_reward": 0.21683929860591888,
      "step": 125
    },
    {
      "completion_length": 22.484375,
      "epoch": 0.3870967741935484,
      "grad_norm": 4.5467848777771,
      "kl": 0.3125,
      "learning_rate": 9.038461538461538e-07,
      "loss": -0.1295,
      "reward": 0.558679610490799,
      "reward_std": 0.23126907646656036,
      "reward_std_all": 1.2455057501792908,
      "rewards-max/forensics_sinkhorn_reward": 0.9144828617572784,
      "rewards-max/forensics_strict_edge_reward": 0.09039061213843524,
      "rewards-max/forensics_tiou_reward": 0.8049405515193939,
      "rewards/forensics_sinkhorn_reward": 0.4198473393917084,
      "rewards/forensics_strict_edge_reward": 0.0037752377102151513,
      "rewards/forensics_tiou_reward": 0.1350570097565651,
      "step": 126
    },
    {
      "completion_length": 23.25,
      "epoch": 0.3901689708141321,
      "grad_norm": 3.6252903938293457,
      "kl": 0.302734375,
      "learning_rate": 9.030769230769231e-07,
      "loss": -0.0913,
      "reward": 0.6887175738811493,
      "reward_std": 0.24768107384443283,
      "reward_std_all": 1.1843312978744507,
      "rewards-max/forensics_sinkhorn_reward": 0.8714476227760315,
      "rewards-max/forensics_strict_edge_reward": 0.03265289713453967,
      "rewards-max/forensics_tiou_reward": 0.6945176124572754,
      "rewards/forensics_sinkhorn_reward": 0.49755948781967163,
      "rewards/forensics_strict_edge_reward": 0.002556734521931503,
      "rewards/forensics_tiou_reward": 0.1886013299226761,
      "step": 127
    },
    {
      "completion_length": 23.40625,
      "epoch": 0.3932411674347158,
      "grad_norm": 4.549530029296875,
      "kl": 0.2783203125,
      "learning_rate": 9.023076923076922e-07,
      "loss": -0.0628,
      "reward": 0.8235666155815125,
      "reward_std": 0.31153377890586853,
      "reward_std_all": 1.145847201347351,
      "rewards-max/forensics_sinkhorn_reward": 0.9706829190254211,
      "rewards-max/forensics_strict_edge_reward": 0.5070041641592979,
      "rewards-max/forensics_tiou_reward": 0.9541267156600952,
      "rewards/forensics_sinkhorn_reward": 0.5466060936450958,
      "rewards/forensics_strict_edge_reward": 0.021717559546232224,
      "rewards/forensics_tiou_reward": 0.2552429363131523,
      "step": 128
    },
    {
      "completion_length": 22.53125,
      "epoch": 0.39631336405529954,
      "grad_norm": 4.342728614807129,
      "kl": 0.3935546875,
      "learning_rate": 9.015384615384614e-07,
      "loss": -0.1198,
      "reward": 0.7066934704780579,
      "reward_std": 0.19981276243925095,
      "reward_std_all": 1.1776378154754639,
      "rewards-max/forensics_sinkhorn_reward": 0.9350731670856476,
      "rewards-max/forensics_strict_edge_reward": 0.23188477009534836,
      "rewards-max/forensics_tiou_reward": 0.873697966337204,
      "rewards/forensics_sinkhorn_reward": 0.48523950576782227,
      "rewards/forensics_strict_edge_reward": 0.00970132602378726,
      "rewards/forensics_tiou_reward": 0.21175266057252884,
      "step": 129
    },
    {
      "completion_length": 22.609375,
      "epoch": 0.39938556067588327,
      "grad_norm": 4.8860321044921875,
      "kl": 0.24951171875,
      "learning_rate": 9.007692307692307e-07,
      "loss": -0.1448,
      "reward": 0.6850240230560303,
      "reward_std": 0.36039167642593384,
      "reward_std_all": 1.2247260808944702,
      "rewards-max/forensics_sinkhorn_reward": 0.9476099908351898,
      "rewards-max/forensics_strict_edge_reward": 0.44547153264284134,
      "rewards-max/forensics_tiou_reward": 0.8987785875797272,
      "rewards/forensics_sinkhorn_reward": 0.4688042104244232,
      "rewards/forensics_strict_edge_reward": 0.02175949001684785,
      "rewards/forensics_tiou_reward": 0.19446032494306564,
      "step": 130
    },
    {
      "completion_length": 25.90625,
      "epoch": 0.402457757296467,
      "grad_norm": 4.645657539367676,
      "kl": 0.27978515625,
      "learning_rate": 9e-07,
      "loss": -0.0659,
      "reward": 0.7648924887180328,
      "reward_std": 0.17251989245414734,
      "reward_std_all": 1.1300933957099915,
      "rewards-max/forensics_sinkhorn_reward": 0.869452714920044,
      "rewards-max/forensics_strict_edge_reward": 0.008634242927655578,
      "rewards-max/forensics_tiou_reward": 0.7176218330860138,
      "rewards/forensics_sinkhorn_reward": 0.5243403315544128,
      "rewards/forensics_strict_edge_reward": 0.0005507011373993009,
      "rewards/forensics_tiou_reward": 0.24000146985054016,
      "step": 131
    },
    {
      "completion_length": 26.734375,
      "epoch": 0.4055299539170507,
      "grad_norm": 5.257702827453613,
      "kl": 0.24169921875,
      "learning_rate": 8.992307692307693e-07,
      "loss": -0.0956,
      "reward": 0.5480955392122269,
      "reward_std": 0.3008357286453247,
      "reward_std_all": 1.2780286073684692,
      "rewards-max/forensics_sinkhorn_reward": 0.8195546865463257,
      "rewards-max/forensics_strict_edge_reward": 0.19662752782869575,
      "rewards-max/forensics_tiou_reward": 0.6258113086223602,
      "rewards/forensics_sinkhorn_reward": 0.3730577230453491,
      "rewards/forensics_strict_edge_reward": 0.01369680380773275,
      "rewards/forensics_tiou_reward": 0.16134104877710342,
      "step": 132
    },
    {
      "completion_length": 27.0625,
      "epoch": 0.40860215053763443,
      "grad_norm": 3.8903565406799316,
      "kl": 0.19677734375,
      "learning_rate": 8.984615384615384e-07,
      "loss": -0.1552,
      "reward": 0.6894303560256958,
      "reward_std": 0.28725358843803406,
      "reward_std_all": 1.197732925415039,
      "rewards-max/forensics_sinkhorn_reward": 0.9262080490589142,
      "rewards-max/forensics_strict_edge_reward": 0.1843648375943303,
      "rewards-max/forensics_tiou_reward": 0.8472641706466675,
      "rewards/forensics_sinkhorn_reward": 0.4486805647611618,
      "rewards/forensics_strict_edge_reward": 0.013436253124382347,
      "rewards/forensics_tiou_reward": 0.22731353342533112,
      "step": 133
    },
    {
      "completion_length": 28.0625,
      "epoch": 0.4116743471582181,
      "grad_norm": 4.61454439163208,
      "kl": 0.248046875,
      "learning_rate": 8.976923076923076e-07,
      "loss": -0.1238,
      "reward": 0.6233134269714355,
      "reward_std": 0.19316386058926582,
      "reward_std_all": 1.2042524218559265,
      "rewards-max/forensics_sinkhorn_reward": 0.8293669521808624,
      "rewards-max/forensics_strict_edge_reward": 0.2099164631217718,
      "rewards-max/forensics_tiou_reward": 0.6827705353498459,
      "rewards/forensics_sinkhorn_reward": 0.40581974387168884,
      "rewards/forensics_strict_edge_reward": 0.009778927662409842,
      "rewards/forensics_tiou_reward": 0.20771470665931702,
      "step": 134
    },
    {
      "completion_length": 23.828125,
      "epoch": 0.4147465437788018,
      "grad_norm": 5.28393030166626,
      "kl": 0.2509765625,
      "learning_rate": 8.969230769230769e-07,
      "loss": -0.1498,
      "reward": 0.7499740719795227,
      "reward_std": 0.2955710291862488,
      "reward_std_all": 1.1828174591064453,
      "rewards-max/forensics_sinkhorn_reward": 0.9640179574489594,
      "rewards-max/forensics_strict_edge_reward": 0.2861167974770069,
      "rewards-max/forensics_tiou_reward": 0.9280403256416321,
      "rewards/forensics_sinkhorn_reward": 0.4750156104564667,
      "rewards/forensics_strict_edge_reward": 0.02143721270840615,
      "rewards/forensics_tiou_reward": 0.253521203994751,
      "step": 135
    },
    {
      "completion_length": 29.109375,
      "epoch": 0.41781874039938555,
      "grad_norm": 4.326157093048096,
      "kl": 0.23486328125,
      "learning_rate": 8.961538461538461e-07,
      "loss": -0.1815,
      "reward": 0.8516052067279816,
      "reward_std": 0.5313636958599091,
      "reward_std_all": 1.1990684270858765,
      "rewards-max/forensics_sinkhorn_reward": 0.9708117842674255,
      "rewards-max/forensics_strict_edge_reward": 0.3400110676884651,
      "rewards-max/forensics_tiou_reward": 0.9525156319141388,
      "rewards/forensics_sinkhorn_reward": 0.49324357509613037,
      "rewards/forensics_strict_edge_reward": 0.03197515895590186,
      "rewards/forensics_tiou_reward": 0.3263864517211914,
      "step": 136
    },
    {
      "completion_length": 26.296875,
      "epoch": 0.42089093701996927,
      "grad_norm": 4.67550802230835,
      "kl": 0.291015625,
      "learning_rate": 8.953846153846152e-07,
      "loss": -0.1701,
      "reward": 0.7202610969543457,
      "reward_std": 0.4331205487251282,
      "reward_std_all": 1.2165712714195251,
      "rewards-max/forensics_sinkhorn_reward": 0.9525524973869324,
      "rewards-max/forensics_strict_edge_reward": 0.07824757974594831,
      "rewards-max/forensics_tiou_reward": 0.894849956035614,
      "rewards/forensics_sinkhorn_reward": 0.46531669795513153,
      "rewards/forensics_strict_edge_reward": 0.003636801615357399,
      "rewards/forensics_tiou_reward": 0.25130754709243774,
      "step": 137
    },
    {
      "completion_length": 24.921875,
      "epoch": 0.423963133640553,
      "grad_norm": 4.528850555419922,
      "kl": 0.26171875,
      "learning_rate": 8.946153846153846e-07,
      "loss": -0.0901,
      "reward": 0.5662120878696442,
      "reward_std": 0.2912634015083313,
      "reward_std_all": 1.2653028964996338,
      "rewards-max/forensics_sinkhorn_reward": 0.9808138012886047,
      "rewards-max/forensics_strict_edge_reward": 0.18540387495886534,
      "rewards-max/forensics_tiou_reward": 0.9737517237663269,
      "rewards/forensics_sinkhorn_reward": 0.4004177451133728,
      "rewards/forensics_strict_edge_reward": 0.007819157333869953,
      "rewards/forensics_tiou_reward": 0.1579752415418625,
      "step": 138
    },
    {
      "completion_length": 26.140625,
      "epoch": 0.4270353302611367,
      "grad_norm": 3.8712735176086426,
      "kl": 0.279296875,
      "learning_rate": 8.938461538461538e-07,
      "loss": -0.103,
      "reward": 0.7491935789585114,
      "reward_std": 0.3358059227466583,
      "reward_std_all": 1.1862850785255432,
      "rewards-max/forensics_sinkhorn_reward": 0.9752895534038544,
      "rewards-max/forensics_strict_edge_reward": 0.14995709992945194,
      "rewards-max/forensics_tiou_reward": 0.9491800367832184,
      "rewards/forensics_sinkhorn_reward": 0.4859208017587662,
      "rewards/forensics_strict_edge_reward": 0.00861624738899991,
      "rewards/forensics_tiou_reward": 0.25465652346611023,
      "step": 139
    },
    {
      "completion_length": 25.96875,
      "epoch": 0.43010752688172044,
      "grad_norm": 4.9785332679748535,
      "kl": 0.205078125,
      "learning_rate": 8.930769230769231e-07,
      "loss": -0.1165,
      "reward": 0.6815605163574219,
      "reward_std": 0.22387675195932388,
      "reward_std_all": 1.1923640966415405,
      "rewards-max/forensics_sinkhorn_reward": 0.7892767190933228,
      "rewards-max/forensics_strict_edge_reward": 0.03480181814302341,
      "rewards-max/forensics_tiou_reward": 0.622837170958519,
      "rewards/forensics_sinkhorn_reward": 0.46345245838165283,
      "rewards/forensics_strict_edge_reward": 0.0034059047993650893,
      "rewards/forensics_tiou_reward": 0.21470215916633606,
      "step": 140
    },
    {
      "completion_length": 25.890625,
      "epoch": 0.43317972350230416,
      "grad_norm": 4.510084629058838,
      "kl": 0.26611328125,
      "learning_rate": 8.923076923076923e-07,
      "loss": -0.103,
      "reward": 0.7887702584266663,
      "reward_std": 0.39596162736415863,
      "reward_std_all": 1.1897811889648438,
      "rewards-max/forensics_sinkhorn_reward": 0.9694201350212097,
      "rewards-max/forensics_strict_edge_reward": 0.43308156356215477,
      "rewards-max/forensics_tiou_reward": 0.9353354871273041,
      "rewards/forensics_sinkhorn_reward": 0.4645799398422241,
      "rewards/forensics_strict_edge_reward": 0.02513998094946146,
      "rewards/forensics_tiou_reward": 0.29905037581920624,
      "step": 141
    },
    {
      "completion_length": 25.9375,
      "epoch": 0.4362519201228879,
      "grad_norm": 4.4788618087768555,
      "kl": 0.2958984375,
      "learning_rate": 8.915384615384614e-07,
      "loss": -0.1412,
      "reward": 0.8809260427951813,
      "reward_std": 0.21952469646930695,
      "reward_std_all": 1.085250973701477,
      "rewards-max/forensics_sinkhorn_reward": 0.9237326681613922,
      "rewards-max/forensics_strict_edge_reward": 0.20259709702804685,
      "rewards-max/forensics_tiou_reward": 0.7914251387119293,
      "rewards/forensics_sinkhorn_reward": 0.5811861753463745,
      "rewards/forensics_strict_edge_reward": 0.008469457185128704,
      "rewards/forensics_tiou_reward": 0.2912704348564148,
      "step": 142
    },
    {
      "completion_length": 22.96875,
      "epoch": 0.4393241167434716,
      "grad_norm": 3.6419854164123535,
      "kl": 0.2451171875,
      "learning_rate": 8.907692307692307e-07,
      "loss": -0.084,
      "reward": 0.8031800091266632,
      "reward_std": 0.25683191418647766,
      "reward_std_all": 1.1362384557724,
      "rewards-max/forensics_sinkhorn_reward": 0.8707898259162903,
      "rewards-max/forensics_strict_edge_reward": 0.20560714229941368,
      "rewards-max/forensics_tiou_reward": 0.7448304295539856,
      "rewards/forensics_sinkhorn_reward": 0.5557458400726318,
      "rewards/forensics_strict_edge_reward": 0.012835367815569043,
      "rewards/forensics_tiou_reward": 0.2345988154411316,
      "step": 143
    },
    {
      "completion_length": 24.546875,
      "epoch": 0.4423963133640553,
      "grad_norm": 4.264729022979736,
      "kl": 0.2470703125,
      "learning_rate": 8.9e-07,
      "loss": -0.1153,
      "reward": 0.7047838270664215,
      "reward_std": 0.19484104961156845,
      "reward_std_all": 1.1630414724349976,
      "rewards-max/forensics_sinkhorn_reward": 0.883504718542099,
      "rewards-max/forensics_strict_edge_reward": 0.008557238426874392,
      "rewards-max/forensics_tiou_reward": 0.707074910402298,
      "rewards/forensics_sinkhorn_reward": 0.49969956278800964,
      "rewards/forensics_strict_edge_reward": 0.00036401387387741124,
      "rewards/forensics_tiou_reward": 0.20472022145986557,
      "step": 144
    },
    {
      "completion_length": 25.1875,
      "epoch": 0.445468509984639,
      "grad_norm": 4.550893783569336,
      "kl": 0.3232421875,
      "learning_rate": 8.892307692307693e-07,
      "loss": 0.0131,
      "reward": 0.7006351053714752,
      "reward_std": 0.24981867522001266,
      "reward_std_all": 1.1815746426582336,
      "rewards-max/forensics_sinkhorn_reward": 0.8565356731414795,
      "rewards-max/forensics_strict_edge_reward": 0.0478106327354908,
      "rewards-max/forensics_tiou_reward": 0.745545506477356,
      "rewards/forensics_sinkhorn_reward": 0.46692174673080444,
      "rewards/forensics_strict_edge_reward": 0.0020397239131852984,
      "rewards/forensics_tiou_reward": 0.2316737025976181,
      "step": 145
    },
    {
      "completion_length": 24.671875,
      "epoch": 0.4485407066052227,
      "grad_norm": 3.867794990539551,
      "kl": 0.25390625,
      "learning_rate": 8.884615384615384e-07,
      "loss": -0.1265,
      "reward": 0.7655594944953918,
      "reward_std": 0.1881629079580307,
      "reward_std_all": 1.1383123397827148,
      "rewards-max/forensics_sinkhorn_reward": 0.9398943781852722,
      "rewards-max/forensics_strict_edge_reward": 0.08819708600640297,
      "rewards-max/forensics_tiou_reward": 0.8819754123687744,
      "rewards/forensics_sinkhorn_reward": 0.5140043795108795,
      "rewards/forensics_strict_edge_reward": 0.004774587927386165,
      "rewards/forensics_tiou_reward": 0.24678055942058563,
      "step": 146
    },
    {
      "completion_length": 23.234375,
      "epoch": 0.45161290322580644,
      "grad_norm": 5.168213367462158,
      "kl": 0.24755859375,
      "learning_rate": 8.876923076923076e-07,
      "loss": -0.0827,
      "reward": 0.7860512733459473,
      "reward_std": 0.26459167897701263,
      "reward_std_all": 1.1486444473266602,
      "rewards-max/forensics_sinkhorn_reward": 0.9617542028427124,
      "rewards-max/forensics_strict_edge_reward": 0.0946718882478308,
      "rewards-max/forensics_tiou_reward": 0.9210569560527802,
      "rewards/forensics_sinkhorn_reward": 0.5101306438446045,
      "rewards/forensics_strict_edge_reward": 0.007069918909110129,
      "rewards/forensics_tiou_reward": 0.2688506841659546,
      "step": 147
    },
    {
      "completion_length": 23.21875,
      "epoch": 0.45468509984639016,
      "grad_norm": 4.370171546936035,
      "kl": 0.2939453125,
      "learning_rate": 8.869230769230769e-07,
      "loss": -0.1274,
      "reward": 0.8476579189300537,
      "reward_std": 0.24577338993549347,
      "reward_std_all": 1.1177857518196106,
      "rewards-max/forensics_sinkhorn_reward": 0.9587924480438232,
      "rewards-max/forensics_strict_edge_reward": 0.2800660580396652,
      "rewards-max/forensics_tiou_reward": 0.9233255982398987,
      "rewards/forensics_sinkhorn_reward": 0.57341268658638,
      "rewards/forensics_strict_edge_reward": 0.013388250954449177,
      "rewards/forensics_tiou_reward": 0.2608569860458374,
      "step": 148
    },
    {
      "completion_length": 22.9375,
      "epoch": 0.4577572964669739,
      "grad_norm": 4.690672874450684,
      "kl": 0.26806640625,
      "learning_rate": 8.861538461538461e-07,
      "loss": -0.1279,
      "reward": 0.8598142862319946,
      "reward_std": 0.34978094696998596,
      "reward_std_all": 1.1294074058532715,
      "rewards-max/forensics_sinkhorn_reward": 0.9737042784690857,
      "rewards-max/forensics_strict_edge_reward": 0.12716589123010635,
      "rewards-max/forensics_tiou_reward": 0.9608586728572845,
      "rewards/forensics_sinkhorn_reward": 0.5719039440155029,
      "rewards/forensics_strict_edge_reward": 0.005693165818229318,
      "rewards/forensics_tiou_reward": 0.2822171747684479,
      "step": 149
    },
    {
      "completion_length": 24.5625,
      "epoch": 0.4608294930875576,
      "grad_norm": 3.7272210121154785,
      "kl": 0.2587890625,
      "learning_rate": 8.853846153846154e-07,
      "loss": -0.122,
      "reward": 0.8369947671890259,
      "reward_std": 0.27997250854969025,
      "reward_std_all": 1.1257892847061157,
      "rewards-max/forensics_sinkhorn_reward": 0.942573219537735,
      "rewards-max/forensics_strict_edge_reward": 0.10797113599255681,
      "rewards-max/forensics_tiou_reward": 0.87333545088768,
      "rewards/forensics_sinkhorn_reward": 0.5463328659534454,
      "rewards/forensics_strict_edge_reward": 0.00686380744446069,
      "rewards/forensics_tiou_reward": 0.28379808366298676,
      "step": 150
    },
    {
      "completion_length": 23.625,
      "epoch": 0.46390168970814133,
      "grad_norm": 4.803853511810303,
      "kl": 0.25927734375,
      "learning_rate": 8.846153846153846e-07,
      "loss": -0.0987,
      "reward": 0.7849233150482178,
      "reward_std": 0.16641337797045708,
      "reward_std_all": 1.1278470158576965,
      "rewards-max/forensics_sinkhorn_reward": 0.8585269451141357,
      "rewards-max/forensics_strict_edge_reward": 0.2654098246119716,
      "rewards-max/forensics_tiou_reward": 0.7249194383621216,
      "rewards/forensics_sinkhorn_reward": 0.5327029675245285,
      "rewards/forensics_strict_edge_reward": 0.016375591869354444,
      "rewards/forensics_tiou_reward": 0.23584473133087158,
      "step": 151
    },
    {
      "completion_length": 22.265625,
      "epoch": 0.46697388632872505,
      "grad_norm": 4.573609352111816,
      "kl": 0.236328125,
      "learning_rate": 8.838461538461538e-07,
      "loss": -0.1415,
      "reward": 0.7782382667064667,
      "reward_std": 0.28211213648319244,
      "reward_std_all": 1.1524839401245117,
      "rewards-max/forensics_sinkhorn_reward": 0.9148044586181641,
      "rewards-max/forensics_strict_edge_reward": 0.2102193998907751,
      "rewards-max/forensics_tiou_reward": 0.805196076631546,
      "rewards/forensics_sinkhorn_reward": 0.5385528802871704,
      "rewards/forensics_strict_edge_reward": 0.0094234581050614,
      "rewards/forensics_tiou_reward": 0.23026196658611298,
      "step": 152
    },
    {
      "completion_length": 26.28125,
      "epoch": 0.4700460829493088,
      "grad_norm": 4.7659149169921875,
      "kl": 0.18798828125,
      "learning_rate": 8.830769230769231e-07,
      "loss": -0.0907,
      "reward": 0.8633321225643158,
      "reward_std": 0.30235908180475235,
      "reward_std_all": 1.1163545846939087,
      "rewards-max/forensics_sinkhorn_reward": 0.9439549148082733,
      "rewards-max/forensics_strict_edge_reward": 0.11969014792703092,
      "rewards-max/forensics_tiou_reward": 0.8825271427631378,
      "rewards/forensics_sinkhorn_reward": 0.5436960607767105,
      "rewards/forensics_strict_edge_reward": 0.00739546704426175,
      "rewards/forensics_tiou_reward": 0.3122406527400017,
      "step": 153
    },
    {
      "completion_length": 23.109375,
      "epoch": 0.4731182795698925,
      "grad_norm": 4.768528938293457,
      "kl": 0.3056640625,
      "learning_rate": 8.823076923076923e-07,
      "loss": -0.1371,
      "reward": 0.8270591795444489,
      "reward_std": 0.2680944353342056,
      "reward_std_all": 1.128389596939087,
      "rewards-max/forensics_sinkhorn_reward": 0.9605436325073242,
      "rewards-max/forensics_strict_edge_reward": 0.23333654552698135,
      "rewards-max/forensics_tiou_reward": 0.9310926795005798,
      "rewards/forensics_sinkhorn_reward": 0.5441185384988785,
      "rewards/forensics_strict_edge_reward": 0.013521926943212748,
      "rewards/forensics_tiou_reward": 0.2694186717271805,
      "step": 154
    },
    {
      "completion_length": 23.84375,
      "epoch": 0.47619047619047616,
      "grad_norm": 4.870229721069336,
      "kl": 0.27099609375,
      "learning_rate": 8.815384615384614e-07,
      "loss": -0.0925,
      "reward": 0.9888716042041779,
      "reward_std": 0.3370710164308548,
      "reward_std_all": 1.0635263919830322,
      "rewards-max/forensics_sinkhorn_reward": 0.9450205862522125,
      "rewards-max/forensics_strict_edge_reward": 0.18580593541264534,
      "rewards-max/forensics_tiou_reward": 0.888910323381424,
      "rewards/forensics_sinkhorn_reward": 0.6192063987255096,
      "rewards/forensics_strict_edge_reward": 0.012224590871483088,
      "rewards/forensics_tiou_reward": 0.3574405312538147,
      "step": 155
    },
    {
      "completion_length": 23.40625,
      "epoch": 0.4792626728110599,
      "grad_norm": 4.192846775054932,
      "kl": 0.27392578125,
      "learning_rate": 8.807692307692307e-07,
      "loss": -0.101,
      "reward": 0.7799392938613892,
      "reward_std": 0.264115646481514,
      "reward_std_all": 1.1539208889007568,
      "rewards-max/forensics_sinkhorn_reward": 0.9339863657951355,
      "rewards-max/forensics_strict_edge_reward": 0.13262751325964928,
      "rewards-max/forensics_tiou_reward": 0.8651376962661743,
      "rewards/forensics_sinkhorn_reward": 0.4953411668539047,
      "rewards/forensics_strict_edge_reward": 0.005909523169975728,
      "rewards/forensics_tiou_reward": 0.2786886692047119,
      "step": 156
    },
    {
      "completion_length": 23.78125,
      "epoch": 0.4823348694316436,
      "grad_norm": 4.70831298828125,
      "kl": 0.2822265625,
      "learning_rate": 8.799999999999999e-07,
      "loss": -0.1074,
      "reward": 1.0144730806350708,
      "reward_std": 0.3548958748579025,
      "reward_std_all": 1.0638062357902527,
      "rewards-max/forensics_sinkhorn_reward": 0.9662930071353912,
      "rewards-max/forensics_strict_edge_reward": 0.45834553241729736,
      "rewards-max/forensics_tiou_reward": 0.942305326461792,
      "rewards/forensics_sinkhorn_reward": 0.605441689491272,
      "rewards/forensics_strict_edge_reward": 0.023247352801263332,
      "rewards/forensics_tiou_reward": 0.3857840597629547,
      "step": 157
    },
    {
      "completion_length": 26.3125,
      "epoch": 0.48540706605222733,
      "grad_norm": 4.837888717651367,
      "kl": 0.3115234375,
      "learning_rate": 8.792307692307693e-07,
      "loss": -0.1297,
      "reward": 0.8243967592716217,
      "reward_std": 0.2712562680244446,
      "reward_std_all": 1.1411787271499634,
      "rewards-max/forensics_sinkhorn_reward": 0.9471203684806824,
      "rewards-max/forensics_strict_edge_reward": 0.5062990635633469,
      "rewards-max/forensics_tiou_reward": 0.9026075601577759,
      "rewards/forensics_sinkhorn_reward": 0.5062156915664673,
      "rewards/forensics_strict_edge_reward": 0.03707176074385643,
      "rewards/forensics_tiou_reward": 0.2811092734336853,
      "step": 158
    },
    {
      "completion_length": 32.046875,
      "epoch": 0.48847926267281105,
      "grad_norm": 4.687170028686523,
      "kl": 0.3837890625,
      "learning_rate": 8.784615384615385e-07,
      "loss": -0.0767,
      "reward": 0.611161857843399,
      "reward_std": 0.15785662084817886,
      "reward_std_all": 1.2059077024459839,
      "rewards-max/forensics_sinkhorn_reward": 0.8693009316921234,
      "rewards-max/forensics_strict_edge_reward": 0.005594438684738634,
      "rewards-max/forensics_tiou_reward": 0.6792598366737366,
      "rewards/forensics_sinkhorn_reward": 0.43073245882987976,
      "rewards/forensics_strict_edge_reward": 0.00027755886874558655,
      "rewards/forensics_tiou_reward": 0.18015186488628387,
      "step": 159
    },
    {
      "completion_length": 24.3125,
      "epoch": 0.4915514592933948,
      "grad_norm": 4.3844170570373535,
      "kl": 0.25048828125,
      "learning_rate": 8.776923076923076e-07,
      "loss": -0.0993,
      "reward": 0.8218656480312347,
      "reward_std": 0.25580695271492004,
      "reward_std_all": 1.1221050024032593,
      "rewards-max/forensics_sinkhorn_reward": 0.9139907956123352,
      "rewards-max/forensics_strict_edge_reward": 0.20891199074685574,
      "rewards-max/forensics_tiou_reward": 0.8667020201683044,
      "rewards/forensics_sinkhorn_reward": 0.5284810364246368,
      "rewards/forensics_strict_edge_reward": 0.012193646631203592,
      "rewards/forensics_tiou_reward": 0.28119100630283356,
      "step": 160
    },
    {
      "completion_length": 22.5625,
      "epoch": 0.4946236559139785,
      "grad_norm": 4.26392126083374,
      "kl": 0.236328125,
      "learning_rate": 8.769230769230769e-07,
      "loss": -0.1076,
      "reward": 0.7785255610942841,
      "reward_std": 0.23114314675331116,
      "reward_std_all": 1.143733024597168,
      "rewards-max/forensics_sinkhorn_reward": 0.897028923034668,
      "rewards-max/forensics_strict_edge_reward": 0.3959572598505474,
      "rewards-max/forensics_tiou_reward": 0.7692686021327972,
      "rewards/forensics_sinkhorn_reward": 0.5178789496421814,
      "rewards/forensics_strict_edge_reward": 0.016842436888623524,
      "rewards/forensics_tiou_reward": 0.24380414187908173,
      "step": 161
    },
    {
      "completion_length": 29.9375,
      "epoch": 0.4976958525345622,
      "grad_norm": 5.108768939971924,
      "kl": 0.27734375,
      "learning_rate": 8.761538461538461e-07,
      "loss": -0.1091,
      "reward": 0.8285436034202576,
      "reward_std": 0.4062800407409668,
      "reward_std_all": 1.1762645244598389,
      "rewards-max/forensics_sinkhorn_reward": 0.9858494997024536,
      "rewards-max/forensics_strict_edge_reward": 0.8582656681537628,
      "rewards-max/forensics_tiou_reward": 0.9913457930088043,
      "rewards/forensics_sinkhorn_reward": 0.4584468603134155,
      "rewards/forensics_strict_edge_reward": 0.04062551632523537,
      "rewards/forensics_tiou_reward": 0.3294712156057358,
      "step": 162
    },
    {
      "completion_length": 24.84375,
      "epoch": 0.500768049155146,
      "grad_norm": 4.510804653167725,
      "kl": 0.267578125,
      "learning_rate": 8.753846153846154e-07,
      "loss": -0.1682,
      "reward": 0.705319344997406,
      "reward_std": 0.26246458292007446,
      "reward_std_all": 1.189061462879181,
      "rewards-max/forensics_sinkhorn_reward": 0.9276272654533386,
      "rewards-max/forensics_strict_edge_reward": 0.04439874645322561,
      "rewards-max/forensics_tiou_reward": 0.8433044850826263,
      "rewards/forensics_sinkhorn_reward": 0.47029952704906464,
      "rewards/forensics_strict_edge_reward": 0.0024909939093049616,
      "rewards/forensics_tiou_reward": 0.23252878338098526,
      "step": 163
    },
    {
      "completion_length": 27.390625,
      "epoch": 0.5038402457757296,
      "grad_norm": 5.026584148406982,
      "kl": 0.25,
      "learning_rate": 8.746153846153845e-07,
      "loss": -0.1325,
      "reward": 0.7074089050292969,
      "reward_std": 0.23965083807706833,
      "reward_std_all": 1.1783323287963867,
      "rewards-max/forensics_sinkhorn_reward": 0.8302134275436401,
      "rewards-max/forensics_strict_edge_reward": 0.23884377348555264,
      "rewards-max/forensics_tiou_reward": 0.7467916309833527,
      "rewards/forensics_sinkhorn_reward": 0.4840824753046036,
      "rewards/forensics_strict_edge_reward": 0.015687164128180095,
      "rewards/forensics_tiou_reward": 0.20763925462961197,
      "step": 164
    },
    {
      "completion_length": 25.5625,
      "epoch": 0.5069124423963134,
      "grad_norm": 4.593482494354248,
      "kl": 0.2568359375,
      "learning_rate": 8.738461538461538e-07,
      "loss": -0.104,
      "reward": 0.8611282706260681,
      "reward_std": 0.33256058394908905,
      "reward_std_all": 1.1367348432540894,
      "rewards-max/forensics_sinkhorn_reward": 0.9849580526351929,
      "rewards-max/forensics_strict_edge_reward": 0.3975098611554131,
      "rewards-max/forensics_tiou_reward": 0.986568033695221,
      "rewards/forensics_sinkhorn_reward": 0.5035146772861481,
      "rewards/forensics_strict_edge_reward": 0.03785982762929052,
      "rewards/forensics_tiou_reward": 0.3197537660598755,
      "step": 165
    },
    {
      "completion_length": 24.40625,
      "epoch": 0.5099846390168971,
      "grad_norm": 4.744661331176758,
      "kl": 0.2900390625,
      "learning_rate": 8.730769230769231e-07,
      "loss": -0.0866,
      "reward": 0.6947951912879944,
      "reward_std": 0.2358735352754593,
      "reward_std_all": 1.176601767539978,
      "rewards-max/forensics_sinkhorn_reward": 0.9067689180374146,
      "rewards-max/forensics_strict_edge_reward": 0.05088130757212639,
      "rewards-max/forensics_tiou_reward": 0.8067516982555389,
      "rewards/forensics_sinkhorn_reward": 0.4731053411960602,
      "rewards/forensics_strict_edge_reward": 0.002207943529356271,
      "rewards/forensics_tiou_reward": 0.2194819152355194,
      "step": 166
    },
    {
      "completion_length": 23.953125,
      "epoch": 0.5130568356374808,
      "grad_norm": 6.17315149307251,
      "kl": 0.5703125,
      "learning_rate": 8.723076923076923e-07,
      "loss": -0.1192,
      "reward": 0.9608145356178284,
      "reward_std": 0.3031836003065109,
      "reward_std_all": 1.0716734528541565,
      "rewards-max/forensics_sinkhorn_reward": 0.950868159532547,
      "rewards-max/forensics_strict_edge_reward": 0.3151310682296753,
      "rewards-max/forensics_tiou_reward": 0.8871904015541077,
      "rewards/forensics_sinkhorn_reward": 0.6054002642631531,
      "rewards/forensics_strict_edge_reward": 0.01609461335465312,
      "rewards/forensics_tiou_reward": 0.3393196761608124,
      "step": 167
    },
    {
      "completion_length": 25.734375,
      "epoch": 0.5161290322580645,
      "grad_norm": 4.442384243011475,
      "kl": 0.26171875,
      "learning_rate": 8.715384615384614e-07,
      "loss": -0.1362,
      "reward": 0.7973589897155762,
      "reward_std": 0.43166249990463257,
      "reward_std_all": 1.1968193650245667,
      "rewards-max/forensics_sinkhorn_reward": 0.9776023030281067,
      "rewards-max/forensics_strict_edge_reward": 0.5007071495056152,
      "rewards-max/forensics_tiou_reward": 0.9702623188495636,
      "rewards/forensics_sinkhorn_reward": 0.48149095475673676,
      "rewards/forensics_strict_edge_reward": 0.02231659647077322,
      "rewards/forensics_tiou_reward": 0.2935514897108078,
      "step": 168
    },
    {
      "completion_length": 26.75,
      "epoch": 0.5192012288786483,
      "grad_norm": 4.181142330169678,
      "kl": 0.291015625,
      "learning_rate": 8.707692307692307e-07,
      "loss": -0.1253,
      "reward": 0.7346883714199066,
      "reward_std": 0.3648436740040779,
      "reward_std_all": 1.1993877291679382,
      "rewards-max/forensics_sinkhorn_reward": 0.9359368681907654,
      "rewards-max/forensics_strict_edge_reward": 0.3117646835744381,
      "rewards-max/forensics_tiou_reward": 0.8723784983158112,
      "rewards/forensics_sinkhorn_reward": 0.4729645252227783,
      "rewards/forensics_strict_edge_reward": 0.014118625083938241,
      "rewards/forensics_tiou_reward": 0.24760521203279495,
      "step": 169
    },
    {
      "completion_length": 25.890625,
      "epoch": 0.522273425499232,
      "grad_norm": 4.284564971923828,
      "kl": 0.24755859375,
      "learning_rate": 8.699999999999999e-07,
      "loss": -0.0405,
      "reward": 0.7777348458766937,
      "reward_std": 0.24559679627418518,
      "reward_std_all": 1.1413964629173279,
      "rewards-max/forensics_sinkhorn_reward": 0.9013598561286926,
      "rewards-max/forensics_strict_edge_reward": 0.1420992836356163,
      "rewards-max/forensics_tiou_reward": 0.7692685425281525,
      "rewards/forensics_sinkhorn_reward": 0.5229098498821259,
      "rewards/forensics_strict_edge_reward": 0.0059658559039235115,
      "rewards/forensics_tiou_reward": 0.24885915964841843,
      "step": 170
    },
    {
      "completion_length": 28.234375,
      "epoch": 0.5253456221198156,
      "grad_norm": 4.065848350524902,
      "kl": 0.26953125,
      "learning_rate": 8.692307692307692e-07,
      "loss": -0.1519,
      "reward": 0.7234622836112976,
      "reward_std": 0.1886250525712967,
      "reward_std_all": 1.1649238467216492,
      "rewards-max/forensics_sinkhorn_reward": 0.9331470727920532,
      "rewards-max/forensics_strict_edge_reward": 0.13724735274445266,
      "rewards-max/forensics_tiou_reward": 0.8429512679576874,
      "rewards/forensics_sinkhorn_reward": 0.48518913984298706,
      "rewards/forensics_strict_edge_reward": 0.009346584112790879,
      "rewards/forensics_tiou_reward": 0.22892656177282333,
      "step": 171
    },
    {
      "completion_length": 24.078125,
      "epoch": 0.5284178187403994,
      "grad_norm": 4.883062839508057,
      "kl": 0.2724609375,
      "learning_rate": 8.684615384615385e-07,
      "loss": -0.0766,
      "reward": 0.9394369125366211,
      "reward_std": 0.26684580370783806,
      "reward_std_all": 1.0790588855743408,
      "rewards-max/forensics_sinkhorn_reward": 0.9451457858085632,
      "rewards-max/forensics_strict_edge_reward": 0.46775948339200113,
      "rewards-max/forensics_tiou_reward": 0.8801674246788025,
      "rewards/forensics_sinkhorn_reward": 0.5975174307823181,
      "rewards/forensics_strict_edge_reward": 0.020911241858101448,
      "rewards/forensics_tiou_reward": 0.32100826501846313,
      "step": 172
    },
    {
      "completion_length": 23.1875,
      "epoch": 0.5314900153609831,
      "grad_norm": 4.229218482971191,
      "kl": 0.2607421875,
      "learning_rate": 8.676923076923076e-07,
      "loss": -0.0566,
      "reward": 0.8818002939224243,
      "reward_std": 0.2482231929898262,
      "reward_std_all": 1.0909552574157715,
      "rewards-max/forensics_sinkhorn_reward": 0.9285366237163544,
      "rewards-max/forensics_strict_edge_reward": 0.07871073111891747,
      "rewards-max/forensics_tiou_reward": 0.8613412082195282,
      "rewards/forensics_sinkhorn_reward": 0.5798790752887726,
      "rewards/forensics_strict_edge_reward": 0.005131048848852515,
      "rewards/forensics_tiou_reward": 0.2967901974916458,
      "step": 173
    },
    {
      "completion_length": 25.328125,
      "epoch": 0.5345622119815668,
      "grad_norm": 4.18862771987915,
      "kl": 0.2333984375,
      "learning_rate": 8.669230769230769e-07,
      "loss": -0.079,
      "reward": 0.8944336771965027,
      "reward_std": 0.2567204609513283,
      "reward_std_all": 1.0886000394821167,
      "rewards-max/forensics_sinkhorn_reward": 0.8935434818267822,
      "rewards-max/forensics_strict_edge_reward": 0.23227659240365028,
      "rewards-max/forensics_tiou_reward": 0.7833570539951324,
      "rewards/forensics_sinkhorn_reward": 0.5580437481403351,
      "rewards/forensics_strict_edge_reward": 0.0125827775336802,
      "rewards/forensics_tiou_reward": 0.32380713522434235,
      "step": 174
    },
    {
      "completion_length": 24.90625,
      "epoch": 0.5376344086021505,
      "grad_norm": 4.306415557861328,
      "kl": 0.3212890625,
      "learning_rate": 8.661538461538461e-07,
      "loss": -0.1651,
      "reward": 0.8063282668590546,
      "reward_std": 0.23638597130775452,
      "reward_std_all": 1.1317601203918457,
      "rewards-max/forensics_sinkhorn_reward": 0.9718124270439148,
      "rewards-max/forensics_strict_edge_reward": 0.37195122241973877,
      "rewards-max/forensics_tiou_reward": 0.9525494575500488,
      "rewards/forensics_sinkhorn_reward": 0.5216940939426422,
      "rewards/forensics_strict_edge_reward": 0.016902542673051357,
      "rewards/forensics_tiou_reward": 0.2677316442131996,
      "step": 175
    },
    {
      "completion_length": 24.828125,
      "epoch": 0.5407066052227343,
      "grad_norm": 4.080478191375732,
      "kl": 0.310546875,
      "learning_rate": 8.653846153846154e-07,
      "loss": -0.0935,
      "reward": 0.7895412445068359,
      "reward_std": 0.18489285558462143,
      "reward_std_all": 1.1247704029083252,
      "rewards-max/forensics_sinkhorn_reward": 0.9176152050495148,
      "rewards-max/forensics_strict_edge_reward": 0.09710565535351634,
      "rewards-max/forensics_tiou_reward": 0.8278100192546844,
      "rewards/forensics_sinkhorn_reward": 0.5178332775831223,
      "rewards/forensics_strict_edge_reward": 0.004076721932506189,
      "rewards/forensics_tiou_reward": 0.2676312178373337,
      "step": 176
    },
    {
      "completion_length": 25.734375,
      "epoch": 0.543778801843318,
      "grad_norm": 3.895785093307495,
      "kl": 0.201171875,
      "learning_rate": 8.646153846153845e-07,
      "loss": -0.0624,
      "reward": 0.7692124843597412,
      "reward_std": 0.3237196207046509,
      "reward_std_all": 1.1770890951156616,
      "rewards-max/forensics_sinkhorn_reward": 0.8542375862598419,
      "rewards-max/forensics_strict_edge_reward": 0.4524692344639334,
      "rewards-max/forensics_tiou_reward": 0.7084557414054871,
      "rewards/forensics_sinkhorn_reward": 0.4852423071861267,
      "rewards/forensics_strict_edge_reward": 0.03429915006199735,
      "rewards/forensics_tiou_reward": 0.24967099726200104,
      "step": 177
    },
    {
      "completion_length": 23.765625,
      "epoch": 0.5468509984639017,
      "grad_norm": 4.770997047424316,
      "kl": 0.3076171875,
      "learning_rate": 8.638461538461537e-07,
      "loss": -0.1092,
      "reward": 0.753463089466095,
      "reward_std": 0.19948755204677582,
      "reward_std_all": 1.1529423594474792,
      "rewards-max/forensics_sinkhorn_reward": 0.8280006647109985,
      "rewards-max/forensics_strict_edge_reward": 0.1720769106649902,
      "rewards-max/forensics_tiou_reward": 0.6941013783216476,
      "rewards/forensics_sinkhorn_reward": 0.5129094272851944,
      "rewards/forensics_strict_edge_reward": 0.008306481658495346,
      "rewards/forensics_tiou_reward": 0.23224721103906631,
      "step": 178
    },
    {
      "completion_length": 26.0625,
      "epoch": 0.5499231950844854,
      "grad_norm": 4.622056484222412,
      "kl": 0.21875,
      "learning_rate": 8.630769230769231e-07,
      "loss": -0.1309,
      "reward": 0.7888284027576447,
      "reward_std": 0.19073579460382462,
      "reward_std_all": 1.1261337399482727,
      "rewards-max/forensics_sinkhorn_reward": 0.8821874558925629,
      "rewards-max/forensics_strict_edge_reward": 0.11253332866613164,
      "rewards-max/forensics_tiou_reward": 0.7393158674240112,
      "rewards/forensics_sinkhorn_reward": 0.5223821997642517,
      "rewards/forensics_strict_edge_reward": 0.004819011172415344,
      "rewards/forensics_tiou_reward": 0.2616272196173668,
      "step": 179
    },
    {
      "completion_length": 26.484375,
      "epoch": 0.5529953917050692,
      "grad_norm": 5.966556072235107,
      "kl": 0.412109375,
      "learning_rate": 8.623076923076923e-07,
      "loss": -0.1665,
      "reward": 0.9494032561779022,
      "reward_std": 0.3822217956185341,
      "reward_std_all": 1.1273955702781677,
      "rewards-max/forensics_sinkhorn_reward": 0.974536657333374,
      "rewards-max/forensics_strict_edge_reward": 0.5992296040058136,
      "rewards-max/forensics_tiou_reward": 0.9633634388446808,
      "rewards/forensics_sinkhorn_reward": 0.5471587479114532,
      "rewards/forensics_strict_edge_reward": 0.057836294174194336,
      "rewards/forensics_tiou_reward": 0.3444082662463188,
      "step": 180
    },
    {
      "completion_length": 25.234375,
      "epoch": 0.5560675883256528,
      "grad_norm": 4.657877445220947,
      "kl": 0.2783203125,
      "learning_rate": 8.615384615384616e-07,
      "loss": -0.0686,
      "reward": 0.908699244260788,
      "reward_std": 0.2589832693338394,
      "reward_std_all": 1.0884023904800415,
      "rewards-max/forensics_sinkhorn_reward": 0.9647814631462097,
      "rewards-max/forensics_strict_edge_reward": 0.16549075860530138,
      "rewards-max/forensics_tiou_reward": 0.9250280857086182,
      "rewards/forensics_sinkhorn_reward": 0.5665029883384705,
      "rewards/forensics_strict_edge_reward": 0.011512363445945084,
      "rewards/forensics_tiou_reward": 0.3306839317083359,
      "step": 181
    },
    {
      "completion_length": 24.625,
      "epoch": 0.5591397849462365,
      "grad_norm": 4.337339878082275,
      "kl": 0.34765625,
      "learning_rate": 8.607692307692307e-07,
      "loss": -0.0794,
      "reward": 0.8898859024047852,
      "reward_std": 0.4418390393257141,
      "reward_std_all": 1.1545074582099915,
      "rewards-max/forensics_sinkhorn_reward": 0.9733414947986603,
      "rewards-max/forensics_strict_edge_reward": 0.51341712474823,
      "rewards-max/forensics_tiou_reward": 0.9577240645885468,
      "rewards/forensics_sinkhorn_reward": 0.5267239212989807,
      "rewards/forensics_strict_edge_reward": 0.0343441991135478,
      "rewards/forensics_tiou_reward": 0.3288177251815796,
      "step": 182
    },
    {
      "completion_length": 23.59375,
      "epoch": 0.5622119815668203,
      "grad_norm": 3.7195816040039062,
      "kl": 0.2900390625,
      "learning_rate": 8.599999999999999e-07,
      "loss": -0.1232,
      "reward": 0.8151361644268036,
      "reward_std": 0.34153470396995544,
      "reward_std_all": 1.1484107971191406,
      "rewards-max/forensics_sinkhorn_reward": 0.9465181231498718,
      "rewards-max/forensics_strict_edge_reward": 0.16866089403629303,
      "rewards-max/forensics_tiou_reward": 0.8921728432178497,
      "rewards/forensics_sinkhorn_reward": 0.5322007834911346,
      "rewards/forensics_strict_edge_reward": 0.007092086365446448,
      "rewards/forensics_tiou_reward": 0.2758433073759079,
      "step": 183
    },
    {
      "completion_length": 25.125,
      "epoch": 0.565284178187404,
      "grad_norm": 3.810185432434082,
      "kl": 0.3505859375,
      "learning_rate": 8.592307692307692e-07,
      "loss": -0.1642,
      "reward": 0.6891230940818787,
      "reward_std": 0.298999086022377,
      "reward_std_all": 1.1992342472076416,
      "rewards-max/forensics_sinkhorn_reward": 0.9291128516197205,
      "rewards-max/forensics_strict_edge_reward": 0.065740080550313,
      "rewards-max/forensics_tiou_reward": 0.8616300523281097,
      "rewards/forensics_sinkhorn_reward": 0.48096078634262085,
      "rewards/forensics_strict_edge_reward": 0.003539575729519129,
      "rewards/forensics_tiou_reward": 0.20462271571159363,
      "step": 184
    },
    {
      "completion_length": 24.265625,
      "epoch": 0.5683563748079877,
      "grad_norm": 4.491569995880127,
      "kl": 0.28125,
      "learning_rate": 8.584615384615385e-07,
      "loss": -0.16,
      "reward": 0.8952232301235199,
      "reward_std": 0.41595253348350525,
      "reward_std_all": 1.1329618096351624,
      "rewards-max/forensics_sinkhorn_reward": 0.9682766199111938,
      "rewards-max/forensics_strict_edge_reward": 0.36102597415447235,
      "rewards-max/forensics_tiou_reward": 0.9427472949028015,
      "rewards/forensics_sinkhorn_reward": 0.5654435157775879,
      "rewards/forensics_strict_edge_reward": 0.02318065706640482,
      "rewards/forensics_tiou_reward": 0.3065989911556244,
      "step": 185
    },
    {
      "completion_length": 24.390625,
      "epoch": 0.5714285714285714,
      "grad_norm": 5.458090782165527,
      "kl": 0.3759765625,
      "learning_rate": 8.576923076923076e-07,
      "loss": -0.1062,
      "reward": 0.9033361375331879,
      "reward_std": 0.2611348405480385,
      "reward_std_all": 1.092639684677124,
      "rewards-max/forensics_sinkhorn_reward": 0.9262221157550812,
      "rewards-max/forensics_strict_edge_reward": 0.3334755219257204,
      "rewards-max/forensics_tiou_reward": 0.8228123486042023,
      "rewards/forensics_sinkhorn_reward": 0.582247793674469,
      "rewards/forensics_strict_edge_reward": 0.026165695144300116,
      "rewards/forensics_tiou_reward": 0.29492270946502686,
      "step": 186
    },
    {
      "completion_length": 29.359375,
      "epoch": 0.5745007680491552,
      "grad_norm": 3.9939870834350586,
      "kl": 0.32421875,
      "learning_rate": 8.569230769230769e-07,
      "loss": -0.1052,
      "reward": 0.6641294956207275,
      "reward_std": 0.25180739164352417,
      "reward_std_all": 1.2011775970458984,
      "rewards-max/forensics_sinkhorn_reward": 0.871180534362793,
      "rewards-max/forensics_strict_edge_reward": 0.019144159865390975,
      "rewards-max/forensics_tiou_reward": 0.7209641933441162,
      "rewards/forensics_sinkhorn_reward": 0.4328637272119522,
      "rewards/forensics_strict_edge_reward": 0.0012930307343594905,
      "rewards/forensics_tiou_reward": 0.2299727201461792,
      "step": 187
    },
    {
      "completion_length": 26.140625,
      "epoch": 0.5775729646697388,
      "grad_norm": 5.474437236785889,
      "kl": 0.302734375,
      "learning_rate": 8.561538461538461e-07,
      "loss": -0.1213,
      "reward": 0.8001088500022888,
      "reward_std": 0.21597123891115189,
      "reward_std_all": 1.1349202990531921,
      "rewards-max/forensics_sinkhorn_reward": 0.9587967991828918,
      "rewards-max/forensics_strict_edge_reward": 0.3217402696609497,
      "rewards-max/forensics_tiou_reward": 0.9281457662582397,
      "rewards/forensics_sinkhorn_reward": 0.4878656566143036,
      "rewards/forensics_strict_edge_reward": 0.016772370785474777,
      "rewards/forensics_tiou_reward": 0.29547081887722015,
      "step": 188
    },
    {
      "completion_length": 25.96875,
      "epoch": 0.5806451612903226,
      "grad_norm": 4.932538032531738,
      "kl": 0.3017578125,
      "learning_rate": 8.553846153846154e-07,
      "loss": -0.1025,
      "reward": 0.880863606929779,
      "reward_std": 0.4360043704509735,
      "reward_std_all": 1.1508561372756958,
      "rewards-max/forensics_sinkhorn_reward": 0.9785999059677124,
      "rewards-max/forensics_strict_edge_reward": 0.35114636272192,
      "rewards-max/forensics_tiou_reward": 0.9704371392726898,
      "rewards/forensics_sinkhorn_reward": 0.5413084626197815,
      "rewards/forensics_strict_edge_reward": 0.015195764601230621,
      "rewards/forensics_tiou_reward": 0.32435938715934753,
      "step": 189
    },
    {
      "completion_length": 27.59375,
      "epoch": 0.5837173579109063,
      "grad_norm": 3.832533359527588,
      "kl": 0.41015625,
      "learning_rate": 8.546153846153846e-07,
      "loss": -0.0881,
      "reward": 0.8283287286758423,
      "reward_std": 0.24313978105783463,
      "reward_std_all": 1.118506371974945,
      "rewards-max/forensics_sinkhorn_reward": 0.9136538803577423,
      "rewards-max/forensics_strict_edge_reward": 0.15978487022221088,
      "rewards-max/forensics_tiou_reward": 0.8232714831829071,
      "rewards/forensics_sinkhorn_reward": 0.5328133255243301,
      "rewards/forensics_strict_edge_reward": 0.014891286031343043,
      "rewards/forensics_tiou_reward": 0.280624121427536,
      "step": 190
    },
    {
      "completion_length": 26.171875,
      "epoch": 0.5867895545314901,
      "grad_norm": 4.500908851623535,
      "kl": 0.2880859375,
      "learning_rate": 8.538461538461537e-07,
      "loss": -0.0992,
      "reward": 0.8183819055557251,
      "reward_std": 0.3691839873790741,
      "reward_std_all": 1.172379493713379,
      "rewards-max/forensics_sinkhorn_reward": 0.9725374281406403,
      "rewards-max/forensics_strict_edge_reward": 0.4529060879140161,
      "rewards-max/forensics_tiou_reward": 0.9611684679985046,
      "rewards/forensics_sinkhorn_reward": 0.4872268736362457,
      "rewards/forensics_strict_edge_reward": 0.03550565700970765,
      "rewards/forensics_tiou_reward": 0.29564930498600006,
      "step": 191
    },
    {
      "completion_length": 27.515625,
      "epoch": 0.5898617511520737,
      "grad_norm": 4.981331825256348,
      "kl": 0.2763671875,
      "learning_rate": 8.530769230769231e-07,
      "loss": -0.0934,
      "reward": 0.9104979634284973,
      "reward_std": 0.40678006410598755,
      "reward_std_all": 1.124891221523285,
      "rewards-max/forensics_sinkhorn_reward": 0.9741673767566681,
      "rewards-max/forensics_strict_edge_reward": 0.4152904450893402,
      "rewards-max/forensics_tiou_reward": 0.9573462605476379,
      "rewards/forensics_sinkhorn_reward": 0.5249128639698029,
      "rewards/forensics_strict_edge_reward": 0.022016581147909164,
      "rewards/forensics_tiou_reward": 0.3635684698820114,
      "step": 192
    },
    {
      "completion_length": 25.265625,
      "epoch": 0.5929339477726574,
      "grad_norm": 3.9800333976745605,
      "kl": 0.30859375,
      "learning_rate": 8.523076923076923e-07,
      "loss": -0.0423,
      "reward": 0.8594448864459991,
      "reward_std": 0.20104893296957016,
      "reward_std_all": 1.088805079460144,
      "rewards-max/forensics_sinkhorn_reward": 0.8765860199928284,
      "rewards-max/forensics_strict_edge_reward": 0.05707456124946475,
      "rewards-max/forensics_tiou_reward": 0.7273355424404144,
      "rewards/forensics_sinkhorn_reward": 0.5408943891525269,
      "rewards/forensics_strict_edge_reward": 0.005223717336775735,
      "rewards/forensics_tiou_reward": 0.31332671642303467,
      "step": 193
    },
    {
      "completion_length": 29.171875,
      "epoch": 0.5960061443932412,
      "grad_norm": 4.528804779052734,
      "kl": 0.31396484375,
      "learning_rate": 8.515384615384616e-07,
      "loss": -0.1042,
      "reward": 0.7382984161376953,
      "reward_std": 0.2784116454422474,
      "reward_std_all": 1.1741108298301697,
      "rewards-max/forensics_sinkhorn_reward": 0.8928335011005402,
      "rewards-max/forensics_strict_edge_reward": 0.4912954866886139,
      "rewards-max/forensics_tiou_reward": 0.8159725964069366,
      "rewards/forensics_sinkhorn_reward": 0.41754959523677826,
      "rewards/forensics_strict_edge_reward": 0.03617868386209011,
      "rewards/forensics_tiou_reward": 0.2845701277256012,
      "step": 194
    },
    {
      "completion_length": 27.453125,
      "epoch": 0.5990783410138248,
      "grad_norm": 3.8813583850860596,
      "kl": 0.291015625,
      "learning_rate": 8.507692307692307e-07,
      "loss": -0.0916,
      "reward": 0.7846606075763702,
      "reward_std": 0.21722017228603363,
      "reward_std_all": 1.1347708702087402,
      "rewards-max/forensics_sinkhorn_reward": 0.9563677906990051,
      "rewards-max/forensics_strict_edge_reward": 0.15995130129158497,
      "rewards-max/forensics_tiou_reward": 0.9096525013446808,
      "rewards/forensics_sinkhorn_reward": 0.47293177247047424,
      "rewards/forensics_strict_edge_reward": 0.007039332878775895,
      "rewards/forensics_tiou_reward": 0.30468951165676117,
      "step": 195
    },
    {
      "completion_length": 29.796875,
      "epoch": 0.6021505376344086,
      "grad_norm": 4.437498569488525,
      "kl": 0.22265625,
      "learning_rate": 8.499999999999999e-07,
      "loss": -0.1299,
      "reward": 0.6386480033397675,
      "reward_std": 0.18974794447422028,
      "reward_std_all": 1.2023093700408936,
      "rewards-max/forensics_sinkhorn_reward": 0.907133549451828,
      "rewards-max/forensics_strict_edge_reward": 0.12340368266450241,
      "rewards-max/forensics_tiou_reward": 0.8085299134254456,
      "rewards/forensics_sinkhorn_reward": 0.4335896223783493,
      "rewards/forensics_strict_edge_reward": 0.00843810900914832,
      "rewards/forensics_tiou_reward": 0.19662027060985565,
      "step": 196
    },
    {
      "completion_length": 31.40625,
      "epoch": 0.6052227342549923,
      "grad_norm": 4.104007720947266,
      "kl": 0.271484375,
      "learning_rate": 8.492307692307692e-07,
      "loss": -0.114,
      "reward": 0.7209722697734833,
      "reward_std": 0.25963571667671204,
      "reward_std_all": 1.18337082862854,
      "rewards-max/forensics_sinkhorn_reward": 0.9328944683074951,
      "rewards-max/forensics_strict_edge_reward": 0.19289823132567108,
      "rewards-max/forensics_tiou_reward": 0.8642460107803345,
      "rewards/forensics_sinkhorn_reward": 0.4518475979566574,
      "rewards/forensics_strict_edge_reward": 0.013598591482150368,
      "rewards/forensics_tiou_reward": 0.2555260881781578,
      "step": 197
    },
    {
      "completion_length": 28.234375,
      "epoch": 0.6082949308755761,
      "grad_norm": 5.030844688415527,
      "kl": 0.2607421875,
      "learning_rate": 8.484615384615384e-07,
      "loss": -0.0779,
      "reward": 0.7385421991348267,
      "reward_std": 0.2525029554963112,
      "reward_std_all": 1.1634504795074463,
      "rewards-max/forensics_sinkhorn_reward": 0.9456263184547424,
      "rewards-max/forensics_strict_edge_reward": 0.17091454565525055,
      "rewards-max/forensics_tiou_reward": 0.8927602767944336,
      "rewards/forensics_sinkhorn_reward": 0.4957275837659836,
      "rewards/forensics_strict_edge_reward": 0.008989592548459768,
      "rewards/forensics_tiou_reward": 0.23382504284381866,
      "step": 198
    },
    {
      "completion_length": 24.765625,
      "epoch": 0.6113671274961597,
      "grad_norm": 5.284968376159668,
      "kl": 0.2734375,
      "learning_rate": 8.476923076923077e-07,
      "loss": -0.0658,
      "reward": 0.8895304203033447,
      "reward_std": 0.24046015739440918,
      "reward_std_all": 1.0791064500808716,
      "rewards-max/forensics_sinkhorn_reward": 0.9176963269710541,
      "rewards-max/forensics_strict_edge_reward": 0.11205768212676048,
      "rewards-max/forensics_tiou_reward": 0.8310703337192535,
      "rewards/forensics_sinkhorn_reward": 0.5729221105575562,
      "rewards/forensics_strict_edge_reward": 0.005369826685637236,
      "rewards/forensics_tiou_reward": 0.311238557100296,
      "step": 199
    },
    {
      "completion_length": 24.28125,
      "epoch": 0.6144393241167435,
      "grad_norm": 4.793516635894775,
      "kl": 0.3466796875,
      "learning_rate": 8.469230769230769e-07,
      "loss": -0.136,
      "reward": 0.7652366757392883,
      "reward_std": 0.23712879419326782,
      "reward_std_all": 1.1443977355957031,
      "rewards-max/forensics_sinkhorn_reward": 0.9215649366378784,
      "rewards-max/forensics_strict_edge_reward": 0.00622333592036739,
      "rewards-max/forensics_tiou_reward": 0.8220097124576569,
      "rewards/forensics_sinkhorn_reward": 0.4892473518848419,
      "rewards/forensics_strict_edge_reward": 0.0002834110455296468,
      "rewards/forensics_tiou_reward": 0.2757059186697006,
      "step": 200
    },
    {
      "completion_length": 26.125,
      "epoch": 0.6175115207373272,
      "grad_norm": 5.2502875328063965,
      "kl": 0.2607421875,
      "learning_rate": 8.461538461538461e-07,
      "loss": -0.0351,
      "reward": 0.8595246076583862,
      "reward_std": 0.3449467420578003,
      "reward_std_all": 1.1377846002578735,
      "rewards-max/forensics_sinkhorn_reward": 0.967322438955307,
      "rewards-max/forensics_strict_edge_reward": 0.3001431114971638,
      "rewards-max/forensics_tiou_reward": 0.9391074776649475,
      "rewards/forensics_sinkhorn_reward": 0.48616519570350647,
      "rewards/forensics_strict_edge_reward": 0.017628455301746726,
      "rewards/forensics_tiou_reward": 0.35573092103004456,
      "step": 201
    },
    {
      "completion_length": 27.546875,
      "epoch": 0.620583717357911,
      "grad_norm": 4.6229634284973145,
      "kl": 0.302734375,
      "learning_rate": 8.453846153846154e-07,
      "loss": -0.088,
      "reward": 0.7428737580776215,
      "reward_std": 0.2670065611600876,
      "reward_std_all": 1.1770319938659668,
      "rewards-max/forensics_sinkhorn_reward": 0.9416160583496094,
      "rewards-max/forensics_strict_edge_reward": 0.08469283278100193,
      "rewards-max/forensics_tiou_reward": 0.8916150033473969,
      "rewards/forensics_sinkhorn_reward": 0.471801295876503,
      "rewards/forensics_strict_edge_reward": 0.0037382873342721723,
      "rewards/forensics_tiou_reward": 0.2673341631889343,
      "step": 202
    },
    {
      "completion_length": 22.75,
      "epoch": 0.6236559139784946,
      "grad_norm": 5.325316429138184,
      "kl": 0.39453125,
      "learning_rate": 8.446153846153846e-07,
      "loss": -0.156,
      "reward": 0.8647566437721252,
      "reward_std": 0.2717088609933853,
      "reward_std_all": 1.1163280606269836,
      "rewards-max/forensics_sinkhorn_reward": 0.9246201813220978,
      "rewards-max/forensics_strict_edge_reward": 0.3070678152143955,
      "rewards-max/forensics_tiou_reward": 0.851790577173233,
      "rewards/forensics_sinkhorn_reward": 0.5841813385486603,
      "rewards/forensics_strict_edge_reward": 0.019749418832361698,
      "rewards/forensics_tiou_reward": 0.2608259543776512,
      "step": 203
    },
    {
      "completion_length": 25.71875,
      "epoch": 0.6267281105990783,
      "grad_norm": 4.700697422027588,
      "kl": 0.2724609375,
      "learning_rate": 8.438461538461537e-07,
      "loss": -0.0734,
      "reward": 0.7697105705738068,
      "reward_std": 0.24850867688655853,
      "reward_std_all": 1.1499398946762085,
      "rewards-max/forensics_sinkhorn_reward": 0.8601302206516266,
      "rewards-max/forensics_strict_edge_reward": 0.21071381703950465,
      "rewards-max/forensics_tiou_reward": 0.7239552736282349,
      "rewards/forensics_sinkhorn_reward": 0.48605482280254364,
      "rewards/forensics_strict_edge_reward": 0.011303284412861103,
      "rewards/forensics_tiou_reward": 0.2723524644970894,
      "step": 204
    },
    {
      "completion_length": 25.171875,
      "epoch": 0.6298003072196621,
      "grad_norm": 4.9787797927856445,
      "kl": 0.2705078125,
      "learning_rate": 8.43076923076923e-07,
      "loss": -0.1501,
      "reward": 1.0074484944343567,
      "reward_std": 0.30857138335704803,
      "reward_std_all": 1.044135957956314,
      "rewards-max/forensics_sinkhorn_reward": 0.9461232423782349,
      "rewards-max/forensics_strict_edge_reward": 0.1767077073454857,
      "rewards-max/forensics_tiou_reward": 0.8966825008392334,
      "rewards/forensics_sinkhorn_reward": 0.6193855702877045,
      "rewards/forensics_strict_edge_reward": 0.01739064184948802,
      "rewards/forensics_tiou_reward": 0.3706723302602768,
      "step": 205
    },
    {
      "completion_length": 26.5625,
      "epoch": 0.6328725038402457,
      "grad_norm": 3.6098008155822754,
      "kl": 0.2490234375,
      "learning_rate": 8.423076923076923e-07,
      "loss": -0.072,
      "reward": 0.8426041305065155,
      "reward_std": 0.3920823186635971,
      "reward_std_all": 1.1673002243041992,
      "rewards-max/forensics_sinkhorn_reward": 0.9591656625270844,
      "rewards-max/forensics_strict_edge_reward": 0.25635693594813347,
      "rewards-max/forensics_tiou_reward": 0.9142042398452759,
      "rewards/forensics_sinkhorn_reward": 0.5229005068540573,
      "rewards/forensics_strict_edge_reward": 0.022914121160283685,
      "rewards/forensics_tiou_reward": 0.2967894971370697,
      "step": 206
    },
    {
      "completion_length": 27.28125,
      "epoch": 0.6359447004608295,
      "grad_norm": 3.939335823059082,
      "kl": 0.35546875,
      "learning_rate": 8.415384615384616e-07,
      "loss": -0.0762,
      "reward": 0.9026505351066589,
      "reward_std": 0.22161592543125153,
      "reward_std_all": 1.0699161887168884,
      "rewards-max/forensics_sinkhorn_reward": 0.9007805585861206,
      "rewards-max/forensics_strict_edge_reward": 0.08281766809523106,
      "rewards-max/forensics_tiou_reward": 0.7811729609966278,
      "rewards/forensics_sinkhorn_reward": 0.561803936958313,
      "rewards/forensics_strict_edge_reward": 0.003589090541936457,
      "rewards/forensics_tiou_reward": 0.337257519364357,
      "step": 207
    },
    {
      "completion_length": 23.796875,
      "epoch": 0.6390168970814132,
      "grad_norm": 4.980459690093994,
      "kl": 0.2939453125,
      "learning_rate": 8.407692307692308e-07,
      "loss": -0.1376,
      "reward": 0.8629139065742493,
      "reward_std": 0.3850683271884918,
      "reward_std_all": 1.1382495760917664,
      "rewards-max/forensics_sinkhorn_reward": 0.9632821083068848,
      "rewards-max/forensics_strict_edge_reward": 0.3872668445110321,
      "rewards-max/forensics_tiou_reward": 0.935208797454834,
      "rewards/forensics_sinkhorn_reward": 0.5312789082527161,
      "rewards/forensics_strict_edge_reward": 0.03226853162050247,
      "rewards/forensics_tiou_reward": 0.29936645925045013,
      "step": 208
    },
    {
      "completion_length": 25.4375,
      "epoch": 0.642089093701997,
      "grad_norm": 4.7972493171691895,
      "kl": 0.32421875,
      "learning_rate": 8.399999999999999e-07,
      "loss": -0.1155,
      "reward": 0.7475718557834625,
      "reward_std": 0.26200374960899353,
      "reward_std_all": 1.1676336526870728,
      "rewards-max/forensics_sinkhorn_reward": 0.9264851808547974,
      "rewards-max/forensics_strict_edge_reward": 0.17945101010263897,
      "rewards-max/forensics_tiou_reward": 0.8321640491485596,
      "rewards/forensics_sinkhorn_reward": 0.4994538724422455,
      "rewards/forensics_strict_edge_reward": 0.01610280801105546,
      "rewards/forensics_tiou_reward": 0.23201517760753632,
      "step": 209
    },
    {
      "completion_length": 27.25,
      "epoch": 0.6451612903225806,
      "grad_norm": 4.571760654449463,
      "kl": 0.25537109375,
      "learning_rate": 8.392307692307692e-07,
      "loss": -0.1121,
      "reward": 0.7578628957271576,
      "reward_std": 0.16915585845708847,
      "reward_std_all": 1.1350890398025513,
      "rewards-max/forensics_sinkhorn_reward": 0.9023438394069672,
      "rewards-max/forensics_strict_edge_reward": 0.02206558908801526,
      "rewards-max/forensics_tiou_reward": 0.7840267717838287,
      "rewards/forensics_sinkhorn_reward": 0.5077823847532272,
      "rewards/forensics_strict_edge_reward": 0.0009952576947398484,
      "rewards/forensics_tiou_reward": 0.24908524006605148,
      "step": 210
    },
    {
      "completion_length": 22.9375,
      "epoch": 0.6482334869431644,
      "grad_norm": 5.521291255950928,
      "kl": 0.2734375,
      "learning_rate": 8.384615384615384e-07,
      "loss": -0.1943,
      "reward": 0.9032341539859772,
      "reward_std": 0.298727810382843,
      "reward_std_all": 1.0963470339775085,
      "rewards-max/forensics_sinkhorn_reward": 0.9621805548667908,
      "rewards-max/forensics_strict_edge_reward": 0.24516794085502625,
      "rewards-max/forensics_tiou_reward": 0.9222531616687775,
      "rewards/forensics_sinkhorn_reward": 0.5785905122756958,
      "rewards/forensics_strict_edge_reward": 0.010451170848682523,
      "rewards/forensics_tiou_reward": 0.314192533493042,
      "step": 211
    },
    {
      "completion_length": 23.296875,
      "epoch": 0.6513056835637481,
      "grad_norm": 4.718843936920166,
      "kl": 0.2841796875,
      "learning_rate": 8.376923076923076e-07,
      "loss": -0.0819,
      "reward": 0.8572453856468201,
      "reward_std": 0.3003925085067749,
      "reward_std_all": 1.1251021027565002,
      "rewards-max/forensics_sinkhorn_reward": 0.9441857039928436,
      "rewards-max/forensics_strict_edge_reward": 0.1724927257746458,
      "rewards-max/forensics_tiou_reward": 0.8763463199138641,
      "rewards/forensics_sinkhorn_reward": 0.5651476979255676,
      "rewards/forensics_strict_edge_reward": 0.011094241839600727,
      "rewards/forensics_tiou_reward": 0.2810034453868866,
      "step": 212
    },
    {
      "completion_length": 23.984375,
      "epoch": 0.6543778801843319,
      "grad_norm": 4.241907119750977,
      "kl": 0.27294921875,
      "learning_rate": 8.369230769230769e-07,
      "loss": -0.1109,
      "reward": 0.7656335234642029,
      "reward_std": 0.26747262477874756,
      "reward_std_all": 1.157536506652832,
      "rewards-max/forensics_sinkhorn_reward": 0.9507568776607513,
      "rewards-max/forensics_strict_edge_reward": 0.07219875603914261,
      "rewards-max/forensics_tiou_reward": 0.8947097361087799,
      "rewards/forensics_sinkhorn_reward": 0.510130375623703,
      "rewards/forensics_strict_edge_reward": 0.003980603767558932,
      "rewards/forensics_tiou_reward": 0.25152261555194855,
      "step": 213
    },
    {
      "completion_length": 22.59375,
      "epoch": 0.6574500768049155,
      "grad_norm": 5.8849287033081055,
      "kl": 0.27734375,
      "learning_rate": 8.361538461538461e-07,
      "loss": -0.1423,
      "reward": 0.86318239569664,
      "reward_std": 0.25042034685611725,
      "reward_std_all": 1.107697308063507,
      "rewards-max/forensics_sinkhorn_reward": 0.9489326775074005,
      "rewards-max/forensics_strict_edge_reward": 0.45087406039237976,
      "rewards-max/forensics_tiou_reward": 0.8977662026882172,
      "rewards/forensics_sinkhorn_reward": 0.5594295114278793,
      "rewards/forensics_strict_edge_reward": 0.01937139593064785,
      "rewards/forensics_tiou_reward": 0.2843814939260483,
      "step": 214
    },
    {
      "completion_length": 25.015625,
      "epoch": 0.6605222734254992,
      "grad_norm": 4.522496700286865,
      "kl": 0.22509765625,
      "learning_rate": 8.353846153846154e-07,
      "loss": -0.076,
      "reward": 0.7811177372932434,
      "reward_std": 0.19053030014038086,
      "reward_std_all": 1.1250008344650269,
      "rewards-max/forensics_sinkhorn_reward": 0.8828202188014984,
      "rewards-max/forensics_strict_edge_reward": 0.029405052540823817,
      "rewards-max/forensics_tiou_reward": 0.7195963561534882,
      "rewards/forensics_sinkhorn_reward": 0.5069347321987152,
      "rewards/forensics_strict_edge_reward": 0.001410489232512191,
      "rewards/forensics_tiou_reward": 0.27277253568172455,
      "step": 215
    },
    {
      "completion_length": 24.71875,
      "epoch": 0.663594470046083,
      "grad_norm": 5.466601848602295,
      "kl": 0.2822265625,
      "learning_rate": 8.346153846153846e-07,
      "loss": -0.0775,
      "reward": 0.7709912359714508,
      "reward_std": 0.22203299030661583,
      "reward_std_all": 1.1493175625801086,
      "rewards-max/forensics_sinkhorn_reward": 0.9495449662208557,
      "rewards-max/forensics_strict_edge_reward": 0.07479662541300058,
      "rewards-max/forensics_tiou_reward": 0.877543717622757,
      "rewards/forensics_sinkhorn_reward": 0.513613149523735,
      "rewards/forensics_strict_edge_reward": 0.004528416320681572,
      "rewards/forensics_tiou_reward": 0.25284966081380844,
      "step": 216
    },
    {
      "completion_length": 24.890625,
      "epoch": 0.6666666666666666,
      "grad_norm": 4.738059043884277,
      "kl": 0.259765625,
      "learning_rate": 8.338461538461537e-07,
      "loss": -0.0719,
      "reward": 0.8697051405906677,
      "reward_std": 0.2983345091342926,
      "reward_std_all": 1.1174196004867554,
      "rewards-max/forensics_sinkhorn_reward": 0.9595288336277008,
      "rewards-max/forensics_strict_edge_reward": 0.29105816036462784,
      "rewards-max/forensics_tiou_reward": 0.9245380461215973,
      "rewards/forensics_sinkhorn_reward": 0.5492069572210312,
      "rewards/forensics_strict_edge_reward": 0.014893571846187115,
      "rewards/forensics_tiou_reward": 0.30560460686683655,
      "step": 217
    },
    {
      "completion_length": 25.796875,
      "epoch": 0.6697388632872504,
      "grad_norm": 4.50336217880249,
      "kl": 0.220703125,
      "learning_rate": 8.33076923076923e-07,
      "loss": -0.1691,
      "reward": 0.9547181725502014,
      "reward_std": 0.3984352797269821,
      "reward_std_all": 1.1056800484657288,
      "rewards-max/forensics_sinkhorn_reward": 0.9754603207111359,
      "rewards-max/forensics_strict_edge_reward": 0.3560166209936142,
      "rewards-max/forensics_tiou_reward": 0.9620888829231262,
      "rewards/forensics_sinkhorn_reward": 0.5691647231578827,
      "rewards/forensics_strict_edge_reward": 0.029012045823037624,
      "rewards/forensics_tiou_reward": 0.3565414249897003,
      "step": 218
    },
    {
      "completion_length": 23.203125,
      "epoch": 0.6728110599078341,
      "grad_norm": 4.380436897277832,
      "kl": 0.20166015625,
      "learning_rate": 8.323076923076922e-07,
      "loss": -0.1301,
      "reward": 0.8324571251869202,
      "reward_std": 0.21249091625213623,
      "reward_std_all": 1.1113938689231873,
      "rewards-max/forensics_sinkhorn_reward": 0.9053060710430145,
      "rewards-max/forensics_strict_edge_reward": 0.035197666758904234,
      "rewards-max/forensics_tiou_reward": 0.8050480782985687,
      "rewards/forensics_sinkhorn_reward": 0.5380880832672119,
      "rewards/forensics_strict_edge_reward": 0.003180847252224339,
      "rewards/forensics_tiou_reward": 0.2911882549524307,
      "step": 219
    },
    {
      "completion_length": 27.984375,
      "epoch": 0.6758832565284179,
      "grad_norm": 4.372664451599121,
      "kl": 0.2646484375,
      "learning_rate": 8.315384615384616e-07,
      "loss": -0.0811,
      "reward": 0.6542287766933441,
      "reward_std": 0.27325767278671265,
      "reward_std_all": 1.2184452414512634,
      "rewards-max/forensics_sinkhorn_reward": 0.939271867275238,
      "rewards-max/forensics_strict_edge_reward": 0.4058148264884949,
      "rewards-max/forensics_tiou_reward": 0.8820386528968811,
      "rewards/forensics_sinkhorn_reward": 0.4076067954301834,
      "rewards/forensics_strict_edge_reward": 0.023017432540655136,
      "rewards/forensics_tiou_reward": 0.22360452264547348,
      "step": 220
    },
    {
      "completion_length": 23.28125,
      "epoch": 0.6789554531490015,
      "grad_norm": 4.006892681121826,
      "kl": 0.314453125,
      "learning_rate": 8.307692307692308e-07,
      "loss": -0.1393,
      "reward": 0.7419914901256561,
      "reward_std": 0.33394691348075867,
      "reward_std_all": 1.1913447976112366,
      "rewards-max/forensics_sinkhorn_reward": 0.9188189208507538,
      "rewards-max/forensics_strict_edge_reward": 0.13303794153034687,
      "rewards-max/forensics_tiou_reward": 0.8444943726062775,
      "rewards/forensics_sinkhorn_reward": 0.4709184616804123,
      "rewards/forensics_strict_edge_reward": 0.010224242359981872,
      "rewards/forensics_tiou_reward": 0.2608487755060196,
      "step": 221
    },
    {
      "completion_length": 24.984375,
      "epoch": 0.6820276497695853,
      "grad_norm": 3.364161729812622,
      "kl": 0.2568359375,
      "learning_rate": 8.299999999999999e-07,
      "loss": -0.0165,
      "reward": 0.7411478459835052,
      "reward_std": 0.2394830584526062,
      "reward_std_all": 1.1748182773590088,
      "rewards-max/forensics_sinkhorn_reward": 0.9681531488895416,
      "rewards-max/forensics_strict_edge_reward": 0.516239482909441,
      "rewards-max/forensics_tiou_reward": 0.9484804570674896,
      "rewards/forensics_sinkhorn_reward": 0.46682611107826233,
      "rewards/forensics_strict_edge_reward": 0.02239693282172084,
      "rewards/forensics_tiou_reward": 0.2519247680902481,
      "step": 222
    },
    {
      "completion_length": 23.375,
      "epoch": 0.685099846390169,
      "grad_norm": 3.7653968334198,
      "kl": 0.2802734375,
      "learning_rate": 8.292307692307692e-07,
      "loss": -0.1207,
      "reward": 0.999596893787384,
      "reward_std": 0.2567973881959915,
      "reward_std_all": 1.036914885044098,
      "rewards-max/forensics_sinkhorn_reward": 0.942087858915329,
      "rewards-max/forensics_strict_edge_reward": 0.26918745040893555,
      "rewards-max/forensics_tiou_reward": 0.8761129379272461,
      "rewards/forensics_sinkhorn_reward": 0.6060350239276886,
      "rewards/forensics_strict_edge_reward": 0.013123481534421444,
      "rewards/forensics_tiou_reward": 0.3804383873939514,
      "step": 223
    },
    {
      "completion_length": 25.296875,
      "epoch": 0.6881720430107527,
      "grad_norm": 5.123507022857666,
      "kl": 0.287109375,
      "learning_rate": 8.284615384615384e-07,
      "loss": -0.158,
      "reward": 0.7586570978164673,
      "reward_std": 0.2748127579689026,
      "reward_std_all": 1.1628727912902832,
      "rewards-max/forensics_sinkhorn_reward": 0.9126087427139282,
      "rewards-max/forensics_strict_edge_reward": 0.190044938120991,
      "rewards-max/forensics_tiou_reward": 0.841981053352356,
      "rewards/forensics_sinkhorn_reward": 0.48508627712726593,
      "rewards/forensics_strict_edge_reward": 0.009035023162141442,
      "rewards/forensics_tiou_reward": 0.2645357847213745,
      "step": 224
    },
    {
      "completion_length": 25.890625,
      "epoch": 0.6912442396313364,
      "grad_norm": 6.286050796508789,
      "kl": 0.521484375,
      "learning_rate": 8.276923076923077e-07,
      "loss": -0.1455,
      "reward": 0.8092421293258667,
      "reward_std": 0.3478693217039108,
      "reward_std_all": 1.1706358194351196,
      "rewards-max/forensics_sinkhorn_reward": 0.9720444083213806,
      "rewards-max/forensics_strict_edge_reward": 0.3596789318835363,
      "rewards-max/forensics_tiou_reward": 0.9552342593669891,
      "rewards/forensics_sinkhorn_reward": 0.5170714259147644,
      "rewards/forensics_strict_edge_reward": 0.0318679143783811,
      "rewards/forensics_tiou_reward": 0.26030272245407104,
      "step": 225
    },
    {
      "completion_length": 25.609375,
      "epoch": 0.6943164362519201,
      "grad_norm": 4.6332855224609375,
      "kl": 0.2626953125,
      "learning_rate": 8.269230769230768e-07,
      "loss": -0.0673,
      "reward": 0.6716846227645874,
      "reward_std": 0.24853311479091644,
      "reward_std_all": 1.2019945979118347,
      "rewards-max/forensics_sinkhorn_reward": 0.938847005367279,
      "rewards-max/forensics_strict_edge_reward": 0.29931730777025223,
      "rewards-max/forensics_tiou_reward": 0.8808535635471344,
      "rewards/forensics_sinkhorn_reward": 0.41728758811950684,
      "rewards/forensics_strict_edge_reward": 0.023223808617331088,
      "rewards/forensics_tiou_reward": 0.23117320984601974,
      "step": 226
    },
    {
      "completion_length": 21.96875,
      "epoch": 0.6973886328725039,
      "grad_norm": 5.3792572021484375,
      "kl": 0.24609375,
      "learning_rate": 8.261538461538461e-07,
      "loss": -0.1163,
      "reward": 0.8616739511489868,
      "reward_std": 0.32112377882003784,
      "reward_std_all": 1.1215733885765076,
      "rewards-max/forensics_sinkhorn_reward": 0.9474954009056091,
      "rewards-max/forensics_strict_edge_reward": 0.26828642189502716,
      "rewards-max/forensics_tiou_reward": 0.8944098651409149,
      "rewards/forensics_sinkhorn_reward": 0.5595496296882629,
      "rewards/forensics_strict_edge_reward": 0.011845937930047512,
      "rewards/forensics_tiou_reward": 0.29027843475341797,
      "step": 227
    },
    {
      "completion_length": 23.5625,
      "epoch": 0.7004608294930875,
      "grad_norm": 4.215519905090332,
      "kl": 0.240234375,
      "learning_rate": 8.253846153846154e-07,
      "loss": -0.1185,
      "reward": 0.8668889701366425,
      "reward_std": 0.3206329792737961,
      "reward_std_all": 1.125680387020111,
      "rewards-max/forensics_sinkhorn_reward": 0.9582951664924622,
      "rewards-max/forensics_strict_edge_reward": 0.33871981501579285,
      "rewards-max/forensics_tiou_reward": 0.9232236444950104,
      "rewards/forensics_sinkhorn_reward": 0.5247557461261749,
      "rewards/forensics_strict_edge_reward": 0.023592907935380936,
      "rewards/forensics_tiou_reward": 0.3185403347015381,
      "step": 228
    },
    {
      "completion_length": 25.046875,
      "epoch": 0.7035330261136713,
      "grad_norm": 4.71560525894165,
      "kl": 0.28515625,
      "learning_rate": 8.246153846153846e-07,
      "loss": -0.1101,
      "reward": 0.9582063555717468,
      "reward_std": 0.40529845654964447,
      "reward_std_all": 1.1165766716003418,
      "rewards-max/forensics_sinkhorn_reward": 0.9781126379966736,
      "rewards-max/forensics_strict_edge_reward": 0.45258962363004684,
      "rewards-max/forensics_tiou_reward": 0.9717772305011749,
      "rewards/forensics_sinkhorn_reward": 0.556260496377945,
      "rewards/forensics_strict_edge_reward": 0.045083037577569485,
      "rewards/forensics_tiou_reward": 0.3568628132343292,
      "step": 229
    },
    {
      "completion_length": 22.203125,
      "epoch": 0.706605222734255,
      "grad_norm": 4.762278079986572,
      "kl": 0.25048828125,
      "learning_rate": 8.238461538461538e-07,
      "loss": -0.1504,
      "reward": 0.7587206065654755,
      "reward_std": 0.32918211817741394,
      "reward_std_all": 1.17396080493927,
      "rewards-max/forensics_sinkhorn_reward": 0.9295639097690582,
      "rewards-max/forensics_strict_edge_reward": 0.14988363906741142,
      "rewards-max/forensics_tiou_reward": 0.8508151173591614,
      "rewards/forensics_sinkhorn_reward": 0.5379521250724792,
      "rewards/forensics_strict_edge_reward": 0.011157713714055717,
      "rewards/forensics_tiou_reward": 0.209610715508461,
      "step": 230
    },
    {
      "completion_length": 25.203125,
      "epoch": 0.7096774193548387,
      "grad_norm": 3.8750312328338623,
      "kl": 0.380859375,
      "learning_rate": 8.23076923076923e-07,
      "loss": -0.1173,
      "reward": 0.8794668018817902,
      "reward_std": 0.2783610075712204,
      "reward_std_all": 1.0990782976150513,
      "rewards-max/forensics_sinkhorn_reward": 0.9343242943286896,
      "rewards-max/forensics_strict_edge_reward": 0.38479528576135635,
      "rewards-max/forensics_tiou_reward": 0.8735891878604889,
      "rewards/forensics_sinkhorn_reward": 0.5412168800830841,
      "rewards/forensics_strict_edge_reward": 0.02124424185603857,
      "rewards/forensics_tiou_reward": 0.3170057088136673,
      "step": 231
    },
    {
      "completion_length": 22.84375,
      "epoch": 0.7127496159754224,
      "grad_norm": 4.225371360778809,
      "kl": 0.3544921875,
      "learning_rate": 8.223076923076922e-07,
      "loss": -0.1351,
      "reward": 0.8151653110980988,
      "reward_std": 0.32225392758846283,
      "reward_std_all": 1.1473880410194397,
      "rewards-max/forensics_sinkhorn_reward": 0.9711139500141144,
      "rewards-max/forensics_strict_edge_reward": 0.22691167145967484,
      "rewards-max/forensics_tiou_reward": 0.9526216387748718,
      "rewards/forensics_sinkhorn_reward": 0.5171855390071869,
      "rewards/forensics_strict_edge_reward": 0.01901789102703333,
      "rewards/forensics_tiou_reward": 0.2789618968963623,
      "step": 232
    },
    {
      "completion_length": 24.40625,
      "epoch": 0.7158218125960062,
      "grad_norm": 4.3494672775268555,
      "kl": 0.29296875,
      "learning_rate": 8.215384615384616e-07,
      "loss": -0.0732,
      "reward": 0.8107130229473114,
      "reward_std": 0.25898703187704086,
      "reward_std_all": 1.1335029006004333,
      "rewards-max/forensics_sinkhorn_reward": 0.9403650462627411,
      "rewards-max/forensics_strict_edge_reward": 0.0946253165602684,
      "rewards-max/forensics_tiou_reward": 0.8522273004055023,
      "rewards/forensics_sinkhorn_reward": 0.5340627282857895,
      "rewards/forensics_strict_edge_reward": 0.005320434225723147,
      "rewards/forensics_tiou_reward": 0.2713299095630646,
      "step": 233
    },
    {
      "completion_length": 26.453125,
      "epoch": 0.7188940092165899,
      "grad_norm": 5.681633472442627,
      "kl": 0.3740234375,
      "learning_rate": 8.207692307692308e-07,
      "loss": -0.1327,
      "reward": 0.840159684419632,
      "reward_std": 0.31865711510181427,
      "reward_std_all": 1.149975061416626,
      "rewards-max/forensics_sinkhorn_reward": 0.9724470973014832,
      "rewards-max/forensics_strict_edge_reward": 0.31723683699965477,
      "rewards-max/forensics_tiou_reward": 0.9624163806438446,
      "rewards/forensics_sinkhorn_reward": 0.4995233863592148,
      "rewards/forensics_strict_edge_reward": 0.023676004260778427,
      "rewards/forensics_tiou_reward": 0.3169602304697037,
      "step": 234
    },
    {
      "completion_length": 30.625,
      "epoch": 0.7219662058371735,
      "grad_norm": 3.959278106689453,
      "kl": 0.359375,
      "learning_rate": 8.199999999999999e-07,
      "loss": -0.0941,
      "reward": 0.6898341774940491,
      "reward_std": 0.2081679254770279,
      "reward_std_all": 1.1783562898635864,
      "rewards-max/forensics_sinkhorn_reward": 0.9643835127353668,
      "rewards-max/forensics_strict_edge_reward": 0.02673580749615212,
      "rewards-max/forensics_tiou_reward": 0.9365235865116119,
      "rewards/forensics_sinkhorn_reward": 0.4589940756559372,
      "rewards/forensics_strict_edge_reward": 0.0015665524122141505,
      "rewards/forensics_tiou_reward": 0.2292734757065773,
      "step": 235
    },
    {
      "completion_length": 22.84375,
      "epoch": 0.7250384024577573,
      "grad_norm": 4.346083164215088,
      "kl": 0.3271484375,
      "learning_rate": 8.192307692307692e-07,
      "loss": -0.1015,
      "reward": 0.7136711180210114,
      "reward_std": 0.2876356691122055,
      "reward_std_all": 1.1791507601737976,
      "rewards-max/forensics_sinkhorn_reward": 0.9014514684677124,
      "rewards-max/forensics_strict_edge_reward": 0.09006082732230425,
      "rewards-max/forensics_tiou_reward": 0.7949304580688477,
      "rewards/forensics_sinkhorn_reward": 0.4952799081802368,
      "rewards/forensics_strict_edge_reward": 0.004457376431673765,
      "rewards/forensics_tiou_reward": 0.2139338105916977,
      "step": 236
    },
    {
      "completion_length": 27.21875,
      "epoch": 0.728110599078341,
      "grad_norm": 4.565394401550293,
      "kl": 0.30859375,
      "learning_rate": 8.184615384615384e-07,
      "loss": -0.0702,
      "reward": 0.8006795942783356,
      "reward_std": 0.315093956887722,
      "reward_std_all": 1.1454010605812073,
      "rewards-max/forensics_sinkhorn_reward": 0.9180381000041962,
      "rewards-max/forensics_strict_edge_reward": 0.06184565951116383,
      "rewards-max/forensics_tiou_reward": 0.8101256489753723,
      "rewards/forensics_sinkhorn_reward": 0.5163331627845764,
      "rewards/forensics_strict_edge_reward": 0.0034532465069787577,
      "rewards/forensics_tiou_reward": 0.2808931916952133,
      "step": 237
    },
    {
      "completion_length": 22.84375,
      "epoch": 0.7311827956989247,
      "grad_norm": 4.594593524932861,
      "kl": 0.3193359375,
      "learning_rate": 8.176923076923077e-07,
      "loss": -0.1018,
      "reward": 0.9198157787322998,
      "reward_std": 0.35332363843917847,
      "reward_std_all": 1.1139017343521118,
      "rewards-max/forensics_sinkhorn_reward": 0.9805011451244354,
      "rewards-max/forensics_strict_edge_reward": 0.42307616770267487,
      "rewards-max/forensics_tiou_reward": 0.9773870706558228,
      "rewards/forensics_sinkhorn_reward": 0.584371417760849,
      "rewards/forensics_strict_edge_reward": 0.024548371322453022,
      "rewards/forensics_tiou_reward": 0.31089602410793304,
      "step": 238
    },
    {
      "completion_length": 27.5625,
      "epoch": 0.7342549923195084,
      "grad_norm": 4.046504497528076,
      "kl": 0.302734375,
      "learning_rate": 8.169230769230769e-07,
      "loss": -0.1272,
      "reward": 0.9735001921653748,
      "reward_std": 0.4024689793586731,
      "reward_std_all": 1.091102659702301,
      "rewards-max/forensics_sinkhorn_reward": 0.9629843235015869,
      "rewards-max/forensics_strict_edge_reward": 0.21826475113630295,
      "rewards-max/forensics_tiou_reward": 0.934400349855423,
      "rewards/forensics_sinkhorn_reward": 0.5591196715831757,
      "rewards/forensics_strict_edge_reward": 0.015831964556127787,
      "rewards/forensics_tiou_reward": 0.39854857325553894,
      "step": 239
    },
    {
      "completion_length": 26.703125,
      "epoch": 0.7373271889400922,
      "grad_norm": 4.531025409698486,
      "kl": 0.267578125,
      "learning_rate": 8.16153846153846e-07,
      "loss": -0.1465,
      "reward": 0.7276324927806854,
      "reward_std": 0.2968790829181671,
      "reward_std_all": 1.1852651834487915,
      "rewards-max/forensics_sinkhorn_reward": 0.9590588808059692,
      "rewards-max/forensics_strict_edge_reward": 0.09946984797716141,
      "rewards-max/forensics_tiou_reward": 0.9231529235839844,
      "rewards/forensics_sinkhorn_reward": 0.47515174746513367,
      "rewards/forensics_strict_edge_reward": 0.0059577529318630695,
      "rewards/forensics_tiou_reward": 0.24652297794818878,
      "step": 240
    },
    {
      "completion_length": 27.78125,
      "epoch": 0.7403993855606759,
      "grad_norm": 4.329019069671631,
      "kl": 0.380859375,
      "learning_rate": 8.153846153846154e-07,
      "loss": -0.1132,
      "reward": 0.9204943776130676,
      "reward_std": 0.41728678345680237,
      "reward_std_all": 1.1216511130332947,
      "rewards-max/forensics_sinkhorn_reward": 0.9582274258136749,
      "rewards-max/forensics_strict_edge_reward": 0.31122827529907227,
      "rewards-max/forensics_tiou_reward": 0.9175656735897064,
      "rewards/forensics_sinkhorn_reward": 0.5625290870666504,
      "rewards/forensics_strict_edge_reward": 0.024878171971067786,
      "rewards/forensics_tiou_reward": 0.33308714628219604,
      "step": 241
    },
    {
      "completion_length": 24.1875,
      "epoch": 0.7434715821812596,
      "grad_norm": 4.634286880493164,
      "kl": 0.2724609375,
      "learning_rate": 8.146153846153846e-07,
      "loss": -0.0698,
      "reward": 0.8114778399467468,
      "reward_std": 0.30552926659584045,
      "reward_std_all": 1.143004059791565,
      "rewards-max/forensics_sinkhorn_reward": 0.9591270983219147,
      "rewards-max/forensics_strict_edge_reward": 0.2655768468976021,
      "rewards-max/forensics_tiou_reward": 0.9284596145153046,
      "rewards/forensics_sinkhorn_reward": 0.5208335220813751,
      "rewards/forensics_strict_edge_reward": 0.016576615627855062,
      "rewards/forensics_tiou_reward": 0.2740677446126938,
      "step": 242
    },
    {
      "completion_length": 22.765625,
      "epoch": 0.7465437788018433,
      "grad_norm": 5.091953277587891,
      "kl": 0.41650390625,
      "learning_rate": 8.138461538461538e-07,
      "loss": -0.0492,
      "reward": 0.8976596593856812,
      "reward_std": 0.23429382592439651,
      "reward_std_all": 1.0725091099739075,
      "rewards-max/forensics_sinkhorn_reward": 0.9387962222099304,
      "rewards-max/forensics_strict_edge_reward": 0.009369631166919135,
      "rewards-max/forensics_tiou_reward": 0.8805751800537109,
      "rewards/forensics_sinkhorn_reward": 0.5654723942279816,
      "rewards/forensics_strict_edge_reward": 0.0004923111573589267,
      "rewards/forensics_tiou_reward": 0.33169493079185486,
      "step": 243
    },
    {
      "completion_length": 28.203125,
      "epoch": 0.7496159754224271,
      "grad_norm": 3.9763407707214355,
      "kl": 0.23486328125,
      "learning_rate": 8.13076923076923e-07,
      "loss": -0.1223,
      "reward": 0.9014030396938324,
      "reward_std": 0.23155157268047333,
      "reward_std_all": 1.0760093331336975,
      "rewards-max/forensics_sinkhorn_reward": 0.9239047765731812,
      "rewards-max/forensics_strict_edge_reward": 0.20248192735016346,
      "rewards-max/forensics_tiou_reward": 0.848781406879425,
      "rewards/forensics_sinkhorn_reward": 0.5430537760257721,
      "rewards/forensics_strict_edge_reward": 0.011049626162275672,
      "rewards/forensics_tiou_reward": 0.34729963541030884,
      "step": 244
    },
    {
      "completion_length": 24.28125,
      "epoch": 0.7526881720430108,
      "grad_norm": 4.612999439239502,
      "kl": 0.2294921875,
      "learning_rate": 8.123076923076922e-07,
      "loss": -0.1093,
      "reward": 0.8461525738239288,
      "reward_std": 0.359273761510849,
      "reward_std_all": 1.1415750980377197,
      "rewards-max/forensics_sinkhorn_reward": 0.9526768922805786,
      "rewards-max/forensics_strict_edge_reward": 0.1315745897591114,
      "rewards-max/forensics_tiou_reward": 0.908525288105011,
      "rewards/forensics_sinkhorn_reward": 0.5208559334278107,
      "rewards/forensics_strict_edge_reward": 0.011710039805620909,
      "rewards/forensics_tiou_reward": 0.31358662247657776,
      "step": 245
    },
    {
      "completion_length": 22.609375,
      "epoch": 0.7557603686635944,
      "grad_norm": 5.11297607421875,
      "kl": 0.275390625,
      "learning_rate": 8.115384615384615e-07,
      "loss": -0.1455,
      "reward": 0.9705581665039062,
      "reward_std": 0.22934003546833992,
      "reward_std_all": 1.0549458265304565,
      "rewards-max/forensics_sinkhorn_reward": 0.894741415977478,
      "rewards-max/forensics_strict_edge_reward": 0.32682548370212317,
      "rewards-max/forensics_tiou_reward": 0.8257866203784943,
      "rewards/forensics_sinkhorn_reward": 0.5855263769626617,
      "rewards/forensics_strict_edge_reward": 0.022200945430085994,
      "rewards/forensics_tiou_reward": 0.36283084750175476,
      "step": 246
    },
    {
      "completion_length": 24.1875,
      "epoch": 0.7588325652841782,
      "grad_norm": 4.242636203765869,
      "kl": 0.2490234375,
      "learning_rate": 8.107692307692308e-07,
      "loss": -0.1411,
      "reward": 1.0086023211479187,
      "reward_std": 0.36962364614009857,
      "reward_std_all": 1.0737630724906921,
      "rewards-max/forensics_sinkhorn_reward": 0.9791767001152039,
      "rewards-max/forensics_strict_edge_reward": 0.519582062959671,
      "rewards-max/forensics_tiou_reward": 0.9758830070495605,
      "rewards/forensics_sinkhorn_reward": 0.5904597640037537,
      "rewards/forensics_strict_edge_reward": 0.040336912497878075,
      "rewards/forensics_tiou_reward": 0.3778056502342224,
      "step": 247
    },
    {
      "completion_length": 24.28125,
      "epoch": 0.7619047619047619,
      "grad_norm": 4.430719375610352,
      "kl": 0.287109375,
      "learning_rate": 8.1e-07,
      "loss": -0.1067,
      "reward": 0.9919994473457336,
      "reward_std": 0.1902378797531128,
      "reward_std_all": 1.0266950130462646,
      "rewards-max/forensics_sinkhorn_reward": 0.9463971555233002,
      "rewards-max/forensics_strict_edge_reward": 0.11754394881427288,
      "rewards-max/forensics_tiou_reward": 0.8897075951099396,
      "rewards/forensics_sinkhorn_reward": 0.6155721545219421,
      "rewards/forensics_strict_edge_reward": 0.0059258749242872,
      "rewards/forensics_tiou_reward": 0.37050144374370575,
      "step": 248
    },
    {
      "completion_length": 24.453125,
      "epoch": 0.7649769585253456,
      "grad_norm": 4.090976715087891,
      "kl": 0.3779296875,
      "learning_rate": 8.092307692307692e-07,
      "loss": -0.0742,
      "reward": 0.9325572550296783,
      "reward_std": 0.35085681080818176,
      "reward_std_all": 1.102457582950592,
      "rewards-max/forensics_sinkhorn_reward": 0.9714506566524506,
      "rewards-max/forensics_strict_edge_reward": 0.4028276577591896,
      "rewards-max/forensics_tiou_reward": 0.951324462890625,
      "rewards/forensics_sinkhorn_reward": 0.5676758289337158,
      "rewards/forensics_strict_edge_reward": 0.03204940212890506,
      "rewards/forensics_tiou_reward": 0.33283206820487976,
      "step": 249
    },
    {
      "completion_length": 25.265625,
      "epoch": 0.7680491551459293,
      "grad_norm": 4.747076511383057,
      "kl": 0.3876953125,
      "learning_rate": 8.084615384615384e-07,
      "loss": -0.095,
      "reward": 0.8635147213935852,
      "reward_std": 0.1658456064760685,
      "reward_std_all": 1.077807366847992,
      "rewards-max/forensics_sinkhorn_reward": 0.8596821129322052,
      "rewards-max/forensics_strict_edge_reward": 0.03836771659553051,
      "rewards-max/forensics_tiou_reward": 0.7140798270702362,
      "rewards/forensics_sinkhorn_reward": 0.5636745989322662,
      "rewards/forensics_strict_edge_reward": 0.0031664028065279126,
      "rewards/forensics_tiou_reward": 0.2966737002134323,
      "step": 250
    },
    {
      "completion_length": 23.96875,
      "epoch": 0.7711213517665131,
      "grad_norm": 5.03303337097168,
      "kl": 0.3720703125,
      "learning_rate": 8.076923076923077e-07,
      "loss": -0.1252,
      "reward": 0.7403236329555511,
      "reward_std": 0.15503288060426712,
      "reward_std_all": 1.1394474506378174,
      "rewards-max/forensics_sinkhorn_reward": 0.901801198720932,
      "rewards-max/forensics_strict_edge_reward": 0.0023511714534834027,
      "rewards-max/forensics_tiou_reward": 0.7721073925495148,
      "rewards/forensics_sinkhorn_reward": 0.5063533782958984,
      "rewards/forensics_strict_edge_reward": 0.0001537811549496837,
      "rewards/forensics_tiou_reward": 0.2338164746761322,
      "step": 251
    },
    {
      "completion_length": 24.296875,
      "epoch": 0.7741935483870968,
      "grad_norm": 5.229014873504639,
      "kl": 0.31640625,
      "learning_rate": 8.069230769230769e-07,
      "loss": -0.0982,
      "reward": 0.7939391136169434,
      "reward_std": 0.3289753347635269,
      "reward_std_all": 1.1577808856964111,
      "rewards-max/forensics_sinkhorn_reward": 0.9276094138622284,
      "rewards-max/forensics_strict_edge_reward": 0.1507657654583454,
      "rewards-max/forensics_tiou_reward": 0.8523748219013214,
      "rewards/forensics_sinkhorn_reward": 0.513959676027298,
      "rewards/forensics_strict_edge_reward": 0.010263160744216293,
      "rewards/forensics_tiou_reward": 0.26971621811389923,
      "step": 252
    },
    {
      "completion_length": 23.921875,
      "epoch": 0.7772657450076805,
      "grad_norm": 5.120702743530273,
      "kl": 0.330078125,
      "learning_rate": 8.06153846153846e-07,
      "loss": -0.051,
      "reward": 0.947537362575531,
      "reward_std": 0.28371594846248627,
      "reward_std_all": 1.0745900869369507,
      "rewards-max/forensics_sinkhorn_reward": 0.9285009205341339,
      "rewards-max/forensics_strict_edge_reward": 0.5965227484703064,
      "rewards-max/forensics_tiou_reward": 0.8680705428123474,
      "rewards/forensics_sinkhorn_reward": 0.5894558131694794,
      "rewards/forensics_strict_edge_reward": 0.025065378285944462,
      "rewards/forensics_tiou_reward": 0.33301614224910736,
      "step": 253
    },
    {
      "completion_length": 26.03125,
      "epoch": 0.7803379416282642,
      "grad_norm": 3.837529182434082,
      "kl": 0.3203125,
      "learning_rate": 8.053846153846154e-07,
      "loss": -0.1456,
      "reward": 0.8037522435188293,
      "reward_std": 0.15057943761348724,
      "reward_std_all": 1.1062230467796326,
      "rewards-max/forensics_sinkhorn_reward": 0.8763388395309448,
      "rewards-max/forensics_strict_edge_reward": 0.17389732878655195,
      "rewards-max/forensics_tiou_reward": 0.7193983197212219,
      "rewards/forensics_sinkhorn_reward": 0.5307757556438446,
      "rewards/forensics_strict_edge_reward": 0.007457268395228311,
      "rewards/forensics_tiou_reward": 0.2655192241072655,
      "step": 254
    },
    {
      "completion_length": 25.59375,
      "epoch": 0.783410138248848,
      "grad_norm": 4.4849042892456055,
      "kl": 0.40625,
      "learning_rate": 8.046153846153846e-07,
      "loss": -0.0671,
      "reward": 0.9348688721656799,
      "reward_std": 0.3623153418302536,
      "reward_std_all": 1.0905067920684814,
      "rewards-max/forensics_sinkhorn_reward": 0.9298736155033112,
      "rewards-max/forensics_strict_edge_reward": 0.3259113132953644,
      "rewards-max/forensics_tiou_reward": 0.8644978702068329,
      "rewards/forensics_sinkhorn_reward": 0.5718681812286377,
      "rewards/forensics_strict_edge_reward": 0.017881023697555065,
      "rewards/forensics_tiou_reward": 0.3451196849346161,
      "step": 255
    },
    {
      "completion_length": 29.296875,
      "epoch": 0.7864823348694316,
      "grad_norm": 4.796127796173096,
      "kl": 0.2216796875,
      "learning_rate": 8.038461538461538e-07,
      "loss": -0.0803,
      "reward": 0.8542866408824921,
      "reward_std": 0.31971028447151184,
      "reward_std_all": 1.1217607855796814,
      "rewards-max/forensics_sinkhorn_reward": 0.9180476665496826,
      "rewards-max/forensics_strict_edge_reward": 0.21814831718802452,
      "rewards-max/forensics_tiou_reward": 0.8209607005119324,
      "rewards/forensics_sinkhorn_reward": 0.5341042578220367,
      "rewards/forensics_strict_edge_reward": 0.012859964277595282,
      "rewards/forensics_tiou_reward": 0.3073224276304245,
      "step": 256
    },
    {
      "completion_length": 26.9375,
      "epoch": 0.7895545314900153,
      "grad_norm": 4.923451900482178,
      "kl": 0.24755859375,
      "learning_rate": 8.03076923076923e-07,
      "loss": -0.0855,
      "reward": 0.7759303152561188,
      "reward_std": 0.23504352569580078,
      "reward_std_all": 1.1417972445487976,
      "rewards-max/forensics_sinkhorn_reward": 0.9223407506942749,
      "rewards-max/forensics_strict_edge_reward": 0.15220642706844956,
      "rewards-max/forensics_tiou_reward": 0.8257465362548828,
      "rewards/forensics_sinkhorn_reward": 0.4583190977573395,
      "rewards/forensics_strict_edge_reward": 0.006392434297595173,
      "rewards/forensics_tiou_reward": 0.311218798160553,
      "step": 257
    },
    {
      "completion_length": 27.390625,
      "epoch": 0.7926267281105991,
      "grad_norm": 4.138204574584961,
      "kl": 0.25341796875,
      "learning_rate": 8.023076923076922e-07,
      "loss": -0.1317,
      "reward": 0.8667551875114441,
      "reward_std": 0.3726358935236931,
      "reward_std_all": 1.1453104615211487,
      "rewards-max/forensics_sinkhorn_reward": 0.9486428499221802,
      "rewards-max/forensics_strict_edge_reward": 0.36152931302785873,
      "rewards-max/forensics_tiou_reward": 0.9038645327091217,
      "rewards/forensics_sinkhorn_reward": 0.5214844346046448,
      "rewards/forensics_strict_edge_reward": 0.030623536091297865,
      "rewards/forensics_tiou_reward": 0.31464728713035583,
      "step": 258
    },
    {
      "completion_length": 28.734375,
      "epoch": 0.7956989247311828,
      "grad_norm": 4.97873592376709,
      "kl": 0.3359375,
      "learning_rate": 8.015384615384615e-07,
      "loss": -0.0903,
      "reward": 0.741553008556366,
      "reward_std": 0.334484800696373,
      "reward_std_all": 1.1933629512786865,
      "rewards-max/forensics_sinkhorn_reward": 0.962056577205658,
      "rewards-max/forensics_strict_edge_reward": 0.43677301704883575,
      "rewards-max/forensics_tiou_reward": 0.9296753406524658,
      "rewards/forensics_sinkhorn_reward": 0.4483480006456375,
      "rewards/forensics_strict_edge_reward": 0.029847711324691772,
      "rewards/forensics_tiou_reward": 0.2633573114871979,
      "step": 259
    },
    {
      "completion_length": 26.53125,
      "epoch": 0.7987711213517665,
      "grad_norm": 4.888911247253418,
      "kl": 0.26953125,
      "learning_rate": 8.007692307692308e-07,
      "loss": -0.0856,
      "reward": 0.8097848296165466,
      "reward_std": 0.1797875463962555,
      "reward_std_all": 1.1110671758651733,
      "rewards-max/forensics_sinkhorn_reward": 0.9183475375175476,
      "rewards-max/forensics_strict_edge_reward": 0.07489891932345927,
      "rewards-max/forensics_tiou_reward": 0.7981404662132263,
      "rewards/forensics_sinkhorn_reward": 0.524253785610199,
      "rewards/forensics_strict_edge_reward": 0.004721620352938771,
      "rewards/forensics_tiou_reward": 0.2808094322681427,
      "step": 260
    },
    {
      "completion_length": 24.71875,
      "epoch": 0.8018433179723502,
      "grad_norm": 4.625534534454346,
      "kl": 0.3115234375,
      "learning_rate": 8e-07,
      "loss": -0.1049,
      "reward": 0.9328953325748444,
      "reward_std": 0.29339128732681274,
      "reward_std_all": 1.0752549767494202,
      "rewards-max/forensics_sinkhorn_reward": 0.9615466892719269,
      "rewards-max/forensics_strict_edge_reward": 0.2224796563386917,
      "rewards-max/forensics_tiou_reward": 0.9240294992923737,
      "rewards/forensics_sinkhorn_reward": 0.5855396687984467,
      "rewards/forensics_strict_edge_reward": 0.010795491049066186,
      "rewards/forensics_tiou_reward": 0.3365601599216461,
      "step": 261
    },
    {
      "completion_length": 23.640625,
      "epoch": 0.804915514592934,
      "grad_norm": 4.621260166168213,
      "kl": 0.32568359375,
      "learning_rate": 7.992307692307692e-07,
      "loss": -0.0667,
      "reward": 0.8269758522510529,
      "reward_std": 0.2327316477894783,
      "reward_std_all": 1.1116633415222168,
      "rewards-max/forensics_sinkhorn_reward": 0.8705346584320068,
      "rewards-max/forensics_strict_edge_reward": 0.07101825275458395,
      "rewards-max/forensics_tiou_reward": 0.7037323117256165,
      "rewards/forensics_sinkhorn_reward": 0.5421695113182068,
      "rewards/forensics_strict_edge_reward": 0.003184010085533373,
      "rewards/forensics_tiou_reward": 0.28162233531475067,
      "step": 262
    },
    {
      "completion_length": 26.890625,
      "epoch": 0.8079877112135176,
      "grad_norm": 3.780357837677002,
      "kl": 0.3212890625,
      "learning_rate": 7.984615384615384e-07,
      "loss": -0.0864,
      "reward": 0.9284782707691193,
      "reward_std": 0.29387012869119644,
      "reward_std_all": 1.079995036125183,
      "rewards-max/forensics_sinkhorn_reward": 0.9283607006072998,
      "rewards-max/forensics_strict_edge_reward": 0.06965029332786798,
      "rewards-max/forensics_tiou_reward": 0.8387004137039185,
      "rewards/forensics_sinkhorn_reward": 0.5739695131778717,
      "rewards/forensics_strict_edge_reward": 0.004028983850730583,
      "rewards/forensics_tiou_reward": 0.35047978162765503,
      "step": 263
    },
    {
      "completion_length": 28.078125,
      "epoch": 0.8110599078341014,
      "grad_norm": 4.009171962738037,
      "kl": 0.3515625,
      "learning_rate": 7.976923076923077e-07,
      "loss": -0.0551,
      "reward": 0.7935514450073242,
      "reward_std": 0.25635358691215515,
      "reward_std_all": 1.1370488405227661,
      "rewards-max/forensics_sinkhorn_reward": 0.9302574098110199,
      "rewards-max/forensics_strict_edge_reward": 0.06201888539362699,
      "rewards-max/forensics_tiou_reward": 0.8489102423191071,
      "rewards/forensics_sinkhorn_reward": 0.4914189279079437,
      "rewards/forensics_strict_edge_reward": 0.0027668463153531775,
      "rewards/forensics_tiou_reward": 0.29936572909355164,
      "step": 264
    },
    {
      "completion_length": 25.875,
      "epoch": 0.8141321044546851,
      "grad_norm": 3.901254892349243,
      "kl": 0.26171875,
      "learning_rate": 7.969230769230769e-07,
      "loss": -0.1242,
      "reward": 0.6514708697795868,
      "reward_std": 0.2561115622520447,
      "reward_std_all": 1.2073689103126526,
      "rewards-max/forensics_sinkhorn_reward": 0.9194429218769073,
      "rewards-max/forensics_strict_edge_reward": 0.21718691289424896,
      "rewards-max/forensics_tiou_reward": 0.8287798762321472,
      "rewards/forensics_sinkhorn_reward": 0.4477752149105072,
      "rewards/forensics_strict_edge_reward": 0.009063505567610264,
      "rewards/forensics_tiou_reward": 0.1946321576833725,
      "step": 265
    },
    {
      "completion_length": 27.703125,
      "epoch": 0.8172043010752689,
      "grad_norm": 5.920890808105469,
      "kl": 0.3359375,
      "learning_rate": 7.96153846153846e-07,
      "loss": -0.1372,
      "reward": 0.8751713633537292,
      "reward_std": 0.3825642466545105,
      "reward_std_all": 1.137131154537201,
      "rewards-max/forensics_sinkhorn_reward": 0.9749702215194702,
      "rewards-max/forensics_strict_edge_reward": 0.396661214530468,
      "rewards-max/forensics_tiou_reward": 0.9548647701740265,
      "rewards/forensics_sinkhorn_reward": 0.5309174656867981,
      "rewards/forensics_strict_edge_reward": 0.01834943052381277,
      "rewards/forensics_tiou_reward": 0.3259044885635376,
      "step": 266
    },
    {
      "completion_length": 36.078125,
      "epoch": 0.8202764976958525,
      "grad_norm": 3.403386116027832,
      "kl": 0.263671875,
      "learning_rate": 7.953846153846154e-07,
      "loss": -0.1057,
      "reward": 0.6403444707393646,
      "reward_std": 0.213466577231884,
      "reward_std_all": 1.2044499516487122,
      "rewards-max/forensics_sinkhorn_reward": 0.7365062534809113,
      "rewards-max/forensics_strict_edge_reward": 0.0988439826760441,
      "rewards-max/forensics_tiou_reward": 0.708213284611702,
      "rewards/forensics_sinkhorn_reward": 0.39987029135227203,
      "rewards/forensics_strict_edge_reward": 0.007169321703258902,
      "rewards/forensics_tiou_reward": 0.23330488801002502,
      "step": 267
    },
    {
      "completion_length": 28.34375,
      "epoch": 0.8233486943164362,
      "grad_norm": 4.95294713973999,
      "kl": 0.291015625,
      "learning_rate": 7.946153846153846e-07,
      "loss": -0.1107,
      "reward": 0.6669174432754517,
      "reward_std": 0.22904933243989944,
      "reward_std_all": 1.1966350078582764,
      "rewards-max/forensics_sinkhorn_reward": 0.8820779025554657,
      "rewards-max/forensics_strict_edge_reward": 0.1912420685403049,
      "rewards-max/forensics_tiou_reward": 0.7432375848293304,
      "rewards/forensics_sinkhorn_reward": 0.4277718663215637,
      "rewards/forensics_strict_edge_reward": 0.008175415627192706,
      "rewards/forensics_tiou_reward": 0.23097018897533417,
      "step": 268
    },
    {
      "completion_length": 29.421875,
      "epoch": 0.82642089093702,
      "grad_norm": 5.559000492095947,
      "kl": 0.3291015625,
      "learning_rate": 7.938461538461538e-07,
      "loss": -0.085,
      "reward": 0.8460853397846222,
      "reward_std": 0.28655147552490234,
      "reward_std_all": 1.1253414154052734,
      "rewards-max/forensics_sinkhorn_reward": 0.9431018233299255,
      "rewards-max/forensics_strict_edge_reward": 0.1496355626732111,
      "rewards-max/forensics_tiou_reward": 0.8759882748126984,
      "rewards/forensics_sinkhorn_reward": 0.4802332669496536,
      "rewards/forensics_strict_edge_reward": 0.01585628418251872,
      "rewards/forensics_tiou_reward": 0.34999576210975647,
      "step": 269
    },
    {
      "completion_length": 27.828125,
      "epoch": 0.8294930875576036,
      "grad_norm": 4.249082088470459,
      "kl": 0.2666015625,
      "learning_rate": 7.930769230769231e-07,
      "loss": -0.1106,
      "reward": 0.8879426717758179,
      "reward_std": 0.3118748962879181,
      "reward_std_all": 1.1007386445999146,
      "rewards-max/forensics_sinkhorn_reward": 0.9679370820522308,
      "rewards-max/forensics_strict_edge_reward": 0.1505085453391075,
      "rewards-max/forensics_tiou_reward": 0.9452257454395294,
      "rewards/forensics_sinkhorn_reward": 0.5321611613035202,
      "rewards/forensics_strict_edge_reward": 0.013169094454497099,
      "rewards/forensics_tiou_reward": 0.34261247515678406,
      "step": 270
    },
    {
      "completion_length": 27.078125,
      "epoch": 0.8325652841781874,
      "grad_norm": 4.205349445343018,
      "kl": 0.3212890625,
      "learning_rate": 7.923076923076922e-07,
      "loss": -0.1496,
      "reward": 0.8336701989173889,
      "reward_std": 0.3445402532815933,
      "reward_std_all": 1.1601451635360718,
      "rewards-max/forensics_sinkhorn_reward": 0.9653686285018921,
      "rewards-max/forensics_strict_edge_reward": 0.5238595008850098,
      "rewards-max/forensics_tiou_reward": 0.9446300864219666,
      "rewards/forensics_sinkhorn_reward": 0.5219338834285736,
      "rewards/forensics_strict_edge_reward": 0.032829614356160164,
      "rewards/forensics_tiou_reward": 0.2789067476987839,
      "step": 271
    },
    {
      "completion_length": 26.484375,
      "epoch": 0.8356374807987711,
      "grad_norm": 3.9184370040893555,
      "kl": 0.31591796875,
      "learning_rate": 7.915384615384615e-07,
      "loss": -0.0724,
      "reward": 1.0620926022529602,
      "reward_std": 0.3440081477165222,
      "reward_std_all": 1.0228005051612854,
      "rewards-max/forensics_sinkhorn_reward": 0.9628486633300781,
      "rewards-max/forensics_strict_edge_reward": 0.07930911635048687,
      "rewards-max/forensics_tiou_reward": 0.9186752438545227,
      "rewards/forensics_sinkhorn_reward": 0.635153740644455,
      "rewards/forensics_strict_edge_reward": 0.005604137571936008,
      "rewards/forensics_tiou_reward": 0.42133474349975586,
      "step": 272
    },
    {
      "completion_length": 25.484375,
      "epoch": 0.8387096774193549,
      "grad_norm": 5.650501728057861,
      "kl": 0.388671875,
      "learning_rate": 7.907692307692307e-07,
      "loss": -0.0511,
      "reward": 0.8831664025783539,
      "reward_std": 0.3034338355064392,
      "reward_std_all": 1.1019103527069092,
      "rewards-max/forensics_sinkhorn_reward": 0.9338535666465759,
      "rewards-max/forensics_strict_edge_reward": 0.1391203124076128,
      "rewards-max/forensics_tiou_reward": 0.8664791285991669,
      "rewards/forensics_sinkhorn_reward": 0.5421087145805359,
      "rewards/forensics_strict_edge_reward": 0.008649447234347463,
      "rewards/forensics_tiou_reward": 0.33240823447704315,
      "step": 273
    },
    {
      "completion_length": 28.0625,
      "epoch": 0.8417818740399385,
      "grad_norm": 4.975771903991699,
      "kl": 0.4072265625,
      "learning_rate": 7.9e-07,
      "loss": -0.1354,
      "reward": 0.7898664474487305,
      "reward_std": 0.28443264216184616,
      "reward_std_all": 1.1566558480262756,
      "rewards-max/forensics_sinkhorn_reward": 0.9570043385028839,
      "rewards-max/forensics_strict_edge_reward": 0.15804020687937737,
      "rewards-max/forensics_tiou_reward": 0.9166755378246307,
      "rewards/forensics_sinkhorn_reward": 0.4931506812572479,
      "rewards/forensics_strict_edge_reward": 0.011462456779554486,
      "rewards/forensics_tiou_reward": 0.2852533757686615,
      "step": 274
    },
    {
      "completion_length": 25.640625,
      "epoch": 0.8448540706605223,
      "grad_norm": 4.769527912139893,
      "kl": 0.2890625,
      "learning_rate": 7.892307692307692e-07,
      "loss": -0.0936,
      "reward": 0.9931166172027588,
      "reward_std": 0.1871299333870411,
      "reward_std_all": 1.026325285434723,
      "rewards-max/forensics_sinkhorn_reward": 0.9067807197570801,
      "rewards-max/forensics_strict_edge_reward": 0.03202316735405475,
      "rewards-max/forensics_tiou_reward": 0.8040892779827118,
      "rewards/forensics_sinkhorn_reward": 0.6131389141082764,
      "rewards/forensics_strict_edge_reward": 0.001987567258765921,
      "rewards/forensics_tiou_reward": 0.37799011170864105,
      "step": 275
    },
    {
      "completion_length": 25.578125,
      "epoch": 0.847926267281106,
      "grad_norm": 4.6387505531311035,
      "kl": 0.375,
      "learning_rate": 7.884615384615384e-07,
      "loss": -0.1112,
      "reward": 1.0012320280075073,
      "reward_std": 0.36768826842308044,
      "reward_std_all": 1.07891845703125,
      "rewards-max/forensics_sinkhorn_reward": 0.9640684127807617,
      "rewards-max/forensics_strict_edge_reward": 0.5067096650600433,
      "rewards-max/forensics_tiou_reward": 0.9411028921604156,
      "rewards/forensics_sinkhorn_reward": 0.5984495282173157,
      "rewards/forensics_strict_edge_reward": 0.05474369041621685,
      "rewards/forensics_tiou_reward": 0.34803882241249084,
      "step": 276
    },
    {
      "completion_length": 27.15625,
      "epoch": 0.8509984639016898,
      "grad_norm": 4.940577507019043,
      "kl": 0.3212890625,
      "learning_rate": 7.876923076923077e-07,
      "loss": -0.1128,
      "reward": 0.9212069511413574,
      "reward_std": 0.2895997315645218,
      "reward_std_all": 1.093327820301056,
      "rewards-max/forensics_sinkhorn_reward": 0.9579691886901855,
      "rewards-max/forensics_strict_edge_reward": 0.349606916308403,
      "rewards-max/forensics_tiou_reward": 0.9238159656524658,
      "rewards/forensics_sinkhorn_reward": 0.5839072465896606,
      "rewards/forensics_strict_edge_reward": 0.026075984351336956,
      "rewards/forensics_tiou_reward": 0.31122376024723053,
      "step": 277
    },
    {
      "completion_length": 26.734375,
      "epoch": 0.8540706605222734,
      "grad_norm": 4.783517360687256,
      "kl": 0.23095703125,
      "learning_rate": 7.869230769230769e-07,
      "loss": -0.1055,
      "reward": 0.8223170042037964,
      "reward_std": 0.21746106445789337,
      "reward_std_all": 1.1063839793205261,
      "rewards-max/forensics_sinkhorn_reward": 0.8347434401512146,
      "rewards-max/forensics_strict_edge_reward": 0.23886488907010062,
      "rewards-max/forensics_tiou_reward": 0.6262710392475128,
      "rewards/forensics_sinkhorn_reward": 0.5436070710420609,
      "rewards/forensics_strict_edge_reward": 0.011273073322172422,
      "rewards/forensics_tiou_reward": 0.2674368619918823,
      "step": 278
    },
    {
      "completion_length": 23.3125,
      "epoch": 0.8571428571428571,
      "grad_norm": 4.449802398681641,
      "kl": 0.34765625,
      "learning_rate": 7.861538461538461e-07,
      "loss": -0.0624,
      "reward": 0.8777587413787842,
      "reward_std": 0.2248949408531189,
      "reward_std_all": 1.0893449187278748,
      "rewards-max/forensics_sinkhorn_reward": 0.9323569536209106,
      "rewards-max/forensics_strict_edge_reward": 0.169258713722229,
      "rewards-max/forensics_tiou_reward": 0.8585125505924225,
      "rewards/forensics_sinkhorn_reward": 0.563774585723877,
      "rewards/forensics_strict_edge_reward": 0.008534598629921675,
      "rewards/forensics_tiou_reward": 0.30544960498809814,
      "step": 279
    },
    {
      "completion_length": 26.34375,
      "epoch": 0.8602150537634409,
      "grad_norm": 4.683992385864258,
      "kl": 0.2802734375,
      "learning_rate": 7.853846153846153e-07,
      "loss": -0.1136,
      "reward": 1.0519801378250122,
      "reward_std": 0.41262345016002655,
      "reward_std_all": 1.0604192018508911,
      "rewards-max/forensics_sinkhorn_reward": 0.9741476476192474,
      "rewards-max/forensics_strict_edge_reward": 0.42794930934906006,
      "rewards-max/forensics_tiou_reward": 0.9631880223751068,
      "rewards/forensics_sinkhorn_reward": 0.615602046251297,
      "rewards/forensics_strict_edge_reward": 0.03725466504693031,
      "rewards/forensics_tiou_reward": 0.3991233557462692,
      "step": 280
    },
    {
      "completion_length": 23.5625,
      "epoch": 0.8632872503840245,
      "grad_norm": 4.852132320404053,
      "kl": 0.298828125,
      "learning_rate": 7.846153846153846e-07,
      "loss": -0.1066,
      "reward": 1.0893646478652954,
      "reward_std": 0.3016444519162178,
      "reward_std_all": 1.0076687633991241,
      "rewards-max/forensics_sinkhorn_reward": 0.9747424125671387,
      "rewards-max/forensics_strict_edge_reward": 0.716046154499054,
      "rewards-max/forensics_tiou_reward": 0.9664590358734131,
      "rewards/forensics_sinkhorn_reward": 0.6492708325386047,
      "rewards/forensics_strict_edge_reward": 0.03135205712169409,
      "rewards/forensics_tiou_reward": 0.4087417721748352,
      "step": 281
    },
    {
      "completion_length": 23.90625,
      "epoch": 0.8663594470046083,
      "grad_norm": 4.905045509338379,
      "kl": 0.298828125,
      "learning_rate": 7.838461538461539e-07,
      "loss": -0.0862,
      "reward": 0.8922614455223083,
      "reward_std": 0.2658926621079445,
      "reward_std_all": 1.0973901152610779,
      "rewards-max/forensics_sinkhorn_reward": 0.8970165252685547,
      "rewards-max/forensics_strict_edge_reward": 0.24028923535661306,
      "rewards-max/forensics_tiou_reward": 0.7604039013385773,
      "rewards/forensics_sinkhorn_reward": 0.5545039772987366,
      "rewards/forensics_strict_edge_reward": 0.02531155219367065,
      "rewards/forensics_tiou_reward": 0.3124459311366081,
      "step": 282
    },
    {
      "completion_length": 27.625,
      "epoch": 0.869431643625192,
      "grad_norm": 4.843585968017578,
      "kl": 0.3583984375,
      "learning_rate": 7.830769230769231e-07,
      "loss": -0.0649,
      "reward": 0.9060111045837402,
      "reward_std": 0.34070080518722534,
      "reward_std_all": 1.1207736730575562,
      "rewards-max/forensics_sinkhorn_reward": 0.9811470210552216,
      "rewards-max/forensics_strict_edge_reward": 0.447728231549263,
      "rewards-max/forensics_tiou_reward": 0.9778619110584259,
      "rewards/forensics_sinkhorn_reward": 0.5060573071241379,
      "rewards/forensics_strict_edge_reward": 0.03788466192781925,
      "rewards/forensics_tiou_reward": 0.36206911504268646,
      "step": 283
    },
    {
      "completion_length": 23.9375,
      "epoch": 0.8725038402457758,
      "grad_norm": 4.238224029541016,
      "kl": 0.3369140625,
      "learning_rate": 7.823076923076922e-07,
      "loss": -0.0363,
      "reward": 0.9888312816619873,
      "reward_std": 0.30096281319856644,
      "reward_std_all": 1.0607223510742188,
      "rewards-max/forensics_sinkhorn_reward": 0.9644052386283875,
      "rewards-max/forensics_strict_edge_reward": 0.4231441020965576,
      "rewards-max/forensics_tiou_reward": 0.9406760931015015,
      "rewards/forensics_sinkhorn_reward": 0.6107066571712494,
      "rewards/forensics_strict_edge_reward": 0.024593293201178312,
      "rewards/forensics_tiou_reward": 0.3535313308238983,
      "step": 284
    },
    {
      "completion_length": 26.25,
      "epoch": 0.8755760368663594,
      "grad_norm": 4.903867244720459,
      "kl": 0.3388671875,
      "learning_rate": 7.815384615384615e-07,
      "loss": -0.156,
      "reward": 0.9004891812801361,
      "reward_std": 0.1483253836631775,
      "reward_std_all": 1.06354421377182,
      "rewards-max/forensics_sinkhorn_reward": 0.9383205473423004,
      "rewards-max/forensics_strict_edge_reward": 0.046185266226530075,
      "rewards-max/forensics_tiou_reward": 0.8765358030796051,
      "rewards/forensics_sinkhorn_reward": 0.5924093425273895,
      "rewards/forensics_strict_edge_reward": 0.002184803714044392,
      "rewards/forensics_tiou_reward": 0.30589500069618225,
      "step": 285
    },
    {
      "completion_length": 24.671875,
      "epoch": 0.8786482334869432,
      "grad_norm": 4.6122941970825195,
      "kl": 0.3076171875,
      "learning_rate": 7.807692307692307e-07,
      "loss": -0.0565,
      "reward": 0.7729721665382385,
      "reward_std": 0.2837854027748108,
      "reward_std_all": 1.1522504091262817,
      "rewards-max/forensics_sinkhorn_reward": 0.9541668891906738,
      "rewards-max/forensics_strict_edge_reward": 0.2851364314556122,
      "rewards-max/forensics_tiou_reward": 0.9088840484619141,
      "rewards/forensics_sinkhorn_reward": 0.49087417125701904,
      "rewards/forensics_strict_edge_reward": 0.014126032590866089,
      "rewards/forensics_tiou_reward": 0.267971932888031,
      "step": 286
    },
    {
      "completion_length": 24.34375,
      "epoch": 0.8817204301075269,
      "grad_norm": 4.8898210525512695,
      "kl": 0.34765625,
      "learning_rate": 7.799999999999999e-07,
      "loss": -0.1475,
      "reward": 0.9124596118927002,
      "reward_std": 0.3596213161945343,
      "reward_std_all": 1.1144522428512573,
      "rewards-max/forensics_sinkhorn_reward": 0.9780560731887817,
      "rewards-max/forensics_strict_edge_reward": 0.37899377197027206,
      "rewards-max/forensics_tiou_reward": 0.9697883129119873,
      "rewards/forensics_sinkhorn_reward": 0.5850443989038467,
      "rewards/forensics_strict_edge_reward": 0.019970493856817484,
      "rewards/forensics_tiou_reward": 0.3074447736144066,
      "step": 287
    },
    {
      "completion_length": 27.484375,
      "epoch": 0.8847926267281107,
      "grad_norm": 4.3585968017578125,
      "kl": 0.3896484375,
      "learning_rate": 7.792307692307693e-07,
      "loss": -0.1093,
      "reward": 0.8018427193164825,
      "reward_std": 0.17197918891906738,
      "reward_std_all": 1.1131932735443115,
      "rewards-max/forensics_sinkhorn_reward": 0.8846079707145691,
      "rewards-max/forensics_strict_edge_reward": 0.21322976669762284,
      "rewards-max/forensics_tiou_reward": 0.7578215003013611,
      "rewards/forensics_sinkhorn_reward": 0.5308541357517242,
      "rewards/forensics_strict_edge_reward": 0.010625380353303626,
      "rewards/forensics_tiou_reward": 0.26036323606967926,
      "step": 288
    },
    {
      "completion_length": 24.015625,
      "epoch": 0.8878648233486943,
      "grad_norm": 4.845871448516846,
      "kl": 0.33056640625,
      "learning_rate": 7.784615384615384e-07,
      "loss": -0.1014,
      "reward": 0.8763305842876434,
      "reward_std": 0.2679719403386116,
      "reward_std_all": 1.1197728514671326,
      "rewards-max/forensics_sinkhorn_reward": 0.9809812009334564,
      "rewards-max/forensics_strict_edge_reward": 0.6065583229064941,
      "rewards-max/forensics_tiou_reward": 0.9755196273326874,
      "rewards/forensics_sinkhorn_reward": 0.5468646734952927,
      "rewards/forensics_strict_edge_reward": 0.027378917671740055,
      "rewards/forensics_tiou_reward": 0.30208709090948105,
      "step": 289
    },
    {
      "completion_length": 29.9375,
      "epoch": 0.890937019969278,
      "grad_norm": 3.4287831783294678,
      "kl": 0.3095703125,
      "learning_rate": 7.776923076923077e-07,
      "loss": -0.1063,
      "reward": 0.9816868901252747,
      "reward_std": 0.2860771417617798,
      "reward_std_all": 1.0625211596488953,
      "rewards-max/forensics_sinkhorn_reward": 0.9258279800415039,
      "rewards-max/forensics_strict_edge_reward": 0.3093508780002594,
      "rewards-max/forensics_tiou_reward": 0.8554763197898865,
      "rewards/forensics_sinkhorn_reward": 0.5569173097610474,
      "rewards/forensics_strict_edge_reward": 0.030293859541416168,
      "rewards/forensics_tiou_reward": 0.3944757729768753,
      "step": 290
    },
    {
      "completion_length": 27.0,
      "epoch": 0.8940092165898618,
      "grad_norm": 5.105864524841309,
      "kl": 0.275390625,
      "learning_rate": 7.769230769230769e-07,
      "loss": -0.0093,
      "reward": 1.0233350098133087,
      "reward_std": 0.3299413174390793,
      "reward_std_all": 1.0462568998336792,
      "rewards-max/forensics_sinkhorn_reward": 0.9743745923042297,
      "rewards-max/forensics_strict_edge_reward": 0.403884619474411,
      "rewards-max/forensics_tiou_reward": 0.9665978252887726,
      "rewards/forensics_sinkhorn_reward": 0.5983610153198242,
      "rewards/forensics_strict_edge_reward": 0.021846733056008816,
      "rewards/forensics_tiou_reward": 0.4031272381544113,
      "step": 291
    },
    {
      "completion_length": 23.515625,
      "epoch": 0.8970814132104454,
      "grad_norm": 4.898962497711182,
      "kl": 0.28515625,
      "learning_rate": 7.761538461538461e-07,
      "loss": -0.1002,
      "reward": 0.9290638864040375,
      "reward_std": 0.3215360641479492,
      "reward_std_all": 1.0989359021186829,
      "rewards-max/forensics_sinkhorn_reward": 0.9402048289775848,
      "rewards-max/forensics_strict_edge_reward": 0.27504963614046574,
      "rewards-max/forensics_tiou_reward": 0.8764608502388,
      "rewards/forensics_sinkhorn_reward": 0.5517770051956177,
      "rewards/forensics_strict_edge_reward": 0.03443750843871385,
      "rewards/forensics_tiou_reward": 0.34284935891628265,
      "step": 292
    },
    {
      "completion_length": 27.3125,
      "epoch": 0.9001536098310292,
      "grad_norm": 3.6127207279205322,
      "kl": 0.2841796875,
      "learning_rate": 7.753846153846153e-07,
      "loss": -0.0429,
      "reward": 0.9093067944049835,
      "reward_std": 0.23445269465446472,
      "reward_std_all": 1.07175612449646,
      "rewards-max/forensics_sinkhorn_reward": 0.8765429258346558,
      "rewards-max/forensics_strict_edge_reward": 0.27456915807852056,
      "rewards-max/forensics_tiou_reward": 0.750227153301239,
      "rewards/forensics_sinkhorn_reward": 0.5673338174819946,
      "rewards/forensics_strict_edge_reward": 0.012322681116529566,
      "rewards/forensics_tiou_reward": 0.32965028285980225,
      "step": 293
    },
    {
      "completion_length": 29.40625,
      "epoch": 0.9032258064516129,
      "grad_norm": 5.183175086975098,
      "kl": 0.3251953125,
      "learning_rate": 7.746153846153845e-07,
      "loss": -0.1749,
      "reward": 0.6932768225669861,
      "reward_std": 0.22508858144283295,
      "reward_std_all": 1.1762155890464783,
      "rewards-max/forensics_sinkhorn_reward": 0.8676880598068237,
      "rewards-max/forensics_strict_edge_reward": 0.016320371069014072,
      "rewards-max/forensics_tiou_reward": 0.7134474217891693,
      "rewards/forensics_sinkhorn_reward": 0.43102777004241943,
      "rewards/forensics_strict_edge_reward": 0.0010949831630568951,
      "rewards/forensics_tiou_reward": 0.2611541450023651,
      "step": 294
    },
    {
      "completion_length": 25.140625,
      "epoch": 0.9062980030721967,
      "grad_norm": 4.143625736236572,
      "kl": 0.3076171875,
      "learning_rate": 7.738461538461539e-07,
      "loss": -0.0972,
      "reward": 0.9124976098537445,
      "reward_std": 0.2145184427499771,
      "reward_std_all": 1.077205240726471,
      "rewards-max/forensics_sinkhorn_reward": 0.9657674729824066,
      "rewards-max/forensics_strict_edge_reward": 0.19450300931930542,
      "rewards-max/forensics_tiou_reward": 0.9464421570301056,
      "rewards/forensics_sinkhorn_reward": 0.5589975118637085,
      "rewards/forensics_strict_edge_reward": 0.011081448756158352,
      "rewards/forensics_tiou_reward": 0.34241868555545807,
      "step": 295
    },
    {
      "completion_length": 25.28125,
      "epoch": 0.9093701996927803,
      "grad_norm": 4.623172283172607,
      "kl": 0.318359375,
      "learning_rate": 7.730769230769231e-07,
      "loss": -0.0676,
      "reward": 0.7799955904483795,
      "reward_std": 0.19868887215852737,
      "reward_std_all": 1.1370510458946228,
      "rewards-max/forensics_sinkhorn_reward": 0.965482234954834,
      "rewards-max/forensics_strict_edge_reward": 0.22504481673240662,
      "rewards-max/forensics_tiou_reward": 0.9318813979625702,
      "rewards/forensics_sinkhorn_reward": 0.5033314526081085,
      "rewards/forensics_strict_edge_reward": 0.015185816213488579,
      "rewards/forensics_tiou_reward": 0.26147831231355667,
      "step": 296
    },
    {
      "completion_length": 27.828125,
      "epoch": 0.9124423963133641,
      "grad_norm": 6.0742340087890625,
      "kl": 0.2734375,
      "learning_rate": 7.723076923076923e-07,
      "loss": -0.0852,
      "reward": 0.7482559084892273,
      "reward_std": 0.25245320424437523,
      "reward_std_all": 1.161653459072113,
      "rewards-max/forensics_sinkhorn_reward": 0.8578179478645325,
      "rewards-max/forensics_strict_edge_reward": 0.4093669630426575,
      "rewards-max/forensics_tiou_reward": 0.7799372971057892,
      "rewards/forensics_sinkhorn_reward": 0.45965906977653503,
      "rewards/forensics_strict_edge_reward": 0.026195664501472038,
      "rewards/forensics_tiou_reward": 0.26240120828151703,
      "step": 297
    },
    {
      "completion_length": 28.1875,
      "epoch": 0.9155145929339478,
      "grad_norm": 4.187925338745117,
      "kl": 0.3291015625,
      "learning_rate": 7.715384615384615e-07,
      "loss": -0.0983,
      "reward": 0.9330911934375763,
      "reward_std": 0.1945476457476616,
      "reward_std_all": 1.0628196001052856,
      "rewards-max/forensics_sinkhorn_reward": 0.9543177783489227,
      "rewards-max/forensics_strict_edge_reward": 0.4565845765173435,
      "rewards-max/forensics_tiou_reward": 0.9167695045471191,
      "rewards/forensics_sinkhorn_reward": 0.5715844631195068,
      "rewards/forensics_strict_edge_reward": 0.021013862686231732,
      "rewards/forensics_tiou_reward": 0.340492844581604,
      "step": 298
    },
    {
      "completion_length": 27.65625,
      "epoch": 0.9185867895545314,
      "grad_norm": 4.410325050354004,
      "kl": 0.2958984375,
      "learning_rate": 7.707692307692307e-07,
      "loss": -0.0398,
      "reward": 0.8984405398368835,
      "reward_std": 0.22871489822864532,
      "reward_std_all": 1.0825964212417603,
      "rewards-max/forensics_sinkhorn_reward": 0.9746473431587219,
      "rewards-max/forensics_strict_edge_reward": 0.22481869161128998,
      "rewards-max/forensics_tiou_reward": 0.9623843431472778,
      "rewards/forensics_sinkhorn_reward": 0.4966981112957001,
      "rewards/forensics_strict_edge_reward": 0.013076734729111195,
      "rewards/forensics_tiou_reward": 0.38866569101810455,
      "step": 299
    },
    {
      "completion_length": 23.578125,
      "epoch": 0.9216589861751152,
      "grad_norm": 5.736368179321289,
      "kl": 0.3154296875,
      "learning_rate": 7.699999999999999e-07,
      "loss": -0.0918,
      "reward": 0.9727967977523804,
      "reward_std": 0.3732113838195801,
      "reward_std_all": 1.096364676952362,
      "rewards-max/forensics_sinkhorn_reward": 0.9794124662876129,
      "rewards-max/forensics_strict_edge_reward": 0.562952071428299,
      "rewards-max/forensics_tiou_reward": 0.9777698814868927,
      "rewards/forensics_sinkhorn_reward": 0.5917738378047943,
      "rewards/forensics_strict_edge_reward": 0.04661675915122032,
      "rewards/forensics_tiou_reward": 0.3344062566757202,
      "step": 300
    },
    {
      "completion_length": 24.171875,
      "epoch": 0.9247311827956989,
      "grad_norm": 3.6640472412109375,
      "kl": 0.34765625,
      "learning_rate": 7.692307692307693e-07,
      "loss": -0.0559,
      "reward": 0.9995774626731873,
      "reward_std": 0.2814953699707985,
      "reward_std_all": 1.0422917604446411,
      "rewards-max/forensics_sinkhorn_reward": 0.9155278205871582,
      "rewards-max/forensics_strict_edge_reward": 0.11245313502149656,
      "rewards-max/forensics_tiou_reward": 0.7643933594226837,
      "rewards/forensics_sinkhorn_reward": 0.6186588406562805,
      "rewards/forensics_strict_edge_reward": 0.007511125608289149,
      "rewards/forensics_tiou_reward": 0.3734074532985687,
      "step": 301
    },
    {
      "completion_length": 23.671875,
      "epoch": 0.9278033794162827,
      "grad_norm": 5.838191032409668,
      "kl": 0.408203125,
      "learning_rate": 7.684615384615384e-07,
      "loss": -0.0663,
      "reward": 0.8778693675994873,
      "reward_std": 0.40895894169807434,
      "reward_std_all": 1.1411190032958984,
      "rewards-max/forensics_sinkhorn_reward": 0.9615216553211212,
      "rewards-max/forensics_strict_edge_reward": 0.21724942326545715,
      "rewards-max/forensics_tiou_reward": 0.931436687707901,
      "rewards/forensics_sinkhorn_reward": 0.517892524600029,
      "rewards/forensics_strict_edge_reward": 0.016395412385463715,
      "rewards/forensics_tiou_reward": 0.343581423163414,
      "step": 302
    },
    {
      "completion_length": 23.78125,
      "epoch": 0.9308755760368663,
      "grad_norm": 4.245938301086426,
      "kl": 0.30078125,
      "learning_rate": 7.676923076923077e-07,
      "loss": -0.1153,
      "reward": 0.9485360383987427,
      "reward_std": 0.25066301226615906,
      "reward_std_all": 1.0711358785629272,
      "rewards-max/forensics_sinkhorn_reward": 0.9391940236091614,
      "rewards-max/forensics_strict_edge_reward": 0.22107651457190514,
      "rewards-max/forensics_tiou_reward": 0.8801177442073822,
      "rewards/forensics_sinkhorn_reward": 0.5957318246364594,
      "rewards/forensics_strict_edge_reward": 0.015479423804208636,
      "rewards/forensics_tiou_reward": 0.3373248279094696,
      "step": 303
    },
    {
      "completion_length": 22.8125,
      "epoch": 0.9339477726574501,
      "grad_norm": 5.299281120300293,
      "kl": 0.3359375,
      "learning_rate": 7.669230769230769e-07,
      "loss": -0.108,
      "reward": 0.9099738001823425,
      "reward_std": 0.30187568068504333,
      "reward_std_all": 1.087962806224823,
      "rewards-max/forensics_sinkhorn_reward": 0.921009361743927,
      "rewards-max/forensics_strict_edge_reward": 0.14106393238762394,
      "rewards-max/forensics_tiou_reward": 0.838336855173111,
      "rewards/forensics_sinkhorn_reward": 0.5643413066864014,
      "rewards/forensics_strict_edge_reward": 0.010842503301319084,
      "rewards/forensics_tiou_reward": 0.3347900062799454,
      "step": 304
    },
    {
      "completion_length": 23.15625,
      "epoch": 0.9370199692780338,
      "grad_norm": 4.63437557220459,
      "kl": 0.3349609375,
      "learning_rate": 7.661538461538461e-07,
      "loss": -0.1487,
      "reward": 0.8517610728740692,
      "reward_std": 0.2598474770784378,
      "reward_std_all": 1.1201353669166565,
      "rewards-max/forensics_sinkhorn_reward": 0.9674860239028931,
      "rewards-max/forensics_strict_edge_reward": 0.2605837807059288,
      "rewards-max/forensics_tiou_reward": 0.9207866787910461,
      "rewards/forensics_sinkhorn_reward": 0.5572574734687805,
      "rewards/forensics_strict_edge_reward": 0.014506318606436253,
      "rewards/forensics_tiou_reward": 0.2799972891807556,
      "step": 305
    },
    {
      "completion_length": 25.125,
      "epoch": 0.9400921658986175,
      "grad_norm": 4.659698486328125,
      "kl": 0.29296875,
      "learning_rate": 7.653846153846153e-07,
      "loss": -0.1005,
      "reward": 0.8349727690219879,
      "reward_std": 0.20624073594808578,
      "reward_std_all": 1.10604590177536,
      "rewards-max/forensics_sinkhorn_reward": 0.9027131795883179,
      "rewards-max/forensics_strict_edge_reward": 0.11172286607325077,
      "rewards-max/forensics_tiou_reward": 0.801064521074295,
      "rewards/forensics_sinkhorn_reward": 0.5283372104167938,
      "rewards/forensics_strict_edge_reward": 0.0047625910374335945,
      "rewards/forensics_tiou_reward": 0.30187298357486725,
      "step": 306
    },
    {
      "completion_length": 24.625,
      "epoch": 0.9431643625192012,
      "grad_norm": 4.545106887817383,
      "kl": 0.2744140625,
      "learning_rate": 7.646153846153845e-07,
      "loss": -0.1078,
      "reward": 1.028578907251358,
      "reward_std": 0.2891122102737427,
      "reward_std_all": 1.0304561853408813,
      "rewards-max/forensics_sinkhorn_reward": 0.9405032098293304,
      "rewards-max/forensics_strict_edge_reward": 0.1075429804623127,
      "rewards-max/forensics_tiou_reward": 0.8821506202220917,
      "rewards/forensics_sinkhorn_reward": 0.634237676858902,
      "rewards/forensics_strict_edge_reward": 0.008778303512372077,
      "rewards/forensics_tiou_reward": 0.3855629861354828,
      "step": 307
    },
    {
      "completion_length": 26.703125,
      "epoch": 0.946236559139785,
      "grad_norm": 4.044839859008789,
      "kl": 0.3330078125,
      "learning_rate": 7.638461538461539e-07,
      "loss": -0.093,
      "reward": 0.8559718430042267,
      "reward_std": 0.27925756573677063,
      "reward_std_all": 1.1152735948562622,
      "rewards-max/forensics_sinkhorn_reward": 0.9522126615047455,
      "rewards-max/forensics_strict_edge_reward": 0.33416298031806946,
      "rewards-max/forensics_tiou_reward": 0.9010786712169647,
      "rewards/forensics_sinkhorn_reward": 0.5188919901847839,
      "rewards/forensics_strict_edge_reward": 0.01789339166134596,
      "rewards/forensics_tiou_reward": 0.3191864341497421,
      "step": 308
    },
    {
      "completion_length": 26.5,
      "epoch": 0.9493087557603687,
      "grad_norm": 5.001770496368408,
      "kl": 0.314453125,
      "learning_rate": 7.630769230769231e-07,
      "loss": -0.0591,
      "reward": 0.8020766079425812,
      "reward_std": 0.2104227915406227,
      "reward_std_all": 1.1314910650253296,
      "rewards-max/forensics_sinkhorn_reward": 0.8588776886463165,
      "rewards-max/forensics_strict_edge_reward": 0.44714050740003586,
      "rewards-max/forensics_tiou_reward": 0.8012469708919525,
      "rewards/forensics_sinkhorn_reward": 0.5108894556760788,
      "rewards/forensics_strict_edge_reward": 0.018731935299001634,
      "rewards/forensics_tiou_reward": 0.27245520055294037,
      "step": 309
    },
    {
      "completion_length": 25.453125,
      "epoch": 0.9523809523809523,
      "grad_norm": 4.985769748687744,
      "kl": 0.3388671875,
      "learning_rate": 7.623076923076923e-07,
      "loss": -0.0263,
      "reward": 0.9436202049255371,
      "reward_std": 0.20017697662115097,
      "reward_std_all": 1.046428769826889,
      "rewards-max/forensics_sinkhorn_reward": 0.8926744759082794,
      "rewards-max/forensics_strict_edge_reward": 0.1795838661491871,
      "rewards-max/forensics_tiou_reward": 0.8185398280620575,
      "rewards/forensics_sinkhorn_reward": 0.6037986278533936,
      "rewards/forensics_strict_edge_reward": 0.007524179760366678,
      "rewards/forensics_tiou_reward": 0.33229736238718033,
      "step": 310
    },
    {
      "completion_length": 26.25,
      "epoch": 0.9554531490015361,
      "grad_norm": 5.396651744842529,
      "kl": 0.2822265625,
      "learning_rate": 7.615384615384615e-07,
      "loss": -0.14,
      "reward": 0.8394463062286377,
      "reward_std": 0.39215949177742004,
      "reward_std_all": 1.1574167609214783,
      "rewards-max/forensics_sinkhorn_reward": 0.9526298642158508,
      "rewards-max/forensics_strict_edge_reward": 0.3141753673553467,
      "rewards-max/forensics_tiou_reward": 0.9123837053775787,
      "rewards/forensics_sinkhorn_reward": 0.49433718621730804,
      "rewards/forensics_strict_edge_reward": 0.018412345089018345,
      "rewards/forensics_tiou_reward": 0.32669679820537567,
      "step": 311
    },
    {
      "completion_length": 26.09375,
      "epoch": 0.9585253456221198,
      "grad_norm": 4.360405445098877,
      "kl": 0.314453125,
      "learning_rate": 7.607692307692307e-07,
      "loss": -0.1282,
      "reward": 1.0459496974945068,
      "reward_std": 0.2147621214389801,
      "reward_std_all": 1.0020561218261719,
      "rewards-max/forensics_sinkhorn_reward": 0.957583874464035,
      "rewards-max/forensics_strict_edge_reward": 0.09393114969134331,
      "rewards-max/forensics_tiou_reward": 0.9059258997440338,
      "rewards/forensics_sinkhorn_reward": 0.6295415163040161,
      "rewards/forensics_strict_edge_reward": 0.007142302813008428,
      "rewards/forensics_tiou_reward": 0.4092658907175064,
      "step": 312
    },
    {
      "completion_length": 24.703125,
      "epoch": 0.9615975422427036,
      "grad_norm": 5.980050563812256,
      "kl": 0.3212890625,
      "learning_rate": 7.599999999999999e-07,
      "loss": -0.1537,
      "reward": 0.9008910655975342,
      "reward_std": 0.3021530658006668,
      "reward_std_all": 1.0942994356155396,
      "rewards-max/forensics_sinkhorn_reward": 0.9421965777873993,
      "rewards-max/forensics_strict_edge_reward": 0.35838583111763,
      "rewards-max/forensics_tiou_reward": 0.8808012008666992,
      "rewards/forensics_sinkhorn_reward": 0.5406795591115952,
      "rewards/forensics_strict_edge_reward": 0.022331105545163155,
      "rewards/forensics_tiou_reward": 0.3378803879022598,
      "step": 313
    },
    {
      "completion_length": 22.78125,
      "epoch": 0.9646697388632872,
      "grad_norm": 4.679298400878906,
      "kl": 0.3076171875,
      "learning_rate": 7.592307692307692e-07,
      "loss": -0.1438,
      "reward": 0.8880117535591125,
      "reward_std": 0.17502547055482864,
      "reward_std_all": 1.0720492601394653,
      "rewards-max/forensics_sinkhorn_reward": 0.9351915121078491,
      "rewards-max/forensics_strict_edge_reward": 0.041153513826429844,
      "rewards-max/forensics_tiou_reward": 0.8561306893825531,
      "rewards/forensics_sinkhorn_reward": 0.5722833275794983,
      "rewards/forensics_strict_edge_reward": 0.002925786073319614,
      "rewards/forensics_tiou_reward": 0.3128025680780411,
      "step": 314
    },
    {
      "completion_length": 26.125,
      "epoch": 0.967741935483871,
      "grad_norm": 4.234024524688721,
      "kl": 0.37109375,
      "learning_rate": 7.584615384615384e-07,
      "loss": -0.1075,
      "reward": 0.7632505297660828,
      "reward_std": 0.28274182975292206,
      "reward_std_all": 1.1571680307388306,
      "rewards-max/forensics_sinkhorn_reward": 0.9387097656726837,
      "rewards-max/forensics_strict_edge_reward": 0.0562007415574044,
      "rewards-max/forensics_tiou_reward": 0.8819010853767395,
      "rewards/forensics_sinkhorn_reward": 0.48888763785362244,
      "rewards/forensics_strict_edge_reward": 0.002382691043749219,
      "rewards/forensics_tiou_reward": 0.2719801589846611,
      "step": 315
    },
    {
      "completion_length": 27.625,
      "epoch": 0.9708141321044547,
      "grad_norm": 5.646679878234863,
      "kl": 0.3369140625,
      "learning_rate": 7.576923076923077e-07,
      "loss": -0.0896,
      "reward": 0.7540002763271332,
      "reward_std": 0.24222221225500107,
      "reward_std_all": 1.1570104956626892,
      "rewards-max/forensics_sinkhorn_reward": 0.9447867572307587,
      "rewards-max/forensics_strict_edge_reward": 0.20859580487012863,
      "rewards-max/forensics_tiou_reward": 0.8899800181388855,
      "rewards/forensics_sinkhorn_reward": 0.44772571325302124,
      "rewards/forensics_strict_edge_reward": 0.013004542328417301,
      "rewards/forensics_tiou_reward": 0.2932700514793396,
      "step": 316
    },
    {
      "completion_length": 22.515625,
      "epoch": 0.9738863287250384,
      "grad_norm": 3.6209254264831543,
      "kl": 0.3876953125,
      "learning_rate": 7.569230769230769e-07,
      "loss": -0.1095,
      "reward": 0.9521148204803467,
      "reward_std": 0.35150080919265747,
      "reward_std_all": 1.0878141522407532,
      "rewards-max/forensics_sinkhorn_reward": 0.9405035972595215,
      "rewards-max/forensics_strict_edge_reward": 0.2007068693637848,
      "rewards-max/forensics_tiou_reward": 0.882539302110672,
      "rewards/forensics_sinkhorn_reward": 0.592970609664917,
      "rewards/forensics_strict_edge_reward": 0.022674459032714367,
      "rewards/forensics_tiou_reward": 0.33646972477436066,
      "step": 317
    },
    {
      "completion_length": 26.609375,
      "epoch": 0.9769585253456221,
      "grad_norm": 4.3596601486206055,
      "kl": 0.2958984375,
      "learning_rate": 7.561538461538461e-07,
      "loss": -0.0984,
      "reward": 0.9925290048122406,
      "reward_std": 0.19050341099500656,
      "reward_std_all": 1.0248719453811646,
      "rewards-max/forensics_sinkhorn_reward": 0.9532072842121124,
      "rewards-max/forensics_strict_edge_reward": 0.25821408932097256,
      "rewards-max/forensics_tiou_reward": 0.9109714031219482,
      "rewards/forensics_sinkhorn_reward": 0.6035896837711334,
      "rewards/forensics_strict_edge_reward": 0.011189750817720778,
      "rewards/forensics_tiou_reward": 0.37774965167045593,
      "step": 318
    },
    {
      "completion_length": 24.234375,
      "epoch": 0.9800307219662059,
      "grad_norm": 4.263552665710449,
      "kl": 0.2666015625,
      "learning_rate": 7.553846153846154e-07,
      "loss": -0.1364,
      "reward": 0.8783624172210693,
      "reward_std": 0.16787533462047577,
      "reward_std_all": 1.0822962522506714,
      "rewards-max/forensics_sinkhorn_reward": 0.895775556564331,
      "rewards-max/forensics_strict_edge_reward": 0.26681816880591214,
      "rewards-max/forensics_tiou_reward": 0.8240850269794464,
      "rewards/forensics_sinkhorn_reward": 0.5260324329137802,
      "rewards/forensics_strict_edge_reward": 0.022055491026549134,
      "rewards/forensics_tiou_reward": 0.3302745074033737,
      "step": 319
    },
    {
      "completion_length": 23.8125,
      "epoch": 0.9831029185867896,
      "grad_norm": 4.2515387535095215,
      "kl": 0.3271484375,
      "learning_rate": 7.546153846153845e-07,
      "loss": -0.0914,
      "reward": 1.1901761293411255,
      "reward_std": 0.25979869067668915,
      "reward_std_all": 0.9436314404010773,
      "rewards-max/forensics_sinkhorn_reward": 0.9836159348487854,
      "rewards-max/forensics_strict_edge_reward": 0.5414712727069855,
      "rewards-max/forensics_tiou_reward": 0.9835453927516937,
      "rewards/forensics_sinkhorn_reward": 0.6918651163578033,
      "rewards/forensics_strict_edge_reward": 0.02806155662983656,
      "rewards/forensics_tiou_reward": 0.47024936974048615,
      "step": 320
    },
    {
      "completion_length": 26.78125,
      "epoch": 0.9861751152073732,
      "grad_norm": 4.8571391105651855,
      "kl": 0.2353515625,
      "learning_rate": 7.538461538461538e-07,
      "loss": -0.1039,
      "reward": 0.9371052384376526,
      "reward_std": 0.33673371374607086,
      "reward_std_all": 1.0935041308403015,
      "rewards-max/forensics_sinkhorn_reward": 0.979315310716629,
      "rewards-max/forensics_strict_edge_reward": 0.417969286441803,
      "rewards-max/forensics_tiou_reward": 0.9729335010051727,
      "rewards/forensics_sinkhorn_reward": 0.5749413967132568,
      "rewards/forensics_strict_edge_reward": 0.02283503394573927,
      "rewards/forensics_tiou_reward": 0.3393288254737854,
      "step": 321
    },
    {
      "completion_length": 26.15625,
      "epoch": 0.989247311827957,
      "grad_norm": 4.32968807220459,
      "kl": 0.3046875,
      "learning_rate": 7.530769230769231e-07,
      "loss": -0.1529,
      "reward": 0.8943194448947906,
      "reward_std": 0.3204796239733696,
      "reward_std_all": 1.113946259021759,
      "rewards-max/forensics_sinkhorn_reward": 0.9667708575725555,
      "rewards-max/forensics_strict_edge_reward": 0.24126016907393932,
      "rewards-max/forensics_tiou_reward": 0.9415104985237122,
      "rewards/forensics_sinkhorn_reward": 0.553615003824234,
      "rewards/forensics_strict_edge_reward": 0.025286071933805943,
      "rewards/forensics_tiou_reward": 0.31541839987039566,
      "step": 322
    },
    {
      "completion_length": 22.671875,
      "epoch": 0.9923195084485407,
      "grad_norm": 5.030013561248779,
      "kl": 0.3388671875,
      "learning_rate": 7.523076923076923e-07,
      "loss": -0.1101,
      "reward": 0.7802143692970276,
      "reward_std": 0.22878247499465942,
      "reward_std_all": 1.145675003528595,
      "rewards-max/forensics_sinkhorn_reward": 0.9260946810245514,
      "rewards-max/forensics_strict_edge_reward": 0.22483221080619842,
      "rewards-max/forensics_tiou_reward": 0.8400426805019379,
      "rewards/forensics_sinkhorn_reward": 0.5276477932929993,
      "rewards/forensics_strict_edge_reward": 0.011028212923520186,
      "rewards/forensics_tiou_reward": 0.24153831601142883,
      "step": 323
    },
    {
      "completion_length": 24.578125,
      "epoch": 0.9953917050691244,
      "grad_norm": 5.766621112823486,
      "kl": 0.337890625,
      "learning_rate": 7.515384615384615e-07,
      "loss": -0.096,
      "reward": 1.1206214427947998,
      "reward_std": 0.41749241948127747,
      "reward_std_all": 1.0219463109970093,
      "rewards-max/forensics_sinkhorn_reward": 0.9736431837081909,
      "rewards-max/forensics_strict_edge_reward": 0.4533289223909378,
      "rewards-max/forensics_tiou_reward": 0.9594495892524719,
      "rewards/forensics_sinkhorn_reward": 0.6338802576065063,
      "rewards/forensics_strict_edge_reward": 0.032151756808161736,
      "rewards/forensics_tiou_reward": 0.4545893967151642,
      "step": 324
    },
    {
      "completion_length": 28.03125,
      "epoch": 0.9984639016897081,
      "grad_norm": 5.02206563949585,
      "kl": 0.3466796875,
      "learning_rate": 7.507692307692307e-07,
      "loss": -0.0485,
      "reward": 0.9096269607543945,
      "reward_std": 0.19620470702648163,
      "reward_std_all": 1.0649044513702393,
      "rewards-max/forensics_sinkhorn_reward": 0.888017863035202,
      "rewards-max/forensics_strict_edge_reward": 0.14557395968586206,
      "rewards-max/forensics_tiou_reward": 0.77904212474823,
      "rewards/forensics_sinkhorn_reward": 0.5617242753505707,
      "rewards/forensics_strict_edge_reward": 0.008954749442636967,
      "rewards/forensics_tiou_reward": 0.3389480412006378,
      "step": 325
    },
    {
      "completion_length": 21.600000381469727,
      "epoch": 1.0,
      "grad_norm": 4.083863735198975,
      "kl": 0.32421875,
      "learning_rate": 7.5e-07,
      "loss": -0.0591,
      "reward": 0.5209042429924011,
      "reward_std": 0.1968398243188858,
      "reward_std_all": 1.240326166152954,
      "rewards-max/forensics_sinkhorn_reward": 0.5524111390113831,
      "rewards-max/forensics_strict_edge_reward": 2.6816840545507148e-05,
      "rewards-max/forensics_tiou_reward": 0.5355797410011292,
      "rewards/forensics_sinkhorn_reward": 0.31098538637161255,
      "rewards/forensics_strict_edge_reward": 5.3641406339011155e-06,
      "rewards/forensics_tiou_reward": 0.2099134773015976,
      "step": 326
    },
    {
      "completion_length": 28.0625,
      "epoch": 1.0030721966205838,
      "grad_norm": 4.4424028396606445,
      "kl": 0.3828125,
      "learning_rate": 7.492307692307692e-07,
      "loss": -0.0854,
      "reward": 0.9423555731773376,
      "reward_std": 0.21988604962825775,
      "reward_std_all": 1.059902310371399,
      "rewards-max/forensics_sinkhorn_reward": 0.9504386186599731,
      "rewards-max/forensics_strict_edge_reward": 0.3569474369287491,
      "rewards-max/forensics_tiou_reward": 0.9042980968952179,
      "rewards/forensics_sinkhorn_reward": 0.5570574700832367,
      "rewards/forensics_strict_edge_reward": 0.028467849362641573,
      "rewards/forensics_tiou_reward": 0.3568302094936371,
      "step": 327
    },
    {
      "completion_length": 23.125,
      "epoch": 1.0061443932411673,
      "grad_norm": 4.5585103034973145,
      "kl": 0.3564453125,
      "learning_rate": 7.484615384615384e-07,
      "loss": -0.1165,
      "reward": 1.0995625853538513,
      "reward_std": 0.3420964777469635,
      "reward_std_all": 1.0209780037403107,
      "rewards-max/forensics_sinkhorn_reward": 0.9764328002929688,
      "rewards-max/forensics_strict_edge_reward": 0.40034955739974976,
      "rewards-max/forensics_tiou_reward": 0.9654651880264282,
      "rewards/forensics_sinkhorn_reward": 0.6547608971595764,
      "rewards/forensics_strict_edge_reward": 0.027139502577483654,
      "rewards/forensics_tiou_reward": 0.41766226291656494,
      "step": 328
    },
    {
      "completion_length": 24.71875,
      "epoch": 1.0092165898617511,
      "grad_norm": 5.292283535003662,
      "kl": 0.3701171875,
      "learning_rate": 7.476923076923077e-07,
      "loss": -0.1105,
      "reward": 1.1204979419708252,
      "reward_std": 0.37773318588733673,
      "reward_std_all": 1.0184550881385803,
      "rewards-max/forensics_sinkhorn_reward": 0.9789023697376251,
      "rewards-max/forensics_strict_edge_reward": 0.5294082351028919,
      "rewards-max/forensics_tiou_reward": 0.9780527353286743,
      "rewards/forensics_sinkhorn_reward": 0.6490703225135803,
      "rewards/forensics_strict_edge_reward": 0.03968116897158325,
      "rewards/forensics_tiou_reward": 0.4317464232444763,
      "step": 329
    },
    {
      "completion_length": 24.390625,
      "epoch": 1.012288786482335,
      "grad_norm": 4.085367202758789,
      "kl": 0.3828125,
      "learning_rate": 7.469230769230769e-07,
      "loss": -0.0889,
      "reward": 0.9989916682243347,
      "reward_std": 0.35466910898685455,
      "reward_std_all": 1.0679094791412354,
      "rewards-max/forensics_sinkhorn_reward": 0.9703821539878845,
      "rewards-max/forensics_strict_edge_reward": 0.42878545820713043,
      "rewards-max/forensics_tiou_reward": 0.9509406983852386,
      "rewards/forensics_sinkhorn_reward": 0.5998814105987549,
      "rewards/forensics_strict_edge_reward": 0.029256383888423443,
      "rewards/forensics_tiou_reward": 0.36985382437705994,
      "step": 330
    },
    {
      "completion_length": 27.5,
      "epoch": 1.0153609831029187,
      "grad_norm": 4.369935512542725,
      "kl": 0.3212890625,
      "learning_rate": 7.461538461538461e-07,
      "loss": -0.135,
      "reward": 1.0949256420135498,
      "reward_std": 0.22044365853071213,
      "reward_std_all": 0.9883352220058441,
      "rewards-max/forensics_sinkhorn_reward": 0.9289074242115021,
      "rewards-max/forensics_strict_edge_reward": 0.2200531642884016,
      "rewards-max/forensics_tiou_reward": 0.869920402765274,
      "rewards/forensics_sinkhorn_reward": 0.613365113735199,
      "rewards/forensics_strict_edge_reward": 0.024625356192700565,
      "rewards/forensics_tiou_reward": 0.45693516731262207,
      "step": 331
    },
    {
      "completion_length": 27.3125,
      "epoch": 1.0184331797235022,
      "grad_norm": 4.239010334014893,
      "kl": 0.28125,
      "learning_rate": 7.453846153846154e-07,
      "loss": -0.1467,
      "reward": 0.9174475073814392,
      "reward_std": 0.2269844263792038,
      "reward_std_all": 1.0896235704421997,
      "rewards-max/forensics_sinkhorn_reward": 0.9530283510684967,
      "rewards-max/forensics_strict_edge_reward": 0.35910215950571,
      "rewards-max/forensics_tiou_reward": 0.9047370553016663,
      "rewards/forensics_sinkhorn_reward": 0.5371798574924469,
      "rewards/forensics_strict_edge_reward": 0.030016555574547965,
      "rewards/forensics_tiou_reward": 0.350251168012619,
      "step": 332
    },
    {
      "completion_length": 24.828125,
      "epoch": 1.021505376344086,
      "grad_norm": 4.334512233734131,
      "kl": 0.4169921875,
      "learning_rate": 7.446153846153845e-07,
      "loss": -0.183,
      "reward": 1.2089297473430634,
      "reward_std": 0.475542277097702,
      "reward_std_all": 1.0167721509933472,
      "rewards-max/forensics_sinkhorn_reward": 0.981095939874649,
      "rewards-max/forensics_strict_edge_reward": 0.6188668608665466,
      "rewards-max/forensics_tiou_reward": 0.9768908619880676,
      "rewards/forensics_sinkhorn_reward": 0.6667744219303131,
      "rewards/forensics_strict_edge_reward": 0.06785644218325615,
      "rewards/forensics_tiou_reward": 0.4742989242076874,
      "step": 333
    },
    {
      "completion_length": 27.296875,
      "epoch": 1.0245775729646698,
      "grad_norm": 4.6217041015625,
      "kl": 0.328125,
      "learning_rate": 7.438461538461538e-07,
      "loss": -0.1985,
      "reward": 0.9945201575756073,
      "reward_std": 0.2745755985379219,
      "reward_std_all": 1.0490394830703735,
      "rewards-max/forensics_sinkhorn_reward": 0.9640869200229645,
      "rewards-max/forensics_strict_edge_reward": 0.44343748688697815,
      "rewards-max/forensics_tiou_reward": 0.93923619389534,
      "rewards/forensics_sinkhorn_reward": 0.5539006292819977,
      "rewards/forensics_strict_edge_reward": 0.05387124605476856,
      "rewards/forensics_tiou_reward": 0.3867482393980026,
      "step": 334
    },
    {
      "completion_length": 27.21875,
      "epoch": 1.0276497695852536,
      "grad_norm": 5.487829685211182,
      "kl": 0.3603515625,
      "learning_rate": 7.43076923076923e-07,
      "loss": -0.1295,
      "reward": 0.8980677127838135,
      "reward_std": 0.2135108783841133,
      "reward_std_all": 1.0818119049072266,
      "rewards-max/forensics_sinkhorn_reward": 0.9300386309623718,
      "rewards-max/forensics_strict_edge_reward": 0.15874998085200787,
      "rewards-max/forensics_tiou_reward": 0.8711429536342621,
      "rewards/forensics_sinkhorn_reward": 0.5481389760971069,
      "rewards/forensics_strict_edge_reward": 0.006828678771853447,
      "rewards/forensics_tiou_reward": 0.34310004115104675,
      "step": 335
    },
    {
      "completion_length": 30.9375,
      "epoch": 1.0307219662058371,
      "grad_norm": 5.895161151885986,
      "kl": 0.357421875,
      "learning_rate": 7.423076923076923e-07,
      "loss": -0.1204,
      "reward": 1.059654414653778,
      "reward_std": 0.3536916673183441,
      "reward_std_all": 1.0371981859207153,
      "rewards-max/forensics_sinkhorn_reward": 0.9771153032779694,
      "rewards-max/forensics_strict_edge_reward": 0.27666913717985153,
      "rewards-max/forensics_tiou_reward": 0.9680440127849579,
      "rewards/forensics_sinkhorn_reward": 0.5928589105606079,
      "rewards/forensics_strict_edge_reward": 0.022973651066422462,
      "rewards/forensics_tiou_reward": 0.4438219219446182,
      "step": 336
    },
    {
      "completion_length": 22.34375,
      "epoch": 1.033794162826421,
      "grad_norm": 5.355310440063477,
      "kl": 0.294921875,
      "learning_rate": 7.415384615384616e-07,
      "loss": -0.173,
      "reward": 1.1961288452148438,
      "reward_std": 0.2979005128145218,
      "reward_std_all": 0.9749631285667419,
      "rewards-max/forensics_sinkhorn_reward": 0.9741449952125549,
      "rewards-max/forensics_strict_edge_reward": 0.7224518060684204,
      "rewards-max/forensics_tiou_reward": 0.9548793733119965,
      "rewards/forensics_sinkhorn_reward": 0.6936337649822235,
      "rewards/forensics_strict_edge_reward": 0.06267064996063709,
      "rewards/forensics_tiou_reward": 0.4398244023323059,
      "step": 337
    },
    {
      "completion_length": 25.5625,
      "epoch": 1.0368663594470047,
      "grad_norm": 4.862210273742676,
      "kl": 0.3447265625,
      "learning_rate": 7.407692307692307e-07,
      "loss": -0.1417,
      "reward": 0.8702147901058197,
      "reward_std": 0.2554307281970978,
      "reward_std_all": 1.1002869009971619,
      "rewards-max/forensics_sinkhorn_reward": 0.9423157870769501,
      "rewards-max/forensics_strict_edge_reward": 0.11936007812619209,
      "rewards-max/forensics_tiou_reward": 0.8691056072711945,
      "rewards/forensics_sinkhorn_reward": 0.5681394636631012,
      "rewards/forensics_strict_edge_reward": 0.005023032543249428,
      "rewards/forensics_tiou_reward": 0.297052338719368,
      "step": 338
    },
    {
      "completion_length": 26.3125,
      "epoch": 1.0399385560675882,
      "grad_norm": 5.187408447265625,
      "kl": 0.4072265625,
      "learning_rate": 7.4e-07,
      "loss": -0.1084,
      "reward": 1.1601552367210388,
      "reward_std": 0.37208719551563263,
      "reward_std_all": 1.0036316514015198,
      "rewards-max/forensics_sinkhorn_reward": 0.9868373870849609,
      "rewards-max/forensics_strict_edge_reward": 0.580880731344223,
      "rewards-max/forensics_tiou_reward": 0.992062121629715,
      "rewards/forensics_sinkhorn_reward": 0.6179052591323853,
      "rewards/forensics_strict_edge_reward": 0.057550251483917236,
      "rewards/forensics_tiou_reward": 0.4846997857093811,
      "step": 339
    },
    {
      "completion_length": 27.453125,
      "epoch": 1.043010752688172,
      "grad_norm": 4.145178318023682,
      "kl": 0.3125,
      "learning_rate": 7.392307692307692e-07,
      "loss": -0.0759,
      "reward": 1.0603811144828796,
      "reward_std": 0.2722393721342087,
      "reward_std_all": 1.0069347023963928,
      "rewards-max/forensics_sinkhorn_reward": 0.9580907225608826,
      "rewards-max/forensics_strict_edge_reward": 0.3132822811603546,
      "rewards-max/forensics_tiou_reward": 0.9181358218193054,
      "rewards/forensics_sinkhorn_reward": 0.62043297290802,
      "rewards/forensics_strict_edge_reward": 0.024821268394589424,
      "rewards/forensics_tiou_reward": 0.4151268005371094,
      "step": 340
    },
    {
      "completion_length": 22.78125,
      "epoch": 1.0460829493087558,
      "grad_norm": 4.77217435836792,
      "kl": 0.3974609375,
      "learning_rate": 7.384615384615384e-07,
      "loss": -0.0595,
      "reward": 1.0175846815109253,
      "reward_std": 0.2669730558991432,
      "reward_std_all": 1.0325007140636444,
      "rewards-max/forensics_sinkhorn_reward": 0.9582551717758179,
      "rewards-max/forensics_strict_edge_reward": 0.0733795166015625,
      "rewards-max/forensics_tiou_reward": 0.91724893450737,
      "rewards/forensics_sinkhorn_reward": 0.5993670523166656,
      "rewards/forensics_strict_edge_reward": 0.007708147633820772,
      "rewards/forensics_tiou_reward": 0.41050949692726135,
      "step": 341
    },
    {
      "completion_length": 27.90625,
      "epoch": 1.0491551459293396,
      "grad_norm": 4.483903884887695,
      "kl": 0.326171875,
      "learning_rate": 7.376923076923076e-07,
      "loss": -0.0922,
      "reward": 1.0322844982147217,
      "reward_std": 0.3201068788766861,
      "reward_std_all": 1.037424087524414,
      "rewards-max/forensics_sinkhorn_reward": 0.9545769989490509,
      "rewards-max/forensics_strict_edge_reward": 0.40157462656497955,
      "rewards-max/forensics_tiou_reward": 0.9200276732444763,
      "rewards/forensics_sinkhorn_reward": 0.5479768216609955,
      "rewards/forensics_strict_edge_reward": 0.04437179770320654,
      "rewards/forensics_tiou_reward": 0.43993590772151947,
      "step": 342
    },
    {
      "completion_length": 26.0625,
      "epoch": 1.0522273425499231,
      "grad_norm": 4.866328239440918,
      "kl": 0.349609375,
      "learning_rate": 7.369230769230769e-07,
      "loss": 0.0147,
      "reward": 1.0633065104484558,
      "reward_std": 0.4940255284309387,
      "reward_std_all": 1.104549527168274,
      "rewards-max/forensics_sinkhorn_reward": 0.9795007705688477,
      "rewards-max/forensics_strict_edge_reward": 0.7676310539245605,
      "rewards-max/forensics_tiou_reward": 0.9787550568580627,
      "rewards/forensics_sinkhorn_reward": 0.562514990568161,
      "rewards/forensics_strict_edge_reward": 0.07524865120649338,
      "rewards/forensics_tiou_reward": 0.4255428910255432,
      "step": 343
    },
    {
      "completion_length": 26.4375,
      "epoch": 1.055299539170507,
      "grad_norm": 5.14832067489624,
      "kl": 0.3759765625,
      "learning_rate": 7.361538461538461e-07,
      "loss": -0.0353,
      "reward": 1.0673953294754028,
      "reward_std": 0.33025412261486053,
      "reward_std_all": 1.0355998873710632,
      "rewards-max/forensics_sinkhorn_reward": 0.9687609076499939,
      "rewards-max/forensics_strict_edge_reward": 0.726275771856308,
      "rewards-max/forensics_tiou_reward": 0.9491109549999237,
      "rewards/forensics_sinkhorn_reward": 0.5915644466876984,
      "rewards/forensics_strict_edge_reward": 0.0396293131634593,
      "rewards/forensics_tiou_reward": 0.43620161712169647,
      "step": 344
    },
    {
      "completion_length": 26.6875,
      "epoch": 1.0583717357910907,
      "grad_norm": 5.345242500305176,
      "kl": 0.373046875,
      "learning_rate": 7.353846153846154e-07,
      "loss": -0.0619,
      "reward": 1.095258891582489,
      "reward_std": 0.4332202672958374,
      "reward_std_all": 1.0500726699829102,
      "rewards-max/forensics_sinkhorn_reward": 0.9708067774772644,
      "rewards-max/forensics_strict_edge_reward": 0.49256695806980133,
      "rewards-max/forensics_tiou_reward": 0.9546462595462799,
      "rewards/forensics_sinkhorn_reward": 0.5895902812480927,
      "rewards/forensics_strict_edge_reward": 0.039872742258012295,
      "rewards/forensics_tiou_reward": 0.4657958596944809,
      "step": 345
    },
    {
      "completion_length": 26.125,
      "epoch": 1.0614439324116745,
      "grad_norm": 4.856149196624756,
      "kl": 0.384765625,
      "learning_rate": 7.346153846153846e-07,
      "loss": -0.0605,
      "reward": 1.1536383032798767,
      "reward_std": 0.24794326722621918,
      "reward_std_all": 0.9640580713748932,
      "rewards-max/forensics_sinkhorn_reward": 0.9576427936553955,
      "rewards-max/forensics_strict_edge_reward": 0.3696242682635784,
      "rewards-max/forensics_tiou_reward": 0.9157149791717529,
      "rewards/forensics_sinkhorn_reward": 0.6644531190395355,
      "rewards/forensics_strict_edge_reward": 0.017395790433511138,
      "rewards/forensics_tiou_reward": 0.4717893451452255,
      "step": 346
    },
    {
      "completion_length": 24.71875,
      "epoch": 1.064516129032258,
      "grad_norm": 4.6915740966796875,
      "kl": 0.3115234375,
      "learning_rate": 7.338461538461538e-07,
      "loss": -0.0351,
      "reward": 1.0435721278190613,
      "reward_std": 0.2844468653202057,
      "reward_std_all": 1.0279814004898071,
      "rewards-max/forensics_sinkhorn_reward": 0.9639443755149841,
      "rewards-max/forensics_strict_edge_reward": 0.3418336343020201,
      "rewards-max/forensics_tiou_reward": 0.9417973756790161,
      "rewards/forensics_sinkhorn_reward": 0.6589291989803314,
      "rewards/forensics_strict_edge_reward": 0.017421486787497997,
      "rewards/forensics_tiou_reward": 0.3672214150428772,
      "step": 347
    },
    {
      "completion_length": 24.640625,
      "epoch": 1.0675883256528418,
      "grad_norm": 5.462399959564209,
      "kl": 0.4111328125,
      "learning_rate": 7.33076923076923e-07,
      "loss": -0.0809,
      "reward": 1.0093753933906555,
      "reward_std": 0.23579613119363785,
      "reward_std_all": 1.0330644249916077,
      "rewards-max/forensics_sinkhorn_reward": 0.961747020483017,
      "rewards-max/forensics_strict_edge_reward": 0.37542198598384857,
      "rewards-max/forensics_tiou_reward": 0.9303239583969116,
      "rewards/forensics_sinkhorn_reward": 0.6152678430080414,
      "rewards/forensics_strict_edge_reward": 0.023391379974782467,
      "rewards/forensics_tiou_reward": 0.37071622908115387,
      "step": 348
    },
    {
      "completion_length": 26.34375,
      "epoch": 1.0706605222734256,
      "grad_norm": 8.089534759521484,
      "kl": 0.595703125,
      "learning_rate": 7.323076923076923e-07,
      "loss": -0.0573,
      "reward": 1.113528847694397,
      "reward_std": 0.27344369888305664,
      "reward_std_all": 0.99139004945755,
      "rewards-max/forensics_sinkhorn_reward": 0.9581794738769531,
      "rewards-max/forensics_strict_edge_reward": 0.37554237991571426,
      "rewards-max/forensics_tiou_reward": 0.9331567883491516,
      "rewards/forensics_sinkhorn_reward": 0.6236256957054138,
      "rewards/forensics_strict_edge_reward": 0.03534040134400129,
      "rewards/forensics_tiou_reward": 0.45456282794475555,
      "step": 349
    },
    {
      "completion_length": 24.75,
      "epoch": 1.0737327188940091,
      "grad_norm": 6.883322238922119,
      "kl": 0.376953125,
      "learning_rate": 7.315384615384616e-07,
      "loss": -0.1082,
      "reward": 1.3466341495513916,
      "reward_std": 0.5062820166349411,
      "reward_std_all": 0.9539633095264435,
      "rewards-max/forensics_sinkhorn_reward": 0.9739909172058105,
      "rewards-max/forensics_strict_edge_reward": 0.49006423354148865,
      "rewards-max/forensics_tiou_reward": 0.9632175862789154,
      "rewards/forensics_sinkhorn_reward": 0.74421626329422,
      "rewards/forensics_strict_edge_reward": 0.06619865074753761,
      "rewards/forensics_tiou_reward": 0.5362192392349243,
      "step": 350
    },
    {
      "completion_length": 25.59375,
      "epoch": 1.076804915514593,
      "grad_norm": 6.461907386779785,
      "kl": 0.326171875,
      "learning_rate": 7.307692307692307e-07,
      "loss": -0.085,
      "reward": 1.0584933161735535,
      "reward_std": 0.32311202585697174,
      "reward_std_all": 1.0472490191459656,
      "rewards-max/forensics_sinkhorn_reward": 0.9697158634662628,
      "rewards-max/forensics_strict_edge_reward": 0.5422053933143616,
      "rewards-max/forensics_tiou_reward": 0.9547611474990845,
      "rewards/forensics_sinkhorn_reward": 0.5963753163814545,
      "rewards/forensics_strict_edge_reward": 0.05478985421359539,
      "rewards/forensics_tiou_reward": 0.4073280841112137,
      "step": 351
    },
    {
      "completion_length": 25.90625,
      "epoch": 1.0798771121351767,
      "grad_norm": 9.985260963439941,
      "kl": 0.580078125,
      "learning_rate": 7.3e-07,
      "loss": -0.0308,
      "reward": 0.8434188067913055,
      "reward_std": 0.2071419358253479,
      "reward_std_all": 1.106321096420288,
      "rewards-max/forensics_sinkhorn_reward": 0.9627311825752258,
      "rewards-max/forensics_strict_edge_reward": 0.2521332090254873,
      "rewards-max/forensics_tiou_reward": 0.9272781610488892,
      "rewards/forensics_sinkhorn_reward": 0.5142785012722015,
      "rewards/forensics_strict_edge_reward": 0.011678224851493724,
      "rewards/forensics_tiou_reward": 0.3174620717763901,
      "step": 352
    },
    {
      "completion_length": 23.578125,
      "epoch": 1.0829493087557605,
      "grad_norm": 5.474032878875732,
      "kl": 0.28515625,
      "learning_rate": 7.292307692307692e-07,
      "loss": -0.1124,
      "reward": 1.0544868111610413,
      "reward_std": 0.2019813284277916,
      "reward_std_all": 1.0025851428508759,
      "rewards-max/forensics_sinkhorn_reward": 0.9135596752166748,
      "rewards-max/forensics_strict_edge_reward": 0.5630159378051758,
      "rewards-max/forensics_tiou_reward": 0.8161448240280151,
      "rewards/forensics_sinkhorn_reward": 0.6586639285087585,
      "rewards/forensics_strict_edge_reward": 0.028257761150598526,
      "rewards/forensics_tiou_reward": 0.3675651401281357,
      "step": 353
    },
    {
      "completion_length": 25.765625,
      "epoch": 1.086021505376344,
      "grad_norm": 5.60471773147583,
      "kl": 0.296875,
      "learning_rate": 7.284615384615384e-07,
      "loss": -0.1264,
      "reward": 0.8951634764671326,
      "reward_std": 0.2474251538515091,
      "reward_std_all": 1.0863953828811646,
      "rewards-max/forensics_sinkhorn_reward": 0.9353732764720917,
      "rewards-max/forensics_strict_edge_reward": 0.31149575114250183,
      "rewards-max/forensics_tiou_reward": 0.866447389125824,
      "rewards/forensics_sinkhorn_reward": 0.5340012311935425,
      "rewards/forensics_strict_edge_reward": 0.017029627226293087,
      "rewards/forensics_tiou_reward": 0.3441326469182968,
      "step": 354
    },
    {
      "completion_length": 26.1875,
      "epoch": 1.0890937019969278,
      "grad_norm": 5.116363048553467,
      "kl": 0.505859375,
      "learning_rate": 7.276923076923076e-07,
      "loss": -0.0616,
      "reward": 1.0883253812789917,
      "reward_std": 0.28350065648555756,
      "reward_std_all": 1.0037369132041931,
      "rewards-max/forensics_sinkhorn_reward": 0.9719467759132385,
      "rewards-max/forensics_strict_edge_reward": 0.26869526505470276,
      "rewards-max/forensics_tiou_reward": 0.9528653621673584,
      "rewards/forensics_sinkhorn_reward": 0.6334556639194489,
      "rewards/forensics_strict_edge_reward": 0.027694005519151688,
      "rewards/forensics_tiou_reward": 0.4271757900714874,
      "step": 355
    },
    {
      "completion_length": 28.53125,
      "epoch": 1.0921658986175116,
      "grad_norm": 4.857796669006348,
      "kl": 0.322265625,
      "learning_rate": 7.269230769230769e-07,
      "loss": -0.0861,
      "reward": 0.9253820180892944,
      "reward_std": 0.24521219730377197,
      "reward_std_all": 1.0779160261154175,
      "rewards-max/forensics_sinkhorn_reward": 0.9759056270122528,
      "rewards-max/forensics_strict_edge_reward": 0.44921496510505676,
      "rewards-max/forensics_tiou_reward": 0.9679652452468872,
      "rewards/forensics_sinkhorn_reward": 0.5449183285236359,
      "rewards/forensics_strict_edge_reward": 0.02705393172800541,
      "rewards/forensics_tiou_reward": 0.3534097969532013,
      "step": 356
    },
    {
      "completion_length": 28.65625,
      "epoch": 1.0952380952380953,
      "grad_norm": 6.049169540405273,
      "kl": 0.390625,
      "learning_rate": 7.261538461538461e-07,
      "loss": -0.0809,
      "reward": 1.1149623394012451,
      "reward_std": 0.2477143332362175,
      "reward_std_all": 0.9680010676383972,
      "rewards-max/forensics_sinkhorn_reward": 0.9490185379981995,
      "rewards-max/forensics_strict_edge_reward": 0.34747011214494705,
      "rewards-max/forensics_tiou_reward": 0.9019206464290619,
      "rewards/forensics_sinkhorn_reward": 0.6304331719875336,
      "rewards/forensics_strict_edge_reward": 0.019224549178034067,
      "rewards/forensics_tiou_reward": 0.46530458331108093,
      "step": 357
    },
    {
      "completion_length": 28.9375,
      "epoch": 1.098310291858679,
      "grad_norm": 5.126999855041504,
      "kl": 0.3818359375,
      "learning_rate": 7.253846153846154e-07,
      "loss": -0.1316,
      "reward": 1.0026320219039917,
      "reward_std": 0.35382863879203796,
      "reward_std_all": 1.0881275534629822,
      "rewards-max/forensics_sinkhorn_reward": 0.9796803593635559,
      "rewards-max/forensics_strict_edge_reward": 0.6515887677669525,
      "rewards-max/forensics_tiou_reward": 0.9799536466598511,
      "rewards/forensics_sinkhorn_reward": 0.5519275069236755,
      "rewards/forensics_strict_edge_reward": 0.06916064023971558,
      "rewards/forensics_tiou_reward": 0.3815438449382782,
      "step": 358
    },
    {
      "completion_length": 27.234375,
      "epoch": 1.1013824884792627,
      "grad_norm": 5.015071392059326,
      "kl": 0.3974609375,
      "learning_rate": 7.246153846153846e-07,
      "loss": -0.1173,
      "reward": 1.064328521490097,
      "reward_std": 0.3098451793193817,
      "reward_std_all": 1.0277278125286102,
      "rewards-max/forensics_sinkhorn_reward": 0.9760836660861969,
      "rewards-max/forensics_strict_edge_reward": 0.47400766611099243,
      "rewards-max/forensics_tiou_reward": 0.9655532240867615,
      "rewards/forensics_sinkhorn_reward": 0.5996914207935333,
      "rewards/forensics_strict_edge_reward": 0.039042212069034576,
      "rewards/forensics_tiou_reward": 0.42559483647346497,
      "step": 359
    },
    {
      "completion_length": 24.265625,
      "epoch": 1.1044546850998465,
      "grad_norm": 4.964437484741211,
      "kl": 0.4140625,
      "learning_rate": 7.238461538461538e-07,
      "loss": -0.1183,
      "reward": 1.2682315111160278,
      "reward_std": 0.4486081004142761,
      "reward_std_all": 0.9629926979541779,
      "rewards-max/forensics_sinkhorn_reward": 0.9870286285877228,
      "rewards-max/forensics_strict_edge_reward": 0.5380023866891861,
      "rewards-max/forensics_tiou_reward": 0.9934344589710236,
      "rewards/forensics_sinkhorn_reward": 0.7032278478145599,
      "rewards/forensics_strict_edge_reward": 0.04177376814186573,
      "rewards/forensics_tiou_reward": 0.5232297480106354,
      "step": 360
    },
    {
      "completion_length": 23.234375,
      "epoch": 1.10752688172043,
      "grad_norm": 5.746985912322998,
      "kl": 0.349609375,
      "learning_rate": 7.23076923076923e-07,
      "loss": -0.1447,
      "reward": 1.1739716529846191,
      "reward_std": 0.20443467795848846,
      "reward_std_all": 0.9406186044216156,
      "rewards-max/forensics_sinkhorn_reward": 0.9717989265918732,
      "rewards-max/forensics_strict_edge_reward": 0.38056008517742157,
      "rewards-max/forensics_tiou_reward": 0.9587574303150177,
      "rewards/forensics_sinkhorn_reward": 0.678178459405899,
      "rewards/forensics_strict_edge_reward": 0.028790080919861794,
      "rewards/forensics_tiou_reward": 0.46700307726860046,
      "step": 361
    },
    {
      "completion_length": 26.9375,
      "epoch": 1.1105990783410138,
      "grad_norm": 5.742724895477295,
      "kl": 0.296875,
      "learning_rate": 7.223076923076922e-07,
      "loss": 0.0283,
      "reward": 0.9406249225139618,
      "reward_std": 0.2824447900056839,
      "reward_std_all": 1.071951150894165,
      "rewards-max/forensics_sinkhorn_reward": 0.9352097809314728,
      "rewards-max/forensics_strict_edge_reward": 0.3166445940732956,
      "rewards-max/forensics_tiou_reward": 0.8617836534976959,
      "rewards/forensics_sinkhorn_reward": 0.5376244783401489,
      "rewards/forensics_strict_edge_reward": 0.016650584060698748,
      "rewards/forensics_tiou_reward": 0.3863498568534851,
      "step": 362
    },
    {
      "completion_length": 27.359375,
      "epoch": 1.1136712749615976,
      "grad_norm": 5.0766777992248535,
      "kl": 0.3701171875,
      "learning_rate": 7.215384615384616e-07,
      "loss": -0.0926,
      "reward": 1.1490164995193481,
      "reward_std": 0.39663103222846985,
      "reward_std_all": 1.0080939531326294,
      "rewards-max/forensics_sinkhorn_reward": 0.9725812077522278,
      "rewards-max/forensics_strict_edge_reward": 0.5042749345302582,
      "rewards-max/forensics_tiou_reward": 0.958157867193222,
      "rewards/forensics_sinkhorn_reward": 0.6343306303024292,
      "rewards/forensics_strict_edge_reward": 0.05142340809106827,
      "rewards/forensics_tiou_reward": 0.46326231956481934,
      "step": 363
    },
    {
      "completion_length": 29.53125,
      "epoch": 1.1167434715821813,
      "grad_norm": 5.321078777313232,
      "kl": 0.357421875,
      "learning_rate": 7.207692307692307e-07,
      "loss": -0.1053,
      "reward": 1.1118272542953491,
      "reward_std": 0.26705125719308853,
      "reward_std_all": 0.9930239915847778,
      "rewards-max/forensics_sinkhorn_reward": 0.9733926057815552,
      "rewards-max/forensics_strict_edge_reward": 0.4974842667579651,
      "rewards-max/forensics_tiou_reward": 0.9625000059604645,
      "rewards/forensics_sinkhorn_reward": 0.6389552056789398,
      "rewards/forensics_strict_edge_reward": 0.029444251209497452,
      "rewards/forensics_tiou_reward": 0.4434278607368469,
      "step": 364
    },
    {
      "completion_length": 22.953125,
      "epoch": 1.119815668202765,
      "grad_norm": 7.269176959991455,
      "kl": 0.3779296875,
      "learning_rate": 7.2e-07,
      "loss": -0.0765,
      "reward": 0.996745377779007,
      "reward_std": 0.20131108909845352,
      "reward_std_all": 1.0247333645820618,
      "rewards-max/forensics_sinkhorn_reward": 0.9648179113864899,
      "rewards-max/forensics_strict_edge_reward": 0.352790892124176,
      "rewards-max/forensics_tiou_reward": 0.9390817582607269,
      "rewards/forensics_sinkhorn_reward": 0.6467311978340149,
      "rewards/forensics_strict_edge_reward": 0.02258377242833376,
      "rewards/forensics_tiou_reward": 0.3274304270744324,
      "step": 365
    },
    {
      "completion_length": 23.6875,
      "epoch": 1.1228878648233487,
      "grad_norm": 7.114365100860596,
      "kl": 0.3486328125,
      "learning_rate": 7.192307692307692e-07,
      "loss": -0.1501,
      "reward": 1.241490662097931,
      "reward_std": 0.44609442353248596,
      "reward_std_all": 0.9799650609493256,
      "rewards-max/forensics_sinkhorn_reward": 0.9838344156742096,
      "rewards-max/forensics_strict_edge_reward": 0.5404154062271118,
      "rewards-max/forensics_tiou_reward": 0.9853864014148712,
      "rewards/forensics_sinkhorn_reward": 0.6984366774559021,
      "rewards/forensics_strict_edge_reward": 0.042649658396840096,
      "rewards/forensics_tiou_reward": 0.5004043877124786,
      "step": 366
    },
    {
      "completion_length": 23.375,
      "epoch": 1.1259600614439325,
      "grad_norm": 5.066861629486084,
      "kl": 0.4384765625,
      "learning_rate": 7.184615384615384e-07,
      "loss": -0.1747,
      "reward": 1.0399115681648254,
      "reward_std": 0.39066460728645325,
      "reward_std_all": 1.0508030652999878,
      "rewards-max/forensics_sinkhorn_reward": 0.9379311501979828,
      "rewards-max/forensics_strict_edge_reward": 0.29834122955799103,
      "rewards-max/forensics_tiou_reward": 0.8892067074775696,
      "rewards/forensics_sinkhorn_reward": 0.617089033126831,
      "rewards/forensics_strict_edge_reward": 0.022824236191809177,
      "rewards/forensics_tiou_reward": 0.3999982923269272,
      "step": 367
    },
    {
      "completion_length": 26.734375,
      "epoch": 1.129032258064516,
      "grad_norm": 5.212182521820068,
      "kl": 0.3974609375,
      "learning_rate": 7.176923076923077e-07,
      "loss": -0.1111,
      "reward": 0.9563020169734955,
      "reward_std": 0.29045942425727844,
      "reward_std_all": 1.0640584230422974,
      "rewards-max/forensics_sinkhorn_reward": 0.9489140212535858,
      "rewards-max/forensics_strict_edge_reward": 0.2908714935183525,
      "rewards-max/forensics_tiou_reward": 0.8958525359630585,
      "rewards/forensics_sinkhorn_reward": 0.5813425779342651,
      "rewards/forensics_strict_edge_reward": 0.016179899452254176,
      "rewards/forensics_tiou_reward": 0.35877957940101624,
      "step": 368
    },
    {
      "completion_length": 25.984375,
      "epoch": 1.1321044546850998,
      "grad_norm": 5.379411220550537,
      "kl": 0.3916015625,
      "learning_rate": 7.169230769230768e-07,
      "loss": -0.1635,
      "reward": 1.1150933802127838,
      "reward_std": 0.20109765231609344,
      "reward_std_all": 0.9704323410987854,
      "rewards-max/forensics_sinkhorn_reward": 0.9713538289070129,
      "rewards-max/forensics_strict_edge_reward": 0.39829444885253906,
      "rewards-max/forensics_tiou_reward": 0.9459713399410248,
      "rewards/forensics_sinkhorn_reward": 0.6092172265052795,
      "rewards/forensics_strict_edge_reward": 0.01972819585353136,
      "rewards/forensics_tiou_reward": 0.486147940158844,
      "step": 369
    },
    {
      "completion_length": 26.90625,
      "epoch": 1.1351766513056836,
      "grad_norm": 6.5929131507873535,
      "kl": 0.3447265625,
      "learning_rate": 7.161538461538461e-07,
      "loss": -0.0959,
      "reward": 0.9112446904182434,
      "reward_std": 0.327040895819664,
      "reward_std_all": 1.1002082228660583,
      "rewards-max/forensics_sinkhorn_reward": 0.9594809114933014,
      "rewards-max/forensics_strict_edge_reward": 0.28488798439502716,
      "rewards-max/forensics_tiou_reward": 0.917336642742157,
      "rewards/forensics_sinkhorn_reward": 0.5155935287475586,
      "rewards/forensics_strict_edge_reward": 0.015606972854584455,
      "rewards/forensics_tiou_reward": 0.38004420697689056,
      "step": 370
    },
    {
      "completion_length": 23.171875,
      "epoch": 1.1382488479262673,
      "grad_norm": 5.216623306274414,
      "kl": 0.34765625,
      "learning_rate": 7.153846153846154e-07,
      "loss": -0.1471,
      "reward": 1.0460591316223145,
      "reward_std": 0.26878871768713,
      "reward_std_all": 1.0186520516872406,
      "rewards-max/forensics_sinkhorn_reward": 0.9623335301876068,
      "rewards-max/forensics_strict_edge_reward": 0.4322323799133301,
      "rewards-max/forensics_tiou_reward": 0.9354950785636902,
      "rewards/forensics_sinkhorn_reward": 0.6325017809867859,
      "rewards/forensics_strict_edge_reward": 0.02730189636349678,
      "rewards/forensics_tiou_reward": 0.3862554579973221,
      "step": 371
    },
    {
      "completion_length": 27.765625,
      "epoch": 1.141321044546851,
      "grad_norm": 5.223265171051025,
      "kl": 0.3408203125,
      "learning_rate": 7.146153846153846e-07,
      "loss": -0.0451,
      "reward": 1.0378650426864624,
      "reward_std": 0.3272901251912117,
      "reward_std_all": 1.0467382669448853,
      "rewards-max/forensics_sinkhorn_reward": 0.9739534854888916,
      "rewards-max/forensics_strict_edge_reward": 0.4736122563481331,
      "rewards-max/forensics_tiou_reward": 0.9630127847194672,
      "rewards/forensics_sinkhorn_reward": 0.5955173075199127,
      "rewards/forensics_strict_edge_reward": 0.02905904036015272,
      "rewards/forensics_tiou_reward": 0.4132887274026871,
      "step": 372
    },
    {
      "completion_length": 25.1875,
      "epoch": 1.1443932411674347,
      "grad_norm": 5.449169635772705,
      "kl": 0.529296875,
      "learning_rate": 7.138461538461538e-07,
      "loss": -0.1077,
      "reward": 1.09396630525589,
      "reward_std": 0.4400653839111328,
      "reward_std_all": 1.0490976572036743,
      "rewards-max/forensics_sinkhorn_reward": 0.9703558087348938,
      "rewards-max/forensics_strict_edge_reward": 0.5317803770303726,
      "rewards-max/forensics_tiou_reward": 0.9537352323532104,
      "rewards/forensics_sinkhorn_reward": 0.6237663626670837,
      "rewards/forensics_strict_edge_reward": 0.045259201899170876,
      "rewards/forensics_tiou_reward": 0.42494073510169983,
      "step": 373
    },
    {
      "completion_length": 23.78125,
      "epoch": 1.1474654377880185,
      "grad_norm": 5.9413933753967285,
      "kl": 0.4384765625,
      "learning_rate": 7.13076923076923e-07,
      "loss": -0.0719,
      "reward": 1.0341627299785614,
      "reward_std": 0.2811990976333618,
      "reward_std_all": 1.0336153209209442,
      "rewards-max/forensics_sinkhorn_reward": 0.9323677122592926,
      "rewards-max/forensics_strict_edge_reward": 0.48834480345249176,
      "rewards-max/forensics_tiou_reward": 0.869379848241806,
      "rewards/forensics_sinkhorn_reward": 0.6349761188030243,
      "rewards/forensics_strict_edge_reward": 0.02764033805578947,
      "rewards/forensics_tiou_reward": 0.37154632806777954,
      "step": 374
    },
    {
      "completion_length": 25.125,
      "epoch": 1.1505376344086022,
      "grad_norm": 4.992783069610596,
      "kl": 0.24560546875,
      "learning_rate": 7.123076923076922e-07,
      "loss": -0.0617,
      "reward": 1.0327211320400238,
      "reward_std": 0.229448102414608,
      "reward_std_all": 1.022417277097702,
      "rewards-max/forensics_sinkhorn_reward": 0.9720015525817871,
      "rewards-max/forensics_strict_edge_reward": 0.6658726632595062,
      "rewards-max/forensics_tiou_reward": 0.9618130028247833,
      "rewards/forensics_sinkhorn_reward": 0.5852696299552917,
      "rewards/forensics_strict_edge_reward": 0.046524547040462494,
      "rewards/forensics_tiou_reward": 0.40092700719833374,
      "step": 375
    },
    {
      "completion_length": 26.1875,
      "epoch": 1.1536098310291858,
      "grad_norm": 5.9419169425964355,
      "kl": 0.306640625,
      "learning_rate": 7.115384615384616e-07,
      "loss": -0.1334,
      "reward": 1.0407627820968628,
      "reward_std": 0.4137721210718155,
      "reward_std_all": 1.0766574144363403,
      "rewards-max/forensics_sinkhorn_reward": 0.9785434603691101,
      "rewards-max/forensics_strict_edge_reward": 0.5027406811714172,
      "rewards-max/forensics_tiou_reward": 0.9706897437572479,
      "rewards/forensics_sinkhorn_reward": 0.5656384825706482,
      "rewards/forensics_strict_edge_reward": 0.06196959316730499,
      "rewards/forensics_tiou_reward": 0.4131546914577484,
      "step": 376
    },
    {
      "completion_length": 25.65625,
      "epoch": 1.1566820276497696,
      "grad_norm": 5.5776848793029785,
      "kl": 0.3837890625,
      "learning_rate": 7.107692307692308e-07,
      "loss": -0.0333,
      "reward": 0.9147242605686188,
      "reward_std": 0.19442608952522278,
      "reward_std_all": 1.0572914481163025,
      "rewards-max/forensics_sinkhorn_reward": 0.9060337841510773,
      "rewards-max/forensics_strict_edge_reward": 0.13221295550465584,
      "rewards-max/forensics_tiou_reward": 0.7757317423820496,
      "rewards/forensics_sinkhorn_reward": 0.5894588530063629,
      "rewards/forensics_strict_edge_reward": 0.006746202823705971,
      "rewards/forensics_tiou_reward": 0.3185192048549652,
      "step": 377
    },
    {
      "completion_length": 24.859375,
      "epoch": 1.1597542242703534,
      "grad_norm": 5.081758499145508,
      "kl": 0.3916015625,
      "learning_rate": 7.1e-07,
      "loss": -0.035,
      "reward": 1.067895770072937,
      "reward_std": 0.3702305257320404,
      "reward_std_all": 1.04586523771286,
      "rewards-max/forensics_sinkhorn_reward": 0.9716649353504181,
      "rewards-max/forensics_strict_edge_reward": 0.4904603958129883,
      "rewards-max/forensics_tiou_reward": 0.9556110203266144,
      "rewards/forensics_sinkhorn_reward": 0.5989950299263,
      "rewards/forensics_strict_edge_reward": 0.04436040669679642,
      "rewards/forensics_tiou_reward": 0.42454028129577637,
      "step": 378
    },
    {
      "completion_length": 21.9375,
      "epoch": 1.1628264208909371,
      "grad_norm": 5.653350830078125,
      "kl": 0.4326171875,
      "learning_rate": 7.092307692307692e-07,
      "loss": -0.114,
      "reward": 0.9307125210762024,
      "reward_std": 0.3443124443292618,
      "reward_std_all": 1.0982434749603271,
      "rewards-max/forensics_sinkhorn_reward": 0.9517509341239929,
      "rewards-max/forensics_strict_edge_reward": 0.31600624322891235,
      "rewards-max/forensics_tiou_reward": 0.9075409770011902,
      "rewards/forensics_sinkhorn_reward": 0.5701783299446106,
      "rewards/forensics_strict_edge_reward": 0.020020035095512867,
      "rewards/forensics_tiou_reward": 0.3405141681432724,
      "step": 379
    },
    {
      "completion_length": 21.640625,
      "epoch": 1.1658986175115207,
      "grad_norm": 6.389430522918701,
      "kl": 0.3271484375,
      "learning_rate": 7.084615384615384e-07,
      "loss": -0.1062,
      "reward": 0.9835515320301056,
      "reward_std": 0.2546045407652855,
      "reward_std_all": 1.0464373230934143,
      "rewards-max/forensics_sinkhorn_reward": 0.9800664782524109,
      "rewards-max/forensics_strict_edge_reward": 0.2856918200850487,
      "rewards-max/forensics_tiou_reward": 0.9722311198711395,
      "rewards/forensics_sinkhorn_reward": 0.5835417211055756,
      "rewards/forensics_strict_edge_reward": 0.01413748413324356,
      "rewards/forensics_tiou_reward": 0.38587237894535065,
      "step": 380
    },
    {
      "completion_length": 22.6875,
      "epoch": 1.1689708141321045,
      "grad_norm": 5.384720325469971,
      "kl": 0.3759765625,
      "learning_rate": 7.076923076923077e-07,
      "loss": -0.0891,
      "reward": 1.0742386281490326,
      "reward_std": 0.35464996099472046,
      "reward_std_all": 1.027034044265747,
      "rewards-max/forensics_sinkhorn_reward": 0.9551670849323273,
      "rewards-max/forensics_strict_edge_reward": 0.4067516513168812,
      "rewards-max/forensics_tiou_reward": 0.9196825325489044,
      "rewards/forensics_sinkhorn_reward": 0.6432402729988098,
      "rewards/forensics_strict_edge_reward": 0.0319530563428998,
      "rewards/forensics_tiou_reward": 0.3990452438592911,
      "step": 381
    },
    {
      "completion_length": 23.609375,
      "epoch": 1.1720430107526882,
      "grad_norm": 5.255063533782959,
      "kl": 0.3701171875,
      "learning_rate": 7.069230769230768e-07,
      "loss": -0.0851,
      "reward": 0.8744516372680664,
      "reward_std": 0.2959437966346741,
      "reward_std_all": 1.1078692078590393,
      "rewards-max/forensics_sinkhorn_reward": 0.9459072947502136,
      "rewards-max/forensics_strict_edge_reward": 0.16221941635012627,
      "rewards-max/forensics_tiou_reward": 0.8856639564037323,
      "rewards/forensics_sinkhorn_reward": 0.5373579263687134,
      "rewards/forensics_strict_edge_reward": 0.007097170106135309,
      "rewards/forensics_tiou_reward": 0.32999657094478607,
      "step": 382
    },
    {
      "completion_length": 26.03125,
      "epoch": 1.1751152073732718,
      "grad_norm": 5.028295516967773,
      "kl": 0.2958984375,
      "learning_rate": 7.061538461538462e-07,
      "loss": -0.0554,
      "reward": 1.0952961444854736,
      "reward_std": 0.2060280367732048,
      "reward_std_all": 0.9861088693141937,
      "rewards-max/forensics_sinkhorn_reward": 0.9369677305221558,
      "rewards-max/forensics_strict_edge_reward": 0.3920994885265827,
      "rewards-max/forensics_tiou_reward": 0.8710989952087402,
      "rewards/forensics_sinkhorn_reward": 0.6290952265262604,
      "rewards/forensics_strict_edge_reward": 0.03969921125099063,
      "rewards/forensics_tiou_reward": 0.42650173604488373,
      "step": 383
    },
    {
      "completion_length": 27.40625,
      "epoch": 1.1781874039938556,
      "grad_norm": 5.770643711090088,
      "kl": 0.3974609375,
      "learning_rate": 7.053846153846154e-07,
      "loss": -0.1322,
      "reward": 1.1350588202476501,
      "reward_std": 0.29227402061223984,
      "reward_std_all": 0.9806513786315918,
      "rewards-max/forensics_sinkhorn_reward": 0.979376882314682,
      "rewards-max/forensics_strict_edge_reward": 0.5753126740455627,
      "rewards-max/forensics_tiou_reward": 0.9752444922924042,
      "rewards/forensics_sinkhorn_reward": 0.6403089761734009,
      "rewards/forensics_strict_edge_reward": 0.033116767182946205,
      "rewards/forensics_tiou_reward": 0.4616330862045288,
      "step": 384
    },
    {
      "completion_length": 22.8125,
      "epoch": 1.1812596006144394,
      "grad_norm": 5.587128639221191,
      "kl": 0.38671875,
      "learning_rate": 7.046153846153846e-07,
      "loss": -0.0796,
      "reward": 1.049795150756836,
      "reward_std": 0.32881394773721695,
      "reward_std_all": 1.0365872383117676,
      "rewards-max/forensics_sinkhorn_reward": 0.9710931479930878,
      "rewards-max/forensics_strict_edge_reward": 0.5054102391004562,
      "rewards-max/forensics_tiou_reward": 0.9511373937129974,
      "rewards/forensics_sinkhorn_reward": 0.6258595883846283,
      "rewards/forensics_strict_edge_reward": 0.038406046107411385,
      "rewards/forensics_tiou_reward": 0.3855295777320862,
      "step": 385
    },
    {
      "completion_length": 22.0,
      "epoch": 1.1843317972350231,
      "grad_norm": 5.033735752105713,
      "kl": 0.30419921875,
      "learning_rate": 7.038461538461539e-07,
      "loss": -0.096,
      "reward": 1.0406670272350311,
      "reward_std": 0.2522580847144127,
      "reward_std_all": 1.0180512368679047,
      "rewards-max/forensics_sinkhorn_reward": 0.9541000127792358,
      "rewards-max/forensics_strict_edge_reward": 0.19946620613336563,
      "rewards-max/forensics_tiou_reward": 0.9058766961097717,
      "rewards/forensics_sinkhorn_reward": 0.6262538135051727,
      "rewards/forensics_strict_edge_reward": 0.011149983620271087,
      "rewards/forensics_tiou_reward": 0.403263196349144,
      "step": 386
    },
    {
      "completion_length": 25.21875,
      "epoch": 1.1874039938556067,
      "grad_norm": 5.389885425567627,
      "kl": 0.3642578125,
      "learning_rate": 7.03076923076923e-07,
      "loss": -0.0571,
      "reward": 1.001161515712738,
      "reward_std": 0.32316139340400696,
      "reward_std_all": 1.058720588684082,
      "rewards-max/forensics_sinkhorn_reward": 0.9797261655330658,
      "rewards-max/forensics_strict_edge_reward": 0.31380394101142883,
      "rewards-max/forensics_tiou_reward": 0.9726492762565613,
      "rewards/forensics_sinkhorn_reward": 0.5958132147789001,
      "rewards/forensics_strict_edge_reward": 0.019951868802309036,
      "rewards/forensics_tiou_reward": 0.38539643585681915,
      "step": 387
    },
    {
      "completion_length": 23.28125,
      "epoch": 1.1904761904761905,
      "grad_norm": 5.305826663970947,
      "kl": 0.31640625,
      "learning_rate": 7.023076923076922e-07,
      "loss": -0.1028,
      "reward": 1.1385578513145447,
      "reward_std": 0.2984528988599777,
      "reward_std_all": 0.9822390079498291,
      "rewards-max/forensics_sinkhorn_reward": 0.9638718962669373,
      "rewards-max/forensics_strict_edge_reward": 0.26359419524669647,
      "rewards-max/forensics_tiou_reward": 0.9205706715583801,
      "rewards/forensics_sinkhorn_reward": 0.6884680986404419,
      "rewards/forensics_strict_edge_reward": 0.015276819467544556,
      "rewards/forensics_tiou_reward": 0.4348129332065582,
      "step": 388
    },
    {
      "completion_length": 27.71875,
      "epoch": 1.1935483870967742,
      "grad_norm": 4.698121070861816,
      "kl": 0.33203125,
      "learning_rate": 7.015384615384615e-07,
      "loss": -0.0973,
      "reward": 1.0507579445838928,
      "reward_std": 0.21740011870861053,
      "reward_std_all": 0.9998310804367065,
      "rewards-max/forensics_sinkhorn_reward": 0.9322199523448944,
      "rewards-max/forensics_strict_edge_reward": 0.11459068581461906,
      "rewards-max/forensics_tiou_reward": 0.856803834438324,
      "rewards/forensics_sinkhorn_reward": 0.6298349797725677,
      "rewards/forensics_strict_edge_reward": 0.007726170006208122,
      "rewards/forensics_tiou_reward": 0.41319677233695984,
      "step": 389
    },
    {
      "completion_length": 27.5625,
      "epoch": 1.1966205837173578,
      "grad_norm": 4.505696773529053,
      "kl": 0.3642578125,
      "learning_rate": 7.007692307692308e-07,
      "loss": -0.0378,
      "reward": 0.9864150881767273,
      "reward_std": 0.37171822786331177,
      "reward_std_all": 1.0765134692192078,
      "rewards-max/forensics_sinkhorn_reward": 0.9700039923191071,
      "rewards-max/forensics_strict_edge_reward": 0.19483278691768646,
      "rewards-max/forensics_tiou_reward": 0.9504991769790649,
      "rewards/forensics_sinkhorn_reward": 0.5686869025230408,
      "rewards/forensics_strict_edge_reward": 0.02451776433736086,
      "rewards/forensics_tiou_reward": 0.3932104557752609,
      "step": 390
    },
    {
      "completion_length": 29.0625,
      "epoch": 1.1996927803379416,
      "grad_norm": 5.267972469329834,
      "kl": 0.359375,
      "learning_rate": 7e-07,
      "loss": -0.0632,
      "reward": 1.0244807302951813,
      "reward_std": 0.2544983848929405,
      "reward_std_all": 1.0324645638465881,
      "rewards-max/forensics_sinkhorn_reward": 0.9620870053768158,
      "rewards-max/forensics_strict_edge_reward": 0.508862778544426,
      "rewards-max/forensics_tiou_reward": 0.9375229179859161,
      "rewards/forensics_sinkhorn_reward": 0.5833854675292969,
      "rewards/forensics_strict_edge_reward": 0.027178840711712837,
      "rewards/forensics_tiou_reward": 0.4139164835214615,
      "step": 391
    },
    {
      "completion_length": 28.328125,
      "epoch": 1.2027649769585254,
      "grad_norm": 4.940494060516357,
      "kl": 0.345703125,
      "learning_rate": 6.992307692307692e-07,
      "loss": -0.0975,
      "reward": 1.0990769267082214,
      "reward_std": 0.20044362545013428,
      "reward_std_all": 0.9768674969673157,
      "rewards-max/forensics_sinkhorn_reward": 0.9364610016345978,
      "rewards-max/forensics_strict_edge_reward": 0.13189373842760688,
      "rewards-max/forensics_tiou_reward": 0.8693302869796753,
      "rewards/forensics_sinkhorn_reward": 0.6468501389026642,
      "rewards/forensics_strict_edge_reward": 0.012016496352771355,
      "rewards/forensics_tiou_reward": 0.4402102828025818,
      "step": 392
    },
    {
      "completion_length": 28.390625,
      "epoch": 1.2058371735791091,
      "grad_norm": 5.354687690734863,
      "kl": 0.630859375,
      "learning_rate": 6.984615384615384e-07,
      "loss": -0.0905,
      "reward": 0.9153189659118652,
      "reward_std": 0.2680150121450424,
      "reward_std_all": 1.0912526845932007,
      "rewards-max/forensics_sinkhorn_reward": 0.9704027473926544,
      "rewards-max/forensics_strict_edge_reward": 0.3146090656518936,
      "rewards-max/forensics_tiou_reward": 0.952604353427887,
      "rewards/forensics_sinkhorn_reward": 0.5641078054904938,
      "rewards/forensics_strict_edge_reward": 0.018237571232020855,
      "rewards/forensics_tiou_reward": 0.33297350257635117,
      "step": 393
    },
    {
      "completion_length": 22.578125,
      "epoch": 1.2089093701996927,
      "grad_norm": 5.863177299499512,
      "kl": 0.3427734375,
      "learning_rate": 6.976923076923077e-07,
      "loss": -0.1533,
      "reward": 1.257658064365387,
      "reward_std": 0.3401682674884796,
      "reward_std_all": 0.9420417845249176,
      "rewards-max/forensics_sinkhorn_reward": 0.9679446816444397,
      "rewards-max/forensics_strict_edge_reward": 0.47831887006759644,
      "rewards-max/forensics_tiou_reward": 0.9489893019199371,
      "rewards/forensics_sinkhorn_reward": 0.7016923725605011,
      "rewards/forensics_strict_edge_reward": 0.05091972555965185,
      "rewards/forensics_tiou_reward": 0.505045935511589,
      "step": 394
    },
    {
      "completion_length": 24.890625,
      "epoch": 1.2119815668202765,
      "grad_norm": 4.80891227722168,
      "kl": 0.23388671875,
      "learning_rate": 6.969230769230769e-07,
      "loss": -0.141,
      "reward": 1.2472285032272339,
      "reward_std": 0.3315589278936386,
      "reward_std_all": 0.9641154110431671,
      "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": 0.6704965829849243,
      "rewards/forensics_strict_edge_reward": 0.0760113075375557,
      "rewards/forensics_tiou_reward": 0.5007207095623016,
      "step": 395
    },
    {
      "completion_length": 28.765625,
      "epoch": 1.2150537634408602,
      "grad_norm": 8.56091594696045,
      "kl": 0.345703125,
      "learning_rate": 6.961538461538461e-07,
      "loss": -0.1249,
      "reward": 1.1362449526786804,
      "reward_std": 0.4011303633451462,
      "reward_std_all": 1.0262299180030823,
      "rewards-max/forensics_sinkhorn_reward": 0.9664157629013062,
      "rewards-max/forensics_strict_edge_reward": 0.44662074744701385,
      "rewards-max/forensics_tiou_reward": 0.947177529335022,
      "rewards/forensics_sinkhorn_reward": 0.6443175375461578,
      "rewards/forensics_strict_edge_reward": 0.055097127333283424,
      "rewards/forensics_tiou_reward": 0.4368302524089813,
      "step": 396
    },
    {
      "completion_length": 28.15625,
      "epoch": 1.218125960061444,
      "grad_norm": 6.388549327850342,
      "kl": 0.4326171875,
      "learning_rate": 6.953846153846154e-07,
      "loss": -0.1145,
      "reward": 0.910356342792511,
      "reward_std": 0.2847587466239929,
      "reward_std_all": 1.0855141878128052,
      "rewards-max/forensics_sinkhorn_reward": 0.8994293510913849,
      "rewards-max/forensics_strict_edge_reward": 0.10200458252802491,
      "rewards-max/forensics_tiou_reward": 0.7767573893070221,
      "rewards/forensics_sinkhorn_reward": 0.5192265808582306,
      "rewards/forensics_strict_edge_reward": 0.010289667516190093,
      "rewards/forensics_tiou_reward": 0.38084010779857635,
      "step": 397
    },
    {
      "completion_length": 23.96875,
      "epoch": 1.2211981566820276,
      "grad_norm": 7.178316116333008,
      "kl": 0.408203125,
      "learning_rate": 6.946153846153846e-07,
      "loss": 0.0066,
      "reward": 1.002001941204071,
      "reward_std": 0.2989732250571251,
      "reward_std_all": 1.0436384677886963,
      "rewards-max/forensics_sinkhorn_reward": 0.9595151245594025,
      "rewards-max/forensics_strict_edge_reward": 0.20722780004143715,
      "rewards-max/forensics_tiou_reward": 0.9100911617279053,
      "rewards/forensics_sinkhorn_reward": 0.5900208652019501,
      "rewards/forensics_strict_edge_reward": 0.011019902885891497,
      "rewards/forensics_tiou_reward": 0.40096113085746765,
      "step": 398
    },
    {
      "completion_length": 25.921875,
      "epoch": 1.2242703533026114,
      "grad_norm": 5.957186222076416,
      "kl": 0.44140625,
      "learning_rate": 6.938461538461539e-07,
      "loss": -0.0408,
      "reward": 1.2264543175697327,
      "reward_std": 0.2999657988548279,
      "reward_std_all": 0.9351775646209717,
      "rewards-max/forensics_sinkhorn_reward": 0.9676626920700073,
      "rewards-max/forensics_strict_edge_reward": 0.31119438260793686,
      "rewards-max/forensics_tiou_reward": 0.938471257686615,
      "rewards/forensics_sinkhorn_reward": 0.6355240046977997,
      "rewards/forensics_strict_edge_reward": 0.02658720500767231,
      "rewards/forensics_tiou_reward": 0.564343124628067,
      "step": 399
    },
    {
      "completion_length": 28.328125,
      "epoch": 1.2273425499231951,
      "grad_norm": 5.652005195617676,
      "kl": 0.3486328125,
      "learning_rate": 6.93076923076923e-07,
      "loss": -0.1229,
      "reward": 1.2858232855796814,
      "reward_std": 0.3414626568555832,
      "reward_std_all": 0.9262384474277496,
      "rewards-max/forensics_sinkhorn_reward": 0.9785223603248596,
      "rewards-max/forensics_strict_edge_reward": 0.38551829755306244,
      "rewards-max/forensics_tiou_reward": 0.9737612903118134,
      "rewards/forensics_sinkhorn_reward": 0.7061139941215515,
      "rewards/forensics_strict_edge_reward": 0.042968062218278646,
      "rewards/forensics_tiou_reward": 0.536741241812706,
      "step": 400
    },
    {
      "completion_length": 25.015625,
      "epoch": 1.230414746543779,
      "grad_norm": 7.009801387786865,
      "kl": 0.3681640625,
      "learning_rate": 6.923076923076922e-07,
      "loss": -0.1339,
      "reward": 1.1138669848442078,
      "reward_std": 0.3314519077539444,
      "reward_std_all": 1.014562875032425,
      "rewards-max/forensics_sinkhorn_reward": 0.9747835099697113,
      "rewards-max/forensics_strict_edge_reward": 0.5221183001995087,
      "rewards-max/forensics_tiou_reward": 0.9598776698112488,
      "rewards/forensics_sinkhorn_reward": 0.6077185869216919,
      "rewards/forensics_strict_edge_reward": 0.04606838524341583,
      "rewards/forensics_tiou_reward": 0.4600800573825836,
      "step": 401
    },
    {
      "completion_length": 27.09375,
      "epoch": 1.2334869431643625,
      "grad_norm": 6.616031169891357,
      "kl": 0.62890625,
      "learning_rate": 6.915384615384615e-07,
      "loss": -0.0699,
      "reward": 0.8612936735153198,
      "reward_std": 0.2979234606027603,
      "reward_std_all": 1.1143790185451508,
      "rewards-max/forensics_sinkhorn_reward": 0.9593159556388855,
      "rewards-max/forensics_strict_edge_reward": 0.17799849868606543,
      "rewards-max/forensics_tiou_reward": 0.9286996126174927,
      "rewards/forensics_sinkhorn_reward": 0.5243171155452728,
      "rewards/forensics_strict_edge_reward": 0.013769997060080641,
      "rewards/forensics_tiou_reward": 0.3232066035270691,
      "step": 402
    },
    {
      "completion_length": 24.09375,
      "epoch": 1.2365591397849462,
      "grad_norm": 5.942255020141602,
      "kl": 0.310546875,
      "learning_rate": 6.907692307692307e-07,
      "loss": -0.0966,
      "reward": 1.1741278171539307,
      "reward_std": 0.20678269863128662,
      "reward_std_all": 0.9470551908016205,
      "rewards-max/forensics_sinkhorn_reward": 0.9724513292312622,
      "rewards-max/forensics_strict_edge_reward": 0.18227707967162132,
      "rewards-max/forensics_tiou_reward": 0.9613374769687653,
      "rewards/forensics_sinkhorn_reward": 0.7038819789886475,
      "rewards/forensics_strict_edge_reward": 0.008674698299728334,
      "rewards/forensics_tiou_reward": 0.46157118678092957,
      "step": 403
    },
    {
      "completion_length": 26.75,
      "epoch": 1.23963133640553,
      "grad_norm": 3.996500015258789,
      "kl": 0.357421875,
      "learning_rate": 6.9e-07,
      "loss": -0.1693,
      "reward": 1.2904101014137268,
      "reward_std": 0.2962725758552551,
      "reward_std_all": 0.9107685387134552,
      "rewards-max/forensics_sinkhorn_reward": 0.9700607657432556,
      "rewards-max/forensics_strict_edge_reward": 0.4724559336900711,
      "rewards-max/forensics_tiou_reward": 0.9561460018157959,
      "rewards/forensics_sinkhorn_reward": 0.7108466029167175,
      "rewards/forensics_strict_edge_reward": 0.04465927928686142,
      "rewards/forensics_tiou_reward": 0.5349041819572449,
      "step": 404
    },
    {
      "completion_length": 26.296875,
      "epoch": 1.2427035330261136,
      "grad_norm": 5.622199058532715,
      "kl": 0.458984375,
      "learning_rate": 6.892307692307692e-07,
      "loss": -0.1543,
      "reward": 1.037874460220337,
      "reward_std": 0.20068275928497314,
      "reward_std_all": 1.006592035293579,
      "rewards-max/forensics_sinkhorn_reward": 0.9818831980228424,
      "rewards-max/forensics_strict_edge_reward": 0.40814805030822754,
      "rewards-max/forensics_tiou_reward": 0.9834610521793365,
      "rewards/forensics_sinkhorn_reward": 0.5972222685813904,
      "rewards/forensics_strict_edge_reward": 0.024532722309231758,
      "rewards/forensics_tiou_reward": 0.4161195158958435,
      "step": 405
    },
    {
      "completion_length": 23.34375,
      "epoch": 1.2457757296466974,
      "grad_norm": 4.6191582679748535,
      "kl": 0.388671875,
      "learning_rate": 6.884615384615384e-07,
      "loss": -0.102,
      "reward": 1.1571995615959167,
      "reward_std": 0.4506067931652069,
      "reward_std_all": 1.014286756515503,
      "rewards-max/forensics_sinkhorn_reward": 0.9808034896850586,
      "rewards-max/forensics_strict_edge_reward": 0.4680478796362877,
      "rewards-max/forensics_tiou_reward": 0.975511223077774,
      "rewards/forensics_sinkhorn_reward": 0.6674416065216064,
      "rewards/forensics_strict_edge_reward": 0.055094762705266476,
      "rewards/forensics_tiou_reward": 0.43466317653656006,
      "step": 406
    },
    {
      "completion_length": 25.0,
      "epoch": 1.2488479262672811,
      "grad_norm": 4.350093364715576,
      "kl": 0.265625,
      "learning_rate": 6.876923076923077e-07,
      "loss": -0.1769,
      "reward": 1.0103348195552826,
      "reward_std": 0.22677677869796753,
      "reward_std_all": 1.027395486831665,
      "rewards-max/forensics_sinkhorn_reward": 0.9276162385940552,
      "rewards-max/forensics_strict_edge_reward": 0.4746674597263336,
      "rewards-max/forensics_tiou_reward": 0.842084527015686,
      "rewards/forensics_sinkhorn_reward": 0.6272152364253998,
      "rewards/forensics_strict_edge_reward": 0.025840661488473415,
      "rewards/forensics_tiou_reward": 0.357278972864151,
      "step": 407
    },
    {
      "completion_length": 25.1875,
      "epoch": 1.251920122887865,
      "grad_norm": 4.713408470153809,
      "kl": 0.330078125,
      "learning_rate": 6.869230769230769e-07,
      "loss": -0.0593,
      "reward": 0.9045682549476624,
      "reward_std": 0.2673479989171028,
      "reward_std_all": 1.0816174149513245,
      "rewards-max/forensics_sinkhorn_reward": 0.9497115015983582,
      "rewards-max/forensics_strict_edge_reward": 0.20542107336223125,
      "rewards-max/forensics_tiou_reward": 0.9057041108608246,
      "rewards/forensics_sinkhorn_reward": 0.5544607043266296,
      "rewards/forensics_strict_edge_reward": 0.010923554422333837,
      "rewards/forensics_tiou_reward": 0.3391840010881424,
      "step": 408
    },
    {
      "completion_length": 25.34375,
      "epoch": 1.2549923195084485,
      "grad_norm": 5.599226951599121,
      "kl": 0.3486328125,
      "learning_rate": 6.861538461538461e-07,
      "loss": -0.0821,
      "reward": 1.0250828266143799,
      "reward_std": 0.3184124231338501,
      "reward_std_all": 1.045773208141327,
      "rewards-max/forensics_sinkhorn_reward": 0.9734918177127838,
      "rewards-max/forensics_strict_edge_reward": 0.4974842667579651,
      "rewards-max/forensics_tiou_reward": 0.9632557034492493,
      "rewards/forensics_sinkhorn_reward": 0.5960206687450409,
      "rewards/forensics_strict_edge_reward": 0.027934412471950054,
      "rewards/forensics_tiou_reward": 0.40112778544425964,
      "step": 409
    },
    {
      "completion_length": 24.9375,
      "epoch": 1.2580645161290323,
      "grad_norm": 4.7680511474609375,
      "kl": 0.326171875,
      "learning_rate": 6.853846153846153e-07,
      "loss": -0.1173,
      "reward": 1.0772022604942322,
      "reward_std": 0.3483975678682327,
      "reward_std_all": 1.028984695672989,
      "rewards-max/forensics_sinkhorn_reward": 0.9749576449394226,
      "rewards-max/forensics_strict_edge_reward": 0.5311143696308136,
      "rewards-max/forensics_tiou_reward": 0.9675136208534241,
      "rewards/forensics_sinkhorn_reward": 0.6281048953533173,
      "rewards/forensics_strict_edge_reward": 0.03491454757750034,
      "rewards/forensics_tiou_reward": 0.41418278217315674,
      "step": 410
    },
    {
      "completion_length": 26.34375,
      "epoch": 1.261136712749616,
      "grad_norm": 6.1870622634887695,
      "kl": 0.357421875,
      "learning_rate": 6.846153846153846e-07,
      "loss": -0.0615,
      "reward": 1.0674918293952942,
      "reward_std": 0.3323226124048233,
      "reward_std_all": 1.0339370369911194,
      "rewards-max/forensics_sinkhorn_reward": 0.9786326885223389,
      "rewards-max/forensics_strict_edge_reward": 0.46869315207004547,
      "rewards-max/forensics_tiou_reward": 0.9741694927215576,
      "rewards/forensics_sinkhorn_reward": 0.6266728341579437,
      "rewards/forensics_strict_edge_reward": 0.03610539436340332,
      "rewards/forensics_tiou_reward": 0.40471363067626953,
      "step": 411
    },
    {
      "completion_length": 26.65625,
      "epoch": 1.2642089093701996,
      "grad_norm": 5.91439151763916,
      "kl": 0.3017578125,
      "learning_rate": 6.838461538461539e-07,
      "loss": -0.0115,
      "reward": 1.0378987491130829,
      "reward_std": 0.40271593630313873,
      "reward_std_all": 1.0553803443908691,
      "rewards-max/forensics_sinkhorn_reward": 0.9631219804286957,
      "rewards-max/forensics_strict_edge_reward": 0.20265214890241623,
      "rewards-max/forensics_tiou_reward": 0.9243342876434326,
      "rewards/forensics_sinkhorn_reward": 0.5725643783807755,
      "rewards/forensics_strict_edge_reward": 0.023614725098013878,
      "rewards/forensics_tiou_reward": 0.44171957671642303,
      "step": 412
    },
    {
      "completion_length": 25.859375,
      "epoch": 1.2672811059907834,
      "grad_norm": 5.518067836761475,
      "kl": 0.3779296875,
      "learning_rate": 6.83076923076923e-07,
      "loss": -0.0473,
      "reward": 1.0135936737060547,
      "reward_std": 0.30648456513881683,
      "reward_std_all": 1.0398051738739014,
      "rewards-max/forensics_sinkhorn_reward": 0.9505794644355774,
      "rewards-max/forensics_strict_edge_reward": 0.2728862054646015,
      "rewards-max/forensics_tiou_reward": 0.9013469517230988,
      "rewards/forensics_sinkhorn_reward": 0.5889452993869781,
      "rewards/forensics_strict_edge_reward": 0.018847887869924307,
      "rewards/forensics_tiou_reward": 0.4058004766702652,
      "step": 413
    },
    {
      "completion_length": 24.078125,
      "epoch": 1.2703533026113671,
      "grad_norm": 5.7059478759765625,
      "kl": 0.4443359375,
      "learning_rate": 6.823076923076922e-07,
      "loss": -0.0911,
      "reward": 0.9767709374427795,
      "reward_std": 0.39434683322906494,
      "reward_std_all": 1.1066842079162598,
      "rewards-max/forensics_sinkhorn_reward": 0.9785346686840057,
      "rewards-max/forensics_strict_edge_reward": 0.6149742305278778,
      "rewards-max/forensics_tiou_reward": 0.9738822877407074,
      "rewards/forensics_sinkhorn_reward": 0.5508450120687485,
      "rewards/forensics_strict_edge_reward": 0.04454851523041725,
      "rewards/forensics_tiou_reward": 0.38137732446193695,
      "step": 414
    },
    {
      "completion_length": 27.421875,
      "epoch": 1.273425499231951,
      "grad_norm": 6.022080898284912,
      "kl": 0.5,
      "learning_rate": 6.815384615384615e-07,
      "loss": -0.0691,
      "reward": 1.062913864850998,
      "reward_std": 0.37143993377685547,
      "reward_std_all": 1.0364636778831482,
      "rewards-max/forensics_sinkhorn_reward": 0.9768050014972687,
      "rewards-max/forensics_strict_edge_reward": 0.3666907325387001,
      "rewards-max/forensics_tiou_reward": 0.9729275107383728,
      "rewards/forensics_sinkhorn_reward": 0.6013589799404144,
      "rewards/forensics_strict_edge_reward": 0.02549159061163664,
      "rewards/forensics_tiou_reward": 0.4360632747411728,
      "step": 415
    },
    {
      "completion_length": 28.984375,
      "epoch": 1.2764976958525347,
      "grad_norm": 5.306396007537842,
      "kl": 0.345703125,
      "learning_rate": 6.807692307692307e-07,
      "loss": -0.073,
      "reward": 0.9373859167098999,
      "reward_std": 0.3501080572605133,
      "reward_std_all": 1.0882798433303833,
      "rewards-max/forensics_sinkhorn_reward": 0.9581810235977173,
      "rewards-max/forensics_strict_edge_reward": 0.182683527469635,
      "rewards-max/forensics_tiou_reward": 0.9123392403125763,
      "rewards/forensics_sinkhorn_reward": 0.5491493940353394,
      "rewards/forensics_strict_edge_reward": 0.012790616136044264,
      "rewards/forensics_tiou_reward": 0.3754459172487259,
      "step": 416
    },
    {
      "completion_length": 23.921875,
      "epoch": 1.2795698924731183,
      "grad_norm": 5.665843486785889,
      "kl": 0.2734375,
      "learning_rate": 6.800000000000001e-07,
      "loss": -0.1464,
      "reward": 1.0196166634559631,
      "reward_std": 0.3229382038116455,
      "reward_std_all": 1.0483102202415466,
      "rewards-max/forensics_sinkhorn_reward": 0.9656273126602173,
      "rewards-max/forensics_strict_edge_reward": 0.40283074975013733,
      "rewards-max/forensics_tiou_reward": 0.9420878887176514,
      "rewards/forensics_sinkhorn_reward": 0.6090449094772339,
      "rewards/forensics_strict_edge_reward": 0.038087693974375725,
      "rewards/forensics_tiou_reward": 0.37248410284519196,
      "step": 417
    },
    {
      "completion_length": 26.375,
      "epoch": 1.282642089093702,
      "grad_norm": 5.892461776733398,
      "kl": 0.4541015625,
      "learning_rate": 6.792307692307692e-07,
      "loss": -0.1294,
      "reward": 0.9986247718334198,
      "reward_std": 0.3716515302658081,
      "reward_std_all": 1.0774862468242645,
      "rewards-max/forensics_sinkhorn_reward": 0.9619835615158081,
      "rewards-max/forensics_strict_edge_reward": 0.20439643040299416,
      "rewards-max/forensics_tiou_reward": 0.9292812645435333,
      "rewards/forensics_sinkhorn_reward": 0.5583356469869614,
      "rewards/forensics_strict_edge_reward": 0.025001075817272067,
      "rewards/forensics_tiou_reward": 0.4152880609035492,
      "step": 418
    },
    {
      "completion_length": 21.421875,
      "epoch": 1.2857142857142856,
      "grad_norm": 5.118348598480225,
      "kl": 0.3115234375,
      "learning_rate": 6.784615384615384e-07,
      "loss": -0.0636,
      "reward": 1.0836535692214966,
      "reward_std": 0.235909104347229,
      "reward_std_all": 0.9894147217273712,
      "rewards-max/forensics_sinkhorn_reward": 0.9543246924877167,
      "rewards-max/forensics_strict_edge_reward": 0.2967786379158497,
      "rewards-max/forensics_tiou_reward": 0.9140336811542511,
      "rewards/forensics_sinkhorn_reward": 0.6516087353229523,
      "rewards/forensics_strict_edge_reward": 0.02379266917705536,
      "rewards/forensics_tiou_reward": 0.40825219452381134,
      "step": 419
    },
    {
      "completion_length": 28.28125,
      "epoch": 1.2887864823348694,
      "grad_norm": 5.208121299743652,
      "kl": 0.30078125,
      "learning_rate": 6.776923076923077e-07,
      "loss": -0.0823,
      "reward": 0.9787676632404327,
      "reward_std": 0.37195374071598053,
      "reward_std_all": 1.0844590663909912,
      "rewards-max/forensics_sinkhorn_reward": 0.9729769825935364,
      "rewards-max/forensics_strict_edge_reward": 0.57436802983284,
      "rewards-max/forensics_tiou_reward": 0.9575320482254028,
      "rewards/forensics_sinkhorn_reward": 0.5598516464233398,
      "rewards/forensics_strict_edge_reward": 0.03325771912932396,
      "rewards/forensics_tiou_reward": 0.38565830886363983,
      "step": 420
    },
    {
      "completion_length": 27.453125,
      "epoch": 1.2918586789554531,
      "grad_norm": 5.524505138397217,
      "kl": 0.3515625,
      "learning_rate": 6.769230769230769e-07,
      "loss": -0.0363,
      "reward": 1.0602984428405762,
      "reward_std": 0.18658673018217087,
      "reward_std_all": 0.9980809092521667,
      "rewards-max/forensics_sinkhorn_reward": 0.9667796194553375,
      "rewards-max/forensics_strict_edge_reward": 0.26760164089500904,
      "rewards-max/forensics_tiou_reward": 0.9408605694770813,
      "rewards/forensics_sinkhorn_reward": 0.6186588704586029,
      "rewards/forensics_strict_edge_reward": 0.021445168182253838,
      "rewards/forensics_tiou_reward": 0.4201944172382355,
      "step": 421
    },
    {
      "completion_length": 24.3125,
      "epoch": 1.294930875576037,
      "grad_norm": 4.466542720794678,
      "kl": 0.443359375,
      "learning_rate": 6.761538461538461e-07,
      "loss": -0.066,
      "reward": 1.0182490944862366,
      "reward_std": 0.34588785469532013,
      "reward_std_all": 1.0508608222007751,
      "rewards-max/forensics_sinkhorn_reward": 0.9475760459899902,
      "rewards-max/forensics_strict_edge_reward": 0.26321638375520706,
      "rewards-max/forensics_tiou_reward": 0.8986808955669403,
      "rewards/forensics_sinkhorn_reward": 0.6112946569919586,
      "rewards/forensics_strict_edge_reward": 0.016943348105996847,
      "rewards/forensics_tiou_reward": 0.3900110721588135,
      "step": 422
    },
    {
      "completion_length": 24.5,
      "epoch": 1.2980030721966207,
      "grad_norm": 7.003297328948975,
      "kl": 0.58203125,
      "learning_rate": 6.753846153846153e-07,
      "loss": -0.0831,
      "reward": 1.1749613881111145,
      "reward_std": 0.31944385170936584,
      "reward_std_all": 0.9678595662117004,
      "rewards-max/forensics_sinkhorn_reward": 0.9517969191074371,
      "rewards-max/forensics_strict_edge_reward": 0.46695858240127563,
      "rewards-max/forensics_tiou_reward": 0.9152610003948212,
      "rewards/forensics_sinkhorn_reward": 0.6850824654102325,
      "rewards/forensics_strict_edge_reward": 0.0332259563729167,
      "rewards/forensics_tiou_reward": 0.4566529393196106,
      "step": 423
    },
    {
      "completion_length": 24.140625,
      "epoch": 1.3010752688172043,
      "grad_norm": 5.459794044494629,
      "kl": 0.3291015625,
      "learning_rate": 6.746153846153845e-07,
      "loss": -0.0337,
      "reward": 1.0487747192382812,
      "reward_std": 0.2665755897760391,
      "reward_std_all": 1.0122016370296478,
      "rewards-max/forensics_sinkhorn_reward": 0.9410355985164642,
      "rewards-max/forensics_strict_edge_reward": 0.09549374133348465,
      "rewards-max/forensics_tiou_reward": 0.8691928088665009,
      "rewards/forensics_sinkhorn_reward": 0.6160432398319244,
      "rewards/forensics_strict_edge_reward": 0.0055689000291749835,
      "rewards/forensics_tiou_reward": 0.4271625876426697,
      "step": 424
    },
    {
      "completion_length": 25.015625,
      "epoch": 1.304147465437788,
      "grad_norm": 5.367672443389893,
      "kl": 0.302734375,
      "learning_rate": 6.738461538461539e-07,
      "loss": -0.0898,
      "reward": 1.148272693157196,
      "reward_std": 0.34590378403663635,
      "reward_std_all": 0.9865761995315552,
      "rewards-max/forensics_sinkhorn_reward": 0.9670530259609222,
      "rewards-max/forensics_strict_edge_reward": 0.4117985963821411,
      "rewards-max/forensics_tiou_reward": 0.9390880167484283,
      "rewards/forensics_sinkhorn_reward": 0.6734912395477295,
      "rewards/forensics_strict_edge_reward": 0.03876871429383755,
      "rewards/forensics_tiou_reward": 0.43601273000240326,
      "step": 425
    },
    {
      "completion_length": 22.65625,
      "epoch": 1.3072196620583718,
      "grad_norm": 6.473650932312012,
      "kl": 0.3173828125,
      "learning_rate": 6.730769230769231e-07,
      "loss": -0.0581,
      "reward": 1.018636405467987,
      "reward_std": 0.1472989097237587,
      "reward_std_all": 1.0001685321331024,
      "rewards-max/forensics_sinkhorn_reward": 0.8991037905216217,
      "rewards-max/forensics_strict_edge_reward": 0.1175652202218771,
      "rewards-max/forensics_tiou_reward": 0.7782965898513794,
      "rewards/forensics_sinkhorn_reward": 0.6465914249420166,
      "rewards/forensics_strict_edge_reward": 0.005957024404779077,
      "rewards/forensics_tiou_reward": 0.366087943315506,
      "step": 426
    },
    {
      "completion_length": 25.515625,
      "epoch": 1.3102918586789554,
      "grad_norm": 4.8842034339904785,
      "kl": 0.4345703125,
      "learning_rate": 6.723076923076922e-07,
      "loss": -0.0819,
      "reward": 1.3928070068359375,
      "reward_std": 0.29805080592632294,
      "reward_std_all": 0.8519833385944366,
      "rewards-max/forensics_sinkhorn_reward": 0.9770533442497253,
      "rewards-max/forensics_strict_edge_reward": 0.30067282915115356,
      "rewards-max/forensics_tiou_reward": 0.9686298072338104,
      "rewards/forensics_sinkhorn_reward": 0.7757484018802643,
      "rewards/forensics_strict_edge_reward": 0.026392404921352863,
      "rewards/forensics_tiou_reward": 0.5906661152839661,
      "step": 427
    },
    {
      "completion_length": 24.1875,
      "epoch": 1.3133640552995391,
      "grad_norm": 7.5467209815979,
      "kl": 0.3837890625,
      "learning_rate": 6.715384615384615e-07,
      "loss": -0.1147,
      "reward": 1.263374924659729,
      "reward_std": 0.3268241584300995,
      "reward_std_all": 0.9335523545742035,
      "rewards-max/forensics_sinkhorn_reward": 0.9688872992992401,
      "rewards-max/forensics_strict_edge_reward": 0.4982926547527313,
      "rewards-max/forensics_tiou_reward": 0.9461445808410645,
      "rewards/forensics_sinkhorn_reward": 0.6941404044628143,
      "rewards/forensics_strict_edge_reward": 0.05226134695112705,
      "rewards/forensics_tiou_reward": 0.5169731974601746,
      "step": 428
    },
    {
      "completion_length": 25.1875,
      "epoch": 1.316436251920123,
      "grad_norm": 5.894674777984619,
      "kl": 0.419921875,
      "learning_rate": 6.707692307692307e-07,
      "loss": -0.0758,
      "reward": 1.2132096886634827,
      "reward_std": 0.3311910629272461,
      "reward_std_all": 0.9554855227470398,
      "rewards-max/forensics_sinkhorn_reward": 0.966854065656662,
      "rewards-max/forensics_strict_edge_reward": 0.5313882380723953,
      "rewards-max/forensics_tiou_reward": 0.9416230022907257,
      "rewards/forensics_sinkhorn_reward": 0.6960171461105347,
      "rewards/forensics_strict_edge_reward": 0.032996928319334984,
      "rewards/forensics_tiou_reward": 0.48419561982154846,
      "step": 429
    },
    {
      "completion_length": 28.578125,
      "epoch": 1.3195084485407067,
      "grad_norm": 4.758193492889404,
      "kl": 0.349609375,
      "learning_rate": 6.7e-07,
      "loss": -0.0987,
      "reward": 1.1805209517478943,
      "reward_std": 0.23656989634037018,
      "reward_std_all": 0.9545217156410217,
      "rewards-max/forensics_sinkhorn_reward": 0.9791232049465179,
      "rewards-max/forensics_strict_edge_reward": 0.5676676407456398,
      "rewards-max/forensics_tiou_reward": 0.9765526354312897,
      "rewards/forensics_sinkhorn_reward": 0.6613509654998779,
      "rewards/forensics_strict_edge_reward": 0.038312291726469994,
      "rewards/forensics_tiou_reward": 0.4808577001094818,
      "step": 430
    },
    {
      "completion_length": 25.96875,
      "epoch": 1.3225806451612903,
      "grad_norm": 5.620765209197998,
      "kl": 0.4775390625,
      "learning_rate": 6.692307692307692e-07,
      "loss": -0.1316,
      "reward": 1.1555036306381226,
      "reward_std": 0.46467773616313934,
      "reward_std_all": 1.0339066982269287,
      "rewards-max/forensics_sinkhorn_reward": 0.9841019213199615,
      "rewards-max/forensics_strict_edge_reward": 0.564324826002121,
      "rewards-max/forensics_tiou_reward": 0.9861505627632141,
      "rewards/forensics_sinkhorn_reward": 0.6411682367324829,
      "rewards/forensics_strict_edge_reward": 0.06865765526890755,
      "rewards/forensics_tiou_reward": 0.4456777572631836,
      "step": 431
    },
    {
      "completion_length": 25.34375,
      "epoch": 1.325652841781874,
      "grad_norm": 5.769843101501465,
      "kl": 0.3232421875,
      "learning_rate": 6.684615384615384e-07,
      "loss": -0.0839,
      "reward": 1.0428507328033447,
      "reward_std": 0.34440408647060394,
      "reward_std_all": 1.044481873512268,
      "rewards-max/forensics_sinkhorn_reward": 0.968422532081604,
      "rewards-max/forensics_strict_edge_reward": 0.42156922072172165,
      "rewards-max/forensics_tiou_reward": 0.9483910501003265,
      "rewards/forensics_sinkhorn_reward": 0.6170158982276917,
      "rewards/forensics_strict_edge_reward": 0.023712913040071726,
      "rewards/forensics_tiou_reward": 0.4021219462156296,
      "step": 432
    },
    {
      "completion_length": 25.421875,
      "epoch": 1.3287250384024578,
      "grad_norm": 5.758223533630371,
      "kl": 0.314453125,
      "learning_rate": 6.676923076923077e-07,
      "loss": -0.1566,
      "reward": 1.2242311835289001,
      "reward_std": 0.3739432841539383,
      "reward_std_all": 0.9656614065170288,
      "rewards-max/forensics_sinkhorn_reward": 0.9655021131038666,
      "rewards-max/forensics_strict_edge_reward": 0.3668999159708619,
      "rewards-max/forensics_tiou_reward": 0.935403972864151,
      "rewards/forensics_sinkhorn_reward": 0.665033757686615,
      "rewards/forensics_strict_edge_reward": 0.04108126857317984,
      "rewards/forensics_tiou_reward": 0.5181161761283875,
      "step": 433
    },
    {
      "completion_length": 30.234375,
      "epoch": 1.3317972350230414,
      "grad_norm": 10.085370063781738,
      "kl": 0.880859375,
      "learning_rate": 6.669230769230769e-07,
      "loss": -0.0696,
      "reward": 1.1778995394706726,
      "reward_std": 0.3273473083972931,
      "reward_std_all": 0.9739426374435425,
      "rewards-max/forensics_sinkhorn_reward": 0.9635456204414368,
      "rewards-max/forensics_strict_edge_reward": 0.2597980201244354,
      "rewards-max/forensics_tiou_reward": 0.9429921507835388,
      "rewards/forensics_sinkhorn_reward": 0.6529793441295624,
      "rewards/forensics_strict_edge_reward": 0.02633414790034294,
      "rewards/forensics_tiou_reward": 0.49858608841896057,
      "step": 434
    },
    {
      "completion_length": 27.328125,
      "epoch": 1.3348694316436251,
      "grad_norm": 5.384067535400391,
      "kl": 0.3076171875,
      "learning_rate": 6.661538461538462e-07,
      "loss": -0.1527,
      "reward": 1.2064940929412842,
      "reward_std": 0.3555985540151596,
      "reward_std_all": 0.9653327763080597,
      "rewards-max/forensics_sinkhorn_reward": 0.9777461588382721,
      "rewards-max/forensics_strict_edge_reward": 0.4046127200126648,
      "rewards-max/forensics_tiou_reward": 0.9713387191295624,
      "rewards/forensics_sinkhorn_reward": 0.6726377904415131,
      "rewards/forensics_strict_edge_reward": 0.04562687408179045,
      "rewards/forensics_tiou_reward": 0.4882293939590454,
      "step": 435
    },
    {
      "completion_length": 25.875,
      "epoch": 1.337941628264209,
      "grad_norm": 5.16735315322876,
      "kl": 0.3828125,
      "learning_rate": 6.653846153846153e-07,
      "loss": -0.1288,
      "reward": 1.135844349861145,
      "reward_std": 0.314035564661026,
      "reward_std_all": 0.9872426092624664,
      "rewards-max/forensics_sinkhorn_reward": 0.9595915973186493,
      "rewards-max/forensics_strict_edge_reward": 0.32768531888723373,
      "rewards-max/forensics_tiou_reward": 0.9294065535068512,
      "rewards/forensics_sinkhorn_reward": 0.6600291132926941,
      "rewards/forensics_strict_edge_reward": 0.02147760707885027,
      "rewards/forensics_tiou_reward": 0.4543377012014389,
      "step": 436
    },
    {
      "completion_length": 25.90625,
      "epoch": 1.3410138248847927,
      "grad_norm": 6.610569477081299,
      "kl": 0.51171875,
      "learning_rate": 6.646153846153845e-07,
      "loss": -0.0987,
      "reward": 1.1943270564079285,
      "reward_std": 0.3421112447977066,
      "reward_std_all": 0.9632294476032257,
      "rewards-max/forensics_sinkhorn_reward": 0.9801358878612518,
      "rewards-max/forensics_strict_edge_reward": 0.4409833997488022,
      "rewards-max/forensics_tiou_reward": 0.9770318567752838,
      "rewards/forensics_sinkhorn_reward": 0.694523274898529,
      "rewards/forensics_strict_edge_reward": 0.03101264126598835,
      "rewards/forensics_tiou_reward": 0.4687911570072174,
      "step": 437
    },
    {
      "completion_length": 24.5,
      "epoch": 1.3440860215053765,
      "grad_norm": 7.048917293548584,
      "kl": 0.4091796875,
      "learning_rate": 6.638461538461539e-07,
      "loss": -0.1121,
      "reward": 0.7818309366703033,
      "reward_std": 0.11448604986071587,
      "reward_std_all": 1.113900065422058,
      "rewards-max/forensics_sinkhorn_reward": 0.9278853833675385,
      "rewards-max/forensics_strict_edge_reward": 0.0677612759245676,
      "rewards-max/forensics_tiou_reward": 0.8575157225131989,
      "rewards/forensics_sinkhorn_reward": 0.5261073410511017,
      "rewards/forensics_strict_edge_reward": 0.0028513452066363243,
      "rewards/forensics_tiou_reward": 0.2528722435235977,
      "step": 438
    },
    {
      "completion_length": 25.9375,
      "epoch": 1.34715821812596,
      "grad_norm": 6.405598163604736,
      "kl": 0.3505859375,
      "learning_rate": 6.630769230769231e-07,
      "loss": -0.1526,
      "reward": 0.9759821593761444,
      "reward_std": 0.39489229023456573,
      "reward_std_all": 1.0992809534072876,
      "rewards-max/forensics_sinkhorn_reward": 0.9780255854129791,
      "rewards-max/forensics_strict_edge_reward": 0.3818294107913971,
      "rewards-max/forensics_tiou_reward": 0.9751888513565063,
      "rewards/forensics_sinkhorn_reward": 0.5472847819328308,
      "rewards/forensics_strict_edge_reward": 0.04877546243369579,
      "rewards/forensics_tiou_reward": 0.3799218386411667,
      "step": 439
    },
    {
      "completion_length": 30.9375,
      "epoch": 1.3502304147465438,
      "grad_norm": 5.259155750274658,
      "kl": 0.3505859375,
      "learning_rate": 6.623076923076923e-07,
      "loss": -0.0735,
      "reward": 1.1125788986682892,
      "reward_std": 0.32044754922389984,
      "reward_std_all": 1.0023049712181091,
      "rewards-max/forensics_sinkhorn_reward": 0.9733379185199738,
      "rewards-max/forensics_strict_edge_reward": 0.35115350782871246,
      "rewards-max/forensics_tiou_reward": 0.9612908661365509,
      "rewards/forensics_sinkhorn_reward": 0.6167361438274384,
      "rewards/forensics_strict_edge_reward": 0.02103571593761444,
      "rewards/forensics_tiou_reward": 0.4748070240020752,
      "step": 440
    },
    {
      "completion_length": 32.34375,
      "epoch": 1.3533026113671274,
      "grad_norm": 7.988787651062012,
      "kl": 0.3662109375,
      "learning_rate": 6.615384615384615e-07,
      "loss": -0.1645,
      "reward": 0.8829271793365479,
      "reward_std": 0.2872109115123749,
      "reward_std_all": 1.1033757328987122,
      "rewards-max/forensics_sinkhorn_reward": 0.9618127048015594,
      "rewards-max/forensics_strict_edge_reward": 0.1457061916589737,
      "rewards-max/forensics_tiou_reward": 0.9239453673362732,
      "rewards/forensics_sinkhorn_reward": 0.480448842048645,
      "rewards/forensics_strict_edge_reward": 0.013217894360423088,
      "rewards/forensics_tiou_reward": 0.3892603665590286,
      "step": 441
    },
    {
      "completion_length": 23.0625,
      "epoch": 1.3563748079877112,
      "grad_norm": 5.559391975402832,
      "kl": 0.4150390625,
      "learning_rate": 6.607692307692307e-07,
      "loss": -0.0621,
      "reward": 1.0265946984291077,
      "reward_std": 0.2944379597902298,
      "reward_std_all": 1.0349942445755005,
      "rewards-max/forensics_sinkhorn_reward": 0.9823865592479706,
      "rewards-max/forensics_strict_edge_reward": 0.3274329677224159,
      "rewards-max/forensics_tiou_reward": 0.9812949597835541,
      "rewards/forensics_sinkhorn_reward": 0.5941833555698395,
      "rewards/forensics_strict_edge_reward": 0.029646670445799828,
      "rewards/forensics_tiou_reward": 0.4027646481990814,
      "step": 442
    },
    {
      "completion_length": 30.15625,
      "epoch": 1.359447004608295,
      "grad_norm": 5.02514123916626,
      "kl": 0.36328125,
      "learning_rate": 6.6e-07,
      "loss": -0.0472,
      "reward": 0.9523180723190308,
      "reward_std": 0.2253732681274414,
      "reward_std_all": 1.0592070817947388,
      "rewards-max/forensics_sinkhorn_reward": 0.9733032882213593,
      "rewards-max/forensics_strict_edge_reward": 0.34856952726840973,
      "rewards-max/forensics_tiou_reward": 0.958917111158371,
      "rewards/forensics_sinkhorn_reward": 0.5129000395536423,
      "rewards/forensics_strict_edge_reward": 0.025349842384457588,
      "rewards/forensics_tiou_reward": 0.41406822204589844,
      "step": 443
    },
    {
      "completion_length": 24.71875,
      "epoch": 1.3625192012288787,
      "grad_norm": 4.735873222351074,
      "kl": 0.3447265625,
      "learning_rate": 6.592307692307691e-07,
      "loss": -0.0344,
      "reward": 1.1929932832717896,
      "reward_std": 0.3016159161925316,
      "reward_std_all": 0.9581858515739441,
      "rewards-max/forensics_sinkhorn_reward": 0.9620496332645416,
      "rewards-max/forensics_strict_edge_reward": 0.3885071277618408,
      "rewards-max/forensics_tiou_reward": 0.9366568922996521,
      "rewards/forensics_sinkhorn_reward": 0.6773350536823273,
      "rewards/forensics_strict_edge_reward": 0.03867691848427057,
      "rewards/forensics_tiou_reward": 0.4769813120365143,
      "step": 444
    },
    {
      "completion_length": 25.078125,
      "epoch": 1.3655913978494625,
      "grad_norm": 5.899199485778809,
      "kl": 0.373046875,
      "learning_rate": 6.584615384615384e-07,
      "loss": -0.1004,
      "reward": 1.1387419700622559,
      "reward_std": 0.33371327817440033,
      "reward_std_all": 0.9984843730926514,
      "rewards-max/forensics_sinkhorn_reward": 0.9723558127880096,
      "rewards-max/forensics_strict_edge_reward": 0.6128445863723755,
      "rewards-max/forensics_tiou_reward": 0.9581635594367981,
      "rewards/forensics_sinkhorn_reward": 0.6238389909267426,
      "rewards/forensics_strict_edge_reward": 0.043755076825618744,
      "rewards/forensics_tiou_reward": 0.4711478650569916,
      "step": 445
    },
    {
      "completion_length": 25.296875,
      "epoch": 1.368663594470046,
      "grad_norm": 6.503596782684326,
      "kl": 0.318359375,
      "learning_rate": 6.576923076923077e-07,
      "loss": -0.0876,
      "reward": 1.2704382538795471,
      "reward_std": 0.41127850115299225,
      "reward_std_all": 0.9724371433258057,
      "rewards-max/forensics_sinkhorn_reward": 0.9859519302845001,
      "rewards-max/forensics_strict_edge_reward": 0.7237316966056824,
      "rewards-max/forensics_tiou_reward": 0.9939340353012085,
      "rewards/forensics_sinkhorn_reward": 0.6525042951107025,
      "rewards/forensics_strict_edge_reward": 0.0651160255074501,
      "rewards/forensics_tiou_reward": 0.5528179109096527,
      "step": 446
    },
    {
      "completion_length": 25.140625,
      "epoch": 1.3717357910906298,
      "grad_norm": 5.513314723968506,
      "kl": 0.3330078125,
      "learning_rate": 6.569230769230769e-07,
      "loss": -0.0475,
      "reward": 0.8781727254390717,
      "reward_std": 0.3945115953683853,
      "reward_std_all": 1.1462148427963257,
      "rewards-max/forensics_sinkhorn_reward": 0.9836119711399078,
      "rewards-max/forensics_strict_edge_reward": 0.38277086429297924,
      "rewards-max/forensics_tiou_reward": 0.9845564663410187,
      "rewards/forensics_sinkhorn_reward": 0.5217662751674652,
      "rewards/forensics_strict_edge_reward": 0.04163796751527116,
      "rewards/forensics_tiou_reward": 0.3147684335708618,
      "step": 447
    },
    {
      "completion_length": 22.484375,
      "epoch": 1.3748079877112136,
      "grad_norm": 6.56773042678833,
      "kl": 0.384765625,
      "learning_rate": 6.561538461538462e-07,
      "loss": -0.0455,
      "reward": 0.9818848967552185,
      "reward_std": 0.22701546549797058,
      "reward_std_all": 1.0449936985969543,
      "rewards-max/forensics_sinkhorn_reward": 0.9564832150936127,
      "rewards-max/forensics_strict_edge_reward": 0.379068061709404,
      "rewards-max/forensics_tiou_reward": 0.9222281873226166,
      "rewards/forensics_sinkhorn_reward": 0.6135188341140747,
      "rewards/forensics_strict_edge_reward": 0.01757444581016898,
      "rewards/forensics_tiou_reward": 0.35079168528318405,
      "step": 448
    },
    {
      "completion_length": 23.171875,
      "epoch": 1.3778801843317972,
      "grad_norm": 9.122733116149902,
      "kl": 0.328125,
      "learning_rate": 6.553846153846153e-07,
      "loss": -0.1016,
      "reward": 1.094723880290985,
      "reward_std": 0.3158980682492256,
      "reward_std_all": 1.0234123170375824,
      "rewards-max/forensics_sinkhorn_reward": 0.9865457713603973,
      "rewards-max/forensics_strict_edge_reward": 0.7659632563591003,
      "rewards-max/forensics_tiou_reward": 0.9906265437602997,
      "rewards/forensics_sinkhorn_reward": 0.5766116082668304,
      "rewards/forensics_strict_edge_reward": 0.04905294254422188,
      "rewards/forensics_tiou_reward": 0.4690593183040619,
      "step": 449
    },
    {
      "completion_length": 27.90625,
      "epoch": 1.380952380952381,
      "grad_norm": 7.447605609893799,
      "kl": 0.51953125,
      "learning_rate": 6.546153846153845e-07,
      "loss": -0.1062,
      "reward": 0.9936131536960602,
      "reward_std": 0.26877300441265106,
      "reward_std_all": 1.0442149639129639,
      "rewards-max/forensics_sinkhorn_reward": 0.9691745340824127,
      "rewards-max/forensics_strict_edge_reward": 0.3345223665237427,
      "rewards-max/forensics_tiou_reward": 0.9507956802845001,
      "rewards/forensics_sinkhorn_reward": 0.5665380954742432,
      "rewards/forensics_strict_edge_reward": 0.02575333882123232,
      "rewards/forensics_tiou_reward": 0.4013216942548752,
      "step": 450
    },
    {
      "completion_length": 26.53125,
      "epoch": 1.3840245775729647,
      "grad_norm": 6.209841251373291,
      "kl": 0.3125,
      "learning_rate": 6.538461538461538e-07,
      "loss": -0.1566,
      "reward": 1.080210566520691,
      "reward_std": 0.30726358294487,
      "reward_std_all": 1.013994038105011,
      "rewards-max/forensics_sinkhorn_reward": 0.9709666967391968,
      "rewards-max/forensics_strict_edge_reward": 0.42878545820713043,
      "rewards-max/forensics_tiou_reward": 0.9518179595470428,
      "rewards/forensics_sinkhorn_reward": 0.6224929094314575,
      "rewards/forensics_strict_edge_reward": 0.025508791208267212,
      "rewards/forensics_tiou_reward": 0.4322088658809662,
      "step": 451
    },
    {
      "completion_length": 24.484375,
      "epoch": 1.3870967741935485,
      "grad_norm": 5.501370429992676,
      "kl": 0.3173828125,
      "learning_rate": 6.530769230769231e-07,
      "loss": -0.0311,
      "reward": 0.976230263710022,
      "reward_std": 0.31493186950683594,
      "reward_std_all": 1.0590470433235168,
      "rewards-max/forensics_sinkhorn_reward": 0.9614676833152771,
      "rewards-max/forensics_strict_edge_reward": 0.4012388288974762,
      "rewards-max/forensics_tiou_reward": 0.9314727187156677,
      "rewards/forensics_sinkhorn_reward": 0.5748290419578552,
      "rewards/forensics_strict_edge_reward": 0.030052660033106804,
      "rewards/forensics_tiou_reward": 0.371348574757576,
      "step": 452
    },
    {
      "completion_length": 22.96875,
      "epoch": 1.390168970814132,
      "grad_norm": 5.584518909454346,
      "kl": 0.5146484375,
      "learning_rate": 6.523076923076923e-07,
      "loss": -0.0793,
      "reward": 1.1378116607666016,
      "reward_std": 0.36586837470531464,
      "reward_std_all": 1.0109086632728577,
      "rewards-max/forensics_sinkhorn_reward": 0.974993497133255,
      "rewards-max/forensics_strict_edge_reward": 0.6076286435127258,
      "rewards-max/forensics_tiou_reward": 0.9669677019119263,
      "rewards/forensics_sinkhorn_reward": 0.651206761598587,
      "rewards/forensics_strict_edge_reward": 0.04598606750369072,
      "rewards/forensics_tiou_reward": 0.4406188428401947,
      "step": 453
    },
    {
      "completion_length": 22.609375,
      "epoch": 1.3932411674347158,
      "grad_norm": 5.625608921051025,
      "kl": 0.384765625,
      "learning_rate": 6.515384615384615e-07,
      "loss": -0.0412,
      "reward": 1.2709466218948364,
      "reward_std": 0.4144793301820755,
      "reward_std_all": 0.9564228355884552,
      "rewards-max/forensics_sinkhorn_reward": 0.9700353741645813,
      "rewards-max/forensics_strict_edge_reward": 0.6919131278991699,
      "rewards-max/forensics_tiou_reward": 0.9582618772983551,
      "rewards/forensics_sinkhorn_reward": 0.7065285444259644,
      "rewards/forensics_strict_edge_reward": 0.046344462782144547,
      "rewards/forensics_tiou_reward": 0.5180736780166626,
      "step": 454
    },
    {
      "completion_length": 22.28125,
      "epoch": 1.3963133640552996,
      "grad_norm": 5.329391002655029,
      "kl": 0.318359375,
      "learning_rate": 6.507692307692307e-07,
      "loss": -0.0861,
      "reward": 1.1775770783424377,
      "reward_std": 0.23384513705968857,
      "reward_std_all": 0.94664266705513,
      "rewards-max/forensics_sinkhorn_reward": 0.9788098931312561,
      "rewards-max/forensics_strict_edge_reward": 0.32368214428424835,
      "rewards-max/forensics_tiou_reward": 0.9738913178443909,
      "rewards/forensics_sinkhorn_reward": 0.6590935289859772,
      "rewards/forensics_strict_edge_reward": 0.0351314852014184,
      "rewards/forensics_tiou_reward": 0.4833521097898483,
      "step": 455
    },
    {
      "completion_length": 22.65625,
      "epoch": 1.3993855606758832,
      "grad_norm": 5.439212322235107,
      "kl": 0.4072265625,
      "learning_rate": 6.5e-07,
      "loss": -0.1013,
      "reward": 1.3312261700630188,
      "reward_std": 0.30627796798944473,
      "reward_std_all": 0.8865374326705933,
      "rewards-max/forensics_sinkhorn_reward": 0.9804507791996002,
      "rewards-max/forensics_strict_edge_reward": 0.5948129594326019,
      "rewards-max/forensics_tiou_reward": 0.9686829149723053,
      "rewards/forensics_sinkhorn_reward": 0.7460798025131226,
      "rewards/forensics_strict_edge_reward": 0.03729342296719551,
      "rewards/forensics_tiou_reward": 0.5478528887033463,
      "step": 456
    },
    {
      "completion_length": 24.484375,
      "epoch": 1.402457757296467,
      "grad_norm": 5.495832443237305,
      "kl": 0.5224609375,
      "learning_rate": 6.492307692307692e-07,
      "loss": -0.1146,
      "reward": 0.958755224943161,
      "reward_std": 0.3070104345679283,
      "reward_std_all": 1.076198697090149,
      "rewards-max/forensics_sinkhorn_reward": 0.9620510637760162,
      "rewards-max/forensics_strict_edge_reward": 0.49870672915130854,
      "rewards-max/forensics_tiou_reward": 0.9235678911209106,
      "rewards/forensics_sinkhorn_reward": 0.5642147958278656,
      "rewards/forensics_strict_edge_reward": 0.028599250596016645,
      "rewards/forensics_tiou_reward": 0.36594122648239136,
      "step": 457
    },
    {
      "completion_length": 23.40625,
      "epoch": 1.4055299539170507,
      "grad_norm": 6.506850719451904,
      "kl": 0.4228515625,
      "learning_rate": 6.484615384615383e-07,
      "loss": -0.0814,
      "reward": 0.9651127457618713,
      "reward_std": 0.26829420030117035,
      "reward_std_all": 1.0672272443771362,
      "rewards-max/forensics_sinkhorn_reward": 0.9378946423530579,
      "rewards-max/forensics_strict_edge_reward": 0.37054420899949037,
      "rewards-max/forensics_tiou_reward": 0.8773853480815887,
      "rewards/forensics_sinkhorn_reward": 0.6086642146110535,
      "rewards/forensics_strict_edge_reward": 0.018096664860422607,
      "rewards/forensics_tiou_reward": 0.3383518159389496,
      "step": 458
    },
    {
      "completion_length": 23.265625,
      "epoch": 1.4086021505376345,
      "grad_norm": 8.28860855102539,
      "kl": 0.8154296875,
      "learning_rate": 6.476923076923077e-07,
      "loss": -0.1267,
      "reward": 1.299282431602478,
      "reward_std": 0.4104697108268738,
      "reward_std_all": 0.9540000259876251,
      "rewards-max/forensics_sinkhorn_reward": 0.9867258071899414,
      "rewards-max/forensics_strict_edge_reward": 0.784106969833374,
      "rewards-max/forensics_tiou_reward": 0.9949381649494171,
      "rewards/forensics_sinkhorn_reward": 0.6931912899017334,
      "rewards/forensics_strict_edge_reward": 0.08235275745391846,
      "rewards/forensics_tiou_reward": 0.5237383544445038,
      "step": 459
    },
    {
      "completion_length": 24.09375,
      "epoch": 1.411674347158218,
      "grad_norm": 6.72247314453125,
      "kl": 0.369140625,
      "learning_rate": 6.469230769230769e-07,
      "loss": -0.1172,
      "reward": 1.046253502368927,
      "reward_std": 0.23838762938976288,
      "reward_std_all": 1.01595801115036,
      "rewards-max/forensics_sinkhorn_reward": 0.9758905470371246,
      "rewards-max/forensics_strict_edge_reward": 0.24417344480752945,
      "rewards-max/forensics_tiou_reward": 0.9635263085365295,
      "rewards/forensics_sinkhorn_reward": 0.6099286377429962,
      "rewards/forensics_strict_edge_reward": 0.01917232573032379,
      "rewards/forensics_tiou_reward": 0.417152538895607,
      "step": 460
    },
    {
      "completion_length": 22.203125,
      "epoch": 1.4147465437788018,
      "grad_norm": 5.2249908447265625,
      "kl": 0.3955078125,
      "learning_rate": 6.461538461538462e-07,
      "loss": -0.072,
      "reward": 1.0779124200344086,
      "reward_std": 0.43103501200675964,
      "reward_std_all": 1.0548632144927979,
      "rewards-max/forensics_sinkhorn_reward": 0.9818449914455414,
      "rewards-max/forensics_strict_edge_reward": 0.4180193245410919,
      "rewards-max/forensics_tiou_reward": 0.9805226027965546,
      "rewards/forensics_sinkhorn_reward": 0.6426236629486084,
      "rewards/forensics_strict_edge_reward": 0.03446673508733511,
      "rewards/forensics_tiou_reward": 0.4008219987154007,
      "step": 461
    },
    {
      "completion_length": 23.40625,
      "epoch": 1.4178187403993856,
      "grad_norm": 5.473618030548096,
      "kl": 0.4384765625,
      "learning_rate": 6.453846153846153e-07,
      "loss": -0.1265,
      "reward": 1.2962702512741089,
      "reward_std": 0.3132123202085495,
      "reward_std_all": 0.9019056558609009,
      "rewards-max/forensics_sinkhorn_reward": 0.9616810083389282,
      "rewards-max/forensics_strict_edge_reward": 0.2913331761956215,
      "rewards-max/forensics_tiou_reward": 0.9345157742500305,
      "rewards/forensics_sinkhorn_reward": 0.725782185792923,
      "rewards/forensics_strict_edge_reward": 0.03385998494923115,
      "rewards/forensics_tiou_reward": 0.5366281270980835,
      "step": 462
    },
    {
      "completion_length": 26.5625,
      "epoch": 1.4208909370199692,
      "grad_norm": 5.225039958953857,
      "kl": 0.3828125,
      "learning_rate": 6.446153846153845e-07,
      "loss": -0.1429,
      "reward": 1.0825680196285248,
      "reward_std": 0.24769581854343414,
      "reward_std_all": 0.9986494779586792,
      "rewards-max/forensics_sinkhorn_reward": 0.9837609827518463,
      "rewards-max/forensics_strict_edge_reward": 0.415291503071785,
      "rewards-max/forensics_tiou_reward": 0.9862236678600311,
      "rewards/forensics_sinkhorn_reward": 0.625414103269577,
      "rewards/forensics_strict_edge_reward": 0.025637022219598293,
      "rewards/forensics_tiou_reward": 0.43151693046092987,
      "step": 463
    },
    {
      "completion_length": 26.46875,
      "epoch": 1.423963133640553,
      "grad_norm": 7.269131183624268,
      "kl": 0.4736328125,
      "learning_rate": 6.438461538461538e-07,
      "loss": -0.1032,
      "reward": 1.079497367143631,
      "reward_std": 0.34274428337812424,
      "reward_std_all": 1.0363797545433044,
      "rewards-max/forensics_sinkhorn_reward": 0.9732749164104462,
      "rewards-max/forensics_strict_edge_reward": 0.5179875791072845,
      "rewards-max/forensics_tiou_reward": 0.9593774974346161,
      "rewards/forensics_sinkhorn_reward": 0.6111682653427124,
      "rewards/forensics_strict_edge_reward": 0.03362613171339035,
      "rewards/forensics_tiou_reward": 0.4347030073404312,
      "step": 464
    },
    {
      "completion_length": 25.703125,
      "epoch": 1.4270353302611367,
      "grad_norm": 6.67659330368042,
      "kl": 0.634765625,
      "learning_rate": 6.43076923076923e-07,
      "loss": -0.0439,
      "reward": 1.2719652652740479,
      "reward_std": 0.32261285185813904,
      "reward_std_all": 0.930074006319046,
      "rewards-max/forensics_sinkhorn_reward": 0.978581577539444,
      "rewards-max/forensics_strict_edge_reward": 0.4235446900129318,
      "rewards-max/forensics_tiou_reward": 0.9719615578651428,
      "rewards/forensics_sinkhorn_reward": 0.7014157772064209,
      "rewards/forensics_strict_edge_reward": 0.03651039861142635,
      "rewards/forensics_tiou_reward": 0.5340391248464584,
      "step": 465
    },
    {
      "completion_length": 27.34375,
      "epoch": 1.4301075268817205,
      "grad_norm": 5.33545446395874,
      "kl": 0.3203125,
      "learning_rate": 6.423076923076924e-07,
      "loss": -0.0961,
      "reward": 1.025099366903305,
      "reward_std": 0.250259418040514,
      "reward_std_all": 1.030052363872528,
      "rewards-max/forensics_sinkhorn_reward": 0.9548197984695435,
      "rewards-max/forensics_strict_edge_reward": 0.23941841907799244,
      "rewards-max/forensics_tiou_reward": 0.8991336822509766,
      "rewards/forensics_sinkhorn_reward": 0.5867636799812317,
      "rewards/forensics_strict_edge_reward": 0.022384147392585874,
      "rewards/forensics_tiou_reward": 0.41595160961151123,
      "step": 466
    },
    {
      "completion_length": 25.921875,
      "epoch": 1.4331797235023043,
      "grad_norm": 6.780632972717285,
      "kl": 0.4375,
      "learning_rate": 6.415384615384615e-07,
      "loss": -0.0633,
      "reward": 1.049637496471405,
      "reward_std": 0.3570665717124939,
      "reward_std_all": 1.0451892614364624,
      "rewards-max/forensics_sinkhorn_reward": 0.9582929313182831,
      "rewards-max/forensics_strict_edge_reward": 0.2359465277986601,
      "rewards-max/forensics_tiou_reward": 0.8898281455039978,
      "rewards/forensics_sinkhorn_reward": 0.6259586811065674,
      "rewards/forensics_strict_edge_reward": 0.02827905553567689,
      "rewards/forensics_tiou_reward": 0.3953997492790222,
      "step": 467
    },
    {
      "completion_length": 23.78125,
      "epoch": 1.4362519201228878,
      "grad_norm": 7.355682849884033,
      "kl": 0.3173828125,
      "learning_rate": 6.407692307692307e-07,
      "loss": -0.1705,
      "reward": 0.9325850009918213,
      "reward_std": 0.21545576304197311,
      "reward_std_all": 1.0828899145126343,
      "rewards-max/forensics_sinkhorn_reward": 0.9818919003009796,
      "rewards-max/forensics_strict_edge_reward": 0.7868733704090118,
      "rewards-max/forensics_tiou_reward": 0.9808088839054108,
      "rewards/forensics_sinkhorn_reward": 0.5722388923168182,
      "rewards/forensics_strict_edge_reward": 0.04417544789612293,
      "rewards/forensics_tiou_reward": 0.31617072224617004,
      "step": 468
    },
    {
      "completion_length": 28.53125,
      "epoch": 1.4393241167434716,
      "grad_norm": 4.105649471282959,
      "kl": 0.3447265625,
      "learning_rate": 6.4e-07,
      "loss": -0.1152,
      "reward": 0.9544437825679779,
      "reward_std": 0.22769670188426971,
      "reward_std_all": 1.0510764718055725,
      "rewards-max/forensics_sinkhorn_reward": 0.9652887582778931,
      "rewards-max/forensics_strict_edge_reward": 0.32617029547691345,
      "rewards-max/forensics_tiou_reward": 0.9401328563690186,
      "rewards/forensics_sinkhorn_reward": 0.5631473511457443,
      "rewards/forensics_strict_edge_reward": 0.018673240207135677,
      "rewards/forensics_tiou_reward": 0.37262319028377533,
      "step": 469
    },
    {
      "completion_length": 23.515625,
      "epoch": 1.4423963133640554,
      "grad_norm": 5.375589847564697,
      "kl": 0.3349609375,
      "learning_rate": 6.392307692307692e-07,
      "loss": -0.0785,
      "reward": 1.2279800772666931,
      "reward_std": 0.343151330947876,
      "reward_std_all": 0.9549994766712189,
      "rewards-max/forensics_sinkhorn_reward": 0.9789892137050629,
      "rewards-max/forensics_strict_edge_reward": 0.6149742305278778,
      "rewards-max/forensics_tiou_reward": 0.9744687378406525,
      "rewards/forensics_sinkhorn_reward": 0.6619686782360077,
      "rewards/forensics_strict_edge_reward": 0.05770052224397659,
      "rewards/forensics_tiou_reward": 0.508310854434967,
      "step": 470
    },
    {
      "completion_length": 24.828125,
      "epoch": 1.445468509984639,
      "grad_norm": 5.484652996063232,
      "kl": 0.37109375,
      "learning_rate": 6.384615384615383e-07,
      "loss": -0.0423,
      "reward": 1.1563695073127747,
      "reward_std": 0.22980600595474243,
      "reward_std_all": 0.954681932926178,
      "rewards-max/forensics_sinkhorn_reward": 0.9535127580165863,
      "rewards-max/forensics_strict_edge_reward": 0.0944858961738646,
      "rewards-max/forensics_tiou_reward": 0.8997537195682526,
      "rewards/forensics_sinkhorn_reward": 0.6774930953979492,
      "rewards/forensics_strict_edge_reward": 0.008177493000403047,
      "rewards/forensics_tiou_reward": 0.47069890797138214,
      "step": 471
    },
    {
      "completion_length": 25.75,
      "epoch": 1.4485407066052227,
      "grad_norm": 5.108868598937988,
      "kl": 0.505859375,
      "learning_rate": 6.376923076923077e-07,
      "loss": -0.0985,
      "reward": 1.2702574729919434,
      "reward_std": 0.42909303307533264,
      "reward_std_all": 0.9785707890987396,
      "rewards-max/forensics_sinkhorn_reward": 0.9730237424373627,
      "rewards-max/forensics_strict_edge_reward": 0.4895991235971451,
      "rewards-max/forensics_tiou_reward": 0.9543229937553406,
      "rewards/forensics_sinkhorn_reward": 0.6926893591880798,
      "rewards/forensics_strict_edge_reward": 0.05175567604601383,
      "rewards/forensics_tiou_reward": 0.5258124470710754,
      "step": 472
    },
    {
      "completion_length": 25.78125,
      "epoch": 1.4516129032258065,
      "grad_norm": 6.287455081939697,
      "kl": 0.31494140625,
      "learning_rate": 6.369230769230769e-07,
      "loss": -0.1539,
      "reward": 1.184185951948166,
      "reward_std": 0.2993013560771942,
      "reward_std_all": 0.9770082831382751,
      "rewards-max/forensics_sinkhorn_reward": 0.96306973695755,
      "rewards-max/forensics_strict_edge_reward": 0.36990174278616905,
      "rewards-max/forensics_tiou_reward": 0.9341029822826385,
      "rewards/forensics_sinkhorn_reward": 0.6644911766052246,
      "rewards/forensics_strict_edge_reward": 0.060592255322262645,
      "rewards/forensics_tiou_reward": 0.4591025114059448,
      "step": 473
    },
    {
      "completion_length": 25.296875,
      "epoch": 1.4546850998463903,
      "grad_norm": 6.128228187561035,
      "kl": 0.3125,
      "learning_rate": 6.361538461538462e-07,
      "loss": -0.126,
      "reward": 1.1611079573631287,
      "reward_std": 0.3629666566848755,
      "reward_std_all": 0.9927085340023041,
      "rewards-max/forensics_sinkhorn_reward": 0.977124959230423,
      "rewards-max/forensics_strict_edge_reward": 0.525653138756752,
      "rewards-max/forensics_tiou_reward": 0.9694746136665344,
      "rewards/forensics_sinkhorn_reward": 0.631231963634491,
      "rewards/forensics_strict_edge_reward": 0.05073983781039715,
      "rewards/forensics_tiou_reward": 0.47913625836372375,
      "step": 474
    },
    {
      "completion_length": 27.46875,
      "epoch": 1.4577572964669738,
      "grad_norm": 4.846154689788818,
      "kl": 0.291015625,
      "learning_rate": 6.353846153846154e-07,
      "loss": -0.0905,
      "reward": 1.0137838125228882,
      "reward_std": 0.34809519350528717,
      "reward_std_all": 1.0593964457511902,
      "rewards-max/forensics_sinkhorn_reward": 0.9827958345413208,
      "rewards-max/forensics_strict_edge_reward": 0.5362234264612198,
      "rewards-max/forensics_tiou_reward": 0.9839296042919159,
      "rewards/forensics_sinkhorn_reward": 0.590374082326889,
      "rewards/forensics_strict_edge_reward": 0.03681703470647335,
      "rewards/forensics_tiou_reward": 0.38659264147281647,
      "step": 475
    },
    {
      "completion_length": 26.390625,
      "epoch": 1.4608294930875576,
      "grad_norm": 5.283424377441406,
      "kl": 0.3125,
      "learning_rate": 6.346153846153845e-07,
      "loss": -0.092,
      "reward": 1.0650560855865479,
      "reward_std": 0.30669455230236053,
      "reward_std_all": 1.0223258435726166,
      "rewards-max/forensics_sinkhorn_reward": 0.9840061962604523,
      "rewards-max/forensics_strict_edge_reward": 0.5168293714523315,
      "rewards-max/forensics_tiou_reward": 0.9817801415920258,
      "rewards/forensics_sinkhorn_reward": 0.6084644198417664,
      "rewards/forensics_strict_edge_reward": 0.026204699650406837,
      "rewards/forensics_tiou_reward": 0.430387020111084,
      "step": 476
    },
    {
      "completion_length": 24.734375,
      "epoch": 1.4639016897081414,
      "grad_norm": 5.948146820068359,
      "kl": 0.326171875,
      "learning_rate": 6.338461538461538e-07,
      "loss": -0.0946,
      "reward": 1.1477175951004028,
      "reward_std": 0.21767458319664001,
      "reward_std_all": 0.9743021428585052,
      "rewards-max/forensics_sinkhorn_reward": 0.9790915846824646,
      "rewards-max/forensics_strict_edge_reward": 0.6253509670495987,
      "rewards-max/forensics_tiou_reward": 0.9722709655761719,
      "rewards/forensics_sinkhorn_reward": 0.6415328085422516,
      "rewards/forensics_strict_edge_reward": 0.039586243219673634,
      "rewards/forensics_tiou_reward": 0.4665985256433487,
      "step": 477
    },
    {
      "completion_length": 24.828125,
      "epoch": 1.466973886328725,
      "grad_norm": 5.814935684204102,
      "kl": 0.4521484375,
      "learning_rate": 6.33076923076923e-07,
      "loss": -0.0718,
      "reward": 1.1933993101119995,
      "reward_std": 0.496533066034317,
      "reward_std_all": 1.0185266435146332,
      "rewards-max/forensics_sinkhorn_reward": 0.9668322503566742,
      "rewards-max/forensics_strict_edge_reward": 0.5820994228124619,
      "rewards-max/forensics_tiou_reward": 0.9510542452335358,
      "rewards/forensics_sinkhorn_reward": 0.6527540981769562,
      "rewards/forensics_strict_edge_reward": 0.059467388316988945,
      "rewards/forensics_tiou_reward": 0.4811778664588928,
      "step": 478
    },
    {
      "completion_length": 25.25,
      "epoch": 1.4700460829493087,
      "grad_norm": 5.315987586975098,
      "kl": 0.3916015625,
      "learning_rate": 6.323076923076924e-07,
      "loss": -0.1171,
      "reward": 1.0799134373664856,
      "reward_std": 0.28475144505500793,
      "reward_std_all": 1.0169469118118286,
      "rewards-max/forensics_sinkhorn_reward": 0.9729135930538177,
      "rewards-max/forensics_strict_edge_reward": 0.4499741829931736,
      "rewards-max/forensics_tiou_reward": 0.9533911943435669,
      "rewards/forensics_sinkhorn_reward": 0.643682211637497,
      "rewards/forensics_strict_edge_reward": 0.024123020470142365,
      "rewards/forensics_tiou_reward": 0.4121082276105881,
      "step": 479
    },
    {
      "completion_length": 23.953125,
      "epoch": 1.4731182795698925,
      "grad_norm": 6.811128616333008,
      "kl": 0.3828125,
      "learning_rate": 6.315384615384615e-07,
      "loss": -0.0855,
      "reward": 0.9753484129905701,
      "reward_std": 0.35525673627853394,
      "reward_std_all": 1.0872313380241394,
      "rewards-max/forensics_sinkhorn_reward": 0.9783430695533752,
      "rewards-max/forensics_strict_edge_reward": 0.5536971986293793,
      "rewards-max/forensics_tiou_reward": 0.9706065356731415,
      "rewards/forensics_sinkhorn_reward": 0.5843017995357513,
      "rewards/forensics_strict_edge_reward": 0.03265446051955223,
      "rewards/forensics_tiou_reward": 0.358392134308815,
      "step": 480
    },
    {
      "completion_length": 28.3125,
      "epoch": 1.4761904761904763,
      "grad_norm": 6.432690143585205,
      "kl": 0.3916015625,
      "learning_rate": 6.307692307692307e-07,
      "loss": -0.1406,
      "reward": 1.1461405754089355,
      "reward_std": 0.3652620315551758,
      "reward_std_all": 1.0053077936172485,
      "rewards-max/forensics_sinkhorn_reward": 0.9620943963527679,
      "rewards-max/forensics_strict_edge_reward": 0.3254972919821739,
      "rewards-max/forensics_tiou_reward": 0.9362193942070007,
      "rewards/forensics_sinkhorn_reward": 0.5998857021331787,
      "rewards/forensics_strict_edge_reward": 0.03198607359081507,
      "rewards/forensics_tiou_reward": 0.5142687857151031,
      "step": 481
    },
    {
      "completion_length": 24.828125,
      "epoch": 1.4792626728110598,
      "grad_norm": 7.469559192657471,
      "kl": 0.3232421875,
      "learning_rate": 6.3e-07,
      "loss": -0.0524,
      "reward": 0.9100137948989868,
      "reward_std": 0.25835883617401123,
      "reward_std_all": 1.076859474182129,
      "rewards-max/forensics_sinkhorn_reward": 0.8858307301998138,
      "rewards-max/forensics_strict_edge_reward": 0.09990896750241518,
      "rewards-max/forensics_tiou_reward": 0.7568050026893616,
      "rewards/forensics_sinkhorn_reward": 0.5501250624656677,
      "rewards/forensics_strict_edge_reward": 0.007447502110153437,
      "rewards/forensics_tiou_reward": 0.35244128108024597,
      "step": 482
    },
    {
      "completion_length": 25.984375,
      "epoch": 1.4823348694316436,
      "grad_norm": 6.02021598815918,
      "kl": 0.505859375,
      "learning_rate": 6.292307692307692e-07,
      "loss": -0.0849,
      "reward": 1.0574873089790344,
      "reward_std": 0.48172619938850403,
      "reward_std_all": 1.0830698609352112,
      "rewards-max/forensics_sinkhorn_reward": 0.9770629107952118,
      "rewards-max/forensics_strict_edge_reward": 0.48898807168006897,
      "rewards-max/forensics_tiou_reward": 0.9612907469272614,
      "rewards/forensics_sinkhorn_reward": 0.6097322404384613,
      "rewards/forensics_strict_edge_reward": 0.0325787803158164,
      "rewards/forensics_tiou_reward": 0.4151763319969177,
      "step": 483
    },
    {
      "completion_length": 24.953125,
      "epoch": 1.4854070660522274,
      "grad_norm": 6.84666109085083,
      "kl": 0.294921875,
      "learning_rate": 6.284615384615385e-07,
      "loss": -0.0443,
      "reward": 1.152782142162323,
      "reward_std": 0.34542280435562134,
      "reward_std_all": 0.9973948001861572,
      "rewards-max/forensics_sinkhorn_reward": 0.9826106429100037,
      "rewards-max/forensics_strict_edge_reward": 0.490813285112381,
      "rewards-max/forensics_tiou_reward": 0.9813283085823059,
      "rewards/forensics_sinkhorn_reward": 0.6713097989559174,
      "rewards/forensics_strict_edge_reward": 0.036081256344914436,
      "rewards/forensics_tiou_reward": 0.445390984416008,
      "step": 484
    },
    {
      "completion_length": 27.171875,
      "epoch": 1.488479262672811,
      "grad_norm": 5.997928619384766,
      "kl": 0.4296875,
      "learning_rate": 6.276923076923076e-07,
      "loss": -0.107,
      "reward": 1.2146395444869995,
      "reward_std": 0.45523785054683685,
      "reward_std_all": 0.9975908696651459,
      "rewards-max/forensics_sinkhorn_reward": 0.9808964133262634,
      "rewards-max/forensics_strict_edge_reward": 0.7958523333072662,
      "rewards-max/forensics_tiou_reward": 0.9806285798549652,
      "rewards/forensics_sinkhorn_reward": 0.6875085234642029,
      "rewards/forensics_strict_edge_reward": 0.06160323694348335,
      "rewards/forensics_tiou_reward": 0.46552774310112,
      "step": 485
    },
    {
      "completion_length": 26.328125,
      "epoch": 1.4915514592933947,
      "grad_norm": 7.732631206512451,
      "kl": 0.6328125,
      "learning_rate": 6.269230769230769e-07,
      "loss": -0.0247,
      "reward": 1.0146677494049072,
      "reward_std": 0.24076636880636215,
      "reward_std_all": 1.0300092101097107,
      "rewards-max/forensics_sinkhorn_reward": 0.899238258600235,
      "rewards-max/forensics_strict_edge_reward": 0.22167667374014854,
      "rewards-max/forensics_tiou_reward": 0.8216981589794159,
      "rewards/forensics_sinkhorn_reward": 0.6035981178283691,
      "rewards/forensics_strict_edge_reward": 0.017434266570489854,
      "rewards/forensics_tiou_reward": 0.3936353772878647,
      "step": 486
    },
    {
      "completion_length": 26.3125,
      "epoch": 1.4946236559139785,
      "grad_norm": 6.076907634735107,
      "kl": 0.373046875,
      "learning_rate": 6.261538461538462e-07,
      "loss": -0.1027,
      "reward": 1.1857054233551025,
      "reward_std": 0.32282260060310364,
      "reward_std_all": 0.9618095457553864,
      "rewards-max/forensics_sinkhorn_reward": 0.9728951752185822,
      "rewards-max/forensics_strict_edge_reward": 0.4759427011013031,
      "rewards-max/forensics_tiou_reward": 0.9614493250846863,
      "rewards/forensics_sinkhorn_reward": 0.6886282563209534,
      "rewards/forensics_strict_edge_reward": 0.04019095841795206,
      "rewards/forensics_tiou_reward": 0.45688608288764954,
      "step": 487
    },
    {
      "completion_length": 25.046875,
      "epoch": 1.4976958525345623,
      "grad_norm": 6.889554023742676,
      "kl": 0.35546875,
      "learning_rate": 6.253846153846154e-07,
      "loss": -0.0792,
      "reward": 1.1249330043792725,
      "reward_std": 0.3126942068338394,
      "reward_std_all": 1.002293884754181,
      "rewards-max/forensics_sinkhorn_reward": 0.956346720457077,
      "rewards-max/forensics_strict_edge_reward": 0.3243762031197548,
      "rewards-max/forensics_tiou_reward": 0.9226622879505157,
      "rewards/forensics_sinkhorn_reward": 0.6432221531867981,
      "rewards/forensics_strict_edge_reward": 0.027473663445562124,
      "rewards/forensics_tiou_reward": 0.45423723757267,
      "step": 488
    },
    {
      "completion_length": 26.25,
      "epoch": 1.500768049155146,
      "grad_norm": 8.979106903076172,
      "kl": 0.4697265625,
      "learning_rate": 6.246153846153845e-07,
      "loss": -0.1299,
      "reward": 1.1426504850387573,
      "reward_std": 0.3270829766988754,
      "reward_std_all": 0.9928881525993347,
      "rewards-max/forensics_sinkhorn_reward": 0.9683702290058136,
      "rewards-max/forensics_strict_edge_reward": 0.5294047147035599,
      "rewards-max/forensics_tiou_reward": 0.9509704411029816,
      "rewards/forensics_sinkhorn_reward": 0.6451945006847382,
      "rewards/forensics_strict_edge_reward": 0.05096811428666115,
      "rewards/forensics_tiou_reward": 0.44648781418800354,
      "step": 489
    },
    {
      "completion_length": 25.328125,
      "epoch": 1.5038402457757296,
      "grad_norm": 7.040536403656006,
      "kl": 0.3955078125,
      "learning_rate": 6.238461538461538e-07,
      "loss": -0.1056,
      "reward": 1.267088770866394,
      "reward_std": 0.21037514507770538,
      "reward_std_all": 0.894784152507782,
      "rewards-max/forensics_sinkhorn_reward": 0.9438402950763702,
      "rewards-max/forensics_strict_edge_reward": 0.20132214110344648,
      "rewards-max/forensics_tiou_reward": 0.8861269056797028,
      "rewards/forensics_sinkhorn_reward": 0.7157434821128845,
      "rewards/forensics_strict_edge_reward": 0.02124796388670802,
      "rewards/forensics_tiou_reward": 0.5300973802804947,
      "step": 490
    },
    {
      "completion_length": 21.984375,
      "epoch": 1.5069124423963134,
      "grad_norm": 6.527215003967285,
      "kl": 0.390625,
      "learning_rate": 6.23076923076923e-07,
      "loss": -0.0716,
      "reward": 1.2474800944328308,
      "reward_std": 0.3014747053384781,
      "reward_std_all": 0.9478430449962616,
      "rewards-max/forensics_sinkhorn_reward": 0.9825448393821716,
      "rewards-max/forensics_strict_edge_reward": 0.7165313363075256,
      "rewards-max/forensics_tiou_reward": 0.9841940402984619,
      "rewards/forensics_sinkhorn_reward": 0.6830684542655945,
      "rewards/forensics_strict_edge_reward": 0.06668735481798649,
      "rewards/forensics_tiou_reward": 0.497724324464798,
      "step": 491
    },
    {
      "completion_length": 23.296875,
      "epoch": 1.509984639016897,
      "grad_norm": 6.477224349975586,
      "kl": 0.419921875,
      "learning_rate": 6.223076923076924e-07,
      "loss": -0.0995,
      "reward": 1.2466524839401245,
      "reward_std": 0.3222253620624542,
      "reward_std_all": 0.9413602948188782,
      "rewards-max/forensics_sinkhorn_reward": 0.9828637838363647,
      "rewards-max/forensics_strict_edge_reward": 0.6479053497314453,
      "rewards-max/forensics_tiou_reward": 0.9817169904708862,
      "rewards/forensics_sinkhorn_reward": 0.7029463648796082,
      "rewards/forensics_strict_edge_reward": 0.04931413196027279,
      "rewards/forensics_tiou_reward": 0.49439190328121185,
      "step": 492
    },
    {
      "completion_length": 25.875,
      "epoch": 1.5130568356374807,
      "grad_norm": 6.5051374435424805,
      "kl": 0.3798828125,
      "learning_rate": 6.215384615384615e-07,
      "loss": -0.1521,
      "reward": 1.4134976863861084,
      "reward_std": 0.3909781575202942,
      "reward_std_all": 0.875466376543045,
      "rewards-max/forensics_sinkhorn_reward": 0.9734943807125092,
      "rewards-max/forensics_strict_edge_reward": 0.3404887020587921,
      "rewards-max/forensics_tiou_reward": 0.9557886719703674,
      "rewards/forensics_sinkhorn_reward": 0.7690230011940002,
      "rewards/forensics_strict_edge_reward": 0.04032081738114357,
      "rewards/forensics_tiou_reward": 0.6041539311408997,
      "step": 493
    },
    {
      "completion_length": 24.171875,
      "epoch": 1.5161290322580645,
      "grad_norm": 6.1628217697143555,
      "kl": 0.3388671875,
      "learning_rate": 6.207692307692307e-07,
      "loss": -0.1478,
      "reward": 1.4097919464111328,
      "reward_std": 0.3970678150653839,
      "reward_std_all": 0.8797923028469086,
      "rewards-max/forensics_sinkhorn_reward": 0.9792630672454834,
      "rewards-max/forensics_strict_edge_reward": 0.5143956542015076,
      "rewards-max/forensics_tiou_reward": 0.9676483273506165,
      "rewards/forensics_sinkhorn_reward": 0.7798339426517487,
      "rewards/forensics_strict_edge_reward": 0.04936102032661438,
      "rewards/forensics_tiou_reward": 0.5805968940258026,
      "step": 494
    },
    {
      "completion_length": 24.484375,
      "epoch": 1.5192012288786483,
      "grad_norm": 9.016290664672852,
      "kl": 0.5751953125,
      "learning_rate": 6.2e-07,
      "loss": -0.1449,
      "reward": 1.358696162700653,
      "reward_std": 0.401115745306015,
      "reward_std_all": 0.9045718610286713,
      "rewards-max/forensics_sinkhorn_reward": 0.976790577173233,
      "rewards-max/forensics_strict_edge_reward": 0.43953123688697815,
      "rewards-max/forensics_tiou_reward": 0.9658930599689484,
      "rewards/forensics_sinkhorn_reward": 0.7267106771469116,
      "rewards/forensics_strict_edge_reward": 0.057266319170594215,
      "rewards/forensics_tiou_reward": 0.5747191309928894,
      "step": 495
    },
    {
      "completion_length": 22.140625,
      "epoch": 1.522273425499232,
      "grad_norm": 7.110507011413574,
      "kl": 0.4013671875,
      "learning_rate": 6.192307692307692e-07,
      "loss": -0.1673,
      "reward": 1.2924635410308838,
      "reward_std": 0.3658412992954254,
      "reward_std_all": 0.9423758685588837,
      "rewards-max/forensics_sinkhorn_reward": 0.9778114855289459,
      "rewards-max/forensics_strict_edge_reward": 0.6169246137142181,
      "rewards-max/forensics_tiou_reward": 0.9704156816005707,
      "rewards/forensics_sinkhorn_reward": 0.7341441214084625,
      "rewards/forensics_strict_edge_reward": 0.055459748953580856,
      "rewards/forensics_tiou_reward": 0.5028596818447113,
      "step": 496
    },
    {
      "completion_length": 28.390625,
      "epoch": 1.5253456221198156,
      "grad_norm": 6.758914947509766,
      "kl": 0.4150390625,
      "learning_rate": 6.184615384615385e-07,
      "loss": -0.0907,
      "reward": 0.9745307564735413,
      "reward_std": 0.30877785384655,
      "reward_std_all": 1.059689223766327,
      "rewards-max/forensics_sinkhorn_reward": 0.9475480914115906,
      "rewards-max/forensics_strict_edge_reward": 0.3341735601425171,
      "rewards-max/forensics_tiou_reward": 0.893887460231781,
      "rewards/forensics_sinkhorn_reward": 0.5745554864406586,
      "rewards/forensics_strict_edge_reward": 0.02742040902376175,
      "rewards/forensics_tiou_reward": 0.37255486845970154,
      "step": 497
    },
    {
      "completion_length": 24.796875,
      "epoch": 1.5284178187403994,
      "grad_norm": 6.860656261444092,
      "kl": 0.3662109375,
      "learning_rate": 6.176923076923076e-07,
      "loss": -0.139,
      "reward": 1.3065625429153442,
      "reward_std": 0.27542610466480255,
      "reward_std_all": 0.8982671201229095,
      "rewards-max/forensics_sinkhorn_reward": 0.9766572117805481,
      "rewards-max/forensics_strict_edge_reward": 0.5226984620094299,
      "rewards-max/forensics_tiou_reward": 0.9669657945632935,
      "rewards/forensics_sinkhorn_reward": 0.7012142539024353,
      "rewards/forensics_strict_edge_reward": 0.06272505223751068,
      "rewards/forensics_tiou_reward": 0.5426231324672699,
      "step": 498
    },
    {
      "completion_length": 26.09375,
      "epoch": 1.531490015360983,
      "grad_norm": 7.869510650634766,
      "kl": 0.3642578125,
      "learning_rate": 6.169230769230768e-07,
      "loss": -0.1612,
      "reward": 1.009513795375824,
      "reward_std": 0.2995031923055649,
      "reward_std_all": 1.0556854605674744,
      "rewards-max/forensics_sinkhorn_reward": 0.9216452240943909,
      "rewards-max/forensics_strict_edge_reward": 0.24143456472302205,
      "rewards-max/forensics_tiou_reward": 0.8078919351100922,
      "rewards/forensics_sinkhorn_reward": 0.6184262335300446,
      "rewards/forensics_strict_edge_reward": 0.024765557485238787,
      "rewards/forensics_tiou_reward": 0.36632193624973297,
      "step": 499
    },
    {
      "completion_length": 28.265625,
      "epoch": 1.5345622119815667,
      "grad_norm": 5.7954535484313965,
      "kl": 0.4111328125,
      "learning_rate": 6.161538461538462e-07,
      "loss": -0.0938,
      "reward": 1.0096421837806702,
      "reward_std": 0.32820361852645874,
      "reward_std_all": 1.0641085803508759,
      "rewards-max/forensics_sinkhorn_reward": 0.9624958336353302,
      "rewards-max/forensics_strict_edge_reward": 0.36739736050367355,
      "rewards-max/forensics_tiou_reward": 0.9345546066761017,
      "rewards/forensics_sinkhorn_reward": 0.5813853442668915,
      "rewards/forensics_strict_edge_reward": 0.0306572075933218,
      "rewards/forensics_tiou_reward": 0.39759959280490875,
      "step": 500
    },
    {
      "completion_length": 26.9375,
      "epoch": 1.5376344086021505,
      "grad_norm": 5.843976020812988,
      "kl": 0.3896484375,
      "learning_rate": 6.153846153846154e-07,
      "loss": -0.094,
      "reward": 0.8823262453079224,
      "reward_std": 0.39733098447322845,
      "reward_std_all": 1.1340919733047485,
      "rewards-max/forensics_sinkhorn_reward": 0.9575692713260651,
      "rewards-max/forensics_strict_edge_reward": 0.2647802084684372,
      "rewards-max/forensics_tiou_reward": 0.9173105657100677,
      "rewards/forensics_sinkhorn_reward": 0.5104219019412994,
      "rewards/forensics_strict_edge_reward": 0.01669260673224926,
      "rewards/forensics_tiou_reward": 0.3552118092775345,
      "step": 501
    },
    {
      "completion_length": 27.3125,
      "epoch": 1.5407066052227343,
      "grad_norm": 6.650833606719971,
      "kl": 0.4326171875,
      "learning_rate": 6.146153846153845e-07,
      "loss": -0.0666,
      "reward": 1.1093939542770386,
      "reward_std": 0.44314536452293396,
      "reward_std_all": 1.0404909253120422,
      "rewards-max/forensics_sinkhorn_reward": 0.9716038405895233,
      "rewards-max/forensics_strict_edge_reward": 0.46283967792987823,
      "rewards-max/forensics_tiou_reward": 0.9557536840438843,
      "rewards/forensics_sinkhorn_reward": 0.6166406571865082,
      "rewards/forensics_strict_edge_reward": 0.03298019431531429,
      "rewards/forensics_tiou_reward": 0.45977313816547394,
      "step": 502
    },
    {
      "completion_length": 28.34375,
      "epoch": 1.543778801843318,
      "grad_norm": 6.377840995788574,
      "kl": 0.5126953125,
      "learning_rate": 6.138461538461538e-07,
      "loss": -0.0706,
      "reward": 1.3630759716033936,
      "reward_std": 0.2519710436463356,
      "reward_std_all": 0.862174928188324,
      "rewards-max/forensics_sinkhorn_reward": 0.9698775708675385,
      "rewards-max/forensics_strict_edge_reward": 0.4973767101764679,
      "rewards-max/forensics_tiou_reward": 0.9537148177623749,
      "rewards/forensics_sinkhorn_reward": 0.7277058959007263,
      "rewards/forensics_strict_edge_reward": 0.054557790979743004,
      "rewards/forensics_tiou_reward": 0.5808123350143433,
      "step": 503
    },
    {
      "completion_length": 27.140625,
      "epoch": 1.5468509984639018,
      "grad_norm": 6.005796432495117,
      "kl": 0.44140625,
      "learning_rate": 6.13076923076923e-07,
      "loss": -0.1197,
      "reward": 1.1870418787002563,
      "reward_std": 0.37825939804315567,
      "reward_std_all": 0.9792210459709167,
      "rewards-max/forensics_sinkhorn_reward": 0.9807290136814117,
      "rewards-max/forensics_strict_edge_reward": 0.31268319487571716,
      "rewards-max/forensics_tiou_reward": 0.9786524176597595,
      "rewards/forensics_sinkhorn_reward": 0.6654663681983948,
      "rewards/forensics_strict_edge_reward": 0.02509539481252432,
      "rewards/forensics_tiou_reward": 0.4964800328016281,
      "step": 504
    },
    {
      "completion_length": 22.96875,
      "epoch": 1.5499231950844854,
      "grad_norm": 5.372918605804443,
      "kl": 0.4228515625,
      "learning_rate": 6.123076923076923e-07,
      "loss": -0.1098,
      "reward": 1.2105580568313599,
      "reward_std": 0.29169078171253204,
      "reward_std_all": 0.953592836856842,
      "rewards-max/forensics_sinkhorn_reward": 0.9759199321269989,
      "rewards-max/forensics_strict_edge_reward": 0.49116502702236176,
      "rewards-max/forensics_tiou_reward": 0.9629727005958557,
      "rewards/forensics_sinkhorn_reward": 0.6942136883735657,
      "rewards/forensics_strict_edge_reward": 0.047981103882193565,
      "rewards/forensics_tiou_reward": 0.46836312115192413,
      "step": 505
    },
    {
      "completion_length": 24.5625,
      "epoch": 1.5529953917050692,
      "grad_norm": 7.157955646514893,
      "kl": 0.2978515625,
      "learning_rate": 6.115384615384616e-07,
      "loss": -0.0388,
      "reward": 1.2020267248153687,
      "reward_std": 0.18540330976247787,
      "reward_std_all": 0.92702716588974,
      "rewards-max/forensics_sinkhorn_reward": 0.9572993814945221,
      "rewards-max/forensics_strict_edge_reward": 0.23789025470614433,
      "rewards-max/forensics_tiou_reward": 0.9265175461769104,
      "rewards/forensics_sinkhorn_reward": 0.6663588881492615,
      "rewards/forensics_strict_edge_reward": 0.019034191500395536,
      "rewards/forensics_tiou_reward": 0.5166335850954056,
      "step": 506
    },
    {
      "completion_length": 27.0625,
      "epoch": 1.5560675883256527,
      "grad_norm": 6.480349540710449,
      "kl": 0.41796875,
      "learning_rate": 6.107692307692307e-07,
      "loss": -0.1254,
      "reward": 1.224743902683258,
      "reward_std": 0.4229758083820343,
      "reward_std_all": 0.9736044406890869,
      "rewards-max/forensics_sinkhorn_reward": 0.9801827371120453,
      "rewards-max/forensics_strict_edge_reward": 0.37559792399406433,
      "rewards-max/forensics_tiou_reward": 0.971320390701294,
      "rewards/forensics_sinkhorn_reward": 0.6517775356769562,
      "rewards/forensics_strict_edge_reward": 0.03972861869260669,
      "rewards/forensics_tiou_reward": 0.5332376956939697,
      "step": 507
    },
    {
      "completion_length": 30.3125,
      "epoch": 1.5591397849462365,
      "grad_norm": 5.57194185256958,
      "kl": 0.2958984375,
      "learning_rate": 6.1e-07,
      "loss": -0.0915,
      "reward": 1.1464561223983765,
      "reward_std": 0.20263304561376572,
      "reward_std_all": 0.9489372074604034,
      "rewards-max/forensics_sinkhorn_reward": 0.9387572109699249,
      "rewards-max/forensics_strict_edge_reward": 0.32402659952640533,
      "rewards-max/forensics_tiou_reward": 0.8812069594860077,
      "rewards/forensics_sinkhorn_reward": 0.6556445062160492,
      "rewards/forensics_strict_edge_reward": 0.023162774741649628,
      "rewards/forensics_tiou_reward": 0.4676489681005478,
      "step": 508
    },
    {
      "completion_length": 24.0,
      "epoch": 1.5622119815668203,
      "grad_norm": 7.290805339813232,
      "kl": 0.4658203125,
      "learning_rate": 6.092307692307692e-07,
      "loss": -0.1136,
      "reward": 1.2219324111938477,
      "reward_std": 0.3921106457710266,
      "reward_std_all": 0.9762994945049286,
      "rewards-max/forensics_sinkhorn_reward": 0.9821906983852386,
      "rewards-max/forensics_strict_edge_reward": 0.4322323799133301,
      "rewards-max/forensics_tiou_reward": 0.9828334748744965,
      "rewards/forensics_sinkhorn_reward": 0.67200767993927,
      "rewards/forensics_strict_edge_reward": 0.04474516957998276,
      "rewards/forensics_tiou_reward": 0.5051795989274979,
      "step": 509
    },
    {
      "completion_length": 23.171875,
      "epoch": 1.565284178187404,
      "grad_norm": 6.315638065338135,
      "kl": 0.3486328125,
      "learning_rate": 6.084615384615385e-07,
      "loss": -0.1022,
      "reward": 1.1534453630447388,
      "reward_std": 0.4071638584136963,
      "reward_std_all": 1.0017668902873993,
      "rewards-max/forensics_sinkhorn_reward": 0.9536769986152649,
      "rewards-max/forensics_strict_edge_reward": 0.578405112028122,
      "rewards-max/forensics_tiou_reward": 0.9144206047058105,
      "rewards/forensics_sinkhorn_reward": 0.624345600605011,
      "rewards/forensics_strict_edge_reward": 0.05412205122411251,
      "rewards/forensics_tiou_reward": 0.4749777466058731,
      "step": 510
    },
    {
      "completion_length": 25.078125,
      "epoch": 1.5683563748079878,
      "grad_norm": 7.240414142608643,
      "kl": 0.4931640625,
      "learning_rate": 6.076923076923076e-07,
      "loss": -0.0381,
      "reward": 1.1240413784980774,
      "reward_std": 0.3015870600938797,
      "reward_std_all": 0.9952654242515564,
      "rewards-max/forensics_sinkhorn_reward": 0.9647365212440491,
      "rewards-max/forensics_strict_edge_reward": 0.3355930745601654,
      "rewards-max/forensics_tiou_reward": 0.9351403415203094,
      "rewards/forensics_sinkhorn_reward": 0.639804869890213,
      "rewards/forensics_strict_edge_reward": 0.0211164653301239,
      "rewards/forensics_tiou_reward": 0.4631200432777405,
      "step": 511
    },
    {
      "completion_length": 23.671875,
      "epoch": 1.5714285714285714,
      "grad_norm": 7.113353729248047,
      "kl": 0.453125,
      "learning_rate": 6.069230769230768e-07,
      "loss": -0.0741,
      "reward": 1.08370840549469,
      "reward_std": 0.454568088054657,
      "reward_std_all": 1.0590373873710632,
      "rewards-max/forensics_sinkhorn_reward": 0.9776313006877899,
      "rewards-max/forensics_strict_edge_reward": 0.6862294971942902,
      "rewards-max/forensics_tiou_reward": 0.9737209975719452,
      "rewards/forensics_sinkhorn_reward": 0.5978790819644928,
      "rewards/forensics_strict_edge_reward": 0.045687125995755196,
      "rewards/forensics_tiou_reward": 0.4401422142982483,
      "step": 512
    },
    {
      "completion_length": 26.921875,
      "epoch": 1.5745007680491552,
      "grad_norm": 6.119558811187744,
      "kl": 0.3779296875,
      "learning_rate": 6.061538461538462e-07,
      "loss": -0.0627,
      "reward": 0.9917477965354919,
      "reward_std": 0.313786581158638,
      "reward_std_all": 1.0579919815063477,
      "rewards-max/forensics_sinkhorn_reward": 0.9651297926902771,
      "rewards-max/forensics_strict_edge_reward": 0.21636603027582169,
      "rewards-max/forensics_tiou_reward": 0.9428102672100067,
      "rewards/forensics_sinkhorn_reward": 0.5865695476531982,
      "rewards/forensics_strict_edge_reward": 0.012949440162628889,
      "rewards/forensics_tiou_reward": 0.3922288119792938,
      "step": 513
    },
    {
      "completion_length": 26.671875,
      "epoch": 1.5775729646697387,
      "grad_norm": 5.846378803253174,
      "kl": 0.4169921875,
      "learning_rate": 6.053846153846154e-07,
      "loss": -0.0749,
      "reward": 1.075520932674408,
      "reward_std": 0.35761086642742157,
      "reward_std_all": 1.0623275935649872,
      "rewards-max/forensics_sinkhorn_reward": 0.9838723838329315,
      "rewards-max/forensics_strict_edge_reward": 0.5248935334384441,
      "rewards-max/forensics_tiou_reward": 0.9830979108810425,
      "rewards/forensics_sinkhorn_reward": 0.589748352766037,
      "rewards/forensics_strict_edge_reward": 0.0525763425976038,
      "rewards/forensics_tiou_reward": 0.4331963062286377,
      "step": 514
    },
    {
      "completion_length": 26.34375,
      "epoch": 1.5806451612903225,
      "grad_norm": 6.718051910400391,
      "kl": 0.3720703125,
      "learning_rate": 6.046153846153847e-07,
      "loss": -0.0671,
      "reward": 0.8874135613441467,
      "reward_std": 0.19802630692720413,
      "reward_std_all": 1.0742442607879639,
      "rewards-max/forensics_sinkhorn_reward": 0.9197719693183899,
      "rewards-max/forensics_strict_edge_reward": 0.07515854388475418,
      "rewards-max/forensics_tiou_reward": 0.8481709957122803,
      "rewards/forensics_sinkhorn_reward": 0.5556734800338745,
      "rewards/forensics_strict_edge_reward": 0.0041302782483398914,
      "rewards/forensics_tiou_reward": 0.3276097774505615,
      "step": 515
    },
    {
      "completion_length": 24.046875,
      "epoch": 1.5837173579109063,
      "grad_norm": 6.799293041229248,
      "kl": 0.4482421875,
      "learning_rate": 6.038461538461538e-07,
      "loss": -0.116,
      "reward": 1.2070772647857666,
      "reward_std": 0.2640908434987068,
      "reward_std_all": 0.9384708404541016,
      "rewards-max/forensics_sinkhorn_reward": 0.9812850952148438,
      "rewards-max/forensics_strict_edge_reward": 0.4855715036392212,
      "rewards-max/forensics_tiou_reward": 0.9814274311065674,
      "rewards/forensics_sinkhorn_reward": 0.6725416481494904,
      "rewards/forensics_strict_edge_reward": 0.04388982988893986,
      "rewards/forensics_tiou_reward": 0.4906458556652069,
      "step": 516
    },
    {
      "completion_length": 22.8125,
      "epoch": 1.58678955453149,
      "grad_norm": 7.135873794555664,
      "kl": 0.4921875,
      "learning_rate": 6.03076923076923e-07,
      "loss": -0.0795,
      "reward": 1.0083626210689545,
      "reward_std": 0.2679886519908905,
      "reward_std_all": 1.0307809114456177,
      "rewards-max/forensics_sinkhorn_reward": 0.9163213670253754,
      "rewards-max/forensics_strict_edge_reward": 0.24249917641282082,
      "rewards-max/forensics_tiou_reward": 0.8463538289070129,
      "rewards/forensics_sinkhorn_reward": 0.6068571209907532,
      "rewards/forensics_strict_edge_reward": 0.016734858276322484,
      "rewards/forensics_tiou_reward": 0.3847706764936447,
      "step": 517
    },
    {
      "completion_length": 25.734375,
      "epoch": 1.5898617511520738,
      "grad_norm": 6.432401180267334,
      "kl": 0.2763671875,
      "learning_rate": 6.023076923076923e-07,
      "loss": -0.05,
      "reward": 0.7586999237537384,
      "reward_std": 0.3268561363220215,
      "reward_std_all": 1.1779481768608093,
      "rewards-max/forensics_sinkhorn_reward": 0.9665871262550354,
      "rewards-max/forensics_strict_edge_reward": 0.34909766912460327,
      "rewards-max/forensics_tiou_reward": 0.9465848207473755,
      "rewards/forensics_sinkhorn_reward": 0.4663413465023041,
      "rewards/forensics_strict_edge_reward": 0.01885797269642353,
      "rewards/forensics_tiou_reward": 0.27350060641765594,
      "step": 518
    },
    {
      "completion_length": 22.484375,
      "epoch": 1.5929339477726574,
      "grad_norm": 6.130648612976074,
      "kl": 0.5302734375,
      "learning_rate": 6.015384615384615e-07,
      "loss": -0.0974,
      "reward": 1.1004670858383179,
      "reward_std": 0.26955416798591614,
      "reward_std_all": 0.9941178560256958,
      "rewards-max/forensics_sinkhorn_reward": 0.9797606766223907,
      "rewards-max/forensics_strict_edge_reward": 0.43584804236888885,
      "rewards-max/forensics_tiou_reward": 0.9734403192996979,
      "rewards/forensics_sinkhorn_reward": 0.6626292169094086,
      "rewards/forensics_strict_edge_reward": 0.024413133040070534,
      "rewards/forensics_tiou_reward": 0.4134247601032257,
      "step": 519
    },
    {
      "completion_length": 24.640625,
      "epoch": 1.5960061443932412,
      "grad_norm": 6.331103324890137,
      "kl": 0.333984375,
      "learning_rate": 6.007692307692307e-07,
      "loss": -0.1118,
      "reward": 0.9462096989154816,
      "reward_std": 0.29322783648967743,
      "reward_std_all": 1.0731963515281677,
      "rewards-max/forensics_sinkhorn_reward": 0.9651730358600616,
      "rewards-max/forensics_strict_edge_reward": 0.2649586796760559,
      "rewards-max/forensics_tiou_reward": 0.9421728253364563,
      "rewards/forensics_sinkhorn_reward": 0.5815526247024536,
      "rewards/forensics_strict_edge_reward": 0.013938755728304386,
      "rewards/forensics_tiou_reward": 0.35071828961372375,
      "step": 520
    },
    {
      "completion_length": 25.640625,
      "epoch": 1.5990783410138247,
      "grad_norm": 5.644484519958496,
      "kl": 0.384765625,
      "learning_rate": 6e-07,
      "loss": -0.1364,
      "reward": 1.039471447467804,
      "reward_std": 0.36189429461956024,
      "reward_std_all": 1.0505146086215973,
      "rewards-max/forensics_sinkhorn_reward": 0.9815424084663391,
      "rewards-max/forensics_strict_edge_reward": 0.6149742305278778,
      "rewards-max/forensics_tiou_reward": 0.9738483726978302,
      "rewards/forensics_sinkhorn_reward": 0.6240208745002747,
      "rewards/forensics_strict_edge_reward": 0.028625331819057465,
      "rewards/forensics_tiou_reward": 0.38682518899440765,
      "step": 521
    },
    {
      "completion_length": 22.171875,
      "epoch": 1.6021505376344085,
      "grad_norm": 5.609747886657715,
      "kl": 0.298828125,
      "learning_rate": 5.992307692307692e-07,
      "loss": -0.1082,
      "reward": 1.0533998608589172,
      "reward_std": 0.3677772432565689,
      "reward_std_all": 1.078635334968567,
      "rewards-max/forensics_sinkhorn_reward": 0.9784592688083649,
      "rewards-max/forensics_strict_edge_reward": 0.5336027555167675,
      "rewards-max/forensics_tiou_reward": 0.9647858142852783,
      "rewards/forensics_sinkhorn_reward": 0.5933961272239685,
      "rewards/forensics_strict_edge_reward": 0.062337269773706794,
      "rewards/forensics_tiou_reward": 0.39766645431518555,
      "step": 522
    },
    {
      "completion_length": 27.28125,
      "epoch": 1.6052227342549923,
      "grad_norm": 6.30155086517334,
      "kl": 0.4169921875,
      "learning_rate": 5.984615384615385e-07,
      "loss": -0.0994,
      "reward": 1.0040660500526428,
      "reward_std": 0.2661936283111572,
      "reward_std_all": 1.0477626323699951,
      "rewards-max/forensics_sinkhorn_reward": 0.9673635959625244,
      "rewards-max/forensics_strict_edge_reward": 0.31268224120140076,
      "rewards-max/forensics_tiou_reward": 0.9432847797870636,
      "rewards/forensics_sinkhorn_reward": 0.5635202378034592,
      "rewards/forensics_strict_edge_reward": 0.021875777281820774,
      "rewards/forensics_tiou_reward": 0.41867007315158844,
      "step": 523
    },
    {
      "completion_length": 22.953125,
      "epoch": 1.608294930875576,
      "grad_norm": 4.701626300811768,
      "kl": 0.3505859375,
      "learning_rate": 5.976923076923077e-07,
      "loss": 0.0152,
      "reward": 1.1653844118118286,
      "reward_std": 0.22213665395975113,
      "reward_std_all": 0.9402602910995483,
      "rewards-max/forensics_sinkhorn_reward": 0.9560542106628418,
      "rewards-max/forensics_strict_edge_reward": 0.11359046772122383,
      "rewards-max/forensics_tiou_reward": 0.9061916470527649,
      "rewards/forensics_sinkhorn_reward": 0.6816249787807465,
      "rewards/forensics_strict_edge_reward": 0.009661386720836163,
      "rewards/forensics_tiou_reward": 0.4740980565547943,
      "step": 524
    },
    {
      "completion_length": 23.953125,
      "epoch": 1.6113671274961598,
      "grad_norm": 6.413716793060303,
      "kl": 0.3037109375,
      "learning_rate": 5.969230769230768e-07,
      "loss": -0.0832,
      "reward": 0.7915970683097839,
      "reward_std": 0.24620606750249863,
      "reward_std_all": 1.136875867843628,
      "rewards-max/forensics_sinkhorn_reward": 0.9160980582237244,
      "rewards-max/forensics_strict_edge_reward": 0.295737074688077,
      "rewards-max/forensics_tiou_reward": 0.80662402510643,
      "rewards/forensics_sinkhorn_reward": 0.49723444879055023,
      "rewards/forensics_strict_edge_reward": 0.016491034322825726,
      "rewards/forensics_tiou_reward": 0.27787162363529205,
      "step": 525
    },
    {
      "completion_length": 26.21875,
      "epoch": 1.6144393241167436,
      "grad_norm": 6.288809776306152,
      "kl": 0.3994140625,
      "learning_rate": 5.961538461538461e-07,
      "loss": -0.0303,
      "reward": 1.1609699130058289,
      "reward_std": 0.24345102161169052,
      "reward_std_all": 0.9464722275733948,
      "rewards-max/forensics_sinkhorn_reward": 0.9663023948669434,
      "rewards-max/forensics_strict_edge_reward": 0.06275854585692286,
      "rewards-max/forensics_tiou_reward": 0.9405975639820099,
      "rewards/forensics_sinkhorn_reward": 0.7040437757968903,
      "rewards/forensics_strict_edge_reward": 0.007954122265800834,
      "rewards/forensics_tiou_reward": 0.4489719420671463,
      "step": 526
    },
    {
      "completion_length": 22.171875,
      "epoch": 1.6175115207373272,
      "grad_norm": 6.116084098815918,
      "kl": 0.365234375,
      "learning_rate": 5.953846153846154e-07,
      "loss": -0.0641,
      "reward": 1.2809637188911438,
      "reward_std": 0.26934191584587097,
      "reward_std_all": 0.9080806374549866,
      "rewards-max/forensics_sinkhorn_reward": 0.9677160680294037,
      "rewards-max/forensics_strict_edge_reward": 0.39497099816799164,
      "rewards-max/forensics_tiou_reward": 0.9203869700431824,
      "rewards/forensics_sinkhorn_reward": 0.7298740148544312,
      "rewards/forensics_strict_edge_reward": 0.03419929835945368,
      "rewards/forensics_tiou_reward": 0.5168904066085815,
      "step": 527
    },
    {
      "completion_length": 21.953125,
      "epoch": 1.620583717357911,
      "grad_norm": 4.856773376464844,
      "kl": 0.28515625,
      "learning_rate": 5.946153846153847e-07,
      "loss": -0.0378,
      "reward": 1.1527068614959717,
      "reward_std": 0.4081765413284302,
      "reward_std_all": 1.02132910490036,
      "rewards-max/forensics_sinkhorn_reward": 0.9718294143676758,
      "rewards-max/forensics_strict_edge_reward": 0.5060040131211281,
      "rewards-max/forensics_tiou_reward": 0.9465264976024628,
      "rewards/forensics_sinkhorn_reward": 0.6542496681213379,
      "rewards/forensics_strict_edge_reward": 0.04886004235595465,
      "rewards/forensics_tiou_reward": 0.449597105383873,
      "step": 528
    },
    {
      "completion_length": 22.453125,
      "epoch": 1.6236559139784945,
      "grad_norm": 6.144923210144043,
      "kl": 0.423828125,
      "learning_rate": 5.938461538461538e-07,
      "loss": -0.1523,
      "reward": 1.212767779827118,
      "reward_std": 0.34626591205596924,
      "reward_std_all": 0.954139232635498,
      "rewards-max/forensics_sinkhorn_reward": 0.9766551554203033,
      "rewards-max/forensics_strict_edge_reward": 0.4827088415622711,
      "rewards-max/forensics_tiou_reward": 0.9702271521091461,
      "rewards/forensics_sinkhorn_reward": 0.7005468308925629,
      "rewards/forensics_strict_edge_reward": 0.03490776848047972,
      "rewards/forensics_tiou_reward": 0.47731316089630127,
      "step": 529
    },
    {
      "completion_length": 22.203125,
      "epoch": 1.6267281105990783,
      "grad_norm": 5.946403503417969,
      "kl": 0.3271484375,
      "learning_rate": 5.93076923076923e-07,
      "loss": -0.0874,
      "reward": 1.2047486901283264,
      "reward_std": 0.3078523874282837,
      "reward_std_all": 0.9658293724060059,
      "rewards-max/forensics_sinkhorn_reward": 0.979913055896759,
      "rewards-max/forensics_strict_edge_reward": 0.5139312744140625,
      "rewards-max/forensics_tiou_reward": 0.971168041229248,
      "rewards/forensics_sinkhorn_reward": 0.7037605047225952,
      "rewards/forensics_strict_edge_reward": 0.027862737886607647,
      "rewards/forensics_tiou_reward": 0.47312548756599426,
      "step": 530
    },
    {
      "completion_length": 27.265625,
      "epoch": 1.629800307219662,
      "grad_norm": 5.502708435058594,
      "kl": 0.32421875,
      "learning_rate": 5.923076923076923e-07,
      "loss": -0.1004,
      "reward": 1.0355307757854462,
      "reward_std": 0.3499760627746582,
      "reward_std_all": 1.0791810750961304,
      "rewards-max/forensics_sinkhorn_reward": 0.9752820730209351,
      "rewards-max/forensics_strict_edge_reward": 0.6779094636440277,
      "rewards-max/forensics_tiou_reward": 0.9657518565654755,
      "rewards/forensics_sinkhorn_reward": 0.5620384216308594,
      "rewards/forensics_strict_edge_reward": 0.0747328707948327,
      "rewards/forensics_tiou_reward": 0.39875948429107666,
      "step": 531
    },
    {
      "completion_length": 28.96875,
      "epoch": 1.6328725038402458,
      "grad_norm": 5.50420618057251,
      "kl": 0.3466796875,
      "learning_rate": 5.915384615384615e-07,
      "loss": -0.129,
      "reward": 1.122454285621643,
      "reward_std": 0.3119807727634907,
      "reward_std_all": 1.0009559392929077,
      "rewards-max/forensics_sinkhorn_reward": 0.9775584042072296,
      "rewards-max/forensics_strict_edge_reward": 0.25206760317087173,
      "rewards-max/forensics_tiou_reward": 0.9696859121322632,
      "rewards/forensics_sinkhorn_reward": 0.6077950894832611,
      "rewards/forensics_strict_edge_reward": 0.0433482633670792,
      "rewards/forensics_tiou_reward": 0.47131092846393585,
      "step": 532
    },
    {
      "completion_length": 26.984375,
      "epoch": 1.6359447004608296,
      "grad_norm": 7.057748317718506,
      "kl": 0.4306640625,
      "learning_rate": 5.907692307692307e-07,
      "loss": -0.1013,
      "reward": 1.1912416815757751,
      "reward_std": 0.27673661708831787,
      "reward_std_all": 0.9530562460422516,
      "rewards-max/forensics_sinkhorn_reward": 0.9772481620311737,
      "rewards-max/forensics_strict_edge_reward": 0.5744507163763046,
      "rewards-max/forensics_tiou_reward": 0.966188371181488,
      "rewards/forensics_sinkhorn_reward": 0.6902697086334229,
      "rewards/forensics_strict_edge_reward": 0.030820440500974655,
      "rewards/forensics_tiou_reward": 0.470151424407959,
      "step": 533
    },
    {
      "completion_length": 22.21875,
      "epoch": 1.6390168970814132,
      "grad_norm": 6.89379358291626,
      "kl": 0.41015625,
      "learning_rate": 5.9e-07,
      "loss": -0.1109,
      "reward": 0.9871230125427246,
      "reward_std": 0.38976840674877167,
      "reward_std_all": 1.0786859393119812,
      "rewards-max/forensics_sinkhorn_reward": 0.9673031866550446,
      "rewards-max/forensics_strict_edge_reward": 0.41901181638240814,
      "rewards-max/forensics_tiou_reward": 0.940630704164505,
      "rewards/forensics_sinkhorn_reward": 0.5836232304573059,
      "rewards/forensics_strict_edge_reward": 0.02687640395015478,
      "rewards/forensics_tiou_reward": 0.37662340700626373,
      "step": 534
    },
    {
      "completion_length": 24.078125,
      "epoch": 1.642089093701997,
      "grad_norm": 5.657297134399414,
      "kl": 0.4072265625,
      "learning_rate": 5.892307692307692e-07,
      "loss": -0.0709,
      "reward": 1.0848557949066162,
      "reward_std": 0.32675398886203766,
      "reward_std_all": 1.0194880366325378,
      "rewards-max/forensics_sinkhorn_reward": 0.9804419279098511,
      "rewards-max/forensics_strict_edge_reward": 0.5858925580978394,
      "rewards-max/forensics_tiou_reward": 0.9708846509456635,
      "rewards/forensics_sinkhorn_reward": 0.6098580956459045,
      "rewards/forensics_strict_edge_reward": 0.054875850677490234,
      "rewards/forensics_tiou_reward": 0.42012183368206024,
      "step": 535
    },
    {
      "completion_length": 25.4375,
      "epoch": 1.6451612903225805,
      "grad_norm": 6.2005743980407715,
      "kl": 0.3759765625,
      "learning_rate": 5.884615384615385e-07,
      "loss": -0.0729,
      "reward": 1.1411415934562683,
      "reward_std": 0.24680838733911514,
      "reward_std_all": 0.9624104797840118,
      "rewards-max/forensics_sinkhorn_reward": 0.9615939259529114,
      "rewards-max/forensics_strict_edge_reward": 0.28734029456973076,
      "rewards-max/forensics_tiou_reward": 0.9242352843284607,
      "rewards/forensics_sinkhorn_reward": 0.6447109878063202,
      "rewards/forensics_strict_edge_reward": 0.019991938723251224,
      "rewards/forensics_tiou_reward": 0.4764386713504791,
      "step": 536
    },
    {
      "completion_length": 25.125,
      "epoch": 1.6482334869431643,
      "grad_norm": 5.426187992095947,
      "kl": 0.421875,
      "learning_rate": 5.876923076923077e-07,
      "loss": -0.1645,
      "reward": 1.2987406253814697,
      "reward_std": 0.500198245048523,
      "reward_std_all": 0.9989216923713684,
      "rewards-max/forensics_sinkhorn_reward": 0.9827874600887299,
      "rewards-max/forensics_strict_edge_reward": 0.7514603435993195,
      "rewards-max/forensics_tiou_reward": 0.9819150567054749,
      "rewards/forensics_sinkhorn_reward": 0.6806720495223999,
      "rewards/forensics_strict_edge_reward": 0.09216750413179398,
      "rewards/forensics_tiou_reward": 0.5259011685848236,
      "step": 537
    },
    {
      "completion_length": 28.8125,
      "epoch": 1.651305683563748,
      "grad_norm": 9.079395294189453,
      "kl": 0.57421875,
      "learning_rate": 5.869230769230768e-07,
      "loss": 0.0152,
      "reward": 1.182852864265442,
      "reward_std": 0.4573708772659302,
      "reward_std_all": 1.0216213464736938,
      "rewards-max/forensics_sinkhorn_reward": 0.982032984495163,
      "rewards-max/forensics_strict_edge_reward": 0.45074111223220825,
      "rewards-max/forensics_tiou_reward": 0.9832020699977875,
      "rewards/forensics_sinkhorn_reward": 0.6487699747085571,
      "rewards/forensics_strict_edge_reward": 0.047298360615968704,
      "rewards/forensics_tiou_reward": 0.48678451776504517,
      "step": 538
    },
    {
      "completion_length": 33.515625,
      "epoch": 1.6543778801843319,
      "grad_norm": 6.033576965332031,
      "kl": 0.3662109375,
      "learning_rate": 5.861538461538461e-07,
      "loss": -0.106,
      "reward": 1.110927164554596,
      "reward_std": 0.2866029441356659,
      "reward_std_all": 0.9959877729415894,
      "rewards-max/forensics_sinkhorn_reward": 0.9817070960998535,
      "rewards-max/forensics_strict_edge_reward": 0.5775986015796661,
      "rewards-max/forensics_tiou_reward": 0.9802006185054779,
      "rewards/forensics_sinkhorn_reward": 0.6075096428394318,
      "rewards/forensics_strict_edge_reward": 0.04289861395955086,
      "rewards/forensics_tiou_reward": 0.46051888167858124,
      "step": 539
    },
    {
      "completion_length": 28.28125,
      "epoch": 1.6574500768049156,
      "grad_norm": 6.380727291107178,
      "kl": 0.4150390625,
      "learning_rate": 5.853846153846154e-07,
      "loss": 0.0008,
      "reward": 1.2608936429023743,
      "reward_std": 0.3448403775691986,
      "reward_std_all": 0.9481605887413025,
      "rewards-max/forensics_sinkhorn_reward": 0.9837192893028259,
      "rewards-max/forensics_strict_edge_reward": 0.41292841732501984,
      "rewards-max/forensics_tiou_reward": 0.9877577424049377,
      "rewards/forensics_sinkhorn_reward": 0.6870093047618866,
      "rewards/forensics_strict_edge_reward": 0.04456003662198782,
      "rewards/forensics_tiou_reward": 0.5293242335319519,
      "step": 540
    },
    {
      "completion_length": 24.078125,
      "epoch": 1.6605222734254992,
      "grad_norm": 6.603221893310547,
      "kl": 0.41015625,
      "learning_rate": 5.846153846153847e-07,
      "loss": -0.1174,
      "reward": 1.4277573823928833,
      "reward_std": 0.3205687701702118,
      "reward_std_all": 0.8507031798362732,
      "rewards-max/forensics_sinkhorn_reward": 0.9819689095020294,
      "rewards-max/forensics_strict_edge_reward": 0.5321442186832428,
      "rewards-max/forensics_tiou_reward": 0.9799306690692902,
      "rewards/forensics_sinkhorn_reward": 0.7673640847206116,
      "rewards/forensics_strict_edge_reward": 0.05671199411153793,
      "rewards/forensics_tiou_reward": 0.6036812365055084,
      "step": 541
    },
    {
      "completion_length": 24.28125,
      "epoch": 1.663594470046083,
      "grad_norm": 7.202409267425537,
      "kl": 0.4072265625,
      "learning_rate": 5.838461538461538e-07,
      "loss": -0.1813,
      "reward": 1.4012545347213745,
      "reward_std": 0.3137727305293083,
      "reward_std_all": 0.8611657321453094,
      "rewards-max/forensics_sinkhorn_reward": 0.9875510931015015,
      "rewards-max/forensics_strict_edge_reward": 0.7093025743961334,
      "rewards-max/forensics_tiou_reward": 0.9968555271625519,
      "rewards/forensics_sinkhorn_reward": 0.7640962302684784,
      "rewards/forensics_strict_edge_reward": 0.055458398535847664,
      "rewards/forensics_tiou_reward": 0.581699788570404,
      "step": 542
    },
    {
      "completion_length": 25.46875,
      "epoch": 1.6666666666666665,
      "grad_norm": 5.83840799331665,
      "kl": 0.3779296875,
      "learning_rate": 5.83076923076923e-07,
      "loss": -0.0739,
      "reward": 1.0663886666297913,
      "reward_std": 0.2471182718873024,
      "reward_std_all": 1.0134334564208984,
      "rewards-max/forensics_sinkhorn_reward": 0.9460619688034058,
      "rewards-max/forensics_strict_edge_reward": 0.2561907023191452,
      "rewards-max/forensics_tiou_reward": 0.8986762166023254,
      "rewards/forensics_sinkhorn_reward": 0.6202865242958069,
      "rewards/forensics_strict_edge_reward": 0.02493313793092966,
      "rewards/forensics_tiou_reward": 0.4211690276861191,
      "step": 543
    },
    {
      "completion_length": 30.625,
      "epoch": 1.6697388632872503,
      "grad_norm": 6.418644428253174,
      "kl": 0.3759765625,
      "learning_rate": 5.823076923076923e-07,
      "loss": -0.0833,
      "reward": 1.1865162253379822,
      "reward_std": 0.3181096315383911,
      "reward_std_all": 0.9576747417449951,
      "rewards-max/forensics_sinkhorn_reward": 0.9338402450084686,
      "rewards-max/forensics_strict_edge_reward": 0.24664020538330078,
      "rewards-max/forensics_tiou_reward": 0.8749483227729797,
      "rewards/forensics_sinkhorn_reward": 0.6423783600330353,
      "rewards/forensics_strict_edge_reward": 0.030053730122745037,
      "rewards/forensics_tiou_reward": 0.5140841454267502,
      "step": 544
    },
    {
      "completion_length": 25.78125,
      "epoch": 1.672811059907834,
      "grad_norm": 7.709845066070557,
      "kl": 0.3408203125,
      "learning_rate": 5.815384615384615e-07,
      "loss": -0.0857,
      "reward": 1.0804770588874817,
      "reward_std": 0.3130692094564438,
      "reward_std_all": 1.0190727412700653,
      "rewards-max/forensics_sinkhorn_reward": 0.969901978969574,
      "rewards-max/forensics_strict_edge_reward": 0.4676465690135956,
      "rewards-max/forensics_tiou_reward": 0.9512568414211273,
      "rewards/forensics_sinkhorn_reward": 0.6177161931991577,
      "rewards/forensics_strict_edge_reward": 0.023798292502760887,
      "rewards/forensics_tiou_reward": 0.43896251916885376,
      "step": 545
    },
    {
      "completion_length": 27.015625,
      "epoch": 1.6758832565284179,
      "grad_norm": 6.126577377319336,
      "kl": 0.322265625,
      "learning_rate": 5.807692307692307e-07,
      "loss": -0.0692,
      "reward": 1.0982116758823395,
      "reward_std": 0.3100244849920273,
      "reward_std_all": 1.0017262697219849,
      "rewards-max/forensics_sinkhorn_reward": 0.9694804549217224,
      "rewards-max/forensics_strict_edge_reward": 0.4268413633108139,
      "rewards-max/forensics_tiou_reward": 0.9531735181808472,
      "rewards/forensics_sinkhorn_reward": 0.5820195376873016,
      "rewards/forensics_strict_edge_reward": 0.03672408498823643,
      "rewards/forensics_tiou_reward": 0.47946797311306,
      "step": 546
    },
    {
      "completion_length": 24.734375,
      "epoch": 1.6789554531490016,
      "grad_norm": 6.644845008850098,
      "kl": 0.4755859375,
      "learning_rate": 5.8e-07,
      "loss": -0.0238,
      "reward": 1.0596231818199158,
      "reward_std": 0.3559540957212448,
      "reward_std_all": 1.0516435503959656,
      "rewards-max/forensics_sinkhorn_reward": 0.9677689969539642,
      "rewards-max/forensics_strict_edge_reward": 0.538444846868515,
      "rewards-max/forensics_tiou_reward": 0.9464327096939087,
      "rewards/forensics_sinkhorn_reward": 0.6251238584518433,
      "rewards/forensics_strict_edge_reward": 0.03160545043647289,
      "rewards/forensics_tiou_reward": 0.40289390087127686,
      "step": 547
    },
    {
      "completion_length": 23.4375,
      "epoch": 1.6820276497695854,
      "grad_norm": 6.72635555267334,
      "kl": 0.302734375,
      "learning_rate": 5.792307692307692e-07,
      "loss": -0.0933,
      "reward": 0.9976440668106079,
      "reward_std": 0.4734819233417511,
      "reward_std_all": 1.0982365608215332,
      "rewards-max/forensics_sinkhorn_reward": 0.9564709365367889,
      "rewards-max/forensics_strict_edge_reward": 0.4697008728981018,
      "rewards-max/forensics_tiou_reward": 0.9250199794769287,
      "rewards/forensics_sinkhorn_reward": 0.5826177000999451,
      "rewards/forensics_strict_edge_reward": 0.04057169705629349,
      "rewards/forensics_tiou_reward": 0.37445466220378876,
      "step": 548
    },
    {
      "completion_length": 25.15625,
      "epoch": 1.685099846390169,
      "grad_norm": 6.287343978881836,
      "kl": 0.3056640625,
      "learning_rate": 5.784615384615385e-07,
      "loss": -0.0949,
      "reward": 1.2572245001792908,
      "reward_std": 0.31671836972236633,
      "reward_std_all": 0.9272087812423706,
      "rewards-max/forensics_sinkhorn_reward": 0.9780814945697784,
      "rewards-max/forensics_strict_edge_reward": 0.4132660925388336,
      "rewards-max/forensics_tiou_reward": 0.971581369638443,
      "rewards/forensics_sinkhorn_reward": 0.7079965472221375,
      "rewards/forensics_strict_edge_reward": 0.04146555811166763,
      "rewards/forensics_tiou_reward": 0.5077624320983887,
      "step": 549
    },
    {
      "completion_length": 25.9375,
      "epoch": 1.6881720430107527,
      "grad_norm": 7.076443195343018,
      "kl": 0.2890625,
      "learning_rate": 5.776923076923077e-07,
      "loss": -0.0679,
      "reward": 1.13001948595047,
      "reward_std": 0.2553075775504112,
      "reward_std_all": 0.9743199944496155,
      "rewards-max/forensics_sinkhorn_reward": 0.9442188143730164,
      "rewards-max/forensics_strict_edge_reward": 0.43628542125225067,
      "rewards-max/forensics_tiou_reward": 0.8958526253700256,
      "rewards/forensics_sinkhorn_reward": 0.6870556473731995,
      "rewards/forensics_strict_edge_reward": 0.023202226497232914,
      "rewards/forensics_tiou_reward": 0.4197615534067154,
      "step": 550
    },
    {
      "completion_length": 23.375,
      "epoch": 1.6912442396313363,
      "grad_norm": 6.124660968780518,
      "kl": 0.3095703125,
      "learning_rate": 5.769230769230768e-07,
      "loss": -0.1103,
      "reward": 1.279638648033142,
      "reward_std": 0.34714657068252563,
      "reward_std_all": 0.9227503836154938,
      "rewards-max/forensics_sinkhorn_reward": 0.9735982716083527,
      "rewards-max/forensics_strict_edge_reward": 0.384300097823143,
      "rewards-max/forensics_tiou_reward": 0.9549532532691956,
      "rewards/forensics_sinkhorn_reward": 0.7158968150615692,
      "rewards/forensics_strict_edge_reward": 0.0370451956987381,
      "rewards/forensics_tiou_reward": 0.526696652173996,
      "step": 551
    },
    {
      "completion_length": 26.328125,
      "epoch": 1.69431643625192,
      "grad_norm": 5.753393650054932,
      "kl": 0.33203125,
      "learning_rate": 5.761538461538461e-07,
      "loss": -0.064,
      "reward": 1.3710802793502808,
      "reward_std": 0.17406391352415085,
      "reward_std_all": 0.8326611816883087,
      "rewards-max/forensics_sinkhorn_reward": 0.9630318582057953,
      "rewards-max/forensics_strict_edge_reward": 0.3298750054091215,
      "rewards-max/forensics_tiou_reward": 0.9287252128124237,
      "rewards/forensics_sinkhorn_reward": 0.7513275444507599,
      "rewards/forensics_strict_edge_reward": 0.022675607819110155,
      "rewards/forensics_tiou_reward": 0.5970771312713623,
      "step": 552
    },
    {
      "completion_length": 22.484375,
      "epoch": 1.6973886328725039,
      "grad_norm": 7.652938365936279,
      "kl": 0.3232421875,
      "learning_rate": 5.753846153846153e-07,
      "loss": -0.1069,
      "reward": 1.108830451965332,
      "reward_std": 0.2518090158700943,
      "reward_std_all": 1.0002628564834595,
      "rewards-max/forensics_sinkhorn_reward": 0.9612719118595123,
      "rewards-max/forensics_strict_edge_reward": 0.5483185648918152,
      "rewards-max/forensics_tiou_reward": 0.9362600147724152,
      "rewards/forensics_sinkhorn_reward": 0.6754977703094482,
      "rewards/forensics_strict_edge_reward": 0.039797297678887844,
      "rewards/forensics_tiou_reward": 0.39353540539741516,
      "step": 553
    },
    {
      "completion_length": 26.421875,
      "epoch": 1.7004608294930876,
      "grad_norm": 6.063847541809082,
      "kl": 0.3544921875,
      "learning_rate": 5.746153846153847e-07,
      "loss": -0.1182,
      "reward": 1.4188008308410645,
      "reward_std": 0.3876076340675354,
      "reward_std_all": 0.8774018883705139,
      "rewards-max/forensics_sinkhorn_reward": 0.981873631477356,
      "rewards-max/forensics_strict_edge_reward": 0.7309097647666931,
      "rewards-max/forensics_tiou_reward": 0.9818507134914398,
      "rewards/forensics_sinkhorn_reward": 0.7737614810466766,
      "rewards/forensics_strict_edge_reward": 0.0837374497205019,
      "rewards/forensics_tiou_reward": 0.5613017976284027,
      "step": 554
    },
    {
      "completion_length": 26.046875,
      "epoch": 1.7035330261136714,
      "grad_norm": 6.503593444824219,
      "kl": 0.51953125,
      "learning_rate": 5.738461538461539e-07,
      "loss": -0.1049,
      "reward": 1.0040486752986908,
      "reward_std": 0.3835919201374054,
      "reward_std_all": 1.0905380845069885,
      "rewards-max/forensics_sinkhorn_reward": 0.9715321063995361,
      "rewards-max/forensics_strict_edge_reward": 0.5931393504142761,
      "rewards-max/forensics_tiou_reward": 0.9563575983047485,
      "rewards/forensics_sinkhorn_reward": 0.5884060263633728,
      "rewards/forensics_strict_edge_reward": 0.03456384316086769,
      "rewards/forensics_tiou_reward": 0.381078839302063,
      "step": 555
    },
    {
      "completion_length": 24.71875,
      "epoch": 1.706605222734255,
      "grad_norm": 8.00336742401123,
      "kl": 0.4052734375,
      "learning_rate": 5.73076923076923e-07,
      "loss": -0.1316,
      "reward": 1.157826155424118,
      "reward_std": 0.3010431230068207,
      "reward_std_all": 0.9815273582935333,
      "rewards-max/forensics_sinkhorn_reward": 0.9766835570335388,
      "rewards-max/forensics_strict_edge_reward": 0.43584154546260834,
      "rewards-max/forensics_tiou_reward": 0.9689382314682007,
      "rewards/forensics_sinkhorn_reward": 0.6548289060592651,
      "rewards/forensics_strict_edge_reward": 0.03491080738604069,
      "rewards/forensics_tiou_reward": 0.468086376786232,
      "step": 556
    },
    {
      "completion_length": 24.65625,
      "epoch": 1.7096774193548387,
      "grad_norm": 8.737018585205078,
      "kl": 0.4853515625,
      "learning_rate": 5.723076923076923e-07,
      "loss": -0.1755,
      "reward": 1.0923653841018677,
      "reward_std": 0.3014342039823532,
      "reward_std_all": 1.0085600018501282,
      "rewards-max/forensics_sinkhorn_reward": 0.9653366506099701,
      "rewards-max/forensics_strict_edge_reward": 0.44491301476955414,
      "rewards-max/forensics_tiou_reward": 0.9395598769187927,
      "rewards/forensics_sinkhorn_reward": 0.6440135836601257,
      "rewards/forensics_strict_edge_reward": 0.03561055846512318,
      "rewards/forensics_tiou_reward": 0.4127412587404251,
      "step": 557
    },
    {
      "completion_length": 23.03125,
      "epoch": 1.7127496159754223,
      "grad_norm": 7.518070697784424,
      "kl": 0.419921875,
      "learning_rate": 5.715384615384615e-07,
      "loss": -0.1185,
      "reward": 1.194920003414154,
      "reward_std": 0.30956892669200897,
      "reward_std_all": 0.9561932682991028,
      "rewards-max/forensics_sinkhorn_reward": 0.9674814641475677,
      "rewards-max/forensics_strict_edge_reward": 0.4160068482160568,
      "rewards-max/forensics_tiou_reward": 0.9457299709320068,
      "rewards/forensics_sinkhorn_reward": 0.6844825744628906,
      "rewards/forensics_strict_edge_reward": 0.03516455739736557,
      "rewards/forensics_tiou_reward": 0.47527292370796204,
      "step": 558
    },
    {
      "completion_length": 26.140625,
      "epoch": 1.715821812596006,
      "grad_norm": 6.839488506317139,
      "kl": 0.462890625,
      "learning_rate": 5.707692307692307e-07,
      "loss": -0.106,
      "reward": 1.2371712923049927,
      "reward_std": 0.2825222760438919,
      "reward_std_all": 0.9409446120262146,
      "rewards-max/forensics_sinkhorn_reward": 0.9785034954547882,
      "rewards-max/forensics_strict_edge_reward": 0.30445480067282915,
      "rewards-max/forensics_tiou_reward": 0.9630981385707855,
      "rewards/forensics_sinkhorn_reward": 0.6866651177406311,
      "rewards/forensics_strict_edge_reward": 0.03194955503568053,
      "rewards/forensics_tiou_reward": 0.5185566842556,
      "step": 559
    },
    {
      "completion_length": 29.515625,
      "epoch": 1.7188940092165899,
      "grad_norm": 6.960942268371582,
      "kl": 0.4873046875,
      "learning_rate": 5.699999999999999e-07,
      "loss": -0.1618,
      "reward": 1.1301352381706238,
      "reward_std": 0.32403604686260223,
      "reward_std_all": 0.9959626793861389,
      "rewards-max/forensics_sinkhorn_reward": 0.943503737449646,
      "rewards-max/forensics_strict_edge_reward": 0.2740815691649914,
      "rewards-max/forensics_tiou_reward": 0.8886873722076416,
      "rewards/forensics_sinkhorn_reward": 0.640196293592453,
      "rewards/forensics_strict_edge_reward": 0.035061418544501066,
      "rewards/forensics_tiou_reward": 0.454877570271492,
      "step": 560
    },
    {
      "completion_length": 25.03125,
      "epoch": 1.7219662058371736,
      "grad_norm": 7.089996337890625,
      "kl": 0.4052734375,
      "learning_rate": 5.692307692307692e-07,
      "loss": -0.0851,
      "reward": 1.0765368938446045,
      "reward_std": 0.2519307881593704,
      "reward_std_all": 0.9982563853263855,
      "rewards-max/forensics_sinkhorn_reward": 0.9645750522613525,
      "rewards-max/forensics_strict_edge_reward": 0.2245134636759758,
      "rewards-max/forensics_tiou_reward": 0.9352526962757111,
      "rewards/forensics_sinkhorn_reward": 0.608276754617691,
      "rewards/forensics_strict_edge_reward": 0.023377922363579273,
      "rewards/forensics_tiou_reward": 0.4448821246623993,
      "step": 561
    },
    {
      "completion_length": 25.890625,
      "epoch": 1.7250384024577574,
      "grad_norm": 6.42067813873291,
      "kl": 0.404296875,
      "learning_rate": 5.684615384615385e-07,
      "loss": -0.1542,
      "reward": 1.222041130065918,
      "reward_std": 0.29125525057315826,
      "reward_std_all": 0.9418302178382874,
      "rewards-max/forensics_sinkhorn_reward": 0.9601953625679016,
      "rewards-max/forensics_strict_edge_reward": 0.34655629098415375,
      "rewards-max/forensics_tiou_reward": 0.9285517930984497,
      "rewards/forensics_sinkhorn_reward": 0.6985825896263123,
      "rewards/forensics_strict_edge_reward": 0.029959130100905895,
      "rewards/forensics_tiou_reward": 0.49349944293498993,
      "step": 562
    },
    {
      "completion_length": 27.296875,
      "epoch": 1.728110599078341,
      "grad_norm": 6.695307731628418,
      "kl": 0.3505859375,
      "learning_rate": 5.676923076923077e-07,
      "loss": -0.1174,
      "reward": 1.1600852012634277,
      "reward_std": 0.2588965892791748,
      "reward_std_all": 0.9659197926521301,
      "rewards-max/forensics_sinkhorn_reward": 0.974875271320343,
      "rewards-max/forensics_strict_edge_reward": 0.5006477534770966,
      "rewards-max/forensics_tiou_reward": 0.965610146522522,
      "rewards/forensics_sinkhorn_reward": 0.63495734333992,
      "rewards/forensics_strict_edge_reward": 0.049182353541255,
      "rewards/forensics_tiou_reward": 0.4759455770254135,
      "step": 563
    },
    {
      "completion_length": 27.8125,
      "epoch": 1.7311827956989247,
      "grad_norm": 7.874819278717041,
      "kl": 0.5390625,
      "learning_rate": 5.669230769230769e-07,
      "loss": -0.0654,
      "reward": 1.0550076365470886,
      "reward_std": 0.18784581869840622,
      "reward_std_all": 0.9893518388271332,
      "rewards-max/forensics_sinkhorn_reward": 0.9679137170314789,
      "rewards-max/forensics_strict_edge_reward": 0.3172316625714302,
      "rewards-max/forensics_tiou_reward": 0.9431418180465698,
      "rewards/forensics_sinkhorn_reward": 0.6160647571086884,
      "rewards/forensics_strict_edge_reward": 0.02059067413210869,
      "rewards/forensics_tiou_reward": 0.41835226118564606,
      "step": 564
    },
    {
      "completion_length": 27.109375,
      "epoch": 1.7342549923195083,
      "grad_norm": 5.997791767120361,
      "kl": 0.3857421875,
      "learning_rate": 5.661538461538461e-07,
      "loss": -0.0354,
      "reward": 1.335485816001892,
      "reward_std": 0.3683185875415802,
      "reward_std_all": 0.9247865378856659,
      "rewards-max/forensics_sinkhorn_reward": 0.9838157296180725,
      "rewards-max/forensics_strict_edge_reward": 0.6763986051082611,
      "rewards-max/forensics_tiou_reward": 0.9868182837963104,
      "rewards/forensics_sinkhorn_reward": 0.7254525125026703,
      "rewards/forensics_strict_edge_reward": 0.05776698887348175,
      "rewards/forensics_tiou_reward": 0.5522662848234177,
      "step": 565
    },
    {
      "completion_length": 24.34375,
      "epoch": 1.737327188940092,
      "grad_norm": 6.657999515533447,
      "kl": 0.3525390625,
      "learning_rate": 5.653846153846153e-07,
      "loss": -0.1561,
      "reward": 1.1022295951843262,
      "reward_std": 0.29225916415452957,
      "reward_std_all": 1.0106816589832306,
      "rewards-max/forensics_sinkhorn_reward": 0.9828861057758331,
      "rewards-max/forensics_strict_edge_reward": 0.5430071502923965,
      "rewards-max/forensics_tiou_reward": 0.9846460223197937,
      "rewards/forensics_sinkhorn_reward": 0.6272172927856445,
      "rewards/forensics_strict_edge_reward": 0.03790340945124626,
      "rewards/forensics_tiou_reward": 0.43710897862911224,
      "step": 566
    },
    {
      "completion_length": 24.234375,
      "epoch": 1.7403993855606759,
      "grad_norm": 6.2179694175720215,
      "kl": 0.3330078125,
      "learning_rate": 5.646153846153847e-07,
      "loss": -0.0746,
      "reward": 1.2401517033576965,
      "reward_std": 0.27040714025497437,
      "reward_std_all": 0.9212148487567902,
      "rewards-max/forensics_sinkhorn_reward": 0.9742621779441833,
      "rewards-max/forensics_strict_edge_reward": 0.49006423354148865,
      "rewards-max/forensics_tiou_reward": 0.962230384349823,
      "rewards/forensics_sinkhorn_reward": 0.6826375424861908,
      "rewards/forensics_strict_edge_reward": 0.030365245416760445,
      "rewards/forensics_tiou_reward": 0.5271490216255188,
      "step": 567
    },
    {
      "completion_length": 26.515625,
      "epoch": 1.7434715821812596,
      "grad_norm": 8.25594425201416,
      "kl": 0.7236328125,
      "learning_rate": 5.638461538461539e-07,
      "loss": -0.0959,
      "reward": 1.2556201815605164,
      "reward_std": 0.38297608494758606,
      "reward_std_all": 0.9568455815315247,
      "rewards-max/forensics_sinkhorn_reward": 0.9762970805168152,
      "rewards-max/forensics_strict_edge_reward": 0.359408937394619,
      "rewards-max/forensics_tiou_reward": 0.9601904451847076,
      "rewards/forensics_sinkhorn_reward": 0.7099114060401917,
      "rewards/forensics_strict_edge_reward": 0.034910306334495544,
      "rewards/forensics_tiou_reward": 0.5107984393835068,
      "step": 568
    },
    {
      "completion_length": 33.046875,
      "epoch": 1.7465437788018434,
      "grad_norm": 5.917353630065918,
      "kl": 0.33203125,
      "learning_rate": 5.63076923076923e-07,
      "loss": -0.0892,
      "reward": 0.9433115124702454,
      "reward_std": 0.30698157846927643,
      "reward_std_all": 1.0829196572303772,
      "rewards-max/forensics_sinkhorn_reward": 0.9749549627304077,
      "rewards-max/forensics_strict_edge_reward": 0.3312747925519943,
      "rewards-max/forensics_tiou_reward": 0.9614629745483398,
      "rewards/forensics_sinkhorn_reward": 0.5505341291427612,
      "rewards/forensics_strict_edge_reward": 0.02147734258323908,
      "rewards/forensics_tiou_reward": 0.37130001187324524,
      "step": 569
    },
    {
      "completion_length": 28.03125,
      "epoch": 1.7496159754224272,
      "grad_norm": 5.2476301193237305,
      "kl": 0.279296875,
      "learning_rate": 5.623076923076923e-07,
      "loss": -0.0812,
      "reward": 1.3001266121864319,
      "reward_std": 0.2876586616039276,
      "reward_std_all": 0.9091318249702454,
      "rewards-max/forensics_sinkhorn_reward": 0.9759174287319183,
      "rewards-max/forensics_strict_edge_reward": 0.5984586775302887,
      "rewards-max/forensics_tiou_reward": 0.9628211557865143,
      "rewards/forensics_sinkhorn_reward": 0.7008219361305237,
      "rewards/forensics_strict_edge_reward": 0.06695988029241562,
      "rewards/forensics_tiou_reward": 0.5323447585105896,
      "step": 570
    },
    {
      "completion_length": 26.46875,
      "epoch": 1.7526881720430108,
      "grad_norm": 5.68591833114624,
      "kl": 0.4072265625,
      "learning_rate": 5.615384615384615e-07,
      "loss": -0.0978,
      "reward": 1.3248764276504517,
      "reward_std": 0.2744532451033592,
      "reward_std_all": 0.8895182907581329,
      "rewards-max/forensics_sinkhorn_reward": 0.9802449643611908,
      "rewards-max/forensics_strict_edge_reward": 0.3732265755534172,
      "rewards-max/forensics_tiou_reward": 0.9770586788654327,
      "rewards/forensics_sinkhorn_reward": 0.7218732237815857,
      "rewards/forensics_strict_edge_reward": 0.03800721745938063,
      "rewards/forensics_tiou_reward": 0.5649960041046143,
      "step": 571
    },
    {
      "completion_length": 28.375,
      "epoch": 1.7557603686635943,
      "grad_norm": 6.5691819190979,
      "kl": 0.4833984375,
      "learning_rate": 5.607692307692308e-07,
      "loss": -0.1833,
      "reward": 1.081031084060669,
      "reward_std": 0.3523138165473938,
      "reward_std_all": 1.0575987100601196,
      "rewards-max/forensics_sinkhorn_reward": 0.9867229759693146,
      "rewards-max/forensics_strict_edge_reward": 0.8465203046798706,
      "rewards-max/forensics_tiou_reward": 0.9938656687736511,
      "rewards/forensics_sinkhorn_reward": 0.5886672735214233,
      "rewards/forensics_strict_edge_reward": 0.07699928060173988,
      "rewards/forensics_tiou_reward": 0.41536447405815125,
      "step": 572
    },
    {
      "completion_length": 27.734375,
      "epoch": 1.758832565284178,
      "grad_norm": 5.906762599945068,
      "kl": 0.3115234375,
      "learning_rate": 5.6e-07,
      "loss": -0.1536,
      "reward": 1.277132511138916,
      "reward_std": 0.2774088680744171,
      "reward_std_all": 0.9125400483608246,
      "rewards-max/forensics_sinkhorn_reward": 0.9627264738082886,
      "rewards-max/forensics_strict_edge_reward": 0.5064183175563812,
      "rewards-max/forensics_tiou_reward": 0.9335745573043823,
      "rewards/forensics_sinkhorn_reward": 0.6872278153896332,
      "rewards/forensics_strict_edge_reward": 0.04558498412370682,
      "rewards/forensics_tiou_reward": 0.5443196892738342,
      "step": 573
    },
    {
      "completion_length": 25.84375,
      "epoch": 1.7619047619047619,
      "grad_norm": 5.915266036987305,
      "kl": 0.375,
      "learning_rate": 5.592307692307691e-07,
      "loss": -0.0452,
      "reward": 1.0676720142364502,
      "reward_std": 0.3598517030477524,
      "reward_std_all": 1.0256676077842712,
      "rewards-max/forensics_sinkhorn_reward": 0.9584014713764191,
      "rewards-max/forensics_strict_edge_reward": 0.3779352605342865,
      "rewards-max/forensics_tiou_reward": 0.9270402491092682,
      "rewards/forensics_sinkhorn_reward": 0.6201643347740173,
      "rewards/forensics_strict_edge_reward": 0.02568572387099266,
      "rewards/forensics_tiou_reward": 0.4218219220638275,
      "step": 574
    },
    {
      "completion_length": 27.71875,
      "epoch": 1.7649769585253456,
      "grad_norm": 5.431387424468994,
      "kl": 0.3271484375,
      "learning_rate": 5.584615384615385e-07,
      "loss": -0.1412,
      "reward": 1.302925169467926,
      "reward_std": 0.28713977336883545,
      "reward_std_all": 0.9054451584815979,
      "rewards-max/forensics_sinkhorn_reward": 0.9695973694324493,
      "rewards-max/forensics_strict_edge_reward": 0.5579498410224915,
      "rewards-max/forensics_tiou_reward": 0.9578326046466827,
      "rewards/forensics_sinkhorn_reward": 0.7024468779563904,
      "rewards/forensics_strict_edge_reward": 0.05836310051381588,
      "rewards/forensics_tiou_reward": 0.5421152114868164,
      "step": 575
    },
    {
      "completion_length": 24.265625,
      "epoch": 1.7680491551459294,
      "grad_norm": 5.558907508850098,
      "kl": 0.3681640625,
      "learning_rate": 5.576923076923077e-07,
      "loss": -0.1392,
      "reward": 1.213454246520996,
      "reward_std": 0.29969364404678345,
      "reward_std_all": 0.9531927704811096,
      "rewards-max/forensics_sinkhorn_reward": 0.9719736874103546,
      "rewards-max/forensics_strict_edge_reward": 0.4099443256855011,
      "rewards-max/forensics_tiou_reward": 0.9538725316524506,
      "rewards/forensics_sinkhorn_reward": 0.6861618161201477,
      "rewards/forensics_strict_edge_reward": 0.04286195756867528,
      "rewards/forensics_tiou_reward": 0.4844304323196411,
      "step": 576
    },
    {
      "completion_length": 24.453125,
      "epoch": 1.7711213517665132,
      "grad_norm": 6.980266571044922,
      "kl": 0.3505859375,
      "learning_rate": 5.569230769230769e-07,
      "loss": -0.1343,
      "reward": 1.0197463035583496,
      "reward_std": 0.30013276636600494,
      "reward_std_all": 1.04666268825531,
      "rewards-max/forensics_sinkhorn_reward": 0.9714508652687073,
      "rewards-max/forensics_strict_edge_reward": 0.33664843440055847,
      "rewards-max/forensics_tiou_reward": 0.955091267824173,
      "rewards/forensics_sinkhorn_reward": 0.5887677669525146,
      "rewards/forensics_strict_edge_reward": 0.030177103355526924,
      "rewards/forensics_tiou_reward": 0.40080155432224274,
      "step": 577
    },
    {
      "completion_length": 25.71875,
      "epoch": 1.7741935483870968,
      "grad_norm": 5.743545055389404,
      "kl": 0.4140625,
      "learning_rate": 5.561538461538461e-07,
      "loss": -0.0799,
      "reward": 1.1885274648666382,
      "reward_std": 0.3166142851114273,
      "reward_std_all": 0.9703895747661591,
      "rewards-max/forensics_sinkhorn_reward": 0.9726101458072662,
      "rewards-max/forensics_strict_edge_reward": 0.5100106298923492,
      "rewards-max/forensics_tiou_reward": 0.9566938579082489,
      "rewards/forensics_sinkhorn_reward": 0.6722377240657806,
      "rewards/forensics_strict_edge_reward": 0.046557892113924026,
      "rewards/forensics_tiou_reward": 0.46973174810409546,
      "step": 578
    },
    {
      "completion_length": 23.484375,
      "epoch": 1.7772657450076805,
      "grad_norm": 5.839752197265625,
      "kl": 0.3427734375,
      "learning_rate": 5.553846153846153e-07,
      "loss": -0.1754,
      "reward": 1.120245337486267,
      "reward_std": 0.2846360206604004,
      "reward_std_all": 1.0007692575454712,
      "rewards-max/forensics_sinkhorn_reward": 0.9571633636951447,
      "rewards-max/forensics_strict_edge_reward": 0.5089184939861298,
      "rewards-max/forensics_tiou_reward": 0.9275119304656982,
      "rewards/forensics_sinkhorn_reward": 0.6638640463352203,
      "rewards/forensics_strict_edge_reward": 0.042927260510623455,
      "rewards/forensics_tiou_reward": 0.41345401108264923,
      "step": 579
    },
    {
      "completion_length": 27.796875,
      "epoch": 1.780337941628264,
      "grad_norm": 6.283278465270996,
      "kl": 0.3525390625,
      "learning_rate": 5.546153846153846e-07,
      "loss": -0.1343,
      "reward": 1.2154630422592163,
      "reward_std": 0.4047578275203705,
      "reward_std_all": 0.9924466609954834,
      "rewards-max/forensics_sinkhorn_reward": 0.9858105778694153,
      "rewards-max/forensics_strict_edge_reward": 0.5965746641159058,
      "rewards-max/forensics_tiou_reward": 0.9911240339279175,
      "rewards/forensics_sinkhorn_reward": 0.6558010578155518,
      "rewards/forensics_strict_edge_reward": 0.07816777750849724,
      "rewards/forensics_tiou_reward": 0.48149412870407104,
      "step": 580
    },
    {
      "completion_length": 21.90625,
      "epoch": 1.7834101382488479,
      "grad_norm": 10.571197509765625,
      "kl": 0.4296875,
      "learning_rate": 5.538461538461539e-07,
      "loss": -0.124,
      "reward": 1.1568084955215454,
      "reward_std": 0.2295239418745041,
      "reward_std_all": 0.9586905837059021,
      "rewards-max/forensics_sinkhorn_reward": 0.9769107401371002,
      "rewards-max/forensics_strict_edge_reward": 0.469832107424736,
      "rewards-max/forensics_tiou_reward": 0.9703585803508759,
      "rewards/forensics_sinkhorn_reward": 0.6715533137321472,
      "rewards/forensics_strict_edge_reward": 0.031026361044496298,
      "rewards/forensics_tiou_reward": 0.45422878861427307,
      "step": 581
    },
    {
      "completion_length": 26.96875,
      "epoch": 1.7864823348694316,
      "grad_norm": 5.91635799407959,
      "kl": 0.4306640625,
      "learning_rate": 5.53076923076923e-07,
      "loss": -0.1012,
      "reward": 1.2031298875808716,
      "reward_std": 0.4421883523464203,
      "reward_std_all": 0.9941230416297913,
      "rewards-max/forensics_sinkhorn_reward": 0.9679594337940216,
      "rewards-max/forensics_strict_edge_reward": 0.44714050740003586,
      "rewards-max/forensics_tiou_reward": 0.944961667060852,
      "rewards/forensics_sinkhorn_reward": 0.6738175749778748,
      "rewards/forensics_strict_edge_reward": 0.03777779499068856,
      "rewards/forensics_tiou_reward": 0.4915345460176468,
      "step": 582
    },
    {
      "completion_length": 25.9375,
      "epoch": 1.7895545314900154,
      "grad_norm": 8.654838562011719,
      "kl": 0.46484375,
      "learning_rate": 5.523076923076923e-07,
      "loss": -0.1502,
      "reward": 1.1211100816726685,
      "reward_std": 0.29857584834098816,
      "reward_std_all": 0.9921431541442871,
      "rewards-max/forensics_sinkhorn_reward": 0.9580983519554138,
      "rewards-max/forensics_strict_edge_reward": 0.37120091915130615,
      "rewards-max/forensics_tiou_reward": 0.9267309904098511,
      "rewards/forensics_sinkhorn_reward": 0.6469597518444061,
      "rewards/forensics_strict_edge_reward": 0.029699556529521942,
      "rewards/forensics_tiou_reward": 0.444450780749321,
      "step": 583
    },
    {
      "completion_length": 27.265625,
      "epoch": 1.7926267281105992,
      "grad_norm": 6.187768459320068,
      "kl": 0.521484375,
      "learning_rate": 5.515384615384615e-07,
      "loss": -0.0599,
      "reward": 1.2393938302993774,
      "reward_std": 0.18384479731321335,
      "reward_std_all": 0.9042117595672607,
      "rewards-max/forensics_sinkhorn_reward": 0.9590089023113251,
      "rewards-max/forensics_strict_edge_reward": 0.33071269094944,
      "rewards-max/forensics_tiou_reward": 0.926531046628952,
      "rewards/forensics_sinkhorn_reward": 0.6889119148254395,
      "rewards/forensics_strict_edge_reward": 0.01821128837764263,
      "rewards/forensics_tiou_reward": 0.532270610332489,
      "step": 584
    },
    {
      "completion_length": 23.59375,
      "epoch": 1.7956989247311828,
      "grad_norm": 6.998341083526611,
      "kl": 0.427734375,
      "learning_rate": 5.507692307692308e-07,
      "loss": -0.1144,
      "reward": 1.059550404548645,
      "reward_std": 0.23918800801038742,
      "reward_std_all": 1.0015423893928528,
      "rewards-max/forensics_sinkhorn_reward": 0.9373658895492554,
      "rewards-max/forensics_strict_edge_reward": 0.15698032081127167,
      "rewards-max/forensics_tiou_reward": 0.8597742021083832,
      "rewards/forensics_sinkhorn_reward": 0.6284967362880707,
      "rewards/forensics_strict_edge_reward": 0.009540995466522872,
      "rewards/forensics_tiou_reward": 0.421512708067894,
      "step": 585
    },
    {
      "completion_length": 23.078125,
      "epoch": 1.7987711213517665,
      "grad_norm": 6.823706150054932,
      "kl": 0.337890625,
      "learning_rate": 5.5e-07,
      "loss": -0.1057,
      "reward": 1.0913066267967224,
      "reward_std": 0.2563057020306587,
      "reward_std_all": 0.9892896413803101,
      "rewards-max/forensics_sinkhorn_reward": 0.9794514179229736,
      "rewards-max/forensics_strict_edge_reward": 0.2574504744261503,
      "rewards-max/forensics_tiou_reward": 0.9708624482154846,
      "rewards/forensics_sinkhorn_reward": 0.6393900513648987,
      "rewards/forensics_strict_edge_reward": 0.014647835167124867,
      "rewards/forensics_tiou_reward": 0.43726876378059387,
      "step": 586
    },
    {
      "completion_length": 22.65625,
      "epoch": 1.80184331797235,
      "grad_norm": 7.658453464508057,
      "kl": 0.39453125,
      "learning_rate": 5.492307692307691e-07,
      "loss": -0.0288,
      "reward": 1.15943443775177,
      "reward_std": 0.2660549134016037,
      "reward_std_all": 0.9603774547576904,
      "rewards-max/forensics_sinkhorn_reward": 0.9691976308822632,
      "rewards-max/forensics_strict_edge_reward": 0.42371489852666855,
      "rewards-max/forensics_tiou_reward": 0.9464199542999268,
      "rewards/forensics_sinkhorn_reward": 0.6757884919643402,
      "rewards/forensics_strict_edge_reward": 0.022597257047891617,
      "rewards/forensics_tiou_reward": 0.4610486179590225,
      "step": 587
    },
    {
      "completion_length": 27.6875,
      "epoch": 1.8049155145929339,
      "grad_norm": 6.86970329284668,
      "kl": 0.4111328125,
      "learning_rate": 5.484615384615385e-07,
      "loss": -0.0294,
      "reward": 1.2091909646987915,
      "reward_std": 0.3048274517059326,
      "reward_std_all": 0.9572744071483612,
      "rewards-max/forensics_sinkhorn_reward": 0.977707028388977,
      "rewards-max/forensics_strict_edge_reward": 0.716116189956665,
      "rewards-max/forensics_tiou_reward": 0.9730053246021271,
      "rewards/forensics_sinkhorn_reward": 0.6603765785694122,
      "rewards/forensics_strict_edge_reward": 0.04360014945268631,
      "rewards/forensics_tiou_reward": 0.505214273929596,
      "step": 588
    },
    {
      "completion_length": 23.6875,
      "epoch": 1.8079877112135176,
      "grad_norm": 6.520449638366699,
      "kl": 0.572265625,
      "learning_rate": 5.476923076923077e-07,
      "loss": -0.1363,
      "reward": 1.2340351939201355,
      "reward_std": 0.3103528618812561,
      "reward_std_all": 0.9531566500663757,
      "rewards-max/forensics_sinkhorn_reward": 0.9520136415958405,
      "rewards-max/forensics_strict_edge_reward": 0.4604835510253906,
      "rewards-max/forensics_tiou_reward": 0.9149430394172668,
      "rewards/forensics_sinkhorn_reward": 0.6980190873146057,
      "rewards/forensics_strict_edge_reward": 0.044591122306883335,
      "rewards/forensics_tiou_reward": 0.4914250671863556,
      "step": 589
    },
    {
      "completion_length": 25.578125,
      "epoch": 1.8110599078341014,
      "grad_norm": 5.7312140464782715,
      "kl": 0.39453125,
      "learning_rate": 5.469230769230769e-07,
      "loss": -0.097,
      "reward": 1.423300564289093,
      "reward_std": 0.2816171497106552,
      "reward_std_all": 0.8605496883392334,
      "rewards-max/forensics_sinkhorn_reward": 0.9857997894287109,
      "rewards-max/forensics_strict_edge_reward": 0.835382878780365,
      "rewards-max/forensics_tiou_reward": 0.993116945028305,
      "rewards/forensics_sinkhorn_reward": 0.7468345463275909,
      "rewards/forensics_strict_edge_reward": 0.0755735170096159,
      "rewards/forensics_tiou_reward": 0.6008925139904022,
      "step": 590
    },
    {
      "completion_length": 25.171875,
      "epoch": 1.8141321044546852,
      "grad_norm": 7.507875442504883,
      "kl": 0.517578125,
      "learning_rate": 5.461538461538461e-07,
      "loss": -0.1407,
      "reward": 1.0845720767974854,
      "reward_std": 0.3167651444673538,
      "reward_std_all": 1.0236309468746185,
      "rewards-max/forensics_sinkhorn_reward": 0.96717369556427,
      "rewards-max/forensics_strict_edge_reward": 0.3893759995698929,
      "rewards-max/forensics_tiou_reward": 0.9464695155620575,
      "rewards/forensics_sinkhorn_reward": 0.610198974609375,
      "rewards/forensics_strict_edge_reward": 0.026863092556595802,
      "rewards/forensics_tiou_reward": 0.4475100040435791,
      "step": 591
    },
    {
      "completion_length": 26.265625,
      "epoch": 1.817204301075269,
      "grad_norm": 7.087900638580322,
      "kl": 0.4365234375,
      "learning_rate": 5.453846153846153e-07,
      "loss": -0.1519,
      "reward": 1.34157794713974,
      "reward_std": 0.3925427794456482,
      "reward_std_all": 0.9248104095458984,
      "rewards-max/forensics_sinkhorn_reward": 0.9788861870765686,
      "rewards-max/forensics_strict_edge_reward": 0.613072544336319,
      "rewards-max/forensics_tiou_reward": 0.9740468263626099,
      "rewards/forensics_sinkhorn_reward": 0.6769096851348877,
      "rewards/forensics_strict_edge_reward": 0.0814583832398057,
      "rewards/forensics_tiou_reward": 0.5832099616527557,
      "step": 592
    },
    {
      "completion_length": 26.140625,
      "epoch": 1.8202764976958525,
      "grad_norm": 9.80290699005127,
      "kl": 0.447265625,
      "learning_rate": 5.446153846153846e-07,
      "loss": -0.1113,
      "reward": 1.1058666408061981,
      "reward_std": 0.34833763539791107,
      "reward_std_all": 1.0241917073726654,
      "rewards-max/forensics_sinkhorn_reward": 0.9714553952217102,
      "rewards-max/forensics_strict_edge_reward": 0.5826494470238686,
      "rewards-max/forensics_tiou_reward": 0.9656017124652863,
      "rewards/forensics_sinkhorn_reward": 0.6091729402542114,
      "rewards/forensics_strict_edge_reward": 0.06546937627717853,
      "rewards/forensics_tiou_reward": 0.4312243163585663,
      "step": 593
    },
    {
      "completion_length": 24.375,
      "epoch": 1.823348694316436,
      "grad_norm": 7.4673943519592285,
      "kl": 0.587890625,
      "learning_rate": 5.438461538461538e-07,
      "loss": -0.0874,
      "reward": 1.2220518589019775,
      "reward_std": 0.19849933311343193,
      "reward_std_all": 0.9237395524978638,
      "rewards-max/forensics_sinkhorn_reward": 0.9797850251197815,
      "rewards-max/forensics_strict_edge_reward": 0.3809451572597027,
      "rewards-max/forensics_tiou_reward": 0.9743312299251556,
      "rewards/forensics_sinkhorn_reward": 0.6830215454101562,
      "rewards/forensics_strict_edge_reward": 0.049464610405266285,
      "rewards/forensics_tiou_reward": 0.4895656555891037,
      "step": 594
    },
    {
      "completion_length": 27.4375,
      "epoch": 1.8264208909370199,
      "grad_norm": 6.740311145782471,
      "kl": 0.48828125,
      "learning_rate": 5.430769230769231e-07,
      "loss": -0.0681,
      "reward": 1.1597123742103577,
      "reward_std": 0.2572580426931381,
      "reward_std_all": 0.9738678932189941,
      "rewards-max/forensics_sinkhorn_reward": 0.9692011773586273,
      "rewards-max/forensics_strict_edge_reward": 0.5107879042625427,
      "rewards-max/forensics_tiou_reward": 0.954477995634079,
      "rewards/forensics_sinkhorn_reward": 0.6673821210861206,
      "rewards/forensics_strict_edge_reward": 0.03373275324702263,
      "rewards/forensics_tiou_reward": 0.45859748125076294,
      "step": 595
    },
    {
      "completion_length": 22.8125,
      "epoch": 1.8294930875576036,
      "grad_norm": 7.480429649353027,
      "kl": 0.3681640625,
      "learning_rate": 5.423076923076923e-07,
      "loss": -0.1036,
      "reward": 0.9574683904647827,
      "reward_std": 0.2665209323167801,
      "reward_std_all": 1.0676513314247131,
      "rewards-max/forensics_sinkhorn_reward": 0.9703827798366547,
      "rewards-max/forensics_strict_edge_reward": 0.51341712474823,
      "rewards-max/forensics_tiou_reward": 0.9571358859539032,
      "rewards/forensics_sinkhorn_reward": 0.5610454678535461,
      "rewards/forensics_strict_edge_reward": 0.02958158403635025,
      "rewards/forensics_tiou_reward": 0.3668413609266281,
      "step": 596
    },
    {
      "completion_length": 23.828125,
      "epoch": 1.8325652841781874,
      "grad_norm": 18.9442195892334,
      "kl": 0.7998046875,
      "learning_rate": 5.415384615384615e-07,
      "loss": -0.0577,
      "reward": 1.126282811164856,
      "reward_std": 0.4767339825630188,
      "reward_std_all": 1.0438692569732666,
      "rewards-max/forensics_sinkhorn_reward": 0.9676951169967651,
      "rewards-max/forensics_strict_edge_reward": 0.4424546957015991,
      "rewards-max/forensics_tiou_reward": 0.9414475858211517,
      "rewards/forensics_sinkhorn_reward": 0.6400291323661804,
      "rewards/forensics_strict_edge_reward": 0.04078393802046776,
      "rewards/forensics_tiou_reward": 0.4454697072505951,
      "step": 597
    },
    {
      "completion_length": 21.328125,
      "epoch": 1.8356374807987712,
      "grad_norm": 5.715350151062012,
      "kl": 0.33203125,
      "learning_rate": 5.407692307692308e-07,
      "loss": -0.116,
      "reward": 1.2277752161026,
      "reward_std": 0.3922811299562454,
      "reward_std_all": 0.9821025729179382,
      "rewards-max/forensics_sinkhorn_reward": 0.973516583442688,
      "rewards-max/forensics_strict_edge_reward": 0.51341712474823,
      "rewards-max/forensics_tiou_reward": 0.963881105184555,
      "rewards/forensics_sinkhorn_reward": 0.6830236613750458,
      "rewards/forensics_strict_edge_reward": 0.045161642134189606,
      "rewards/forensics_tiou_reward": 0.49958983063697815,
      "step": 598
    },
    {
      "completion_length": 26.359375,
      "epoch": 1.838709677419355,
      "grad_norm": 6.495640277862549,
      "kl": 0.3701171875,
      "learning_rate": 5.4e-07,
      "loss": -0.1052,
      "reward": 1.0586601197719574,
      "reward_std": 0.4247974157333374,
      "reward_std_all": 1.0791733860969543,
      "rewards-max/forensics_sinkhorn_reward": 0.9626563787460327,
      "rewards-max/forensics_strict_edge_reward": 0.6040194034576416,
      "rewards-max/forensics_tiou_reward": 0.9395579993724823,
      "rewards/forensics_sinkhorn_reward": 0.6007892489433289,
      "rewards/forensics_strict_edge_reward": 0.05464468291029334,
      "rewards/forensics_tiou_reward": 0.4032263159751892,
      "step": 599
    },
    {
      "completion_length": 26.609375,
      "epoch": 1.8417818740399385,
      "grad_norm": 6.8075947761535645,
      "kl": 0.3662109375,
      "learning_rate": 5.392307692307691e-07,
      "loss": -0.0801,
      "reward": 1.0954643487930298,
      "reward_std": 0.3277808576822281,
      "reward_std_all": 1.0142972469329834,
      "rewards-max/forensics_sinkhorn_reward": 0.9765402972698212,
      "rewards-max/forensics_strict_edge_reward": 0.2831524144858122,
      "rewards-max/forensics_tiou_reward": 0.9730018377304077,
      "rewards/forensics_sinkhorn_reward": 0.6228675544261932,
      "rewards/forensics_strict_edge_reward": 0.04180975677445531,
      "rewards/forensics_tiou_reward": 0.4307870864868164,
      "step": 600
    },
    {
      "completion_length": 22.796875,
      "epoch": 1.8448540706605223,
      "grad_norm": 6.120138168334961,
      "kl": 0.3603515625,
      "learning_rate": 5.384615384615384e-07,
      "loss": -0.1499,
      "reward": 1.2110437154769897,
      "reward_std": 0.5539613962173462,
      "reward_std_all": 1.0255616009235382,
      "rewards-max/forensics_sinkhorn_reward": 0.9833499789237976,
      "rewards-max/forensics_strict_edge_reward": 0.6503761410713196,
      "rewards-max/forensics_tiou_reward": 0.9825429618358612,
      "rewards/forensics_sinkhorn_reward": 0.6634068787097931,
      "rewards/forensics_strict_edge_reward": 0.06869168020784855,
      "rewards/forensics_tiou_reward": 0.47894513607025146,
      "step": 601
    },
    {
      "completion_length": 24.609375,
      "epoch": 1.8479262672811059,
      "grad_norm": 6.290087699890137,
      "kl": 0.3955078125,
      "learning_rate": 5.376923076923077e-07,
      "loss": -0.0776,
      "reward": 1.3485766649246216,
      "reward_std": 0.23148836195468903,
      "reward_std_all": 0.8659927248954773,
      "rewards-max/forensics_sinkhorn_reward": 0.9827918112277985,
      "rewards-max/forensics_strict_edge_reward": 0.5581351518630981,
      "rewards-max/forensics_tiou_reward": 0.9811595380306244,
      "rewards/forensics_sinkhorn_reward": 0.7520215511322021,
      "rewards/forensics_strict_edge_reward": 0.03488667495548725,
      "rewards/forensics_tiou_reward": 0.5616684257984161,
      "step": 602
    },
    {
      "completion_length": 24.078125,
      "epoch": 1.8509984639016897,
      "grad_norm": 6.110663890838623,
      "kl": 0.3701171875,
      "learning_rate": 5.369230769230769e-07,
      "loss": -0.1236,
      "reward": 1.1305719017982483,
      "reward_std": 0.32841427624225616,
      "reward_std_all": 1.0048880875110626,
      "rewards-max/forensics_sinkhorn_reward": 0.9828491508960724,
      "rewards-max/forensics_strict_edge_reward": 0.756708562374115,
      "rewards-max/forensics_tiou_reward": 0.9848736822605133,
      "rewards/forensics_sinkhorn_reward": 0.654967188835144,
      "rewards/forensics_strict_edge_reward": 0.05626790598034859,
      "rewards/forensics_tiou_reward": 0.41933678090572357,
      "step": 603
    },
    {
      "completion_length": 24.640625,
      "epoch": 1.8540706605222734,
      "grad_norm": 6.113715171813965,
      "kl": 0.43359375,
      "learning_rate": 5.361538461538462e-07,
      "loss": -0.0038,
      "reward": 1.2917611002922058,
      "reward_std": 0.3797365874052048,
      "reward_std_all": 0.9289953708648682,
      "rewards-max/forensics_sinkhorn_reward": 0.9769232273101807,
      "rewards-max/forensics_strict_edge_reward": 0.4462650418281555,
      "rewards-max/forensics_tiou_reward": 0.9650183320045471,
      "rewards/forensics_sinkhorn_reward": 0.719683825969696,
      "rewards/forensics_strict_edge_reward": 0.03384610824286938,
      "rewards/forensics_tiou_reward": 0.53823122382164,
      "step": 604
    },
    {
      "completion_length": 22.328125,
      "epoch": 1.8571428571428572,
      "grad_norm": 5.803893566131592,
      "kl": 0.447265625,
      "learning_rate": 5.353846153846153e-07,
      "loss": -0.0684,
      "reward": 1.2095761895179749,
      "reward_std": 0.3018147200345993,
      "reward_std_all": 0.9545870423316956,
      "rewards-max/forensics_sinkhorn_reward": 0.9731651246547699,
      "rewards-max/forensics_strict_edge_reward": 0.5050012171268463,
      "rewards-max/forensics_tiou_reward": 0.9612603783607483,
      "rewards/forensics_sinkhorn_reward": 0.6994588971138,
      "rewards/forensics_strict_edge_reward": 0.03194513916969299,
      "rewards/forensics_tiou_reward": 0.478172242641449,
      "step": 605
    },
    {
      "completion_length": 26.703125,
      "epoch": 1.860215053763441,
      "grad_norm": 7.254394054412842,
      "kl": 0.32421875,
      "learning_rate": 5.346153846153846e-07,
      "loss": -0.1411,
      "reward": 1.0722513794898987,
      "reward_std": 0.27442028373479843,
      "reward_std_all": 1.0212929844856262,
      "rewards-max/forensics_sinkhorn_reward": 0.9839516878128052,
      "rewards-max/forensics_strict_edge_reward": 0.48913954198360443,
      "rewards-max/forensics_tiou_reward": 0.9870983958244324,
      "rewards/forensics_sinkhorn_reward": 0.6147019267082214,
      "rewards/forensics_strict_edge_reward": 0.04189242608845234,
      "rewards/forensics_tiou_reward": 0.41565701365470886,
      "step": 606
    },
    {
      "completion_length": 25.125,
      "epoch": 1.8632872503840245,
      "grad_norm": 7.564494609832764,
      "kl": 0.31689453125,
      "learning_rate": 5.338461538461538e-07,
      "loss": -0.1116,
      "reward": 1.3109055757522583,
      "reward_std": 0.29010628163814545,
      "reward_std_all": 0.9035955965518951,
      "rewards-max/forensics_sinkhorn_reward": 0.9752289950847626,
      "rewards-max/forensics_strict_edge_reward": 0.304876372218132,
      "rewards-max/forensics_tiou_reward": 0.9518454372882843,
      "rewards/forensics_sinkhorn_reward": 0.7306505143642426,
      "rewards/forensics_strict_edge_reward": 0.05428261938504875,
      "rewards/forensics_tiou_reward": 0.525972381234169,
      "step": 607
    },
    {
      "completion_length": 25.140625,
      "epoch": 1.8663594470046083,
      "grad_norm": 6.644888401031494,
      "kl": 0.40234375,
      "learning_rate": 5.33076923076923e-07,
      "loss": -0.0754,
      "reward": 1.382542371749878,
      "reward_std": 0.33590687811374664,
      "reward_std_all": 0.8836407661437988,
      "rewards-max/forensics_sinkhorn_reward": 0.9871773421764374,
      "rewards-max/forensics_strict_edge_reward": 0.776590883731842,
      "rewards-max/forensics_tiou_reward": 0.9946021139621735,
      "rewards/forensics_sinkhorn_reward": 0.7110548615455627,
      "rewards/forensics_strict_edge_reward": 0.0826716236770153,
      "rewards/forensics_tiou_reward": 0.5888160169124603,
      "step": 608
    },
    {
      "completion_length": 27.25,
      "epoch": 1.8694316436251919,
      "grad_norm": 7.014779567718506,
      "kl": 0.3603515625,
      "learning_rate": 5.323076923076923e-07,
      "loss": -0.0603,
      "reward": 1.1482455134391785,
      "reward_std": 0.2538567781448364,
      "reward_std_all": 0.9583121538162231,
      "rewards-max/forensics_sinkhorn_reward": 0.9569645524024963,
      "rewards-max/forensics_strict_edge_reward": 0.21873006224632263,
      "rewards-max/forensics_tiou_reward": 0.9195154309272766,
      "rewards/forensics_sinkhorn_reward": 0.6480141580104828,
      "rewards/forensics_strict_edge_reward": 0.021385583095252514,
      "rewards/forensics_tiou_reward": 0.4788457453250885,
      "step": 609
    },
    {
      "completion_length": 24.59375,
      "epoch": 1.8725038402457757,
      "grad_norm": 7.178742408752441,
      "kl": 0.4609375,
      "learning_rate": 5.315384615384615e-07,
      "loss": -0.0856,
      "reward": 1.1085216403007507,
      "reward_std": 0.27721698582172394,
      "reward_std_all": 0.9929401278495789,
      "rewards-max/forensics_sinkhorn_reward": 0.9688286483287811,
      "rewards-max/forensics_strict_edge_reward": 0.5294047147035599,
      "rewards-max/forensics_tiou_reward": 0.9469747543334961,
      "rewards/forensics_sinkhorn_reward": 0.6516738831996918,
      "rewards/forensics_strict_edge_reward": 0.02635860349982977,
      "rewards/forensics_tiou_reward": 0.4304891526699066,
      "step": 610
    },
    {
      "completion_length": 22.140625,
      "epoch": 1.8755760368663594,
      "grad_norm": 6.881528854370117,
      "kl": 0.384765625,
      "learning_rate": 5.307692307692308e-07,
      "loss": -0.0741,
      "reward": 1.2690398693084717,
      "reward_std": 0.28845715522766113,
      "reward_std_all": 0.9157896637916565,
      "rewards-max/forensics_sinkhorn_reward": 0.9697778522968292,
      "rewards-max/forensics_strict_edge_reward": 0.2578420117497444,
      "rewards-max/forensics_tiou_reward": 0.9498206973075867,
      "rewards/forensics_sinkhorn_reward": 0.7162547707557678,
      "rewards/forensics_strict_edge_reward": 0.027547134086489677,
      "rewards/forensics_tiou_reward": 0.5252380520105362,
      "step": 611
    },
    {
      "completion_length": 22.421875,
      "epoch": 1.8786482334869432,
      "grad_norm": 6.413933753967285,
      "kl": 0.390625,
      "learning_rate": 5.3e-07,
      "loss": -0.1449,
      "reward": 0.9933457374572754,
      "reward_std": 0.3096945732831955,
      "reward_std_all": 1.0600839853286743,
      "rewards-max/forensics_sinkhorn_reward": 0.9718863368034363,
      "rewards-max/forensics_strict_edge_reward": 0.2485707327723503,
      "rewards-max/forensics_tiou_reward": 0.9569547772407532,
      "rewards/forensics_sinkhorn_reward": 0.5878589153289795,
      "rewards/forensics_strict_edge_reward": 0.01906943181529641,
      "rewards/forensics_tiou_reward": 0.3864174038171768,
      "step": 612
    },
    {
      "completion_length": 24.015625,
      "epoch": 1.881720430107527,
      "grad_norm": 7.1213483810424805,
      "kl": 0.49609375,
      "learning_rate": 5.292307692307692e-07,
      "loss": -0.0807,
      "reward": 1.339374601840973,
      "reward_std": 0.3152284473180771,
      "reward_std_all": 0.8927810788154602,
      "rewards-max/forensics_sinkhorn_reward": 0.9748803675174713,
      "rewards-max/forensics_strict_edge_reward": 0.4590318948030472,
      "rewards-max/forensics_tiou_reward": 0.9605498015880585,
      "rewards/forensics_sinkhorn_reward": 0.7341490387916565,
      "rewards/forensics_strict_edge_reward": 0.04086281545460224,
      "rewards/forensics_tiou_reward": 0.5643627047538757,
      "step": 613
    },
    {
      "completion_length": 26.734375,
      "epoch": 1.8847926267281108,
      "grad_norm": 10.481358528137207,
      "kl": 0.5244140625,
      "learning_rate": 5.284615384615384e-07,
      "loss": 0.0105,
      "reward": 0.8939893841743469,
      "reward_std": 0.21785176545381546,
      "reward_std_all": 1.0905548334121704,
      "rewards-max/forensics_sinkhorn_reward": 0.9794517159461975,
      "rewards-max/forensics_strict_edge_reward": 0.2873457036912441,
      "rewards-max/forensics_tiou_reward": 0.9734693467617035,
      "rewards/forensics_sinkhorn_reward": 0.5092860162258148,
      "rewards/forensics_strict_edge_reward": 0.026201992528513074,
      "rewards/forensics_tiou_reward": 0.3585013225674629,
      "step": 614
    },
    {
      "completion_length": 27.78125,
      "epoch": 1.8878648233486943,
      "grad_norm": 6.647071838378906,
      "kl": 0.3603515625,
      "learning_rate": 5.276923076923076e-07,
      "loss": -0.1435,
      "reward": 1.1157997250556946,
      "reward_std": 0.2839937210083008,
      "reward_std_all": 0.987711489200592,
      "rewards-max/forensics_sinkhorn_reward": 0.9717845320701599,
      "rewards-max/forensics_strict_edge_reward": 0.2789544239640236,
      "rewards-max/forensics_tiou_reward": 0.9596096873283386,
      "rewards/forensics_sinkhorn_reward": 0.6410315036773682,
      "rewards/forensics_strict_edge_reward": 0.020546465180814266,
      "rewards/forensics_tiou_reward": 0.45422177016735077,
      "step": 615
    },
    {
      "completion_length": 29.046875,
      "epoch": 1.8909370199692779,
      "grad_norm": 6.547812461853027,
      "kl": 0.41015625,
      "learning_rate": 5.269230769230769e-07,
      "loss": -0.0936,
      "reward": 1.375313401222229,
      "reward_std": 0.3559695929288864,
      "reward_std_all": 0.8806068897247314,
      "rewards-max/forensics_sinkhorn_reward": 0.9731094241142273,
      "rewards-max/forensics_strict_edge_reward": 0.33620673790574074,
      "rewards-max/forensics_tiou_reward": 0.9593549072742462,
      "rewards/forensics_sinkhorn_reward": 0.7223782241344452,
      "rewards/forensics_strict_edge_reward": 0.039064702577888966,
      "rewards/forensics_tiou_reward": 0.6138704121112823,
      "step": 616
    },
    {
      "completion_length": 22.453125,
      "epoch": 1.8940092165898617,
      "grad_norm": 7.644073009490967,
      "kl": 0.396484375,
      "learning_rate": 5.261538461538462e-07,
      "loss": -0.0387,
      "reward": 1.302489995956421,
      "reward_std": 0.19734830036759377,
      "reward_std_all": 0.8841122984886169,
      "rewards-max/forensics_sinkhorn_reward": 0.969040185213089,
      "rewards-max/forensics_strict_edge_reward": 0.23630925454199314,
      "rewards-max/forensics_tiou_reward": 0.950425922870636,
      "rewards/forensics_sinkhorn_reward": 0.7494936585426331,
      "rewards/forensics_strict_edge_reward": 0.02427565597463399,
      "rewards/forensics_tiou_reward": 0.5287207365036011,
      "step": 617
    },
    {
      "completion_length": 27.4375,
      "epoch": 1.8970814132104454,
      "grad_norm": 6.295856475830078,
      "kl": 0.3837890625,
      "learning_rate": 5.253846153846153e-07,
      "loss": -0.0808,
      "reward": 1.3763580322265625,
      "reward_std": 0.41958075761795044,
      "reward_std_all": 0.9174180626869202,
      "rewards-max/forensics_sinkhorn_reward": 0.9819861054420471,
      "rewards-max/forensics_strict_edge_reward": 0.7657178044319153,
      "rewards-max/forensics_tiou_reward": 0.976681649684906,
      "rewards/forensics_sinkhorn_reward": 0.7367155849933624,
      "rewards/forensics_strict_edge_reward": 0.06826266273856163,
      "rewards/forensics_tiou_reward": 0.5713796317577362,
      "step": 618
    },
    {
      "completion_length": 23.0625,
      "epoch": 1.9001536098310292,
      "grad_norm": 6.453517913818359,
      "kl": 0.416015625,
      "learning_rate": 5.246153846153846e-07,
      "loss": -0.1188,
      "reward": 1.0623133182525635,
      "reward_std": 0.3018921762704849,
      "reward_std_all": 1.0218992829322815,
      "rewards-max/forensics_sinkhorn_reward": 0.9803623855113983,
      "rewards-max/forensics_strict_edge_reward": 0.5728720277547836,
      "rewards-max/forensics_tiou_reward": 0.9783714115619659,
      "rewards/forensics_sinkhorn_reward": 0.6131852269172668,
      "rewards/forensics_strict_edge_reward": 0.03981271758675575,
      "rewards/forensics_tiou_reward": 0.4093153774738312,
      "step": 619
    },
    {
      "completion_length": 26.359375,
      "epoch": 1.903225806451613,
      "grad_norm": 5.648045539855957,
      "kl": 0.369140625,
      "learning_rate": 5.238461538461538e-07,
      "loss": -0.0686,
      "reward": 1.1350993514060974,
      "reward_std": 0.44279225170612335,
      "reward_std_all": 1.0290881991386414,
      "rewards-max/forensics_sinkhorn_reward": 0.9830005168914795,
      "rewards-max/forensics_strict_edge_reward": 0.3885071277618408,
      "rewards-max/forensics_tiou_reward": 0.9827870428562164,
      "rewards/forensics_sinkhorn_reward": 0.622382789850235,
      "rewards/forensics_strict_edge_reward": 0.04424671363085508,
      "rewards/forensics_tiou_reward": 0.46846991777420044,
      "step": 620
    },
    {
      "completion_length": 25.3125,
      "epoch": 1.9062980030721968,
      "grad_norm": 6.940499782562256,
      "kl": 0.4658203125,
      "learning_rate": 5.23076923076923e-07,
      "loss": -0.0765,
      "reward": 1.3075052499771118,
      "reward_std": 0.2428298443555832,
      "reward_std_all": 0.8926438689231873,
      "rewards-max/forensics_sinkhorn_reward": 0.9808945655822754,
      "rewards-max/forensics_strict_edge_reward": 0.5855723917484283,
      "rewards-max/forensics_tiou_reward": 0.9750840067863464,
      "rewards/forensics_sinkhorn_reward": 0.7180888056755066,
      "rewards/forensics_strict_edge_reward": 0.05712096951901913,
      "rewards/forensics_tiou_reward": 0.5322954505681992,
      "step": 621
    },
    {
      "completion_length": 25.0625,
      "epoch": 1.9093701996927803,
      "grad_norm": 8.444374084472656,
      "kl": 0.771484375,
      "learning_rate": 5.223076923076924e-07,
      "loss": -0.053,
      "reward": 1.1347014904022217,
      "reward_std": 0.17773882672190666,
      "reward_std_all": 0.9582637548446655,
      "rewards-max/forensics_sinkhorn_reward": 0.9521101713180542,
      "rewards-max/forensics_strict_edge_reward": 0.41717062797397375,
      "rewards-max/forensics_tiou_reward": 0.8967104554176331,
      "rewards/forensics_sinkhorn_reward": 0.6674686372280121,
      "rewards/forensics_strict_edge_reward": 0.02315228001680225,
      "rewards/forensics_tiou_reward": 0.44408050179481506,
      "step": 622
    },
    {
      "completion_length": 25.265625,
      "epoch": 1.912442396313364,
      "grad_norm": 7.0939483642578125,
      "kl": 0.388671875,
      "learning_rate": 5.215384615384615e-07,
      "loss": -0.1242,
      "reward": 1.3291393518447876,
      "reward_std": 0.1849275790154934,
      "reward_std_all": 0.860632598400116,
      "rewards-max/forensics_sinkhorn_reward": 0.9687615633010864,
      "rewards-max/forensics_strict_edge_reward": 0.21414172649383545,
      "rewards-max/forensics_tiou_reward": 0.9498632550239563,
      "rewards/forensics_sinkhorn_reward": 0.7520531117916107,
      "rewards/forensics_strict_edge_reward": 0.018189620226621628,
      "rewards/forensics_tiou_reward": 0.5588965713977814,
      "step": 623
    },
    {
      "completion_length": 26.9375,
      "epoch": 1.9155145929339477,
      "grad_norm": 6.731479167938232,
      "kl": 0.306640625,
      "learning_rate": 5.207692307692308e-07,
      "loss": -0.0811,
      "reward": 1.1196328997612,
      "reward_std": 0.0956527516245842,
      "reward_std_all": 0.9401088953018188,
      "rewards-max/forensics_sinkhorn_reward": 0.914144366979599,
      "rewards-max/forensics_strict_edge_reward": 0.046137843281030655,
      "rewards-max/forensics_tiou_reward": 0.7916399240493774,
      "rewards/forensics_sinkhorn_reward": 0.6576745212078094,
      "rewards/forensics_strict_edge_reward": 0.0035374534782022238,
      "rewards/forensics_tiou_reward": 0.4584209620952606,
      "step": 624
    },
    {
      "completion_length": 28.40625,
      "epoch": 1.9185867895545314,
      "grad_norm": 7.33116340637207,
      "kl": 0.3662109375,
      "learning_rate": 5.2e-07,
      "loss": -0.0673,
      "reward": 1.046980619430542,
      "reward_std": 0.2682073563337326,
      "reward_std_all": 1.021960198879242,
      "rewards-max/forensics_sinkhorn_reward": 0.9711688756942749,
      "rewards-max/forensics_strict_edge_reward": 0.37979935854673386,
      "rewards-max/forensics_tiou_reward": 0.955598920583725,
      "rewards/forensics_sinkhorn_reward": 0.5853100717067719,
      "rewards/forensics_strict_edge_reward": 0.03555223671719432,
      "rewards/forensics_tiou_reward": 0.42611831426620483,
      "step": 625
    },
    {
      "completion_length": 23.375,
      "epoch": 1.9216589861751152,
      "grad_norm": 7.628832817077637,
      "kl": 0.3603515625,
      "learning_rate": 5.192307692307692e-07,
      "loss": -0.0501,
      "reward": 1.1819279789924622,
      "reward_std": 0.22550667822360992,
      "reward_std_all": 0.9326519072055817,
      "rewards-max/forensics_sinkhorn_reward": 0.9786946475505829,
      "rewards-max/forensics_strict_edge_reward": 0.226064283400774,
      "rewards-max/forensics_tiou_reward": 0.9742232263088226,
      "rewards/forensics_sinkhorn_reward": 0.7042449712753296,
      "rewards/forensics_strict_edge_reward": 0.017082540085539222,
      "rewards/forensics_tiou_reward": 0.46060046553611755,
      "step": 626
    },
    {
      "completion_length": 24.859375,
      "epoch": 1.924731182795699,
      "grad_norm": 6.873016834259033,
      "kl": 0.4150390625,
      "learning_rate": 5.184615384615384e-07,
      "loss": -0.1363,
      "reward": 1.335483729839325,
      "reward_std": 0.2639290988445282,
      "reward_std_all": 0.8829630017280579,
      "rewards-max/forensics_sinkhorn_reward": 0.9843825697898865,
      "rewards-max/forensics_strict_edge_reward": 0.3364075720310211,
      "rewards-max/forensics_tiou_reward": 0.9883401095867157,
      "rewards/forensics_sinkhorn_reward": 0.7188040316104889,
      "rewards/forensics_strict_edge_reward": 0.0327409403398633,
      "rewards/forensics_tiou_reward": 0.5839387476444244,
      "step": 627
    },
    {
      "completion_length": 25.359375,
      "epoch": 1.9278033794162828,
      "grad_norm": 14.466730117797852,
      "kl": 1.162109375,
      "learning_rate": 5.176923076923076e-07,
      "loss": -0.0669,
      "reward": 1.2733812928199768,
      "reward_std": 0.3488917946815491,
      "reward_std_all": 0.9364998638629913,
      "rewards-max/forensics_sinkhorn_reward": 0.9811844229698181,
      "rewards-max/forensics_strict_edge_reward": 0.5421086549758911,
      "rewards-max/forensics_tiou_reward": 0.9798877239227295,
      "rewards/forensics_sinkhorn_reward": 0.7027748823165894,
      "rewards/forensics_strict_edge_reward": 0.06936766020953655,
      "rewards/forensics_tiou_reward": 0.5012387484312057,
      "step": 628
    },
    {
      "completion_length": 23.828125,
      "epoch": 1.9308755760368663,
      "grad_norm": 9.303394317626953,
      "kl": 0.708984375,
      "learning_rate": 5.16923076923077e-07,
      "loss": -0.1766,
      "reward": 1.177081048488617,
      "reward_std": 0.35248005390167236,
      "reward_std_all": 0.9830735325813293,
      "rewards-max/forensics_sinkhorn_reward": 0.978615015745163,
      "rewards-max/forensics_strict_edge_reward": 0.3543699085712433,
      "rewards-max/forensics_tiou_reward": 0.9730973541736603,
      "rewards/forensics_sinkhorn_reward": 0.653028815984726,
      "rewards/forensics_strict_edge_reward": 0.03436936065554619,
      "rewards/forensics_tiou_reward": 0.4896828532218933,
      "step": 629
    },
    {
      "completion_length": 26.0625,
      "epoch": 1.93394777265745,
      "grad_norm": 6.890493392944336,
      "kl": 0.4208984375,
      "learning_rate": 5.161538461538462e-07,
      "loss": -0.0331,
      "reward": 1.1305588483810425,
      "reward_std": 0.28419850766658783,
      "reward_std_all": 0.9832196235656738,
      "rewards-max/forensics_sinkhorn_reward": 0.9767383933067322,
      "rewards-max/forensics_strict_edge_reward": 0.45954063534736633,
      "rewards-max/forensics_tiou_reward": 0.9704999923706055,
      "rewards/forensics_sinkhorn_reward": 0.6386580765247345,
      "rewards/forensics_strict_edge_reward": 0.026134207844734192,
      "rewards/forensics_tiou_reward": 0.46576663851737976,
      "step": 630
    },
    {
      "completion_length": 27.28125,
      "epoch": 1.9370199692780337,
      "grad_norm": 8.109560012817383,
      "kl": 0.4228515625,
      "learning_rate": 5.153846153846153e-07,
      "loss": -0.066,
      "reward": 1.1644687056541443,
      "reward_std": 0.18759963661432266,
      "reward_std_all": 0.9473176598548889,
      "rewards-max/forensics_sinkhorn_reward": 0.9796337485313416,
      "rewards-max/forensics_strict_edge_reward": 0.44855010509490967,
      "rewards-max/forensics_tiou_reward": 0.978709489107132,
      "rewards/forensics_sinkhorn_reward": 0.6380189657211304,
      "rewards/forensics_strict_edge_reward": 0.03244753181934357,
      "rewards/forensics_tiou_reward": 0.49400220811367035,
      "step": 631
    },
    {
      "completion_length": 24.203125,
      "epoch": 1.9400921658986174,
      "grad_norm": 6.289647102355957,
      "kl": 0.353515625,
      "learning_rate": 5.146153846153846e-07,
      "loss": -0.0833,
      "reward": 1.0527976751327515,
      "reward_std": 0.23096486181020737,
      "reward_std_all": 1.008477807044983,
      "rewards-max/forensics_sinkhorn_reward": 0.9176406562328339,
      "rewards-max/forensics_strict_edge_reward": 0.2846794556826353,
      "rewards-max/forensics_tiou_reward": 0.8413772881031036,
      "rewards/forensics_sinkhorn_reward": 0.6301566064357758,
      "rewards/forensics_strict_edge_reward": 0.020507657434791327,
      "rewards/forensics_tiou_reward": 0.40213336050510406,
      "step": 632
    },
    {
      "completion_length": 24.25,
      "epoch": 1.9431643625192012,
      "grad_norm": 6.0186614990234375,
      "kl": 0.396484375,
      "learning_rate": 5.138461538461538e-07,
      "loss": -0.0965,
      "reward": 0.8681427538394928,
      "reward_std": 0.2846049964427948,
      "reward_std_all": 1.1112509369850159,
      "rewards-max/forensics_sinkhorn_reward": 0.9575245678424835,
      "rewards-max/forensics_strict_edge_reward": 0.33063653111457825,
      "rewards-max/forensics_tiou_reward": 0.9150174856185913,
      "rewards/forensics_sinkhorn_reward": 0.515985831618309,
      "rewards/forensics_strict_edge_reward": 0.016077613458037376,
      "rewards/forensics_tiou_reward": 0.33607926964759827,
      "step": 633
    },
    {
      "completion_length": 26.359375,
      "epoch": 1.946236559139785,
      "grad_norm": 5.817234516143799,
      "kl": 0.396484375,
      "learning_rate": 5.13076923076923e-07,
      "loss": -0.085,
      "reward": 1.111059308052063,
      "reward_std": 0.33364176750183105,
      "reward_std_all": 1.0009576380252838,
      "rewards-max/forensics_sinkhorn_reward": 0.9681576490402222,
      "rewards-max/forensics_strict_edge_reward": 0.5851681530475616,
      "rewards-max/forensics_tiou_reward": 0.9486859440803528,
      "rewards/forensics_sinkhorn_reward": 0.6255208551883698,
      "rewards/forensics_strict_edge_reward": 0.03710456006228924,
      "rewards/forensics_tiou_reward": 0.44843390583992004,
      "step": 634
    },
    {
      "completion_length": 27.609375,
      "epoch": 1.9493087557603688,
      "grad_norm": 6.90401029586792,
      "kl": 0.34375,
      "learning_rate": 5.123076923076923e-07,
      "loss": -0.119,
      "reward": 1.0854023396968842,
      "reward_std": 0.3693382441997528,
      "reward_std_all": 1.0484195947647095,
      "rewards-max/forensics_sinkhorn_reward": 0.9803579449653625,
      "rewards-max/forensics_strict_edge_reward": 0.6331912577152252,
      "rewards-max/forensics_tiou_reward": 0.9805746376514435,
      "rewards/forensics_sinkhorn_reward": 0.6268393695354462,
      "rewards/forensics_strict_edge_reward": 0.0538768763653934,
      "rewards/forensics_tiou_reward": 0.4046861529350281,
      "step": 635
    },
    {
      "completion_length": 25.453125,
      "epoch": 1.9523809523809523,
      "grad_norm": 7.508499622344971,
      "kl": 0.3330078125,
      "learning_rate": 5.115384615384615e-07,
      "loss": -0.0631,
      "reward": 1.1661879420280457,
      "reward_std": 0.3379913866519928,
      "reward_std_all": 0.9755354523658752,
      "rewards-max/forensics_sinkhorn_reward": 0.9740683138370514,
      "rewards-max/forensics_strict_edge_reward": 0.08166956075001508,
      "rewards-max/forensics_tiou_reward": 0.9521548449993134,
      "rewards/forensics_sinkhorn_reward": 0.6421913802623749,
      "rewards/forensics_strict_edge_reward": 0.004164352823863737,
      "rewards/forensics_tiou_reward": 0.5198321789503098,
      "step": 636
    },
    {
      "completion_length": 28.421875,
      "epoch": 1.955453149001536,
      "grad_norm": 6.984659671783447,
      "kl": 0.4228515625,
      "learning_rate": 5.107692307692308e-07,
      "loss": -0.067,
      "reward": 1.0846398770809174,
      "reward_std": 0.3015797734260559,
      "reward_std_all": 1.0128644108772278,
      "rewards-max/forensics_sinkhorn_reward": 0.9694241285324097,
      "rewards-max/forensics_strict_edge_reward": 0.4941990226507187,
      "rewards-max/forensics_tiou_reward": 0.9517737030982971,
      "rewards/forensics_sinkhorn_reward": 0.5965385735034943,
      "rewards/forensics_strict_edge_reward": 0.035614192485809326,
      "rewards/forensics_tiou_reward": 0.45248711109161377,
      "step": 637
    },
    {
      "completion_length": 29.890625,
      "epoch": 1.9585253456221197,
      "grad_norm": 6.572402477264404,
      "kl": 0.4443359375,
      "learning_rate": 5.1e-07,
      "loss": -0.046,
      "reward": 1.2384807467460632,
      "reward_std": 0.4829753190279007,
      "reward_std_all": 0.9977886378765106,
      "rewards-max/forensics_sinkhorn_reward": 0.9799482226371765,
      "rewards-max/forensics_strict_edge_reward": 0.5890537798404694,
      "rewards-max/forensics_tiou_reward": 0.9786963164806366,
      "rewards/forensics_sinkhorn_reward": 0.6610087156295776,
      "rewards/forensics_strict_edge_reward": 0.06834690645337105,
      "rewards/forensics_tiou_reward": 0.5091250687837601,
      "step": 638
    },
    {
      "completion_length": 26.53125,
      "epoch": 1.9615975422427034,
      "grad_norm": 5.9750752449035645,
      "kl": 0.3203125,
      "learning_rate": 5.092307692307692e-07,
      "loss": -0.1258,
      "reward": 1.3738020658493042,
      "reward_std": 0.268270306289196,
      "reward_std_all": 0.869053840637207,
      "rewards-max/forensics_sinkhorn_reward": 0.9762526154518127,
      "rewards-max/forensics_strict_edge_reward": 0.7505797147750854,
      "rewards-max/forensics_tiou_reward": 0.9734634459018707,
      "rewards/forensics_sinkhorn_reward": 0.7418510317802429,
      "rewards/forensics_strict_edge_reward": 0.0918816402554512,
      "rewards/forensics_tiou_reward": 0.5400694459676743,
      "step": 639
    },
    {
      "completion_length": 23.734375,
      "epoch": 1.9646697388632872,
      "grad_norm": 6.038504600524902,
      "kl": 0.576171875,
      "learning_rate": 5.084615384615384e-07,
      "loss": -0.0252,
      "reward": 1.1227200627326965,
      "reward_std": 0.3872060924768448,
      "reward_std_all": 1.018719732761383,
      "rewards-max/forensics_sinkhorn_reward": 0.9748225808143616,
      "rewards-max/forensics_strict_edge_reward": 0.5237197279930115,
      "rewards-max/forensics_tiou_reward": 0.9623795449733734,
      "rewards/forensics_sinkhorn_reward": 0.6425738334655762,
      "rewards/forensics_strict_edge_reward": 0.03829440101981163,
      "rewards/forensics_tiou_reward": 0.4418519139289856,
      "step": 640
    },
    {
      "completion_length": 25.15625,
      "epoch": 1.967741935483871,
      "grad_norm": 5.839356899261475,
      "kl": 0.3271484375,
      "learning_rate": 5.076923076923076e-07,
      "loss": -0.0975,
      "reward": 1.185076117515564,
      "reward_std": 0.3572799861431122,
      "reward_std_all": 0.9813253879547119,
      "rewards-max/forensics_sinkhorn_reward": 0.9739222526550293,
      "rewards-max/forensics_strict_edge_reward": 0.4589645564556122,
      "rewards-max/forensics_tiou_reward": 0.9627168476581573,
      "rewards/forensics_sinkhorn_reward": 0.6577890515327454,
      "rewards/forensics_strict_edge_reward": 0.04740532860159874,
      "rewards/forensics_tiou_reward": 0.4798817038536072,
      "step": 641
    },
    {
      "completion_length": 25.53125,
      "epoch": 1.9708141321044548,
      "grad_norm": 5.874617099761963,
      "kl": 0.5,
      "learning_rate": 5.06923076923077e-07,
      "loss": -0.0034,
      "reward": 1.2276738286018372,
      "reward_std": 0.32919226586818695,
      "reward_std_all": 0.9563078880310059,
      "rewards-max/forensics_sinkhorn_reward": 0.967227578163147,
      "rewards-max/forensics_strict_edge_reward": 0.5548860281705856,
      "rewards-max/forensics_tiou_reward": 0.9470222294330597,
      "rewards/forensics_sinkhorn_reward": 0.7025453448295593,
      "rewards/forensics_strict_edge_reward": 0.0337991937994957,
      "rewards/forensics_tiou_reward": 0.4913293421268463,
      "step": 642
    },
    {
      "completion_length": 24.703125,
      "epoch": 1.9738863287250386,
      "grad_norm": 7.618697643280029,
      "kl": 0.3984375,
      "learning_rate": 5.061538461538462e-07,
      "loss": -0.1081,
      "reward": 1.0171266794204712,
      "reward_std": 0.3112982511520386,
      "reward_std_all": 1.047679841518402,
      "rewards-max/forensics_sinkhorn_reward": 0.9638548493385315,
      "rewards-max/forensics_strict_edge_reward": 0.3699144087731838,
      "rewards-max/forensics_tiou_reward": 0.9421840906143188,
      "rewards/forensics_sinkhorn_reward": 0.5845911502838135,
      "rewards/forensics_strict_edge_reward": 0.02982229262124747,
      "rewards/forensics_tiou_reward": 0.40271322429180145,
      "step": 643
    },
    {
      "completion_length": 25.015625,
      "epoch": 1.976958525345622,
      "grad_norm": 7.662853717803955,
      "kl": 0.41015625,
      "learning_rate": 5.053846153846154e-07,
      "loss": -0.0623,
      "reward": 1.1735795140266418,
      "reward_std": 0.34862928092479706,
      "reward_std_all": 0.9867183566093445,
      "rewards-max/forensics_sinkhorn_reward": 0.9742115139961243,
      "rewards-max/forensics_strict_edge_reward": 0.43710191547870636,
      "rewards-max/forensics_tiou_reward": 0.9670519530773163,
      "rewards/forensics_sinkhorn_reward": 0.6548206210136414,
      "rewards/forensics_strict_edge_reward": 0.040245214477181435,
      "rewards/forensics_tiou_reward": 0.47851358354091644,
      "step": 644
    },
    {
      "completion_length": 28.96875,
      "epoch": 1.9800307219662059,
      "grad_norm": 6.409226417541504,
      "kl": 0.34765625,
      "learning_rate": 5.046153846153846e-07,
      "loss": -0.071,
      "reward": 1.2108256220817566,
      "reward_std": 0.25747884809970856,
      "reward_std_all": 0.9348888397216797,
      "rewards-max/forensics_sinkhorn_reward": 0.9820020496845245,
      "rewards-max/forensics_strict_edge_reward": 0.5102156698703766,
      "rewards-max/forensics_tiou_reward": 0.9810146987438202,
      "rewards/forensics_sinkhorn_reward": 0.6441316306591034,
      "rewards/forensics_strict_edge_reward": 0.038667239248752594,
      "rewards/forensics_tiou_reward": 0.5280267000198364,
      "step": 645
    },
    {
      "completion_length": 26.5625,
      "epoch": 1.9831029185867894,
      "grad_norm": 6.546656608581543,
      "kl": 0.37109375,
      "learning_rate": 5.038461538461538e-07,
      "loss": -0.1103,
      "reward": 1.369869589805603,
      "reward_std": 0.3598264753818512,
      "reward_std_all": 0.9101850688457489,
      "rewards-max/forensics_sinkhorn_reward": 0.982771247625351,
      "rewards-max/forensics_strict_edge_reward": 0.7246644794940948,
      "rewards-max/forensics_tiou_reward": 0.9845055639743805,
      "rewards/forensics_sinkhorn_reward": 0.7320771515369415,
      "rewards/forensics_strict_edge_reward": 0.07521812804043293,
      "rewards/forensics_tiou_reward": 0.5625742971897125,
      "step": 646
    },
    {
      "completion_length": 26.234375,
      "epoch": 1.9861751152073732,
      "grad_norm": 7.066744327545166,
      "kl": 0.4267578125,
      "learning_rate": 5.03076923076923e-07,
      "loss": -0.0732,
      "reward": 1.3182119131088257,
      "reward_std": 0.3664795905351639,
      "reward_std_all": 0.9233562350273132,
      "rewards-max/forensics_sinkhorn_reward": 0.9804932475090027,
      "rewards-max/forensics_strict_edge_reward": 0.5122979953885078,
      "rewards-max/forensics_tiou_reward": 0.9786939024925232,
      "rewards/forensics_sinkhorn_reward": 0.7071215510368347,
      "rewards/forensics_strict_edge_reward": 0.05728939455002546,
      "rewards/forensics_tiou_reward": 0.5538009405136108,
      "step": 647
    },
    {
      "completion_length": 26.09375,
      "epoch": 1.989247311827957,
      "grad_norm": 7.848992824554443,
      "kl": 0.5458984375,
      "learning_rate": 5.023076923076923e-07,
      "loss": -0.0148,
      "reward": 1.1911039352416992,
      "reward_std": 0.3874622732400894,
      "reward_std_all": 0.9975956678390503,
      "rewards-max/forensics_sinkhorn_reward": 0.9721669256687164,
      "rewards-max/forensics_strict_edge_reward": 0.5522415935993195,
      "rewards-max/forensics_tiou_reward": 0.959103137254715,
      "rewards/forensics_sinkhorn_reward": 0.6427423357963562,
      "rewards/forensics_strict_edge_reward": 0.06533329095691442,
      "rewards/forensics_tiou_reward": 0.48302826285362244,
      "step": 648
    },
    {
      "completion_length": 28.40625,
      "epoch": 1.9923195084485408,
      "grad_norm": 6.672125339508057,
      "kl": 0.359375,
      "learning_rate": 5.015384615384614e-07,
      "loss": -0.1514,
      "reward": 1.360614538192749,
      "reward_std": 0.342021182179451,
      "reward_std_all": 0.8945128321647644,
      "rewards-max/forensics_sinkhorn_reward": 0.9734531342983246,
      "rewards-max/forensics_strict_edge_reward": 0.3747613877058029,
      "rewards-max/forensics_tiou_reward": 0.9519511163234711,
      "rewards/forensics_sinkhorn_reward": 0.7255959510803223,
      "rewards/forensics_strict_edge_reward": 0.03984912484884262,
      "rewards/forensics_tiou_reward": 0.5951694250106812,
      "step": 649
    },
    {
      "completion_length": 22.046875,
      "epoch": 1.9953917050691246,
      "grad_norm": 7.171604156494141,
      "kl": 0.505859375,
      "learning_rate": 5.007692307692308e-07,
      "loss": -0.1152,
      "reward": 1.1752285957336426,
      "reward_std": 0.3486242890357971,
      "reward_std_all": 0.9970671534538269,
      "rewards-max/forensics_sinkhorn_reward": 0.9796849489212036,
      "rewards-max/forensics_strict_edge_reward": 0.6175843328237534,
      "rewards-max/forensics_tiou_reward": 0.9739839136600494,
      "rewards/forensics_sinkhorn_reward": 0.661706417798996,
      "rewards/forensics_strict_edge_reward": 0.0493305791169405,
      "rewards/forensics_tiou_reward": 0.46419157087802887,
      "step": 650
    },
    {
      "completion_length": 26.953125,
      "epoch": 1.998463901689708,
      "grad_norm": 7.888006210327148,
      "kl": 0.453125,
      "learning_rate": 5e-07,
      "loss": 0.0351,
      "reward": 1.056178092956543,
      "reward_std": 0.26394712924957275,
      "reward_std_all": 1.0130636096000671,
      "rewards-max/forensics_sinkhorn_reward": 0.9639127254486084,
      "rewards-max/forensics_strict_edge_reward": 0.2206837348639965,
      "rewards-max/forensics_tiou_reward": 0.9433318674564362,
      "rewards/forensics_sinkhorn_reward": 0.5760360658168793,
      "rewards/forensics_strict_edge_reward": 0.017703427467495203,
      "rewards/forensics_tiou_reward": 0.46243853867053986,
      "step": 651
    },
    {
      "completion_length": 25.200000762939453,
      "epoch": 2.0,
      "grad_norm": 4.06988525390625,
      "kl": 0.671875,
      "learning_rate": 4.992307692307692e-07,
      "loss": -0.0591,
      "reward": 1.416530728340149,
      "reward_std": 0.4660460948944092,
      "reward_std_all": 0.9785410165786743,
      "rewards-max/forensics_sinkhorn_reward": 0.9750109314918518,
      "rewards-max/forensics_strict_edge_reward": 0.740818202495575,
      "rewards-max/forensics_tiou_reward": 0.9681087136268616,
      "rewards/forensics_sinkhorn_reward": 0.7043331265449524,
      "rewards/forensics_strict_edge_reward": 0.15456269681453705,
      "rewards/forensics_tiou_reward": 0.5576348900794983,
      "step": 652
    },
    {
      "completion_length": 22.578125,
      "epoch": 2.003072196620584,
      "grad_norm": 7.907129764556885,
      "kl": 0.572265625,
      "learning_rate": 4.984615384615384e-07,
      "loss": -0.1577,
      "reward": 1.232509970664978,
      "reward_std": 0.2632924020290375,
      "reward_std_all": 0.9313126802444458,
      "rewards-max/forensics_sinkhorn_reward": 0.9637713730335236,
      "rewards-max/forensics_strict_edge_reward": 0.4270339459180832,
      "rewards-max/forensics_tiou_reward": 0.9391975998878479,
      "rewards/forensics_sinkhorn_reward": 0.680557906627655,
      "rewards/forensics_strict_edge_reward": 0.04060113709419966,
      "rewards/forensics_tiou_reward": 0.5113508701324463,
      "step": 653
    },
    {
      "completion_length": 24.375,
      "epoch": 2.0061443932411676,
      "grad_norm": 8.522977828979492,
      "kl": 0.478515625,
      "learning_rate": 4.976923076923077e-07,
      "loss": -0.0756,
      "reward": 1.3500626683235168,
      "reward_std": 0.41043123602867126,
      "reward_std_all": 0.9230304658412933,
      "rewards-max/forensics_sinkhorn_reward": 0.9815065264701843,
      "rewards-max/forensics_strict_edge_reward": 0.7201169431209564,
      "rewards-max/forensics_tiou_reward": 0.9792087376117706,
      "rewards/forensics_sinkhorn_reward": 0.7400445640087128,
      "rewards/forensics_strict_edge_reward": 0.0829140953719616,
      "rewards/forensics_tiou_reward": 0.5271039605140686,
      "step": 654
    },
    {
      "completion_length": 27.328125,
      "epoch": 2.0092165898617513,
      "grad_norm": 6.216756820678711,
      "kl": 0.4853515625,
      "learning_rate": 4.969230769230769e-07,
      "loss": -0.0656,
      "reward": 1.428507685661316,
      "reward_std": 0.394294872879982,
      "reward_std_all": 0.8866145610809326,
      "rewards-max/forensics_sinkhorn_reward": 0.9667572379112244,
      "rewards-max/forensics_strict_edge_reward": 0.7624467611312866,
      "rewards-max/forensics_tiou_reward": 0.9519924223423004,
      "rewards/forensics_sinkhorn_reward": 0.7507487833499908,
      "rewards/forensics_strict_edge_reward": 0.10202514193952084,
      "rewards/forensics_tiou_reward": 0.5757337808609009,
      "step": 655
    },
    {
      "completion_length": 24.328125,
      "epoch": 2.0122887864823347,
      "grad_norm": 6.931584358215332,
      "kl": 0.4716796875,
      "learning_rate": 4.961538461538462e-07,
      "loss": -0.1365,
      "reward": 1.4511579275131226,
      "reward_std": 0.3541509509086609,
      "reward_std_all": 0.8466099798679352,
      "rewards-max/forensics_sinkhorn_reward": 0.9803006649017334,
      "rewards-max/forensics_strict_edge_reward": 0.5171787291765213,
      "rewards-max/forensics_tiou_reward": 0.9757716357707977,
      "rewards/forensics_sinkhorn_reward": 0.7543181777000427,
      "rewards/forensics_strict_edge_reward": 0.06587209179997444,
      "rewards/forensics_tiou_reward": 0.6309677064418793,
      "step": 656
    },
    {
      "completion_length": 27.609375,
      "epoch": 2.0153609831029184,
      "grad_norm": 7.793954372406006,
      "kl": 0.6181640625,
      "learning_rate": 4.953846153846154e-07,
      "loss": -0.0893,
      "reward": 1.3785158395767212,
      "reward_std": 0.4585207998752594,
      "reward_std_all": 0.9372773170471191,
      "rewards-max/forensics_sinkhorn_reward": 0.9873659610748291,
      "rewards-max/forensics_strict_edge_reward": 0.7166809141635895,
      "rewards-max/forensics_tiou_reward": 0.9953301846981049,
      "rewards/forensics_sinkhorn_reward": 0.6775970160961151,
      "rewards/forensics_strict_edge_reward": 0.11099811643362045,
      "rewards/forensics_tiou_reward": 0.5899206697940826,
      "step": 657
    },
    {
      "completion_length": 26.40625,
      "epoch": 2.0184331797235022,
      "grad_norm": 7.220152378082275,
      "kl": 0.462890625,
      "learning_rate": 4.946153846153846e-07,
      "loss": -0.0607,
      "reward": 1.2461327910423279,
      "reward_std": 0.1395852528512478,
      "reward_std_all": 0.8881086707115173,
      "rewards-max/forensics_sinkhorn_reward": 0.9831445813179016,
      "rewards-max/forensics_strict_edge_reward": 0.5302301347255707,
      "rewards-max/forensics_tiou_reward": 0.9840450286865234,
      "rewards/forensics_sinkhorn_reward": 0.7005641758441925,
      "rewards/forensics_strict_edge_reward": 0.04157351516187191,
      "rewards/forensics_tiou_reward": 0.5039951354265213,
      "step": 658
    },
    {
      "completion_length": 23.953125,
      "epoch": 2.021505376344086,
      "grad_norm": 8.613859176635742,
      "kl": 0.6689453125,
      "learning_rate": 4.938461538461538e-07,
      "loss": -0.0784,
      "reward": 1.2014504075050354,
      "reward_std": 0.34802567958831787,
      "reward_std_all": 0.9770957827568054,
      "rewards-max/forensics_sinkhorn_reward": 0.9668132364749908,
      "rewards-max/forensics_strict_edge_reward": 0.6110004484653473,
      "rewards-max/forensics_tiou_reward": 0.9522514939308167,
      "rewards/forensics_sinkhorn_reward": 0.6479346752166748,
      "rewards/forensics_strict_edge_reward": 0.06988603621721268,
      "rewards/forensics_tiou_reward": 0.4836297631263733,
      "step": 659
    },
    {
      "completion_length": 25.46875,
      "epoch": 2.02457757296467,
      "grad_norm": 8.098487854003906,
      "kl": 0.6640625,
      "learning_rate": 4.93076923076923e-07,
      "loss": -0.0225,
      "reward": 1.301952838897705,
      "reward_std": 0.4023004174232483,
      "reward_std_all": 0.9466465711593628,
      "rewards-max/forensics_sinkhorn_reward": 0.9871052205562592,
      "rewards-max/forensics_strict_edge_reward": 0.8301283419132233,
      "rewards-max/forensics_tiou_reward": 0.9953585863113403,
      "rewards/forensics_sinkhorn_reward": 0.7031881511211395,
      "rewards/forensics_strict_edge_reward": 0.08211034908890724,
      "rewards/forensics_tiou_reward": 0.5166543126106262,
      "step": 660
    },
    {
      "completion_length": 28.78125,
      "epoch": 2.0276497695852536,
      "grad_norm": 7.420882701873779,
      "kl": 0.470703125,
      "learning_rate": 4.923076923076923e-07,
      "loss": -0.1652,
      "reward": 1.2819794416427612,
      "reward_std": 0.2985483705997467,
      "reward_std_all": 0.9230906665325165,
      "rewards-max/forensics_sinkhorn_reward": 0.9805633425712585,
      "rewards-max/forensics_strict_edge_reward": 0.6293275952339172,
      "rewards-max/forensics_tiou_reward": 0.9773761928081512,
      "rewards/forensics_sinkhorn_reward": 0.7025628387928009,
      "rewards/forensics_strict_edge_reward": 0.06792321056127548,
      "rewards/forensics_tiou_reward": 0.5114934593439102,
      "step": 661
    },
    {
      "completion_length": 24.03125,
      "epoch": 2.0307219662058373,
      "grad_norm": 7.090987682342529,
      "kl": 0.4228515625,
      "learning_rate": 4.915384615384615e-07,
      "loss": -0.1229,
      "reward": 1.4769843816757202,
      "reward_std": 0.2208634316921234,
      "reward_std_all": 0.8116861879825592,
      "rewards-max/forensics_sinkhorn_reward": 0.9837638139724731,
      "rewards-max/forensics_strict_edge_reward": 0.7165313363075256,
      "rewards-max/forensics_tiou_reward": 0.9865723252296448,
      "rewards/forensics_sinkhorn_reward": 0.7816917300224304,
      "rewards/forensics_strict_edge_reward": 0.06567893922328949,
      "rewards/forensics_tiou_reward": 0.6296137571334839,
      "step": 662
    },
    {
      "completion_length": 26.765625,
      "epoch": 2.033794162826421,
      "grad_norm": 8.162544250488281,
      "kl": 0.4599609375,
      "learning_rate": 4.907692307692308e-07,
      "loss": -0.1003,
      "reward": 1.379513919353485,
      "reward_std": 0.23827750980854034,
      "reward_std_all": 0.8472115397453308,
      "rewards-max/forensics_sinkhorn_reward": 0.9798130393028259,
      "rewards-max/forensics_strict_edge_reward": 0.36248624697327614,
      "rewards-max/forensics_tiou_reward": 0.9714936316013336,
      "rewards/forensics_sinkhorn_reward": 0.7527213990688324,
      "rewards/forensics_strict_edge_reward": 0.02269398234784603,
      "rewards/forensics_tiou_reward": 0.6040984690189362,
      "step": 663
    },
    {
      "completion_length": 28.421875,
      "epoch": 2.0368663594470044,
      "grad_norm": 7.002873420715332,
      "kl": 0.384765625,
      "learning_rate": 4.9e-07,
      "loss": -0.0871,
      "reward": 1.34750235080719,
      "reward_std": 0.4589154124259949,
      "reward_std_all": 0.9451836347579956,
      "rewards-max/forensics_sinkhorn_reward": 0.9734811186790466,
      "rewards-max/forensics_strict_edge_reward": 0.756708562374115,
      "rewards-max/forensics_tiou_reward": 0.9648283421993256,
      "rewards/forensics_sinkhorn_reward": 0.7167003750801086,
      "rewards/forensics_strict_edge_reward": 0.08828585036098957,
      "rewards/forensics_tiou_reward": 0.5425161570310593,
      "step": 664
    },
    {
      "completion_length": 29.078125,
      "epoch": 2.0399385560675882,
      "grad_norm": 6.718451976776123,
      "kl": 0.5107421875,
      "learning_rate": 4.892307692307692e-07,
      "loss": -0.051,
      "reward": 1.5448073148727417,
      "reward_std": 0.3347768187522888,
      "reward_std_all": 0.8061342239379883,
      "rewards-max/forensics_sinkhorn_reward": 0.9801850020885468,
      "rewards-max/forensics_strict_edge_reward": 0.7744058072566986,
      "rewards-max/forensics_tiou_reward": 0.9793041944503784,
      "rewards/forensics_sinkhorn_reward": 0.7732470035552979,
      "rewards/forensics_strict_edge_reward": 0.11337216943502426,
      "rewards/forensics_tiou_reward": 0.658188134431839,
      "step": 665
    },
    {
      "completion_length": 24.71875,
      "epoch": 2.043010752688172,
      "grad_norm": 12.922892570495605,
      "kl": 1.009765625,
      "learning_rate": 4.884615384615384e-07,
      "loss": -0.0271,
      "reward": 1.2886539697647095,
      "reward_std": 0.29480116814374924,
      "reward_std_all": 0.9092158079147339,
      "rewards-max/forensics_sinkhorn_reward": 0.9725785255432129,
      "rewards-max/forensics_strict_edge_reward": 0.5500316619873047,
      "rewards-max/forensics_tiou_reward": 0.9627954959869385,
      "rewards/forensics_sinkhorn_reward": 0.7059318423271179,
      "rewards/forensics_strict_edge_reward": 0.04116593860089779,
      "rewards/forensics_tiou_reward": 0.5415560901165009,
      "step": 666
    },
    {
      "completion_length": 25.671875,
      "epoch": 2.046082949308756,
      "grad_norm": 7.01169490814209,
      "kl": 0.365234375,
      "learning_rate": 4.876923076923077e-07,
      "loss": -0.0573,
      "reward": 1.2872667908668518,
      "reward_std": 0.3494075834751129,
      "reward_std_all": 0.9336965084075928,
      "rewards-max/forensics_sinkhorn_reward": 0.9652649760246277,
      "rewards-max/forensics_strict_edge_reward": 0.5374098420143127,
      "rewards-max/forensics_tiou_reward": 0.9358754456043243,
      "rewards/forensics_sinkhorn_reward": 0.6895288527011871,
      "rewards/forensics_strict_edge_reward": 0.05333683267235756,
      "rewards/forensics_tiou_reward": 0.5444011092185974,
      "step": 667
    },
    {
      "completion_length": 24.140625,
      "epoch": 2.0491551459293396,
      "grad_norm": 9.642937660217285,
      "kl": 0.396484375,
      "learning_rate": 4.869230769230769e-07,
      "loss": -0.103,
      "reward": 1.5128851532936096,
      "reward_std": 0.3417707085609436,
      "reward_std_all": 0.8366807997226715,
      "rewards-max/forensics_sinkhorn_reward": 0.9887476861476898,
      "rewards-max/forensics_strict_edge_reward": 0.8959447741508484,
      "rewards-max/forensics_tiou_reward": 0.9992177784442902,
      "rewards/forensics_sinkhorn_reward": 0.7550733387470245,
      "rewards/forensics_strict_edge_reward": 0.09969836473464966,
      "rewards/forensics_tiou_reward": 0.6581133604049683,
      "step": 668
    },
    {
      "completion_length": 24.8125,
      "epoch": 2.0522273425499233,
      "grad_norm": 7.8865227699279785,
      "kl": 0.5400390625,
      "learning_rate": 4.861538461538461e-07,
      "loss": -0.1027,
      "reward": 1.3520002961158752,
      "reward_std": 0.45951223373413086,
      "reward_std_all": 0.9500489830970764,
      "rewards-max/forensics_sinkhorn_reward": 0.9838365614414215,
      "rewards-max/forensics_strict_edge_reward": 0.7875787317752838,
      "rewards-max/forensics_tiou_reward": 0.9852180778980255,
      "rewards/forensics_sinkhorn_reward": 0.7131746113300323,
      "rewards/forensics_strict_edge_reward": 0.07854346558451653,
      "rewards/forensics_tiou_reward": 0.5602823197841644,
      "step": 669
    },
    {
      "completion_length": 25.390625,
      "epoch": 2.055299539170507,
      "grad_norm": 7.963527679443359,
      "kl": 0.408203125,
      "learning_rate": 4.853846153846154e-07,
      "loss": -0.1283,
      "reward": 1.3052340149879456,
      "reward_std": 0.29874956607818604,
      "reward_std_all": 0.9182126224040985,
      "rewards-max/forensics_sinkhorn_reward": 0.9881217777729034,
      "rewards-max/forensics_strict_edge_reward": 0.8540171384811401,
      "rewards-max/forensics_tiou_reward": 0.9976402223110199,
      "rewards/forensics_sinkhorn_reward": 0.6938395202159882,
      "rewards/forensics_strict_edge_reward": 0.0675537995994091,
      "rewards/forensics_tiou_reward": 0.5438407063484192,
      "step": 670
    },
    {
      "completion_length": 23.25,
      "epoch": 2.0583717357910905,
      "grad_norm": 6.774673938751221,
      "kl": 0.482421875,
      "learning_rate": 4.846153846153846e-07,
      "loss": -0.1489,
      "reward": 1.2980612516403198,
      "reward_std": 0.4063135087490082,
      "reward_std_all": 0.9544178247451782,
      "rewards-max/forensics_sinkhorn_reward": 0.9846198260784149,
      "rewards-max/forensics_strict_edge_reward": 0.6515887677669525,
      "rewards-max/forensics_tiou_reward": 0.9870551824569702,
      "rewards/forensics_sinkhorn_reward": 0.6887533664703369,
      "rewards/forensics_strict_edge_reward": 0.07808881253004074,
      "rewards/forensics_tiou_reward": 0.5312191247940063,
      "step": 671
    },
    {
      "completion_length": 28.890625,
      "epoch": 2.0614439324116742,
      "grad_norm": 9.240812301635742,
      "kl": 0.4453125,
      "learning_rate": 4.838461538461539e-07,
      "loss": -0.0642,
      "reward": 1.245207130908966,
      "reward_std": 0.39644839614629745,
      "reward_std_all": 0.9769793450832367,
      "rewards-max/forensics_sinkhorn_reward": 0.9720290005207062,
      "rewards-max/forensics_strict_edge_reward": 0.8333333432674408,
      "rewards-max/forensics_tiou_reward": 0.9530636668205261,
      "rewards/forensics_sinkhorn_reward": 0.6273481249809265,
      "rewards/forensics_strict_edge_reward": 0.08412563242018223,
      "rewards/forensics_tiou_reward": 0.5337333530187607,
      "step": 672
    },
    {
      "completion_length": 29.578125,
      "epoch": 2.064516129032258,
      "grad_norm": 7.9847612380981445,
      "kl": 0.4404296875,
      "learning_rate": 4.83076923076923e-07,
      "loss": -0.0973,
      "reward": 1.4447003602981567,
      "reward_std": 0.3265375345945358,
      "reward_std_all": 0.8692824244499207,
      "rewards-max/forensics_sinkhorn_reward": 0.9866258800029755,
      "rewards-max/forensics_strict_edge_reward": 0.7829082906246185,
      "rewards-max/forensics_tiou_reward": 0.9924359321594238,
      "rewards/forensics_sinkhorn_reward": 0.7124428153038025,
      "rewards/forensics_strict_edge_reward": 0.11838169023394585,
      "rewards/forensics_tiou_reward": 0.6138758361339569,
      "step": 673
    },
    {
      "completion_length": 23.59375,
      "epoch": 2.067588325652842,
      "grad_norm": 15.703649520874023,
      "kl": 0.4921875,
      "learning_rate": 4.823076923076923e-07,
      "loss": -0.0748,
      "reward": 1.2876284718513489,
      "reward_std": 0.3540109544992447,
      "reward_std_all": 0.9353923201560974,
      "rewards-max/forensics_sinkhorn_reward": 0.9766563773155212,
      "rewards-max/forensics_strict_edge_reward": 0.5708463788032532,
      "rewards-max/forensics_tiou_reward": 0.966937929391861,
      "rewards/forensics_sinkhorn_reward": 0.6815122663974762,
      "rewards/forensics_strict_edge_reward": 0.05781814083456993,
      "rewards/forensics_tiou_reward": 0.5482981503009796,
      "step": 674
    },
    {
      "completion_length": 25.96875,
      "epoch": 2.0706605222734256,
      "grad_norm": 7.410651683807373,
      "kl": 0.4365234375,
      "learning_rate": 4.815384615384615e-07,
      "loss": -0.1047,
      "reward": 1.2615975141525269,
      "reward_std": 0.20026522129774094,
      "reward_std_all": 0.9002167284488678,
      "rewards-max/forensics_sinkhorn_reward": 0.9201103448867798,
      "rewards-max/forensics_strict_edge_reward": 0.3628445779904723,
      "rewards-max/forensics_tiou_reward": 0.8392974436283112,
      "rewards/forensics_sinkhorn_reward": 0.7185377776622772,
      "rewards/forensics_strict_edge_reward": 0.04226788075175136,
      "rewards/forensics_tiou_reward": 0.5007917732000351,
      "step": 675
    },
    {
      "completion_length": 27.53125,
      "epoch": 2.0737327188940093,
      "grad_norm": 7.438793659210205,
      "kl": 0.4921875,
      "learning_rate": 4.807692307692307e-07,
      "loss": -0.0545,
      "reward": 1.4037386775016785,
      "reward_std": 0.4038923680782318,
      "reward_std_all": 0.8957858085632324,
      "rewards-max/forensics_sinkhorn_reward": 0.9819574952125549,
      "rewards-max/forensics_strict_edge_reward": 0.7489434033632278,
      "rewards-max/forensics_tiou_reward": 0.9850119948387146,
      "rewards/forensics_sinkhorn_reward": 0.6950364410877228,
      "rewards/forensics_strict_edge_reward": 0.07394945994019508,
      "rewards/forensics_tiou_reward": 0.6347528100013733,
      "step": 676
    },
    {
      "completion_length": 25.359375,
      "epoch": 2.076804915514593,
      "grad_norm": 8.269720077514648,
      "kl": 0.453125,
      "learning_rate": 4.8e-07,
      "loss": -0.1292,
      "reward": 1.500756561756134,
      "reward_std": 0.3740976005792618,
      "reward_std_all": 0.8509421944618225,
      "rewards-max/forensics_sinkhorn_reward": 0.9857129156589508,
      "rewards-max/forensics_strict_edge_reward": 0.8419698476791382,
      "rewards-max/forensics_tiou_reward": 0.9914473593235016,
      "rewards/forensics_sinkhorn_reward": 0.7796477675437927,
      "rewards/forensics_strict_edge_reward": 0.11389889568090439,
      "rewards/forensics_tiou_reward": 0.6072098016738892,
      "step": 677
    },
    {
      "completion_length": 26.328125,
      "epoch": 2.0798771121351765,
      "grad_norm": 8.53193187713623,
      "kl": 0.744140625,
      "learning_rate": 4.792307692307692e-07,
      "loss": -0.0367,
      "reward": 1.3702203631401062,
      "reward_std": 0.4894121289253235,
      "reward_std_all": 0.9377756416797638,
      "rewards-max/forensics_sinkhorn_reward": 0.9803282618522644,
      "rewards-max/forensics_strict_edge_reward": 0.5557698607444763,
      "rewards-max/forensics_tiou_reward": 0.9764481484889984,
      "rewards/forensics_sinkhorn_reward": 0.7181700170040131,
      "rewards/forensics_strict_edge_reward": 0.07691311836242676,
      "rewards/forensics_tiou_reward": 0.5751373171806335,
      "step": 678
    },
    {
      "completion_length": 26.96875,
      "epoch": 2.0829493087557602,
      "grad_norm": 9.059444427490234,
      "kl": 0.4697265625,
      "learning_rate": 4.784615384615384e-07,
      "loss": -0.1132,
      "reward": 0.8966721892356873,
      "reward_std": 0.18651579320430756,
      "reward_std_all": 1.068469524383545,
      "rewards-max/forensics_sinkhorn_reward": 0.9531027674674988,
      "rewards-max/forensics_strict_edge_reward": 0.15399670973420143,
      "rewards-max/forensics_tiou_reward": 0.9193972945213318,
      "rewards/forensics_sinkhorn_reward": 0.524910032749176,
      "rewards/forensics_strict_edge_reward": 0.013415929861366749,
      "rewards/forensics_tiou_reward": 0.35834625363349915,
      "step": 679
    },
    {
      "completion_length": 29.703125,
      "epoch": 2.086021505376344,
      "grad_norm": 7.935710906982422,
      "kl": 0.4521484375,
      "learning_rate": 4.776923076923077e-07,
      "loss": -0.0736,
      "reward": 1.4136798977851868,
      "reward_std": 0.3580692410469055,
      "reward_std_all": 0.8827133774757385,
      "rewards-max/forensics_sinkhorn_reward": 0.9840721487998962,
      "rewards-max/forensics_strict_edge_reward": 0.882964164018631,
      "rewards-max/forensics_tiou_reward": 0.9845281541347504,
      "rewards/forensics_sinkhorn_reward": 0.7071000337600708,
      "rewards/forensics_strict_edge_reward": 0.13872363418340683,
      "rewards/forensics_tiou_reward": 0.5678562521934509,
      "step": 680
    },
    {
      "completion_length": 26.34375,
      "epoch": 2.089093701996928,
      "grad_norm": 14.883148193359375,
      "kl": 0.85546875,
      "learning_rate": 4.769230769230769e-07,
      "loss": -0.0854,
      "reward": 1.5907090902328491,
      "reward_std": 0.42908793687820435,
      "reward_std_all": 0.8142779767513275,
      "rewards-max/forensics_sinkhorn_reward": 0.9840444028377533,
      "rewards-max/forensics_strict_edge_reward": 0.7972188889980316,
      "rewards-max/forensics_tiou_reward": 0.9833860397338867,
      "rewards/forensics_sinkhorn_reward": 0.7807352542877197,
      "rewards/forensics_strict_edge_reward": 0.1491037830710411,
      "rewards/forensics_tiou_reward": 0.6608700454235077,
      "step": 681
    },
    {
      "completion_length": 23.140625,
      "epoch": 2.0921658986175116,
      "grad_norm": 9.412311553955078,
      "kl": 0.4931640625,
      "learning_rate": 4.7615384615384613e-07,
      "loss": -0.0892,
      "reward": 1.3888286352157593,
      "reward_std": 0.2674441710114479,
      "reward_std_all": 0.8688530921936035,
      "rewards-max/forensics_sinkhorn_reward": 0.9872491955757141,
      "rewards-max/forensics_strict_edge_reward": 0.7939751744270325,
      "rewards-max/forensics_tiou_reward": 0.9970301985740662,
      "rewards/forensics_sinkhorn_reward": 0.759248286485672,
      "rewards/forensics_strict_edge_reward": 0.06961193308234215,
      "rewards/forensics_tiou_reward": 0.5599684119224548,
      "step": 682
    },
    {
      "completion_length": 24.4375,
      "epoch": 2.0952380952380953,
      "grad_norm": 6.765923976898193,
      "kl": 0.5556640625,
      "learning_rate": 4.753846153846154e-07,
      "loss": -0.1185,
      "reward": 1.4647470712661743,
      "reward_std": 0.47868241369724274,
      "reward_std_all": 0.8927721083164215,
      "rewards-max/forensics_sinkhorn_reward": 0.9784228205680847,
      "rewards-max/forensics_strict_edge_reward": 0.756708562374115,
      "rewards-max/forensics_tiou_reward": 0.9749158322811127,
      "rewards/forensics_sinkhorn_reward": 0.7605774700641632,
      "rewards/forensics_strict_edge_reward": 0.10269515216350555,
      "rewards/forensics_tiou_reward": 0.6014743745326996,
      "step": 683
    },
    {
      "completion_length": 23.859375,
      "epoch": 2.098310291858679,
      "grad_norm": 8.062091827392578,
      "kl": 0.451171875,
      "learning_rate": 4.7461538461538456e-07,
      "loss": -0.137,
      "reward": 1.32379150390625,
      "reward_std": 0.28762395679950714,
      "reward_std_all": 0.8986221849918365,
      "rewards-max/forensics_sinkhorn_reward": 0.9847451448440552,
      "rewards-max/forensics_strict_edge_reward": 0.4502716660499573,
      "rewards-max/forensics_tiou_reward": 0.9872325658798218,
      "rewards/forensics_sinkhorn_reward": 0.712684154510498,
      "rewards/forensics_strict_edge_reward": 0.05601485073566437,
      "rewards/forensics_tiou_reward": 0.555092453956604,
      "step": 684
    },
    {
      "completion_length": 23.15625,
      "epoch": 2.1013824884792625,
      "grad_norm": 8.389103889465332,
      "kl": 0.416015625,
      "learning_rate": 4.738461538461538e-07,
      "loss": -0.1721,
      "reward": 1.3429315686225891,
      "reward_std": 0.4243089556694031,
      "reward_std_all": 0.9367881119251251,
      "rewards-max/forensics_sinkhorn_reward": 0.9813966453075409,
      "rewards-max/forensics_strict_edge_reward": 0.49006423354148865,
      "rewards-max/forensics_tiou_reward": 0.9826200604438782,
      "rewards/forensics_sinkhorn_reward": 0.7113028764724731,
      "rewards/forensics_strict_edge_reward": 0.08495227247476578,
      "rewards/forensics_tiou_reward": 0.5466764569282532,
      "step": 685
    },
    {
      "completion_length": 25.03125,
      "epoch": 2.1044546850998462,
      "grad_norm": 26.415008544921875,
      "kl": 0.791015625,
      "learning_rate": 4.7307692307692304e-07,
      "loss": -0.0706,
      "reward": 1.4362032413482666,
      "reward_std": 0.28813034296035767,
      "reward_std_all": 0.8295526504516602,
      "rewards-max/forensics_sinkhorn_reward": 0.9824323356151581,
      "rewards-max/forensics_strict_edge_reward": 0.6236754059791565,
      "rewards-max/forensics_tiou_reward": 0.9807863533496857,
      "rewards/forensics_sinkhorn_reward": 0.7846222221851349,
      "rewards/forensics_strict_edge_reward": 0.056569114327430725,
      "rewards/forensics_tiou_reward": 0.5950120091438293,
      "step": 686
    },
    {
      "completion_length": 24.078125,
      "epoch": 2.10752688172043,
      "grad_norm": 10.497247695922852,
      "kl": 0.544921875,
      "learning_rate": 4.723076923076923e-07,
      "loss": -0.1327,
      "reward": 1.4842325448989868,
      "reward_std": 0.40677863359451294,
      "reward_std_all": 0.860677570104599,
      "rewards-max/forensics_sinkhorn_reward": 0.9863832294940948,
      "rewards-max/forensics_strict_edge_reward": 0.6966203600168228,
      "rewards-max/forensics_tiou_reward": 0.9909363687038422,
      "rewards/forensics_sinkhorn_reward": 0.7869687378406525,
      "rewards/forensics_strict_edge_reward": 0.06431625038385391,
      "rewards/forensics_tiou_reward": 0.6329476833343506,
      "step": 687
    },
    {
      "completion_length": 28.640625,
      "epoch": 2.110599078341014,
      "grad_norm": 8.991619110107422,
      "kl": 0.4033203125,
      "learning_rate": 4.715384615384615e-07,
      "loss": -0.0892,
      "reward": 1.427584946155548,
      "reward_std": 0.3806896507740021,
      "reward_std_all": 0.8754864037036896,
      "rewards-max/forensics_sinkhorn_reward": 0.9860863089561462,
      "rewards-max/forensics_strict_edge_reward": 0.7808269262313843,
      "rewards-max/forensics_tiou_reward": 0.9920969307422638,
      "rewards/forensics_sinkhorn_reward": 0.7157062888145447,
      "rewards/forensics_strict_edge_reward": 0.0956394299864769,
      "rewards/forensics_tiou_reward": 0.6162392497062683,
      "step": 688
    },
    {
      "completion_length": 24.96875,
      "epoch": 2.1136712749615976,
      "grad_norm": 9.25455093383789,
      "kl": 0.5556640625,
      "learning_rate": 4.7076923076923073e-07,
      "loss": -0.0502,
      "reward": 1.1436623334884644,
      "reward_std": 0.25015632808208466,
      "reward_std_all": 0.9809261560440063,
      "rewards-max/forensics_sinkhorn_reward": 0.9827893674373627,
      "rewards-max/forensics_strict_edge_reward": 0.4739979952573776,
      "rewards-max/forensics_tiou_reward": 0.9835222661495209,
      "rewards/forensics_sinkhorn_reward": 0.6204113960266113,
      "rewards/forensics_strict_edge_reward": 0.031930265948176384,
      "rewards/forensics_tiou_reward": 0.4913206994533539,
      "step": 689
    },
    {
      "completion_length": 26.46875,
      "epoch": 2.1167434715821813,
      "grad_norm": 8.282952308654785,
      "kl": 0.3955078125,
      "learning_rate": 4.6999999999999995e-07,
      "loss": -0.1367,
      "reward": 1.465498924255371,
      "reward_std": 0.35276012122631073,
      "reward_std_all": 0.8428711295127869,
      "rewards-max/forensics_sinkhorn_reward": 0.9865416288375854,
      "rewards-max/forensics_strict_edge_reward": 0.7229099273681641,
      "rewards-max/forensics_tiou_reward": 0.9930181801319122,
      "rewards/forensics_sinkhorn_reward": 0.7789054214954376,
      "rewards/forensics_strict_edge_reward": 0.08003845810890198,
      "rewards/forensics_tiou_reward": 0.6065550446510315,
      "step": 690
    },
    {
      "completion_length": 22.109375,
      "epoch": 2.119815668202765,
      "grad_norm": 7.952264785766602,
      "kl": 0.5234375,
      "learning_rate": 4.692307692307692e-07,
      "loss": -0.0416,
      "reward": 1.5566022396087646,
      "reward_std": 0.3646991550922394,
      "reward_std_all": 0.8133472502231598,
      "rewards-max/forensics_sinkhorn_reward": 0.9839889705181122,
      "rewards-max/forensics_strict_edge_reward": 0.711992472410202,
      "rewards-max/forensics_tiou_reward": 0.9845450520515442,
      "rewards/forensics_sinkhorn_reward": 0.8174785077571869,
      "rewards/forensics_strict_edge_reward": 0.10121342539787292,
      "rewards/forensics_tiou_reward": 0.6379103064537048,
      "step": 691
    },
    {
      "completion_length": 25.53125,
      "epoch": 2.122887864823349,
      "grad_norm": 9.022695541381836,
      "kl": 0.5439453125,
      "learning_rate": 4.684615384615385e-07,
      "loss": -0.1251,
      "reward": 1.338828682899475,
      "reward_std": 0.33534862101078033,
      "reward_std_all": 0.9198825061321259,
      "rewards-max/forensics_sinkhorn_reward": 0.9797752499580383,
      "rewards-max/forensics_strict_edge_reward": 0.7835284769535065,
      "rewards-max/forensics_tiou_reward": 0.9780256152153015,
      "rewards/forensics_sinkhorn_reward": 0.7144719064235687,
      "rewards/forensics_strict_edge_reward": 0.06531047448515892,
      "rewards/forensics_tiou_reward": 0.5590463280677795,
      "step": 692
    },
    {
      "completion_length": 25.21875,
      "epoch": 2.1259600614439322,
      "grad_norm": 10.178776741027832,
      "kl": 0.4462890625,
      "learning_rate": 4.6769230769230764e-07,
      "loss": -0.1332,
      "reward": 1.5010111927986145,
      "reward_std": 0.4636109322309494,
      "reward_std_all": 0.8903307616710663,
      "rewards-max/forensics_sinkhorn_reward": 0.9891026318073273,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.7607974410057068,
      "rewards/forensics_strict_edge_reward": 0.13732727617025375,
      "rewards/forensics_tiou_reward": 0.6028863787651062,
      "step": 693
    },
    {
      "completion_length": 26.78125,
      "epoch": 2.129032258064516,
      "grad_norm": 8.501218795776367,
      "kl": 0.451171875,
      "learning_rate": 4.669230769230769e-07,
      "loss": -0.1188,
      "reward": 1.173111915588379,
      "reward_std": 0.35084068775177,
      "reward_std_all": 0.9892472624778748,
      "rewards-max/forensics_sinkhorn_reward": 0.9756044745445251,
      "rewards-max/forensics_strict_edge_reward": 0.5038763880729675,
      "rewards-max/forensics_tiou_reward": 0.9661096930503845,
      "rewards/forensics_sinkhorn_reward": 0.6391905546188354,
      "rewards/forensics_strict_edge_reward": 0.06703827902674675,
      "rewards/forensics_tiou_reward": 0.4668830931186676,
      "step": 694
    },
    {
      "completion_length": 24.375,
      "epoch": 2.1321044546851,
      "grad_norm": 8.024014472961426,
      "kl": 0.4619140625,
      "learning_rate": 4.661538461538461e-07,
      "loss": -0.0875,
      "reward": 1.3857786655426025,
      "reward_std": 0.3062521368265152,
      "reward_std_all": 0.8816676139831543,
      "rewards-max/forensics_sinkhorn_reward": 0.9826421439647675,
      "rewards-max/forensics_strict_edge_reward": 0.882964164018631,
      "rewards-max/forensics_tiou_reward": 0.9855519533157349,
      "rewards/forensics_sinkhorn_reward": 0.7238539755344391,
      "rewards/forensics_strict_edge_reward": 0.08947771042585373,
      "rewards/forensics_tiou_reward": 0.5724470019340515,
      "step": 695
    },
    {
      "completion_length": 24.96875,
      "epoch": 2.1351766513056836,
      "grad_norm": 8.895771026611328,
      "kl": 0.47265625,
      "learning_rate": 4.653846153846154e-07,
      "loss": 0.0014,
      "reward": 1.2233731746673584,
      "reward_std": 0.28928936272859573,
      "reward_std_all": 0.9379670023918152,
      "rewards-max/forensics_sinkhorn_reward": 0.9755114018917084,
      "rewards-max/forensics_strict_edge_reward": 0.5303425639867783,
      "rewards-max/forensics_tiou_reward": 0.9648600518703461,
      "rewards/forensics_sinkhorn_reward": 0.6688485741615295,
      "rewards/forensics_strict_edge_reward": 0.05884680524468422,
      "rewards/forensics_tiou_reward": 0.49567772448062897,
      "step": 696
    },
    {
      "completion_length": 25.734375,
      "epoch": 2.1382488479262673,
      "grad_norm": 7.236288547515869,
      "kl": 0.5166015625,
      "learning_rate": 4.646153846153846e-07,
      "loss": -0.1049,
      "reward": 1.367587387561798,
      "reward_std": 0.28115780651569366,
      "reward_std_all": 0.8607342839241028,
      "rewards-max/forensics_sinkhorn_reward": 0.9661993086338043,
      "rewards-max/forensics_strict_edge_reward": 0.23545214720070362,
      "rewards-max/forensics_tiou_reward": 0.944582462310791,
      "rewards/forensics_sinkhorn_reward": 0.7661848068237305,
      "rewards/forensics_strict_edge_reward": 0.03170101344585419,
      "rewards/forensics_tiou_reward": 0.5697014629840851,
      "step": 697
    },
    {
      "completion_length": 23.328125,
      "epoch": 2.141321044546851,
      "grad_norm": 6.68424654006958,
      "kl": 0.521484375,
      "learning_rate": 4.638461538461538e-07,
      "loss": -0.0947,
      "reward": 1.2936864495277405,
      "reward_std": 0.41032999753952026,
      "reward_std_all": 0.9590333700180054,
      "rewards-max/forensics_sinkhorn_reward": 0.96942999958992,
      "rewards-max/forensics_strict_edge_reward": 0.6317985653877258,
      "rewards-max/forensics_tiou_reward": 0.9443305432796478,
      "rewards/forensics_sinkhorn_reward": 0.7148416340351105,
      "rewards/forensics_strict_edge_reward": 0.05458617024123669,
      "rewards/forensics_tiou_reward": 0.5242585241794586,
      "step": 698
    },
    {
      "completion_length": 25.578125,
      "epoch": 2.144393241167435,
      "grad_norm": 8.92497444152832,
      "kl": 0.4990234375,
      "learning_rate": 4.6307692307692303e-07,
      "loss": -0.075,
      "reward": 1.5427055954933167,
      "reward_std": 0.36080189049243927,
      "reward_std_all": 0.8095324635505676,
      "rewards-max/forensics_sinkhorn_reward": 0.9841428101062775,
      "rewards-max/forensics_strict_edge_reward": 0.7187938392162323,
      "rewards-max/forensics_tiou_reward": 0.9866903722286224,
      "rewards/forensics_sinkhorn_reward": 0.8163878917694092,
      "rewards/forensics_strict_edge_reward": 0.07780388370156288,
      "rewards/forensics_tiou_reward": 0.6485138535499573,
      "step": 699
    },
    {
      "completion_length": 22.1875,
      "epoch": 2.1474654377880182,
      "grad_norm": 9.873583793640137,
      "kl": 0.4287109375,
      "learning_rate": 4.623076923076923e-07,
      "loss": -0.0677,
      "reward": 1.3216928243637085,
      "reward_std": 0.4731205254793167,
      "reward_std_all": 0.9696642160415649,
      "rewards-max/forensics_sinkhorn_reward": 0.9852422177791595,
      "rewards-max/forensics_strict_edge_reward": 0.5747906565666199,
      "rewards-max/forensics_tiou_reward": 0.9905986487865448,
      "rewards/forensics_sinkhorn_reward": 0.7236037850379944,
      "rewards/forensics_strict_edge_reward": 0.07327678799629211,
      "rewards/forensics_tiou_reward": 0.5248121917247772,
      "step": 700
    },
    {
      "completion_length": 21.9375,
      "epoch": 2.150537634408602,
      "grad_norm": 6.630171775817871,
      "kl": 0.31689453125,
      "learning_rate": 4.6153846153846156e-07,
      "loss": -0.0549,
      "reward": 1.4332979917526245,
      "reward_std": 0.3729725033044815,
      "reward_std_all": 0.8681196570396423,
      "rewards-max/forensics_sinkhorn_reward": 0.9777340292930603,
      "rewards-max/forensics_strict_edge_reward": 0.5755647718906403,
      "rewards-max/forensics_tiou_reward": 0.9730607569217682,
      "rewards/forensics_sinkhorn_reward": 0.7720547318458557,
      "rewards/forensics_strict_edge_reward": 0.08170418068766594,
      "rewards/forensics_tiou_reward": 0.5795390605926514,
      "step": 701
    },
    {
      "completion_length": 26.6875,
      "epoch": 2.153609831029186,
      "grad_norm": 8.255520820617676,
      "kl": 0.51171875,
      "learning_rate": 4.607692307692307e-07,
      "loss": -0.0774,
      "reward": 1.3541251420974731,
      "reward_std": 0.3344305604696274,
      "reward_std_all": 0.8993154466152191,
      "rewards-max/forensics_sinkhorn_reward": 0.986379861831665,
      "rewards-max/forensics_strict_edge_reward": 0.8351600170135498,
      "rewards-max/forensics_tiou_reward": 0.9944777488708496,
      "rewards/forensics_sinkhorn_reward": 0.6993331015110016,
      "rewards/forensics_strict_edge_reward": 0.09608705714344978,
      "rewards/forensics_tiou_reward": 0.5587048828601837,
      "step": 702
    },
    {
      "completion_length": 28.515625,
      "epoch": 2.1566820276497696,
      "grad_norm": 9.8878812789917,
      "kl": 0.416015625,
      "learning_rate": 4.6e-07,
      "loss": -0.0853,
      "reward": 1.1287569999694824,
      "reward_std": 0.29555609822273254,
      "reward_std_all": 0.9816561937332153,
      "rewards-max/forensics_sinkhorn_reward": 0.9783493280410767,
      "rewards-max/forensics_strict_edge_reward": 0.5684406757354736,
      "rewards-max/forensics_tiou_reward": 0.9730542600154877,
      "rewards/forensics_sinkhorn_reward": 0.6352753937244415,
      "rewards/forensics_strict_edge_reward": 0.03436149097979069,
      "rewards/forensics_tiou_reward": 0.45912010967731476,
      "step": 703
    },
    {
      "completion_length": 25.328125,
      "epoch": 2.1597542242703534,
      "grad_norm": 9.737817764282227,
      "kl": 0.466796875,
      "learning_rate": 4.592307692307692e-07,
      "loss": -0.0264,
      "reward": 1.2705861926078796,
      "reward_std": 0.3593057245016098,
      "reward_std_all": 0.9377983212471008,
      "rewards-max/forensics_sinkhorn_reward": 0.9824605584144592,
      "rewards-max/forensics_strict_edge_reward": 0.5853398740291595,
      "rewards-max/forensics_tiou_reward": 0.9787168800830841,
      "rewards/forensics_sinkhorn_reward": 0.6850345730781555,
      "rewards/forensics_strict_edge_reward": 0.04843073524534702,
      "rewards/forensics_tiou_reward": 0.5371208190917969,
      "step": 704
    },
    {
      "completion_length": 23.390625,
      "epoch": 2.162826420890937,
      "grad_norm": 8.388933181762695,
      "kl": 0.357421875,
      "learning_rate": 4.5846153846153847e-07,
      "loss": -0.1238,
      "reward": 1.2390645742416382,
      "reward_std": 0.4531307816505432,
      "reward_std_all": 0.9997736215591431,
      "rewards-max/forensics_sinkhorn_reward": 0.9816287755966187,
      "rewards-max/forensics_strict_edge_reward": 0.756708562374115,
      "rewards-max/forensics_tiou_reward": 0.9826121628284454,
      "rewards/forensics_sinkhorn_reward": 0.6613086760044098,
      "rewards/forensics_strict_edge_reward": 0.07294481247663498,
      "rewards/forensics_tiou_reward": 0.504811018705368,
      "step": 705
    },
    {
      "completion_length": 23.390625,
      "epoch": 2.165898617511521,
      "grad_norm": 8.603558540344238,
      "kl": 0.373046875,
      "learning_rate": 4.5769230769230763e-07,
      "loss": -0.1097,
      "reward": 1.2352584600448608,
      "reward_std": 0.27097440510988235,
      "reward_std_all": 0.9362012445926666,
      "rewards-max/forensics_sinkhorn_reward": 0.9885624945163727,
      "rewards-max/forensics_strict_edge_reward": 0.6720768958330154,
      "rewards-max/forensics_tiou_reward": 0.9983647763729095,
      "rewards/forensics_sinkhorn_reward": 0.7056559026241302,
      "rewards/forensics_strict_edge_reward": 0.04972602054476738,
      "rewards/forensics_tiou_reward": 0.4798765480518341,
      "step": 706
    },
    {
      "completion_length": 22.0625,
      "epoch": 2.1689708141321047,
      "grad_norm": 9.390406608581543,
      "kl": 0.4814453125,
      "learning_rate": 4.569230769230769e-07,
      "loss": -0.0527,
      "reward": 1.4500449895858765,
      "reward_std": 0.25489185750484467,
      "reward_std_all": 0.8178008794784546,
      "rewards-max/forensics_sinkhorn_reward": 0.9728746116161346,
      "rewards-max/forensics_strict_edge_reward": 0.46164050698280334,
      "rewards-max/forensics_tiou_reward": 0.9577442407608032,
      "rewards/forensics_sinkhorn_reward": 0.7947331070899963,
      "rewards/forensics_strict_edge_reward": 0.04644072242081165,
      "rewards/forensics_tiou_reward": 0.6088712215423584,
      "step": 707
    },
    {
      "completion_length": 24.0625,
      "epoch": 2.172043010752688,
      "grad_norm": 8.67657470703125,
      "kl": 0.580078125,
      "learning_rate": 4.561538461538461e-07,
      "loss": -0.122,
      "reward": 1.4587138891220093,
      "reward_std": 0.2686241865158081,
      "reward_std_all": 0.8336107730865479,
      "rewards-max/forensics_sinkhorn_reward": 0.9797287881374359,
      "rewards-max/forensics_strict_edge_reward": 0.5383405238389969,
      "rewards-max/forensics_tiou_reward": 0.9755989015102386,
      "rewards/forensics_sinkhorn_reward": 0.7885033786296844,
      "rewards/forensics_strict_edge_reward": 0.07415442168712616,
      "rewards/forensics_tiou_reward": 0.5960560441017151,
      "step": 708
    },
    {
      "completion_length": 25.09375,
      "epoch": 2.175115207373272,
      "grad_norm": 8.619145393371582,
      "kl": 0.53515625,
      "learning_rate": 4.553846153846154e-07,
      "loss": -0.1172,
      "reward": 1.284562885761261,
      "reward_std": 0.23533181846141815,
      "reward_std_all": 0.8972225785255432,
      "rewards-max/forensics_sinkhorn_reward": 0.9774394929409027,
      "rewards-max/forensics_strict_edge_reward": 0.2981753796339035,
      "rewards-max/forensics_tiou_reward": 0.9638663530349731,
      "rewards/forensics_sinkhorn_reward": 0.7307357788085938,
      "rewards/forensics_strict_edge_reward": 0.033589474856853485,
      "rewards/forensics_tiou_reward": 0.5202375650405884,
      "step": 709
    },
    {
      "completion_length": 22.84375,
      "epoch": 2.1781874039938556,
      "grad_norm": 9.270590782165527,
      "kl": 0.3818359375,
      "learning_rate": 4.546153846153846e-07,
      "loss": -0.0798,
      "reward": 1.4060571789741516,
      "reward_std": 0.2799892947077751,
      "reward_std_all": 0.8753357231616974,
      "rewards-max/forensics_sinkhorn_reward": 0.9738514423370361,
      "rewards-max/forensics_strict_edge_reward": 0.594437800347805,
      "rewards-max/forensics_tiou_reward": 0.9602330327033997,
      "rewards/forensics_sinkhorn_reward": 0.7322856187820435,
      "rewards/forensics_strict_edge_reward": 0.08868906274437904,
      "rewards/forensics_tiou_reward": 0.5850825011730194,
      "step": 710
    },
    {
      "completion_length": 25.65625,
      "epoch": 2.1812596006144394,
      "grad_norm": 8.915480613708496,
      "kl": 0.4375,
      "learning_rate": 4.538461538461538e-07,
      "loss": -0.0708,
      "reward": 1.491503357887268,
      "reward_std": 0.2113036923110485,
      "reward_std_all": 0.7952800989151001,
      "rewards-max/forensics_sinkhorn_reward": 0.9827666878700256,
      "rewards-max/forensics_strict_edge_reward": 0.8582656681537628,
      "rewards-max/forensics_tiou_reward": 0.9840187132358551,
      "rewards/forensics_sinkhorn_reward": 0.7637803852558136,
      "rewards/forensics_strict_edge_reward": 0.09033144637942314,
      "rewards/forensics_tiou_reward": 0.6373915672302246,
      "step": 711
    },
    {
      "completion_length": 27.046875,
      "epoch": 2.184331797235023,
      "grad_norm": 11.52426815032959,
      "kl": 0.46875,
      "learning_rate": 4.5307692307692307e-07,
      "loss": -0.1307,
      "reward": 1.2490744292736053,
      "reward_std": 0.31289346516132355,
      "reward_std_all": 0.9325122237205505,
      "rewards-max/forensics_sinkhorn_reward": 0.979963481426239,
      "rewards-max/forensics_strict_edge_reward": 0.48878854513168335,
      "rewards-max/forensics_tiou_reward": 0.9789462387561798,
      "rewards/forensics_sinkhorn_reward": 0.7003623247146606,
      "rewards/forensics_strict_edge_reward": 0.057790257036685944,
      "rewards/forensics_tiou_reward": 0.49092186987400055,
      "step": 712
    },
    {
      "completion_length": 22.296875,
      "epoch": 2.187403993855607,
      "grad_norm": 9.813535690307617,
      "kl": 0.5234375,
      "learning_rate": 4.523076923076923e-07,
      "loss": -0.0011,
      "reward": 1.3998138308525085,
      "reward_std": 0.3665712922811508,
      "reward_std_all": 0.8912535309791565,
      "rewards-max/forensics_sinkhorn_reward": 0.9818027019500732,
      "rewards-max/forensics_strict_edge_reward": 0.6778962016105652,
      "rewards-max/forensics_tiou_reward": 0.9824877977371216,
      "rewards/forensics_sinkhorn_reward": 0.7379098534584045,
      "rewards/forensics_strict_edge_reward": 0.08589251339435577,
      "rewards/forensics_tiou_reward": 0.5760115087032318,
      "step": 713
    },
    {
      "completion_length": 25.15625,
      "epoch": 2.1904761904761907,
      "grad_norm": 9.7713623046875,
      "kl": 0.4853515625,
      "learning_rate": 4.5153846153846155e-07,
      "loss": -0.1145,
      "reward": 1.4883546829223633,
      "reward_std": 0.43121929466724396,
      "reward_std_all": 0.8679312169551849,
      "rewards-max/forensics_sinkhorn_reward": 0.9795077741146088,
      "rewards-max/forensics_strict_edge_reward": 0.784106969833374,
      "rewards-max/forensics_tiou_reward": 0.9784001410007477,
      "rewards/forensics_sinkhorn_reward": 0.7800812125205994,
      "rewards/forensics_strict_edge_reward": 0.11288338899612427,
      "rewards/forensics_tiou_reward": 0.5953900516033173,
      "step": 714
    },
    {
      "completion_length": 23.65625,
      "epoch": 2.193548387096774,
      "grad_norm": 9.631152153015137,
      "kl": 0.5263671875,
      "learning_rate": 4.507692307692307e-07,
      "loss": -0.075,
      "reward": 1.2273595333099365,
      "reward_std": 0.3882693648338318,
      "reward_std_all": 0.9851557016372681,
      "rewards-max/forensics_sinkhorn_reward": 0.9835556149482727,
      "rewards-max/forensics_strict_edge_reward": 0.882964164018631,
      "rewards-max/forensics_tiou_reward": 0.9881386160850525,
      "rewards/forensics_sinkhorn_reward": 0.6843280792236328,
      "rewards/forensics_strict_edge_reward": 0.07370702922344208,
      "rewards/forensics_tiou_reward": 0.46932438015937805,
      "step": 715
    },
    {
      "completion_length": 28.0,
      "epoch": 2.196620583717358,
      "grad_norm": 9.377581596374512,
      "kl": 0.5,
      "learning_rate": 4.5e-07,
      "loss": -0.1344,
      "reward": 1.4593900442123413,
      "reward_std": 0.349305123090744,
      "reward_std_all": 0.8509940207004547,
      "rewards-max/forensics_sinkhorn_reward": 0.9862580001354218,
      "rewards-max/forensics_strict_edge_reward": 0.8120969831943512,
      "rewards-max/forensics_tiou_reward": 0.9930903911590576,
      "rewards/forensics_sinkhorn_reward": 0.7491150498390198,
      "rewards/forensics_strict_edge_reward": 0.09997500851750374,
      "rewards/forensics_tiou_reward": 0.6103000640869141,
      "step": 716
    },
    {
      "completion_length": 26.75,
      "epoch": 2.1996927803379416,
      "grad_norm": 12.961669921875,
      "kl": 0.5322265625,
      "learning_rate": 4.492307692307692e-07,
      "loss": -0.073,
      "reward": 1.4987863898277283,
      "reward_std": 0.2994988113641739,
      "reward_std_all": 0.8275758326053619,
      "rewards-max/forensics_sinkhorn_reward": 0.9865580201148987,
      "rewards-max/forensics_strict_edge_reward": 0.8582656681537628,
      "rewards-max/forensics_tiou_reward": 0.9955674409866333,
      "rewards/forensics_sinkhorn_reward": 0.7652303576469421,
      "rewards/forensics_strict_edge_reward": 0.09122231230139732,
      "rewards/forensics_tiou_reward": 0.6423337459564209,
      "step": 717
    },
    {
      "completion_length": 25.390625,
      "epoch": 2.2027649769585254,
      "grad_norm": 9.203060150146484,
      "kl": 0.578125,
      "learning_rate": 4.4846153846153846e-07,
      "loss": -0.0563,
      "reward": 1.3116015791893005,
      "reward_std": 0.32169902324676514,
      "reward_std_all": 0.9245288074016571,
      "rewards-max/forensics_sinkhorn_reward": 0.9748094975948334,
      "rewards-max/forensics_strict_edge_reward": 0.6660739481449127,
      "rewards-max/forensics_tiou_reward": 0.9649489521980286,
      "rewards/forensics_sinkhorn_reward": 0.7198043763637543,
      "rewards/forensics_strict_edge_reward": 0.05332907661795616,
      "rewards/forensics_tiou_reward": 0.5384681820869446,
      "step": 718
    },
    {
      "completion_length": 23.59375,
      "epoch": 2.205837173579109,
      "grad_norm": 9.186872482299805,
      "kl": 0.46875,
      "learning_rate": 4.476923076923076e-07,
      "loss": -0.0763,
      "reward": 1.6028060913085938,
      "reward_std": 0.39581039547920227,
      "reward_std_all": 0.7972365319728851,
      "rewards-max/forensics_sinkhorn_reward": 0.9845390021800995,
      "rewards-max/forensics_strict_edge_reward": 0.6335065066814423,
      "rewards-max/forensics_tiou_reward": 0.9868107438087463,
      "rewards/forensics_sinkhorn_reward": 0.8235418200492859,
      "rewards/forensics_strict_edge_reward": 0.09718738496303558,
      "rewards/forensics_tiou_reward": 0.6820769309997559,
      "step": 719
    },
    {
      "completion_length": 24.375,
      "epoch": 2.208909370199693,
      "grad_norm": 9.821907043457031,
      "kl": 0.5146484375,
      "learning_rate": 4.469230769230769e-07,
      "loss": -0.0654,
      "reward": 1.6312814354896545,
      "reward_std": 0.37476886808872223,
      "reward_std_all": 0.7866117358207703,
      "rewards-max/forensics_sinkhorn_reward": 0.9847068190574646,
      "rewards-max/forensics_strict_edge_reward": 0.5736877918243408,
      "rewards-max/forensics_tiou_reward": 0.9897494316101074,
      "rewards/forensics_sinkhorn_reward": 0.8487509191036224,
      "rewards/forensics_strict_edge_reward": 0.06007879972457886,
      "rewards/forensics_tiou_reward": 0.7224516868591309,
      "step": 720
    },
    {
      "completion_length": 23.796875,
      "epoch": 2.2119815668202767,
      "grad_norm": 9.341938018798828,
      "kl": 0.419921875,
      "learning_rate": 4.4615384615384615e-07,
      "loss": -0.0682,
      "reward": 1.5190943479537964,
      "reward_std": 0.4125650227069855,
      "reward_std_all": 0.8432237505912781,
      "rewards-max/forensics_sinkhorn_reward": 0.9856599271297455,
      "rewards-max/forensics_strict_edge_reward": 0.863698273897171,
      "rewards-max/forensics_tiou_reward": 0.9916059374809265,
      "rewards/forensics_sinkhorn_reward": 0.7746969163417816,
      "rewards/forensics_strict_edge_reward": 0.11023428663611412,
      "rewards/forensics_tiou_reward": 0.6341631710529327,
      "step": 721
    },
    {
      "completion_length": 24.328125,
      "epoch": 2.21505376344086,
      "grad_norm": 10.078389167785645,
      "kl": 0.3818359375,
      "learning_rate": 4.4538461538461537e-07,
      "loss": -0.0222,
      "reward": 1.5418944358825684,
      "reward_std": 0.4663079082965851,
      "reward_std_all": 0.8660881519317627,
      "rewards-max/forensics_sinkhorn_reward": 0.9808447360992432,
      "rewards-max/forensics_strict_edge_reward": 0.6863083839416504,
      "rewards-max/forensics_tiou_reward": 0.9802402853965759,
      "rewards/forensics_sinkhorn_reward": 0.7860659956932068,
      "rewards/forensics_strict_edge_reward": 0.10945180431008339,
      "rewards/forensics_tiou_reward": 0.6463765203952789,
      "step": 722
    },
    {
      "completion_length": 25.40625,
      "epoch": 2.218125960061444,
      "grad_norm": 9.868444442749023,
      "kl": 0.4365234375,
      "learning_rate": 4.4461538461538463e-07,
      "loss": -0.0409,
      "reward": 1.3942186832427979,
      "reward_std": 0.49733616411685944,
      "reward_std_all": 0.9231811165809631,
      "rewards-max/forensics_sinkhorn_reward": 0.9828250408172607,
      "rewards-max/forensics_strict_edge_reward": 0.6054387986660004,
      "rewards-max/forensics_tiou_reward": 0.9832195043563843,
      "rewards/forensics_sinkhorn_reward": 0.7459093332290649,
      "rewards/forensics_strict_edge_reward": 0.08163043111562729,
      "rewards/forensics_tiou_reward": 0.5666788816452026,
      "step": 723
    },
    {
      "completion_length": 24.671875,
      "epoch": 2.2211981566820276,
      "grad_norm": 9.492467880249023,
      "kl": 0.421875,
      "learning_rate": 4.438461538461538e-07,
      "loss": -0.0115,
      "reward": 1.459094226360321,
      "reward_std": 0.33333054184913635,
      "reward_std_all": 0.8512110114097595,
      "rewards-max/forensics_sinkhorn_reward": 0.9888028204441071,
      "rewards-max/forensics_strict_edge_reward": 0.8704091012477875,
      "rewards-max/forensics_tiou_reward": 0.9993914365768433,
      "rewards/forensics_sinkhorn_reward": 0.7628014981746674,
      "rewards/forensics_strict_edge_reward": 0.08990712836384773,
      "rewards/forensics_tiou_reward": 0.6063855886459351,
      "step": 724
    },
    {
      "completion_length": 28.5,
      "epoch": 2.2242703533026114,
      "grad_norm": 9.342171669006348,
      "kl": 0.619140625,
      "learning_rate": 4.4307692307692306e-07,
      "loss": -0.0559,
      "reward": 1.3082412481307983,
      "reward_std": 0.3248424157500267,
      "reward_std_all": 0.9129742980003357,
      "rewards-max/forensics_sinkhorn_reward": 0.9787891209125519,
      "rewards-max/forensics_strict_edge_reward": 0.5484400391578674,
      "rewards-max/forensics_tiou_reward": 0.9712516963481903,
      "rewards/forensics_sinkhorn_reward": 0.69170281291008,
      "rewards/forensics_strict_edge_reward": 0.06976715661585331,
      "rewards/forensics_tiou_reward": 0.5467712134122849,
      "step": 725
    },
    {
      "completion_length": 25.71875,
      "epoch": 2.227342549923195,
      "grad_norm": 8.886561393737793,
      "kl": 0.42578125,
      "learning_rate": 4.423076923076923e-07,
      "loss": -0.1073,
      "reward": 1.640869915485382,
      "reward_std": 0.25671374797821045,
      "reward_std_all": 0.7293732762336731,
      "rewards-max/forensics_sinkhorn_reward": 0.9747950732707977,
      "rewards-max/forensics_strict_edge_reward": 0.6002442985773087,
      "rewards-max/forensics_tiou_reward": 0.9652150869369507,
      "rewards/forensics_sinkhorn_reward": 0.8628271520137787,
      "rewards/forensics_strict_edge_reward": 0.06888183578848839,
      "rewards/forensics_tiou_reward": 0.7091608345508575,
      "step": 726
    },
    {
      "completion_length": 28.15625,
      "epoch": 2.230414746543779,
      "grad_norm": 8.512308120727539,
      "kl": 0.4599609375,
      "learning_rate": 4.4153846153846154e-07,
      "loss": -0.0795,
      "reward": 1.4512333273887634,
      "reward_std": 0.2973150312900543,
      "reward_std_all": 0.8300958275794983,
      "rewards-max/forensics_sinkhorn_reward": 0.9878605008125305,
      "rewards-max/forensics_strict_edge_reward": 0.5172730684280396,
      "rewards-max/forensics_tiou_reward": 0.9962085485458374,
      "rewards/forensics_sinkhorn_reward": 0.7464653849601746,
      "rewards/forensics_strict_edge_reward": 0.0737716555595398,
      "rewards/forensics_tiou_reward": 0.6309963762760162,
      "step": 727
    },
    {
      "completion_length": 25.34375,
      "epoch": 2.2334869431643627,
      "grad_norm": 8.942986488342285,
      "kl": 0.4921875,
      "learning_rate": 4.407692307692307e-07,
      "loss": -0.054,
      "reward": 1.2751991152763367,
      "reward_std": 0.315801277756691,
      "reward_std_all": 0.9239278435707092,
      "rewards-max/forensics_sinkhorn_reward": 0.9752146005630493,
      "rewards-max/forensics_strict_edge_reward": 0.478833943605423,
      "rewards-max/forensics_tiou_reward": 0.9651684165000916,
      "rewards/forensics_sinkhorn_reward": 0.6818685233592987,
      "rewards/forensics_strict_edge_reward": 0.0383031889796257,
      "rewards/forensics_tiou_reward": 0.5550273656845093,
      "step": 728
    },
    {
      "completion_length": 29.828125,
      "epoch": 2.236559139784946,
      "grad_norm": 10.37962532043457,
      "kl": 0.4208984375,
      "learning_rate": 4.3999999999999997e-07,
      "loss": -0.0839,
      "reward": 1.203674554824829,
      "reward_std": 0.45651693642139435,
      "reward_std_all": 1.0209277272224426,
      "rewards-max/forensics_sinkhorn_reward": 0.979596883058548,
      "rewards-max/forensics_strict_edge_reward": 0.757034033536911,
      "rewards-max/forensics_tiou_reward": 0.9760035574436188,
      "rewards/forensics_sinkhorn_reward": 0.5644891560077667,
      "rewards/forensics_strict_edge_reward": 0.11497131362557411,
      "rewards/forensics_tiou_reward": 0.524214118719101,
      "step": 729
    },
    {
      "completion_length": 27.9375,
      "epoch": 2.23963133640553,
      "grad_norm": 8.763976097106934,
      "kl": 0.45703125,
      "learning_rate": 4.3923076923076924e-07,
      "loss": -0.1339,
      "reward": 1.4060528874397278,
      "reward_std": 0.27699314057826996,
      "reward_std_all": 0.8641302287578583,
      "rewards-max/forensics_sinkhorn_reward": 0.9785547256469727,
      "rewards-max/forensics_strict_edge_reward": 0.6711390465497971,
      "rewards-max/forensics_tiou_reward": 0.9727406799793243,
      "rewards/forensics_sinkhorn_reward": 0.7533847987651825,
      "rewards/forensics_strict_edge_reward": 0.10199281200766563,
      "rewards/forensics_tiou_reward": 0.5506752282381058,
      "step": 730
    },
    {
      "completion_length": 24.234375,
      "epoch": 2.2427035330261136,
      "grad_norm": 9.690962791442871,
      "kl": 0.51953125,
      "learning_rate": 4.3846153846153845e-07,
      "loss": -0.1307,
      "reward": 1.4288026094436646,
      "reward_std": 0.2916707694530487,
      "reward_std_all": 0.8444800972938538,
      "rewards-max/forensics_sinkhorn_reward": 0.9578866958618164,
      "rewards-max/forensics_strict_edge_reward": 0.5527456998825073,
      "rewards-max/forensics_tiou_reward": 0.9270589649677277,
      "rewards/forensics_sinkhorn_reward": 0.7854350507259369,
      "rewards/forensics_strict_edge_reward": 0.05168730020523071,
      "rewards/forensics_tiou_reward": 0.5916802287101746,
      "step": 731
    },
    {
      "completion_length": 24.90625,
      "epoch": 2.2457757296466974,
      "grad_norm": 10.064297676086426,
      "kl": 0.74609375,
      "learning_rate": 4.376923076923077e-07,
      "loss": -0.0173,
      "reward": 1.268139362335205,
      "reward_std": 0.33240120112895966,
      "reward_std_all": 0.9443095922470093,
      "rewards-max/forensics_sinkhorn_reward": 0.9793305993080139,
      "rewards-max/forensics_strict_edge_reward": 0.5021075010299683,
      "rewards-max/forensics_tiou_reward": 0.9768244326114655,
      "rewards/forensics_sinkhorn_reward": 0.70059934258461,
      "rewards/forensics_strict_edge_reward": 0.05278895050287247,
      "rewards/forensics_tiou_reward": 0.5147510766983032,
      "step": 732
    },
    {
      "completion_length": 27.328125,
      "epoch": 2.248847926267281,
      "grad_norm": 9.280224800109863,
      "kl": 0.521484375,
      "learning_rate": 4.369230769230769e-07,
      "loss": -0.1534,
      "reward": 1.4790154099464417,
      "reward_std": 0.43755707144737244,
      "reward_std_all": 0.872821182012558,
      "rewards-max/forensics_sinkhorn_reward": 0.9879035651683807,
      "rewards-max/forensics_strict_edge_reward": 0.7837068438529968,
      "rewards-max/forensics_tiou_reward": 0.9964079856872559,
      "rewards/forensics_sinkhorn_reward": 0.7655705809593201,
      "rewards/forensics_strict_edge_reward": 0.13158930465579033,
      "rewards/forensics_tiou_reward": 0.5818555355072021,
      "step": 733
    },
    {
      "completion_length": 25.046875,
      "epoch": 2.251920122887865,
      "grad_norm": 10.972970008850098,
      "kl": 0.5126953125,
      "learning_rate": 4.3615384615384614e-07,
      "loss": -0.1039,
      "reward": 1.4457436800003052,
      "reward_std": 0.44066187739372253,
      "reward_std_all": 0.8950771987438202,
      "rewards-max/forensics_sinkhorn_reward": 0.9854199588298798,
      "rewards-max/forensics_strict_edge_reward": 0.7689784169197083,
      "rewards-max/forensics_tiou_reward": 0.9916107654571533,
      "rewards/forensics_sinkhorn_reward": 0.7190842926502228,
      "rewards/forensics_strict_edge_reward": 0.12368382513523102,
      "rewards/forensics_tiou_reward": 0.602975606918335,
      "step": 734
    },
    {
      "completion_length": 26.484375,
      "epoch": 2.2549923195084487,
      "grad_norm": 39.71937561035156,
      "kl": 1.318359375,
      "learning_rate": 4.3538461538461536e-07,
      "loss": -0.0085,
      "reward": 1.335272490978241,
      "reward_std": 0.34112709760665894,
      "reward_std_all": 0.9014963507652283,
      "rewards-max/forensics_sinkhorn_reward": 0.9833152890205383,
      "rewards-max/forensics_strict_edge_reward": 0.5848344266414642,
      "rewards-max/forensics_tiou_reward": 0.9843896925449371,
      "rewards/forensics_sinkhorn_reward": 0.7038902044296265,
      "rewards/forensics_strict_edge_reward": 0.0816219225525856,
      "rewards/forensics_tiou_reward": 0.5497603416442871,
      "step": 735
    },
    {
      "completion_length": 29.859375,
      "epoch": 2.258064516129032,
      "grad_norm": 15.056288719177246,
      "kl": 0.53515625,
      "learning_rate": 4.346153846153846e-07,
      "loss": -0.1799,
      "reward": 1.4998416900634766,
      "reward_std": 0.1745157167315483,
      "reward_std_all": 0.7772514224052429,
      "rewards-max/forensics_sinkhorn_reward": 0.9745798110961914,
      "rewards-max/forensics_strict_edge_reward": 0.506708562374115,
      "rewards-max/forensics_tiou_reward": 0.9620333313941956,
      "rewards/forensics_sinkhorn_reward": 0.772845447063446,
      "rewards/forensics_strict_edge_reward": 0.049137188121676445,
      "rewards/forensics_tiou_reward": 0.6778590679168701,
      "step": 736
    },
    {
      "completion_length": 26.4375,
      "epoch": 2.261136712749616,
      "grad_norm": 10.212711334228516,
      "kl": 0.48828125,
      "learning_rate": 4.338461538461538e-07,
      "loss": -0.0587,
      "reward": 1.660903811454773,
      "reward_std": 0.4169432520866394,
      "reward_std_all": 0.7847756743431091,
      "rewards-max/forensics_sinkhorn_reward": 0.9880993068218231,
      "rewards-max/forensics_strict_edge_reward": 0.952418714761734,
      "rewards-max/forensics_tiou_reward": 0.9974964559078217,
      "rewards/forensics_sinkhorn_reward": 0.80988410115242,
      "rewards/forensics_strict_edge_reward": 0.14811045676469803,
      "rewards/forensics_tiou_reward": 0.7029092609882355,
      "step": 737
    },
    {
      "completion_length": 24.734375,
      "epoch": 2.2642089093701996,
      "grad_norm": 8.458017349243164,
      "kl": 0.4873046875,
      "learning_rate": 4.3307692307692305e-07,
      "loss": -0.1126,
      "reward": 1.3665825128555298,
      "reward_std": 0.3000097721815109,
      "reward_std_all": 0.8681045174598694,
      "rewards-max/forensics_sinkhorn_reward": 0.9765715599060059,
      "rewards-max/forensics_strict_edge_reward": 0.5787998139858246,
      "rewards-max/forensics_tiou_reward": 0.9670127332210541,
      "rewards/forensics_sinkhorn_reward": 0.712143212556839,
      "rewards/forensics_strict_edge_reward": 0.06882071867585182,
      "rewards/forensics_tiou_reward": 0.5856186151504517,
      "step": 738
    },
    {
      "completion_length": 22.6875,
      "epoch": 2.2672811059907834,
      "grad_norm": 8.820732116699219,
      "kl": 0.482421875,
      "learning_rate": 4.3230769230769227e-07,
      "loss": -0.081,
      "reward": 1.3844558596611023,
      "reward_std": 0.5208579152822495,
      "reward_std_all": 0.9629120230674744,
      "rewards-max/forensics_sinkhorn_reward": 0.9886306822299957,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.7236124277114868,
      "rewards/forensics_strict_edge_reward": 0.14399366080760956,
      "rewards/forensics_tiou_reward": 0.5168498456478119,
      "step": 739
    },
    {
      "completion_length": 27.359375,
      "epoch": 2.270353302611367,
      "grad_norm": 8.2578125,
      "kl": 0.462890625,
      "learning_rate": 4.3153846153846153e-07,
      "loss": -0.0679,
      "reward": 1.11768639087677,
      "reward_std": 0.33011581003665924,
      "reward_std_all": 1.0065166652202606,
      "rewards-max/forensics_sinkhorn_reward": 0.9834336042404175,
      "rewards-max/forensics_strict_edge_reward": 0.43310546875,
      "rewards-max/forensics_tiou_reward": 0.9866116344928741,
      "rewards/forensics_sinkhorn_reward": 0.6463609039783478,
      "rewards/forensics_strict_edge_reward": 0.038584284484386444,
      "rewards/forensics_tiou_reward": 0.4327411651611328,
      "step": 740
    },
    {
      "completion_length": 26.625,
      "epoch": 2.273425499231951,
      "grad_norm": 9.125911712646484,
      "kl": 0.5244140625,
      "learning_rate": 4.307692307692308e-07,
      "loss": -0.0732,
      "reward": 1.3796150088310242,
      "reward_std": 0.24699626863002777,
      "reward_std_all": 0.8525733053684235,
      "rewards-max/forensics_sinkhorn_reward": 0.962588906288147,
      "rewards-max/forensics_strict_edge_reward": 0.5159293413162231,
      "rewards-max/forensics_tiou_reward": 0.9356477558612823,
      "rewards/forensics_sinkhorn_reward": 0.7336361110210419,
      "rewards/forensics_strict_edge_reward": 0.05762738548219204,
      "rewards/forensics_tiou_reward": 0.5883516073226929,
      "step": 741
    },
    {
      "completion_length": 22.265625,
      "epoch": 2.2764976958525347,
      "grad_norm": 10.977656364440918,
      "kl": 0.56640625,
      "learning_rate": 4.2999999999999996e-07,
      "loss": -0.0991,
      "reward": 1.277479350566864,
      "reward_std": 0.4008454531431198,
      "reward_std_all": 0.9871174097061157,
      "rewards-max/forensics_sinkhorn_reward": 0.9883980453014374,
      "rewards-max/forensics_strict_edge_reward": 0.9232408702373505,
      "rewards-max/forensics_tiou_reward": 0.998704195022583,
      "rewards/forensics_sinkhorn_reward": 0.6810417473316193,
      "rewards/forensics_strict_edge_reward": 0.10740051418542862,
      "rewards/forensics_tiou_reward": 0.4890371263027191,
      "step": 742
    },
    {
      "completion_length": 26.515625,
      "epoch": 2.279569892473118,
      "grad_norm": 8.608863830566406,
      "kl": 0.568359375,
      "learning_rate": 4.292307692307692e-07,
      "loss": -0.0435,
      "reward": 1.4214235544204712,
      "reward_std": 0.42063644528388977,
      "reward_std_all": 0.9012318253517151,
      "rewards-max/forensics_sinkhorn_reward": 0.9886692762374878,
      "rewards-max/forensics_strict_edge_reward": 0.952418714761734,
      "rewards-max/forensics_tiou_reward": 0.9987038373947144,
      "rewards/forensics_sinkhorn_reward": 0.7372746169567108,
      "rewards/forensics_strict_edge_reward": 0.09657957404851913,
      "rewards/forensics_tiou_reward": 0.5875692963600159,
      "step": 743
    },
    {
      "completion_length": 25.25,
      "epoch": 2.282642089093702,
      "grad_norm": 13.544486045837402,
      "kl": 0.6865234375,
      "learning_rate": 4.2846153846153844e-07,
      "loss": -0.076,
      "reward": 1.210924506187439,
      "reward_std": 0.3391580730676651,
      "reward_std_all": 0.9683062434196472,
      "rewards-max/forensics_sinkhorn_reward": 0.9782361090183258,
      "rewards-max/forensics_strict_edge_reward": 0.6573699712753296,
      "rewards-max/forensics_tiou_reward": 0.9724690318107605,
      "rewards/forensics_sinkhorn_reward": 0.6484487652778625,
      "rewards/forensics_strict_edge_reward": 0.059915754944086075,
      "rewards/forensics_tiou_reward": 0.5025599300861359,
      "step": 744
    },
    {
      "completion_length": 27.734375,
      "epoch": 2.2857142857142856,
      "grad_norm": 9.680173873901367,
      "kl": 0.4658203125,
      "learning_rate": 4.276923076923077e-07,
      "loss": -0.1212,
      "reward": 1.3758193254470825,
      "reward_std": 0.3935772776603699,
      "reward_std_all": 0.9087212085723877,
      "rewards-max/forensics_sinkhorn_reward": 0.9804989993572235,
      "rewards-max/forensics_strict_edge_reward": 0.6742177754640579,
      "rewards-max/forensics_tiou_reward": 0.9765177369117737,
      "rewards/forensics_sinkhorn_reward": 0.6852678656578064,
      "rewards/forensics_strict_edge_reward": 0.09139715880155563,
      "rewards/forensics_tiou_reward": 0.5991543531417847,
      "step": 745
    },
    {
      "completion_length": 26.859375,
      "epoch": 2.2887864823348694,
      "grad_norm": 11.962323188781738,
      "kl": 0.8095703125,
      "learning_rate": 4.2692307692307687e-07,
      "loss": -0.1867,
      "reward": 1.6833419799804688,
      "reward_std": 0.4543868750333786,
      "reward_std_all": 0.8041804432868958,
      "rewards-max/forensics_sinkhorn_reward": 0.9888358414173126,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.7824585437774658,
      "rewards/forensics_strict_edge_reward": 0.17301685363054276,
      "rewards/forensics_tiou_reward": 0.7278665900230408,
      "step": 746
    },
    {
      "completion_length": 25.71875,
      "epoch": 2.291858678955453,
      "grad_norm": 8.802874565124512,
      "kl": 0.48046875,
      "learning_rate": 4.2615384615384613e-07,
      "loss": -0.1106,
      "reward": 1.329775631427765,
      "reward_std": 0.41790226101875305,
      "reward_std_all": 0.9344917237758636,
      "rewards-max/forensics_sinkhorn_reward": 0.9835276007652283,
      "rewards-max/forensics_strict_edge_reward": 0.7453272640705109,
      "rewards-max/forensics_tiou_reward": 0.984655112028122,
      "rewards/forensics_sinkhorn_reward": 0.7021016776561737,
      "rewards/forensics_strict_edge_reward": 0.07271545007824898,
      "rewards/forensics_tiou_reward": 0.5549584627151489,
      "step": 747
    },
    {
      "completion_length": 26.265625,
      "epoch": 2.294930875576037,
      "grad_norm": 9.394081115722656,
      "kl": 0.3974609375,
      "learning_rate": 4.2538461538461535e-07,
      "loss": -0.0893,
      "reward": 1.3930736780166626,
      "reward_std": 0.38430458307266235,
      "reward_std_all": 0.8932085335254669,
      "rewards-max/forensics_sinkhorn_reward": 0.9790775179862976,
      "rewards-max/forensics_strict_edge_reward": 0.5949569344520569,
      "rewards-max/forensics_tiou_reward": 0.9737732112407684,
      "rewards/forensics_sinkhorn_reward": 0.7276298999786377,
      "rewards/forensics_strict_edge_reward": 0.06955535337328911,
      "rewards/forensics_tiou_reward": 0.5958884060382843,
      "step": 748
    },
    {
      "completion_length": 22.6875,
      "epoch": 2.2980030721966207,
      "grad_norm": 8.595417976379395,
      "kl": 0.3916015625,
      "learning_rate": 4.246153846153846e-07,
      "loss": -0.0943,
      "reward": 1.4430485367774963,
      "reward_std": 0.32109886407852173,
      "reward_std_all": 0.8688272833824158,
      "rewards-max/forensics_sinkhorn_reward": 0.9743395149707794,
      "rewards-max/forensics_strict_edge_reward": 0.670350581407547,
      "rewards-max/forensics_tiou_reward": 0.9659453928470612,
      "rewards/forensics_sinkhorn_reward": 0.7807194888591766,
      "rewards/forensics_strict_edge_reward": 0.06942435167729855,
      "rewards/forensics_tiou_reward": 0.5929047167301178,
      "step": 749
    },
    {
      "completion_length": 27.71875,
      "epoch": 2.3010752688172045,
      "grad_norm": 10.327815055847168,
      "kl": 0.611328125,
      "learning_rate": 4.2384615384615383e-07,
      "loss": -0.1141,
      "reward": 1.3238674998283386,
      "reward_std": 0.2961469888687134,
      "reward_std_all": 0.9009599685668945,
      "rewards-max/forensics_sinkhorn_reward": 0.9794654548168182,
      "rewards-max/forensics_strict_edge_reward": 0.6644296199083328,
      "rewards-max/forensics_tiou_reward": 0.9638453423976898,
      "rewards/forensics_sinkhorn_reward": 0.7001761794090271,
      "rewards/forensics_strict_edge_reward": 0.07970774546265602,
      "rewards/forensics_tiou_reward": 0.5439836233854294,
      "step": 750
    },
    {
      "completion_length": 25.78125,
      "epoch": 2.3041474654377883,
      "grad_norm": 9.373393058776855,
      "kl": 0.482421875,
      "learning_rate": 4.2307692307692304e-07,
      "loss": -0.075,
      "reward": 1.2985426187515259,
      "reward_std": 0.25782743096351624,
      "reward_std_all": 0.902687281370163,
      "rewards-max/forensics_sinkhorn_reward": 0.9606188237667084,
      "rewards-max/forensics_strict_edge_reward": 0.2700788117945194,
      "rewards-max/forensics_tiou_reward": 0.932555079460144,
      "rewards/forensics_sinkhorn_reward": 0.7065267562866211,
      "rewards/forensics_strict_edge_reward": 0.025699851103127003,
      "rewards/forensics_tiou_reward": 0.5663159787654877,
      "step": 751
    },
    {
      "completion_length": 23.90625,
      "epoch": 2.3072196620583716,
      "grad_norm": 13.017221450805664,
      "kl": 0.4697265625,
      "learning_rate": 4.223076923076923e-07,
      "loss": -0.0483,
      "reward": 1.3353406190872192,
      "reward_std": 0.40659256279468536,
      "reward_std_all": 0.9328305423259735,
      "rewards-max/forensics_sinkhorn_reward": 0.9833566844463348,
      "rewards-max/forensics_strict_edge_reward": 0.6876508891582489,
      "rewards-max/forensics_tiou_reward": 0.9844730794429779,
      "rewards/forensics_sinkhorn_reward": 0.7279418408870697,
      "rewards/forensics_strict_edge_reward": 0.062435269355773926,
      "rewards/forensics_tiou_reward": 0.544963464140892,
      "step": 752
    },
    {
      "completion_length": 26.421875,
      "epoch": 2.3102918586789554,
      "grad_norm": 10.459027290344238,
      "kl": 0.646484375,
      "learning_rate": 4.215384615384615e-07,
      "loss": -0.0333,
      "reward": 1.522257924079895,
      "reward_std": 0.49157701432704926,
      "reward_std_all": 0.9030078649520874,
      "rewards-max/forensics_sinkhorn_reward": 0.9883029758930206,
      "rewards-max/forensics_strict_edge_reward": 0.9414820969104767,
      "rewards-max/forensics_tiou_reward": 0.9976181387901306,
      "rewards/forensics_sinkhorn_reward": 0.7571664452552795,
      "rewards/forensics_strict_edge_reward": 0.14956530183553696,
      "rewards/forensics_tiou_reward": 0.6155261695384979,
      "step": 753
    },
    {
      "completion_length": 24.796875,
      "epoch": 2.313364055299539,
      "grad_norm": 10.448647499084473,
      "kl": 0.654296875,
      "learning_rate": 4.207692307692308e-07,
      "loss": -0.0815,
      "reward": 1.4188175797462463,
      "reward_std": 0.42940467596054077,
      "reward_std_all": 0.8812204599380493,
      "rewards-max/forensics_sinkhorn_reward": 0.9868430495262146,
      "rewards-max/forensics_strict_edge_reward": 0.4181677848100662,
      "rewards-max/forensics_tiou_reward": 0.9923907518386841,
      "rewards/forensics_sinkhorn_reward": 0.7707571983337402,
      "rewards/forensics_strict_edge_reward": 0.06366412900388241,
      "rewards/forensics_tiou_reward": 0.5843961834907532,
      "step": 754
    },
    {
      "completion_length": 22.359375,
      "epoch": 2.316436251920123,
      "grad_norm": 8.496058464050293,
      "kl": 0.3486328125,
      "learning_rate": 4.1999999999999995e-07,
      "loss": -0.1455,
      "reward": 1.4358618259429932,
      "reward_std": 0.31266364455223083,
      "reward_std_all": 0.8509153425693512,
      "rewards-max/forensics_sinkhorn_reward": 0.9837261438369751,
      "rewards-max/forensics_strict_edge_reward": 0.5829301476478577,
      "rewards-max/forensics_tiou_reward": 0.9852228462696075,
      "rewards/forensics_sinkhorn_reward": 0.7704601287841797,
      "rewards/forensics_strict_edge_reward": 0.06283017434179783,
      "rewards/forensics_tiou_reward": 0.6025715172290802,
      "step": 755
    },
    {
      "completion_length": 24.09375,
      "epoch": 2.3195084485407067,
      "grad_norm": 7.539974212646484,
      "kl": 0.392578125,
      "learning_rate": 4.192307692307692e-07,
      "loss": -0.096,
      "reward": 1.6538888812065125,
      "reward_std": 0.44237080216407776,
      "reward_std_all": 0.8038236200809479,
      "rewards-max/forensics_sinkhorn_reward": 0.9856009185314178,
      "rewards-max/forensics_strict_edge_reward": 0.8038780093193054,
      "rewards-max/forensics_tiou_reward": 0.9926804006099701,
      "rewards/forensics_sinkhorn_reward": 0.8204229772090912,
      "rewards/forensics_strict_edge_reward": 0.14524231851100922,
      "rewards/forensics_tiou_reward": 0.6882235407829285,
      "step": 756
    },
    {
      "completion_length": 25.890625,
      "epoch": 2.3225806451612905,
      "grad_norm": 8.173919677734375,
      "kl": 0.3388671875,
      "learning_rate": 4.1846153846153843e-07,
      "loss": -0.0675,
      "reward": 1.1417554020881653,
      "reward_std": 0.3671667128801346,
      "reward_std_all": 1.0190108120441437,
      "rewards-max/forensics_sinkhorn_reward": 0.9782677888870239,
      "rewards-max/forensics_strict_edge_reward": 0.6168098747730255,
      "rewards-max/forensics_tiou_reward": 0.9728926718235016,
      "rewards/forensics_sinkhorn_reward": 0.6312857270240784,
      "rewards/forensics_strict_edge_reward": 0.06210678815841675,
      "rewards/forensics_tiou_reward": 0.44836294651031494,
      "step": 757
    },
    {
      "completion_length": 21.0,
      "epoch": 2.3256528417818743,
      "grad_norm": 8.1654052734375,
      "kl": 0.4775390625,
      "learning_rate": 4.176923076923077e-07,
      "loss": -0.1204,
      "reward": 1.3137478828430176,
      "reward_std": 0.4357532113790512,
      "reward_std_all": 0.9714813828468323,
      "rewards-max/forensics_sinkhorn_reward": 0.985452800989151,
      "rewards-max/forensics_strict_edge_reward": 0.7542104423046112,
      "rewards-max/forensics_tiou_reward": 0.9897958934307098,
      "rewards/forensics_sinkhorn_reward": 0.7159828841686249,
      "rewards/forensics_strict_edge_reward": 0.12341984361410141,
      "rewards/forensics_tiou_reward": 0.4743451327085495,
      "step": 758
    },
    {
      "completion_length": 24.859375,
      "epoch": 2.3287250384024576,
      "grad_norm": 11.076549530029297,
      "kl": 0.546875,
      "learning_rate": 4.1692307692307686e-07,
      "loss": -0.1383,
      "reward": 1.5096880793571472,
      "reward_std": 0.3553314208984375,
      "reward_std_all": 0.8345248401165009,
      "rewards-max/forensics_sinkhorn_reward": 0.9860130548477173,
      "rewards-max/forensics_strict_edge_reward": 0.8269640505313873,
      "rewards-max/forensics_tiou_reward": 0.9898322522640228,
      "rewards/forensics_sinkhorn_reward": 0.7672444581985474,
      "rewards/forensics_strict_edge_reward": 0.1351815015077591,
      "rewards/forensics_tiou_reward": 0.60726198554039,
      "step": 759
    },
    {
      "completion_length": 26.3125,
      "epoch": 2.3317972350230414,
      "grad_norm": 10.594108581542969,
      "kl": 0.4013671875,
      "learning_rate": 4.161538461538461e-07,
      "loss": -0.1654,
      "reward": 1.2545902132987976,
      "reward_std": 0.3401424139738083,
      "reward_std_all": 0.9528021514415741,
      "rewards-max/forensics_sinkhorn_reward": 0.9873154759407043,
      "rewards-max/forensics_strict_edge_reward": 0.755835086107254,
      "rewards-max/forensics_tiou_reward": 0.9944764077663422,
      "rewards/forensics_sinkhorn_reward": 0.6390854716300964,
      "rewards/forensics_strict_edge_reward": 0.09093903005123138,
      "rewards/forensics_tiou_reward": 0.5245656967163086,
      "step": 760
    },
    {
      "completion_length": 24.578125,
      "epoch": 2.334869431643625,
      "grad_norm": 10.369961738586426,
      "kl": 0.55078125,
      "learning_rate": 4.153846153846154e-07,
      "loss": -0.15,
      "reward": 1.5541167855262756,
      "reward_std": 0.43787774443626404,
      "reward_std_all": 0.8492806255817413,
      "rewards-max/forensics_sinkhorn_reward": 0.9896073043346405,
      "rewards-max/forensics_strict_edge_reward": 0.882964164018631,
      "rewards-max/forensics_tiou_reward": 0.9995909333229065,
      "rewards/forensics_sinkhorn_reward": 0.787976086139679,
      "rewards/forensics_strict_edge_reward": 0.12923721969127655,
      "rewards/forensics_tiou_reward": 0.6369035243988037,
      "step": 761
    },
    {
      "completion_length": 29.625,
      "epoch": 2.337941628264209,
      "grad_norm": 10.835124969482422,
      "kl": 0.4208984375,
      "learning_rate": 4.146153846153846e-07,
      "loss": -0.0057,
      "reward": 1.4527568221092224,
      "reward_std": 0.26782461255788803,
      "reward_std_all": 0.8263329267501831,
      "rewards-max/forensics_sinkhorn_reward": 0.9799852967262268,
      "rewards-max/forensics_strict_edge_reward": 0.7545789182186127,
      "rewards-max/forensics_tiou_reward": 0.9762685000896454,
      "rewards/forensics_sinkhorn_reward": 0.7388191521167755,
      "rewards/forensics_strict_edge_reward": 0.07955650985240936,
      "rewards/forensics_tiou_reward": 0.6343812346458435,
      "step": 762
    },
    {
      "completion_length": 27.234375,
      "epoch": 2.3410138248847927,
      "grad_norm": 9.965180397033691,
      "kl": 0.451171875,
      "learning_rate": 4.1384615384615387e-07,
      "loss": -0.0794,
      "reward": 1.6976088881492615,
      "reward_std": 0.37091830372810364,
      "reward_std_all": 0.7489159405231476,
      "rewards-max/forensics_sinkhorn_reward": 0.9877772331237793,
      "rewards-max/forensics_strict_edge_reward": 0.8057825267314911,
      "rewards-max/forensics_tiou_reward": 0.9985695779323578,
      "rewards/forensics_sinkhorn_reward": 0.8325064182281494,
      "rewards/forensics_strict_edge_reward": 0.1459985300898552,
      "rewards/forensics_tiou_reward": 0.719103991985321,
      "step": 763
    },
    {
      "completion_length": 23.03125,
      "epoch": 2.3440860215053765,
      "grad_norm": 9.629111289978027,
      "kl": 0.5517578125,
      "learning_rate": 4.1307692307692303e-07,
      "loss": -0.149,
      "reward": 1.5912473797798157,
      "reward_std": 0.2913559228181839,
      "reward_std_all": 0.780314713716507,
      "rewards-max/forensics_sinkhorn_reward": 0.9832776188850403,
      "rewards-max/forensics_strict_edge_reward": 0.6762872636318207,
      "rewards-max/forensics_tiou_reward": 0.9895726442337036,
      "rewards/forensics_sinkhorn_reward": 0.8308149576187134,
      "rewards/forensics_strict_edge_reward": 0.09467807412147522,
      "rewards/forensics_tiou_reward": 0.6657543182373047,
      "step": 764
    },
    {
      "completion_length": 22.765625,
      "epoch": 2.3471582181259603,
      "grad_norm": 9.965947151184082,
      "kl": 0.4140625,
      "learning_rate": 4.123076923076923e-07,
      "loss": -0.0938,
      "reward": 1.4661219120025635,
      "reward_std": 0.35831068456172943,
      "reward_std_all": 0.8499142229557037,
      "rewards-max/forensics_sinkhorn_reward": 0.984668642282486,
      "rewards-max/forensics_strict_edge_reward": 0.756708562374115,
      "rewards-max/forensics_tiou_reward": 0.9905723333358765,
      "rewards/forensics_sinkhorn_reward": 0.7967875003814697,
      "rewards/forensics_strict_edge_reward": 0.06162591278553009,
      "rewards/forensics_tiou_reward": 0.6077085733413696,
      "step": 765
    },
    {
      "completion_length": 25.75,
      "epoch": 2.3502304147465436,
      "grad_norm": 20.879690170288086,
      "kl": 0.5546875,
      "learning_rate": 4.115384615384615e-07,
      "loss": -0.1244,
      "reward": 1.2973524928092957,
      "reward_std": 0.3256995603442192,
      "reward_std_all": 0.9146764874458313,
      "rewards-max/forensics_sinkhorn_reward": 0.9710215032100677,
      "rewards-max/forensics_strict_edge_reward": 0.3920080354437232,
      "rewards-max/forensics_tiou_reward": 0.9527901411056519,
      "rewards/forensics_sinkhorn_reward": 0.7356723546981812,
      "rewards/forensics_strict_edge_reward": 0.03255935164634138,
      "rewards/forensics_tiou_reward": 0.5291207730770111,
      "step": 766
    },
    {
      "completion_length": 23.515625,
      "epoch": 2.3533026113671274,
      "grad_norm": 9.796576499938965,
      "kl": 0.4013671875,
      "learning_rate": 4.107692307692308e-07,
      "loss": -0.0244,
      "reward": 1.2936644554138184,
      "reward_std": 0.4158347249031067,
      "reward_std_all": 0.9539608657360077,
      "rewards-max/forensics_sinkhorn_reward": 0.9843770563602448,
      "rewards-max/forensics_strict_edge_reward": 0.6419725120067596,
      "rewards-max/forensics_tiou_reward": 0.98891681432724,
      "rewards/forensics_sinkhorn_reward": 0.6672478914260864,
      "rewards/forensics_strict_edge_reward": 0.08264954760670662,
      "rewards/forensics_tiou_reward": 0.543766975402832,
      "step": 767
    },
    {
      "completion_length": 25.625,
      "epoch": 2.356374807987711,
      "grad_norm": 9.069990158081055,
      "kl": 0.5390625,
      "learning_rate": 4.0999999999999994e-07,
      "loss": -0.0734,
      "reward": 1.4234422445297241,
      "reward_std": 0.3304029032588005,
      "reward_std_all": 0.8672660291194916,
      "rewards-max/forensics_sinkhorn_reward": 0.9867778420448303,
      "rewards-max/forensics_strict_edge_reward": 0.7029964029788971,
      "rewards-max/forensics_tiou_reward": 0.9940633177757263,
      "rewards/forensics_sinkhorn_reward": 0.7498133480548859,
      "rewards/forensics_strict_edge_reward": 0.0557908546179533,
      "rewards/forensics_tiou_reward": 0.6178379952907562,
      "step": 768
    },
    {
      "completion_length": 26.625,
      "epoch": 2.359447004608295,
      "grad_norm": 9.982111930847168,
      "kl": 0.404296875,
      "learning_rate": 4.092307692307692e-07,
      "loss": -0.1255,
      "reward": 1.4677802920341492,
      "reward_std": 0.3852730989456177,
      "reward_std_all": 0.8584195971488953,
      "rewards-max/forensics_sinkhorn_reward": 0.985936164855957,
      "rewards-max/forensics_strict_edge_reward": 0.6305927336215973,
      "rewards-max/forensics_tiou_reward": 0.9919708669185638,
      "rewards/forensics_sinkhorn_reward": 0.7720923721790314,
      "rewards/forensics_strict_edge_reward": 0.08420670032501221,
      "rewards/forensics_tiou_reward": 0.611481249332428,
      "step": 769
    },
    {
      "completion_length": 27.671875,
      "epoch": 2.3625192012288787,
      "grad_norm": 11.239381790161133,
      "kl": 0.6044921875,
      "learning_rate": 4.0846153846153847e-07,
      "loss": -0.0458,
      "reward": 1.428274154663086,
      "reward_std": 0.323624424636364,
      "reward_std_all": 0.866987556219101,
      "rewards-max/forensics_sinkhorn_reward": 0.986592173576355,
      "rewards-max/forensics_strict_edge_reward": 0.8004741370677948,
      "rewards-max/forensics_tiou_reward": 0.9931358993053436,
      "rewards/forensics_sinkhorn_reward": 0.7191494703292847,
      "rewards/forensics_strict_edge_reward": 0.10085582733154297,
      "rewards/forensics_tiou_reward": 0.6082688868045807,
      "step": 770
    },
    {
      "completion_length": 26.640625,
      "epoch": 2.3655913978494625,
      "grad_norm": 10.616249084472656,
      "kl": 0.5244140625,
      "learning_rate": 4.076923076923077e-07,
      "loss": -0.0867,
      "reward": 1.438489556312561,
      "reward_std": 0.4682294577360153,
      "reward_std_all": 0.8998292684555054,
      "rewards-max/forensics_sinkhorn_reward": 0.985507607460022,
      "rewards-max/forensics_strict_edge_reward": 0.7295457720756531,
      "rewards-max/forensics_tiou_reward": 0.989787220954895,
      "rewards/forensics_sinkhorn_reward": 0.7142355740070343,
      "rewards/forensics_strict_edge_reward": 0.1438693180680275,
      "rewards/forensics_tiou_reward": 0.5803846418857574,
      "step": 771
    },
    {
      "completion_length": 24.90625,
      "epoch": 2.3686635944700463,
      "grad_norm": 7.827700138092041,
      "kl": 0.4892578125,
      "learning_rate": 4.069230769230769e-07,
      "loss": -0.0125,
      "reward": 1.429088294506073,
      "reward_std": 0.26877302676439285,
      "reward_std_all": 0.8479827344417572,
      "rewards-max/forensics_sinkhorn_reward": 0.9763851463794708,
      "rewards-max/forensics_strict_edge_reward": 0.7744058072566986,
      "rewards-max/forensics_tiou_reward": 0.970771074295044,
      "rewards/forensics_sinkhorn_reward": 0.6867063343524933,
      "rewards/forensics_strict_edge_reward": 0.18411967158317566,
      "rewards/forensics_tiou_reward": 0.558262288570404,
      "step": 772
    },
    {
      "completion_length": 24.15625,
      "epoch": 2.3717357910906296,
      "grad_norm": 8.578751564025879,
      "kl": 0.3525390625,
      "learning_rate": 4.061538461538461e-07,
      "loss": -0.0646,
      "reward": 1.2001572251319885,
      "reward_std": 0.354301854968071,
      "reward_std_all": 1.0001897513866425,
      "rewards-max/forensics_sinkhorn_reward": 0.9861345589160919,
      "rewards-max/forensics_strict_edge_reward": 0.8827048540115356,
      "rewards-max/forensics_tiou_reward": 0.9920342862606049,
      "rewards/forensics_sinkhorn_reward": 0.6640086770057678,
      "rewards/forensics_strict_edge_reward": 0.09839848056435585,
      "rewards/forensics_tiou_reward": 0.43775007128715515,
      "step": 773
    },
    {
      "completion_length": 26.453125,
      "epoch": 2.3748079877112134,
      "grad_norm": 8.508667945861816,
      "kl": 0.3544921875,
      "learning_rate": 4.053846153846154e-07,
      "loss": -0.0219,
      "reward": 1.4207839965820312,
      "reward_std": 0.43853524327278137,
      "reward_std_all": 0.9031594395637512,
      "rewards-max/forensics_sinkhorn_reward": 0.9750885665416718,
      "rewards-max/forensics_strict_edge_reward": 0.4544609785079956,
      "rewards-max/forensics_tiou_reward": 0.9678515493869781,
      "rewards/forensics_sinkhorn_reward": 0.7296632528305054,
      "rewards/forensics_strict_edge_reward": 0.09676498919725418,
      "rewards/forensics_tiou_reward": 0.5943558216094971,
      "step": 774
    },
    {
      "completion_length": 26.109375,
      "epoch": 2.377880184331797,
      "grad_norm": 13.33022689819336,
      "kl": 0.88671875,
      "learning_rate": 4.046153846153846e-07,
      "loss": -0.0119,
      "reward": 1.3712823390960693,
      "reward_std": 0.40417422354221344,
      "reward_std_all": 0.9204172492027283,
      "rewards-max/forensics_sinkhorn_reward": 0.9861796498298645,
      "rewards-max/forensics_strict_edge_reward": 0.7482926547527313,
      "rewards-max/forensics_tiou_reward": 0.9936057329177856,
      "rewards/forensics_sinkhorn_reward": 0.7408275604248047,
      "rewards/forensics_strict_edge_reward": 0.07455820590257645,
      "rewards/forensics_tiou_reward": 0.5558966398239136,
      "step": 775
    },
    {
      "completion_length": 25.453125,
      "epoch": 2.380952380952381,
      "grad_norm": 8.843409538269043,
      "kl": 0.474609375,
      "learning_rate": 4.0384615384615386e-07,
      "loss": -0.1032,
      "reward": 1.3237882256507874,
      "reward_std": 0.28760090470314026,
      "reward_std_all": 0.887135773897171,
      "rewards-max/forensics_sinkhorn_reward": 0.9732220470905304,
      "rewards-max/forensics_strict_edge_reward": 0.5284619629383087,
      "rewards-max/forensics_tiou_reward": 0.9584282338619232,
      "rewards/forensics_sinkhorn_reward": 0.7115665078163147,
      "rewards/forensics_strict_edge_reward": 0.06877236068248749,
      "rewards/forensics_tiou_reward": 0.5434494465589523,
      "step": 776
    },
    {
      "completion_length": 23.984375,
      "epoch": 2.3840245775729647,
      "grad_norm": 9.993145942687988,
      "kl": 0.501953125,
      "learning_rate": 4.03076923076923e-07,
      "loss": -0.0989,
      "reward": 1.5891226530075073,
      "reward_std": 0.38369689881801605,
      "reward_std_all": 0.8025873303413391,
      "rewards-max/forensics_sinkhorn_reward": 0.9865762591362,
      "rewards-max/forensics_strict_edge_reward": 0.7165313363075256,
      "rewards-max/forensics_tiou_reward": 0.9935867190361023,
      "rewards/forensics_sinkhorn_reward": 0.7964287102222443,
      "rewards/forensics_strict_edge_reward": 0.12027756869792938,
      "rewards/forensics_tiou_reward": 0.6724163889884949,
      "step": 777
    },
    {
      "completion_length": 25.515625,
      "epoch": 2.3870967741935485,
      "grad_norm": 11.013554573059082,
      "kl": 0.396484375,
      "learning_rate": 4.023076923076923e-07,
      "loss": -0.1377,
      "reward": 1.5016078352928162,
      "reward_std": 0.49399900436401367,
      "reward_std_all": 0.8993622660636902,
      "rewards-max/forensics_sinkhorn_reward": 0.9875892400741577,
      "rewards-max/forensics_strict_edge_reward": 0.8523736894130707,
      "rewards-max/forensics_tiou_reward": 0.9951122105121613,
      "rewards/forensics_sinkhorn_reward": 0.7668594121932983,
      "rewards/forensics_strict_edge_reward": 0.12262114137411118,
      "rewards/forensics_tiou_reward": 0.612127423286438,
      "step": 778
    },
    {
      "completion_length": 29.359375,
      "epoch": 2.3901689708141323,
      "grad_norm": 8.657543182373047,
      "kl": 0.4677734375,
      "learning_rate": 4.015384615384615e-07,
      "loss": -0.0662,
      "reward": 1.4213677644729614,
      "reward_std": 0.30771882832050323,
      "reward_std_all": 0.8551509976387024,
      "rewards-max/forensics_sinkhorn_reward": 0.982040137052536,
      "rewards-max/forensics_strict_edge_reward": 0.756708562374115,
      "rewards-max/forensics_tiou_reward": 0.9838771820068359,
      "rewards/forensics_sinkhorn_reward": 0.7661961913108826,
      "rewards/forensics_strict_edge_reward": 0.05665561370551586,
      "rewards/forensics_tiou_reward": 0.5985158681869507,
      "step": 779
    },
    {
      "completion_length": 26.0625,
      "epoch": 2.3932411674347156,
      "grad_norm": 9.805777549743652,
      "kl": 0.57421875,
      "learning_rate": 4.0076923076923077e-07,
      "loss": -0.0203,
      "reward": 1.4385570287704468,
      "reward_std": 0.37040242552757263,
      "reward_std_all": 0.8688070476055145,
      "rewards-max/forensics_sinkhorn_reward": 0.980905681848526,
      "rewards-max/forensics_strict_edge_reward": 0.6232984811067581,
      "rewards-max/forensics_tiou_reward": 0.9816694855690002,
      "rewards/forensics_sinkhorn_reward": 0.7602630257606506,
      "rewards/forensics_strict_edge_reward": 0.06835011020302773,
      "rewards/forensics_tiou_reward": 0.6099438965320587,
      "step": 780
    },
    {
      "completion_length": 27.515625,
      "epoch": 2.3963133640552994,
      "grad_norm": 8.035096168518066,
      "kl": 0.52734375,
      "learning_rate": 4e-07,
      "loss": -0.0249,
      "reward": 1.5102777481079102,
      "reward_std": 0.41142337024211884,
      "reward_std_all": 0.883563369512558,
      "rewards-max/forensics_sinkhorn_reward": 0.983200877904892,
      "rewards-max/forensics_strict_edge_reward": 0.8465203046798706,
      "rewards-max/forensics_tiou_reward": 0.9851847290992737,
      "rewards/forensics_sinkhorn_reward": 0.7330282926559448,
      "rewards/forensics_strict_edge_reward": 0.12932362779974937,
      "rewards/forensics_tiou_reward": 0.647925853729248,
      "step": 781
    },
    {
      "completion_length": 26.9375,
      "epoch": 2.399385560675883,
      "grad_norm": 10.816137313842773,
      "kl": 0.57421875,
      "learning_rate": 3.992307692307692e-07,
      "loss": -0.1008,
      "reward": 1.4525849223136902,
      "reward_std": 0.5418529212474823,
      "reward_std_all": 0.9468045830726624,
      "rewards-max/forensics_sinkhorn_reward": 0.9895579218864441,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.7141009867191315,
      "rewards/forensics_strict_edge_reward": 0.15041543543338776,
      "rewards/forensics_tiou_reward": 0.5880684852600098,
      "step": 782
    },
    {
      "completion_length": 27.703125,
      "epoch": 2.402457757296467,
      "grad_norm": 10.565779685974121,
      "kl": 0.6328125,
      "learning_rate": 3.9846153846153846e-07,
      "loss": -0.0495,
      "reward": 1.4933263659477234,
      "reward_std": 0.32694384455680847,
      "reward_std_all": 0.8413971364498138,
      "rewards-max/forensics_sinkhorn_reward": 0.9854858815670013,
      "rewards-max/forensics_strict_edge_reward": 0.824172168970108,
      "rewards-max/forensics_tiou_reward": 0.991172581911087,
      "rewards/forensics_sinkhorn_reward": 0.7456107139587402,
      "rewards/forensics_strict_edge_reward": 0.11482873931527138,
      "rewards/forensics_tiou_reward": 0.6328869462013245,
      "step": 783
    },
    {
      "completion_length": 25.140625,
      "epoch": 2.4055299539170507,
      "grad_norm": 13.67254638671875,
      "kl": 0.4560546875,
      "learning_rate": 3.976923076923077e-07,
      "loss": -0.1437,
      "reward": 1.570454478263855,
      "reward_std": 0.4087617099285126,
      "reward_std_all": 0.8211449980735779,
      "rewards-max/forensics_sinkhorn_reward": 0.9882107079029083,
      "rewards-max/forensics_strict_edge_reward": 0.7744058072566986,
      "rewards-max/forensics_tiou_reward": 0.9977731108665466,
      "rewards/forensics_sinkhorn_reward": 0.7692415714263916,
      "rewards/forensics_strict_edge_reward": 0.15822644531726837,
      "rewards/forensics_tiou_reward": 0.6429865062236786,
      "step": 784
    },
    {
      "completion_length": 25.359375,
      "epoch": 2.4086021505376345,
      "grad_norm": 43.558048248291016,
      "kl": 0.50390625,
      "learning_rate": 3.969230769230769e-07,
      "loss": -0.0815,
      "reward": 1.4781545400619507,
      "reward_std": 0.34102168679237366,
      "reward_std_all": 0.8455101251602173,
      "rewards-max/forensics_sinkhorn_reward": 0.9880363941192627,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.7374915182590485,
      "rewards/forensics_strict_edge_reward": 0.16345598548650742,
      "rewards/forensics_tiou_reward": 0.5772070288658142,
      "step": 785
    },
    {
      "completion_length": 24.640625,
      "epoch": 2.4116743471582183,
      "grad_norm": 9.36330509185791,
      "kl": 0.4501953125,
      "learning_rate": 3.961538461538461e-07,
      "loss": -0.134,
      "reward": 1.2686491012573242,
      "reward_std": 0.31303533911705017,
      "reward_std_all": 0.9329557418823242,
      "rewards-max/forensics_sinkhorn_reward": 0.9756540358066559,
      "rewards-max/forensics_strict_edge_reward": 0.661531001329422,
      "rewards-max/forensics_tiou_reward": 0.96886146068573,
      "rewards/forensics_sinkhorn_reward": 0.6975679993629456,
      "rewards/forensics_strict_edge_reward": 0.05333182215690613,
      "rewards/forensics_tiou_reward": 0.5177493691444397,
      "step": 786
    },
    {
      "completion_length": 20.390625,
      "epoch": 2.4147465437788016,
      "grad_norm": 9.910545349121094,
      "kl": 0.5234375,
      "learning_rate": 3.9538461538461537e-07,
      "loss": -0.1386,
      "reward": 1.4938631653785706,
      "reward_std": 0.08961362764239311,
      "reward_std_all": 0.7608424425125122,
      "rewards-max/forensics_sinkhorn_reward": 0.9819235801696777,
      "rewards-max/forensics_strict_edge_reward": 0.44892527163028717,
      "rewards-max/forensics_tiou_reward": 0.9752926528453827,
      "rewards/forensics_sinkhorn_reward": 0.8117888569831848,
      "rewards/forensics_strict_edge_reward": 0.05808371491730213,
      "rewards/forensics_tiou_reward": 0.6239905953407288,
      "step": 787
    },
    {
      "completion_length": 24.96875,
      "epoch": 2.4178187403993854,
      "grad_norm": 9.580652236938477,
      "kl": 0.61328125,
      "learning_rate": 3.946153846153846e-07,
      "loss": -0.0728,
      "reward": 1.6448951363563538,
      "reward_std": 0.3904498219490051,
      "reward_std_all": 0.7748039960861206,
      "rewards-max/forensics_sinkhorn_reward": 0.9867022931575775,
      "rewards-max/forensics_strict_edge_reward": 0.7249859720468521,
      "rewards-max/forensics_tiou_reward": 0.9968085587024689,
      "rewards/forensics_sinkhorn_reward": 0.8112781643867493,
      "rewards/forensics_strict_edge_reward": 0.13262970745563507,
      "rewards/forensics_tiou_reward": 0.7009871900081635,
      "step": 788
    },
    {
      "completion_length": 24.46875,
      "epoch": 2.420890937019969,
      "grad_norm": 8.947708129882812,
      "kl": 0.4443359375,
      "learning_rate": 3.9384615384615385e-07,
      "loss": -0.09,
      "reward": 1.2185290455818176,
      "reward_std": 0.21344606578350067,
      "reward_std_all": 0.9258015155792236,
      "rewards-max/forensics_sinkhorn_reward": 0.9709267318248749,
      "rewards-max/forensics_strict_edge_reward": 0.39297446608543396,
      "rewards-max/forensics_tiou_reward": 0.9531786441802979,
      "rewards/forensics_sinkhorn_reward": 0.6475114226341248,
      "rewards/forensics_strict_edge_reward": 0.033967772498726845,
      "rewards/forensics_tiou_reward": 0.5370498895645142,
      "step": 789
    },
    {
      "completion_length": 24.03125,
      "epoch": 2.423963133640553,
      "grad_norm": 8.062515258789062,
      "kl": 0.41015625,
      "learning_rate": 3.9307692307692306e-07,
      "loss": -0.1065,
      "reward": 1.513543963432312,
      "reward_std": 0.2552224472165108,
      "reward_std_all": 0.8064782023429871,
      "rewards-max/forensics_sinkhorn_reward": 0.9869268834590912,
      "rewards-max/forensics_strict_edge_reward": 0.7145891487598419,
      "rewards-max/forensics_tiou_reward": 0.9945447146892548,
      "rewards/forensics_sinkhorn_reward": 0.8084627091884613,
      "rewards/forensics_strict_edge_reward": 0.08781093917787075,
      "rewards/forensics_tiou_reward": 0.6172703206539154,
      "step": 790
    },
    {
      "completion_length": 23.515625,
      "epoch": 2.4270353302611367,
      "grad_norm": 10.63676643371582,
      "kl": 0.466796875,
      "learning_rate": 3.923076923076923e-07,
      "loss": -0.0257,
      "reward": 1.3447244763374329,
      "reward_std": 0.3968384563922882,
      "reward_std_all": 0.947781503200531,
      "rewards-max/forensics_sinkhorn_reward": 0.9812648594379425,
      "rewards-max/forensics_strict_edge_reward": 0.7457418143749237,
      "rewards-max/forensics_tiou_reward": 0.9800754189491272,
      "rewards/forensics_sinkhorn_reward": 0.7118353843688965,
      "rewards/forensics_strict_edge_reward": 0.09926366060972214,
      "rewards/forensics_tiou_reward": 0.5336253941059113,
      "step": 791
    },
    {
      "completion_length": 26.109375,
      "epoch": 2.4301075268817205,
      "grad_norm": 15.315935134887695,
      "kl": 0.505859375,
      "learning_rate": 3.9153846153846154e-07,
      "loss": -0.106,
      "reward": 1.6489237546920776,
      "reward_std": 0.35258111357688904,
      "reward_std_all": 0.7624978423118591,
      "rewards-max/forensics_sinkhorn_reward": 0.9890805780887604,
      "rewards-max/forensics_strict_edge_reward": 0.8706149160861969,
      "rewards-max/forensics_tiou_reward": 0.996916264295578,
      "rewards/forensics_sinkhorn_reward": 0.8224651217460632,
      "rewards/forensics_strict_edge_reward": 0.11795863136649132,
      "rewards/forensics_tiou_reward": 0.7084999680519104,
      "step": 792
    },
    {
      "completion_length": 23.40625,
      "epoch": 2.4331797235023043,
      "grad_norm": 10.169872283935547,
      "kl": 0.412109375,
      "learning_rate": 3.9076923076923076e-07,
      "loss": -0.0779,
      "reward": 1.5586431622505188,
      "reward_std": 0.3958613872528076,
      "reward_std_all": 0.8261208534240723,
      "rewards-max/forensics_sinkhorn_reward": 0.9891653060913086,
      "rewards-max/forensics_strict_edge_reward": 0.9093653857707977,
      "rewards-max/forensics_tiou_reward": 0.9990211427211761,
      "rewards/forensics_sinkhorn_reward": 0.7887440323829651,
      "rewards/forensics_strict_edge_reward": 0.1288875713944435,
      "rewards/forensics_tiou_reward": 0.6410115659236908,
      "step": 793
    },
    {
      "completion_length": 24.859375,
      "epoch": 2.436251920122888,
      "grad_norm": 9.503324508666992,
      "kl": 0.3916015625,
      "learning_rate": 3.8999999999999997e-07,
      "loss": -0.1271,
      "reward": 1.435680866241455,
      "reward_std": 0.20966635644435883,
      "reward_std_all": 0.8291181921958923,
      "rewards-max/forensics_sinkhorn_reward": 0.9870555698871613,
      "rewards-max/forensics_strict_edge_reward": 0.7579931616783142,
      "rewards-max/forensics_tiou_reward": 0.9964974820613861,
      "rewards/forensics_sinkhorn_reward": 0.7506863474845886,
      "rewards/forensics_strict_edge_reward": 0.09328689239919186,
      "rewards/forensics_tiou_reward": 0.591707706451416,
      "step": 794
    },
    {
      "completion_length": 28.78125,
      "epoch": 2.439324116743472,
      "grad_norm": 10.904938697814941,
      "kl": 0.4853515625,
      "learning_rate": 3.892307692307692e-07,
      "loss": -0.0173,
      "reward": 1.4094341397285461,
      "reward_std": 0.37844692170619965,
      "reward_std_all": 0.8932849764823914,
      "rewards-max/forensics_sinkhorn_reward": 0.9882442057132721,
      "rewards-max/forensics_strict_edge_reward": 0.861454963684082,
      "rewards-max/forensics_tiou_reward": 0.9970310926437378,
      "rewards/forensics_sinkhorn_reward": 0.7250702977180481,
      "rewards/forensics_strict_edge_reward": 0.10716541856527328,
      "rewards/forensics_tiou_reward": 0.5771984159946442,
      "step": 795
    },
    {
      "completion_length": 24.671875,
      "epoch": 2.442396313364055,
      "grad_norm": 9.195453643798828,
      "kl": 0.474609375,
      "learning_rate": 3.8846153846153845e-07,
      "loss": -0.0898,
      "reward": 1.4633409976959229,
      "reward_std": 0.43745872378349304,
      "reward_std_all": 0.8870227634906769,
      "rewards-max/forensics_sinkhorn_reward": 0.9858328998088837,
      "rewards-max/forensics_strict_edge_reward": 0.7445254027843475,
      "rewards-max/forensics_tiou_reward": 0.9908585548400879,
      "rewards/forensics_sinkhorn_reward": 0.7607064247131348,
      "rewards/forensics_strict_edge_reward": 0.09779877960681915,
      "rewards/forensics_tiou_reward": 0.6048357486724854,
      "step": 796
    },
    {
      "completion_length": 22.59375,
      "epoch": 2.445468509984639,
      "grad_norm": 9.436415672302246,
      "kl": 0.4775390625,
      "learning_rate": 3.8769230769230767e-07,
      "loss": -0.0729,
      "reward": 1.4499356150627136,
      "reward_std": 0.4864729642868042,
      "reward_std_all": 0.8917645812034607,
      "rewards-max/forensics_sinkhorn_reward": 0.9853987991809845,
      "rewards-max/forensics_strict_edge_reward": 0.5343473553657532,
      "rewards-max/forensics_tiou_reward": 0.9897353947162628,
      "rewards/forensics_sinkhorn_reward": 0.7438374161720276,
      "rewards/forensics_strict_edge_reward": 0.0650776382535696,
      "rewards/forensics_tiou_reward": 0.6410205364227295,
      "step": 797
    },
    {
      "completion_length": 25.859375,
      "epoch": 2.4485407066052227,
      "grad_norm": 8.914809226989746,
      "kl": 0.5712890625,
      "learning_rate": 3.8692307692307693e-07,
      "loss": -0.0938,
      "reward": 1.3478462100028992,
      "reward_std": 0.28404098749160767,
      "reward_std_all": 0.8844737410545349,
      "rewards-max/forensics_sinkhorn_reward": 0.9828438460826874,
      "rewards-max/forensics_strict_edge_reward": 0.6401460766792297,
      "rewards-max/forensics_tiou_reward": 0.9834759533405304,
      "rewards/forensics_sinkhorn_reward": 0.6971704959869385,
      "rewards/forensics_strict_edge_reward": 0.09495629370212555,
      "rewards/forensics_tiou_reward": 0.5557194352149963,
      "step": 798
    },
    {
      "completion_length": 27.984375,
      "epoch": 2.4516129032258065,
      "grad_norm": 10.579130172729492,
      "kl": 0.513671875,
      "learning_rate": 3.8615384615384615e-07,
      "loss": -0.0268,
      "reward": 1.4137942790985107,
      "reward_std": 0.35592110455036163,
      "reward_std_all": 0.8886370956897736,
      "rewards-max/forensics_sinkhorn_reward": 0.988546222448349,
      "rewards-max/forensics_strict_edge_reward": 0.8351600170135498,
      "rewards-max/forensics_tiou_reward": 0.9984984695911407,
      "rewards/forensics_sinkhorn_reward": 0.7214973270893097,
      "rewards/forensics_strict_edge_reward": 0.08637727424502373,
      "rewards/forensics_tiou_reward": 0.605919748544693,
      "step": 799
    },
    {
      "completion_length": 24.390625,
      "epoch": 2.4546850998463903,
      "grad_norm": 8.075135231018066,
      "kl": 0.3955078125,
      "learning_rate": 3.8538461538461536e-07,
      "loss": -0.0902,
      "reward": 1.3526558876037598,
      "reward_std": 0.3751305937767029,
      "reward_std_all": 0.9143992960453033,
      "rewards-max/forensics_sinkhorn_reward": 0.9851736128330231,
      "rewards-max/forensics_strict_edge_reward": 0.756708562374115,
      "rewards-max/forensics_tiou_reward": 0.9893954694271088,
      "rewards/forensics_sinkhorn_reward": 0.7464863061904907,
      "rewards/forensics_strict_edge_reward": 0.059868765994906425,
      "rewards/forensics_tiou_reward": 0.546300858259201,
      "step": 800
    },
    {
      "completion_length": 25.84375,
      "epoch": 2.457757296466974,
      "grad_norm": 10.698159217834473,
      "kl": 0.4248046875,
      "learning_rate": 3.8461538461538463e-07,
      "loss": -0.1923,
      "reward": 1.5147871375083923,
      "reward_std": 0.4496365338563919,
      "reward_std_all": 0.8749067485332489,
      "rewards-max/forensics_sinkhorn_reward": 0.9847871363162994,
      "rewards-max/forensics_strict_edge_reward": 0.9093653857707977,
      "rewards-max/forensics_tiou_reward": 0.9896508753299713,
      "rewards/forensics_sinkhorn_reward": 0.7529173195362091,
      "rewards/forensics_strict_edge_reward": 0.1395232081413269,
      "rewards/forensics_tiou_reward": 0.6223465502262115,
      "step": 801
    },
    {
      "completion_length": 23.796875,
      "epoch": 2.460829493087558,
      "grad_norm": 8.62313175201416,
      "kl": 0.6220703125,
      "learning_rate": 3.8384615384615384e-07,
      "loss": -0.0595,
      "reward": 1.1522539258003235,
      "reward_std": 0.36015184223651886,
      "reward_std_all": 1.0103189945220947,
      "rewards-max/forensics_sinkhorn_reward": 0.9790593981742859,
      "rewards-max/forensics_strict_edge_reward": 0.5986784100532532,
      "rewards-max/forensics_tiou_reward": 0.9754921197891235,
      "rewards/forensics_sinkhorn_reward": 0.6520553231239319,
      "rewards/forensics_strict_edge_reward": 0.06087309308350086,
      "rewards/forensics_tiou_reward": 0.4393254667520523,
      "step": 802
    },
    {
      "completion_length": 29.921875,
      "epoch": 2.463901689708141,
      "grad_norm": 10.199795722961426,
      "kl": 0.52734375,
      "learning_rate": 3.8307692307692305e-07,
      "loss": -0.0869,
      "reward": 1.3931392431259155,
      "reward_std": 0.32623693346977234,
      "reward_std_all": 0.8895868360996246,
      "rewards-max/forensics_sinkhorn_reward": 0.9730410575866699,
      "rewards-max/forensics_strict_edge_reward": 0.5739988833665848,
      "rewards-max/forensics_tiou_reward": 0.9637033641338348,
      "rewards/forensics_sinkhorn_reward": 0.7122982144355774,
      "rewards/forensics_strict_edge_reward": 0.08793210983276367,
      "rewards/forensics_tiou_reward": 0.5929089188575745,
      "step": 803
    },
    {
      "completion_length": 23.921875,
      "epoch": 2.466973886328725,
      "grad_norm": 12.06390380859375,
      "kl": 0.5693359375,
      "learning_rate": 3.8230769230769227e-07,
      "loss": -0.0361,
      "reward": 1.1723722219467163,
      "reward_std": 0.296441912651062,
      "reward_std_all": 0.9803975820541382,
      "rewards-max/forensics_sinkhorn_reward": 0.9850457012653351,
      "rewards-max/forensics_strict_edge_reward": 0.748430609703064,
      "rewards-max/forensics_tiou_reward": 0.9895692765712738,
      "rewards/forensics_sinkhorn_reward": 0.6447398364543915,
      "rewards/forensics_strict_edge_reward": 0.05955936573445797,
      "rewards/forensics_tiou_reward": 0.4680730402469635,
      "step": 804
    },
    {
      "completion_length": 25.859375,
      "epoch": 2.4700460829493087,
      "grad_norm": 9.914285659790039,
      "kl": 0.4169921875,
      "learning_rate": 3.8153846153846153e-07,
      "loss": -0.0646,
      "reward": 1.3752788305282593,
      "reward_std": 0.23119128495454788,
      "reward_std_all": 0.8616660833358765,
      "rewards-max/forensics_sinkhorn_reward": 0.9820095300674438,
      "rewards-max/forensics_strict_edge_reward": 0.8032653331756592,
      "rewards-max/forensics_tiou_reward": 0.9816897809505463,
      "rewards/forensics_sinkhorn_reward": 0.7408091723918915,
      "rewards/forensics_strict_edge_reward": 0.0648858342319727,
      "rewards/forensics_tiou_reward": 0.5695838034152985,
      "step": 805
    },
    {
      "completion_length": 24.328125,
      "epoch": 2.4731182795698925,
      "grad_norm": 8.369131088256836,
      "kl": 0.5439453125,
      "learning_rate": 3.8076923076923075e-07,
      "loss": -0.0977,
      "reward": 1.2868791818618774,
      "reward_std": 0.3278619050979614,
      "reward_std_all": 0.9156723916530609,
      "rewards-max/forensics_sinkhorn_reward": 0.9770516753196716,
      "rewards-max/forensics_strict_edge_reward": 0.40866968780755997,
      "rewards-max/forensics_tiou_reward": 0.967473566532135,
      "rewards/forensics_sinkhorn_reward": 0.7257072329521179,
      "rewards/forensics_strict_edge_reward": 0.047081658616662025,
      "rewards/forensics_tiou_reward": 0.5140902996063232,
      "step": 806
    },
    {
      "completion_length": 26.28125,
      "epoch": 2.4761904761904763,
      "grad_norm": 11.345295906066895,
      "kl": 0.583984375,
      "learning_rate": 3.7999999999999996e-07,
      "loss": -0.0812,
      "reward": 1.3331313729286194,
      "reward_std": 0.28943391144275665,
      "reward_std_all": 0.919133186340332,
      "rewards-max/forensics_sinkhorn_reward": 0.9897684156894684,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.7235396206378937,
      "rewards/forensics_strict_edge_reward": 0.10271404311060905,
      "rewards/forensics_tiou_reward": 0.5068777203559875,
      "step": 807
    },
    {
      "completion_length": 24.171875,
      "epoch": 2.47926267281106,
      "grad_norm": 9.303526878356934,
      "kl": 0.400390625,
      "learning_rate": 3.792307692307692e-07,
      "loss": -0.0645,
      "reward": 1.448747158050537,
      "reward_std": 0.3303525298833847,
      "reward_std_all": 0.8632470071315765,
      "rewards-max/forensics_sinkhorn_reward": 0.986586332321167,
      "rewards-max/forensics_strict_edge_reward": 0.7692224383354187,
      "rewards-max/forensics_tiou_reward": 0.9958298206329346,
      "rewards/forensics_sinkhorn_reward": 0.7840599119663239,
      "rewards/forensics_strict_edge_reward": 0.06525013968348503,
      "rewards/forensics_tiou_reward": 0.5994371175765991,
      "step": 808
    },
    {
      "completion_length": 26.671875,
      "epoch": 2.482334869431644,
      "grad_norm": 10.198705673217773,
      "kl": 0.431640625,
      "learning_rate": 3.7846153846153844e-07,
      "loss": -0.0165,
      "reward": 1.3959966897964478,
      "reward_std": 0.22431198507547379,
      "reward_std_all": 0.8460959792137146,
      "rewards-max/forensics_sinkhorn_reward": 0.9746706485748291,
      "rewards-max/forensics_strict_edge_reward": 0.7673708498477936,
      "rewards-max/forensics_tiou_reward": 0.9657939970493317,
      "rewards/forensics_sinkhorn_reward": 0.7316329479217529,
      "rewards/forensics_strict_edge_reward": 0.05533015914261341,
      "rewards/forensics_tiou_reward": 0.609033614397049,
      "step": 809
    },
    {
      "completion_length": 29.1875,
      "epoch": 2.485407066052227,
      "grad_norm": 10.408923149108887,
      "kl": 0.6328125,
      "learning_rate": 3.776923076923077e-07,
      "loss": -0.1668,
      "reward": 1.4837928414344788,
      "reward_std": 0.4437883049249649,
      "reward_std_all": 0.888228714466095,
      "rewards-max/forensics_sinkhorn_reward": 0.9870782196521759,
      "rewards-max/forensics_strict_edge_reward": 0.8032653331756592,
      "rewards-max/forensics_tiou_reward": 0.9953345358371735,
      "rewards/forensics_sinkhorn_reward": 0.7417938113212585,
      "rewards/forensics_strict_edge_reward": 0.13190284371376038,
      "rewards/forensics_tiou_reward": 0.6100961565971375,
      "step": 810
    },
    {
      "completion_length": 25.125,
      "epoch": 2.488479262672811,
      "grad_norm": 9.863043785095215,
      "kl": 0.4462890625,
      "learning_rate": 3.769230769230769e-07,
      "loss": -0.0266,
      "reward": 1.4312688112258911,
      "reward_std": 0.31934261322021484,
      "reward_std_all": 0.8617672622203827,
      "rewards-max/forensics_sinkhorn_reward": 0.9878199994564056,
      "rewards-max/forensics_strict_edge_reward": 0.8608468174934387,
      "rewards-max/forensics_tiou_reward": 0.9951876699924469,
      "rewards/forensics_sinkhorn_reward": 0.744511753320694,
      "rewards/forensics_strict_edge_reward": 0.09260722622275352,
      "rewards/forensics_tiou_reward": 0.5941498875617981,
      "step": 811
    },
    {
      "completion_length": 26.75,
      "epoch": 2.4915514592933947,
      "grad_norm": 11.15906047821045,
      "kl": 0.55859375,
      "learning_rate": 3.7615384615384614e-07,
      "loss": -0.0161,
      "reward": 0.9926208853721619,
      "reward_std": 0.36604754626750946,
      "reward_std_all": 1.0687109231948853,
      "rewards-max/forensics_sinkhorn_reward": 0.9520543217658997,
      "rewards-max/forensics_strict_edge_reward": 0.32260795682668686,
      "rewards-max/forensics_tiou_reward": 0.9045713543891907,
      "rewards/forensics_sinkhorn_reward": 0.5502973943948746,
      "rewards/forensics_strict_edge_reward": 0.02919198665767908,
      "rewards/forensics_tiou_reward": 0.4131315052509308,
      "step": 812
    },
    {
      "completion_length": 27.28125,
      "epoch": 2.4946236559139785,
      "grad_norm": 9.32443904876709,
      "kl": 0.521484375,
      "learning_rate": 3.7538461538461535e-07,
      "loss": 0.0184,
      "reward": 1.5241286158561707,
      "reward_std": 0.38647493720054626,
      "reward_std_all": 0.8398232460021973,
      "rewards-max/forensics_sinkhorn_reward": 0.9836045205593109,
      "rewards-max/forensics_strict_edge_reward": 0.7536259889602661,
      "rewards-max/forensics_tiou_reward": 0.9865924715995789,
      "rewards/forensics_sinkhorn_reward": 0.7824051678180695,
      "rewards/forensics_strict_edge_reward": 0.08890550211071968,
      "rewards/forensics_tiou_reward": 0.6528180837631226,
      "step": 813
    },
    {
      "completion_length": 31.09375,
      "epoch": 2.4976958525345623,
      "grad_norm": 11.00417423248291,
      "kl": 0.509765625,
      "learning_rate": 3.746153846153846e-07,
      "loss": -0.0489,
      "reward": 1.60801500082016,
      "reward_std": 0.18660122156143188,
      "reward_std_all": 0.7237017750740051,
      "rewards-max/forensics_sinkhorn_reward": 0.9813441932201385,
      "rewards-max/forensics_strict_edge_reward": 0.442320853471756,
      "rewards-max/forensics_tiou_reward": 0.983813464641571,
      "rewards/forensics_sinkhorn_reward": 0.8459278643131256,
      "rewards/forensics_strict_edge_reward": 0.054755743592977524,
      "rewards/forensics_tiou_reward": 0.7073314189910889,
      "step": 814
    },
    {
      "completion_length": 23.296875,
      "epoch": 2.500768049155146,
      "grad_norm": 10.932146072387695,
      "kl": 0.572265625,
      "learning_rate": 3.7384615384615383e-07,
      "loss": -0.0385,
      "reward": 1.4678453207015991,
      "reward_std": 0.29158367216587067,
      "reward_std_all": 0.8165684640407562,
      "rewards-max/forensics_sinkhorn_reward": 0.9548473954200745,
      "rewards-max/forensics_strict_edge_reward": 0.5154097378253937,
      "rewards-max/forensics_tiou_reward": 0.919759601354599,
      "rewards/forensics_sinkhorn_reward": 0.7910730242729187,
      "rewards/forensics_strict_edge_reward": 0.06280802562832832,
      "rewards/forensics_tiou_reward": 0.613964319229126,
      "step": 815
    },
    {
      "completion_length": 24.171875,
      "epoch": 2.50384024577573,
      "grad_norm": 10.372095108032227,
      "kl": 0.513671875,
      "learning_rate": 3.7307692307692304e-07,
      "loss": -0.1656,
      "reward": 1.8129173517227173,
      "reward_std": 0.48816509544849396,
      "reward_std_all": 0.7422972917556763,
      "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": 0.8525471985340118,
      "rewards/forensics_strict_edge_reward": 0.2253614142537117,
      "rewards/forensics_tiou_reward": 0.7350087761878967,
      "step": 816
    },
    {
      "completion_length": 24.21875,
      "epoch": 2.506912442396313,
      "grad_norm": 9.99441909790039,
      "kl": 0.4228515625,
      "learning_rate": 3.7230769230769226e-07,
      "loss": -0.1078,
      "reward": 1.2967884540557861,
      "reward_std": 0.30621880292892456,
      "reward_std_all": 0.9058824777603149,
      "rewards-max/forensics_sinkhorn_reward": 0.9769359230995178,
      "rewards-max/forensics_strict_edge_reward": 0.40513233840465546,
      "rewards-max/forensics_tiou_reward": 0.9671445488929749,
      "rewards/forensics_sinkhorn_reward": 0.7208194136619568,
      "rewards/forensics_strict_edge_reward": 0.044782282784581184,
      "rewards/forensics_tiou_reward": 0.5311867892742157,
      "step": 817
    },
    {
      "completion_length": 26.25,
      "epoch": 2.509984639016897,
      "grad_norm": 8.237133026123047,
      "kl": 0.46875,
      "learning_rate": 3.715384615384615e-07,
      "loss": -0.1059,
      "reward": 1.331237554550171,
      "reward_std": 0.48126381635665894,
      "reward_std_all": 0.9821901619434357,
      "rewards-max/forensics_sinkhorn_reward": 0.9858314990997314,
      "rewards-max/forensics_strict_edge_reward": 0.7711026966571808,
      "rewards-max/forensics_tiou_reward": 0.989331990480423,
      "rewards/forensics_sinkhorn_reward": 0.7073506712913513,
      "rewards/forensics_strict_edge_reward": 0.09586656093597412,
      "rewards/forensics_tiou_reward": 0.5280202031135559,
      "step": 818
    },
    {
      "completion_length": 28.265625,
      "epoch": 2.5130568356374807,
      "grad_norm": 9.150012969970703,
      "kl": 0.6455078125,
      "learning_rate": 3.707692307692308e-07,
      "loss": -0.0692,
      "reward": 1.6460918188095093,
      "reward_std": 0.36411720514297485,
      "reward_std_all": 0.7666225731372833,
      "rewards-max/forensics_sinkhorn_reward": 0.9850800037384033,
      "rewards-max/forensics_strict_edge_reward": 0.5773773044347763,
      "rewards-max/forensics_tiou_reward": 0.9891246259212494,
      "rewards/forensics_sinkhorn_reward": 0.8361402153968811,
      "rewards/forensics_strict_edge_reward": 0.09195315837860107,
      "rewards/forensics_tiou_reward": 0.7179985046386719,
      "step": 819
    },
    {
      "completion_length": 28.875,
      "epoch": 2.5161290322580645,
      "grad_norm": 9.357832908630371,
      "kl": 0.599609375,
      "learning_rate": 3.7e-07,
      "loss": -0.0555,
      "reward": 1.3377931118011475,
      "reward_std": 0.40720854699611664,
      "reward_std_all": 0.9359927177429199,
      "rewards-max/forensics_sinkhorn_reward": 0.9827360808849335,
      "rewards-max/forensics_strict_edge_reward": 0.7559782564640045,
      "rewards-max/forensics_tiou_reward": 0.9830048382282257,
      "rewards/forensics_sinkhorn_reward": 0.6715312302112579,
      "rewards/forensics_strict_edge_reward": 0.10493096336722374,
      "rewards/forensics_tiou_reward": 0.5613308548927307,
      "step": 820
    },
    {
      "completion_length": 28.4375,
      "epoch": 2.5192012288786483,
      "grad_norm": 9.122920036315918,
      "kl": 0.4853515625,
      "learning_rate": 3.692307692307692e-07,
      "loss": -0.0224,
      "reward": 1.5728961825370789,
      "reward_std": 0.3630540147423744,
      "reward_std_all": 0.8184563517570496,
      "rewards-max/forensics_sinkhorn_reward": 0.9830464124679565,
      "rewards-max/forensics_strict_edge_reward": 0.712966650724411,
      "rewards-max/forensics_tiou_reward": 0.9871642291545868,
      "rewards/forensics_sinkhorn_reward": 0.7720848321914673,
      "rewards/forensics_strict_edge_reward": 0.12831304781138897,
      "rewards/forensics_tiou_reward": 0.6724982857704163,
      "step": 821
    },
    {
      "completion_length": 26.4375,
      "epoch": 2.522273425499232,
      "grad_norm": 10.127277374267578,
      "kl": 0.548828125,
      "learning_rate": 3.6846153846153843e-07,
      "loss": -0.0471,
      "reward": 1.6091651916503906,
      "reward_std": 0.44202861189842224,
      "reward_std_all": 0.8369906842708588,
      "rewards-max/forensics_sinkhorn_reward": 0.9885551035404205,
      "rewards-max/forensics_strict_edge_reward": 0.9836080372333527,
      "rewards-max/forensics_tiou_reward": 0.999909520149231,
      "rewards/forensics_sinkhorn_reward": 0.7675655782222748,
      "rewards/forensics_strict_edge_reward": 0.17541904747486115,
      "rewards/forensics_tiou_reward": 0.6661805808544159,
      "step": 822
    },
    {
      "completion_length": 27.234375,
      "epoch": 2.525345622119816,
      "grad_norm": 8.93156623840332,
      "kl": 0.43359375,
      "learning_rate": 3.676923076923077e-07,
      "loss": -0.1048,
      "reward": 1.4899046421051025,
      "reward_std": 0.4067216068506241,
      "reward_std_all": 0.8441019058227539,
      "rewards-max/forensics_sinkhorn_reward": 0.9864300191402435,
      "rewards-max/forensics_strict_edge_reward": 0.6360848695039749,
      "rewards-max/forensics_tiou_reward": 0.9955655932426453,
      "rewards/forensics_sinkhorn_reward": 0.7484873533248901,
      "rewards/forensics_strict_edge_reward": 0.1012221947312355,
      "rewards/forensics_tiou_reward": 0.6401951014995575,
      "step": 823
    },
    {
      "completion_length": 29.421875,
      "epoch": 2.528417818740399,
      "grad_norm": 9.126818656921387,
      "kl": 0.4833984375,
      "learning_rate": 3.669230769230769e-07,
      "loss": -0.0735,
      "reward": 1.3475595712661743,
      "reward_std": 0.23902662098407745,
      "reward_std_all": 0.8814724087715149,
      "rewards-max/forensics_sinkhorn_reward": 0.9789712727069855,
      "rewards-max/forensics_strict_edge_reward": 0.8135445415973663,
      "rewards-max/forensics_tiou_reward": 0.9770887196063995,
      "rewards/forensics_sinkhorn_reward": 0.6925501227378845,
      "rewards/forensics_strict_edge_reward": 0.0916927419602871,
      "rewards/forensics_tiou_reward": 0.5633167624473572,
      "step": 824
    },
    {
      "completion_length": 26.546875,
      "epoch": 2.531490015360983,
      "grad_norm": 9.454221725463867,
      "kl": 0.4912109375,
      "learning_rate": 3.661538461538461e-07,
      "loss": -0.1865,
      "reward": 1.5445352792739868,
      "reward_std": 0.4097803980112076,
      "reward_std_all": 0.8449422121047974,
      "rewards-max/forensics_sinkhorn_reward": 0.9882118701934814,
      "rewards-max/forensics_strict_edge_reward": 0.8605633974075317,
      "rewards-max/forensics_tiou_reward": 0.9954562783241272,
      "rewards/forensics_sinkhorn_reward": 0.7971993982791901,
      "rewards/forensics_strict_edge_reward": 0.11390944197773933,
      "rewards/forensics_tiou_reward": 0.6334264576435089,
      "step": 825
    },
    {
      "completion_length": 29.21875,
      "epoch": 2.5345622119815667,
      "grad_norm": 10.33183479309082,
      "kl": 0.544921875,
      "learning_rate": 3.6538461538461534e-07,
      "loss": -0.0254,
      "reward": 1.2010448575019836,
      "reward_std": 0.6773447692394257,
      "reward_std_all": 1.1034331917762756,
      "rewards-max/forensics_sinkhorn_reward": 0.9883970022201538,
      "rewards-max/forensics_strict_edge_reward": 0.882964164018631,
      "rewards-max/forensics_tiou_reward": 0.9977434873580933,
      "rewards/forensics_sinkhorn_reward": 0.622747927904129,
      "rewards/forensics_strict_edge_reward": 0.08963306993246078,
      "rewards/forensics_tiou_reward": 0.48866385221481323,
      "step": 826
    },
    {
      "completion_length": 23.265625,
      "epoch": 2.5376344086021505,
      "grad_norm": 7.689429759979248,
      "kl": 0.4677734375,
      "learning_rate": 3.646153846153846e-07,
      "loss": -0.006,
      "reward": 1.3757829666137695,
      "reward_std": 0.24203689396381378,
      "reward_std_all": 0.8651747703552246,
      "rewards-max/forensics_sinkhorn_reward": 0.9676474034786224,
      "rewards-max/forensics_strict_edge_reward": 0.5662333369255066,
      "rewards-max/forensics_tiou_reward": 0.948723554611206,
      "rewards/forensics_sinkhorn_reward": 0.7670065760612488,
      "rewards/forensics_strict_edge_reward": 0.04647386260330677,
      "rewards/forensics_tiou_reward": 0.5623023808002472,
      "step": 827
    },
    {
      "completion_length": 24.46875,
      "epoch": 2.5407066052227343,
      "grad_norm": 10.601856231689453,
      "kl": 0.4814453125,
      "learning_rate": 3.638461538461538e-07,
      "loss": -0.0888,
      "reward": 1.4644359350204468,
      "reward_std": 0.4385008215904236,
      "reward_std_all": 0.9001374840736389,
      "rewards-max/forensics_sinkhorn_reward": 0.9882524907588959,
      "rewards-max/forensics_strict_edge_reward": 0.8661397695541382,
      "rewards-max/forensics_tiou_reward": 0.9991316497325897,
      "rewards/forensics_sinkhorn_reward": 0.7490496039390564,
      "rewards/forensics_strict_edge_reward": 0.15037551149725914,
      "rewards/forensics_tiou_reward": 0.5650107860565186,
      "step": 828
    },
    {
      "completion_length": 23.75,
      "epoch": 2.543778801843318,
      "grad_norm": 8.311702728271484,
      "kl": 0.4287109375,
      "learning_rate": 3.6307692307692303e-07,
      "loss": -0.1415,
      "reward": 1.5407096147537231,
      "reward_std": 0.3622756451368332,
      "reward_std_all": 0.8350113034248352,
      "rewards-max/forensics_sinkhorn_reward": 0.9844838976860046,
      "rewards-max/forensics_strict_edge_reward": 0.8205508291721344,
      "rewards-max/forensics_tiou_reward": 0.9870081543922424,
      "rewards/forensics_sinkhorn_reward": 0.7598620653152466,
      "rewards/forensics_strict_edge_reward": 0.14030295610427856,
      "rewards/forensics_tiou_reward": 0.6405446231365204,
      "step": 829
    },
    {
      "completion_length": 25.078125,
      "epoch": 2.546850998463902,
      "grad_norm": 11.476137161254883,
      "kl": 0.431640625,
      "learning_rate": 3.623076923076923e-07,
      "loss": -0.1078,
      "reward": 1.305553138256073,
      "reward_std": 0.3315027505159378,
      "reward_std_all": 0.9205743372440338,
      "rewards-max/forensics_sinkhorn_reward": 0.9863654375076294,
      "rewards-max/forensics_strict_edge_reward": 0.7507568895816803,
      "rewards-max/forensics_tiou_reward": 0.9940849840641022,
      "rewards/forensics_sinkhorn_reward": 0.6994295418262482,
      "rewards/forensics_strict_edge_reward": 0.07571359537541866,
      "rewards/forensics_tiou_reward": 0.5304100215435028,
      "step": 830
    },
    {
      "completion_length": 26.046875,
      "epoch": 2.549923195084485,
      "grad_norm": 10.038811683654785,
      "kl": 0.4296875,
      "learning_rate": 3.615384615384615e-07,
      "loss": -0.1197,
      "reward": 1.2073301076889038,
      "reward_std": 0.3482375741004944,
      "reward_std_all": 0.9900205135345459,
      "rewards-max/forensics_sinkhorn_reward": 0.9837981760501862,
      "rewards-max/forensics_strict_edge_reward": 0.7361641973257065,
      "rewards-max/forensics_tiou_reward": 0.9853841960430145,
      "rewards/forensics_sinkhorn_reward": 0.6475521922111511,
      "rewards/forensics_strict_edge_reward": 0.06101858988404274,
      "rewards/forensics_tiou_reward": 0.49875935912132263,
      "step": 831
    },
    {
      "completion_length": 25.28125,
      "epoch": 2.5529953917050694,
      "grad_norm": 11.940427780151367,
      "kl": 0.5224609375,
      "learning_rate": 3.607692307692308e-07,
      "loss": -0.1333,
      "reward": 1.7273424863815308,
      "reward_std": 0.44959673285484314,
      "reward_std_all": 0.7677730023860931,
      "rewards-max/forensics_sinkhorn_reward": 0.9879885613918304,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.8266918063163757,
      "rewards/forensics_strict_edge_reward": 0.16149187088012695,
      "rewards/forensics_tiou_reward": 0.7391588687896729,
      "step": 832
    },
    {
      "completion_length": 24.140625,
      "epoch": 2.5560675883256527,
      "grad_norm": 10.236939430236816,
      "kl": 0.5302734375,
      "learning_rate": 3.6e-07,
      "loss": -0.0521,
      "reward": 1.3726208209991455,
      "reward_std": 0.4441719800233841,
      "reward_std_all": 0.9172269701957703,
      "rewards-max/forensics_sinkhorn_reward": 0.9798200130462646,
      "rewards-max/forensics_strict_edge_reward": 0.5791139304637909,
      "rewards-max/forensics_tiou_reward": 0.9787819981575012,
      "rewards/forensics_sinkhorn_reward": 0.7409661412239075,
      "rewards/forensics_strict_edge_reward": 0.0697002075612545,
      "rewards/forensics_tiou_reward": 0.5619544386863708,
      "step": 833
    },
    {
      "completion_length": 22.8125,
      "epoch": 2.5591397849462365,
      "grad_norm": 13.392203330993652,
      "kl": 0.4462890625,
      "learning_rate": 3.592307692307692e-07,
      "loss": -0.1088,
      "reward": 1.3960835337638855,
      "reward_std": 0.30704016983509064,
      "reward_std_all": 0.8851033449172974,
      "rewards-max/forensics_sinkhorn_reward": 0.9822401702404022,
      "rewards-max/forensics_strict_edge_reward": 0.5347211100161076,
      "rewards-max/forensics_tiou_reward": 0.9817591905593872,
      "rewards/forensics_sinkhorn_reward": 0.7514649629592896,
      "rewards/forensics_strict_edge_reward": 0.05185235384851694,
      "rewards/forensics_tiou_reward": 0.5927661955356598,
      "step": 834
    },
    {
      "completion_length": 24.859375,
      "epoch": 2.5622119815668203,
      "grad_norm": 24.38628387451172,
      "kl": 1.705078125,
      "learning_rate": 3.584615384615384e-07,
      "loss": -0.064,
      "reward": 1.3298792243003845,
      "reward_std": 0.4470940977334976,
      "reward_std_all": 0.9662792980670929,
      "rewards-max/forensics_sinkhorn_reward": 0.9879054725170135,
      "rewards-max/forensics_strict_edge_reward": 0.9291328191757202,
      "rewards-max/forensics_tiou_reward": 0.9970242083072662,
      "rewards/forensics_sinkhorn_reward": 0.7073929905891418,
      "rewards/forensics_strict_edge_reward": 0.08532825112342834,
      "rewards/forensics_tiou_reward": 0.5371580123901367,
      "step": 835
    },
    {
      "completion_length": 25.09375,
      "epoch": 2.565284178187404,
      "grad_norm": 12.657404899597168,
      "kl": 0.431640625,
      "learning_rate": 3.576923076923077e-07,
      "loss": -0.1498,
      "reward": 1.1831318736076355,
      "reward_std": 0.325104795396328,
      "reward_std_all": 0.978580117225647,
      "rewards-max/forensics_sinkhorn_reward": 0.9856205582618713,
      "rewards-max/forensics_strict_edge_reward": 0.7516844868659973,
      "rewards-max/forensics_tiou_reward": 0.9915767014026642,
      "rewards/forensics_sinkhorn_reward": 0.6482837796211243,
      "rewards/forensics_strict_edge_reward": 0.08254212513566017,
      "rewards/forensics_tiou_reward": 0.4523060619831085,
      "step": 836
    },
    {
      "completion_length": 22.609375,
      "epoch": 2.568356374807988,
      "grad_norm": 10.63211727142334,
      "kl": 0.5517578125,
      "learning_rate": 3.569230769230769e-07,
      "loss": -0.0198,
      "reward": 1.2415051460266113,
      "reward_std": 0.33761435747146606,
      "reward_std_all": 0.9546472728252411,
      "rewards-max/forensics_sinkhorn_reward": 0.9775470197200775,
      "rewards-max/forensics_strict_edge_reward": 0.6149742305278778,
      "rewards-max/forensics_tiou_reward": 0.9700602889060974,
      "rewards/forensics_sinkhorn_reward": 0.6587196588516235,
      "rewards/forensics_strict_edge_reward": 0.07657863199710846,
      "rewards/forensics_tiou_reward": 0.5062067955732346,
      "step": 837
    },
    {
      "completion_length": 23.671875,
      "epoch": 2.571428571428571,
      "grad_norm": 10.549051284790039,
      "kl": 0.505859375,
      "learning_rate": 3.561538461538461e-07,
      "loss": -0.1379,
      "reward": 1.6172863245010376,
      "reward_std": 0.29157721996307373,
      "reward_std_all": 0.7566483318805695,
      "rewards-max/forensics_sinkhorn_reward": 0.9839317798614502,
      "rewards-max/forensics_strict_edge_reward": 0.7703645825386047,
      "rewards-max/forensics_tiou_reward": 0.9860822856426239,
      "rewards/forensics_sinkhorn_reward": 0.8287932276725769,
      "rewards/forensics_strict_edge_reward": 0.08915351331233978,
      "rewards/forensics_tiou_reward": 0.6993396580219269,
      "step": 838
    },
    {
      "completion_length": 31.0,
      "epoch": 2.5745007680491554,
      "grad_norm": 12.748650550842285,
      "kl": 0.650390625,
      "learning_rate": 3.553846153846154e-07,
      "loss": -0.1466,
      "reward": 1.2064149975776672,
      "reward_std": 0.3540450185537338,
      "reward_std_all": 0.9824413657188416,
      "rewards-max/forensics_sinkhorn_reward": 0.9851578772068024,
      "rewards-max/forensics_strict_edge_reward": 0.8419698476791382,
      "rewards-max/forensics_tiou_reward": 0.9899424314498901,
      "rewards/forensics_sinkhorn_reward": 0.6344037353992462,
      "rewards/forensics_strict_edge_reward": 0.08700889348983765,
      "rewards/forensics_tiou_reward": 0.4850023686885834,
      "step": 839
    },
    {
      "completion_length": 23.21875,
      "epoch": 2.5775729646697387,
      "grad_norm": 9.562209129333496,
      "kl": 0.51171875,
      "learning_rate": 3.546153846153846e-07,
      "loss": -0.1087,
      "reward": 1.5651803016662598,
      "reward_std": 0.4487857073545456,
      "reward_std_all": 0.8585265278816223,
      "rewards-max/forensics_sinkhorn_reward": 0.9838782250881195,
      "rewards-max/forensics_strict_edge_reward": 0.8465203046798706,
      "rewards-max/forensics_tiou_reward": 0.9873611629009247,
      "rewards/forensics_sinkhorn_reward": 0.7839192152023315,
      "rewards/forensics_strict_edge_reward": 0.14541762322187424,
      "rewards/forensics_tiou_reward": 0.635843425989151,
      "step": 840
    },
    {
      "completion_length": 25.046875,
      "epoch": 2.5806451612903225,
      "grad_norm": 11.212993621826172,
      "kl": 0.703125,
      "learning_rate": 3.5384615384615386e-07,
      "loss": -0.0598,
      "reward": 1.3439223766326904,
      "reward_std": 0.4439942091703415,
      "reward_std_all": 0.9404868483543396,
      "rewards-max/forensics_sinkhorn_reward": 0.981837272644043,
      "rewards-max/forensics_strict_edge_reward": 0.6736744344234467,
      "rewards-max/forensics_tiou_reward": 0.9800163805484772,
      "rewards/forensics_sinkhorn_reward": 0.7368066608905792,
      "rewards/forensics_strict_edge_reward": 0.06672412157058716,
      "rewards/forensics_tiou_reward": 0.5403916537761688,
      "step": 841
    },
    {
      "completion_length": 20.96875,
      "epoch": 2.5837173579109063,
      "grad_norm": 9.83584213256836,
      "kl": 0.4921875,
      "learning_rate": 3.530769230769231e-07,
      "loss": -0.0643,
      "reward": 1.3134638667106628,
      "reward_std": 0.4127471446990967,
      "reward_std_all": 0.9310185313224792,
      "rewards-max/forensics_sinkhorn_reward": 0.9737913310527802,
      "rewards-max/forensics_strict_edge_reward": 0.5062484890222549,
      "rewards-max/forensics_tiou_reward": 0.9615734815597534,
      "rewards/forensics_sinkhorn_reward": 0.7480255663394928,
      "rewards/forensics_strict_edge_reward": 0.056302087381482124,
      "rewards/forensics_tiou_reward": 0.5091362446546555,
      "step": 842
    },
    {
      "completion_length": 29.65625,
      "epoch": 2.58678955453149,
      "grad_norm": 8.739901542663574,
      "kl": 0.541015625,
      "learning_rate": 3.523076923076923e-07,
      "loss": -0.0567,
      "reward": 1.4575762152671814,
      "reward_std": 0.5520125925540924,
      "reward_std_all": 0.9355786442756653,
      "rewards-max/forensics_sinkhorn_reward": 0.9831136167049408,
      "rewards-max/forensics_strict_edge_reward": 0.7851271331310272,
      "rewards-max/forensics_tiou_reward": 0.9848503172397614,
      "rewards/forensics_sinkhorn_reward": 0.7423162758350372,
      "rewards/forensics_strict_edge_reward": 0.1202898658812046,
      "rewards/forensics_tiou_reward": 0.594970166683197,
      "step": 843
    },
    {
      "completion_length": 26.5625,
      "epoch": 2.589861751152074,
      "grad_norm": 9.39455509185791,
      "kl": 0.630859375,
      "learning_rate": 3.515384615384615e-07,
      "loss": -0.0626,
      "reward": 1.5718948245048523,
      "reward_std": 0.3693888485431671,
      "reward_std_all": 0.7992021441459656,
      "rewards-max/forensics_sinkhorn_reward": 0.9858767092227936,
      "rewards-max/forensics_strict_edge_reward": 0.8582656681537628,
      "rewards-max/forensics_tiou_reward": 0.9908266365528107,
      "rewards/forensics_sinkhorn_reward": 0.7890271246433258,
      "rewards/forensics_strict_edge_reward": 0.11148746311664581,
      "rewards/forensics_tiou_reward": 0.671380341053009,
      "step": 844
    },
    {
      "completion_length": 26.859375,
      "epoch": 2.592933947772657,
      "grad_norm": 14.683533668518066,
      "kl": 0.4873046875,
      "learning_rate": 3.5076923076923077e-07,
      "loss": -0.0776,
      "reward": 1.3950218558311462,
      "reward_std": 0.39956480264663696,
      "reward_std_all": 0.9075792729854584,
      "rewards-max/forensics_sinkhorn_reward": 0.9832356572151184,
      "rewards-max/forensics_strict_edge_reward": 0.6103231012821198,
      "rewards-max/forensics_tiou_reward": 0.9857019186019897,
      "rewards/forensics_sinkhorn_reward": 0.7425695657730103,
      "rewards/forensics_strict_edge_reward": 0.0789586678147316,
      "rewards/forensics_tiou_reward": 0.5734935998916626,
      "step": 845
    },
    {
      "completion_length": 29.453125,
      "epoch": 2.5960061443932414,
      "grad_norm": 65.51496124267578,
      "kl": 2.2275390625,
      "learning_rate": 3.5e-07,
      "loss": 0.0108,
      "reward": 1.6861409544944763,
      "reward_std": 0.2671215534210205,
      "reward_std_all": 0.7198358178138733,
      "rewards-max/forensics_sinkhorn_reward": 0.9845038056373596,
      "rewards-max/forensics_strict_edge_reward": 0.6698123216629028,
      "rewards-max/forensics_tiou_reward": 0.9869026839733124,
      "rewards/forensics_sinkhorn_reward": 0.8557707667350769,
      "rewards/forensics_strict_edge_reward": 0.09111810475587845,
      "rewards/forensics_tiou_reward": 0.739252120256424,
      "step": 846
    },
    {
      "completion_length": 26.234375,
      "epoch": 2.5990783410138247,
      "grad_norm": 9.129587173461914,
      "kl": 0.376953125,
      "learning_rate": 3.492307692307692e-07,
      "loss": -0.0954,
      "reward": 1.4711222648620605,
      "reward_std": 0.3561341315507889,
      "reward_std_all": 0.8585544526576996,
      "rewards-max/forensics_sinkhorn_reward": 0.9844297766685486,
      "rewards-max/forensics_strict_edge_reward": 0.8648126125335693,
      "rewards-max/forensics_tiou_reward": 0.9872324764728546,
      "rewards/forensics_sinkhorn_reward": 0.7622898817062378,
      "rewards/forensics_strict_edge_reward": 0.09410316497087479,
      "rewards/forensics_tiou_reward": 0.6147291958332062,
      "step": 847
    },
    {
      "completion_length": 28.171875,
      "epoch": 2.6021505376344085,
      "grad_norm": 8.425909042358398,
      "kl": 0.4755859375,
      "learning_rate": 3.4846153846153846e-07,
      "loss": -0.0539,
      "reward": 1.2557151317596436,
      "reward_std": 0.3138905465602875,
      "reward_std_all": 0.9476407468318939,
      "rewards-max/forensics_sinkhorn_reward": 0.9864312410354614,
      "rewards-max/forensics_strict_edge_reward": 0.7836964428424835,
      "rewards-max/forensics_tiou_reward": 0.9936107993125916,
      "rewards/forensics_sinkhorn_reward": 0.6912174820899963,
      "rewards/forensics_strict_edge_reward": 0.08383860811591148,
      "rewards/forensics_tiou_reward": 0.48065897822380066,
      "step": 848
    },
    {
      "completion_length": 24.40625,
      "epoch": 2.6052227342549923,
      "grad_norm": 12.213801383972168,
      "kl": 0.701171875,
      "learning_rate": 3.476923076923077e-07,
      "loss": -0.0589,
      "reward": 1.5521699786186218,
      "reward_std": 0.42259564995765686,
      "reward_std_all": 0.8243915140628815,
      "rewards-max/forensics_sinkhorn_reward": 0.9851029813289642,
      "rewards-max/forensics_strict_edge_reward": 0.796095460653305,
      "rewards-max/forensics_tiou_reward": 0.9907054007053375,
      "rewards/forensics_sinkhorn_reward": 0.7710353136062622,
      "rewards/forensics_strict_edge_reward": 0.09003240242600441,
      "rewards/forensics_tiou_reward": 0.6911022663116455,
      "step": 849
    },
    {
      "completion_length": 21.140625,
      "epoch": 2.608294930875576,
      "grad_norm": 10.959440231323242,
      "kl": 0.3720703125,
      "learning_rate": 3.4692307692307694e-07,
      "loss": -0.1137,
      "reward": 1.620132327079773,
      "reward_std": 0.3388683348894119,
      "reward_std_all": 0.7727493643760681,
      "rewards-max/forensics_sinkhorn_reward": 0.9814070165157318,
      "rewards-max/forensics_strict_edge_reward": 0.7514603435993195,
      "rewards-max/forensics_tiou_reward": 0.9806199669837952,
      "rewards/forensics_sinkhorn_reward": 0.8325689435005188,
      "rewards/forensics_strict_edge_reward": 0.11120236665010452,
      "rewards/forensics_tiou_reward": 0.6763610243797302,
      "step": 850
    },
    {
      "completion_length": 24.734375,
      "epoch": 2.61136712749616,
      "grad_norm": 10.121410369873047,
      "kl": 0.3896484375,
      "learning_rate": 3.461538461538461e-07,
      "loss": -0.1405,
      "reward": 1.4227274060249329,
      "reward_std": 0.24407034367322922,
      "reward_std_all": 0.8305774331092834,
      "rewards-max/forensics_sinkhorn_reward": 0.9720501005649567,
      "rewards-max/forensics_strict_edge_reward": 0.44042593240737915,
      "rewards-max/forensics_tiou_reward": 0.9517211318016052,
      "rewards/forensics_sinkhorn_reward": 0.7537966668605804,
      "rewards/forensics_strict_edge_reward": 0.04854828305542469,
      "rewards/forensics_tiou_reward": 0.6203825175762177,
      "step": 851
    },
    {
      "completion_length": 28.328125,
      "epoch": 2.6144393241167436,
      "grad_norm": 10.146053314208984,
      "kl": 0.4404296875,
      "learning_rate": 3.4538461538461537e-07,
      "loss": -0.0538,
      "reward": 1.2843430638313293,
      "reward_std": 0.25711341947317123,
      "reward_std_all": 0.8991884291172028,
      "rewards-max/forensics_sinkhorn_reward": 0.969622790813446,
      "rewards-max/forensics_strict_edge_reward": 0.25842931866645813,
      "rewards-max/forensics_tiou_reward": 0.9407335519790649,
      "rewards/forensics_sinkhorn_reward": 0.7155701816082001,
      "rewards/forensics_strict_edge_reward": 0.033141881227493286,
      "rewards/forensics_tiou_reward": 0.535631000995636,
      "step": 852
    },
    {
      "completion_length": 27.515625,
      "epoch": 2.6175115207373274,
      "grad_norm": 10.364153861999512,
      "kl": 0.4775390625,
      "learning_rate": 3.446153846153846e-07,
      "loss": -0.1056,
      "reward": 1.3298750519752502,
      "reward_std": 0.33664631843566895,
      "reward_std_all": 0.9252131283283234,
      "rewards-max/forensics_sinkhorn_reward": 0.9873495697975159,
      "rewards-max/forensics_strict_edge_reward": 0.6009034514427185,
      "rewards-max/forensics_tiou_reward": 0.9945074617862701,
      "rewards/forensics_sinkhorn_reward": 0.6708261370658875,
      "rewards/forensics_strict_edge_reward": 0.09903239458799362,
      "rewards/forensics_tiou_reward": 0.5600164532661438,
      "step": 853
    },
    {
      "completion_length": 25.875,
      "epoch": 2.6205837173579107,
      "grad_norm": 9.445068359375,
      "kl": 0.38671875,
      "learning_rate": 3.4384615384615385e-07,
      "loss": -0.0529,
      "reward": 1.3774574398994446,
      "reward_std": 0.3962995409965515,
      "reward_std_all": 0.924850583076477,
      "rewards-max/forensics_sinkhorn_reward": 0.9876537919044495,
      "rewards-max/forensics_strict_edge_reward": 0.7487103343009949,
      "rewards-max/forensics_tiou_reward": 0.9974254369735718,
      "rewards/forensics_sinkhorn_reward": 0.7428752183914185,
      "rewards/forensics_strict_edge_reward": 0.07522833533585072,
      "rewards/forensics_tiou_reward": 0.5593537986278534,
      "step": 854
    },
    {
      "completion_length": 23.5,
      "epoch": 2.6236559139784945,
      "grad_norm": 11.773903846740723,
      "kl": 0.619140625,
      "learning_rate": 3.4307692307692307e-07,
      "loss": -0.0657,
      "reward": 1.349144160747528,
      "reward_std": 0.2702820897102356,
      "reward_std_all": 0.8871509432792664,
      "rewards-max/forensics_sinkhorn_reward": 0.9871532917022705,
      "rewards-max/forensics_strict_edge_reward": 0.6334160566329956,
      "rewards-max/forensics_tiou_reward": 0.9936282336711884,
      "rewards/forensics_sinkhorn_reward": 0.7390936315059662,
      "rewards/forensics_strict_edge_reward": 0.05339672788977623,
      "rewards/forensics_tiou_reward": 0.5566538870334625,
      "step": 855
    },
    {
      "completion_length": 29.546875,
      "epoch": 2.6267281105990783,
      "grad_norm": 9.4102201461792,
      "kl": 0.4990234375,
      "learning_rate": 3.423076923076923e-07,
      "loss": 0.0219,
      "reward": 1.2349024415016174,
      "reward_std": 0.2632144093513489,
      "reward_std_all": 0.9237223565578461,
      "rewards-max/forensics_sinkhorn_reward": 0.9637549519538879,
      "rewards-max/forensics_strict_edge_reward": 0.32897473871707916,
      "rewards-max/forensics_tiou_reward": 0.9366768598556519,
      "rewards/forensics_sinkhorn_reward": 0.6763178706169128,
      "rewards/forensics_strict_edge_reward": 0.04257750138640404,
      "rewards/forensics_tiou_reward": 0.5160070657730103,
      "step": 856
    },
    {
      "completion_length": 25.4375,
      "epoch": 2.629800307219662,
      "grad_norm": 9.344189643859863,
      "kl": 0.4541015625,
      "learning_rate": 3.415384615384615e-07,
      "loss": -0.0086,
      "reward": 1.403636872768402,
      "reward_std": 0.5411093831062317,
      "reward_std_all": 0.9782665967941284,
      "rewards-max/forensics_sinkhorn_reward": 0.9899527728557587,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.6982491612434387,
      "rewards/forensics_strict_edge_reward": 0.14521678537130356,
      "rewards/forensics_tiou_reward": 0.5601708889007568,
      "step": 857
    },
    {
      "completion_length": 25.765625,
      "epoch": 2.632872503840246,
      "grad_norm": 11.728137016296387,
      "kl": 0.4326171875,
      "learning_rate": 3.4076923076923076e-07,
      "loss": -0.1166,
      "reward": 1.3828771114349365,
      "reward_std": 0.33017192780971527,
      "reward_std_all": 0.8810387253761292,
      "rewards-max/forensics_sinkhorn_reward": 0.9802501201629639,
      "rewards-max/forensics_strict_edge_reward": 0.8465203046798706,
      "rewards-max/forensics_tiou_reward": 0.977067619562149,
      "rewards/forensics_sinkhorn_reward": 0.7330926060676575,
      "rewards/forensics_strict_edge_reward": 0.08071533963084221,
      "rewards/forensics_tiou_reward": 0.5690691769123077,
      "step": 858
    },
    {
      "completion_length": 24.625,
      "epoch": 2.6359447004608296,
      "grad_norm": 30.21880531311035,
      "kl": 0.669921875,
      "learning_rate": 3.4000000000000003e-07,
      "loss": -0.0785,
      "reward": 1.462177038192749,
      "reward_std": 0.4344089478254318,
      "reward_std_all": 0.8837230503559113,
      "rewards-max/forensics_sinkhorn_reward": 0.9872618615627289,
      "rewards-max/forensics_strict_edge_reward": 0.9375866651535034,
      "rewards-max/forensics_tiou_reward": 0.994912177324295,
      "rewards/forensics_sinkhorn_reward": 0.775491327047348,
      "rewards/forensics_strict_edge_reward": 0.09222982078790665,
      "rewards/forensics_tiou_reward": 0.5944560170173645,
      "step": 859
    },
    {
      "completion_length": 28.640625,
      "epoch": 2.6390168970814134,
      "grad_norm": 9.979339599609375,
      "kl": 0.525390625,
      "learning_rate": 3.392307692307692e-07,
      "loss": -0.0288,
      "reward": 1.2133044600486755,
      "reward_std": 0.3555027097463608,
      "reward_std_all": 0.9660856127738953,
      "rewards-max/forensics_sinkhorn_reward": 0.9695935547351837,
      "rewards-max/forensics_strict_edge_reward": 0.46017518639564514,
      "rewards-max/forensics_tiou_reward": 0.9518357515335083,
      "rewards/forensics_sinkhorn_reward": 0.6497088670730591,
      "rewards/forensics_strict_edge_reward": 0.06501816213130951,
      "rewards/forensics_tiou_reward": 0.49857741594314575,
      "step": 860
    },
    {
      "completion_length": 24.8125,
      "epoch": 2.6420890937019967,
      "grad_norm": 11.053375244140625,
      "kl": 0.5576171875,
      "learning_rate": 3.3846153846153845e-07,
      "loss": -0.1474,
      "reward": 1.6873611211776733,
      "reward_std": 0.4752968102693558,
      "reward_std_all": 0.7860518991947174,
      "rewards-max/forensics_sinkhorn_reward": 0.988125205039978,
      "rewards-max/forensics_strict_edge_reward": 0.8516424298286438,
      "rewards-max/forensics_tiou_reward": 0.9969933032989502,
      "rewards/forensics_sinkhorn_reward": 0.8218503892421722,
      "rewards/forensics_strict_edge_reward": 0.16233095526695251,
      "rewards/forensics_tiou_reward": 0.7031797766685486,
      "step": 861
    },
    {
      "completion_length": 25.234375,
      "epoch": 2.6451612903225805,
      "grad_norm": 9.781190872192383,
      "kl": 0.4814453125,
      "learning_rate": 3.3769230769230767e-07,
      "loss": -0.0307,
      "reward": 1.4118426442146301,
      "reward_std": 0.36766211688518524,
      "reward_std_all": 0.9063674211502075,
      "rewards-max/forensics_sinkhorn_reward": 0.9879404604434967,
      "rewards-max/forensics_strict_edge_reward": 0.8465360999107361,
      "rewards-max/forensics_tiou_reward": 0.997276097536087,
      "rewards/forensics_sinkhorn_reward": 0.6904832124710083,
      "rewards/forensics_strict_edge_reward": 0.16254181787371635,
      "rewards/forensics_tiou_reward": 0.5588176250457764,
      "step": 862
    },
    {
      "completion_length": 24.3125,
      "epoch": 2.6482334869431643,
      "grad_norm": 11.702637672424316,
      "kl": 0.48046875,
      "learning_rate": 3.3692307692307693e-07,
      "loss": -0.1577,
      "reward": 1.4935065507888794,
      "reward_std": 0.32187607884407043,
      "reward_std_all": 0.8212565779685974,
      "rewards-max/forensics_sinkhorn_reward": 0.9876951277256012,
      "rewards-max/forensics_strict_edge_reward": 0.779474675655365,
      "rewards-max/forensics_tiou_reward": 0.9968978464603424,
      "rewards/forensics_sinkhorn_reward": 0.775491327047348,
      "rewards/forensics_strict_edge_reward": 0.0710521787405014,
      "rewards/forensics_tiou_reward": 0.6469630002975464,
      "step": 863
    },
    {
      "completion_length": 26.046875,
      "epoch": 2.651305683563748,
      "grad_norm": 23.544309616088867,
      "kl": 0.84765625,
      "learning_rate": 3.361538461538461e-07,
      "loss": 0.0167,
      "reward": 1.6155397295951843,
      "reward_std": 0.40986084938049316,
      "reward_std_all": 0.7902067303657532,
      "rewards-max/forensics_sinkhorn_reward": 0.9856260418891907,
      "rewards-max/forensics_strict_edge_reward": 0.741247296333313,
      "rewards-max/forensics_tiou_reward": 0.9902293980121613,
      "rewards/forensics_sinkhorn_reward": 0.830095648765564,
      "rewards/forensics_strict_edge_reward": 0.10561211034655571,
      "rewards/forensics_tiou_reward": 0.6798319816589355,
      "step": 864
    },
    {
      "completion_length": 23.984375,
      "epoch": 2.654377880184332,
      "grad_norm": 10.922052383422852,
      "kl": 0.3349609375,
      "learning_rate": 3.3538461538461536e-07,
      "loss": -0.0993,
      "reward": 1.456409752368927,
      "reward_std": 0.34509824216365814,
      "reward_std_all": 0.8846772909164429,
      "rewards-max/forensics_sinkhorn_reward": 0.9888124167919159,
      "rewards-max/forensics_strict_edge_reward": 0.8704091012477875,
      "rewards-max/forensics_tiou_reward": 0.9984637200832367,
      "rewards/forensics_sinkhorn_reward": 0.746894896030426,
      "rewards/forensics_strict_edge_reward": 0.11163939535617828,
      "rewards/forensics_tiou_reward": 0.5978753864765167,
      "step": 865
    },
    {
      "completion_length": 25.640625,
      "epoch": 2.6574500768049156,
      "grad_norm": 10.00260066986084,
      "kl": 0.5166015625,
      "learning_rate": 3.346153846153846e-07,
      "loss": -0.0488,
      "reward": 1.5723770260810852,
      "reward_std": 0.38327430188655853,
      "reward_std_all": 0.8341180980205536,
      "rewards-max/forensics_sinkhorn_reward": 0.9895322620868683,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.7763071656227112,
      "rewards/forensics_strict_edge_reward": 0.14712746441364288,
      "rewards/forensics_tiou_reward": 0.6489424109458923,
      "step": 866
    },
    {
      "completion_length": 25.125,
      "epoch": 2.6605222734254994,
      "grad_norm": 10.170068740844727,
      "kl": 0.51171875,
      "learning_rate": 3.3384615384615384e-07,
      "loss": -0.0617,
      "reward": 1.4906098246574402,
      "reward_std": 0.4735349416732788,
      "reward_std_all": 0.8891203701496124,
      "rewards-max/forensics_sinkhorn_reward": 0.9889309108257294,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.7554827332496643,
      "rewards/forensics_strict_edge_reward": 0.15002267807722092,
      "rewards/forensics_tiou_reward": 0.5851044952869415,
      "step": 867
    },
    {
      "completion_length": 26.640625,
      "epoch": 2.6635944700460827,
      "grad_norm": 12.564750671386719,
      "kl": 0.5107421875,
      "learning_rate": 3.330769230769231e-07,
      "loss": -0.0921,
      "reward": 1.5139241218566895,
      "reward_std": 0.388863205909729,
      "reward_std_all": 0.8535061180591583,
      "rewards-max/forensics_sinkhorn_reward": 0.9871638417243958,
      "rewards-max/forensics_strict_edge_reward": 0.836103081703186,
      "rewards-max/forensics_tiou_reward": 0.9975145161151886,
      "rewards/forensics_sinkhorn_reward": 0.7670524716377258,
      "rewards/forensics_strict_edge_reward": 0.11434697359800339,
      "rewards/forensics_tiou_reward": 0.632524698972702,
      "step": 868
    },
    {
      "completion_length": 25.75,
      "epoch": 2.6666666666666665,
      "grad_norm": 10.322072982788086,
      "kl": 0.486328125,
      "learning_rate": 3.3230769230769227e-07,
      "loss": -0.1371,
      "reward": 1.609326183795929,
      "reward_std": 0.2898523062467575,
      "reward_std_all": 0.7731803059577942,
      "rewards-max/forensics_sinkhorn_reward": 0.985411137342453,
      "rewards-max/forensics_strict_edge_reward": 0.8511950969696045,
      "rewards-max/forensics_tiou_reward": 0.9931611120700836,
      "rewards/forensics_sinkhorn_reward": 0.82114177942276,
      "rewards/forensics_strict_edge_reward": 0.11876209825277328,
      "rewards/forensics_tiou_reward": 0.6694222390651703,
      "step": 869
    },
    {
      "completion_length": 24.796875,
      "epoch": 2.6697388632872503,
      "grad_norm": 15.687163352966309,
      "kl": 1.275390625,
      "learning_rate": 3.3153846153846154e-07,
      "loss": -0.0828,
      "reward": 1.6282910108566284,
      "reward_std": 0.36030814051628113,
      "reward_std_all": 0.7878668308258057,
      "rewards-max/forensics_sinkhorn_reward": 0.9871321022510529,
      "rewards-max/forensics_strict_edge_reward": 0.5826494470238686,
      "rewards-max/forensics_tiou_reward": 0.9936307966709137,
      "rewards/forensics_sinkhorn_reward": 0.8293221592903137,
      "rewards/forensics_strict_edge_reward": 0.10499884933233261,
      "rewards/forensics_tiou_reward": 0.6939700245857239,
      "step": 870
    },
    {
      "completion_length": 26.109375,
      "epoch": 2.672811059907834,
      "grad_norm": 11.330734252929688,
      "kl": 0.5771484375,
      "learning_rate": 3.3076923076923075e-07,
      "loss": -0.0449,
      "reward": 1.4133881330490112,
      "reward_std": 0.2372751384973526,
      "reward_std_all": 0.8531117141246796,
      "rewards-max/forensics_sinkhorn_reward": 0.9894649982452393,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.7163580656051636,
      "rewards/forensics_strict_edge_reward": 0.10380018129944801,
      "rewards/forensics_tiou_reward": 0.5932298898696899,
      "step": 871
    },
    {
      "completion_length": 23.140625,
      "epoch": 2.675883256528418,
      "grad_norm": 10.11542797088623,
      "kl": 0.3671875,
      "learning_rate": 3.3e-07,
      "loss": -0.0854,
      "reward": 1.219681829214096,
      "reward_std": 0.23210129141807556,
      "reward_std_all": 0.9312747120857239,
      "rewards-max/forensics_sinkhorn_reward": 0.9865946173667908,
      "rewards-max/forensics_strict_edge_reward": 0.5314356125891209,
      "rewards-max/forensics_tiou_reward": 0.9931303262710571,
      "rewards/forensics_sinkhorn_reward": 0.6611768305301666,
      "rewards/forensics_strict_edge_reward": 0.058359112590551376,
      "rewards/forensics_tiou_reward": 0.500145897269249,
      "step": 872
    },
    {
      "completion_length": 23.953125,
      "epoch": 2.6789554531490016,
      "grad_norm": 10.154182434082031,
      "kl": 0.5859375,
      "learning_rate": 3.292307692307692e-07,
      "loss": -0.0548,
      "reward": 1.5596749186515808,
      "reward_std": 0.32593652606010437,
      "reward_std_all": 0.7993777692317963,
      "rewards-max/forensics_sinkhorn_reward": 0.9891725778579712,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.7902755737304688,
      "rewards/forensics_strict_edge_reward": 0.10810261964797974,
      "rewards/forensics_tiou_reward": 0.6612967550754547,
      "step": 873
    },
    {
      "completion_length": 26.421875,
      "epoch": 2.6820276497695854,
      "grad_norm": 10.468092918395996,
      "kl": 0.5068359375,
      "learning_rate": 3.2846153846153844e-07,
      "loss": -0.0373,
      "reward": 1.3044618964195251,
      "reward_std": 0.17104700207710266,
      "reward_std_all": 0.8660539090633392,
      "rewards-max/forensics_sinkhorn_reward": 0.9546755850315094,
      "rewards-max/forensics_strict_edge_reward": 0.3809697926044464,
      "rewards-max/forensics_tiou_reward": 0.9241771996021271,
      "rewards/forensics_sinkhorn_reward": 0.7053676247596741,
      "rewards/forensics_strict_edge_reward": 0.033945790491998196,
      "rewards/forensics_tiou_reward": 0.5651485025882721,
      "step": 874
    },
    {
      "completion_length": 24.953125,
      "epoch": 2.6850998463901687,
      "grad_norm": 10.797183990478516,
      "kl": 0.634765625,
      "learning_rate": 3.2769230769230766e-07,
      "loss": -0.0092,
      "reward": 1.467654526233673,
      "reward_std": 0.35266895592212677,
      "reward_std_all": 0.8491893708705902,
      "rewards-max/forensics_sinkhorn_reward": 0.9850964248180389,
      "rewards-max/forensics_strict_edge_reward": 0.7837068438529968,
      "rewards-max/forensics_tiou_reward": 0.9885371327400208,
      "rewards/forensics_sinkhorn_reward": 0.7732572555541992,
      "rewards/forensics_strict_edge_reward": 0.08742566779255867,
      "rewards/forensics_tiou_reward": 0.6069716513156891,
      "step": 875
    },
    {
      "completion_length": 27.859375,
      "epoch": 2.688172043010753,
      "grad_norm": 9.743391036987305,
      "kl": 0.5,
      "learning_rate": 3.269230769230769e-07,
      "loss": -0.1355,
      "reward": 1.4730414152145386,
      "reward_std": 0.4158959984779358,
      "reward_std_all": 0.8834925293922424,
      "rewards-max/forensics_sinkhorn_reward": 0.9864173531532288,
      "rewards-max/forensics_strict_edge_reward": 0.7972188889980316,
      "rewards-max/forensics_tiou_reward": 0.9924898743629456,
      "rewards/forensics_sinkhorn_reward": 0.7617729008197784,
      "rewards/forensics_strict_edge_reward": 0.10106416046619415,
      "rewards/forensics_tiou_reward": 0.6102042496204376,
      "step": 876
    },
    {
      "completion_length": 22.890625,
      "epoch": 2.6912442396313363,
      "grad_norm": 9.445043563842773,
      "kl": 0.595703125,
      "learning_rate": 3.2615384615384614e-07,
      "loss": -0.1095,
      "reward": 1.5066462755203247,
      "reward_std": 0.31477753072977066,
      "reward_std_all": 0.8300197720527649,
      "rewards-max/forensics_sinkhorn_reward": 0.9866838455200195,
      "rewards-max/forensics_strict_edge_reward": 0.7286747694015503,
      "rewards-max/forensics_tiou_reward": 0.9951982796192169,
      "rewards/forensics_sinkhorn_reward": 0.7634882926940918,
      "rewards/forensics_strict_edge_reward": 0.1069633923470974,
      "rewards/forensics_tiou_reward": 0.6361945569515228,
      "step": 877
    },
    {
      "completion_length": 21.78125,
      "epoch": 2.69431643625192,
      "grad_norm": 9.005969047546387,
      "kl": 0.4443359375,
      "learning_rate": 3.2538461538461535e-07,
      "loss": -0.0713,
      "reward": 1.3383708000183105,
      "reward_std": 0.34022022783756256,
      "reward_std_all": 0.8961513936519623,
      "rewards-max/forensics_sinkhorn_reward": 0.9844140112400055,
      "rewards-max/forensics_strict_edge_reward": 0.7047859728336334,
      "rewards-max/forensics_tiou_reward": 0.9879687428474426,
      "rewards/forensics_sinkhorn_reward": 0.7430976927280426,
      "rewards/forensics_strict_edge_reward": 0.05668516270816326,
      "rewards/forensics_tiou_reward": 0.5385879278182983,
      "step": 878
    },
    {
      "completion_length": 24.84375,
      "epoch": 2.697388632872504,
      "grad_norm": 15.27272891998291,
      "kl": 0.734375,
      "learning_rate": 3.246153846153846e-07,
      "loss": -0.1537,
      "reward": 1.5082092881202698,
      "reward_std": 0.5087938010692596,
      "reward_std_all": 0.8868306875228882,
      "rewards-max/forensics_sinkhorn_reward": 0.9838220477104187,
      "rewards-max/forensics_strict_edge_reward": 0.7403165996074677,
      "rewards-max/forensics_tiou_reward": 0.9833351671695709,
      "rewards/forensics_sinkhorn_reward": 0.7794323861598969,
      "rewards/forensics_strict_edge_reward": 0.10726867243647575,
      "rewards/forensics_tiou_reward": 0.6215081512928009,
      "step": 879
    },
    {
      "completion_length": 26.5625,
      "epoch": 2.7004608294930876,
      "grad_norm": 10.196581840515137,
      "kl": 0.484375,
      "learning_rate": 3.2384615384615383e-07,
      "loss": -0.1066,
      "reward": 1.5158353447914124,
      "reward_std": 0.3997703939676285,
      "reward_std_all": 0.8439938426017761,
      "rewards-max/forensics_sinkhorn_reward": 0.9834882616996765,
      "rewards-max/forensics_strict_edge_reward": 0.541214644908905,
      "rewards-max/forensics_tiou_reward": 0.985268771648407,
      "rewards/forensics_sinkhorn_reward": 0.7609286308288574,
      "rewards/forensics_strict_edge_reward": 0.12079265341162682,
      "rewards/forensics_tiou_reward": 0.6341139674186707,
      "step": 880
    },
    {
      "completion_length": 23.578125,
      "epoch": 2.7035330261136714,
      "grad_norm": 10.078664779663086,
      "kl": 0.52734375,
      "learning_rate": 3.230769230769231e-07,
      "loss": -0.0624,
      "reward": 1.322804868221283,
      "reward_std": 0.4241078943014145,
      "reward_std_all": 0.9512680470943451,
      "rewards-max/forensics_sinkhorn_reward": 0.9820519089698792,
      "rewards-max/forensics_strict_edge_reward": 0.750005841255188,
      "rewards-max/forensics_tiou_reward": 0.9820190072059631,
      "rewards/forensics_sinkhorn_reward": 0.7142916321754456,
      "rewards/forensics_strict_edge_reward": 0.09620850533246994,
      "rewards/forensics_tiou_reward": 0.5123047232627869,
      "step": 881
    },
    {
      "completion_length": 26.140625,
      "epoch": 2.7066052227342547,
      "grad_norm": 9.499274253845215,
      "kl": 0.599609375,
      "learning_rate": 3.2230769230769226e-07,
      "loss": -0.1054,
      "reward": 1.619667112827301,
      "reward_std": 0.5031828582286835,
      "reward_std_all": 0.8652732074260712,
      "rewards-max/forensics_sinkhorn_reward": 0.9897510409355164,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.7750382721424103,
      "rewards/forensics_strict_edge_reward": 0.20387908816337585,
      "rewards/forensics_tiou_reward": 0.6407497525215149,
      "step": 882
    },
    {
      "completion_length": 26.46875,
      "epoch": 2.709677419354839,
      "grad_norm": 10.218673706054688,
      "kl": 0.5302734375,
      "learning_rate": 3.215384615384615e-07,
      "loss": -0.0458,
      "reward": 1.5811778903007507,
      "reward_std": 0.4573713093996048,
      "reward_std_all": 0.8468271791934967,
      "rewards-max/forensics_sinkhorn_reward": 0.9851666986942291,
      "rewards-max/forensics_strict_edge_reward": 0.8419698476791382,
      "rewards-max/forensics_tiou_reward": 0.9921571910381317,
      "rewards/forensics_sinkhorn_reward": 0.7628810107707977,
      "rewards/forensics_strict_edge_reward": 0.1594974361360073,
      "rewards/forensics_tiou_reward": 0.6587994396686554,
      "step": 883
    },
    {
      "completion_length": 28.640625,
      "epoch": 2.7127496159754223,
      "grad_norm": 8.712431907653809,
      "kl": 0.732421875,
      "learning_rate": 3.2076923076923074e-07,
      "loss": -0.0746,
      "reward": 1.3902426362037659,
      "reward_std": 0.3123873770236969,
      "reward_std_all": 0.8633329272270203,
      "rewards-max/forensics_sinkhorn_reward": 0.9760600924491882,
      "rewards-max/forensics_strict_edge_reward": 0.507478877902031,
      "rewards-max/forensics_tiou_reward": 0.9675026535987854,
      "rewards/forensics_sinkhorn_reward": 0.7548791468143463,
      "rewards/forensics_strict_edge_reward": 0.05514054745435715,
      "rewards/forensics_tiou_reward": 0.5802229940891266,
      "step": 884
    },
    {
      "completion_length": 27.84375,
      "epoch": 2.715821812596006,
      "grad_norm": 10.79322624206543,
      "kl": 0.46875,
      "learning_rate": 3.2e-07,
      "loss": -0.1154,
      "reward": 1.5290994048118591,
      "reward_std": 0.5113762319087982,
      "reward_std_all": 0.9003468453884125,
      "rewards-max/forensics_sinkhorn_reward": 0.9881769120693207,
      "rewards-max/forensics_strict_edge_reward": 0.7959342896938324,
      "rewards-max/forensics_tiou_reward": 0.9972831308841705,
      "rewards/forensics_sinkhorn_reward": 0.7338458001613617,
      "rewards/forensics_strict_edge_reward": 0.15392664074897766,
      "rewards/forensics_tiou_reward": 0.6413269639015198,
      "step": 885
    },
    {
      "completion_length": 24.3125,
      "epoch": 2.71889400921659,
      "grad_norm": 11.080667495727539,
      "kl": 0.5009765625,
      "learning_rate": 3.1923076923076917e-07,
      "loss": -0.074,
      "reward": 1.3109208345413208,
      "reward_std": 0.49810656905174255,
      "reward_std_all": 0.9527512490749359,
      "rewards-max/forensics_sinkhorn_reward": 0.9763612449169159,
      "rewards-max/forensics_strict_edge_reward": 0.48898807168006897,
      "rewards-max/forensics_tiou_reward": 0.965061753988266,
      "rewards/forensics_sinkhorn_reward": 0.7122645378112793,
      "rewards/forensics_strict_edge_reward": 0.06636297702789307,
      "rewards/forensics_tiou_reward": 0.532293289899826,
      "step": 886
    },
    {
      "completion_length": 25.140625,
      "epoch": 2.7219662058371736,
      "grad_norm": 10.189781188964844,
      "kl": 0.48828125,
      "learning_rate": 3.1846153846153843e-07,
      "loss": -0.0151,
      "reward": 1.3137200474739075,
      "reward_std": 0.39926767349243164,
      "reward_std_all": 0.9421178996562958,
      "rewards-max/forensics_sinkhorn_reward": 0.9814166724681854,
      "rewards-max/forensics_strict_edge_reward": 0.639672726392746,
      "rewards-max/forensics_tiou_reward": 0.9797346591949463,
      "rewards/forensics_sinkhorn_reward": 0.703281432390213,
      "rewards/forensics_strict_edge_reward": 0.07538562640547752,
      "rewards/forensics_tiou_reward": 0.5350530445575714,
      "step": 887
    },
    {
      "completion_length": 22.234375,
      "epoch": 2.7250384024577574,
      "grad_norm": 8.944900512695312,
      "kl": 0.4228515625,
      "learning_rate": 3.176923076923077e-07,
      "loss": -0.123,
      "reward": 1.5007270574569702,
      "reward_std": 0.3424268960952759,
      "reward_std_all": 0.8395143747329712,
      "rewards-max/forensics_sinkhorn_reward": 0.9884275496006012,
      "rewards-max/forensics_strict_edge_reward": 0.952418714761734,
      "rewards-max/forensics_tiou_reward": 0.9971011579036713,
      "rewards/forensics_sinkhorn_reward": 0.7672455906867981,
      "rewards/forensics_strict_edge_reward": 0.11555228009819984,
      "rewards/forensics_tiou_reward": 0.6179291009902954,
      "step": 888
    },
    {
      "completion_length": 27.28125,
      "epoch": 2.7281105990783407,
      "grad_norm": 8.623416900634766,
      "kl": 0.4482421875,
      "learning_rate": 3.169230769230769e-07,
      "loss": -0.0538,
      "reward": 1.4226729273796082,
      "reward_std": 0.5092706382274628,
      "reward_std_all": 0.9341581463813782,
      "rewards-max/forensics_sinkhorn_reward": 0.9878745675086975,
      "rewards-max/forensics_strict_edge_reward": 0.9291328191757202,
      "rewards-max/forensics_tiou_reward": 0.9966971278190613,
      "rewards/forensics_sinkhorn_reward": 0.7220728993415833,
      "rewards/forensics_strict_edge_reward": 0.10603393241763115,
      "rewards/forensics_tiou_reward": 0.5945660769939423,
      "step": 889
    },
    {
      "completion_length": 25.71875,
      "epoch": 2.731182795698925,
      "grad_norm": 9.215575218200684,
      "kl": 0.48828125,
      "learning_rate": 3.161538461538462e-07,
      "loss": -0.0767,
      "reward": 1.4210103154182434,
      "reward_std": 0.5093750059604645,
      "reward_std_all": 0.9577235877513885,
      "rewards-max/forensics_sinkhorn_reward": 0.9882495701313019,
      "rewards-max/forensics_strict_edge_reward": 0.9918040335178375,
      "rewards-max/forensics_tiou_reward": 0.9999438226222992,
      "rewards/forensics_sinkhorn_reward": 0.7186065018177032,
      "rewards/forensics_strict_edge_reward": 0.14478044211864471,
      "rewards/forensics_tiou_reward": 0.5576233118772507,
      "step": 890
    },
    {
      "completion_length": 26.75,
      "epoch": 2.7342549923195083,
      "grad_norm": 10.990300178527832,
      "kl": 0.69140625,
      "learning_rate": 3.1538461538461534e-07,
      "loss": -0.0206,
      "reward": 1.598173201084137,
      "reward_std": 0.4093056321144104,
      "reward_std_all": 0.832686185836792,
      "rewards-max/forensics_sinkhorn_reward": 0.9895638525485992,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.7656292021274567,
      "rewards/forensics_strict_edge_reward": 0.1838700920343399,
      "rewards/forensics_tiou_reward": 0.6486738473176956,
      "step": 891
    },
    {
      "completion_length": 25.359375,
      "epoch": 2.737327188940092,
      "grad_norm": 8.891059875488281,
      "kl": 0.4765625,
      "learning_rate": 3.146153846153846e-07,
      "loss": -0.0764,
      "reward": 1.41237074136734,
      "reward_std": 0.39999300241470337,
      "reward_std_all": 0.889648973941803,
      "rewards-max/forensics_sinkhorn_reward": 0.9818446636199951,
      "rewards-max/forensics_strict_edge_reward": 0.7133049219846725,
      "rewards-max/forensics_tiou_reward": 0.9806376397609711,
      "rewards/forensics_sinkhorn_reward": 0.7449638545513153,
      "rewards/forensics_strict_edge_reward": 0.08493507094681263,
      "rewards/forensics_tiou_reward": 0.5824718177318573,
      "step": 892
    },
    {
      "completion_length": 23.34375,
      "epoch": 2.740399385560676,
      "grad_norm": 10.340804100036621,
      "kl": 0.4384765625,
      "learning_rate": 3.138461538461538e-07,
      "loss": -0.0656,
      "reward": 1.4274319410324097,
      "reward_std": 0.40296003222465515,
      "reward_std_all": 0.8921562135219574,
      "rewards-max/forensics_sinkhorn_reward": 0.9850244224071503,
      "rewards-max/forensics_strict_edge_reward": 0.882964164018631,
      "rewards-max/forensics_tiou_reward": 0.9902187585830688,
      "rewards/forensics_sinkhorn_reward": 0.7479746043682098,
      "rewards/forensics_strict_edge_reward": 0.09093660674989223,
      "rewards/forensics_tiou_reward": 0.5885206460952759,
      "step": 893
    },
    {
      "completion_length": 27.234375,
      "epoch": 2.7434715821812596,
      "grad_norm": 11.015246391296387,
      "kl": 0.6640625,
      "learning_rate": 3.130769230769231e-07,
      "loss": -0.0335,
      "reward": 1.3455243706703186,
      "reward_std": 0.3363625109195709,
      "reward_std_all": 0.9028551578521729,
      "rewards-max/forensics_sinkhorn_reward": 0.9825813472270966,
      "rewards-max/forensics_strict_edge_reward": 0.7888355851173401,
      "rewards-max/forensics_tiou_reward": 0.9868558347225189,
      "rewards/forensics_sinkhorn_reward": 0.6983670592308044,
      "rewards/forensics_strict_edge_reward": 0.10216650366783142,
      "rewards/forensics_tiou_reward": 0.5449909120798111,
      "step": 894
    },
    {
      "completion_length": 27.328125,
      "epoch": 2.7465437788018434,
      "grad_norm": 10.906242370605469,
      "kl": 0.607421875,
      "learning_rate": 3.1230769230769225e-07,
      "loss": 0.021,
      "reward": 1.5383297204971313,
      "reward_std": 0.2716742008924484,
      "reward_std_all": 0.7969347834587097,
      "rewards-max/forensics_sinkhorn_reward": 0.9872576296329498,
      "rewards-max/forensics_strict_edge_reward": 0.8360384404659271,
      "rewards-max/forensics_tiou_reward": 0.993402749300003,
      "rewards/forensics_sinkhorn_reward": 0.7739527523517609,
      "rewards/forensics_strict_edge_reward": 0.14019320160150528,
      "rewards/forensics_tiou_reward": 0.6241838037967682,
      "step": 895
    },
    {
      "completion_length": 26.5625,
      "epoch": 2.749615975422427,
      "grad_norm": 9.27000904083252,
      "kl": 0.4921875,
      "learning_rate": 3.115384615384615e-07,
      "loss": -0.1534,
      "reward": 1.7430734634399414,
      "reward_std": 0.4651806056499481,
      "reward_std_all": 0.7694548964500427,
      "rewards-max/forensics_sinkhorn_reward": 0.9873909950256348,
      "rewards-max/forensics_strict_edge_reward": 0.9291328191757202,
      "rewards-max/forensics_tiou_reward": 0.9964306950569153,
      "rewards/forensics_sinkhorn_reward": 0.8289230465888977,
      "rewards/forensics_strict_edge_reward": 0.17327802628278732,
      "rewards/forensics_tiou_reward": 0.740872323513031,
      "step": 896
    },
    {
      "completion_length": 24.796875,
      "epoch": 2.752688172043011,
      "grad_norm": 17.027740478515625,
      "kl": 0.634765625,
      "learning_rate": 3.1076923076923073e-07,
      "loss": -0.0733,
      "reward": 1.6367450952529907,
      "reward_std": 0.43601737916469574,
      "reward_std_all": 0.8089216649532318,
      "rewards-max/forensics_sinkhorn_reward": 0.9892562031745911,
      "rewards-max/forensics_strict_edge_reward": 0.9918040335178375,
      "rewards-max/forensics_tiou_reward": 0.9996976256370544,
      "rewards/forensics_sinkhorn_reward": 0.8125563561916351,
      "rewards/forensics_strict_edge_reward": 0.1365075223147869,
      "rewards/forensics_tiou_reward": 0.68768110871315,
      "step": 897
    },
    {
      "completion_length": 25.21875,
      "epoch": 2.7557603686635943,
      "grad_norm": 11.065873146057129,
      "kl": 0.408203125,
      "learning_rate": 3.1e-07,
      "loss": -0.0361,
      "reward": 1.4686992168426514,
      "reward_std": 0.2707982808351517,
      "reward_std_all": 0.8278341889381409,
      "rewards-max/forensics_sinkhorn_reward": 0.9887282252311707,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.7386265397071838,
      "rewards/forensics_strict_edge_reward": 0.12684106081724167,
      "rewards/forensics_tiou_reward": 0.6032315492630005,
      "step": 898
    },
    {
      "completion_length": 25.328125,
      "epoch": 2.758832565284178,
      "grad_norm": 11.054676055908203,
      "kl": 0.4169921875,
      "learning_rate": 3.0923076923076926e-07,
      "loss": -0.0689,
      "reward": 1.380910575389862,
      "reward_std": 0.39549732208251953,
      "reward_std_all": 0.91194087266922,
      "rewards-max/forensics_sinkhorn_reward": 0.9797100126743317,
      "rewards-max/forensics_strict_edge_reward": 0.6720768958330154,
      "rewards-max/forensics_tiou_reward": 0.9696988165378571,
      "rewards/forensics_sinkhorn_reward": 0.7300701439380646,
      "rewards/forensics_strict_edge_reward": 0.06768350675702095,
      "rewards/forensics_tiou_reward": 0.5831568539142609,
      "step": 899
    },
    {
      "completion_length": 26.609375,
      "epoch": 2.761904761904762,
      "grad_norm": 9.328287124633789,
      "kl": 0.49609375,
      "learning_rate": 3.084615384615384e-07,
      "loss": -0.1687,
      "reward": 1.5094709992408752,
      "reward_std": 0.3405975103378296,
      "reward_std_all": 0.8479196429252625,
      "rewards-max/forensics_sinkhorn_reward": 0.9770506024360657,
      "rewards-max/forensics_strict_edge_reward": 0.7838338315486908,
      "rewards-max/forensics_tiou_reward": 0.9706015288829803,
      "rewards/forensics_sinkhorn_reward": 0.7579506039619446,
      "rewards/forensics_strict_edge_reward": 0.11779620498418808,
      "rewards/forensics_tiou_reward": 0.6337242126464844,
      "step": 900
    },
    {
      "completion_length": 27.359375,
      "epoch": 2.7649769585253456,
      "grad_norm": 9.120821952819824,
      "kl": 0.5068359375,
      "learning_rate": 3.076923076923077e-07,
      "loss": -0.1533,
      "reward": 1.3776360750198364,
      "reward_std": 0.43978139758110046,
      "reward_std_all": 0.9192144572734833,
      "rewards-max/forensics_sinkhorn_reward": 0.980412095785141,
      "rewards-max/forensics_strict_edge_reward": 0.5500316619873047,
      "rewards-max/forensics_tiou_reward": 0.9751229882240295,
      "rewards/forensics_sinkhorn_reward": 0.7036245465278625,
      "rewards/forensics_strict_edge_reward": 0.08781205117702484,
      "rewards/forensics_tiou_reward": 0.5861994624137878,
      "step": 901
    },
    {
      "completion_length": 27.015625,
      "epoch": 2.7680491551459294,
      "grad_norm": 9.387214660644531,
      "kl": 0.484375,
      "learning_rate": 3.069230769230769e-07,
      "loss": -0.0663,
      "reward": 1.4461705684661865,
      "reward_std": 0.29160215705633163,
      "reward_std_all": 0.8389262557029724,
      "rewards-max/forensics_sinkhorn_reward": 0.9806514978408813,
      "rewards-max/forensics_strict_edge_reward": 0.6760384738445282,
      "rewards-max/forensics_tiou_reward": 0.9767725765705109,
      "rewards/forensics_sinkhorn_reward": 0.7590240240097046,
      "rewards/forensics_strict_edge_reward": 0.07040487229824066,
      "rewards/forensics_tiou_reward": 0.6167417764663696,
      "step": 902
    },
    {
      "completion_length": 29.84375,
      "epoch": 2.771121351766513,
      "grad_norm": 11.531875610351562,
      "kl": 0.857421875,
      "learning_rate": 3.0615384615384617e-07,
      "loss": -0.0288,
      "reward": 1.455496072769165,
      "reward_std": 0.40370576083660126,
      "reward_std_all": 0.8828348815441132,
      "rewards-max/forensics_sinkhorn_reward": 0.9886321425437927,
      "rewards-max/forensics_strict_edge_reward": 0.7606693804264069,
      "rewards-max/forensics_tiou_reward": 0.9971461594104767,
      "rewards/forensics_sinkhorn_reward": 0.7251094579696655,
      "rewards/forensics_strict_edge_reward": 0.12009228765964508,
      "rewards/forensics_tiou_reward": 0.6102942526340485,
      "step": 903
    },
    {
      "completion_length": 29.34375,
      "epoch": 2.774193548387097,
      "grad_norm": 9.821844100952148,
      "kl": 0.583984375,
      "learning_rate": 3.0538461538461533e-07,
      "loss": -0.0639,
      "reward": 1.5777814388275146,
      "reward_std": 0.29873335361480713,
      "reward_std_all": 0.7784412801265717,
      "rewards-max/forensics_sinkhorn_reward": 0.9774268269538879,
      "rewards-max/forensics_strict_edge_reward": 0.7172991037368774,
      "rewards-max/forensics_tiou_reward": 0.9709555804729462,
      "rewards/forensics_sinkhorn_reward": 0.7899241745471954,
      "rewards/forensics_strict_edge_reward": 0.1249200813472271,
      "rewards/forensics_tiou_reward": 0.6629371047019958,
      "step": 904
    },
    {
      "completion_length": 25.640625,
      "epoch": 2.7772657450076803,
      "grad_norm": 11.376750946044922,
      "kl": 0.537109375,
      "learning_rate": 3.046153846153846e-07,
      "loss": -0.0877,
      "reward": 1.3942716717720032,
      "reward_std": 0.36771373450756073,
      "reward_std_all": 0.9023671448230743,
      "rewards-max/forensics_sinkhorn_reward": 0.9883468747138977,
      "rewards-max/forensics_strict_edge_reward": 0.8301283419132233,
      "rewards-max/forensics_tiou_reward": 0.9979783892631531,
      "rewards/forensics_sinkhorn_reward": 0.7071658670902252,
      "rewards/forensics_strict_edge_reward": 0.09947747550904751,
      "rewards/forensics_tiou_reward": 0.5876283049583435,
      "step": 905
    },
    {
      "completion_length": 24.171875,
      "epoch": 2.780337941628264,
      "grad_norm": 10.448132514953613,
      "kl": 0.4912109375,
      "learning_rate": 3.038461538461538e-07,
      "loss": -0.0927,
      "reward": 1.5292960405349731,
      "reward_std": 0.28092263638973236,
      "reward_std_all": 0.7965945601463318,
      "rewards-max/forensics_sinkhorn_reward": 0.9790412783622742,
      "rewards-max/forensics_strict_edge_reward": 0.5472486205399036,
      "rewards-max/forensics_tiou_reward": 0.9729468822479248,
      "rewards/forensics_sinkhorn_reward": 0.8087638914585114,
      "rewards/forensics_strict_edge_reward": 0.0697702169418335,
      "rewards/forensics_tiou_reward": 0.6507619321346283,
      "step": 906
    },
    {
      "completion_length": 23.96875,
      "epoch": 2.783410138248848,
      "grad_norm": 9.063243865966797,
      "kl": 0.498046875,
      "learning_rate": 3.030769230769231e-07,
      "loss": -0.0004,
      "reward": 1.3726303577423096,
      "reward_std": 0.3755231499671936,
      "reward_std_all": 0.908831000328064,
      "rewards-max/forensics_sinkhorn_reward": 0.9884522557258606,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.6858814358711243,
      "rewards/forensics_strict_edge_reward": 0.13837160915136337,
      "rewards/forensics_tiou_reward": 0.5483772456645966,
      "step": 907
    },
    {
      "completion_length": 26.046875,
      "epoch": 2.7864823348694316,
      "grad_norm": 10.992926597595215,
      "kl": 0.5244140625,
      "learning_rate": 3.0230769230769235e-07,
      "loss": -0.1505,
      "reward": 1.4584801197052002,
      "reward_std": 0.4512508511543274,
      "reward_std_all": 0.8970769345760345,
      "rewards-max/forensics_sinkhorn_reward": 0.9862173497676849,
      "rewards-max/forensics_strict_edge_reward": 0.8158992826938629,
      "rewards-max/forensics_tiou_reward": 0.991729199886322,
      "rewards/forensics_sinkhorn_reward": 0.7538000643253326,
      "rewards/forensics_strict_edge_reward": 0.11623623967170715,
      "rewards/forensics_tiou_reward": 0.588443785905838,
      "step": 908
    },
    {
      "completion_length": 25.140625,
      "epoch": 2.7895545314900154,
      "grad_norm": 9.56141471862793,
      "kl": 0.4560546875,
      "learning_rate": 3.015384615384615e-07,
      "loss": -0.0779,
      "reward": 1.740675926208496,
      "reward_std": 0.37426483631134033,
      "reward_std_all": 0.7536633312702179,
      "rewards-max/forensics_sinkhorn_reward": 0.989231675863266,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.8296187818050385,
      "rewards/forensics_strict_edge_reward": 0.2171153798699379,
      "rewards/forensics_tiou_reward": 0.693941742181778,
      "step": 909
    },
    {
      "completion_length": 26.234375,
      "epoch": 2.792626728110599,
      "grad_norm": 11.941071510314941,
      "kl": 0.740234375,
      "learning_rate": 3.0076923076923077e-07,
      "loss": -0.0592,
      "reward": 1.3751615285873413,
      "reward_std": 0.26772846281528473,
      "reward_std_all": 0.8570446074008942,
      "rewards-max/forensics_sinkhorn_reward": 0.9839415550231934,
      "rewards-max/forensics_strict_edge_reward": 0.5690955817699432,
      "rewards-max/forensics_tiou_reward": 0.9855768382549286,
      "rewards/forensics_sinkhorn_reward": 0.7296733558177948,
      "rewards/forensics_strict_edge_reward": 0.06696753762662411,
      "rewards/forensics_tiou_reward": 0.578520730137825,
      "step": 910
    },
    {
      "completion_length": 25.28125,
      "epoch": 2.795698924731183,
      "grad_norm": 11.01291275024414,
      "kl": 0.458984375,
      "learning_rate": 3e-07,
      "loss": -0.1025,
      "reward": 1.5700217485427856,
      "reward_std": 0.42770226299762726,
      "reward_std_all": 0.8321627974510193,
      "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": 0.7672412991523743,
      "rewards/forensics_strict_edge_reward": 0.19934476912021637,
      "rewards/forensics_tiou_reward": 0.6034358143806458,
      "step": 911
    },
    {
      "completion_length": 26.25,
      "epoch": 2.7987711213517663,
      "grad_norm": 178.45355224609375,
      "kl": 7.357421875,
      "learning_rate": 2.9923076923076925e-07,
      "loss": 0.2658,
      "reward": 1.2989516258239746,
      "reward_std": 0.31904904544353485,
      "reward_std_all": 0.9137699604034424,
      "rewards-max/forensics_sinkhorn_reward": 0.983367532491684,
      "rewards-max/forensics_strict_edge_reward": 0.7318825423717499,
      "rewards-max/forensics_tiou_reward": 0.9847415089607239,
      "rewards/forensics_sinkhorn_reward": 0.7032566964626312,
      "rewards/forensics_strict_edge_reward": 0.06522188149392605,
      "rewards/forensics_tiou_reward": 0.530472993850708,
      "step": 912
    },
    {
      "completion_length": 27.234375,
      "epoch": 2.80184331797235,
      "grad_norm": 10.45456314086914,
      "kl": 0.548828125,
      "learning_rate": 2.984615384615384e-07,
      "loss": -0.0136,
      "reward": 1.500346302986145,
      "reward_std": 0.42014312744140625,
      "reward_std_all": 0.8546575009822845,
      "rewards-max/forensics_sinkhorn_reward": 0.9877309203147888,
      "rewards-max/forensics_strict_edge_reward": 0.8994694948196411,
      "rewards-max/forensics_tiou_reward": 0.9986108541488647,
      "rewards/forensics_sinkhorn_reward": 0.7600425183773041,
      "rewards/forensics_strict_edge_reward": 0.1365428790450096,
      "rewards/forensics_tiou_reward": 0.6037608981132507,
      "step": 913
    },
    {
      "completion_length": 26.046875,
      "epoch": 2.804915514592934,
      "grad_norm": 9.314417839050293,
      "kl": 0.4912109375,
      "learning_rate": 2.976923076923077e-07,
      "loss": -0.1031,
      "reward": 1.401708960533142,
      "reward_std": 0.18471477925777435,
      "reward_std_all": 0.8176750540733337,
      "rewards-max/forensics_sinkhorn_reward": 0.9654464423656464,
      "rewards-max/forensics_strict_edge_reward": 0.46365442872047424,
      "rewards-max/forensics_tiou_reward": 0.9362847805023193,
      "rewards/forensics_sinkhorn_reward": 0.741969645023346,
      "rewards/forensics_strict_edge_reward": 0.06527669355273247,
      "rewards/forensics_tiou_reward": 0.5944626331329346,
      "step": 914
    },
    {
      "completion_length": 25.75,
      "epoch": 2.8079877112135176,
      "grad_norm": 11.411078453063965,
      "kl": 0.4501953125,
      "learning_rate": 2.969230769230769e-07,
      "loss": -0.0929,
      "reward": 1.3541207313537598,
      "reward_std": 0.3083234205842018,
      "reward_std_all": 0.9093663096427917,
      "rewards-max/forensics_sinkhorn_reward": 0.9822106659412384,
      "rewards-max/forensics_strict_edge_reward": 0.7473120391368866,
      "rewards-max/forensics_tiou_reward": 0.9810527861118317,
      "rewards/forensics_sinkhorn_reward": 0.7028997540473938,
      "rewards/forensics_strict_edge_reward": 0.11591938510537148,
      "rewards/forensics_tiou_reward": 0.5353016406297684,
      "step": 915
    },
    {
      "completion_length": 24.171875,
      "epoch": 2.8110599078341014,
      "grad_norm": 10.233821868896484,
      "kl": 0.578125,
      "learning_rate": 2.9615384615384616e-07,
      "loss": -0.1135,
      "reward": 1.3562365770339966,
      "reward_std": 0.3030717894434929,
      "reward_std_all": 0.9151913523674011,
      "rewards-max/forensics_sinkhorn_reward": 0.9882254004478455,
      "rewards-max/forensics_strict_edge_reward": 0.882964164018631,
      "rewards-max/forensics_tiou_reward": 0.9971588850021362,
      "rewards/forensics_sinkhorn_reward": 0.720283567905426,
      "rewards/forensics_strict_edge_reward": 0.09670188650488853,
      "rewards/forensics_tiou_reward": 0.5392511487007141,
      "step": 916
    },
    {
      "completion_length": 28.15625,
      "epoch": 2.814132104454685,
      "grad_norm": 12.757143020629883,
      "kl": 0.65625,
      "learning_rate": 2.953846153846154e-07,
      "loss": -0.0369,
      "reward": 1.6514431238174438,
      "reward_std": 0.5364300012588501,
      "reward_std_all": 0.853952556848526,
      "rewards-max/forensics_sinkhorn_reward": 0.9892260134220123,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.7834576964378357,
      "rewards/forensics_strict_edge_reward": 0.20634997636079788,
      "rewards/forensics_tiou_reward": 0.6616355180740356,
      "step": 917
    },
    {
      "completion_length": 24.21875,
      "epoch": 2.817204301075269,
      "grad_norm": 12.029245376586914,
      "kl": 0.392578125,
      "learning_rate": 2.946153846153846e-07,
      "loss": -0.08,
      "reward": 1.4989114999771118,
      "reward_std": 0.2793468236923218,
      "reward_std_all": 0.817194014787674,
      "rewards-max/forensics_sinkhorn_reward": 0.9808837175369263,
      "rewards-max/forensics_strict_edge_reward": 0.6839397251605988,
      "rewards-max/forensics_tiou_reward": 0.9774419665336609,
      "rewards/forensics_sinkhorn_reward": 0.7670428454875946,
      "rewards/forensics_strict_edge_reward": 0.07799728214740753,
      "rewards/forensics_tiou_reward": 0.6538713276386261,
      "step": 918
    },
    {
      "completion_length": 24.3125,
      "epoch": 2.8202764976958523,
      "grad_norm": 9.618914604187012,
      "kl": 0.5400390625,
      "learning_rate": 2.9384615384615385e-07,
      "loss": -0.0436,
      "reward": 1.4679136872291565,
      "reward_std": 0.3411343991756439,
      "reward_std_all": 0.8536551594734192,
      "rewards-max/forensics_sinkhorn_reward": 0.9799305200576782,
      "rewards-max/forensics_strict_edge_reward": 0.7913247346878052,
      "rewards-max/forensics_tiou_reward": 0.9762749969959259,
      "rewards/forensics_sinkhorn_reward": 0.7871915698051453,
      "rewards/forensics_strict_edge_reward": 0.08121154084801674,
      "rewards/forensics_tiou_reward": 0.5995104610919952,
      "step": 919
    },
    {
      "completion_length": 26.875,
      "epoch": 2.823348694316436,
      "grad_norm": 19.198461532592773,
      "kl": 1.2666015625,
      "learning_rate": 2.9307692307692307e-07,
      "loss": -0.0201,
      "reward": 1.53874933719635,
      "reward_std": 0.5739365071058273,
      "reward_std_all": 0.9347832500934601,
      "rewards-max/forensics_sinkhorn_reward": 0.9898907840251923,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.7340022921562195,
      "rewards/forensics_strict_edge_reward": 0.1676332950592041,
      "rewards/forensics_tiou_reward": 0.6371138095855713,
      "step": 920
    },
    {
      "completion_length": 31.3125,
      "epoch": 2.82642089093702,
      "grad_norm": 10.1459379196167,
      "kl": 0.599609375,
      "learning_rate": 2.9230769230769234e-07,
      "loss": 0.0223,
      "reward": 1.6745869517326355,
      "reward_std": 0.45681294798851013,
      "reward_std_all": 0.8012162446975708,
      "rewards-max/forensics_sinkhorn_reward": 0.9855217933654785,
      "rewards-max/forensics_strict_edge_reward": 0.7877196669578552,
      "rewards-max/forensics_tiou_reward": 0.9916348159313202,
      "rewards/forensics_sinkhorn_reward": 0.8122339844703674,
      "rewards/forensics_strict_edge_reward": 0.12322575598955154,
      "rewards/forensics_tiou_reward": 0.7391272187232971,
      "step": 921
    },
    {
      "completion_length": 24.78125,
      "epoch": 2.8294930875576036,
      "grad_norm": 11.001242637634277,
      "kl": 0.458984375,
      "learning_rate": 2.915384615384615e-07,
      "loss": -0.1697,
      "reward": 1.5464235544204712,
      "reward_std": 0.5256364345550537,
      "reward_std_all": 0.8870147168636322,
      "rewards-max/forensics_sinkhorn_reward": 0.9883570969104767,
      "rewards-max/forensics_strict_edge_reward": 0.9836080372333527,
      "rewards-max/forensics_tiou_reward": 0.9994076192378998,
      "rewards/forensics_sinkhorn_reward": 0.7485579252243042,
      "rewards/forensics_strict_edge_reward": 0.1695939376950264,
      "rewards/forensics_tiou_reward": 0.628271609544754,
      "step": 922
    },
    {
      "completion_length": 28.234375,
      "epoch": 2.8325652841781874,
      "grad_norm": 10.199004173278809,
      "kl": 0.375,
      "learning_rate": 2.9076923076923076e-07,
      "loss": -0.0546,
      "reward": 1.552388608455658,
      "reward_std": 0.2657739818096161,
      "reward_std_all": 0.782264232635498,
      "rewards-max/forensics_sinkhorn_reward": 0.9798589646816254,
      "rewards-max/forensics_strict_edge_reward": 0.6490390747785568,
      "rewards-max/forensics_tiou_reward": 0.9778288304805756,
      "rewards/forensics_sinkhorn_reward": 0.7679876685142517,
      "rewards/forensics_strict_edge_reward": 0.08265438675880432,
      "rewards/forensics_tiou_reward": 0.7017465829849243,
      "step": 923
    },
    {
      "completion_length": 28.8125,
      "epoch": 2.835637480798771,
      "grad_norm": 10.705727577209473,
      "kl": 0.4033203125,
      "learning_rate": 2.9e-07,
      "loss": -0.0045,
      "reward": 1.3946951627731323,
      "reward_std": 0.27504579722881317,
      "reward_std_all": 0.8598589301109314,
      "rewards-max/forensics_sinkhorn_reward": 0.9834776520729065,
      "rewards-max/forensics_strict_edge_reward": 0.756708562374115,
      "rewards-max/forensics_tiou_reward": 0.9856830537319183,
      "rewards/forensics_sinkhorn_reward": 0.7245962023735046,
      "rewards/forensics_strict_edge_reward": 0.09519577771425247,
      "rewards/forensics_tiou_reward": 0.5749032199382782,
      "step": 924
    },
    {
      "completion_length": 23.28125,
      "epoch": 2.838709677419355,
      "grad_norm": 12.612796783447266,
      "kl": 0.8720703125,
      "learning_rate": 2.8923076923076924e-07,
      "loss": -0.0336,
      "reward": 1.4334052205085754,
      "reward_std": 0.4487314075231552,
      "reward_std_all": 0.9267371594905853,
      "rewards-max/forensics_sinkhorn_reward": 0.9814547598361969,
      "rewards-max/forensics_strict_edge_reward": 0.7523508965969086,
      "rewards-max/forensics_tiou_reward": 0.9784815311431885,
      "rewards/forensics_sinkhorn_reward": 0.7498884201049805,
      "rewards/forensics_strict_edge_reward": 0.12078250199556351,
      "rewards/forensics_tiou_reward": 0.5627342164516449,
      "step": 925
    },
    {
      "completion_length": 26.453125,
      "epoch": 2.8417818740399383,
      "grad_norm": 9.682393074035645,
      "kl": 0.728515625,
      "learning_rate": 2.884615384615384e-07,
      "loss": -0.0366,
      "reward": 1.450066089630127,
      "reward_std": 0.3969607949256897,
      "reward_std_all": 0.8903115093708038,
      "rewards-max/forensics_sinkhorn_reward": 0.9874024987220764,
      "rewards-max/forensics_strict_edge_reward": 0.8117329776287079,
      "rewards-max/forensics_tiou_reward": 0.9945781826972961,
      "rewards/forensics_sinkhorn_reward": 0.7567416131496429,
      "rewards/forensics_strict_edge_reward": 0.11809351108968258,
      "rewards/forensics_tiou_reward": 0.5752309858798981,
      "step": 926
    },
    {
      "completion_length": 27.203125,
      "epoch": 2.8448540706605225,
      "grad_norm": 10.544719696044922,
      "kl": 0.5048828125,
      "learning_rate": 2.8769230769230767e-07,
      "loss": -0.0888,
      "reward": 1.516343891620636,
      "reward_std": 0.3340274691581726,
      "reward_std_all": 0.8136004209518433,
      "rewards-max/forensics_sinkhorn_reward": 0.9869471490383148,
      "rewards-max/forensics_strict_edge_reward": 0.6702584624290466,
      "rewards-max/forensics_tiou_reward": 0.9949703812599182,
      "rewards/forensics_sinkhorn_reward": 0.7492386400699615,
      "rewards/forensics_strict_edge_reward": 0.11551408097147942,
      "rewards/forensics_tiou_reward": 0.6515912413597107,
      "step": 927
    },
    {
      "completion_length": 23.265625,
      "epoch": 2.847926267281106,
      "grad_norm": 8.968017578125,
      "kl": 0.40234375,
      "learning_rate": 2.8692307692307694e-07,
      "loss": -0.1485,
      "reward": 1.5269718170166016,
      "reward_std": 0.2741761729121208,
      "reward_std_all": 0.7935234308242798,
      "rewards-max/forensics_sinkhorn_reward": 0.9730197489261627,
      "rewards-max/forensics_strict_edge_reward": 0.5548860281705856,
      "rewards-max/forensics_tiou_reward": 0.9603326618671417,
      "rewards/forensics_sinkhorn_reward": 0.8091744482517242,
      "rewards/forensics_strict_edge_reward": 0.0745898000895977,
      "rewards/forensics_tiou_reward": 0.6432076096534729,
      "step": 928
    },
    {
      "completion_length": 25.25,
      "epoch": 2.8509984639016897,
      "grad_norm": 10.959110260009766,
      "kl": 0.5126953125,
      "learning_rate": 2.8615384615384615e-07,
      "loss": -0.0364,
      "reward": 1.4925402998924255,
      "reward_std": 0.2736170291900635,
      "reward_std_all": 0.8005995750427246,
      "rewards-max/forensics_sinkhorn_reward": 0.9626725912094116,
      "rewards-max/forensics_strict_edge_reward": 0.5900419056415558,
      "rewards-max/forensics_tiou_reward": 0.939072996377945,
      "rewards/forensics_sinkhorn_reward": 0.7846216857433319,
      "rewards/forensics_strict_edge_reward": 0.06905439496040344,
      "rewards/forensics_tiou_reward": 0.6388642191886902,
      "step": 929
    },
    {
      "completion_length": 24.6875,
      "epoch": 2.8540706605222734,
      "grad_norm": 10.152902603149414,
      "kl": 0.4501953125,
      "learning_rate": 2.8538461538461536e-07,
      "loss": -0.0717,
      "reward": 1.5339409708976746,
      "reward_std": 0.3255448266863823,
      "reward_std_all": 0.8321562111377716,
      "rewards-max/forensics_sinkhorn_reward": 0.9888482987880707,
      "rewards-max/forensics_strict_edge_reward": 0.9513615369796753,
      "rewards-max/forensics_tiou_reward": 0.9995675086975098,
      "rewards/forensics_sinkhorn_reward": 0.7868682444095612,
      "rewards/forensics_strict_edge_reward": 0.12724273651838303,
      "rewards/forensics_tiou_reward": 0.6198299825191498,
      "step": 930
    },
    {
      "completion_length": 24.03125,
      "epoch": 2.857142857142857,
      "grad_norm": 10.559395790100098,
      "kl": 0.40625,
      "learning_rate": 2.846153846153846e-07,
      "loss": -0.0899,
      "reward": 1.518639326095581,
      "reward_std": 0.4818460941314697,
      "reward_std_all": 0.8768226504325867,
      "rewards-max/forensics_sinkhorn_reward": 0.9886399209499359,
      "rewards-max/forensics_strict_edge_reward": 0.9232408702373505,
      "rewards-max/forensics_tiou_reward": 0.999067485332489,
      "rewards/forensics_sinkhorn_reward": 0.7881501317024231,
      "rewards/forensics_strict_edge_reward": 0.13210366666316986,
      "rewards/forensics_tiou_reward": 0.5983855128288269,
      "step": 931
    },
    {
      "completion_length": 28.4375,
      "epoch": 2.860215053763441,
      "grad_norm": 9.577000617980957,
      "kl": 0.44140625,
      "learning_rate": 2.8384615384615384e-07,
      "loss": -0.1116,
      "reward": 1.3152852654457092,
      "reward_std": 0.4113093614578247,
      "reward_std_all": 0.9402070939540863,
      "rewards-max/forensics_sinkhorn_reward": 0.9818334579467773,
      "rewards-max/forensics_strict_edge_reward": 0.7837068438529968,
      "rewards-max/forensics_tiou_reward": 0.9808412790298462,
      "rewards/forensics_sinkhorn_reward": 0.7002948820590973,
      "rewards/forensics_strict_edge_reward": 0.07931584492325783,
      "rewards/forensics_tiou_reward": 0.5356745570898056,
      "step": 932
    },
    {
      "completion_length": 23.921875,
      "epoch": 2.8632872503840243,
      "grad_norm": 11.459807395935059,
      "kl": 0.4541015625,
      "learning_rate": 2.8307692307692306e-07,
      "loss": -0.0314,
      "reward": 1.4502628445625305,
      "reward_std": 0.3836623281240463,
      "reward_std_all": 0.8733325302600861,
      "rewards-max/forensics_sinkhorn_reward": 0.9826772809028625,
      "rewards-max/forensics_strict_edge_reward": 0.756708562374115,
      "rewards-max/forensics_tiou_reward": 0.9849216639995575,
      "rewards/forensics_sinkhorn_reward": 0.7308264374732971,
      "rewards/forensics_strict_edge_reward": 0.11494498327374458,
      "rewards/forensics_tiou_reward": 0.6044914424419403,
      "step": 933
    },
    {
      "completion_length": 22.296875,
      "epoch": 2.8663594470046085,
      "grad_norm": 10.381409645080566,
      "kl": 0.58203125,
      "learning_rate": 2.823076923076923e-07,
      "loss": -0.0291,
      "reward": 1.565758764743805,
      "reward_std": 0.33022095263004303,
      "reward_std_all": 0.8051146268844604,
      "rewards-max/forensics_sinkhorn_reward": 0.966665118932724,
      "rewards-max/forensics_strict_edge_reward": 0.5284441187977791,
      "rewards-max/forensics_tiou_reward": 0.9467686116695404,
      "rewards/forensics_sinkhorn_reward": 0.8143005967140198,
      "rewards/forensics_strict_edge_reward": 0.08309459965676069,
      "rewards/forensics_tiou_reward": 0.6683636009693146,
      "step": 934
    },
    {
      "completion_length": 28.125,
      "epoch": 2.869431643625192,
      "grad_norm": 10.751776695251465,
      "kl": 0.521484375,
      "learning_rate": 2.815384615384615e-07,
      "loss": -0.0327,
      "reward": 1.3369471430778503,
      "reward_std": 0.4065866619348526,
      "reward_std_all": 0.9358828663825989,
      "rewards-max/forensics_sinkhorn_reward": 0.9872291684150696,
      "rewards-max/forensics_strict_edge_reward": 0.7277118265628815,
      "rewards-max/forensics_tiou_reward": 0.9953778684139252,
      "rewards/forensics_sinkhorn_reward": 0.6939515769481659,
      "rewards/forensics_strict_edge_reward": 0.09456549398601055,
      "rewards/forensics_tiou_reward": 0.548430010676384,
      "step": 935
    },
    {
      "completion_length": 28.40625,
      "epoch": 2.8725038402457757,
      "grad_norm": 12.305672645568848,
      "kl": 0.705078125,
      "learning_rate": 2.8076923076923075e-07,
      "loss": 0.0001,
      "reward": 1.5263246297836304,
      "reward_std": 0.4478929787874222,
      "reward_std_all": 0.8614106774330139,
      "rewards-max/forensics_sinkhorn_reward": 0.9893360137939453,
      "rewards-max/forensics_strict_edge_reward": 0.9067722856998444,
      "rewards-max/forensics_tiou_reward": 0.9990800619125366,
      "rewards/forensics_sinkhorn_reward": 0.7357090711593628,
      "rewards/forensics_strict_edge_reward": 0.12916820868849754,
      "rewards/forensics_tiou_reward": 0.6614474356174469,
      "step": 936
    },
    {
      "completion_length": 24.953125,
      "epoch": 2.8755760368663594,
      "grad_norm": 11.383291244506836,
      "kl": 0.513671875,
      "learning_rate": 2.8e-07,
      "loss": -0.0498,
      "reward": 1.4309676885604858,
      "reward_std": 0.2762216329574585,
      "reward_std_all": 0.8311919867992401,
      "rewards-max/forensics_sinkhorn_reward": 0.9854877889156342,
      "rewards-max/forensics_strict_edge_reward": 0.5106222033500671,
      "rewards-max/forensics_tiou_reward": 0.988166332244873,
      "rewards/forensics_sinkhorn_reward": 0.7587622106075287,
      "rewards/forensics_strict_edge_reward": 0.04211480729281902,
      "rewards/forensics_tiou_reward": 0.6300905644893646,
      "step": 937
    },
    {
      "completion_length": 21.96875,
      "epoch": 2.878648233486943,
      "grad_norm": 10.276823043823242,
      "kl": 0.44921875,
      "learning_rate": 2.7923076923076923e-07,
      "loss": -0.0643,
      "reward": 1.4742506742477417,
      "reward_std": 0.4619511067867279,
      "reward_std_all": 0.889121949672699,
      "rewards-max/forensics_sinkhorn_reward": 0.9826338887214661,
      "rewards-max/forensics_strict_edge_reward": 0.8351600170135498,
      "rewards-max/forensics_tiou_reward": 0.985622227191925,
      "rewards/forensics_sinkhorn_reward": 0.7672776579856873,
      "rewards/forensics_strict_edge_reward": 0.1203397661447525,
      "rewards/forensics_tiou_reward": 0.5866332650184631,
      "step": 938
    },
    {
      "completion_length": 26.09375,
      "epoch": 2.881720430107527,
      "grad_norm": 11.82446575164795,
      "kl": 0.5859375,
      "learning_rate": 2.7846153846153845e-07,
      "loss": -0.081,
      "reward": 1.593321979045868,
      "reward_std": 0.4838501065969467,
      "reward_std_all": 0.8447310626506805,
      "rewards-max/forensics_sinkhorn_reward": 0.9876781702041626,
      "rewards-max/forensics_strict_edge_reward": 0.882964164018631,
      "rewards-max/forensics_tiou_reward": 0.9975452423095703,
      "rewards/forensics_sinkhorn_reward": 0.8127025961875916,
      "rewards/forensics_strict_edge_reward": 0.12140190228819847,
      "rewards/forensics_tiou_reward": 0.6592175364494324,
      "step": 939
    },
    {
      "completion_length": 27.109375,
      "epoch": 2.8847926267281108,
      "grad_norm": 14.458318710327148,
      "kl": 0.6005859375,
      "learning_rate": 2.7769230769230766e-07,
      "loss": -0.1495,
      "reward": 1.5040711164474487,
      "reward_std": 0.4043854922056198,
      "reward_std_all": 0.8520780801773071,
      "rewards-max/forensics_sinkhorn_reward": 0.9873439371585846,
      "rewards-max/forensics_strict_edge_reward": 0.7647041082382202,
      "rewards-max/forensics_tiou_reward": 0.9934761226177216,
      "rewards/forensics_sinkhorn_reward": 0.751713216304779,
      "rewards/forensics_strict_edge_reward": 0.12138323113322258,
      "rewards/forensics_tiou_reward": 0.630974680185318,
      "step": 940
    },
    {
      "completion_length": 24.515625,
      "epoch": 2.8878648233486945,
      "grad_norm": 12.069965362548828,
      "kl": 0.712890625,
      "learning_rate": 2.7692307692307693e-07,
      "loss": -0.0783,
      "reward": 1.3890073895454407,
      "reward_std": 0.31072893738746643,
      "reward_std_all": 0.8849961757659912,
      "rewards-max/forensics_sinkhorn_reward": 0.9833422005176544,
      "rewards-max/forensics_strict_edge_reward": 0.882964164018631,
      "rewards-max/forensics_tiou_reward": 0.9858536720275879,
      "rewards/forensics_sinkhorn_reward": 0.7385903894901276,
      "rewards/forensics_strict_edge_reward": 0.09912057034671307,
      "rewards/forensics_tiou_reward": 0.5512964874505997,
      "step": 941
    },
    {
      "completion_length": 25.015625,
      "epoch": 2.890937019969278,
      "grad_norm": 9.080659866333008,
      "kl": 0.431640625,
      "learning_rate": 2.7615384615384614e-07,
      "loss": -0.0244,
      "reward": 1.5142048597335815,
      "reward_std": 0.31937307119369507,
      "reward_std_all": 0.813483476638794,
      "rewards-max/forensics_sinkhorn_reward": 0.983540415763855,
      "rewards-max/forensics_strict_edge_reward": 0.7875787317752838,
      "rewards-max/forensics_tiou_reward": 0.9866672456264496,
      "rewards/forensics_sinkhorn_reward": 0.7795803248882294,
      "rewards/forensics_strict_edge_reward": 0.11151043325662613,
      "rewards/forensics_tiou_reward": 0.6231141686439514,
      "step": 942
    },
    {
      "completion_length": 25.4375,
      "epoch": 2.8940092165898617,
      "grad_norm": 12.538893699645996,
      "kl": 0.4931640625,
      "learning_rate": 2.753846153846154e-07,
      "loss": -0.0455,
      "reward": 1.5522838234901428,
      "reward_std": 0.3277926743030548,
      "reward_std_all": 0.8043483197689056,
      "rewards-max/forensics_sinkhorn_reward": 0.9855642914772034,
      "rewards-max/forensics_strict_edge_reward": 0.9375866651535034,
      "rewards-max/forensics_tiou_reward": 0.9925883412361145,
      "rewards/forensics_sinkhorn_reward": 0.7516810595989227,
      "rewards/forensics_strict_edge_reward": 0.15826090425252914,
      "rewards/forensics_tiou_reward": 0.6423418521881104,
      "step": 943
    },
    {
      "completion_length": 23.84375,
      "epoch": 2.8970814132104454,
      "grad_norm": 10.504103660583496,
      "kl": 0.595703125,
      "learning_rate": 2.7461538461538457e-07,
      "loss": -0.1667,
      "reward": 1.427509069442749,
      "reward_std": 0.3631257712841034,
      "reward_std_all": 0.8890738487243652,
      "rewards-max/forensics_sinkhorn_reward": 0.989384651184082,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.741190493106842,
      "rewards/forensics_strict_edge_reward": 0.13865602761507034,
      "rewards/forensics_tiou_reward": 0.5476624667644501,
      "step": 944
    },
    {
      "completion_length": 25.0625,
      "epoch": 2.900153609831029,
      "grad_norm": 10.872032165527344,
      "kl": 0.4228515625,
      "learning_rate": 2.7384615384615383e-07,
      "loss": -0.1701,
      "reward": 1.1782805919647217,
      "reward_std": 0.2949020713567734,
      "reward_std_all": 0.9732840359210968,
      "rewards-max/forensics_sinkhorn_reward": 0.9704123139381409,
      "rewards-max/forensics_strict_edge_reward": 0.6008101999759674,
      "rewards-max/forensics_tiou_reward": 0.9552980065345764,
      "rewards/forensics_sinkhorn_reward": 0.6214421391487122,
      "rewards/forensics_strict_edge_reward": 0.08019178360700607,
      "rewards/forensics_tiou_reward": 0.47664666175842285,
      "step": 945
    },
    {
      "completion_length": 23.578125,
      "epoch": 2.903225806451613,
      "grad_norm": 9.51711654663086,
      "kl": 0.681640625,
      "learning_rate": 2.7307692307692305e-07,
      "loss": -0.1344,
      "reward": 1.6696144938468933,
      "reward_std": 0.40521693229675293,
      "reward_std_all": 0.7840515077114105,
      "rewards-max/forensics_sinkhorn_reward": 0.9882997870445251,
      "rewards-max/forensics_strict_edge_reward": 0.9232408702373505,
      "rewards-max/forensics_tiou_reward": 0.9973045885562897,
      "rewards/forensics_sinkhorn_reward": 0.8327673077583313,
      "rewards/forensics_strict_edge_reward": 0.13978039100766182,
      "rewards/forensics_tiou_reward": 0.6970667541027069,
      "step": 946
    },
    {
      "completion_length": 27.546875,
      "epoch": 2.9062980030721968,
      "grad_norm": 11.761460304260254,
      "kl": 0.51171875,
      "learning_rate": 2.723076923076923e-07,
      "loss": -0.0187,
      "reward": 1.6119054555892944,
      "reward_std": 0.46396248042583466,
      "reward_std_all": 0.8493372797966003,
      "rewards-max/forensics_sinkhorn_reward": 0.9891187846660614,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.7851728796958923,
      "rewards/forensics_strict_edge_reward": 0.1766245923936367,
      "rewards/forensics_tiou_reward": 0.6501079499721527,
      "step": 947
    },
    {
      "completion_length": 23.3125,
      "epoch": 2.9093701996927805,
      "grad_norm": 9.427928924560547,
      "kl": 0.53515625,
      "learning_rate": 2.7153846153846153e-07,
      "loss": -0.0201,
      "reward": 1.4902912378311157,
      "reward_std": 0.5065160244703293,
      "reward_std_all": 0.8903568089008331,
      "rewards-max/forensics_sinkhorn_reward": 0.9840108752250671,
      "rewards-max/forensics_strict_edge_reward": 0.7933230996131897,
      "rewards-max/forensics_tiou_reward": 0.9831026792526245,
      "rewards/forensics_sinkhorn_reward": 0.7831299901008606,
      "rewards/forensics_strict_edge_reward": 0.12236202880740166,
      "rewards/forensics_tiou_reward": 0.5847992897033691,
      "step": 948
    },
    {
      "completion_length": 25.890625,
      "epoch": 2.912442396313364,
      "grad_norm": 8.49057388305664,
      "kl": 0.56640625,
      "learning_rate": 2.7076923076923074e-07,
      "loss": -0.1261,
      "reward": 1.6423245072364807,
      "reward_std": 0.43548066914081573,
      "reward_std_all": 0.7890190184116364,
      "rewards-max/forensics_sinkhorn_reward": 0.9780276715755463,
      "rewards-max/forensics_strict_edge_reward": 0.625803679227829,
      "rewards-max/forensics_tiou_reward": 0.9716734886169434,
      "rewards/forensics_sinkhorn_reward": 0.7934566736221313,
      "rewards/forensics_strict_edge_reward": 0.20448060519993305,
      "rewards/forensics_tiou_reward": 0.6443873047828674,
      "step": 949
    },
    {
      "completion_length": 25.578125,
      "epoch": 2.9155145929339477,
      "grad_norm": 11.65039348602295,
      "kl": 0.646484375,
      "learning_rate": 2.7e-07,
      "loss": -0.0904,
      "reward": 1.4138880968093872,
      "reward_std": 0.5092189311981201,
      "reward_std_all": 0.937766969203949,
      "rewards-max/forensics_sinkhorn_reward": 0.9677245616912842,
      "rewards-max/forensics_strict_edge_reward": 0.8333333432674408,
      "rewards-max/forensics_tiou_reward": 0.9368546307086945,
      "rewards/forensics_sinkhorn_reward": 0.7126549184322357,
      "rewards/forensics_strict_edge_reward": 0.1313234306871891,
      "rewards/forensics_tiou_reward": 0.5699098110198975,
      "step": 950
    },
    {
      "completion_length": 28.625,
      "epoch": 2.9185867895545314,
      "grad_norm": 10.170876502990723,
      "kl": 0.52734375,
      "learning_rate": 2.692307692307692e-07,
      "loss": -0.0412,
      "reward": 1.455837607383728,
      "reward_std": 0.38118530809879303,
      "reward_std_all": 0.8735865950584412,
      "rewards-max/forensics_sinkhorn_reward": 0.9877819418907166,
      "rewards-max/forensics_strict_edge_reward": 0.7742429971694946,
      "rewards-max/forensics_tiou_reward": 0.994323194026947,
      "rewards/forensics_sinkhorn_reward": 0.7265004813671112,
      "rewards/forensics_strict_edge_reward": 0.08547822199761868,
      "rewards/forensics_tiou_reward": 0.6438588798046112,
      "step": 951
    },
    {
      "completion_length": 21.59375,
      "epoch": 2.921658986175115,
      "grad_norm": 9.103582382202148,
      "kl": 0.5966796875,
      "learning_rate": 2.6846153846153844e-07,
      "loss": -0.1254,
      "reward": 1.3873414993286133,
      "reward_std": 0.2944198399782181,
      "reward_std_all": 0.8721101880073547,
      "rewards-max/forensics_sinkhorn_reward": 0.9778429269790649,
      "rewards-max/forensics_strict_edge_reward": 0.6409624814987183,
      "rewards-max/forensics_tiou_reward": 0.969135046005249,
      "rewards/forensics_sinkhorn_reward": 0.7451627552509308,
      "rewards/forensics_strict_edge_reward": 0.07036530412733555,
      "rewards/forensics_tiou_reward": 0.5718133449554443,
      "step": 952
    },
    {
      "completion_length": 29.203125,
      "epoch": 2.924731182795699,
      "grad_norm": 8.982972145080566,
      "kl": 0.50390625,
      "learning_rate": 2.6769230769230765e-07,
      "loss": -0.0528,
      "reward": 1.5998419523239136,
      "reward_std": 0.2976543605327606,
      "reward_std_all": 0.7762234508991241,
      "rewards-max/forensics_sinkhorn_reward": 0.9879849851131439,
      "rewards-max/forensics_strict_edge_reward": 0.9130095541477203,
      "rewards-max/forensics_tiou_reward": 0.9950891435146332,
      "rewards/forensics_sinkhorn_reward": 0.7845763564109802,
      "rewards/forensics_strict_edge_reward": 0.12338914722204208,
      "rewards/forensics_tiou_reward": 0.6918765008449554,
      "step": 953
    },
    {
      "completion_length": 23.984375,
      "epoch": 2.9278033794162828,
      "grad_norm": 11.291303634643555,
      "kl": 0.453125,
      "learning_rate": 2.669230769230769e-07,
      "loss": -0.1116,
      "reward": 1.413856863975525,
      "reward_std": 0.44666673243045807,
      "reward_std_all": 0.9286151826381683,
      "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": 0.722114622592926,
      "rewards/forensics_strict_edge_reward": 0.1236409842967987,
      "rewards/forensics_tiou_reward": 0.568101316690445,
      "step": 954
    },
    {
      "completion_length": 27.53125,
      "epoch": 2.9308755760368665,
      "grad_norm": 18.259706497192383,
      "kl": 0.8125,
      "learning_rate": 2.6615384615384613e-07,
      "loss": -0.049,
      "reward": 1.4625110030174255,
      "reward_std": 0.3333699628710747,
      "reward_std_all": 0.8765382766723633,
      "rewards-max/forensics_sinkhorn_reward": 0.9862655401229858,
      "rewards-max/forensics_strict_edge_reward": 0.8704091012477875,
      "rewards-max/forensics_tiou_reward": 0.9912331998348236,
      "rewards/forensics_sinkhorn_reward": 0.7587488889694214,
      "rewards/forensics_strict_edge_reward": 0.11609544418752193,
      "rewards/forensics_tiou_reward": 0.5876666009426117,
      "step": 955
    },
    {
      "completion_length": 27.9375,
      "epoch": 2.93394777265745,
      "grad_norm": 10.391119003295898,
      "kl": 0.578125,
      "learning_rate": 2.653846153846154e-07,
      "loss": -0.0505,
      "reward": 1.5500925183296204,
      "reward_std": 0.30713726580142975,
      "reward_std_all": 0.7897265255451202,
      "rewards-max/forensics_sinkhorn_reward": 0.9872854351997375,
      "rewards-max/forensics_strict_edge_reward": 0.7521992623806,
      "rewards-max/forensics_tiou_reward": 0.9951671063899994,
      "rewards/forensics_sinkhorn_reward": 0.7978770434856415,
      "rewards/forensics_strict_edge_reward": 0.09322734922170639,
      "rewards/forensics_tiou_reward": 0.6589881777763367,
      "step": 956
    },
    {
      "completion_length": 25.53125,
      "epoch": 2.9370199692780337,
      "grad_norm": 10.905538558959961,
      "kl": 0.7587890625,
      "learning_rate": 2.646153846153846e-07,
      "loss": -0.0452,
      "reward": 1.551758050918579,
      "reward_std": 0.4993287920951843,
      "reward_std_all": 0.8688229620456696,
      "rewards-max/forensics_sinkhorn_reward": 0.989237904548645,
      "rewards-max/forensics_strict_edge_reward": 0.8827048540115356,
      "rewards-max/forensics_tiou_reward": 0.9977883696556091,
      "rewards/forensics_sinkhorn_reward": 0.7784021198749542,
      "rewards/forensics_strict_edge_reward": 0.15265380591154099,
      "rewards/forensics_tiou_reward": 0.6207022368907928,
      "step": 957
    },
    {
      "completion_length": 24.859375,
      "epoch": 2.9400921658986174,
      "grad_norm": 9.561492919921875,
      "kl": 0.4033203125,
      "learning_rate": 2.638461538461538e-07,
      "loss": -0.0221,
      "reward": 1.375036895275116,
      "reward_std": 0.31005623936653137,
      "reward_std_all": 0.9001459181308746,
      "rewards-max/forensics_sinkhorn_reward": 0.989445298910141,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.7152158617973328,
      "rewards/forensics_strict_edge_reward": 0.12196462601423264,
      "rewards/forensics_tiou_reward": 0.5378563404083252,
      "step": 958
    },
    {
      "completion_length": 24.484375,
      "epoch": 2.943164362519201,
      "grad_norm": 11.070252418518066,
      "kl": 0.587890625,
      "learning_rate": 2.630769230769231e-07,
      "loss": -0.0909,
      "reward": 1.543611228466034,
      "reward_std": 0.4626454561948776,
      "reward_std_all": 0.858012467622757,
      "rewards-max/forensics_sinkhorn_reward": 0.9812095761299133,
      "rewards-max/forensics_strict_edge_reward": 0.7246644794940948,
      "rewards-max/forensics_tiou_reward": 0.9845250248908997,
      "rewards/forensics_sinkhorn_reward": 0.801420271396637,
      "rewards/forensics_strict_edge_reward": 0.12058775126934052,
      "rewards/forensics_tiou_reward": 0.62160325050354,
      "step": 959
    },
    {
      "completion_length": 29.234375,
      "epoch": 2.946236559139785,
      "grad_norm": 10.963858604431152,
      "kl": 0.6044921875,
      "learning_rate": 2.623076923076923e-07,
      "loss": -0.0017,
      "reward": 1.4117045998573303,
      "reward_std": 0.4372812956571579,
      "reward_std_all": 0.9215923845767975,
      "rewards-max/forensics_sinkhorn_reward": 0.9882851243019104,
      "rewards-max/forensics_strict_edge_reward": 0.9016326665878296,
      "rewards-max/forensics_tiou_reward": 0.9994141161441803,
      "rewards/forensics_sinkhorn_reward": 0.6971510350704193,
      "rewards/forensics_strict_edge_reward": 0.16431833803653717,
      "rewards/forensics_tiou_reward": 0.5502351373434067,
      "step": 960
    },
    {
      "completion_length": 26.40625,
      "epoch": 2.9493087557603688,
      "grad_norm": 9.70680046081543,
      "kl": 0.443359375,
      "learning_rate": 2.615384615384615e-07,
      "loss": -0.1295,
      "reward": 1.2142646312713623,
      "reward_std": 0.5638857483863831,
      "reward_std_all": 1.0566757917404175,
      "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": 0.6354207694530487,
      "rewards/forensics_strict_edge_reward": 0.1151479035615921,
      "rewards/forensics_tiou_reward": 0.4636959731578827,
      "step": 961
    },
    {
      "completion_length": 22.34375,
      "epoch": 2.9523809523809526,
      "grad_norm": 10.032991409301758,
      "kl": 0.376953125,
      "learning_rate": 2.6076923076923073e-07,
      "loss": -0.1159,
      "reward": 1.5638837218284607,
      "reward_std": 0.3593626916408539,
      "reward_std_all": 0.8267956674098969,
      "rewards-max/forensics_sinkhorn_reward": 0.983537495136261,
      "rewards-max/forensics_strict_edge_reward": 0.7837068438529968,
      "rewards-max/forensics_tiou_reward": 0.9837991893291473,
      "rewards/forensics_sinkhorn_reward": 0.7622393369674683,
      "rewards/forensics_strict_edge_reward": 0.143234521150589,
      "rewards/forensics_tiou_reward": 0.6584098041057587,
      "step": 962
    },
    {
      "completion_length": 27.078125,
      "epoch": 2.955453149001536,
      "grad_norm": 20.432758331298828,
      "kl": 0.912109375,
      "learning_rate": 2.6e-07,
      "loss": -0.0527,
      "reward": 1.5363868474960327,
      "reward_std": 0.4700339436531067,
      "reward_std_all": 0.8550460636615753,
      "rewards-max/forensics_sinkhorn_reward": 0.9867478013038635,
      "rewards-max/forensics_strict_edge_reward": 0.7204575538635254,
      "rewards-max/forensics_tiou_reward": 0.9929979741573334,
      "rewards/forensics_sinkhorn_reward": 0.7570982277393341,
      "rewards/forensics_strict_edge_reward": 0.11311191692948341,
      "rewards/forensics_tiou_reward": 0.6661765575408936,
      "step": 963
    },
    {
      "completion_length": 27.65625,
      "epoch": 2.9585253456221197,
      "grad_norm": 9.656635284423828,
      "kl": 0.5595703125,
      "learning_rate": 2.592307692307692e-07,
      "loss": -0.0607,
      "reward": 1.671908974647522,
      "reward_std": 0.4487483501434326,
      "reward_std_all": 0.7942251563072205,
      "rewards-max/forensics_sinkhorn_reward": 0.989323616027832,
      "rewards-max/forensics_strict_edge_reward": 0.9687933325767517,
      "rewards-max/forensics_tiou_reward": 0.999517947435379,
      "rewards/forensics_sinkhorn_reward": 0.8186410069465637,
      "rewards/forensics_strict_edge_reward": 0.17422859370708466,
      "rewards/forensics_tiou_reward": 0.6790393888950348,
      "step": 964
    },
    {
      "completion_length": 26.859375,
      "epoch": 2.9615975422427034,
      "grad_norm": 12.81090259552002,
      "kl": 0.8203125,
      "learning_rate": 2.584615384615385e-07,
      "loss": 0.0407,
      "reward": 1.5586797595024109,
      "reward_std": 0.37631693482398987,
      "reward_std_all": 0.8196202218532562,
      "rewards-max/forensics_sinkhorn_reward": 0.9891857206821442,
      "rewards-max/forensics_strict_edge_reward": 0.9093840420246124,
      "rewards-max/forensics_tiou_reward": 0.999435544013977,
      "rewards/forensics_sinkhorn_reward": 0.7780648469924927,
      "rewards/forensics_strict_edge_reward": 0.11918209493160248,
      "rewards/forensics_tiou_reward": 0.6614328920841217,
      "step": 965
    },
    {
      "completion_length": 28.625,
      "epoch": 2.964669738863287,
      "grad_norm": 10.680171966552734,
      "kl": 0.435546875,
      "learning_rate": 2.5769230769230764e-07,
      "loss": -0.0892,
      "reward": 1.6172122955322266,
      "reward_std": 0.356873095035553,
      "reward_std_all": 0.7866029143333435,
      "rewards-max/forensics_sinkhorn_reward": 0.989765852689743,
      "rewards-max/forensics_strict_edge_reward": 0.9918040335178375,
      "rewards-max/forensics_tiou_reward": 0.999970555305481,
      "rewards/forensics_sinkhorn_reward": 0.8057549297809601,
      "rewards/forensics_strict_edge_reward": 0.1460893601179123,
      "rewards/forensics_tiou_reward": 0.6653680503368378,
      "step": 966
    },
    {
      "completion_length": 24.6875,
      "epoch": 2.967741935483871,
      "grad_norm": 11.316210746765137,
      "kl": 0.4765625,
      "learning_rate": 2.569230769230769e-07,
      "loss": -0.0739,
      "reward": 1.2687098383903503,
      "reward_std": 0.35198456048965454,
      "reward_std_all": 0.9530895948410034,
      "rewards-max/forensics_sinkhorn_reward": 0.982307642698288,
      "rewards-max/forensics_strict_edge_reward": 0.756708562374115,
      "rewards-max/forensics_tiou_reward": 0.983470618724823,
      "rewards/forensics_sinkhorn_reward": 0.6981866657733917,
      "rewards/forensics_strict_edge_reward": 0.049160173162817955,
      "rewards/forensics_tiou_reward": 0.5213630795478821,
      "step": 967
    },
    {
      "completion_length": 22.234375,
      "epoch": 2.9708141321044548,
      "grad_norm": 18.45595932006836,
      "kl": 0.396484375,
      "learning_rate": 2.5615384615384617e-07,
      "loss": -0.1065,
      "reward": 1.5659207105636597,
      "reward_std": 0.3683861941099167,
      "reward_std_all": 0.8012612760066986,
      "rewards-max/forensics_sinkhorn_reward": 0.9811751246452332,
      "rewards-max/forensics_strict_edge_reward": 0.756708562374115,
      "rewards-max/forensics_tiou_reward": 0.9789556562900543,
      "rewards/forensics_sinkhorn_reward": 0.8108171820640564,
      "rewards/forensics_strict_edge_reward": 0.1042372565716505,
      "rewards/forensics_tiou_reward": 0.6508662402629852,
      "step": 968
    },
    {
      "completion_length": 26.3125,
      "epoch": 2.9738863287250386,
      "grad_norm": 13.787668228149414,
      "kl": 0.5400390625,
      "learning_rate": 2.553846153846154e-07,
      "loss": -0.1973,
      "reward": 1.483374834060669,
      "reward_std": 0.481344074010849,
      "reward_std_all": 0.8956575393676758,
      "rewards-max/forensics_sinkhorn_reward": 0.9879507422447205,
      "rewards-max/forensics_strict_edge_reward": 0.8959447741508484,
      "rewards-max/forensics_tiou_reward": 0.9961268305778503,
      "rewards/forensics_sinkhorn_reward": 0.7393209934234619,
      "rewards/forensics_strict_edge_reward": 0.125632181763649,
      "rewards/forensics_tiou_reward": 0.6184215545654297,
      "step": 969
    },
    {
      "completion_length": 27.5625,
      "epoch": 2.976958525345622,
      "grad_norm": 10.652478218078613,
      "kl": 0.5830078125,
      "learning_rate": 2.546153846153846e-07,
      "loss": -0.097,
      "reward": 1.5034103989601135,
      "reward_std": 0.4572731852531433,
      "reward_std_all": 0.8855567872524261,
      "rewards-max/forensics_sinkhorn_reward": 0.9874855875968933,
      "rewards-max/forensics_strict_edge_reward": 0.8667194843292236,
      "rewards-max/forensics_tiou_reward": 0.9952392578125,
      "rewards/forensics_sinkhorn_reward": 0.7494447827339172,
      "rewards/forensics_strict_edge_reward": 0.13471754640340805,
      "rewards/forensics_tiou_reward": 0.6192480623722076,
      "step": 970
    },
    {
      "completion_length": 26.5,
      "epoch": 2.980030721966206,
      "grad_norm": 13.144832611083984,
      "kl": 0.5234375,
      "learning_rate": 2.538461538461538e-07,
      "loss": -0.0412,
      "reward": 1.5441746711730957,
      "reward_std": 0.36215221881866455,
      "reward_std_all": 0.8156729340553284,
      "rewards-max/forensics_sinkhorn_reward": 0.9883559942245483,
      "rewards-max/forensics_strict_edge_reward": 0.9454971551895142,
      "rewards-max/forensics_tiou_reward": 0.9996867179870605,
      "rewards/forensics_sinkhorn_reward": 0.7485300898551941,
      "rewards/forensics_strict_edge_reward": 0.16309376060962677,
      "rewards/forensics_tiou_reward": 0.6325508058071136,
      "step": 971
    },
    {
      "completion_length": 27.765625,
      "epoch": 2.9831029185867894,
      "grad_norm": 10.648205757141113,
      "kl": 0.455078125,
      "learning_rate": 2.530769230769231e-07,
      "loss": -0.1226,
      "reward": 1.6502450704574585,
      "reward_std": 0.44095611572265625,
      "reward_std_all": 0.7981534600257874,
      "rewards-max/forensics_sinkhorn_reward": 0.9892333745956421,
      "rewards-max/forensics_strict_edge_reward": 0.9375866651535034,
      "rewards-max/forensics_tiou_reward": 0.9990410506725311,
      "rewards/forensics_sinkhorn_reward": 0.8080025315284729,
      "rewards/forensics_strict_edge_reward": 0.16933926939964294,
      "rewards/forensics_tiou_reward": 0.672903299331665,
      "step": 972
    },
    {
      "completion_length": 24.140625,
      "epoch": 2.986175115207373,
      "grad_norm": 12.046374320983887,
      "kl": 0.517578125,
      "learning_rate": 2.523076923076923e-07,
      "loss": -0.047,
      "reward": 1.5117390751838684,
      "reward_std": 0.45180559158325195,
      "reward_std_all": 0.8783425986766815,
      "rewards-max/forensics_sinkhorn_reward": 0.9895409345626831,
      "rewards-max/forensics_strict_edge_reward": 0.8419698476791382,
      "rewards-max/forensics_tiou_reward": 0.9983291029930115,
      "rewards/forensics_sinkhorn_reward": 0.756033331155777,
      "rewards/forensics_strict_edge_reward": 0.13652334362268448,
      "rewards/forensics_tiou_reward": 0.6191824078559875,
      "step": 973
    },
    {
      "completion_length": 23.21875,
      "epoch": 2.989247311827957,
      "grad_norm": 9.078128814697266,
      "kl": 0.583984375,
      "learning_rate": 2.515384615384615e-07,
      "loss": -0.0729,
      "reward": 1.4219392538070679,
      "reward_std": 0.386695921421051,
      "reward_std_all": 0.9087755084037781,
      "rewards-max/forensics_sinkhorn_reward": 0.9889647662639618,
      "rewards-max/forensics_strict_edge_reward": 0.9677534997463226,
      "rewards-max/forensics_tiou_reward": 0.9999025762081146,
      "rewards/forensics_sinkhorn_reward": 0.706020176410675,
      "rewards/forensics_strict_edge_reward": 0.12089506909251213,
      "rewards/forensics_tiou_reward": 0.5950240790843964,
      "step": 974
    },
    {
      "completion_length": 24.5,
      "epoch": 2.9923195084485408,
      "grad_norm": 12.084003448486328,
      "kl": 0.52734375,
      "learning_rate": 2.507692307692307e-07,
      "loss": -0.0759,
      "reward": 1.472931146621704,
      "reward_std": 0.4092588722705841,
      "reward_std_all": 0.8659392297267914,
      "rewards-max/forensics_sinkhorn_reward": 0.9873735010623932,
      "rewards-max/forensics_strict_edge_reward": 0.8389547169208527,
      "rewards-max/forensics_tiou_reward": 0.9957634210586548,
      "rewards/forensics_sinkhorn_reward": 0.7660133242607117,
      "rewards/forensics_strict_edge_reward": 0.09738540276885033,
      "rewards/forensics_tiou_reward": 0.6095323860645294,
      "step": 975
    },
    {
      "completion_length": 25.671875,
      "epoch": 2.9953917050691246,
      "grad_norm": 14.902130126953125,
      "kl": 0.58203125,
      "learning_rate": 2.5e-07,
      "loss": -0.1143,
      "reward": 1.661864161491394,
      "reward_std": 0.5025156885385513,
      "reward_std_all": 0.8194456696510315,
      "rewards-max/forensics_sinkhorn_reward": 0.9901103377342224,
      "rewards-max/forensics_strict_edge_reward": 0.9918040335178375,
      "rewards-max/forensics_tiou_reward": 0.9996806085109711,
      "rewards/forensics_sinkhorn_reward": 0.7970858812332153,
      "rewards/forensics_strict_edge_reward": 0.17578765749931335,
      "rewards/forensics_tiou_reward": 0.6889906823635101,
      "step": 976
    },
    {
      "completion_length": 27.09375,
      "epoch": 2.998463901689708,
      "grad_norm": 14.040796279907227,
      "kl": 0.6962890625,
      "learning_rate": 2.492307692307692e-07,
      "loss": -0.0513,
      "reward": 1.4051846265792847,
      "reward_std": 0.5411540567874908,
      "reward_std_all": 0.9468677639961243,
      "rewards-max/forensics_sinkhorn_reward": 0.9827591180801392,
      "rewards-max/forensics_strict_edge_reward": 0.824172168970108,
      "rewards-max/forensics_tiou_reward": 0.979955643415451,
      "rewards/forensics_sinkhorn_reward": 0.7274400293827057,
      "rewards/forensics_strict_edge_reward": 0.10914045572280884,
      "rewards/forensics_tiou_reward": 0.5686041116714478,
      "step": 977
    },
    {
      "completion_length": 32.60000228881836,
      "epoch": 3.0,
      "grad_norm": 8.18755054473877,
      "kl": 0.6171875,
      "learning_rate": 2.4846153846153847e-07,
      "loss": 0.0047,
      "reward": 2.1960229873657227,
      "reward_std": 0.692744255065918,
      "reward_std_all": 0.6943232417106628,
      "rewards-max/forensics_sinkhorn_reward": 0.9896095395088196,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.8377367258071899,
      "rewards/forensics_strict_edge_reward": 0.47126060724258423,
      "rewards/forensics_tiou_reward": 0.8870258331298828,
      "step": 978
    },
    {
      "completion_length": 24.421875,
      "epoch": 3.003072196620584,
      "grad_norm": 26.4471378326416,
      "kl": 1.197265625,
      "learning_rate": 2.476923076923077e-07,
      "loss": -0.0134,
      "reward": 1.5305132865905762,
      "reward_std": 0.4744367152452469,
      "reward_std_all": 0.8538832664489746,
      "rewards-max/forensics_sinkhorn_reward": 0.9825369417667389,
      "rewards-max/forensics_strict_edge_reward": 0.7346639335155487,
      "rewards-max/forensics_tiou_reward": 0.9825093746185303,
      "rewards/forensics_sinkhorn_reward": 0.7852500379085541,
      "rewards/forensics_strict_edge_reward": 0.10749103128910065,
      "rewards/forensics_tiou_reward": 0.6377722322940826,
      "step": 979
    },
    {
      "completion_length": 25.0,
      "epoch": 3.0061443932411676,
      "grad_norm": 9.769305229187012,
      "kl": 0.583984375,
      "learning_rate": 2.469230769230769e-07,
      "loss": -0.0383,
      "reward": 1.4944957494735718,
      "reward_std": 0.36794905364513397,
      "reward_std_all": 0.8585816025733948,
      "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": 0.7417735159397125,
      "rewards/forensics_strict_edge_reward": 0.15516357123851776,
      "rewards/forensics_tiou_reward": 0.5975586175918579,
      "step": 980
    },
    {
      "completion_length": 24.28125,
      "epoch": 3.0092165898617513,
      "grad_norm": 10.089637756347656,
      "kl": 0.556640625,
      "learning_rate": 2.4615384615384616e-07,
      "loss": -0.1566,
      "reward": 1.5976750254631042,
      "reward_std": 0.37606917321681976,
      "reward_std_all": 0.8030774295330048,
      "rewards-max/forensics_sinkhorn_reward": 0.9857948422431946,
      "rewards-max/forensics_strict_edge_reward": 0.7093063294887543,
      "rewards-max/forensics_tiou_reward": 0.9914017021656036,
      "rewards/forensics_sinkhorn_reward": 0.8071529865264893,
      "rewards/forensics_strict_edge_reward": 0.12913822755217552,
      "rewards/forensics_tiou_reward": 0.6613838076591492,
      "step": 981
    },
    {
      "completion_length": 24.09375,
      "epoch": 3.0122887864823347,
      "grad_norm": 10.818264961242676,
      "kl": 0.650390625,
      "learning_rate": 2.453846153846154e-07,
      "loss": -0.0471,
      "reward": 1.6089182496070862,
      "reward_std": 0.26404254883527756,
      "reward_std_all": 0.745782732963562,
      "rewards-max/forensics_sinkhorn_reward": 0.9813347160816193,
      "rewards-max/forensics_strict_edge_reward": 0.6149742305278778,
      "rewards-max/forensics_tiou_reward": 0.9792823493480682,
      "rewards/forensics_sinkhorn_reward": 0.8374170660972595,
      "rewards/forensics_strict_edge_reward": 0.07574432529509068,
      "rewards/forensics_tiou_reward": 0.6957568526268005,
      "step": 982
    },
    {
      "completion_length": 26.296875,
      "epoch": 3.0153609831029184,
      "grad_norm": 12.064068794250488,
      "kl": 0.5634765625,
      "learning_rate": 2.446153846153846e-07,
      "loss": -0.0067,
      "reward": 1.7304467558860779,
      "reward_std": 0.5627752542495728,
      "reward_std_all": 0.8210688829421997,
      "rewards-max/forensics_sinkhorn_reward": 0.9894649982452393,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.8071898818016052,
      "rewards/forensics_strict_edge_reward": 0.22077827900648117,
      "rewards/forensics_tiou_reward": 0.7024786472320557,
      "step": 983
    },
    {
      "completion_length": 24.65625,
      "epoch": 3.0184331797235022,
      "grad_norm": 12.166314125061035,
      "kl": 0.4482421875,
      "learning_rate": 2.4384615384615386e-07,
      "loss": -0.1161,
      "reward": 1.4510239362716675,
      "reward_std": 0.46427127718925476,
      "reward_std_all": 0.9240673184394836,
      "rewards-max/forensics_sinkhorn_reward": 0.9871780872344971,
      "rewards-max/forensics_strict_edge_reward": 0.8704091012477875,
      "rewards-max/forensics_tiou_reward": 0.9941978752613068,
      "rewards/forensics_sinkhorn_reward": 0.7427334785461426,
      "rewards/forensics_strict_edge_reward": 0.1347135454416275,
      "rewards/forensics_tiou_reward": 0.5735768973827362,
      "step": 984
    },
    {
      "completion_length": 24.9375,
      "epoch": 3.021505376344086,
      "grad_norm": 10.8762788772583,
      "kl": 0.546875,
      "learning_rate": 2.4307692307692307e-07,
      "loss": -0.1028,
      "reward": 1.5282108783721924,
      "reward_std": 0.5192771404981613,
      "reward_std_all": 0.8844364285469055,
      "rewards-max/forensics_sinkhorn_reward": 0.98696368932724,
      "rewards-max/forensics_strict_edge_reward": 0.8351600170135498,
      "rewards-max/forensics_tiou_reward": 0.9902297258377075,
      "rewards/forensics_sinkhorn_reward": 0.7268125414848328,
      "rewards/forensics_strict_edge_reward": 0.14464543387293816,
      "rewards/forensics_tiou_reward": 0.6567528545856476,
      "step": 985
    },
    {
      "completion_length": 23.5625,
      "epoch": 3.02457757296467,
      "grad_norm": 11.845519065856934,
      "kl": 0.53125,
      "learning_rate": 2.423076923076923e-07,
      "loss": -0.1857,
      "reward": 1.5939117074012756,
      "reward_std": 0.421427845954895,
      "reward_std_all": 0.8313988447189331,
      "rewards-max/forensics_sinkhorn_reward": 0.9846880733966827,
      "rewards-max/forensics_strict_edge_reward": 0.9093653857707977,
      "rewards-max/forensics_tiou_reward": 0.9900074303150177,
      "rewards/forensics_sinkhorn_reward": 0.7996790707111359,
      "rewards/forensics_strict_edge_reward": 0.12381958961486816,
      "rewards/forensics_tiou_reward": 0.6704130470752716,
      "step": 986
    },
    {
      "completion_length": 25.15625,
      "epoch": 3.0276497695852536,
      "grad_norm": 10.358637809753418,
      "kl": 0.521484375,
      "learning_rate": 2.415384615384615e-07,
      "loss": 0.016,
      "reward": 1.5954223275184631,
      "reward_std": 0.30849967151880264,
      "reward_std_all": 0.765934407711029,
      "rewards-max/forensics_sinkhorn_reward": 0.9784908294677734,
      "rewards-max/forensics_strict_edge_reward": 0.6515953242778778,
      "rewards-max/forensics_tiou_reward": 0.9728094935417175,
      "rewards/forensics_sinkhorn_reward": 0.828742504119873,
      "rewards/forensics_strict_edge_reward": 0.09697289019823074,
      "rewards/forensics_tiou_reward": 0.6697068810462952,
      "step": 987
    },
    {
      "completion_length": 22.90625,
      "epoch": 3.0307219662058373,
      "grad_norm": 10.390329360961914,
      "kl": 0.5185546875,
      "learning_rate": 2.4076923076923076e-07,
      "loss": -0.1317,
      "reward": 1.4898880124092102,
      "reward_std": 0.3552200049161911,
      "reward_std_all": 0.8604251444339752,
      "rewards-max/forensics_sinkhorn_reward": 0.9856565296649933,
      "rewards-max/forensics_strict_edge_reward": 0.8582656681537628,
      "rewards-max/forensics_tiou_reward": 0.9920501708984375,
      "rewards/forensics_sinkhorn_reward": 0.7514427602291107,
      "rewards/forensics_strict_edge_reward": 0.1318584829568863,
      "rewards/forensics_tiou_reward": 0.606586754322052,
      "step": 988
    },
    {
      "completion_length": 28.34375,
      "epoch": 3.033794162826421,
      "grad_norm": 12.083654403686523,
      "kl": 0.5048828125,
      "learning_rate": 2.4e-07,
      "loss": 0.0575,
      "reward": 1.39901402592659,
      "reward_std": 0.32172146439552307,
      "reward_std_all": 0.8887996077537537,
      "rewards-max/forensics_sinkhorn_reward": 0.9813723862171173,
      "rewards-max/forensics_strict_edge_reward": 0.6839397251605988,
      "rewards-max/forensics_tiou_reward": 0.9794863164424896,
      "rewards/forensics_sinkhorn_reward": 0.7089761793613434,
      "rewards/forensics_strict_edge_reward": 0.14481921773403883,
      "rewards/forensics_tiou_reward": 0.5452186912298203,
      "step": 989
    },
    {
      "completion_length": 25.546875,
      "epoch": 3.0368663594470044,
      "grad_norm": 12.282196998596191,
      "kl": 0.5146484375,
      "learning_rate": 2.392307692307692e-07,
      "loss": -0.042,
      "reward": 1.5189430117607117,
      "reward_std": 0.42092905938625336,
      "reward_std_all": 0.858437716960907,
      "rewards-max/forensics_sinkhorn_reward": 0.9881559014320374,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.73755744099617,
      "rewards/forensics_strict_edge_reward": 0.1672845408320427,
      "rewards/forensics_tiou_reward": 0.6141011416912079,
      "step": 990
    },
    {
      "completion_length": 25.953125,
      "epoch": 3.0399385560675882,
      "grad_norm": 312.85369873046875,
      "kl": 1.93359375,
      "learning_rate": 2.3846153846153846e-07,
      "loss": 0.0421,
      "reward": 1.8480549454689026,
      "reward_std": 0.3886575996875763,
      "reward_std_all": 0.6893669664859772,
      "rewards-max/forensics_sinkhorn_reward": 0.988852471113205,
      "rewards-max/forensics_strict_edge_reward": 0.961620420217514,
      "rewards-max/forensics_tiou_reward": 0.9992469251155853,
      "rewards/forensics_sinkhorn_reward": 0.8373792171478271,
      "rewards/forensics_strict_edge_reward": 0.2278277426958084,
      "rewards/forensics_tiou_reward": 0.782848060131073,
      "step": 991
    },
    {
      "completion_length": 27.015625,
      "epoch": 3.043010752688172,
      "grad_norm": 12.098288536071777,
      "kl": 0.62890625,
      "learning_rate": 2.376923076923077e-07,
      "loss": 0.0147,
      "reward": 1.7692167162895203,
      "reward_std": 0.3525103032588959,
      "reward_std_all": 0.7176301777362823,
      "rewards-max/forensics_sinkhorn_reward": 0.9894075989723206,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.8678320050239563,
      "rewards/forensics_strict_edge_reward": 0.14734894782304764,
      "rewards/forensics_tiou_reward": 0.7540357112884521,
      "step": 992
    },
    {
      "completion_length": 26.96875,
      "epoch": 3.046082949308756,
      "grad_norm": 11.675614356994629,
      "kl": 0.5078125,
      "learning_rate": 2.369230769230769e-07,
      "loss": -0.1053,
      "reward": 1.5810956954956055,
      "reward_std": 0.2703644931316376,
      "reward_std_all": 0.7829284965991974,
      "rewards-max/forensics_sinkhorn_reward": 0.9862939715385437,
      "rewards-max/forensics_strict_edge_reward": 0.732279509305954,
      "rewards-max/forensics_tiou_reward": 0.9904446005821228,
      "rewards/forensics_sinkhorn_reward": 0.7639376223087311,
      "rewards/forensics_strict_edge_reward": 0.16159578040242195,
      "rewards/forensics_tiou_reward": 0.6555624008178711,
      "step": 993
    },
    {
      "completion_length": 29.703125,
      "epoch": 3.0491551459293396,
      "grad_norm": 10.902420997619629,
      "kl": 0.609375,
      "learning_rate": 2.3615384615384615e-07,
      "loss": -0.0427,
      "reward": 1.818495750427246,
      "reward_std": 0.6035994589328766,
      "reward_std_all": 0.7957069277763367,
      "rewards-max/forensics_sinkhorn_reward": 0.9900825917720795,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.8226660192012787,
      "rewards/forensics_strict_edge_reward": 0.26531165838241577,
      "rewards/forensics_tiou_reward": 0.7305180132389069,
      "step": 994
    },
    {
      "completion_length": 27.34375,
      "epoch": 3.0522273425499233,
      "grad_norm": 10.152816772460938,
      "kl": 0.44140625,
      "learning_rate": 2.3538461538461537e-07,
      "loss": 0.0295,
      "reward": 1.5612635612487793,
      "reward_std": 0.37687528133392334,
      "reward_std_all": 0.8258282542228699,
      "rewards-max/forensics_sinkhorn_reward": 0.9871278703212738,
      "rewards-max/forensics_strict_edge_reward": 0.9291328191757202,
      "rewards-max/forensics_tiou_reward": 0.9943117201328278,
      "rewards/forensics_sinkhorn_reward": 0.7382485866546631,
      "rewards/forensics_strict_edge_reward": 0.18909412622451782,
      "rewards/forensics_tiou_reward": 0.633920818567276,
      "step": 995
    },
    {
      "completion_length": 24.734375,
      "epoch": 3.055299539170507,
      "grad_norm": 11.020962715148926,
      "kl": 0.5166015625,
      "learning_rate": 2.346153846153846e-07,
      "loss": -0.0721,
      "reward": 1.450144112110138,
      "reward_std": 0.3852480947971344,
      "reward_std_all": 0.8631290793418884,
      "rewards-max/forensics_sinkhorn_reward": 0.9747289419174194,
      "rewards-max/forensics_strict_edge_reward": 0.638425350189209,
      "rewards-max/forensics_tiou_reward": 0.9653831124305725,
      "rewards/forensics_sinkhorn_reward": 0.7531355023384094,
      "rewards/forensics_strict_edge_reward": 0.09549817815423012,
      "rewards/forensics_tiou_reward": 0.6015104055404663,
      "step": 996
    },
    {
      "completion_length": 24.015625,
      "epoch": 3.0583717357910905,
      "grad_norm": 12.52525806427002,
      "kl": 0.5,
      "learning_rate": 2.3384615384615382e-07,
      "loss": -0.0521,
      "reward": 1.6021984219551086,
      "reward_std": 0.2827029228210449,
      "reward_std_all": 0.7641128599643707,
      "rewards-max/forensics_sinkhorn_reward": 0.9875648319721222,
      "rewards-max/forensics_strict_edge_reward": 0.7941735982894897,
      "rewards-max/forensics_tiou_reward": 0.9949154853820801,
      "rewards/forensics_sinkhorn_reward": 0.784641832113266,
      "rewards/forensics_strict_edge_reward": 0.1447458416223526,
      "rewards/forensics_tiou_reward": 0.6728107035160065,
      "step": 997
    },
    {
      "completion_length": 26.796875,
      "epoch": 3.0614439324116742,
      "grad_norm": 11.335254669189453,
      "kl": 0.5546875,
      "learning_rate": 2.3307692307692306e-07,
      "loss": -0.0216,
      "reward": 1.6033034324645996,
      "reward_std": 0.6344729363918304,
      "reward_std_all": 0.9213557839393616,
      "rewards-max/forensics_sinkhorn_reward": 0.9886893630027771,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.7359655499458313,
      "rewards/forensics_strict_edge_reward": 0.22154996544122696,
      "rewards/forensics_tiou_reward": 0.6457878947257996,
      "step": 998
    },
    {
      "completion_length": 25.203125,
      "epoch": 3.064516129032258,
      "grad_norm": 13.011497497558594,
      "kl": 0.568359375,
      "learning_rate": 2.323076923076923e-07,
      "loss": -0.0397,
      "reward": 1.7534470558166504,
      "reward_std": 0.501025602221489,
      "reward_std_all": 0.7931908965110779,
      "rewards-max/forensics_sinkhorn_reward": 0.9892019629478455,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.7898197770118713,
      "rewards/forensics_strict_edge_reward": 0.2493683323264122,
      "rewards/forensics_tiou_reward": 0.7142590880393982,
      "step": 999
    },
    {
      "completion_length": 25.515625,
      "epoch": 3.067588325652842,
      "grad_norm": 10.84512710571289,
      "kl": 0.521484375,
      "learning_rate": 2.3153846153846151e-07,
      "loss": -0.0996,
      "reward": 1.7093919515609741,
      "reward_std": 0.35156378149986267,
      "reward_std_all": 0.7442644834518433,
      "rewards-max/forensics_sinkhorn_reward": 0.9891113936901093,
      "rewards-max/forensics_strict_edge_reward": 0.9838767349720001,
      "rewards-max/forensics_tiou_reward": 0.999814510345459,
      "rewards/forensics_sinkhorn_reward": 0.8328766524791718,
      "rewards/forensics_strict_edge_reward": 0.14288374036550522,
      "rewards/forensics_tiou_reward": 0.7336315810680389,
      "step": 1000
    },
    {
      "completion_length": 24.0,
      "epoch": 3.0706605222734256,
      "grad_norm": 9.997551918029785,
      "kl": 0.4736328125,
      "learning_rate": 2.3076923076923078e-07,
      "loss": -0.0322,
      "reward": 1.6386688947677612,
      "reward_std": 0.3134399354457855,
      "reward_std_all": 0.7521745562553406,
      "rewards-max/forensics_sinkhorn_reward": 0.9819952547550201,
      "rewards-max/forensics_strict_edge_reward": 0.7482926547527313,
      "rewards-max/forensics_tiou_reward": 0.9827145338058472,
      "rewards/forensics_sinkhorn_reward": 0.8067750334739685,
      "rewards/forensics_strict_edge_reward": 0.14084192365407944,
      "rewards/forensics_tiou_reward": 0.6910519599914551,
      "step": 1001
    },
    {
      "completion_length": 24.59375,
      "epoch": 3.0737327188940093,
      "grad_norm": 10.298619270324707,
      "kl": 0.58203125,
      "learning_rate": 2.3e-07,
      "loss": -0.051,
      "reward": 1.7158992290496826,
      "reward_std": 0.4694213271141052,
      "reward_std_all": 0.788678914308548,
      "rewards-max/forensics_sinkhorn_reward": 0.9899584352970123,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.8169083893299103,
      "rewards/forensics_strict_edge_reward": 0.24795307964086533,
      "rewards/forensics_tiou_reward": 0.6510379016399384,
      "step": 1002
    },
    {
      "completion_length": 27.53125,
      "epoch": 3.076804915514593,
      "grad_norm": 12.036187171936035,
      "kl": 0.4375,
      "learning_rate": 2.2923076923076924e-07,
      "loss": -0.0783,
      "reward": 1.7301273941993713,
      "reward_std": 0.39487919211387634,
      "reward_std_all": 0.7578093409538269,
      "rewards-max/forensics_sinkhorn_reward": 0.9847967326641083,
      "rewards-max/forensics_strict_edge_reward": 0.8158992826938629,
      "rewards-max/forensics_tiou_reward": 0.987654834985733,
      "rewards/forensics_sinkhorn_reward": 0.8217032849788666,
      "rewards/forensics_strict_edge_reward": 0.14713601395487785,
      "rewards/forensics_tiou_reward": 0.7612881064414978,
      "step": 1003
    },
    {
      "completion_length": 24.6875,
      "epoch": 3.0798771121351765,
      "grad_norm": 12.9406099319458,
      "kl": 0.61328125,
      "learning_rate": 2.2846153846153845e-07,
      "loss": -0.06,
      "reward": 1.6593360304832458,
      "reward_std": 0.4078741669654846,
      "reward_std_all": 0.8294880092144012,
      "rewards-max/forensics_sinkhorn_reward": 0.9901976585388184,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.8120515942573547,
      "rewards/forensics_strict_edge_reward": 0.2009568139910698,
      "rewards/forensics_tiou_reward": 0.6463276743888855,
      "step": 1004
    },
    {
      "completion_length": 23.84375,
      "epoch": 3.0829493087557602,
      "grad_norm": 13.298952102661133,
      "kl": 0.580078125,
      "learning_rate": 2.276923076923077e-07,
      "loss": -0.0929,
      "reward": 1.656655192375183,
      "reward_std": 0.4436095356941223,
      "reward_std_all": 0.8017288744449615,
      "rewards-max/forensics_sinkhorn_reward": 0.9888993501663208,
      "rewards-max/forensics_strict_edge_reward": 0.9677534997463226,
      "rewards-max/forensics_tiou_reward": 0.9994173943996429,
      "rewards/forensics_sinkhorn_reward": 0.7751258909702301,
      "rewards/forensics_strict_edge_reward": 0.20556487143039703,
      "rewards/forensics_tiou_reward": 0.6759642958641052,
      "step": 1005
    },
    {
      "completion_length": 25.0,
      "epoch": 3.086021505376344,
      "grad_norm": 12.54017162322998,
      "kl": 0.4912109375,
      "learning_rate": 2.269230769230769e-07,
      "loss": -0.0449,
      "reward": 1.4739477634429932,
      "reward_std": 0.33842214941978455,
      "reward_std_all": 0.8388665616512299,
      "rewards-max/forensics_sinkhorn_reward": 0.9886941611766815,
      "rewards-max/forensics_strict_edge_reward": 0.882964164018631,
      "rewards-max/forensics_tiou_reward": 0.9989982843399048,
      "rewards/forensics_sinkhorn_reward": 0.7723254263401031,
      "rewards/forensics_strict_edge_reward": 0.09329776093363762,
      "rewards/forensics_tiou_reward": 0.6083245575428009,
      "step": 1006
    },
    {
      "completion_length": 27.5,
      "epoch": 3.089093701996928,
      "grad_norm": 11.573147773742676,
      "kl": 0.5732421875,
      "learning_rate": 2.2615384615384614e-07,
      "loss": -0.0709,
      "reward": 1.623126745223999,
      "reward_std": 0.4429045766592026,
      "reward_std_all": 0.8314562141895294,
      "rewards-max/forensics_sinkhorn_reward": 0.9837886095046997,
      "rewards-max/forensics_strict_edge_reward": 0.7711026966571808,
      "rewards-max/forensics_tiou_reward": 0.9848856329917908,
      "rewards/forensics_sinkhorn_reward": 0.8215300440788269,
      "rewards/forensics_strict_edge_reward": 0.1345730610191822,
      "rewards/forensics_tiou_reward": 0.6670236587524414,
      "step": 1007
    },
    {
      "completion_length": 23.71875,
      "epoch": 3.0921658986175116,
      "grad_norm": 18.448593139648438,
      "kl": 0.66796875,
      "learning_rate": 2.2538461538461536e-07,
      "loss": -0.0861,
      "reward": 1.6669142246246338,
      "reward_std": 0.416281059384346,
      "reward_std_all": 0.8014905154705048,
      "rewards-max/forensics_sinkhorn_reward": 0.9890819191932678,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.7992637753486633,
      "rewards/forensics_strict_edge_reward": 0.1805676594376564,
      "rewards/forensics_tiou_reward": 0.687082827091217,
      "step": 1008
    },
    {
      "completion_length": 25.03125,
      "epoch": 3.0952380952380953,
      "grad_norm": 12.997706413269043,
      "kl": 0.4619140625,
      "learning_rate": 2.246153846153846e-07,
      "loss": -0.123,
      "reward": 1.6618786454200745,
      "reward_std": 0.37481196224689484,
      "reward_std_all": 0.778412938117981,
      "rewards-max/forensics_sinkhorn_reward": 0.9894293546676636,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.8123503625392914,
      "rewards/forensics_strict_edge_reward": 0.1698625609278679,
      "rewards/forensics_tiou_reward": 0.6796656847000122,
      "step": 1009
    },
    {
      "completion_length": 25.03125,
      "epoch": 3.098310291858679,
      "grad_norm": 10.096071243286133,
      "kl": 0.53125,
      "learning_rate": 2.238461538461538e-07,
      "loss": -0.0679,
      "reward": 1.7255097031593323,
      "reward_std": 0.30831289291381836,
      "reward_std_all": 0.7225699424743652,
      "rewards-max/forensics_sinkhorn_reward": 0.9817095696926117,
      "rewards-max/forensics_strict_edge_reward": 0.756708562374115,
      "rewards-max/forensics_tiou_reward": 0.9805322885513306,
      "rewards/forensics_sinkhorn_reward": 0.861143946647644,
      "rewards/forensics_strict_edge_reward": 0.1260107085108757,
      "rewards/forensics_tiou_reward": 0.7383549511432648,
      "step": 1010
    },
    {
      "completion_length": 29.0625,
      "epoch": 3.1013824884792625,
      "grad_norm": 35.256858825683594,
      "kl": 0.833984375,
      "learning_rate": 2.2307692307692308e-07,
      "loss": 0.0484,
      "reward": 1.5950958132743835,
      "reward_std": 0.32132117450237274,
      "reward_std_all": 0.7856907248497009,
      "rewards-max/forensics_sinkhorn_reward": 0.9873980283737183,
      "rewards-max/forensics_strict_edge_reward": 0.8333333432674408,
      "rewards-max/forensics_tiou_reward": 0.9931516349315643,
      "rewards/forensics_sinkhorn_reward": 0.7682320475578308,
      "rewards/forensics_strict_edge_reward": 0.14671118557453156,
      "rewards/forensics_tiou_reward": 0.6801525354385376,
      "step": 1011
    },
    {
      "completion_length": 22.828125,
      "epoch": 3.1044546850998462,
      "grad_norm": 11.102910995483398,
      "kl": 0.5390625,
      "learning_rate": 2.2230769230769232e-07,
      "loss": -0.1063,
      "reward": 1.6139187812805176,
      "reward_std": 0.502347856760025,
      "reward_std_all": 0.8501689434051514,
      "rewards-max/forensics_sinkhorn_reward": 0.9890260398387909,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.8005035817623138,
      "rewards/forensics_strict_edge_reward": 0.15975820273160934,
      "rewards/forensics_tiou_reward": 0.653656929731369,
      "step": 1012
    },
    {
      "completion_length": 25.296875,
      "epoch": 3.10752688172043,
      "grad_norm": 11.724505424499512,
      "kl": 0.5244140625,
      "learning_rate": 2.2153846153846153e-07,
      "loss": -0.1148,
      "reward": 1.7089510560035706,
      "reward_std": 0.4020850658416748,
      "reward_std_all": 0.7863880395889282,
      "rewards-max/forensics_sinkhorn_reward": 0.9892823994159698,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.816346138715744,
      "rewards/forensics_strict_edge_reward": 0.21037689596414566,
      "rewards/forensics_tiou_reward": 0.6822279989719391,
      "step": 1013
    },
    {
      "completion_length": 27.109375,
      "epoch": 3.110599078341014,
      "grad_norm": 19.61217498779297,
      "kl": 1.12890625,
      "learning_rate": 2.2076923076923077e-07,
      "loss": 0.1406,
      "reward": 1.9513704776763916,
      "reward_std": 0.4491805285215378,
      "reward_std_all": 0.691053032875061,
      "rewards-max/forensics_sinkhorn_reward": 0.9902575314044952,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.8458980023860931,
      "rewards/forensics_strict_edge_reward": 0.33679406344890594,
      "rewards/forensics_tiou_reward": 0.7686783671379089,
      "step": 1014
    },
    {
      "completion_length": 26.109375,
      "epoch": 3.1136712749615976,
      "grad_norm": 14.37463665008545,
      "kl": 0.5966796875,
      "learning_rate": 2.1999999999999998e-07,
      "loss": -0.0016,
      "reward": 1.9177038669586182,
      "reward_std": 0.29294073581695557,
      "reward_std_all": 0.6270566880702972,
      "rewards-max/forensics_sinkhorn_reward": 0.9895638525485992,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.8955664038658142,
      "rewards/forensics_strict_edge_reward": 0.2070995569229126,
      "rewards/forensics_tiou_reward": 0.8150378167629242,
      "step": 1015
    },
    {
      "completion_length": 26.1875,
      "epoch": 3.1167434715821813,
      "grad_norm": 14.712079048156738,
      "kl": 0.486328125,
      "learning_rate": 2.1923076923076922e-07,
      "loss": 0.0243,
      "reward": 1.7771643996238708,
      "reward_std": 0.28852303326129913,
      "reward_std_all": 0.6941796839237213,
      "rewards-max/forensics_sinkhorn_reward": 0.9873685836791992,
      "rewards-max/forensics_strict_edge_reward": 0.8582656681537628,
      "rewards-max/forensics_tiou_reward": 0.9943373501300812,
      "rewards/forensics_sinkhorn_reward": 0.8060016334056854,
      "rewards/forensics_strict_edge_reward": 0.2395711913704872,
      "rewards/forensics_tiou_reward": 0.7315916121006012,
      "step": 1016
    },
    {
      "completion_length": 26.328125,
      "epoch": 3.119815668202765,
      "grad_norm": 12.18547534942627,
      "kl": 0.615234375,
      "learning_rate": 2.1846153846153844e-07,
      "loss": -0.0721,
      "reward": 1.7472845911979675,
      "reward_std": 0.35409919917583466,
      "reward_std_all": 0.731029599905014,
      "rewards-max/forensics_sinkhorn_reward": 0.9889281988143921,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.8405151665210724,
      "rewards/forensics_strict_edge_reward": 0.19218869507312775,
      "rewards/forensics_tiou_reward": 0.7145808637142181,
      "step": 1017
    },
    {
      "completion_length": 24.890625,
      "epoch": 3.122887864823349,
      "grad_norm": 10.525472640991211,
      "kl": 0.537109375,
      "learning_rate": 2.1769230769230768e-07,
      "loss": -0.0955,
      "reward": 1.7283130884170532,
      "reward_std": 0.334855318069458,
      "reward_std_all": 0.7435813248157501,
      "rewards-max/forensics_sinkhorn_reward": 0.9887494444847107,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.833705484867096,
      "rewards/forensics_strict_edge_reward": 0.197307251393795,
      "rewards/forensics_tiou_reward": 0.6973004639148712,
      "step": 1018
    },
    {
      "completion_length": 26.890625,
      "epoch": 3.1259600614439322,
      "grad_norm": 11.092370986938477,
      "kl": 0.576171875,
      "learning_rate": 2.169230769230769e-07,
      "loss": 0.0117,
      "reward": 1.711892545223236,
      "reward_std": 0.3219616413116455,
      "reward_std_all": 0.731757402420044,
      "rewards-max/forensics_sinkhorn_reward": 0.974570244550705,
      "rewards-max/forensics_strict_edge_reward": 0.7246644794940948,
      "rewards-max/forensics_tiou_reward": 0.9631107747554779,
      "rewards/forensics_sinkhorn_reward": 0.8242174983024597,
      "rewards/forensics_strict_edge_reward": 0.19455331563949585,
      "rewards/forensics_tiou_reward": 0.6931218504905701,
      "step": 1019
    },
    {
      "completion_length": 27.875,
      "epoch": 3.129032258064516,
      "grad_norm": 18.03257942199707,
      "kl": 0.89453125,
      "learning_rate": 2.1615384615384613e-07,
      "loss": -0.0177,
      "reward": 1.9000827074050903,
      "reward_std": 0.3976361006498337,
      "reward_std_all": 0.6751872301101685,
      "rewards-max/forensics_sinkhorn_reward": 0.990240603685379,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.8535890579223633,
      "rewards/forensics_strict_edge_reward": 0.2661627382040024,
      "rewards/forensics_tiou_reward": 0.7803309559822083,
      "step": 1020
    },
    {
      "completion_length": 25.84375,
      "epoch": 3.1321044546851,
      "grad_norm": 12.288508415222168,
      "kl": 0.537109375,
      "learning_rate": 2.153846153846154e-07,
      "loss": -0.0342,
      "reward": 1.4641523957252502,
      "reward_std": 0.2791328579187393,
      "reward_std_all": 0.8311580717563629,
      "rewards-max/forensics_sinkhorn_reward": 0.98783740401268,
      "rewards-max/forensics_strict_edge_reward": 0.8351600170135498,
      "rewards-max/forensics_tiou_reward": 0.9975709319114685,
      "rewards/forensics_sinkhorn_reward": 0.7714434266090393,
      "rewards/forensics_strict_edge_reward": 0.08911053091287613,
      "rewards/forensics_tiou_reward": 0.6035984754562378,
      "step": 1021
    },
    {
      "completion_length": 26.125,
      "epoch": 3.1351766513056836,
      "grad_norm": 11.640196800231934,
      "kl": 0.556640625,
      "learning_rate": 2.146153846153846e-07,
      "loss": -0.0965,
      "reward": 1.511767566204071,
      "reward_std": 0.34416329860687256,
      "reward_std_all": 0.8372156322002411,
      "rewards-max/forensics_sinkhorn_reward": 0.9799900949001312,
      "rewards-max/forensics_strict_edge_reward": 0.8135445415973663,
      "rewards-max/forensics_tiou_reward": 0.9778674840927124,
      "rewards/forensics_sinkhorn_reward": 0.7311605513095856,
      "rewards/forensics_strict_edge_reward": 0.14308832213282585,
      "rewards/forensics_tiou_reward": 0.6375187635421753,
      "step": 1022
    },
    {
      "completion_length": 29.71875,
      "epoch": 3.1382488479262673,
      "grad_norm": 15.48478889465332,
      "kl": 1.087890625,
      "learning_rate": 2.1384615384615385e-07,
      "loss": -0.0716,
      "reward": 1.684964895248413,
      "reward_std": 0.45609116554260254,
      "reward_std_all": 0.8104673027992249,
      "rewards-max/forensics_sinkhorn_reward": 0.9887222647666931,
      "rewards-max/forensics_strict_edge_reward": 0.9232408702373505,
      "rewards-max/forensics_tiou_reward": 0.9965264201164246,
      "rewards/forensics_sinkhorn_reward": 0.8006785213947296,
      "rewards/forensics_strict_edge_reward": 0.1994970738887787,
      "rewards/forensics_tiou_reward": 0.68478924036026,
      "step": 1023
    },
    {
      "completion_length": 24.421875,
      "epoch": 3.141321044546851,
      "grad_norm": 12.053864479064941,
      "kl": 0.7109375,
      "learning_rate": 2.1307692307692307e-07,
      "loss": 0.051,
      "reward": 1.4773799777030945,
      "reward_std": 0.4050625413656235,
      "reward_std_all": 0.8678713440895081,
      "rewards-max/forensics_sinkhorn_reward": 0.9827206134796143,
      "rewards-max/forensics_strict_edge_reward": 0.7933230996131897,
      "rewards-max/forensics_tiou_reward": 0.9851051568984985,
      "rewards/forensics_sinkhorn_reward": 0.7542510628700256,
      "rewards/forensics_strict_edge_reward": 0.11554304510354996,
      "rewards/forensics_tiou_reward": 0.607585996389389,
      "step": 1024
    },
    {
      "completion_length": 21.5625,
      "epoch": 3.144393241167435,
      "grad_norm": 13.95728588104248,
      "kl": 0.58203125,
      "learning_rate": 2.123076923076923e-07,
      "loss": -0.0604,
      "reward": 1.5324665904045105,
      "reward_std": 0.450211763381958,
      "reward_std_all": 0.8568673729896545,
      "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": 0.7898440957069397,
      "rewards/forensics_strict_edge_reward": 0.10949227958917618,
      "rewards/forensics_tiou_reward": 0.63313028216362,
      "step": 1025
    },
    {
      "completion_length": 27.578125,
      "epoch": 3.1474654377880182,
      "grad_norm": 13.604779243469238,
      "kl": 0.5078125,
      "learning_rate": 2.1153846153846152e-07,
      "loss": -0.1127,
      "reward": 1.678872287273407,
      "reward_std": 0.3891236186027527,
      "reward_std_all": 0.7974623143672943,
      "rewards-max/forensics_sinkhorn_reward": 0.9895057082176208,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.8044705092906952,
      "rewards/forensics_strict_edge_reward": 0.2114184945821762,
      "rewards/forensics_tiou_reward": 0.6629831492900848,
      "step": 1026
    },
    {
      "completion_length": 27.265625,
      "epoch": 3.150537634408602,
      "grad_norm": 11.254866600036621,
      "kl": 0.56640625,
      "learning_rate": 2.1076923076923076e-07,
      "loss": 0.0096,
      "reward": 1.7183068990707397,
      "reward_std": 0.3531441241502762,
      "reward_std_all": 0.7670765519142151,
      "rewards-max/forensics_sinkhorn_reward": 0.9898141324520111,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.803020179271698,
      "rewards/forensics_strict_edge_reward": 0.21321073174476624,
      "rewards/forensics_tiou_reward": 0.7020760178565979,
      "step": 1027
    },
    {
      "completion_length": 22.859375,
      "epoch": 3.153609831029186,
      "grad_norm": 9.579582214355469,
      "kl": 0.5537109375,
      "learning_rate": 2.0999999999999997e-07,
      "loss": -0.0538,
      "reward": 1.503037929534912,
      "reward_std": 0.4885643869638443,
      "reward_std_all": 0.8764985501766205,
      "rewards-max/forensics_sinkhorn_reward": 0.987040787935257,
      "rewards-max/forensics_strict_edge_reward": 0.8053485751152039,
      "rewards-max/forensics_tiou_reward": 0.9912160336971283,
      "rewards/forensics_sinkhorn_reward": 0.7745863497257233,
      "rewards/forensics_strict_edge_reward": 0.08716888353228569,
      "rewards/forensics_tiou_reward": 0.6412827372550964,
      "step": 1028
    },
    {
      "completion_length": 26.359375,
      "epoch": 3.1566820276497696,
      "grad_norm": 12.97392749786377,
      "kl": 0.5546875,
      "learning_rate": 2.0923076923076921e-07,
      "loss": -0.0289,
      "reward": 1.6214348673820496,
      "reward_std": 0.3976258635520935,
      "reward_std_all": 0.8084445595741272,
      "rewards-max/forensics_sinkhorn_reward": 0.9892774522304535,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.7796821296215057,
      "rewards/forensics_strict_edge_reward": 0.18813197314739227,
      "rewards/forensics_tiou_reward": 0.6536208391189575,
      "step": 1029
    },
    {
      "completion_length": 24.453125,
      "epoch": 3.1597542242703534,
      "grad_norm": 14.889305114746094,
      "kl": 0.4921875,
      "learning_rate": 2.0846153846153843e-07,
      "loss": -0.082,
      "reward": 1.6582555770874023,
      "reward_std": 0.459768682718277,
      "reward_std_all": 0.8182039856910706,
      "rewards-max/forensics_sinkhorn_reward": 0.9870530962944031,
      "rewards-max/forensics_strict_edge_reward": 0.9291328191757202,
      "rewards-max/forensics_tiou_reward": 0.9952380955219269,
      "rewards/forensics_sinkhorn_reward": 0.8112359941005707,
      "rewards/forensics_strict_edge_reward": 0.17059121280908585,
      "rewards/forensics_tiou_reward": 0.6764284372329712,
      "step": 1030
    },
    {
      "completion_length": 25.96875,
      "epoch": 3.162826420890937,
      "grad_norm": 9.724201202392578,
      "kl": 0.5703125,
      "learning_rate": 2.076923076923077e-07,
      "loss": -0.0147,
      "reward": 1.7722105383872986,
      "reward_std": 0.49154865741729736,
      "reward_std_all": 0.772693544626236,
      "rewards-max/forensics_sinkhorn_reward": 0.9891714453697205,
      "rewards-max/forensics_strict_edge_reward": 0.8959447741508484,
      "rewards-max/forensics_tiou_reward": 0.9987995624542236,
      "rewards/forensics_sinkhorn_reward": 0.8329257369041443,
      "rewards/forensics_strict_edge_reward": 0.19867950677871704,
      "rewards/forensics_tiou_reward": 0.7406053245067596,
      "step": 1031
    },
    {
      "completion_length": 27.796875,
      "epoch": 3.165898617511521,
      "grad_norm": 12.56550121307373,
      "kl": 0.720703125,
      "learning_rate": 2.0692307692307694e-07,
      "loss": -0.1285,
      "reward": 1.7376102209091187,
      "reward_std": 0.5601325631141663,
      "reward_std_all": 0.8205634951591492,
      "rewards-max/forensics_sinkhorn_reward": 0.9840421676635742,
      "rewards-max/forensics_strict_edge_reward": 0.7573813796043396,
      "rewards-max/forensics_tiou_reward": 0.9844386577606201,
      "rewards/forensics_sinkhorn_reward": 0.805852860212326,
      "rewards/forensics_strict_edge_reward": 0.23359078913927078,
      "rewards/forensics_tiou_reward": 0.6981664896011353,
      "step": 1032
    },
    {
      "completion_length": 24.65625,
      "epoch": 3.1689708141321047,
      "grad_norm": 11.95958423614502,
      "kl": 0.62109375,
      "learning_rate": 2.0615384615384615e-07,
      "loss": -0.0979,
      "reward": 1.613237977027893,
      "reward_std": 0.3225848525762558,
      "reward_std_all": 0.7907466292381287,
      "rewards-max/forensics_sinkhorn_reward": 0.9828616082668304,
      "rewards-max/forensics_strict_edge_reward": 0.8582656681537628,
      "rewards-max/forensics_tiou_reward": 0.9864159822463989,
      "rewards/forensics_sinkhorn_reward": 0.7998388111591339,
      "rewards/forensics_strict_edge_reward": 0.13729145750403404,
      "rewards/forensics_tiou_reward": 0.6761077642440796,
      "step": 1033
    },
    {
      "completion_length": 27.640625,
      "epoch": 3.172043010752688,
      "grad_norm": 12.201275825500488,
      "kl": 0.6015625,
      "learning_rate": 2.053846153846154e-07,
      "loss": 0.0047,
      "reward": 1.5844785571098328,
      "reward_std": 0.3390810936689377,
      "reward_std_all": 0.8221209645271301,
      "rewards-max/forensics_sinkhorn_reward": 0.9895982444286346,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.7370179295539856,
      "rewards/forensics_strict_edge_reward": 0.22448408603668213,
      "rewards/forensics_tiou_reward": 0.6229766309261322,
      "step": 1034
    },
    {
      "completion_length": 23.65625,
      "epoch": 3.175115207373272,
      "grad_norm": 11.220330238342285,
      "kl": 0.521484375,
      "learning_rate": 2.046153846153846e-07,
      "loss": 0.0009,
      "reward": 1.626902163028717,
      "reward_std": 0.3726799786090851,
      "reward_std_all": 0.7972806096076965,
      "rewards-max/forensics_sinkhorn_reward": 0.9842871427536011,
      "rewards-max/forensics_strict_edge_reward": 0.6901743710041046,
      "rewards-max/forensics_tiou_reward": 0.9859182238578796,
      "rewards/forensics_sinkhorn_reward": 0.823779284954071,
      "rewards/forensics_strict_edge_reward": 0.11838499270379543,
      "rewards/forensics_tiou_reward": 0.6847378611564636,
      "step": 1035
    },
    {
      "completion_length": 27.59375,
      "epoch": 3.1781874039938556,
      "grad_norm": 12.131528854370117,
      "kl": 0.658203125,
      "learning_rate": 2.0384615384615384e-07,
      "loss": -0.0328,
      "reward": 1.4690526127815247,
      "reward_std": 0.5362996757030487,
      "reward_std_all": 0.9430588185787201,
      "rewards-max/forensics_sinkhorn_reward": 0.9899190962314606,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.6766961216926575,
      "rewards/forensics_strict_edge_reward": 0.19555359333753586,
      "rewards/forensics_tiou_reward": 0.5968028604984283,
      "step": 1036
    },
    {
      "completion_length": 31.15625,
      "epoch": 3.1812596006144394,
      "grad_norm": 20.842212677001953,
      "kl": 0.767578125,
      "learning_rate": 2.0307692307692306e-07,
      "loss": -0.0593,
      "reward": 1.6614791750907898,
      "reward_std": 0.3661297559738159,
      "reward_std_all": 0.7807821929454803,
      "rewards-max/forensics_sinkhorn_reward": 0.9881750047206879,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.8173260986804962,
      "rewards/forensics_strict_edge_reward": 0.1728850156068802,
      "rewards/forensics_tiou_reward": 0.6712680459022522,
      "step": 1037
    },
    {
      "completion_length": 30.78125,
      "epoch": 3.184331797235023,
      "grad_norm": 12.793724060058594,
      "kl": 0.4765625,
      "learning_rate": 2.023076923076923e-07,
      "loss": -0.1173,
      "reward": 1.7098649740219116,
      "reward_std": 0.46972620487213135,
      "reward_std_all": 0.820490151643753,
      "rewards-max/forensics_sinkhorn_reward": 0.9900984168052673,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.8022878766059875,
      "rewards/forensics_strict_edge_reward": 0.2082846462726593,
      "rewards/forensics_tiou_reward": 0.6992924511432648,
      "step": 1038
    },
    {
      "completion_length": 26.390625,
      "epoch": 3.187403993855607,
      "grad_norm": 12.85200309753418,
      "kl": 0.53125,
      "learning_rate": 2.015384615384615e-07,
      "loss": -0.0104,
      "reward": 1.6685866117477417,
      "reward_std": 0.32772742211818695,
      "reward_std_all": 0.7606984078884125,
      "rewards-max/forensics_sinkhorn_reward": 0.9891153872013092,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.7976677417755127,
      "rewards/forensics_strict_edge_reward": 0.19887740910053253,
      "rewards/forensics_tiou_reward": 0.6720414757728577,
      "step": 1039
    },
    {
      "completion_length": 27.875,
      "epoch": 3.1904761904761907,
      "grad_norm": 12.7554292678833,
      "kl": 0.591796875,
      "learning_rate": 2.0076923076923075e-07,
      "loss": -0.0325,
      "reward": 1.69658762216568,
      "reward_std": 0.44674816727638245,
      "reward_std_all": 0.7977966368198395,
      "rewards-max/forensics_sinkhorn_reward": 0.9879294633865356,
      "rewards-max/forensics_strict_edge_reward": 0.8158992826938629,
      "rewards-max/forensics_tiou_reward": 0.9952642619609833,
      "rewards/forensics_sinkhorn_reward": 0.8126913011074066,
      "rewards/forensics_strict_edge_reward": 0.21071944385766983,
      "rewards/forensics_tiou_reward": 0.6731767952442169,
      "step": 1040
    },
    {
      "completion_length": 27.046875,
      "epoch": 3.193548387096774,
      "grad_norm": 11.242217063903809,
      "kl": 0.4990234375,
      "learning_rate": 2e-07,
      "loss": -0.0429,
      "reward": 1.725338339805603,
      "reward_std": 0.370526522397995,
      "reward_std_all": 0.7501966059207916,
      "rewards-max/forensics_sinkhorn_reward": 0.9882819056510925,
      "rewards-max/forensics_strict_edge_reward": 0.9464867115020752,
      "rewards-max/forensics_tiou_reward": 0.9984776079654694,
      "rewards/forensics_sinkhorn_reward": 0.8128790259361267,
      "rewards/forensics_strict_edge_reward": 0.18411476910114288,
      "rewards/forensics_tiou_reward": 0.7283445596694946,
      "step": 1041
    },
    {
      "completion_length": 27.578125,
      "epoch": 3.196620583717358,
      "grad_norm": 13.497987747192383,
      "kl": 0.703125,
      "learning_rate": 1.9923076923076923e-07,
      "loss": -0.0297,
      "reward": 1.6674504280090332,
      "reward_std": 0.3179149180650711,
      "reward_std_all": 0.7857360243797302,
      "rewards-max/forensics_sinkhorn_reward": 0.9894404709339142,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.7564733922481537,
      "rewards/forensics_strict_edge_reward": 0.2299266718327999,
      "rewards/forensics_tiou_reward": 0.6810502707958221,
      "step": 1042
    },
    {
      "completion_length": 22.84375,
      "epoch": 3.1996927803379416,
      "grad_norm": 9.691052436828613,
      "kl": 0.4873046875,
      "learning_rate": 1.9846153846153844e-07,
      "loss": -0.0965,
      "reward": 1.7893425822257996,
      "reward_std": 0.47605428099632263,
      "reward_std_all": 0.7680371403694153,
      "rewards-max/forensics_sinkhorn_reward": 0.9898779690265656,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.8418980538845062,
      "rewards/forensics_strict_edge_reward": 0.2512572631239891,
      "rewards/forensics_tiou_reward": 0.6961872279644012,
      "step": 1043
    },
    {
      "completion_length": 30.390625,
      "epoch": 3.2027649769585254,
      "grad_norm": 15.864521026611328,
      "kl": 0.4736328125,
      "learning_rate": 1.9769230769230768e-07,
      "loss": -0.0299,
      "reward": 1.8303226232528687,
      "reward_std": 0.5140136778354645,
      "reward_std_all": 0.7673780918121338,
      "rewards-max/forensics_sinkhorn_reward": 0.9895195364952087,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.8202800154685974,
      "rewards/forensics_strict_edge_reward": 0.27899184823036194,
      "rewards/forensics_tiou_reward": 0.7310506701469421,
      "step": 1044
    },
    {
      "completion_length": 23.296875,
      "epoch": 3.205837173579109,
      "grad_norm": 14.298924446105957,
      "kl": 0.71484375,
      "learning_rate": 1.9692307692307693e-07,
      "loss": -0.111,
      "reward": 1.5801716446876526,
      "reward_std": 0.4486994594335556,
      "reward_std_all": 0.8652571141719818,
      "rewards-max/forensics_sinkhorn_reward": 0.9895395338535309,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.703070729970932,
      "rewards/forensics_strict_edge_reward": 0.2436259463429451,
      "rewards/forensics_tiou_reward": 0.6334749460220337,
      "step": 1045
    },
    {
      "completion_length": 26.4375,
      "epoch": 3.208909370199693,
      "grad_norm": 13.230206489562988,
      "kl": 0.625,
      "learning_rate": 1.9615384615384614e-07,
      "loss": -0.1051,
      "reward": 1.5917739868164062,
      "reward_std": 0.40980902314186096,
      "reward_std_all": 0.806820422410965,
      "rewards-max/forensics_sinkhorn_reward": 0.9846263527870178,
      "rewards-max/forensics_strict_edge_reward": 0.7511901557445526,
      "rewards-max/forensics_tiou_reward": 0.9880677461624146,
      "rewards/forensics_sinkhorn_reward": 0.7742424309253693,
      "rewards/forensics_strict_edge_reward": 0.15340612456202507,
      "rewards/forensics_tiou_reward": 0.6641254127025604,
      "step": 1046
    },
    {
      "completion_length": 23.703125,
      "epoch": 3.2119815668202767,
      "grad_norm": 12.852140426635742,
      "kl": 0.65234375,
      "learning_rate": 1.9538461538461538e-07,
      "loss": -0.1395,
      "reward": 1.7511003613471985,
      "reward_std": 0.4039180278778076,
      "reward_std_all": 0.7495550811290741,
      "rewards-max/forensics_sinkhorn_reward": 0.9897510409355164,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.8290910720825195,
      "rewards/forensics_strict_edge_reward": 0.2239065319299698,
      "rewards/forensics_tiou_reward": 0.6981028020381927,
      "step": 1047
    },
    {
      "completion_length": 23.09375,
      "epoch": 3.21505376344086,
      "grad_norm": 9.976764678955078,
      "kl": 0.59375,
      "learning_rate": 1.946153846153846e-07,
      "loss": -0.1159,
      "reward": 1.7716703414916992,
      "reward_std": 0.41763049364089966,
      "reward_std_all": 0.743031233549118,
      "rewards-max/forensics_sinkhorn_reward": 0.9857252538204193,
      "rewards-max/forensics_strict_edge_reward": 0.882964164018631,
      "rewards-max/forensics_tiou_reward": 0.9917255640029907,
      "rewards/forensics_sinkhorn_reward": 0.8470067381858826,
      "rewards/forensics_strict_edge_reward": 0.19914600998163223,
      "rewards/forensics_tiou_reward": 0.725517600774765,
      "step": 1048
    },
    {
      "completion_length": 30.015625,
      "epoch": 3.218125960061444,
      "grad_norm": 10.973484992980957,
      "kl": 0.4912109375,
      "learning_rate": 1.9384615384615383e-07,
      "loss": -0.0024,
      "reward": 1.5766738057136536,
      "reward_std": 0.3990975022315979,
      "reward_std_all": 0.8302792608737946,
      "rewards-max/forensics_sinkhorn_reward": 0.986581027507782,
      "rewards-max/forensics_strict_edge_reward": 0.7589184939861298,
      "rewards-max/forensics_tiou_reward": 0.9911186397075653,
      "rewards/forensics_sinkhorn_reward": 0.7542430460453033,
      "rewards/forensics_strict_edge_reward": 0.1917431503534317,
      "rewards/forensics_tiou_reward": 0.6306875795125961,
      "step": 1049
    },
    {
      "completion_length": 27.921875,
      "epoch": 3.2211981566820276,
      "grad_norm": 13.08495044708252,
      "kl": 0.4921875,
      "learning_rate": 1.9307692307692307e-07,
      "loss": -0.0358,
      "reward": 1.9686921834945679,
      "reward_std": 0.3418513685464859,
      "reward_std_all": 0.6138782501220703,
      "rewards-max/forensics_sinkhorn_reward": 0.9901119470596313,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.8879197537899017,
      "rewards/forensics_strict_edge_reward": 0.27438129484653473,
      "rewards/forensics_tiou_reward": 0.8063910603523254,
      "step": 1050
    },
    {
      "completion_length": 24.828125,
      "epoch": 3.2242703533026114,
      "grad_norm": 11.381475448608398,
      "kl": 0.607421875,
      "learning_rate": 1.9230769230769231e-07,
      "loss": -0.0481,
      "reward": 1.6820580959320068,
      "reward_std": 0.48916955292224884,
      "reward_std_all": 0.8144723773002625,
      "rewards-max/forensics_sinkhorn_reward": 0.9863258898258209,
      "rewards-max/forensics_strict_edge_reward": 0.861880362033844,
      "rewards-max/forensics_tiou_reward": 0.9930087029933929,
      "rewards/forensics_sinkhorn_reward": 0.8126700818538666,
      "rewards/forensics_strict_edge_reward": 0.1687760166823864,
      "rewards/forensics_tiou_reward": 0.7006120681762695,
      "step": 1051
    },
    {
      "completion_length": 26.0625,
      "epoch": 3.227342549923195,
      "grad_norm": 14.582388877868652,
      "kl": 0.52734375,
      "learning_rate": 1.9153846153846153e-07,
      "loss": -0.0751,
      "reward": 1.664148986339569,
      "reward_std": 0.3598189055919647,
      "reward_std_all": 0.7678360939025879,
      "rewards-max/forensics_sinkhorn_reward": 0.9878577291965485,
      "rewards-max/forensics_strict_edge_reward": 0.8333333432674408,
      "rewards-max/forensics_tiou_reward": 0.9959357380867004,
      "rewards/forensics_sinkhorn_reward": 0.8196987807750702,
      "rewards/forensics_strict_edge_reward": 0.13708391413092613,
      "rewards/forensics_tiou_reward": 0.7073663175106049,
      "step": 1052
    },
    {
      "completion_length": 28.296875,
      "epoch": 3.230414746543779,
      "grad_norm": 12.335739135742188,
      "kl": 0.595703125,
      "learning_rate": 1.9076923076923077e-07,
      "loss": -0.0744,
      "reward": 1.6606032252311707,
      "reward_std": 0.36899157613515854,
      "reward_std_all": 0.8031118512153625,
      "rewards-max/forensics_sinkhorn_reward": 0.9884269833564758,
      "rewards-max/forensics_strict_edge_reward": 0.7759142816066742,
      "rewards-max/forensics_tiou_reward": 0.9954257011413574,
      "rewards/forensics_sinkhorn_reward": 0.7909288108348846,
      "rewards/forensics_strict_edge_reward": 0.20903047546744347,
      "rewards/forensics_tiou_reward": 0.6606439650058746,
      "step": 1053
    },
    {
      "completion_length": 24.640625,
      "epoch": 3.2334869431643627,
      "grad_norm": 14.182327270507812,
      "kl": 0.7236328125,
      "learning_rate": 1.8999999999999998e-07,
      "loss": 0.0183,
      "reward": 1.8370446562767029,
      "reward_std": 0.33987492322921753,
      "reward_std_all": 0.711394339799881,
      "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": 0.8613974750041962,
      "rewards/forensics_strict_edge_reward": 0.2314462661743164,
      "rewards/forensics_tiou_reward": 0.7442007660865784,
      "step": 1054
    },
    {
      "completion_length": 25.25,
      "epoch": 3.236559139784946,
      "grad_norm": 9.633806228637695,
      "kl": 0.5234375,
      "learning_rate": 1.8923076923076922e-07,
      "loss": -0.0167,
      "reward": 1.5591762065887451,
      "reward_std": 0.47224459052085876,
      "reward_std_all": 0.8680641651153564,
      "rewards-max/forensics_sinkhorn_reward": 0.9889281988143921,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.7541206777095795,
      "rewards/forensics_strict_edge_reward": 0.184137225151062,
      "rewards/forensics_tiou_reward": 0.6209183633327484,
      "step": 1055
    },
    {
      "completion_length": 26.390625,
      "epoch": 3.23963133640553,
      "grad_norm": 10.540745735168457,
      "kl": 0.4375,
      "learning_rate": 1.8846153846153846e-07,
      "loss": -0.1314,
      "reward": 1.5765741467475891,
      "reward_std": 0.3283276781439781,
      "reward_std_all": 0.791829913854599,
      "rewards-max/forensics_sinkhorn_reward": 0.9886400401592255,
      "rewards-max/forensics_strict_edge_reward": 0.870158314704895,
      "rewards-max/forensics_tiou_reward": 0.9983645379543304,
      "rewards/forensics_sinkhorn_reward": 0.7909833490848541,
      "rewards/forensics_strict_edge_reward": 0.14428726583719254,
      "rewards/forensics_tiou_reward": 0.641303539276123,
      "step": 1056
    },
    {
      "completion_length": 31.109375,
      "epoch": 3.2427035330261136,
      "grad_norm": 12.06179141998291,
      "kl": 0.623046875,
      "learning_rate": 1.8769230769230767e-07,
      "loss": -0.0426,
      "reward": 1.8390347361564636,
      "reward_std": 0.3696431666612625,
      "reward_std_all": 0.718765914440155,
      "rewards-max/forensics_sinkhorn_reward": 0.9901953935623169,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.8089396357536316,
      "rewards/forensics_strict_edge_reward": 0.30448615550994873,
      "rewards/forensics_tiou_reward": 0.7256089150905609,
      "step": 1057
    },
    {
      "completion_length": 25.8125,
      "epoch": 3.2457757296466974,
      "grad_norm": 10.5711669921875,
      "kl": 0.603515625,
      "learning_rate": 1.8692307692307692e-07,
      "loss": -0.1218,
      "reward": 1.4181005954742432,
      "reward_std": 0.3227885663509369,
      "reward_std_all": 0.8595116436481476,
      "rewards-max/forensics_sinkhorn_reward": 0.977173238992691,
      "rewards-max/forensics_strict_edge_reward": 0.6222964972257614,
      "rewards-max/forensics_tiou_reward": 0.970361739397049,
      "rewards/forensics_sinkhorn_reward": 0.7620856165885925,
      "rewards/forensics_strict_edge_reward": 0.055602455511689186,
      "rewards/forensics_tiou_reward": 0.6004125773906708,
      "step": 1058
    },
    {
      "completion_length": 27.6875,
      "epoch": 3.248847926267281,
      "grad_norm": 11.594693183898926,
      "kl": 0.59375,
      "learning_rate": 1.8615384615384613e-07,
      "loss": -0.0021,
      "reward": 1.771422028541565,
      "reward_std": 0.4298121780157089,
      "reward_std_all": 0.7370804250240326,
      "rewards-max/forensics_sinkhorn_reward": 0.9895264804363251,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.8382349908351898,
      "rewards/forensics_strict_edge_reward": 0.20174028724431992,
      "rewards/forensics_tiou_reward": 0.7314466834068298,
      "step": 1059
    },
    {
      "completion_length": 23.46875,
      "epoch": 3.251920122887865,
      "grad_norm": 11.87570571899414,
      "kl": 0.50390625,
      "learning_rate": 1.853846153846154e-07,
      "loss": -0.1327,
      "reward": 1.6339863538742065,
      "reward_std": 0.5068932920694351,
      "reward_std_all": 0.8586490750312805,
      "rewards-max/forensics_sinkhorn_reward": 0.9894738793373108,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.7867750525474548,
      "rewards/forensics_strict_edge_reward": 0.1732461228966713,
      "rewards/forensics_tiou_reward": 0.6739650964736938,
      "step": 1060
    },
    {
      "completion_length": 26.375,
      "epoch": 3.2549923195084487,
      "grad_norm": 11.096550941467285,
      "kl": 0.72265625,
      "learning_rate": 1.846153846153846e-07,
      "loss": -0.0299,
      "reward": 1.646332859992981,
      "reward_std": 0.3713516741991043,
      "reward_std_all": 0.8008604943752289,
      "rewards-max/forensics_sinkhorn_reward": 0.9890056550502777,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.7800529897212982,
      "rewards/forensics_strict_edge_reward": 0.2226388230919838,
      "rewards/forensics_tiou_reward": 0.643641009926796,
      "step": 1061
    },
    {
      "completion_length": 25.03125,
      "epoch": 3.258064516129032,
      "grad_norm": 13.374114990234375,
      "kl": 0.54296875,
      "learning_rate": 1.8384615384615385e-07,
      "loss": 0.0165,
      "reward": 1.4905259013175964,
      "reward_std": 0.4877210259437561,
      "reward_std_all": 0.8991280496120453,
      "rewards-max/forensics_sinkhorn_reward": 0.9899916648864746,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.7313665747642517,
      "rewards/forensics_strict_edge_reward": 0.14204464852809906,
      "rewards/forensics_tiou_reward": 0.6171147227287292,
      "step": 1062
    },
    {
      "completion_length": 23.40625,
      "epoch": 3.261136712749616,
      "grad_norm": 13.180253028869629,
      "kl": 0.6015625,
      "learning_rate": 1.8307692307692306e-07,
      "loss": -0.0732,
      "reward": 1.4707697629928589,
      "reward_std": 0.4791218936443329,
      "reward_std_all": 0.901365339756012,
      "rewards-max/forensics_sinkhorn_reward": 0.9867212176322937,
      "rewards-max/forensics_strict_edge_reward": 0.9093653857707977,
      "rewards-max/forensics_tiou_reward": 0.9927116632461548,
      "rewards/forensics_sinkhorn_reward": 0.6921634674072266,
      "rewards/forensics_strict_edge_reward": 0.15981198474764824,
      "rewards/forensics_tiou_reward": 0.618794247508049,
      "step": 1063
    },
    {
      "completion_length": 27.015625,
      "epoch": 3.2642089093701996,
      "grad_norm": 10.986602783203125,
      "kl": 0.548828125,
      "learning_rate": 1.823076923076923e-07,
      "loss": -0.1408,
      "reward": 1.7500693202018738,
      "reward_std": 0.4755840003490448,
      "reward_std_all": 0.7870046198368073,
      "rewards-max/forensics_sinkhorn_reward": 0.9902515709400177,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.81260746717453,
      "rewards/forensics_strict_edge_reward": 0.2474854439496994,
      "rewards/forensics_tiou_reward": 0.6899764537811279,
      "step": 1064
    },
    {
      "completion_length": 25.59375,
      "epoch": 3.2672811059907834,
      "grad_norm": 14.628039360046387,
      "kl": 0.4716796875,
      "learning_rate": 1.8153846153846152e-07,
      "loss": -0.0573,
      "reward": 1.6519290208816528,
      "reward_std": 0.44510187208652496,
      "reward_std_all": 0.814012348651886,
      "rewards-max/forensics_sinkhorn_reward": 0.9861707389354706,
      "rewards-max/forensics_strict_edge_reward": 0.8351600170135498,
      "rewards-max/forensics_tiou_reward": 0.9914817214012146,
      "rewards/forensics_sinkhorn_reward": 0.7929540276527405,
      "rewards/forensics_strict_edge_reward": 0.1894720420241356,
      "rewards/forensics_tiou_reward": 0.6695027649402618,
      "step": 1065
    },
    {
      "completion_length": 25.25,
      "epoch": 3.270353302611367,
      "grad_norm": 12.82509994506836,
      "kl": 0.580078125,
      "learning_rate": 1.8076923076923076e-07,
      "loss": -0.0146,
      "reward": 1.5024638175964355,
      "reward_std": 0.4249056726694107,
      "reward_std_all": 0.8701713383197784,
      "rewards-max/forensics_sinkhorn_reward": 0.9852614998817444,
      "rewards-max/forensics_strict_edge_reward": 0.8582656681537628,
      "rewards-max/forensics_tiou_reward": 0.9921477138996124,
      "rewards/forensics_sinkhorn_reward": 0.7492711842060089,
      "rewards/forensics_strict_edge_reward": 0.1280931942164898,
      "rewards/forensics_tiou_reward": 0.625099390745163,
      "step": 1066
    },
    {
      "completion_length": 30.8125,
      "epoch": 3.273425499231951,
      "grad_norm": 10.455354690551758,
      "kl": 0.576171875,
      "learning_rate": 1.8e-07,
      "loss": -0.0647,
      "reward": 1.6023268103599548,
      "reward_std": 0.5247331261634827,
      "reward_std_all": 0.8468112349510193,
      "rewards-max/forensics_sinkhorn_reward": 0.9894301891326904,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.7526750862598419,
      "rewards/forensics_strict_edge_reward": 0.1920892894268036,
      "rewards/forensics_tiou_reward": 0.6575624942779541,
      "step": 1067
    },
    {
      "completion_length": 26.0625,
      "epoch": 3.2764976958525347,
      "grad_norm": 9.480844497680664,
      "kl": 0.548828125,
      "learning_rate": 1.792307692307692e-07,
      "loss": -0.0674,
      "reward": 1.8788467049598694,
      "reward_std": 0.3802470713853836,
      "reward_std_all": 0.674001008272171,
      "rewards-max/forensics_sinkhorn_reward": 0.9877406656742096,
      "rewards-max/forensics_strict_edge_reward": 0.8135445415973663,
      "rewards-max/forensics_tiou_reward": 0.99397873878479,
      "rewards/forensics_sinkhorn_reward": 0.8486593663692474,
      "rewards/forensics_strict_edge_reward": 0.27592408657073975,
      "rewards/forensics_tiou_reward": 0.754263162612915,
      "step": 1068
    },
    {
      "completion_length": 26.234375,
      "epoch": 3.279569892473118,
      "grad_norm": 12.508949279785156,
      "kl": 0.5390625,
      "learning_rate": 1.7846153846153845e-07,
      "loss": -0.0407,
      "reward": 1.6102758646011353,
      "reward_std": 0.4026178866624832,
      "reward_std_all": 0.8211027383804321,
      "rewards-max/forensics_sinkhorn_reward": 0.9890736639499664,
      "rewards-max/forensics_strict_edge_reward": 0.9838767349720001,
      "rewards-max/forensics_tiou_reward": 0.9997832477092743,
      "rewards/forensics_sinkhorn_reward": 0.7592824101448059,
      "rewards/forensics_strict_edge_reward": 0.18636563047766685,
      "rewards/forensics_tiou_reward": 0.6646278500556946,
      "step": 1069
    },
    {
      "completion_length": 28.890625,
      "epoch": 3.282642089093702,
      "grad_norm": 12.642706871032715,
      "kl": 0.552734375,
      "learning_rate": 1.776923076923077e-07,
      "loss": -0.0046,
      "reward": 1.6977869272232056,
      "reward_std": 0.3507421389222145,
      "reward_std_all": 0.7480036914348602,
      "rewards-max/forensics_sinkhorn_reward": 0.988580971956253,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.8056716322898865,
      "rewards/forensics_strict_edge_reward": 0.17134101688861847,
      "rewards/forensics_tiou_reward": 0.7207743525505066,
      "step": 1070
    },
    {
      "completion_length": 28.234375,
      "epoch": 3.2857142857142856,
      "grad_norm": 12.992531776428223,
      "kl": 0.5673828125,
      "learning_rate": 1.7692307692307693e-07,
      "loss": -0.0018,
      "reward": 1.5846242904663086,
      "reward_std": 0.2657839506864548,
      "reward_std_all": 0.7659172117710114,
      "rewards-max/forensics_sinkhorn_reward": 0.9834286570549011,
      "rewards-max/forensics_strict_edge_reward": 0.7899699509143829,
      "rewards-max/forensics_tiou_reward": 0.9844850301742554,
      "rewards/forensics_sinkhorn_reward": 0.8106470108032227,
      "rewards/forensics_strict_edge_reward": 0.1123587042093277,
      "rewards/forensics_tiou_reward": 0.6616186499595642,
      "step": 1071
    },
    {
      "completion_length": 24.859375,
      "epoch": 3.2887864823348694,
      "grad_norm": 11.944741249084473,
      "kl": 0.48046875,
      "learning_rate": 1.7615384615384615e-07,
      "loss": -0.0973,
      "reward": 1.743076205253601,
      "reward_std": 0.4340321272611618,
      "reward_std_all": 0.7548188865184784,
      "rewards-max/forensics_sinkhorn_reward": 0.9869474470615387,
      "rewards-max/forensics_strict_edge_reward": 0.9093653857707977,
      "rewards-max/forensics_tiou_reward": 0.9941761791706085,
      "rewards/forensics_sinkhorn_reward": 0.8465844094753265,
      "rewards/forensics_strict_edge_reward": 0.14639485254883766,
      "rewards/forensics_tiou_reward": 0.7500969171524048,
      "step": 1072
    },
    {
      "completion_length": 26.734375,
      "epoch": 3.291858678955453,
      "grad_norm": 11.998906135559082,
      "kl": 0.5087890625,
      "learning_rate": 1.7538461538461539e-07,
      "loss": -0.1136,
      "reward": 1.8767398595809937,
      "reward_std": 0.268198698759079,
      "reward_std_all": 0.6320516765117645,
      "rewards-max/forensics_sinkhorn_reward": 0.9884939193725586,
      "rewards-max/forensics_strict_edge_reward": 0.8465203046798706,
      "rewards-max/forensics_tiou_reward": 0.9961967766284943,
      "rewards/forensics_sinkhorn_reward": 0.8954921066761017,
      "rewards/forensics_strict_edge_reward": 0.17389608547091484,
      "rewards/forensics_tiou_reward": 0.8073515594005585,
      "step": 1073
    },
    {
      "completion_length": 27.75,
      "epoch": 3.294930875576037,
      "grad_norm": 10.937204360961914,
      "kl": 0.5615234375,
      "learning_rate": 1.746153846153846e-07,
      "loss": -0.1214,
      "reward": 1.7875420451164246,
      "reward_std": 0.39022327959537506,
      "reward_std_all": 0.7425791621208191,
      "rewards-max/forensics_sinkhorn_reward": 0.9891961812973022,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.8429594933986664,
      "rewards/forensics_strict_edge_reward": 0.24045678228139877,
      "rewards/forensics_tiou_reward": 0.7041257321834564,
      "step": 1074
    },
    {
      "completion_length": 31.96875,
      "epoch": 3.2980030721966207,
      "grad_norm": 12.877942085266113,
      "kl": 0.775390625,
      "learning_rate": 1.7384615384615384e-07,
      "loss": -0.0498,
      "reward": 1.3688843250274658,
      "reward_std": 0.47144849598407745,
      "reward_std_all": 0.9634701609611511,
      "rewards-max/forensics_sinkhorn_reward": 0.9870886206626892,
      "rewards-max/forensics_strict_edge_reward": 0.9189028441905975,
      "rewards-max/forensics_tiou_reward": 0.9948513209819794,
      "rewards/forensics_sinkhorn_reward": 0.6435295045375824,
      "rewards/forensics_strict_edge_reward": 0.16574276983737946,
      "rewards/forensics_tiou_reward": 0.5596121549606323,
      "step": 1075
    },
    {
      "completion_length": 22.34375,
      "epoch": 3.3010752688172045,
      "grad_norm": 13.179239273071289,
      "kl": 0.59375,
      "learning_rate": 1.7307692307692305e-07,
      "loss": -0.0089,
      "reward": 1.6122629642486572,
      "reward_std": 0.37688563764095306,
      "reward_std_all": 0.7953788638114929,
      "rewards-max/forensics_sinkhorn_reward": 0.9875006377696991,
      "rewards-max/forensics_strict_edge_reward": 0.9291328191757202,
      "rewards-max/forensics_tiou_reward": 0.9951806366443634,
      "rewards/forensics_sinkhorn_reward": 0.8082680404186249,
      "rewards/forensics_strict_edge_reward": 0.13438869640231133,
      "rewards/forensics_tiou_reward": 0.6696061789989471,
      "step": 1076
    },
    {
      "completion_length": 22.84375,
      "epoch": 3.3041474654377883,
      "grad_norm": 9.554558753967285,
      "kl": 0.4853515625,
      "learning_rate": 1.723076923076923e-07,
      "loss": -0.0681,
      "reward": 1.5564996004104614,
      "reward_std": 0.4116092622280121,
      "reward_std_all": 0.8567355275154114,
      "rewards-max/forensics_sinkhorn_reward": 0.9814861416816711,
      "rewards-max/forensics_strict_edge_reward": 0.7838338315486908,
      "rewards-max/forensics_tiou_reward": 0.9802622199058533,
      "rewards/forensics_sinkhorn_reward": 0.7803656160831451,
      "rewards/forensics_strict_edge_reward": 0.1563793234527111,
      "rewards/forensics_tiou_reward": 0.6197546720504761,
      "step": 1077
    },
    {
      "completion_length": 25.640625,
      "epoch": 3.3072196620583716,
      "grad_norm": 13.244667053222656,
      "kl": 0.4287109375,
      "learning_rate": 1.7153846153846153e-07,
      "loss": -0.0522,
      "reward": 1.4735249280929565,
      "reward_std": 0.38843517005443573,
      "reward_std_all": 0.875568687915802,
      "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": 0.7241699993610382,
      "rewards/forensics_strict_edge_reward": 0.13563872873783112,
      "rewards/forensics_tiou_reward": 0.6137161552906036,
      "step": 1078
    },
    {
      "completion_length": 29.140625,
      "epoch": 3.3102918586789554,
      "grad_norm": 35.85971450805664,
      "kl": 0.86328125,
      "learning_rate": 1.7076923076923075e-07,
      "loss": 0.0305,
      "reward": 1.6676483154296875,
      "reward_std": 0.37441378831863403,
      "reward_std_all": 0.7720150649547577,
      "rewards-max/forensics_sinkhorn_reward": 0.990130603313446,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.8066076636314392,
      "rewards/forensics_strict_edge_reward": 0.1701391413807869,
      "rewards/forensics_tiou_reward": 0.6909015774726868,
      "step": 1079
    },
    {
      "completion_length": 22.59375,
      "epoch": 3.313364055299539,
      "grad_norm": 12.342203140258789,
      "kl": 0.548828125,
      "learning_rate": 1.7000000000000001e-07,
      "loss": 0.0064,
      "reward": 1.3821342587471008,
      "reward_std": 0.4601873904466629,
      "reward_std_all": 0.944507360458374,
      "rewards-max/forensics_sinkhorn_reward": 0.9896026849746704,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.7156099379062653,
      "rewards/forensics_strict_edge_reward": 0.13683662563562393,
      "rewards/forensics_tiou_reward": 0.5296878218650818,
      "step": 1080
    },
    {
      "completion_length": 27.78125,
      "epoch": 3.316436251920123,
      "grad_norm": 14.80681324005127,
      "kl": 0.74609375,
      "learning_rate": 1.6923076923076923e-07,
      "loss": -0.0042,
      "reward": 1.4756341576576233,
      "reward_std": 0.3897838741540909,
      "reward_std_all": 0.8815607726573944,
      "rewards-max/forensics_sinkhorn_reward": 0.9891408979892731,
      "rewards-max/forensics_strict_edge_reward": 0.9414820969104767,
      "rewards-max/forensics_tiou_reward": 0.999706506729126,
      "rewards/forensics_sinkhorn_reward": 0.6960954964160919,
      "rewards/forensics_strict_edge_reward": 0.19132915139198303,
      "rewards/forensics_tiou_reward": 0.5882095396518707,
      "step": 1081
    },
    {
      "completion_length": 26.015625,
      "epoch": 3.3195084485407067,
      "grad_norm": 11.433243751525879,
      "kl": 0.609375,
      "learning_rate": 1.6846153846153847e-07,
      "loss": -0.0688,
      "reward": 1.5654882192611694,
      "reward_std": 0.4533681571483612,
      "reward_std_all": 0.858970582485199,
      "rewards-max/forensics_sinkhorn_reward": 0.9844193458557129,
      "rewards-max/forensics_strict_edge_reward": 0.7972188889980316,
      "rewards-max/forensics_tiou_reward": 0.9872922301292419,
      "rewards/forensics_sinkhorn_reward": 0.7666177451610565,
      "rewards/forensics_strict_edge_reward": 0.12802120298147202,
      "rewards/forensics_tiou_reward": 0.6708492934703827,
      "step": 1082
    },
    {
      "completion_length": 24.734375,
      "epoch": 3.3225806451612905,
      "grad_norm": 14.574947357177734,
      "kl": 0.51953125,
      "learning_rate": 1.6769230769230768e-07,
      "loss": -0.0666,
      "reward": 1.5476107001304626,
      "reward_std": 0.33717505633831024,
      "reward_std_all": 0.8230843245983124,
      "rewards-max/forensics_sinkhorn_reward": 0.9849199652671814,
      "rewards-max/forensics_strict_edge_reward": 0.8959447741508484,
      "rewards-max/forensics_tiou_reward": 0.9871655106544495,
      "rewards/forensics_sinkhorn_reward": 0.724433034658432,
      "rewards/forensics_strict_edge_reward": 0.20410219579935074,
      "rewards/forensics_tiou_reward": 0.6190755069255829,
      "step": 1083
    },
    {
      "completion_length": 29.375,
      "epoch": 3.3256528417818743,
      "grad_norm": 12.597404479980469,
      "kl": 0.62109375,
      "learning_rate": 1.6692307692307692e-07,
      "loss": -0.0915,
      "reward": 1.5967340469360352,
      "reward_std": 0.45158466696739197,
      "reward_std_all": 0.8267759680747986,
      "rewards-max/forensics_sinkhorn_reward": 0.9786586463451385,
      "rewards-max/forensics_strict_edge_reward": 0.8135445415973663,
      "rewards-max/forensics_tiou_reward": 0.9769810736179352,
      "rewards/forensics_sinkhorn_reward": 0.7563689053058624,
      "rewards/forensics_strict_edge_reward": 0.1732499897480011,
      "rewards/forensics_tiou_reward": 0.6671151518821716,
      "step": 1084
    },
    {
      "completion_length": 26.78125,
      "epoch": 3.3287250384024576,
      "grad_norm": 14.095133781433105,
      "kl": 0.666015625,
      "learning_rate": 1.6615384615384613e-07,
      "loss": -0.0569,
      "reward": 1.72504460811615,
      "reward_std": 0.4994398057460785,
      "reward_std_all": 0.8244709670543671,
      "rewards-max/forensics_sinkhorn_reward": 0.986566036939621,
      "rewards-max/forensics_strict_edge_reward": 0.8582656681537628,
      "rewards-max/forensics_tiou_reward": 0.9903719127178192,
      "rewards/forensics_sinkhorn_reward": 0.8115967810153961,
      "rewards/forensics_strict_edge_reward": 0.2000875622034073,
      "rewards/forensics_tiou_reward": 0.7133602499961853,
      "step": 1085
    },
    {
      "completion_length": 30.765625,
      "epoch": 3.3317972350230414,
      "grad_norm": 18.940692901611328,
      "kl": 0.9609375,
      "learning_rate": 1.6538461538461538e-07,
      "loss": -0.0091,
      "reward": 1.8453239798545837,
      "reward_std": 0.5074532926082611,
      "reward_std_all": 0.7578911483287811,
      "rewards-max/forensics_sinkhorn_reward": 0.9897074103355408,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.8205873370170593,
      "rewards/forensics_strict_edge_reward": 0.2613336443901062,
      "rewards/forensics_tiou_reward": 0.7634030878543854,
      "step": 1086
    },
    {
      "completion_length": 28.125,
      "epoch": 3.334869431643625,
      "grad_norm": 13.027496337890625,
      "kl": 0.5859375,
      "learning_rate": 1.646153846153846e-07,
      "loss": -0.0512,
      "reward": 1.7817128896713257,
      "reward_std": 0.4907377064228058,
      "reward_std_all": 0.8053304255008698,
      "rewards-max/forensics_sinkhorn_reward": 0.9902819395065308,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.8021821081638336,
      "rewards/forensics_strict_edge_reward": 0.27393992245197296,
      "rewards/forensics_tiou_reward": 0.7055908143520355,
      "step": 1087
    },
    {
      "completion_length": 25.140625,
      "epoch": 3.337941628264209,
      "grad_norm": 13.151041030883789,
      "kl": 0.603515625,
      "learning_rate": 1.6384615384615383e-07,
      "loss": -0.0413,
      "reward": 1.8010897040367126,
      "reward_std": 0.42930619418621063,
      "reward_std_all": 0.7491810619831085,
      "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": 0.8382073938846588,
      "rewards/forensics_strict_edge_reward": 0.2684737518429756,
      "rewards/forensics_tiou_reward": 0.694408506155014,
      "step": 1088
    },
    {
      "completion_length": 24.125,
      "epoch": 3.3410138248847927,
      "grad_norm": 14.785518646240234,
      "kl": 0.75390625,
      "learning_rate": 1.6307692307692307e-07,
      "loss": -0.0979,
      "reward": 1.7749914526939392,
      "reward_std": 0.3547869771718979,
      "reward_std_all": 0.7051268815994263,
      "rewards-max/forensics_sinkhorn_reward": 0.9881619215011597,
      "rewards-max/forensics_strict_edge_reward": 0.8582656681537628,
      "rewards-max/forensics_tiou_reward": 0.9969955086708069,
      "rewards/forensics_sinkhorn_reward": 0.8480023741722107,
      "rewards/forensics_strict_edge_reward": 0.17225295305252075,
      "rewards/forensics_tiou_reward": 0.7547361254692078,
      "step": 1089
    },
    {
      "completion_length": 29.046875,
      "epoch": 3.3440860215053765,
      "grad_norm": 12.997811317443848,
      "kl": 0.59375,
      "learning_rate": 1.623076923076923e-07,
      "loss": -0.0097,
      "reward": 1.8077049255371094,
      "reward_std": 0.43698565661907196,
      "reward_std_all": 0.7556269466876984,
      "rewards-max/forensics_sinkhorn_reward": 0.9897510409355164,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.8330018222332001,
      "rewards/forensics_strict_edge_reward": 0.22733048349618912,
      "rewards/forensics_tiou_reward": 0.7473725080490112,
      "step": 1090
    },
    {
      "completion_length": 23.609375,
      "epoch": 3.3471582181259603,
      "grad_norm": 12.915979385375977,
      "kl": 0.640625,
      "learning_rate": 1.6153846153846155e-07,
      "loss": 0.0437,
      "reward": 1.6254940032958984,
      "reward_std": 0.36293113231658936,
      "reward_std_all": 0.7924069762229919,
      "rewards-max/forensics_sinkhorn_reward": 0.9856466054916382,
      "rewards-max/forensics_strict_edge_reward": 0.8333333432674408,
      "rewards-max/forensics_tiou_reward": 0.9886056184768677,
      "rewards/forensics_sinkhorn_reward": 0.8203347325325012,
      "rewards/forensics_strict_edge_reward": 0.12399307638406754,
      "rewards/forensics_tiou_reward": 0.6811661124229431,
      "step": 1091
    },
    {
      "completion_length": 23.796875,
      "epoch": 3.3502304147465436,
      "grad_norm": 12.756492614746094,
      "kl": 0.4423828125,
      "learning_rate": 1.6076923076923076e-07,
      "loss": -0.0781,
      "reward": 1.583062767982483,
      "reward_std": 0.37374086678028107,
      "reward_std_all": 0.8098796010017395,
      "rewards-max/forensics_sinkhorn_reward": 0.9895292520523071,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.7482779026031494,
      "rewards/forensics_strict_edge_reward": 0.2000003680586815,
      "rewards/forensics_tiou_reward": 0.634784460067749,
      "step": 1092
    },
    {
      "completion_length": 25.625,
      "epoch": 3.3533026113671274,
      "grad_norm": 14.890593528747559,
      "kl": 0.5849609375,
      "learning_rate": 1.6e-07,
      "loss": -0.0955,
      "reward": 1.8283414840698242,
      "reward_std": 0.44309786707162857,
      "reward_std_all": 0.7311181724071503,
      "rewards-max/forensics_sinkhorn_reward": 0.9888167381286621,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.8602534830570221,
      "rewards/forensics_strict_edge_reward": 0.20869815349578857,
      "rewards/forensics_tiou_reward": 0.7593898475170135,
      "step": 1093
    },
    {
      "completion_length": 26.78125,
      "epoch": 3.356374807987711,
      "grad_norm": 11.9060697555542,
      "kl": 0.54296875,
      "learning_rate": 1.5923076923076922e-07,
      "loss": -0.0732,
      "reward": 1.5612353682518005,
      "reward_std": 0.39571619033813477,
      "reward_std_all": 0.825899213552475,
      "rewards-max/forensics_sinkhorn_reward": 0.9866382479667664,
      "rewards-max/forensics_strict_edge_reward": 0.8582656681537628,
      "rewards-max/forensics_tiou_reward": 0.9943913817405701,
      "rewards/forensics_sinkhorn_reward": 0.7515313029289246,
      "rewards/forensics_strict_edge_reward": 0.18008708208799362,
      "rewards/forensics_tiou_reward": 0.6296170353889465,
      "step": 1094
    },
    {
      "completion_length": 25.015625,
      "epoch": 3.359447004608295,
      "grad_norm": 14.764387130737305,
      "kl": 0.57421875,
      "learning_rate": 1.5846153846153846e-07,
      "loss": -0.1148,
      "reward": 1.6667001247406006,
      "reward_std": 0.38898421823978424,
      "reward_std_all": 0.7998156547546387,
      "rewards-max/forensics_sinkhorn_reward": 0.9895395338535309,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.8083745837211609,
      "rewards/forensics_strict_edge_reward": 0.18323373794555664,
      "rewards/forensics_tiou_reward": 0.6750918924808502,
      "step": 1095
    },
    {
      "completion_length": 30.53125,
      "epoch": 3.3625192012288787,
      "grad_norm": 12.197322845458984,
      "kl": 0.71875,
      "learning_rate": 1.5769230769230767e-07,
      "loss": -0.0914,
      "reward": 1.7628684043884277,
      "reward_std": 0.5054665207862854,
      "reward_std_all": 0.7961930930614471,
      "rewards-max/forensics_sinkhorn_reward": 0.9893902838230133,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.771230548620224,
      "rewards/forensics_strict_edge_reward": 0.2892303913831711,
      "rewards/forensics_tiou_reward": 0.7024074792861938,
      "step": 1096
    },
    {
      "completion_length": 22.390625,
      "epoch": 3.3655913978494625,
      "grad_norm": 11.82132625579834,
      "kl": 0.4267578125,
      "learning_rate": 1.569230769230769e-07,
      "loss": -0.0984,
      "reward": 1.7576530575752258,
      "reward_std": 0.4466354548931122,
      "reward_std_all": 0.7607251405715942,
      "rewards-max/forensics_sinkhorn_reward": 0.9892763793468475,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.8316100537776947,
      "rewards/forensics_strict_edge_reward": 0.20129930973052979,
      "rewards/forensics_tiou_reward": 0.7247436344623566,
      "step": 1097
    },
    {
      "completion_length": 29.203125,
      "epoch": 3.3686635944700463,
      "grad_norm": 14.177084922790527,
      "kl": 0.728515625,
      "learning_rate": 1.5615384615384612e-07,
      "loss": -0.0613,
      "reward": 1.8677974343299866,
      "reward_std": 0.46738119423389435,
      "reward_std_all": 0.7406016290187836,
      "rewards-max/forensics_sinkhorn_reward": 0.9894770681858063,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.8358122706413269,
      "rewards/forensics_strict_edge_reward": 0.2924269735813141,
      "rewards/forensics_tiou_reward": 0.739558219909668,
      "step": 1098
    },
    {
      "completion_length": 22.453125,
      "epoch": 3.3717357910906296,
      "grad_norm": 16.271499633789062,
      "kl": 0.609375,
      "learning_rate": 1.5538461538461536e-07,
      "loss": -0.0538,
      "reward": 1.6325832605361938,
      "reward_std": 0.4249396324157715,
      "reward_std_all": 0.8158118724822998,
      "rewards-max/forensics_sinkhorn_reward": 0.9889446496963501,
      "rewards-max/forensics_strict_edge_reward": 0.9479723870754242,
      "rewards-max/forensics_tiou_reward": 0.998988687992096,
      "rewards/forensics_sinkhorn_reward": 0.820976585149765,
      "rewards/forensics_strict_edge_reward": 0.15347588434815407,
      "rewards/forensics_tiou_reward": 0.6581307351589203,
      "step": 1099
    },
    {
      "completion_length": 24.625,
      "epoch": 3.3748079877112134,
      "grad_norm": 12.7435941696167,
      "kl": 0.607421875,
      "learning_rate": 1.5461538461538463e-07,
      "loss": -0.0596,
      "reward": 1.4359039664268494,
      "reward_std": 0.44973570108413696,
      "reward_std_all": 0.9253043532371521,
      "rewards-max/forensics_sinkhorn_reward": 0.9892392158508301,
      "rewards-max/forensics_strict_edge_reward": 0.9682791531085968,
      "rewards-max/forensics_tiou_reward": 0.9993367493152618,
      "rewards/forensics_sinkhorn_reward": 0.7194110751152039,
      "rewards/forensics_strict_edge_reward": 0.12945177778601646,
      "rewards/forensics_tiou_reward": 0.5870411843061447,
      "step": 1100
    },
    {
      "completion_length": 21.9375,
      "epoch": 3.377880184331797,
      "grad_norm": 13.44908332824707,
      "kl": 0.6865234375,
      "learning_rate": 1.5384615384615385e-07,
      "loss": -0.0453,
      "reward": 1.5886709094047546,
      "reward_std": 0.4695498049259186,
      "reward_std_all": 0.8337770700454712,
      "rewards-max/forensics_sinkhorn_reward": 0.9895250499248505,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.7956196665763855,
      "rewards/forensics_strict_edge_reward": 0.14619942754507065,
      "rewards/forensics_tiou_reward": 0.6468518078327179,
      "step": 1101
    },
    {
      "completion_length": 23.5,
      "epoch": 3.380952380952381,
      "grad_norm": 20.14897346496582,
      "kl": 0.77734375,
      "learning_rate": 1.5307692307692309e-07,
      "loss": -0.0195,
      "reward": 1.4532347321510315,
      "reward_std": 0.34452947974205017,
      "reward_std_all": 0.8477831184864044,
      "rewards-max/forensics_sinkhorn_reward": 0.9861108362674713,
      "rewards-max/forensics_strict_edge_reward": 0.756708562374115,
      "rewards-max/forensics_tiou_reward": 0.9904019236564636,
      "rewards/forensics_sinkhorn_reward": 0.7607150673866272,
      "rewards/forensics_strict_edge_reward": 0.13819586113095284,
      "rewards/forensics_tiou_reward": 0.5543237924575806,
      "step": 1102
    },
    {
      "completion_length": 28.3125,
      "epoch": 3.3840245775729647,
      "grad_norm": 12.420273780822754,
      "kl": 0.5859375,
      "learning_rate": 1.523076923076923e-07,
      "loss": -0.1071,
      "reward": 1.9108139276504517,
      "reward_std": 0.395759716629982,
      "reward_std_all": 0.6813515722751617,
      "rewards-max/forensics_sinkhorn_reward": 0.9904015362262726,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.8833832144737244,
      "rewards/forensics_strict_edge_reward": 0.23335327208042145,
      "rewards/forensics_tiou_reward": 0.7940773665904999,
      "step": 1103
    },
    {
      "completion_length": 26.546875,
      "epoch": 3.3870967741935485,
      "grad_norm": 12.171364784240723,
      "kl": 0.59375,
      "learning_rate": 1.5153846153846154e-07,
      "loss": -0.0137,
      "reward": 1.7395303845405579,
      "reward_std": 0.32480743527412415,
      "reward_std_all": 0.7185968458652496,
      "rewards-max/forensics_sinkhorn_reward": 0.9887506365776062,
      "rewards-max/forensics_strict_edge_reward": 0.9332861304283142,
      "rewards-max/forensics_tiou_reward": 0.9994624555110931,
      "rewards/forensics_sinkhorn_reward": 0.8160218596458435,
      "rewards/forensics_strict_edge_reward": 0.22652926295995712,
      "rewards/forensics_tiou_reward": 0.6969792246818542,
      "step": 1104
    },
    {
      "completion_length": 27.734375,
      "epoch": 3.3901689708141323,
      "grad_norm": 13.001761436462402,
      "kl": 0.67578125,
      "learning_rate": 1.5076923076923075e-07,
      "loss": 0.0289,
      "reward": 1.6063295602798462,
      "reward_std": 0.3397117853164673,
      "reward_std_all": 0.7937802970409393,
      "rewards-max/forensics_sinkhorn_reward": 0.9903029799461365,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.7634318470954895,
      "rewards/forensics_strict_edge_reward": 0.1892210990190506,
      "rewards/forensics_tiou_reward": 0.6536766588687897,
      "step": 1105
    },
    {
      "completion_length": 27.75,
      "epoch": 3.3932411674347156,
      "grad_norm": 10.80063247680664,
      "kl": 0.748046875,
      "learning_rate": 1.5e-07,
      "loss": -0.103,
      "reward": 1.8396711349487305,
      "reward_std": 0.5161329060792923,
      "reward_std_all": 0.7653410136699677,
      "rewards-max/forensics_sinkhorn_reward": 0.9901880919933319,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.8416325449943542,
      "rewards/forensics_strict_edge_reward": 0.27376989275217056,
      "rewards/forensics_tiou_reward": 0.7242687344551086,
      "step": 1106
    },
    {
      "completion_length": 21.75,
      "epoch": 3.3963133640552994,
      "grad_norm": 10.808815002441406,
      "kl": 0.44140625,
      "learning_rate": 1.492307692307692e-07,
      "loss": -0.2327,
      "reward": 1.8065427541732788,
      "reward_std": 0.46242935955524445,
      "reward_std_all": 0.7712644934654236,
      "rewards-max/forensics_sinkhorn_reward": 0.9894598126411438,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.831701785326004,
      "rewards/forensics_strict_edge_reward": 0.26999297738075256,
      "rewards/forensics_tiou_reward": 0.7048479914665222,
      "step": 1107
    },
    {
      "completion_length": 24.859375,
      "epoch": 3.399385560675883,
      "grad_norm": 11.026315689086914,
      "kl": 0.5068359375,
      "learning_rate": 1.4846153846153845e-07,
      "loss": -0.0666,
      "reward": 1.8206067085266113,
      "reward_std": 0.4747975468635559,
      "reward_std_all": 0.7437493801116943,
      "rewards-max/forensics_sinkhorn_reward": 0.9885295033454895,
      "rewards-max/forensics_strict_edge_reward": 0.9253588318824768,
      "rewards-max/forensics_tiou_reward": 0.999299556016922,
      "rewards/forensics_sinkhorn_reward": 0.8168681263923645,
      "rewards/forensics_strict_edge_reward": 0.2562287151813507,
      "rewards/forensics_tiou_reward": 0.7475098371505737,
      "step": 1108
    },
    {
      "completion_length": 28.40625,
      "epoch": 3.402457757296467,
      "grad_norm": 12.05515193939209,
      "kl": 0.701171875,
      "learning_rate": 1.476923076923077e-07,
      "loss": -0.089,
      "reward": 1.707534372806549,
      "reward_std": 0.3794301748275757,
      "reward_std_all": 0.752609521150589,
      "rewards-max/forensics_sinkhorn_reward": 0.9876134097576141,
      "rewards-max/forensics_strict_edge_reward": 0.8582656681537628,
      "rewards-max/forensics_tiou_reward": 0.9946962594985962,
      "rewards/forensics_sinkhorn_reward": 0.8193728029727936,
      "rewards/forensics_strict_edge_reward": 0.15668092295527458,
      "rewards/forensics_tiou_reward": 0.7314806580543518,
      "step": 1109
    },
    {
      "completion_length": 22.234375,
      "epoch": 3.4055299539170507,
      "grad_norm": 85.6888656616211,
      "kl": 2.5322265625,
      "learning_rate": 1.4692307692307693e-07,
      "loss": -0.0149,
      "reward": 1.5353747606277466,
      "reward_std": 0.6240891814231873,
      "reward_std_all": 0.9253793060779572,
      "rewards-max/forensics_sinkhorn_reward": 0.9879677593708038,
      "rewards-max/forensics_strict_edge_reward": 0.9291328191757202,
      "rewards-max/forensics_tiou_reward": 0.9970242083072662,
      "rewards/forensics_sinkhorn_reward": 0.7498134970664978,
      "rewards/forensics_strict_edge_reward": 0.1869787573814392,
      "rewards/forensics_tiou_reward": 0.5985825657844543,
      "step": 1110
    },
    {
      "completion_length": 24.34375,
      "epoch": 3.4086021505376345,
      "grad_norm": 13.325312614440918,
      "kl": 0.701171875,
      "learning_rate": 1.4615384615384617e-07,
      "loss": -0.0385,
      "reward": 1.7752536535263062,
      "reward_std": 0.42771418392658234,
      "reward_std_all": 0.7470544576644897,
      "rewards-max/forensics_sinkhorn_reward": 0.9896519780158997,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.8386017978191376,
      "rewards/forensics_strict_edge_reward": 0.22109727561473846,
      "rewards/forensics_tiou_reward": 0.7155545949935913,
      "step": 1111
    },
    {
      "completion_length": 23.859375,
      "epoch": 3.4116743471582183,
      "grad_norm": 13.440486907958984,
      "kl": 0.552734375,
      "learning_rate": 1.4538461538461538e-07,
      "loss": 0.042,
      "reward": 1.81099534034729,
      "reward_std": 0.281858891248703,
      "reward_std_all": 0.6769377589225769,
      "rewards-max/forensics_sinkhorn_reward": 0.9900084435939789,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.8655106425285339,
      "rewards/forensics_strict_edge_reward": 0.20209653303027153,
      "rewards/forensics_tiou_reward": 0.7433881461620331,
      "step": 1112
    },
    {
      "completion_length": 26.0625,
      "epoch": 3.4147465437788016,
      "grad_norm": 11.566437721252441,
      "kl": 0.4912109375,
      "learning_rate": 1.4461538461538462e-07,
      "loss": 0.029,
      "reward": 1.6440013647079468,
      "reward_std": 0.389120489358902,
      "reward_std_all": 0.7930750548839569,
      "rewards-max/forensics_sinkhorn_reward": 0.9884665906429291,
      "rewards-max/forensics_strict_edge_reward": 0.8021849691867828,
      "rewards-max/forensics_tiou_reward": 0.9982047379016876,
      "rewards/forensics_sinkhorn_reward": 0.790129154920578,
      "rewards/forensics_strict_edge_reward": 0.1507987603545189,
      "rewards/forensics_tiou_reward": 0.7030733823776245,
      "step": 1113
    },
    {
      "completion_length": 26.34375,
      "epoch": 3.4178187403993854,
      "grad_norm": 13.399849891662598,
      "kl": 0.572265625,
      "learning_rate": 1.4384615384615384e-07,
      "loss": -0.0759,
      "reward": 1.649340271949768,
      "reward_std": 0.4204103350639343,
      "reward_std_all": 0.7877168357372284,
      "rewards-max/forensics_sinkhorn_reward": 0.986625075340271,
      "rewards-max/forensics_strict_edge_reward": 0.8418737053871155,
      "rewards-max/forensics_tiou_reward": 0.9926213622093201,
      "rewards/forensics_sinkhorn_reward": 0.8071140050888062,
      "rewards/forensics_strict_edge_reward": 0.15051133930683136,
      "rewards/forensics_tiou_reward": 0.6917149424552917,
      "step": 1114
    },
    {
      "completion_length": 27.453125,
      "epoch": 3.420890937019969,
      "grad_norm": 12.739781379699707,
      "kl": 0.583984375,
      "learning_rate": 1.4307692307692308e-07,
      "loss": -0.0404,
      "reward": 1.82808917760849,
      "reward_std": 0.44984279572963715,
      "reward_std_all": 0.7187422215938568,
      "rewards-max/forensics_sinkhorn_reward": 0.9882856905460358,
      "rewards-max/forensics_strict_edge_reward": 0.882964164018631,
      "rewards-max/forensics_tiou_reward": 0.9960251450538635,
      "rewards/forensics_sinkhorn_reward": 0.839587926864624,
      "rewards/forensics_strict_edge_reward": 0.24248231947422028,
      "rewards/forensics_tiou_reward": 0.746019035577774,
      "step": 1115
    },
    {
      "completion_length": 22.21875,
      "epoch": 3.423963133640553,
      "grad_norm": 11.305578231811523,
      "kl": 0.4853515625,
      "learning_rate": 1.423076923076923e-07,
      "loss": -0.0072,
      "reward": 1.3780509233474731,
      "reward_std": 0.29688624292612076,
      "reward_std_all": 0.8830907344818115,
      "rewards-max/forensics_sinkhorn_reward": 0.9794265925884247,
      "rewards-max/forensics_strict_edge_reward": 0.8135445415973663,
      "rewards-max/forensics_tiou_reward": 0.976302832365036,
      "rewards/forensics_sinkhorn_reward": 0.7245838642120361,
      "rewards/forensics_strict_edge_reward": 0.09667876362800598,
      "rewards/forensics_tiou_reward": 0.5567883849143982,
      "step": 1116
    },
    {
      "completion_length": 26.953125,
      "epoch": 3.4270353302611367,
      "grad_norm": 13.46533489227295,
      "kl": 0.5498046875,
      "learning_rate": 1.4153846153846153e-07,
      "loss": -0.0405,
      "reward": 1.5641983151435852,
      "reward_std": 0.5204586088657379,
      "reward_std_all": 0.8829061985015869,
      "rewards-max/forensics_sinkhorn_reward": 0.9841407537460327,
      "rewards-max/forensics_strict_edge_reward": 0.7921086251735687,
      "rewards-max/forensics_tiou_reward": 0.9842648506164551,
      "rewards/forensics_sinkhorn_reward": 0.7710264623165131,
      "rewards/forensics_strict_edge_reward": 0.16597628220915794,
      "rewards/forensics_tiou_reward": 0.6271955370903015,
      "step": 1117
    },
    {
      "completion_length": 29.484375,
      "epoch": 3.4301075268817205,
      "grad_norm": 12.613755226135254,
      "kl": 0.443359375,
      "learning_rate": 1.4076923076923074e-07,
      "loss": -0.1755,
      "reward": 1.6371091604232788,
      "reward_std": 0.42644479870796204,
      "reward_std_all": 0.7824615836143494,
      "rewards-max/forensics_sinkhorn_reward": 0.9851782619953156,
      "rewards-max/forensics_strict_edge_reward": 0.8033190071582794,
      "rewards-max/forensics_tiou_reward": 0.9922517836093903,
      "rewards/forensics_sinkhorn_reward": 0.7848614752292633,
      "rewards/forensics_strict_edge_reward": 0.15444282442331314,
      "rewards/forensics_tiou_reward": 0.6978048980236053,
      "step": 1118
    },
    {
      "completion_length": 23.703125,
      "epoch": 3.4331797235023043,
      "grad_norm": 9.81672477722168,
      "kl": 0.52734375,
      "learning_rate": 1.4e-07,
      "loss": -0.0252,
      "reward": 1.414303481578827,
      "reward_std": 0.3949365168809891,
      "reward_std_all": 0.8893477320671082,
      "rewards-max/forensics_sinkhorn_reward": 0.984766960144043,
      "rewards-max/forensics_strict_edge_reward": 0.614451140165329,
      "rewards-max/forensics_tiou_reward": 0.989022821187973,
      "rewards/forensics_sinkhorn_reward": 0.7571079134941101,
      "rewards/forensics_strict_edge_reward": 0.08119909837841988,
      "rewards/forensics_tiou_reward": 0.575996458530426,
      "step": 1119
    },
    {
      "completion_length": 23.53125,
      "epoch": 3.436251920122888,
      "grad_norm": 12.270418167114258,
      "kl": 0.52734375,
      "learning_rate": 1.3923076923076922e-07,
      "loss": -0.0971,
      "reward": 1.6428664326667786,
      "reward_std": 0.45455655455589294,
      "reward_std_all": 0.812809407711029,
      "rewards-max/forensics_sinkhorn_reward": 0.9877571761608124,
      "rewards-max/forensics_strict_edge_reward": 0.8959447741508484,
      "rewards-max/forensics_tiou_reward": 0.9966091513633728,
      "rewards/forensics_sinkhorn_reward": 0.8036219477653503,
      "rewards/forensics_strict_edge_reward": 0.18300465494394302,
      "rewards/forensics_tiou_reward": 0.6562398076057434,
      "step": 1120
    },
    {
      "completion_length": 25.171875,
      "epoch": 3.439324116743472,
      "grad_norm": 13.515945434570312,
      "kl": 0.6484375,
      "learning_rate": 1.3846153846153846e-07,
      "loss": -0.0141,
      "reward": 1.5829830765724182,
      "reward_std": 0.5309537947177887,
      "reward_std_all": 0.8876586556434631,
      "rewards-max/forensics_sinkhorn_reward": 0.986222118139267,
      "rewards-max/forensics_strict_edge_reward": 0.8116492331027985,
      "rewards-max/forensics_tiou_reward": 0.9923321008682251,
      "rewards/forensics_sinkhorn_reward": 0.7634471356868744,
      "rewards/forensics_strict_edge_reward": 0.15547319501638412,
      "rewards/forensics_tiou_reward": 0.6640627384185791,
      "step": 1121
    },
    {
      "completion_length": 23.484375,
      "epoch": 3.442396313364055,
      "grad_norm": 12.077794075012207,
      "kl": 0.484375,
      "learning_rate": 1.376923076923077e-07,
      "loss": -0.1667,
      "reward": 1.6551879048347473,
      "reward_std": 0.34752415865659714,
      "reward_std_all": 0.7807277143001556,
      "rewards-max/forensics_sinkhorn_reward": 0.9895904362201691,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.8053683638572693,
      "rewards/forensics_strict_edge_reward": 0.1592235565185547,
      "rewards/forensics_tiou_reward": 0.6905959844589233,
      "step": 1122
    },
    {
      "completion_length": 29.328125,
      "epoch": 3.445468509984639,
      "grad_norm": 10.92894458770752,
      "kl": 0.458984375,
      "learning_rate": 1.3692307692307692e-07,
      "loss": 0.0021,
      "reward": 1.490752100944519,
      "reward_std": 0.30323660373687744,
      "reward_std_all": 0.8243692219257355,
      "rewards-max/forensics_sinkhorn_reward": 0.9880273342132568,
      "rewards-max/forensics_strict_edge_reward": 0.824172168970108,
      "rewards-max/forensics_tiou_reward": 0.9964379668235779,
      "rewards/forensics_sinkhorn_reward": 0.7127878665924072,
      "rewards/forensics_strict_edge_reward": 0.14569071680307388,
      "rewards/forensics_tiou_reward": 0.6322734951972961,
      "step": 1123
    },
    {
      "completion_length": 28.140625,
      "epoch": 3.4485407066052227,
      "grad_norm": 11.029149055480957,
      "kl": 0.580078125,
      "learning_rate": 1.3615384615384616e-07,
      "loss": -0.0756,
      "reward": 1.7372890710830688,
      "reward_std": 0.32831738144159317,
      "reward_std_all": 0.728963166475296,
      "rewards-max/forensics_sinkhorn_reward": 0.9893417358398438,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.8229904174804688,
      "rewards/forensics_strict_edge_reward": 0.18373605608940125,
      "rewards/forensics_tiou_reward": 0.7305626273155212,
      "step": 1124
    },
    {
      "completion_length": 23.9375,
      "epoch": 3.4516129032258065,
      "grad_norm": 15.835536003112793,
      "kl": 0.833984375,
      "learning_rate": 1.3538461538461537e-07,
      "loss": -0.0733,
      "reward": 1.700123369693756,
      "reward_std": 0.49488580226898193,
      "reward_std_all": 0.7941337823867798,
      "rewards-max/forensics_sinkhorn_reward": 0.9895764887332916,
      "rewards-max/forensics_strict_edge_reward": 0.9677534997463226,
      "rewards-max/forensics_tiou_reward": 0.9997292160987854,
      "rewards/forensics_sinkhorn_reward": 0.8178493976593018,
      "rewards/forensics_strict_edge_reward": 0.1770126149058342,
      "rewards/forensics_tiou_reward": 0.7052614092826843,
      "step": 1125
    },
    {
      "completion_length": 26.953125,
      "epoch": 3.4546850998463903,
      "grad_norm": 11.862466812133789,
      "kl": 0.583984375,
      "learning_rate": 1.346153846153846e-07,
      "loss": -0.0389,
      "reward": 2.0407090187072754,
      "reward_std": 0.3424091190099716,
      "reward_std_all": 0.6043398380279541,
      "rewards-max/forensics_sinkhorn_reward": 0.989499568939209,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.8911097049713135,
      "rewards/forensics_strict_edge_reward": 0.3381846994161606,
      "rewards/forensics_tiou_reward": 0.8114146292209625,
      "step": 1126
    },
    {
      "completion_length": 26.421875,
      "epoch": 3.457757296466974,
      "grad_norm": 12.171113967895508,
      "kl": 0.609375,
      "learning_rate": 1.3384615384615383e-07,
      "loss": -0.0252,
      "reward": 1.5288134813308716,
      "reward_std": 0.3692353069782257,
      "reward_std_all": 0.8341425657272339,
      "rewards-max/forensics_sinkhorn_reward": 0.9885491132736206,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.7511108219623566,
      "rewards/forensics_strict_edge_reward": 0.15707407891750336,
      "rewards/forensics_tiou_reward": 0.6206285953521729,
      "step": 1127
    },
    {
      "completion_length": 25.296875,
      "epoch": 3.460829493087558,
      "grad_norm": 13.427774429321289,
      "kl": 0.482421875,
      "learning_rate": 1.3307692307692307e-07,
      "loss": -0.0412,
      "reward": 1.6391311883926392,
      "reward_std": 0.29716189950704575,
      "reward_std_all": 0.7495703399181366,
      "rewards-max/forensics_sinkhorn_reward": 0.9890543222427368,
      "rewards-max/forensics_strict_edge_reward": 0.9677534997463226,
      "rewards-max/forensics_tiou_reward": 0.9986112117767334,
      "rewards/forensics_sinkhorn_reward": 0.7792252004146576,
      "rewards/forensics_strict_edge_reward": 0.1484030894935131,
      "rewards/forensics_tiou_reward": 0.7115028500556946,
      "step": 1128
    },
    {
      "completion_length": 23.28125,
      "epoch": 3.463901689708141,
      "grad_norm": 13.91609001159668,
      "kl": 0.54296875,
      "learning_rate": 1.323076923076923e-07,
      "loss": -0.0094,
      "reward": 1.8273693919181824,
      "reward_std": 0.3160388618707657,
      "reward_std_all": 0.667519599199295,
      "rewards-max/forensics_sinkhorn_reward": 0.9850244224071503,
      "rewards-max/forensics_strict_edge_reward": 0.8135445415973663,
      "rewards-max/forensics_tiou_reward": 0.9891660809516907,
      "rewards/forensics_sinkhorn_reward": 0.8999897539615631,
      "rewards/forensics_strict_edge_reward": 0.1389443799853325,
      "rewards/forensics_tiou_reward": 0.7884352803230286,
      "step": 1129
    },
    {
      "completion_length": 27.25,
      "epoch": 3.466973886328725,
      "grad_norm": 14.816498756408691,
      "kl": 0.830078125,
      "learning_rate": 1.3153846153846155e-07,
      "loss": -0.0513,
      "reward": 1.675107181072235,
      "reward_std": 0.4948301166296005,
      "reward_std_all": 0.8395196497440338,
      "rewards-max/forensics_sinkhorn_reward": 0.9894102811813354,
      "rewards-max/forensics_strict_edge_reward": 0.9677534997463226,
      "rewards-max/forensics_tiou_reward": 0.9997785985469818,
      "rewards/forensics_sinkhorn_reward": 0.770015299320221,
      "rewards/forensics_strict_edge_reward": 0.2207239866256714,
      "rewards/forensics_tiou_reward": 0.6843679547309875,
      "step": 1130
    },
    {
      "completion_length": 24.875,
      "epoch": 3.4700460829493087,
      "grad_norm": 14.129789352416992,
      "kl": 0.49609375,
      "learning_rate": 1.3076923076923076e-07,
      "loss": 0.0009,
      "reward": 1.6451680660247803,
      "reward_std": 0.2938474044203758,
      "reward_std_all": 0.7604873478412628,
      "rewards-max/forensics_sinkhorn_reward": 0.9882356524467468,
      "rewards-max/forensics_strict_edge_reward": 0.824172168970108,
      "rewards-max/forensics_tiou_reward": 0.9976365864276886,
      "rewards/forensics_sinkhorn_reward": 0.8043173849582672,
      "rewards/forensics_strict_edge_reward": 0.16379301995038986,
      "rewards/forensics_tiou_reward": 0.6770576238632202,
      "step": 1131
    },
    {
      "completion_length": 27.359375,
      "epoch": 3.4731182795698925,
      "grad_norm": 14.842601776123047,
      "kl": 0.60546875,
      "learning_rate": 1.3e-07,
      "loss": -0.0374,
      "reward": 1.616683006286621,
      "reward_std": 0.3936566114425659,
      "reward_std_all": 0.8395184576511383,
      "rewards-max/forensics_sinkhorn_reward": 0.989896684885025,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.7710470855236053,
      "rewards/forensics_strict_edge_reward": 0.1650940328836441,
      "rewards/forensics_tiou_reward": 0.6805418431758881,
      "step": 1132
    },
    {
      "completion_length": 25.578125,
      "epoch": 3.4761904761904763,
      "grad_norm": 10.895127296447754,
      "kl": 0.63671875,
      "learning_rate": 1.2923076923076924e-07,
      "loss": 0.0391,
      "reward": 1.7279826402664185,
      "reward_std": 0.4423050284385681,
      "reward_std_all": 0.7699055969715118,
      "rewards-max/forensics_sinkhorn_reward": 0.9888911247253418,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.827147901058197,
      "rewards/forensics_strict_edge_reward": 0.1989327073097229,
      "rewards/forensics_tiou_reward": 0.7019021213054657,
      "step": 1133
    },
    {
      "completion_length": 25.90625,
      "epoch": 3.47926267281106,
      "grad_norm": 15.729001998901367,
      "kl": 0.72265625,
      "learning_rate": 1.2846153846153845e-07,
      "loss": 0.0059,
      "reward": 1.7616754174232483,
      "reward_std": 0.4797807037830353,
      "reward_std_all": 0.7863225340843201,
      "rewards-max/forensics_sinkhorn_reward": 0.9888461232185364,
      "rewards-max/forensics_strict_edge_reward": 0.961620420217514,
      "rewards-max/forensics_tiou_reward": 0.9996077418327332,
      "rewards/forensics_sinkhorn_reward": 0.8194234371185303,
      "rewards/forensics_strict_edge_reward": 0.232997864484787,
      "rewards/forensics_tiou_reward": 0.7092540264129639,
      "step": 1134
    },
    {
      "completion_length": 27.375,
      "epoch": 3.482334869431644,
      "grad_norm": 12.421273231506348,
      "kl": 0.4873046875,
      "learning_rate": 1.276923076923077e-07,
      "loss": -0.0134,
      "reward": 1.5484832525253296,
      "reward_std": 0.39841708540916443,
      "reward_std_all": 0.8415254354476929,
      "rewards-max/forensics_sinkhorn_reward": 0.9884326756000519,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.8003116548061371,
      "rewards/forensics_strict_edge_reward": 0.12096516042947769,
      "rewards/forensics_tiou_reward": 0.6272064745426178,
      "step": 1135
    },
    {
      "completion_length": 25.328125,
      "epoch": 3.485407066052227,
      "grad_norm": 13.451766967773438,
      "kl": 0.755859375,
      "learning_rate": 1.269230769230769e-07,
      "loss": -0.134,
      "reward": 1.690973699092865,
      "reward_std": 0.44090868532657623,
      "reward_std_all": 0.8219662606716156,
      "rewards-max/forensics_sinkhorn_reward": 0.9903688132762909,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.7983319461345673,
      "rewards/forensics_strict_edge_reward": 0.23720648884773254,
      "rewards/forensics_tiou_reward": 0.6554353535175323,
      "step": 1136
    },
    {
      "completion_length": 22.640625,
      "epoch": 3.488479262672811,
      "grad_norm": 13.559475898742676,
      "kl": 0.482421875,
      "learning_rate": 1.2615384615384615e-07,
      "loss": -0.0473,
      "reward": 1.6404579877853394,
      "reward_std": 0.3227471560239792,
      "reward_std_all": 0.7622519135475159,
      "rewards-max/forensics_sinkhorn_reward": 0.9859722852706909,
      "rewards-max/forensics_strict_edge_reward": 0.7767926454544067,
      "rewards-max/forensics_tiou_reward": 0.9925091862678528,
      "rewards/forensics_sinkhorn_reward": 0.8271369636058807,
      "rewards/forensics_strict_edge_reward": 0.13399014435708523,
      "rewards/forensics_tiou_reward": 0.6793309450149536,
      "step": 1137
    },
    {
      "completion_length": 26.078125,
      "epoch": 3.4915514592933947,
      "grad_norm": 11.255064964294434,
      "kl": 0.5400390625,
      "learning_rate": 1.2538461538461536e-07,
      "loss": -0.0156,
      "reward": 1.8018214702606201,
      "reward_std": 0.4879678785800934,
      "reward_std_all": 0.7493935823440552,
      "rewards-max/forensics_sinkhorn_reward": 0.984768271446228,
      "rewards-max/forensics_strict_edge_reward": 0.8351600170135498,
      "rewards-max/forensics_tiou_reward": 0.990034818649292,
      "rewards/forensics_sinkhorn_reward": 0.8313169479370117,
      "rewards/forensics_strict_edge_reward": 0.24779746681451797,
      "rewards/forensics_tiou_reward": 0.7227070927619934,
      "step": 1138
    },
    {
      "completion_length": 25.765625,
      "epoch": 3.4946236559139785,
      "grad_norm": 11.746169090270996,
      "kl": 0.5078125,
      "learning_rate": 1.246153846153846e-07,
      "loss": -0.1396,
      "reward": 1.8397157192230225,
      "reward_std": 0.4511640667915344,
      "reward_std_all": 0.7235228419303894,
      "rewards-max/forensics_sinkhorn_reward": 0.9896633625030518,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.858871728181839,
      "rewards/forensics_strict_edge_reward": 0.24903114140033722,
      "rewards/forensics_tiou_reward": 0.7318128943443298,
      "step": 1139
    },
    {
      "completion_length": 26.640625,
      "epoch": 3.4976958525345623,
      "grad_norm": 17.308361053466797,
      "kl": 0.6875,
      "learning_rate": 1.2384615384615384e-07,
      "loss": -0.1355,
      "reward": 1.725195586681366,
      "reward_std": 0.4197942167520523,
      "reward_std_all": 0.7776007056236267,
      "rewards-max/forensics_sinkhorn_reward": 0.9901740849018097,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.8090746402740479,
      "rewards/forensics_strict_edge_reward": 0.2293211966753006,
      "rewards/forensics_tiou_reward": 0.6867996454238892,
      "step": 1140
    },
    {
      "completion_length": 22.453125,
      "epoch": 3.500768049155146,
      "grad_norm": 12.341349601745605,
      "kl": 0.421875,
      "learning_rate": 1.2307692307692308e-07,
      "loss": -0.0052,
      "reward": 1.5675618052482605,
      "reward_std": 0.39988645911216736,
      "reward_std_all": 0.8241479992866516,
      "rewards-max/forensics_sinkhorn_reward": 0.9871650636196136,
      "rewards-max/forensics_strict_edge_reward": 0.8959447741508484,
      "rewards-max/forensics_tiou_reward": 0.9941001534461975,
      "rewards/forensics_sinkhorn_reward": 0.7578139007091522,
      "rewards/forensics_strict_edge_reward": 0.15303147584199905,
      "rewards/forensics_tiou_reward": 0.656716376543045,
      "step": 1141
    },
    {
      "completion_length": 26.59375,
      "epoch": 3.50384024577573,
      "grad_norm": 14.568336486816406,
      "kl": 0.5869140625,
      "learning_rate": 1.223076923076923e-07,
      "loss": -0.1099,
      "reward": 1.9465201497077942,
      "reward_std": 0.4046171307563782,
      "reward_std_all": 0.6565993130207062,
      "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": 0.8648771047592163,
      "rewards/forensics_strict_edge_reward": 0.26382797956466675,
      "rewards/forensics_tiou_reward": 0.8178150057792664,
      "step": 1142
    },
    {
      "completion_length": 25.78125,
      "epoch": 3.506912442396313,
      "grad_norm": 15.13188648223877,
      "kl": 0.912109375,
      "learning_rate": 1.2153846153846154e-07,
      "loss": -0.0228,
      "reward": 1.8214802742004395,
      "reward_std": 0.48972833156585693,
      "reward_std_all": 0.7543880939483643,
      "rewards-max/forensics_sinkhorn_reward": 0.9893964231014252,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.8559969365596771,
      "rewards/forensics_strict_edge_reward": 0.22572508454322815,
      "rewards/forensics_tiou_reward": 0.7397581934928894,
      "step": 1143
    },
    {
      "completion_length": 23.640625,
      "epoch": 3.509984639016897,
      "grad_norm": 12.603677749633789,
      "kl": 0.56640625,
      "learning_rate": 1.2076923076923075e-07,
      "loss": -0.0501,
      "reward": 1.648651897907257,
      "reward_std": 0.43021051585674286,
      "reward_std_all": 0.7977335751056671,
      "rewards-max/forensics_sinkhorn_reward": 0.9882001578807831,
      "rewards-max/forensics_strict_edge_reward": 0.7719359695911407,
      "rewards-max/forensics_tiou_reward": 0.9969958066940308,
      "rewards/forensics_sinkhorn_reward": 0.8208649456501007,
      "rewards/forensics_strict_edge_reward": 0.1700505055487156,
      "rewards/forensics_tiou_reward": 0.6577363908290863,
      "step": 1144
    },
    {
      "completion_length": 25.6875,
      "epoch": 3.5130568356374807,
      "grad_norm": 11.660343170166016,
      "kl": 0.5537109375,
      "learning_rate": 1.2e-07,
      "loss": -0.044,
      "reward": 1.9139606356620789,
      "reward_std": 0.5628954172134399,
      "reward_std_all": 0.7481600642204285,
      "rewards-max/forensics_sinkhorn_reward": 0.9898345768451691,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.868200421333313,
      "rewards/forensics_strict_edge_reward": 0.27801431715488434,
      "rewards/forensics_tiou_reward": 0.7677457928657532,
      "step": 1145
    },
    {
      "completion_length": 25.28125,
      "epoch": 3.5161290322580645,
      "grad_norm": 11.554789543151855,
      "kl": 0.478515625,
      "learning_rate": 1.1923076923076923e-07,
      "loss": -0.0994,
      "reward": 1.6055716276168823,
      "reward_std": 0.3074697405099869,
      "reward_std_all": 0.7776921689510345,
      "rewards-max/forensics_sinkhorn_reward": 0.9888469278812408,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.761028915643692,
      "rewards/forensics_strict_edge_reward": 0.1912005990743637,
      "rewards/forensics_tiou_reward": 0.6533421576023102,
      "step": 1146
    },
    {
      "completion_length": 27.984375,
      "epoch": 3.5192012288786483,
      "grad_norm": 12.842544555664062,
      "kl": 0.6875,
      "learning_rate": 1.1846153846153846e-07,
      "loss": -0.0701,
      "reward": 1.6858689188957214,
      "reward_std": 0.3732263147830963,
      "reward_std_all": 0.7696755826473236,
      "rewards-max/forensics_sinkhorn_reward": 0.9887555539608002,
      "rewards-max/forensics_strict_edge_reward": 0.9093653857707977,
      "rewards-max/forensics_tiou_reward": 0.9976995587348938,
      "rewards/forensics_sinkhorn_reward": 0.831599771976471,
      "rewards/forensics_strict_edge_reward": 0.13400724157691002,
      "rewards/forensics_tiou_reward": 0.7202619016170502,
      "step": 1147
    },
    {
      "completion_length": 29.96875,
      "epoch": 3.522273425499232,
      "grad_norm": 10.691892623901367,
      "kl": 0.650390625,
      "learning_rate": 1.1769230769230768e-07,
      "loss": -0.0751,
      "reward": 1.5906589031219482,
      "reward_std": 0.35623079538345337,
      "reward_std_all": 0.8101336061954498,
      "rewards-max/forensics_sinkhorn_reward": 0.9875377714633942,
      "rewards-max/forensics_strict_edge_reward": 0.8582656681537628,
      "rewards-max/forensics_tiou_reward": 0.9933891892433167,
      "rewards/forensics_sinkhorn_reward": 0.7656912207603455,
      "rewards/forensics_strict_edge_reward": 0.14366362988948822,
      "rewards/forensics_tiou_reward": 0.6813039183616638,
      "step": 1148
    },
    {
      "completion_length": 27.171875,
      "epoch": 3.525345622119816,
      "grad_norm": 11.816819190979004,
      "kl": 0.4658203125,
      "learning_rate": 1.1692307692307691e-07,
      "loss": -0.0624,
      "reward": 1.6359845399856567,
      "reward_std": 0.40853795409202576,
      "reward_std_all": 0.8006953001022339,
      "rewards-max/forensics_sinkhorn_reward": 0.9868400692939758,
      "rewards-max/forensics_strict_edge_reward": 0.9375866651535034,
      "rewards-max/forensics_tiou_reward": 0.9947711825370789,
      "rewards/forensics_sinkhorn_reward": 0.776138573884964,
      "rewards/forensics_strict_edge_reward": 0.20381945371627808,
      "rewards/forensics_tiou_reward": 0.6560265421867371,
      "step": 1149
    },
    {
      "completion_length": 27.890625,
      "epoch": 3.528417818740399,
      "grad_norm": 127.59810638427734,
      "kl": 3.419921875,
      "learning_rate": 1.1615384615384615e-07,
      "loss": -0.0009,
      "reward": 1.58651202917099,
      "reward_std": 0.43374475836753845,
      "reward_std_all": 0.8405616879463196,
      "rewards-max/forensics_sinkhorn_reward": 0.9888378381729126,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.7168596386909485,
      "rewards/forensics_strict_edge_reward": 0.18472550436854362,
      "rewards/forensics_tiou_reward": 0.684926837682724,
      "step": 1150
    },
    {
      "completion_length": 25.484375,
      "epoch": 3.531490015360983,
      "grad_norm": 12.823586463928223,
      "kl": 0.587890625,
      "learning_rate": 1.1538461538461539e-07,
      "loss": -0.0606,
      "reward": 1.7535873651504517,
      "reward_std": 0.34340573847293854,
      "reward_std_all": 0.722620964050293,
      "rewards-max/forensics_sinkhorn_reward": 0.9887247681617737,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.837959885597229,
      "rewards/forensics_strict_edge_reward": 0.1979673206806183,
      "rewards/forensics_tiou_reward": 0.717660129070282,
      "step": 1151
    },
    {
      "completion_length": 22.78125,
      "epoch": 3.5345622119815667,
      "grad_norm": 12.000078201293945,
      "kl": 0.5087890625,
      "learning_rate": 1.1461538461538462e-07,
      "loss": -0.0379,
      "reward": 1.5402610898017883,
      "reward_std": 0.2945355772972107,
      "reward_std_all": 0.7928120791912079,
      "rewards-max/forensics_sinkhorn_reward": 0.9862715303897858,
      "rewards-max/forensics_strict_edge_reward": 0.7654097378253937,
      "rewards-max/forensics_tiou_reward": 0.989380419254303,
      "rewards/forensics_sinkhorn_reward": 0.763308972120285,
      "rewards/forensics_strict_edge_reward": 0.12672272883355618,
      "rewards/forensics_tiou_reward": 0.6502293944358826,
      "step": 1152
    },
    {
      "completion_length": 26.828125,
      "epoch": 3.5376344086021505,
      "grad_norm": 13.44650650024414,
      "kl": 0.642578125,
      "learning_rate": 1.1384615384615384e-07,
      "loss": -0.082,
      "reward": 1.6255317330360413,
      "reward_std": 0.37322747707366943,
      "reward_std_all": 0.7995437979698181,
      "rewards-max/forensics_sinkhorn_reward": 0.9899567365646362,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.7697484493255615,
      "rewards/forensics_strict_edge_reward": 0.15106834471225739,
      "rewards/forensics_tiou_reward": 0.7047150135040283,
      "step": 1153
    },
    {
      "completion_length": 23.875,
      "epoch": 3.5407066052227343,
      "grad_norm": 12.516221046447754,
      "kl": 0.498046875,
      "learning_rate": 1.1307692307692307e-07,
      "loss": -0.1015,
      "reward": 1.8841911554336548,
      "reward_std": 0.35480089485645294,
      "reward_std_all": 0.6687526106834412,
      "rewards-max/forensics_sinkhorn_reward": 0.9890156686306,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.8858851492404938,
      "rewards/forensics_strict_edge_reward": 0.22977693378925323,
      "rewards/forensics_tiou_reward": 0.7685289978981018,
      "step": 1154
    },
    {
      "completion_length": 25.890625,
      "epoch": 3.543778801843318,
      "grad_norm": 12.728917121887207,
      "kl": 0.533203125,
      "learning_rate": 1.123076923076923e-07,
      "loss": -0.0217,
      "reward": 1.8122979402542114,
      "reward_std": 0.34192679822444916,
      "reward_std_all": 0.7243395149707794,
      "rewards-max/forensics_sinkhorn_reward": 0.9892472326755524,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.8571378290653229,
      "rewards/forensics_strict_edge_reward": 0.2232351079583168,
      "rewards/forensics_tiou_reward": 0.7319249510765076,
      "step": 1155
    },
    {
      "completion_length": 25.90625,
      "epoch": 3.546850998463902,
      "grad_norm": 13.251351356506348,
      "kl": 0.662109375,
      "learning_rate": 1.1153846153846154e-07,
      "loss": -0.1341,
      "reward": 1.746060311794281,
      "reward_std": 0.33487993478775024,
      "reward_std_all": 0.7121981382369995,
      "rewards-max/forensics_sinkhorn_reward": 0.9884612858295441,
      "rewards-max/forensics_strict_edge_reward": 0.9762093424797058,
      "rewards-max/forensics_tiou_reward": 0.9996749758720398,
      "rewards/forensics_sinkhorn_reward": 0.8555421233177185,
      "rewards/forensics_strict_edge_reward": 0.1666480079293251,
      "rewards/forensics_tiou_reward": 0.7238702476024628,
      "step": 1156
    },
    {
      "completion_length": 28.234375,
      "epoch": 3.549923195084485,
      "grad_norm": 16.447874069213867,
      "kl": 0.962890625,
      "learning_rate": 1.1076923076923077e-07,
      "loss": 0.0107,
      "reward": 1.9335364699363708,
      "reward_std": 0.510432168841362,
      "reward_std_all": 0.6966236531734467,
      "rewards-max/forensics_sinkhorn_reward": 0.9901235103607178,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.8692984282970428,
      "rewards/forensics_strict_edge_reward": 0.2787676155567169,
      "rewards/forensics_tiou_reward": 0.7854704260826111,
      "step": 1157
    },
    {
      "completion_length": 25.78125,
      "epoch": 3.5529953917050694,
      "grad_norm": 16.993064880371094,
      "kl": 0.85546875,
      "learning_rate": 1.0999999999999999e-07,
      "loss": 0.0014,
      "reward": 1.829491674900055,
      "reward_std": 0.3743892014026642,
      "reward_std_all": 0.7173265516757965,
      "rewards-max/forensics_sinkhorn_reward": 0.9899958968162537,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.8162793517112732,
      "rewards/forensics_strict_edge_reward": 0.30145568400621414,
      "rewards/forensics_tiou_reward": 0.7117566466331482,
      "step": 1158
    },
    {
      "completion_length": 23.671875,
      "epoch": 3.5560675883256527,
      "grad_norm": 11.942984580993652,
      "kl": 0.556640625,
      "learning_rate": 1.0923076923076922e-07,
      "loss": -0.1474,
      "reward": 1.7235034704208374,
      "reward_std": 0.389600932598114,
      "reward_std_all": 0.739886999130249,
      "rewards-max/forensics_sinkhorn_reward": 0.9886277616024017,
      "rewards-max/forensics_strict_edge_reward": 0.8968863189220428,
      "rewards-max/forensics_tiou_reward": 0.9974812567234039,
      "rewards/forensics_sinkhorn_reward": 0.8286352753639221,
      "rewards/forensics_strict_edge_reward": 0.19336236268281937,
      "rewards/forensics_tiou_reward": 0.7015058994293213,
      "step": 1159
    },
    {
      "completion_length": 29.21875,
      "epoch": 3.5591397849462365,
      "grad_norm": 13.535271644592285,
      "kl": 0.646484375,
      "learning_rate": 1.0846153846153845e-07,
      "loss": 0.0416,
      "reward": 1.5492753386497498,
      "reward_std": 0.3324538394808769,
      "reward_std_all": 0.8183571994304657,
      "rewards-max/forensics_sinkhorn_reward": 0.9889281988143921,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.7714599072933197,
      "rewards/forensics_strict_edge_reward": 0.13654711097478867,
      "rewards/forensics_tiou_reward": 0.6412683129310608,
      "step": 1160
    },
    {
      "completion_length": 22.015625,
      "epoch": 3.5622119815668203,
      "grad_norm": 12.664624214172363,
      "kl": 0.51953125,
      "learning_rate": 1.076923076923077e-07,
      "loss": -0.0319,
      "reward": 1.7355290055274963,
      "reward_std": 0.45909738540649414,
      "reward_std_all": 0.7644559442996979,
      "rewards-max/forensics_sinkhorn_reward": 0.984457403421402,
      "rewards-max/forensics_strict_edge_reward": 0.8419698476791382,
      "rewards-max/forensics_tiou_reward": 0.9881289303302765,
      "rewards/forensics_sinkhorn_reward": 0.8318996131420135,
      "rewards/forensics_strict_edge_reward": 0.1824638769030571,
      "rewards/forensics_tiou_reward": 0.7211655378341675,
      "step": 1161
    },
    {
      "completion_length": 24.890625,
      "epoch": 3.565284178187404,
      "grad_norm": 12.19424057006836,
      "kl": 0.4970703125,
      "learning_rate": 1.0692307692307693e-07,
      "loss": -0.0294,
      "reward": 1.8207327127456665,
      "reward_std": 0.2257574424147606,
      "reward_std_all": 0.6569190621376038,
      "rewards-max/forensics_sinkhorn_reward": 0.9893882870674133,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.8796364963054657,
      "rewards/forensics_strict_edge_reward": 0.1715782731771469,
      "rewards/forensics_tiou_reward": 0.7695178687572479,
      "step": 1162
    },
    {
      "completion_length": 22.15625,
      "epoch": 3.568356374807988,
      "grad_norm": 11.08520793914795,
      "kl": 0.4638671875,
      "learning_rate": 1.0615384615384615e-07,
      "loss": -0.0292,
      "reward": 1.660781741142273,
      "reward_std": 0.3867235481739044,
      "reward_std_all": 0.7839192450046539,
      "rewards-max/forensics_sinkhorn_reward": 0.9771674573421478,
      "rewards-max/forensics_strict_edge_reward": 0.8032653331756592,
      "rewards-max/forensics_tiou_reward": 0.9717675447463989,
      "rewards/forensics_sinkhorn_reward": 0.798366904258728,
      "rewards/forensics_strict_edge_reward": 0.18613766133785248,
      "rewards/forensics_tiou_reward": 0.6762771606445312,
      "step": 1163
    },
    {
      "completion_length": 24.109375,
      "epoch": 3.571428571428571,
      "grad_norm": 13.029989242553711,
      "kl": 0.4990234375,
      "learning_rate": 1.0538461538461538e-07,
      "loss": -0.0447,
      "reward": 1.6522401571273804,
      "reward_std": 0.3849586844444275,
      "reward_std_all": 0.7945069670677185,
      "rewards-max/forensics_sinkhorn_reward": 0.9847553074359894,
      "rewards-max/forensics_strict_edge_reward": 0.8351600170135498,
      "rewards-max/forensics_tiou_reward": 0.9877548515796661,
      "rewards/forensics_sinkhorn_reward": 0.7996029555797577,
      "rewards/forensics_strict_edge_reward": 0.1930166259407997,
      "rewards/forensics_tiou_reward": 0.6596205532550812,
      "step": 1164
    },
    {
      "completion_length": 32.109375,
      "epoch": 3.5745007680491554,
      "grad_norm": 13.102959632873535,
      "kl": 0.548828125,
      "learning_rate": 1.0461538461538461e-07,
      "loss": 0.0041,
      "reward": 1.7624331712722778,
      "reward_std": 0.5173361301422119,
      "reward_std_all": 0.8161836564540863,
      "rewards-max/forensics_sinkhorn_reward": 0.9899047911167145,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.7616400718688965,
      "rewards/forensics_strict_edge_reward": 0.2650938928127289,
      "rewards/forensics_tiou_reward": 0.7356992363929749,
      "step": 1165
    },
    {
      "completion_length": 23.703125,
      "epoch": 3.5775729646697387,
      "grad_norm": 16.007770538330078,
      "kl": 0.5419921875,
      "learning_rate": 1.0384615384615385e-07,
      "loss": -0.0834,
      "reward": 1.5792414546012878,
      "reward_std": 0.4243403673171997,
      "reward_std_all": 0.8463053107261658,
      "rewards-max/forensics_sinkhorn_reward": 0.987288773059845,
      "rewards-max/forensics_strict_edge_reward": 0.756708562374115,
      "rewards-max/forensics_tiou_reward": 0.9956519901752472,
      "rewards/forensics_sinkhorn_reward": 0.7933842241764069,
      "rewards/forensics_strict_edge_reward": 0.170036181807518,
      "rewards/forensics_tiou_reward": 0.6158210933208466,
      "step": 1166
    },
    {
      "completion_length": 23.34375,
      "epoch": 3.5806451612903225,
      "grad_norm": 14.248398780822754,
      "kl": 0.70703125,
      "learning_rate": 1.0307692307692307e-07,
      "loss": -0.0907,
      "reward": 1.735921025276184,
      "reward_std": 0.49534082412719727,
      "reward_std_all": 0.8406427800655365,
      "rewards-max/forensics_sinkhorn_reward": 0.9887205064296722,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.792533665895462,
      "rewards/forensics_strict_edge_reward": 0.25683846324682236,
      "rewards/forensics_tiou_reward": 0.6865488886833191,
      "step": 1167
    },
    {
      "completion_length": 28.625,
      "epoch": 3.5837173579109063,
      "grad_norm": 12.572875022888184,
      "kl": 0.70703125,
      "learning_rate": 1.023076923076923e-07,
      "loss": 0.0769,
      "reward": 1.6163424253463745,
      "reward_std": 0.5162443965673447,
      "reward_std_all": 0.8417775928974152,
      "rewards-max/forensics_sinkhorn_reward": 0.9879305362701416,
      "rewards-max/forensics_strict_edge_reward": 0.8582656681537628,
      "rewards-max/forensics_tiou_reward": 0.9963500797748566,
      "rewards/forensics_sinkhorn_reward": 0.7835960984230042,
      "rewards/forensics_strict_edge_reward": 0.16297589242458344,
      "rewards/forensics_tiou_reward": 0.6697705388069153,
      "step": 1168
    },
    {
      "completion_length": 27.796875,
      "epoch": 3.58678955453149,
      "grad_norm": 12.507611274719238,
      "kl": 0.4990234375,
      "learning_rate": 1.0153846153846153e-07,
      "loss": -0.1333,
      "reward": 1.912856936454773,
      "reward_std": 0.40635594725608826,
      "reward_std_all": 0.6838742792606354,
      "rewards-max/forensics_sinkhorn_reward": 0.9902856647968292,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.8641573786735535,
      "rewards/forensics_strict_edge_reward": 0.26641808450222015,
      "rewards/forensics_tiou_reward": 0.7822814583778381,
      "step": 1169
    },
    {
      "completion_length": 24.40625,
      "epoch": 3.589861751152074,
      "grad_norm": 14.27853012084961,
      "kl": 0.5712890625,
      "learning_rate": 1.0076923076923076e-07,
      "loss": 0.0361,
      "reward": 1.779197096824646,
      "reward_std": 0.30864711105823517,
      "reward_std_all": 0.7161747813224792,
      "rewards-max/forensics_sinkhorn_reward": 0.9897510409355164,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.8576378524303436,
      "rewards/forensics_strict_edge_reward": 0.19886644184589386,
      "rewards/forensics_tiou_reward": 0.7226928472518921,
      "step": 1170
    },
    {
      "completion_length": 24.25,
      "epoch": 3.592933947772657,
      "grad_norm": 14.002820014953613,
      "kl": 0.642578125,
      "learning_rate": 1e-07,
      "loss": -0.1617,
      "reward": 1.8371883630752563,
      "reward_std": 0.44770191609859467,
      "reward_std_all": 0.7382879555225372,
      "rewards-max/forensics_sinkhorn_reward": 0.9902081787586212,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.8564531803131104,
      "rewards/forensics_strict_edge_reward": 0.2577440142631531,
      "rewards/forensics_tiou_reward": 0.7229910492897034,
      "step": 1171
    },
    {
      "completion_length": 24.421875,
      "epoch": 3.5960061443932414,
      "grad_norm": 14.742834091186523,
      "kl": 0.60546875,
      "learning_rate": 9.923076923076922e-08,
      "loss": -0.1366,
      "reward": 1.6985927820205688,
      "reward_std": 0.6220166087150574,
      "reward_std_all": 0.8704561591148376,
      "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": 0.8113567233085632,
      "rewards/forensics_strict_edge_reward": 0.2061689868569374,
      "rewards/forensics_tiou_reward": 0.6810671389102936,
      "step": 1172
    },
    {
      "completion_length": 23.484375,
      "epoch": 3.5990783410138247,
      "grad_norm": 15.230729103088379,
      "kl": 0.689453125,
      "learning_rate": 9.846153846153846e-08,
      "loss": -0.1541,
      "reward": 1.7085010409355164,
      "reward_std": 0.6032043099403381,
      "reward_std_all": 0.8356241881847382,
      "rewards-max/forensics_sinkhorn_reward": 0.9895322620868683,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.8018685877323151,
      "rewards/forensics_strict_edge_reward": 0.20276688039302826,
      "rewards/forensics_tiou_reward": 0.7038656771183014,
      "step": 1173
    },
    {
      "completion_length": 24.09375,
      "epoch": 3.6021505376344085,
      "grad_norm": 11.954303741455078,
      "kl": 0.615234375,
      "learning_rate": 9.769230769230769e-08,
      "loss": -0.0388,
      "reward": 1.6982213854789734,
      "reward_std": 0.4817870706319809,
      "reward_std_all": 0.8038235008716583,
      "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": 0.7915919125080109,
      "rewards/forensics_strict_edge_reward": 0.22503432631492615,
      "rewards/forensics_tiou_reward": 0.6815951764583588,
      "step": 1174
    },
    {
      "completion_length": 23.75,
      "epoch": 3.6052227342549923,
      "grad_norm": 11.80324649810791,
      "kl": 0.658203125,
      "learning_rate": 9.692307692307692e-08,
      "loss": -0.1057,
      "reward": 1.9491273164749146,
      "reward_std": 0.48255524039268494,
      "reward_std_all": 0.7085586786270142,
      "rewards-max/forensics_sinkhorn_reward": 0.9896251857280731,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.8526409268379211,
      "rewards/forensics_strict_edge_reward": 0.31722763180732727,
      "rewards/forensics_tiou_reward": 0.7792588472366333,
      "step": 1175
    },
    {
      "completion_length": 25.8125,
      "epoch": 3.608294930875576,
      "grad_norm": 11.80926513671875,
      "kl": 0.580078125,
      "learning_rate": 9.615384615384616e-08,
      "loss": -0.064,
      "reward": 1.8079134225845337,
      "reward_std": 0.5663247406482697,
      "reward_std_all": 0.8026162981987,
      "rewards-max/forensics_sinkhorn_reward": 0.9901891350746155,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.8480920493602753,
      "rewards/forensics_strict_edge_reward": 0.24174290895462036,
      "rewards/forensics_tiou_reward": 0.7180784344673157,
      "step": 1176
    },
    {
      "completion_length": 26.53125,
      "epoch": 3.61136712749616,
      "grad_norm": 11.281105041503906,
      "kl": 0.4384765625,
      "learning_rate": 9.538461538461538e-08,
      "loss": 0.0059,
      "reward": 1.7375059127807617,
      "reward_std": 0.405622735619545,
      "reward_std_all": 0.7629324197769165,
      "rewards-max/forensics_sinkhorn_reward": 0.9888167381286621,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.8351354002952576,
      "rewards/forensics_strict_edge_reward": 0.18792724609375,
      "rewards/forensics_tiou_reward": 0.7144432961940765,
      "step": 1177
    },
    {
      "completion_length": 26.71875,
      "epoch": 3.6144393241167436,
      "grad_norm": 11.535161972045898,
      "kl": 0.720703125,
      "learning_rate": 9.461538461538461e-08,
      "loss": -0.0113,
      "reward": 1.6226998567581177,
      "reward_std": 0.3202830106019974,
      "reward_std_all": 0.7649396359920502,
      "rewards-max/forensics_sinkhorn_reward": 0.9890427887439728,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.7805852293968201,
      "rewards/forensics_strict_edge_reward": 0.19114084541797638,
      "rewards/forensics_tiou_reward": 0.6509738266468048,
      "step": 1178
    },
    {
      "completion_length": 27.375,
      "epoch": 3.6175115207373274,
      "grad_norm": 13.345359802246094,
      "kl": 0.4560546875,
      "learning_rate": 9.384615384615384e-08,
      "loss": 0.0103,
      "reward": 1.8403555154800415,
      "reward_std": 0.3836069852113724,
      "reward_std_all": 0.6918330788612366,
      "rewards-max/forensics_sinkhorn_reward": 0.9892373085021973,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.8538307249546051,
      "rewards/forensics_strict_edge_reward": 0.20973606407642365,
      "rewards/forensics_tiou_reward": 0.7767886221408844,
      "step": 1179
    },
    {
      "completion_length": 25.28125,
      "epoch": 3.6205837173579107,
      "grad_norm": 11.240792274475098,
      "kl": 0.4677734375,
      "learning_rate": 9.307692307692306e-08,
      "loss": -0.1455,
      "reward": 1.567114531993866,
      "reward_std": 0.40401311218738556,
      "reward_std_all": 0.8329024314880371,
      "rewards-max/forensics_sinkhorn_reward": 0.9846928119659424,
      "rewards-max/forensics_strict_edge_reward": 0.8135445415973663,
      "rewards-max/forensics_tiou_reward": 0.9869040846824646,
      "rewards/forensics_sinkhorn_reward": 0.7884348034858704,
      "rewards/forensics_strict_edge_reward": 0.12814119085669518,
      "rewards/forensics_tiou_reward": 0.6505384743213654,
      "step": 1180
    },
    {
      "completion_length": 23.40625,
      "epoch": 3.6236559139784945,
      "grad_norm": 12.936662673950195,
      "kl": 0.58984375,
      "learning_rate": 9.23076923076923e-08,
      "loss": 0.0253,
      "reward": 1.7624660730361938,
      "reward_std": 0.34907446801662445,
      "reward_std_all": 0.7207514345645905,
      "rewards-max/forensics_sinkhorn_reward": 0.9889231026172638,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.8473931550979614,
      "rewards/forensics_strict_edge_reward": 0.1896081417798996,
      "rewards/forensics_tiou_reward": 0.725464791059494,
      "step": 1181
    },
    {
      "completion_length": 23.890625,
      "epoch": 3.6267281105990783,
      "grad_norm": 12.954031944274902,
      "kl": 0.4609375,
      "learning_rate": 9.153846153846153e-08,
      "loss": -0.0424,
      "reward": 1.7186020612716675,
      "reward_std": 0.43017007410526276,
      "reward_std_all": 0.7759437263011932,
      "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": 0.8343841433525085,
      "rewards/forensics_strict_edge_reward": 0.20848727226257324,
      "rewards/forensics_tiou_reward": 0.6757305860519409,
      "step": 1182
    },
    {
      "completion_length": 25.859375,
      "epoch": 3.629800307219662,
      "grad_norm": 11.685661315917969,
      "kl": 0.564453125,
      "learning_rate": 9.076923076923076e-08,
      "loss": 0.0006,
      "reward": 1.7690773010253906,
      "reward_std": 0.5101606845855713,
      "reward_std_all": 0.7813937067985535,
      "rewards-max/forensics_sinkhorn_reward": 0.9894302487373352,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.8333250880241394,
      "rewards/forensics_strict_edge_reward": 0.20271805673837662,
      "rewards/forensics_tiou_reward": 0.7330341935157776,
      "step": 1183
    },
    {
      "completion_length": 28.359375,
      "epoch": 3.632872503840246,
      "grad_norm": 13.791702270507812,
      "kl": 0.896484375,
      "learning_rate": 9e-08,
      "loss": -0.0285,
      "reward": 1.6451783776283264,
      "reward_std": 0.49694450199604034,
      "reward_std_all": 0.8210523426532745,
      "rewards-max/forensics_sinkhorn_reward": 0.9898709654808044,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.7857419848442078,
      "rewards/forensics_strict_edge_reward": 0.17125405371189117,
      "rewards/forensics_tiou_reward": 0.6881822943687439,
      "step": 1184
    },
    {
      "completion_length": 27.171875,
      "epoch": 3.6359447004608296,
      "grad_norm": 13.154787063598633,
      "kl": 0.654296875,
      "learning_rate": 8.923076923076923e-08,
      "loss": -0.1015,
      "reward": 1.5016579031944275,
      "reward_std": 0.534875363111496,
      "reward_std_all": 0.9213082194328308,
      "rewards-max/forensics_sinkhorn_reward": 0.987248957157135,
      "rewards-max/forensics_strict_edge_reward": 0.8351600170135498,
      "rewards-max/forensics_tiou_reward": 0.9934251010417938,
      "rewards/forensics_sinkhorn_reward": 0.7333764731884003,
      "rewards/forensics_strict_edge_reward": 0.16399423778057098,
      "rewards/forensics_tiou_reward": 0.6042871177196503,
      "step": 1185
    },
    {
      "completion_length": 25.859375,
      "epoch": 3.6390168970814134,
      "grad_norm": 11.814722061157227,
      "kl": 0.5634765625,
      "learning_rate": 8.846153846153847e-08,
      "loss": -0.0845,
      "reward": 1.8134058117866516,
      "reward_std": 0.38148699700832367,
      "reward_std_all": 0.7356133759021759,
      "rewards-max/forensics_sinkhorn_reward": 0.9895815253257751,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.8442012965679169,
      "rewards/forensics_strict_edge_reward": 0.23703204095363617,
      "rewards/forensics_tiou_reward": 0.7321724593639374,
      "step": 1186
    },
    {
      "completion_length": 28.890625,
      "epoch": 3.6420890937019967,
      "grad_norm": 13.447160720825195,
      "kl": 1.060546875,
      "learning_rate": 8.769230769230769e-08,
      "loss": -0.0181,
      "reward": 1.6323469281196594,
      "reward_std": 0.4353943318128586,
      "reward_std_all": 0.8289458453655243,
      "rewards-max/forensics_sinkhorn_reward": 0.9886393547058105,
      "rewards-max/forensics_strict_edge_reward": 0.9375866651535034,
      "rewards-max/forensics_tiou_reward": 0.9973476827144623,
      "rewards/forensics_sinkhorn_reward": 0.7877555191516876,
      "rewards/forensics_strict_edge_reward": 0.20352152734994888,
      "rewards/forensics_tiou_reward": 0.6410698592662811,
      "step": 1187
    },
    {
      "completion_length": 28.5625,
      "epoch": 3.6451612903225805,
      "grad_norm": 12.548276901245117,
      "kl": 0.79296875,
      "learning_rate": 8.692307692307692e-08,
      "loss": -0.0666,
      "reward": 1.4671775698661804,
      "reward_std": 0.4062620848417282,
      "reward_std_all": 0.8921076655387878,
      "rewards-max/forensics_sinkhorn_reward": 0.985426664352417,
      "rewards-max/forensics_strict_edge_reward": 0.7836527824401855,
      "rewards-max/forensics_tiou_reward": 0.9907374382019043,
      "rewards/forensics_sinkhorn_reward": 0.712396502494812,
      "rewards/forensics_strict_edge_reward": 0.12390291690826416,
      "rewards/forensics_tiou_reward": 0.6308781802654266,
      "step": 1188
    },
    {
      "completion_length": 23.75,
      "epoch": 3.6482334869431643,
      "grad_norm": 14.424718856811523,
      "kl": 0.876953125,
      "learning_rate": 8.615384615384615e-08,
      "loss": -0.1782,
      "reward": 1.6863338351249695,
      "reward_std": 0.46602863073349,
      "reward_std_all": 0.7937345504760742,
      "rewards-max/forensics_sinkhorn_reward": 0.9879902601242065,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.8007188439369202,
      "rewards/forensics_strict_edge_reward": 0.18617573380470276,
      "rewards/forensics_tiou_reward": 0.6994391679763794,
      "step": 1189
    },
    {
      "completion_length": 23.15625,
      "epoch": 3.651305683563748,
      "grad_norm": 10.5797119140625,
      "kl": 0.4970703125,
      "learning_rate": 8.538461538461537e-08,
      "loss": -0.1288,
      "reward": 1.3875287771224976,
      "reward_std": 0.35268189013004303,
      "reward_std_all": 0.890530526638031,
      "rewards-max/forensics_sinkhorn_reward": 0.977922797203064,
      "rewards-max/forensics_strict_edge_reward": 0.5581126511096954,
      "rewards-max/forensics_tiou_reward": 0.973993569612503,
      "rewards/forensics_sinkhorn_reward": 0.7425608932971954,
      "rewards/forensics_strict_edge_reward": 0.08589136600494385,
      "rewards/forensics_tiou_reward": 0.5590764880180359,
      "step": 1190
    },
    {
      "completion_length": 22.59375,
      "epoch": 3.654377880184332,
      "grad_norm": 15.075644493103027,
      "kl": 0.64453125,
      "learning_rate": 8.461538461538461e-08,
      "loss": 0.1387,
      "reward": 2.009000539779663,
      "reward_std": 0.3729225695133209,
      "reward_std_all": 0.6054271459579468,
      "rewards-max/forensics_sinkhorn_reward": 0.9895250499248505,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.8988685011863708,
      "rewards/forensics_strict_edge_reward": 0.3023933321237564,
      "rewards/forensics_tiou_reward": 0.8077386617660522,
      "step": 1191
    },
    {
      "completion_length": 22.75,
      "epoch": 3.6574500768049156,
      "grad_norm": 14.137870788574219,
      "kl": 0.51953125,
      "learning_rate": 8.384615384615384e-08,
      "loss": 0.0178,
      "reward": 1.7235293984413147,
      "reward_std": 0.37503843009471893,
      "reward_std_all": 0.734907478094101,
      "rewards-max/forensics_sinkhorn_reward": 0.9885696470737457,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.788692057132721,
      "rewards/forensics_strict_edge_reward": 0.22104902565479279,
      "rewards/forensics_tiou_reward": 0.7137881517410278,
      "step": 1192
    },
    {
      "completion_length": 25.4375,
      "epoch": 3.6605222734254994,
      "grad_norm": 16.356836318969727,
      "kl": 0.6396484375,
      "learning_rate": 8.307692307692307e-08,
      "loss": -0.0052,
      "reward": 1.747568964958191,
      "reward_std": 0.45301133394241333,
      "reward_std_all": 0.7525234818458557,
      "rewards-max/forensics_sinkhorn_reward": 0.9888506531715393,
      "rewards-max/forensics_strict_edge_reward": 0.9291328191757202,
      "rewards-max/forensics_tiou_reward": 0.9976238012313843,
      "rewards/forensics_sinkhorn_reward": 0.8314196467399597,
      "rewards/forensics_strict_edge_reward": 0.2056087926030159,
      "rewards/forensics_tiou_reward": 0.7105405330657959,
      "step": 1193
    },
    {
      "completion_length": 22.234375,
      "epoch": 3.6635944700460827,
      "grad_norm": 26.591503143310547,
      "kl": 0.736328125,
      "learning_rate": 8.23076923076923e-08,
      "loss": -0.071,
      "reward": 1.6064406633377075,
      "reward_std": 0.35698722302913666,
      "reward_std_all": 0.7841201424598694,
      "rewards-max/forensics_sinkhorn_reward": 0.9889346063137054,
      "rewards-max/forensics_strict_edge_reward": 0.862213134765625,
      "rewards-max/forensics_tiou_reward": 0.9983577728271484,
      "rewards/forensics_sinkhorn_reward": 0.8284139037132263,
      "rewards/forensics_strict_edge_reward": 0.10294059664011002,
      "rewards/forensics_tiou_reward": 0.6750861406326294,
      "step": 1194
    },
    {
      "completion_length": 21.6875,
      "epoch": 3.6666666666666665,
      "grad_norm": 12.191243171691895,
      "kl": 0.36328125,
      "learning_rate": 8.153846153846153e-08,
      "loss": -0.0991,
      "reward": 1.5114484429359436,
      "reward_std": 0.4299921542406082,
      "reward_std_all": 0.8793740570545197,
      "rewards-max/forensics_sinkhorn_reward": 0.9895187318325043,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.7504183650016785,
      "rewards/forensics_strict_edge_reward": 0.15480846166610718,
      "rewards/forensics_tiou_reward": 0.6062215566635132,
      "step": 1195
    },
    {
      "completion_length": 24.4375,
      "epoch": 3.6697388632872503,
      "grad_norm": 16.084056854248047,
      "kl": 0.662109375,
      "learning_rate": 8.076923076923077e-08,
      "loss": -0.0454,
      "reward": 1.563032865524292,
      "reward_std": 0.5506992936134338,
      "reward_std_all": 0.9123064875602722,
      "rewards-max/forensics_sinkhorn_reward": 0.9895718097686768,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.7434312701225281,
      "rewards/forensics_strict_edge_reward": 0.21410882472991943,
      "rewards/forensics_tiou_reward": 0.6054928004741669,
      "step": 1196
    },
    {
      "completion_length": 25.59375,
      "epoch": 3.672811059907834,
      "grad_norm": 13.675390243530273,
      "kl": 0.681640625,
      "learning_rate": 8e-08,
      "loss": -0.0865,
      "reward": 1.549721121788025,
      "reward_std": 0.32833317667245865,
      "reward_std_all": 0.8351920247077942,
      "rewards-max/forensics_sinkhorn_reward": 0.9893410503864288,
      "rewards-max/forensics_strict_edge_reward": 0.8304895758628845,
      "rewards-max/forensics_tiou_reward": 0.9985896944999695,
      "rewards/forensics_sinkhorn_reward": 0.7684741914272308,
      "rewards/forensics_strict_edge_reward": 0.1689624823629856,
      "rewards/forensics_tiou_reward": 0.612284392118454,
      "step": 1197
    },
    {
      "completion_length": 25.53125,
      "epoch": 3.675883256528418,
      "grad_norm": 12.93138599395752,
      "kl": 0.693359375,
      "learning_rate": 7.923076923076923e-08,
      "loss": -0.0327,
      "reward": 1.3884915113449097,
      "reward_std": 0.46569232642650604,
      "reward_std_all": 0.9443674385547638,
      "rewards-max/forensics_sinkhorn_reward": 0.9896719455718994,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.6917186677455902,
      "rewards/forensics_strict_edge_reward": 0.15563035756349564,
      "rewards/forensics_tiou_reward": 0.5411424338817596,
      "step": 1198
    },
    {
      "completion_length": 25.234375,
      "epoch": 3.6789554531490016,
      "grad_norm": 13.991519927978516,
      "kl": 0.630859375,
      "learning_rate": 7.846153846153846e-08,
      "loss": 0.0067,
      "reward": 1.8331029415130615,
      "reward_std": 0.34730078279972076,
      "reward_std_all": 0.6810528039932251,
      "rewards-max/forensics_sinkhorn_reward": 0.9885408580303192,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.8633393049240112,
      "rewards/forensics_strict_edge_reward": 0.217544786632061,
      "rewards/forensics_tiou_reward": 0.7522187530994415,
      "step": 1199
    },
    {
      "completion_length": 23.078125,
      "epoch": 3.6820276497695854,
      "grad_norm": 12.03671646118164,
      "kl": 0.53125,
      "learning_rate": 7.769230769230768e-08,
      "loss": -0.0528,
      "reward": 1.719654142856598,
      "reward_std": 0.5157893598079681,
      "reward_std_all": 0.8195173442363739,
      "rewards-max/forensics_sinkhorn_reward": 0.990246593952179,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.8214558959007263,
      "rewards/forensics_strict_edge_reward": 0.1933787539601326,
      "rewards/forensics_tiou_reward": 0.7048193514347076,
      "step": 1200
    },
    {
      "completion_length": 23.015625,
      "epoch": 3.6850998463901687,
      "grad_norm": 15.794657707214355,
      "kl": 0.900390625,
      "learning_rate": 7.692307692307692e-08,
      "loss": 0.0724,
      "reward": 1.5269607305526733,
      "reward_std": 0.6360687613487244,
      "reward_std_all": 0.9544899761676788,
      "rewards-max/forensics_sinkhorn_reward": 0.9900730550289154,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.7649164199829102,
      "rewards/forensics_strict_edge_reward": 0.1650882139801979,
      "rewards/forensics_tiou_reward": 0.5969561338424683,
      "step": 1201
    },
    {
      "completion_length": 23.109375,
      "epoch": 3.688172043010753,
      "grad_norm": 11.14525032043457,
      "kl": 0.5546875,
      "learning_rate": 7.615384615384615e-08,
      "loss": -0.0603,
      "reward": 1.696402132511139,
      "reward_std": 0.3455538973212242,
      "reward_std_all": 0.7443667054176331,
      "rewards-max/forensics_sinkhorn_reward": 0.989284098148346,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.8448244035243988,
      "rewards/forensics_strict_edge_reward": 0.17240767180919647,
      "rewards/forensics_tiou_reward": 0.6791700720787048,
      "step": 1202
    },
    {
      "completion_length": 27.96875,
      "epoch": 3.6912442396313363,
      "grad_norm": 12.897624015808105,
      "kl": 0.6015625,
      "learning_rate": 7.538461538461538e-08,
      "loss": 0.0195,
      "reward": 1.6529556512832642,
      "reward_std": 0.2800179049372673,
      "reward_std_all": 0.7472488284111023,
      "rewards-max/forensics_sinkhorn_reward": 0.9859375357627869,
      "rewards-max/forensics_strict_edge_reward": 0.8582656681537628,
      "rewards-max/forensics_tiou_reward": 0.9904114902019501,
      "rewards/forensics_sinkhorn_reward": 0.810938835144043,
      "rewards/forensics_strict_edge_reward": 0.17204620316624641,
      "rewards/forensics_tiou_reward": 0.6699707210063934,
      "step": 1203
    },
    {
      "completion_length": 23.1875,
      "epoch": 3.69431643625192,
      "grad_norm": 11.118005752563477,
      "kl": 0.4521484375,
      "learning_rate": 7.46153846153846e-08,
      "loss": -0.1822,
      "reward": 1.6903055310249329,
      "reward_std": 0.4686596244573593,
      "reward_std_all": 0.8107292950153351,
      "rewards-max/forensics_sinkhorn_reward": 0.9900156557559967,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.8112593293190002,
      "rewards/forensics_strict_edge_reward": 0.2018430009484291,
      "rewards/forensics_tiou_reward": 0.6772032082080841,
      "step": 1204
    },
    {
      "completion_length": 21.890625,
      "epoch": 3.697388632872504,
      "grad_norm": 12.108508110046387,
      "kl": 0.552734375,
      "learning_rate": 7.384615384615384e-08,
      "loss": -0.0486,
      "reward": 1.9432033896446228,
      "reward_std": 0.3915480077266693,
      "reward_std_all": 0.6672202944755554,
      "rewards-max/forensics_sinkhorn_reward": 0.989905595779419,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.8854562938213348,
      "rewards/forensics_strict_edge_reward": 0.2839144915342331,
      "rewards/forensics_tiou_reward": 0.7738327085971832,
      "step": 1205
    },
    {
      "completion_length": 26.6875,
      "epoch": 3.7004608294930876,
      "grad_norm": 10.434593200683594,
      "kl": 0.693359375,
      "learning_rate": 7.307692307692308e-08,
      "loss": -0.1057,
      "reward": 1.8093842267990112,
      "reward_std": 0.4182214289903641,
      "reward_std_all": 0.7263183295726776,
      "rewards-max/forensics_sinkhorn_reward": 0.9890565872192383,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.8632945418357849,
      "rewards/forensics_strict_edge_reward": 0.20549792051315308,
      "rewards/forensics_tiou_reward": 0.740591824054718,
      "step": 1206
    },
    {
      "completion_length": 25.515625,
      "epoch": 3.7035330261136714,
      "grad_norm": 11.919801712036133,
      "kl": 0.53515625,
      "learning_rate": 7.230769230769231e-08,
      "loss": -0.1406,
      "reward": 1.731320559978485,
      "reward_std": 0.3463796377182007,
      "reward_std_all": 0.7320219278335571,
      "rewards-max/forensics_sinkhorn_reward": 0.985645592212677,
      "rewards-max/forensics_strict_edge_reward": 0.8582656681537628,
      "rewards-max/forensics_tiou_reward": 0.9898173809051514,
      "rewards/forensics_sinkhorn_reward": 0.8127272427082062,
      "rewards/forensics_strict_edge_reward": 0.20514756441116333,
      "rewards/forensics_tiou_reward": 0.713445782661438,
      "step": 1207
    },
    {
      "completion_length": 26.375,
      "epoch": 3.7066052227342547,
      "grad_norm": 14.249067306518555,
      "kl": 0.447265625,
      "learning_rate": 7.153846153846154e-08,
      "loss": -0.0556,
      "reward": 1.8184142708778381,
      "reward_std": 0.33889855444431305,
      "reward_std_all": 0.6947306096553802,
      "rewards-max/forensics_sinkhorn_reward": 0.9901940524578094,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.8299684524536133,
      "rewards/forensics_strict_edge_reward": 0.27128319442272186,
      "rewards/forensics_tiou_reward": 0.7171626389026642,
      "step": 1208
    },
    {
      "completion_length": 25.15625,
      "epoch": 3.709677419354839,
      "grad_norm": 11.931041717529297,
      "kl": 0.5341796875,
      "learning_rate": 7.076923076923076e-08,
      "loss": -0.1117,
      "reward": 1.619127094745636,
      "reward_std": 0.39653944969177246,
      "reward_std_all": 0.7959240674972534,
      "rewards-max/forensics_sinkhorn_reward": 0.984102874994278,
      "rewards-max/forensics_strict_edge_reward": 0.7401526570320129,
      "rewards-max/forensics_tiou_reward": 0.9870678186416626,
      "rewards/forensics_sinkhorn_reward": 0.811417818069458,
      "rewards/forensics_strict_edge_reward": 0.13450009375810623,
      "rewards/forensics_tiou_reward": 0.6732092499732971,
      "step": 1209
    },
    {
      "completion_length": 27.015625,
      "epoch": 3.7127496159754223,
      "grad_norm": 10.572579383850098,
      "kl": 0.53515625,
      "learning_rate": 7e-08,
      "loss": -0.07,
      "reward": 1.660865843296051,
      "reward_std": 0.5349415242671967,
      "reward_std_all": 0.8577792942523956,
      "rewards-max/forensics_sinkhorn_reward": 0.989891529083252,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.8010604977607727,
      "rewards/forensics_strict_edge_reward": 0.19525663554668427,
      "rewards/forensics_tiou_reward": 0.6645486950874329,
      "step": 1210
    },
    {
      "completion_length": 23.828125,
      "epoch": 3.715821812596006,
      "grad_norm": 11.133726119995117,
      "kl": 0.59375,
      "learning_rate": 6.923076923076923e-08,
      "loss": -0.1099,
      "reward": 1.8736697435379028,
      "reward_std": 0.47074584662914276,
      "reward_std_all": 0.7349646687507629,
      "rewards-max/forensics_sinkhorn_reward": 0.9897074103355408,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.8464780449867249,
      "rewards/forensics_strict_edge_reward": 0.26066067069768906,
      "rewards/forensics_tiou_reward": 0.7665309906005859,
      "step": 1211
    },
    {
      "completion_length": 27.9375,
      "epoch": 3.71889400921659,
      "grad_norm": 12.686710357666016,
      "kl": 0.5703125,
      "learning_rate": 6.846153846153846e-08,
      "loss": -0.0037,
      "reward": 1.5826083421707153,
      "reward_std": 0.4329766780138016,
      "reward_std_all": 0.8255025446414948,
      "rewards-max/forensics_sinkhorn_reward": 0.9882242381572723,
      "rewards-max/forensics_strict_edge_reward": 0.8551307618618011,
      "rewards-max/forensics_tiou_reward": 0.9960783123970032,
      "rewards/forensics_sinkhorn_reward": 0.7856385707855225,
      "rewards/forensics_strict_edge_reward": 0.1373039409518242,
      "rewards/forensics_tiou_reward": 0.6596658527851105,
      "step": 1212
    },
    {
      "completion_length": 28.25,
      "epoch": 3.7219662058371736,
      "grad_norm": 15.94315242767334,
      "kl": 0.728515625,
      "learning_rate": 6.769230769230769e-08,
      "loss": -0.0084,
      "reward": 1.5357838869094849,
      "reward_std": 0.43905067443847656,
      "reward_std_all": 0.8885353207588196,
      "rewards-max/forensics_sinkhorn_reward": 0.9895842671394348,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.7178850471973419,
      "rewards/forensics_strict_edge_reward": 0.2072916328907013,
      "rewards/forensics_tiou_reward": 0.610607236623764,
      "step": 1213
    },
    {
      "completion_length": 27.546875,
      "epoch": 3.7250384024577574,
      "grad_norm": 10.446956634521484,
      "kl": 0.564453125,
      "learning_rate": 6.692307692307691e-08,
      "loss": -0.1338,
      "reward": 1.5588395595550537,
      "reward_std": 0.24684032797813416,
      "reward_std_all": 0.774695873260498,
      "rewards-max/forensics_sinkhorn_reward": 0.9891196489334106,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.7453080713748932,
      "rewards/forensics_strict_edge_reward": 0.1814420372247696,
      "rewards/forensics_tiou_reward": 0.6320894956588745,
      "step": 1214
    },
    {
      "completion_length": 24.796875,
      "epoch": 3.7281105990783407,
      "grad_norm": 13.368304252624512,
      "kl": 0.5546875,
      "learning_rate": 6.615384615384615e-08,
      "loss": -0.0136,
      "reward": 1.824824571609497,
      "reward_std": 0.5118314176797867,
      "reward_std_all": 0.76073157787323,
      "rewards-max/forensics_sinkhorn_reward": 0.9895264804363251,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.8246711492538452,
      "rewards/forensics_strict_edge_reward": 0.2470599114894867,
      "rewards/forensics_tiou_reward": 0.7530935406684875,
      "step": 1215
    },
    {
      "completion_length": 23.484375,
      "epoch": 3.731182795698925,
      "grad_norm": 12.583403587341309,
      "kl": 0.6328125,
      "learning_rate": 6.538461538461538e-08,
      "loss": -0.0641,
      "reward": 1.7083109617233276,
      "reward_std": 0.47094428539276123,
      "reward_std_all": 0.8087509274482727,
      "rewards-max/forensics_sinkhorn_reward": 0.9877724051475525,
      "rewards-max/forensics_strict_edge_reward": 0.9291328191757202,
      "rewards-max/forensics_tiou_reward": 0.9958616495132446,
      "rewards/forensics_sinkhorn_reward": 0.8139919638633728,
      "rewards/forensics_strict_edge_reward": 0.1953871175646782,
      "rewards/forensics_tiou_reward": 0.6989317536354065,
      "step": 1216
    },
    {
      "completion_length": 27.984375,
      "epoch": 3.7342549923195083,
      "grad_norm": 13.940295219421387,
      "kl": 0.78125,
      "learning_rate": 6.461538461538462e-08,
      "loss": -0.0393,
      "reward": 1.655286967754364,
      "reward_std": 0.44311827421188354,
      "reward_std_all": 0.8207508027553558,
      "rewards-max/forensics_sinkhorn_reward": 0.9903688132762909,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.7584864795207977,
      "rewards/forensics_strict_edge_reward": 0.22330307960510254,
      "rewards/forensics_tiou_reward": 0.6734974384307861,
      "step": 1217
    },
    {
      "completion_length": 23.515625,
      "epoch": 3.737327188940092,
      "grad_norm": 19.861316680908203,
      "kl": 1.041015625,
      "learning_rate": 6.384615384615385e-08,
      "loss": -0.1424,
      "reward": 1.8459964990615845,
      "reward_std": 0.49747519195079803,
      "reward_std_all": 0.7380769848823547,
      "rewards-max/forensics_sinkhorn_reward": 0.9889942407608032,
      "rewards-max/forensics_strict_edge_reward": 0.9291328191757202,
      "rewards-max/forensics_tiou_reward": 0.9972094297409058,
      "rewards/forensics_sinkhorn_reward": 0.8520554602146149,
      "rewards/forensics_strict_edge_reward": 0.24704700708389282,
      "rewards/forensics_tiou_reward": 0.7468940615653992,
      "step": 1218
    },
    {
      "completion_length": 24.046875,
      "epoch": 3.740399385560676,
      "grad_norm": 9.072315216064453,
      "kl": 0.529296875,
      "learning_rate": 6.307692307692307e-08,
      "loss": -0.0264,
      "reward": 1.603219747543335,
      "reward_std": 0.39590708911418915,
      "reward_std_all": 0.8374265134334564,
      "rewards-max/forensics_sinkhorn_reward": 0.9900505840778351,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.7852272391319275,
      "rewards/forensics_strict_edge_reward": 0.18395091593265533,
      "rewards/forensics_tiou_reward": 0.6340416371822357,
      "step": 1219
    },
    {
      "completion_length": 22.28125,
      "epoch": 3.7434715821812596,
      "grad_norm": 10.991410255432129,
      "kl": 0.615234375,
      "learning_rate": 6.23076923076923e-08,
      "loss": -0.1239,
      "reward": 1.6230233907699585,
      "reward_std": 0.46640847623348236,
      "reward_std_all": 0.8308362364768982,
      "rewards-max/forensics_sinkhorn_reward": 0.9892369508743286,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.7995653748512268,
      "rewards/forensics_strict_edge_reward": 0.17926685512065887,
      "rewards/forensics_tiou_reward": 0.644191175699234,
      "step": 1220
    },
    {
      "completion_length": 27.1875,
      "epoch": 3.7465437788018434,
      "grad_norm": 14.142976760864258,
      "kl": 0.66796875,
      "learning_rate": 6.153846153846154e-08,
      "loss": -0.0641,
      "reward": 1.7123211026191711,
      "reward_std": 0.40643535554409027,
      "reward_std_all": 0.7833930850028992,
      "rewards-max/forensics_sinkhorn_reward": 0.989800900220871,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.8000688254833221,
      "rewards/forensics_strict_edge_reward": 0.24674563854932785,
      "rewards/forensics_tiou_reward": 0.6655065715312958,
      "step": 1221
    },
    {
      "completion_length": 23.609375,
      "epoch": 3.749615975422427,
      "grad_norm": 13.628625869750977,
      "kl": 0.552734375,
      "learning_rate": 6.076923076923077e-08,
      "loss": -0.1464,
      "reward": 1.6120390892028809,
      "reward_std": 0.35573500394821167,
      "reward_std_all": 0.8264379799365997,
      "rewards-max/forensics_sinkhorn_reward": 0.982625424861908,
      "rewards-max/forensics_strict_edge_reward": 0.8074871003627777,
      "rewards-max/forensics_tiou_reward": 0.9824939370155334,
      "rewards/forensics_sinkhorn_reward": 0.7418374717235565,
      "rewards/forensics_strict_edge_reward": 0.22324874252080917,
      "rewards/forensics_tiou_reward": 0.6469527781009674,
      "step": 1222
    },
    {
      "completion_length": 24.671875,
      "epoch": 3.752688172043011,
      "grad_norm": 12.129789352416992,
      "kl": 0.55078125,
      "learning_rate": 6e-08,
      "loss": -0.0501,
      "reward": 1.6046168208122253,
      "reward_std": 0.5339392125606537,
      "reward_std_all": 0.881377100944519,
      "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": 0.7732331454753876,
      "rewards/forensics_strict_edge_reward": 0.18524152785539627,
      "rewards/forensics_tiou_reward": 0.6461422443389893,
      "step": 1223
    },
    {
      "completion_length": 22.15625,
      "epoch": 3.7557603686635943,
      "grad_norm": 13.654383659362793,
      "kl": 0.583984375,
      "learning_rate": 5.923076923076923e-08,
      "loss": -0.0328,
      "reward": 1.7661513686180115,
      "reward_std": 0.533076673746109,
      "reward_std_all": 0.7960252165794373,
      "rewards-max/forensics_sinkhorn_reward": 0.9890056550502777,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.8383677005767822,
      "rewards/forensics_strict_edge_reward": 0.23424261063337326,
      "rewards/forensics_tiou_reward": 0.6935410797595978,
      "step": 1224
    },
    {
      "completion_length": 24.109375,
      "epoch": 3.758832565284178,
      "grad_norm": 11.584436416625977,
      "kl": 0.439453125,
      "learning_rate": 5.8461538461538455e-08,
      "loss": -0.1512,
      "reward": 1.6475275158882141,
      "reward_std": 0.3571476638317108,
      "reward_std_all": 0.799210250377655,
      "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": 0.7652468383312225,
      "rewards/forensics_strict_edge_reward": 0.21211552619934082,
      "rewards/forensics_tiou_reward": 0.6701652407646179,
      "step": 1225
    },
    {
      "completion_length": 27.140625,
      "epoch": 3.761904761904762,
      "grad_norm": 13.24372386932373,
      "kl": 0.87109375,
      "learning_rate": 5.7692307692307695e-08,
      "loss": 0.1134,
      "reward": 1.6277537941932678,
      "reward_std": 0.5782201737165451,
      "reward_std_all": 0.900641679763794,
      "rewards-max/forensics_sinkhorn_reward": 0.9857401549816132,
      "rewards-max/forensics_strict_edge_reward": 0.8608831465244293,
      "rewards-max/forensics_tiou_reward": 0.9897161722183228,
      "rewards/forensics_sinkhorn_reward": 0.7792043685913086,
      "rewards/forensics_strict_edge_reward": 0.19353393092751503,
      "rewards/forensics_tiou_reward": 0.6550155282020569,
      "step": 1226
    },
    {
      "completion_length": 27.90625,
      "epoch": 3.7649769585253456,
      "grad_norm": 11.598155975341797,
      "kl": 0.544921875,
      "learning_rate": 5.692307692307692e-08,
      "loss": -0.0244,
      "reward": 1.6155139207839966,
      "reward_std": 0.4203542023897171,
      "reward_std_all": 0.8085854053497314,
      "rewards-max/forensics_sinkhorn_reward": 0.9878413677215576,
      "rewards-max/forensics_strict_edge_reward": 0.882964164018631,
      "rewards-max/forensics_tiou_reward": 0.9962957501411438,
      "rewards/forensics_sinkhorn_reward": 0.8003528118133545,
      "rewards/forensics_strict_edge_reward": 0.15736309438943863,
      "rewards/forensics_tiou_reward": 0.6577981412410736,
      "step": 1227
    },
    {
      "completion_length": 24.671875,
      "epoch": 3.7680491551459294,
      "grad_norm": 12.947687149047852,
      "kl": 0.55078125,
      "learning_rate": 5.615384615384615e-08,
      "loss": -0.0412,
      "reward": 1.6992263793945312,
      "reward_std": 0.45163969695568085,
      "reward_std_all": 0.7847777307033539,
      "rewards-max/forensics_sinkhorn_reward": 0.9893225133419037,
      "rewards-max/forensics_strict_edge_reward": 0.9291328191757202,
      "rewards-max/forensics_tiou_reward": 0.9987068176269531,
      "rewards/forensics_sinkhorn_reward": 0.7924646437168121,
      "rewards/forensics_strict_edge_reward": 0.2106202095746994,
      "rewards/forensics_tiou_reward": 0.6961415410041809,
      "step": 1228
    },
    {
      "completion_length": 26.359375,
      "epoch": 3.771121351766513,
      "grad_norm": 11.291301727294922,
      "kl": 0.5126953125,
      "learning_rate": 5.538461538461538e-08,
      "loss": 0.0159,
      "reward": 1.513837218284607,
      "reward_std": 0.3978079557418823,
      "reward_std_all": 0.8445185124874115,
      "rewards-max/forensics_sinkhorn_reward": 0.9880737960338593,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.7705405056476593,
      "rewards/forensics_strict_edge_reward": 0.1416989378631115,
      "rewards/forensics_tiou_reward": 0.6015976667404175,
      "step": 1229
    },
    {
      "completion_length": 23.296875,
      "epoch": 3.774193548387097,
      "grad_norm": 12.758527755737305,
      "kl": 0.619140625,
      "learning_rate": 5.461538461538461e-08,
      "loss": -0.1041,
      "reward": 1.7223365902900696,
      "reward_std": 0.5318497121334076,
      "reward_std_all": 0.8231242895126343,
      "rewards-max/forensics_sinkhorn_reward": 0.9893862307071686,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.8000990748405457,
      "rewards/forensics_strict_edge_reward": 0.2543371468782425,
      "rewards/forensics_tiou_reward": 0.6679003834724426,
      "step": 1230
    },
    {
      "completion_length": 27.796875,
      "epoch": 3.7772657450076803,
      "grad_norm": 12.522790908813477,
      "kl": 0.60546875,
      "learning_rate": 5.384615384615385e-08,
      "loss": 0.0741,
      "reward": 1.8303749561309814,
      "reward_std": 0.37972304224967957,
      "reward_std_all": 0.7023417949676514,
      "rewards-max/forensics_sinkhorn_reward": 0.9900062084197998,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.820373147726059,
      "rewards/forensics_strict_edge_reward": 0.2525985389947891,
      "rewards/forensics_tiou_reward": 0.7574032843112946,
      "step": 1231
    },
    {
      "completion_length": 23.734375,
      "epoch": 3.780337941628264,
      "grad_norm": 12.085328102111816,
      "kl": 0.83984375,
      "learning_rate": 5.307692307692308e-08,
      "loss": 0.0139,
      "reward": 1.471071481704712,
      "reward_std": 0.43653184175491333,
      "reward_std_all": 0.9201214909553528,
      "rewards-max/forensics_sinkhorn_reward": 0.9890917241573334,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.7551751732826233,
      "rewards/forensics_strict_edge_reward": 0.1423240229487419,
      "rewards/forensics_tiou_reward": 0.5735722631216049,
      "step": 1232
    },
    {
      "completion_length": 26.9375,
      "epoch": 3.783410138248848,
      "grad_norm": 17.399728775024414,
      "kl": 1.01171875,
      "learning_rate": 5.2307692307692304e-08,
      "loss": -0.0322,
      "reward": 1.7605742812156677,
      "reward_std": 0.5037369281053543,
      "reward_std_all": 0.7804163098335266,
      "rewards-max/forensics_sinkhorn_reward": 0.9874339401721954,
      "rewards-max/forensics_strict_edge_reward": 0.8032653331756592,
      "rewards-max/forensics_tiou_reward": 0.9931369125843048,
      "rewards/forensics_sinkhorn_reward": 0.8357438445091248,
      "rewards/forensics_strict_edge_reward": 0.20165041089057922,
      "rewards/forensics_tiou_reward": 0.7231800556182861,
      "step": 1233
    },
    {
      "completion_length": 24.296875,
      "epoch": 3.7864823348694316,
      "grad_norm": 13.175040245056152,
      "kl": 0.5419921875,
      "learning_rate": 5.153846153846154e-08,
      "loss": -0.0735,
      "reward": 1.757337212562561,
      "reward_std": 0.38764943182468414,
      "reward_std_all": 0.7516116499900818,
      "rewards-max/forensics_sinkhorn_reward": 0.9894649982452393,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.8151529431343079,
      "rewards/forensics_strict_edge_reward": 0.25327926874160767,
      "rewards/forensics_tiou_reward": 0.6889050602912903,
      "step": 1234
    },
    {
      "completion_length": 27.09375,
      "epoch": 3.7895545314900154,
      "grad_norm": 13.433815956115723,
      "kl": 0.529296875,
      "learning_rate": 5.0769230769230764e-08,
      "loss": -0.1606,
      "reward": 1.6500385403633118,
      "reward_std": 0.45643216371536255,
      "reward_std_all": 0.8049279451370239,
      "rewards-max/forensics_sinkhorn_reward": 0.988983154296875,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.768890917301178,
      "rewards/forensics_strict_edge_reward": 0.19442875683307648,
      "rewards/forensics_tiou_reward": 0.6867189407348633,
      "step": 1235
    },
    {
      "completion_length": 25.90625,
      "epoch": 3.792626728110599,
      "grad_norm": 13.320003509521484,
      "kl": 0.4619140625,
      "learning_rate": 5e-08,
      "loss": -0.0163,
      "reward": 1.5997018218040466,
      "reward_std": 0.47365862131118774,
      "reward_std_all": 0.8400543928146362,
      "rewards-max/forensics_sinkhorn_reward": 0.9897410273551941,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.7704694867134094,
      "rewards/forensics_strict_edge_reward": 0.16304141283035278,
      "rewards/forensics_tiou_reward": 0.6661909520626068,
      "step": 1236
    },
    {
      "completion_length": 23.359375,
      "epoch": 3.795698924731183,
      "grad_norm": 12.105475425720215,
      "kl": 0.638671875,
      "learning_rate": 4.923076923076923e-08,
      "loss": -0.008,
      "reward": 1.703907549381256,
      "reward_std": 0.41055063903331757,
      "reward_std_all": 0.753375381231308,
      "rewards-max/forensics_sinkhorn_reward": 0.9874875247478485,
      "rewards-max/forensics_strict_edge_reward": 0.9342645704746246,
      "rewards-max/forensics_tiou_reward": 0.996242344379425,
      "rewards/forensics_sinkhorn_reward": 0.8392004072666168,
      "rewards/forensics_strict_edge_reward": 0.16073672100901604,
      "rewards/forensics_tiou_reward": 0.7039704024791718,
      "step": 1237
    },
    {
      "completion_length": 24.40625,
      "epoch": 3.7987711213517663,
      "grad_norm": 13.945648193359375,
      "kl": 0.771484375,
      "learning_rate": 4.846153846153846e-08,
      "loss": -0.1034,
      "reward": 1.7502611875534058,
      "reward_std": 0.38084404170513153,
      "reward_std_all": 0.7363066077232361,
      "rewards-max/forensics_sinkhorn_reward": 0.9883864521980286,
      "rewards-max/forensics_strict_edge_reward": 0.7877823710441589,
      "rewards-max/forensics_tiou_reward": 0.996392011642456,
      "rewards/forensics_sinkhorn_reward": 0.832831472158432,
      "rewards/forensics_strict_edge_reward": 0.24613525718450546,
      "rewards/forensics_tiou_reward": 0.6712943911552429,
      "step": 1238
    },
    {
      "completion_length": 28.78125,
      "epoch": 3.80184331797235,
      "grad_norm": 11.22603702545166,
      "kl": 0.541015625,
      "learning_rate": 4.769230769230769e-08,
      "loss": -0.055,
      "reward": 1.790736198425293,
      "reward_std": 0.3354029059410095,
      "reward_std_all": 0.720068097114563,
      "rewards-max/forensics_sinkhorn_reward": 0.9886408150196075,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.8320357203483582,
      "rewards/forensics_strict_edge_reward": 0.2385462149977684,
      "rewards/forensics_tiou_reward": 0.7201542854309082,
      "step": 1239
    },
    {
      "completion_length": 21.78125,
      "epoch": 3.804915514592934,
      "grad_norm": 8.866761207580566,
      "kl": 0.482421875,
      "learning_rate": 4.692307692307692e-08,
      "loss": -0.1075,
      "reward": 1.5370423793792725,
      "reward_std": 0.45099736750125885,
      "reward_std_all": 0.8479570150375366,
      "rewards-max/forensics_sinkhorn_reward": 0.9883498847484589,
      "rewards-max/forensics_strict_edge_reward": 0.9677534997463226,
      "rewards-max/forensics_tiou_reward": 0.998239129781723,
      "rewards/forensics_sinkhorn_reward": 0.7928438484668732,
      "rewards/forensics_strict_edge_reward": 0.1307176649570465,
      "rewards/forensics_tiou_reward": 0.6134809255599976,
      "step": 1240
    },
    {
      "completion_length": 23.71875,
      "epoch": 3.8079877112135176,
      "grad_norm": 11.537182807922363,
      "kl": 0.6357421875,
      "learning_rate": 4.615384615384615e-08,
      "loss": -0.1509,
      "reward": 1.490292251110077,
      "reward_std": 0.4618707001209259,
      "reward_std_all": 0.8939889967441559,
      "rewards-max/forensics_sinkhorn_reward": 0.9859094619750977,
      "rewards-max/forensics_strict_edge_reward": 0.7995419204235077,
      "rewards-max/forensics_tiou_reward": 0.9928475916385651,
      "rewards/forensics_sinkhorn_reward": 0.7453055083751678,
      "rewards/forensics_strict_edge_reward": 0.14300092309713364,
      "rewards/forensics_tiou_reward": 0.6019858121871948,
      "step": 1241
    },
    {
      "completion_length": 22.5625,
      "epoch": 3.8110599078341014,
      "grad_norm": 11.641843795776367,
      "kl": 0.521484375,
      "learning_rate": 4.538461538461538e-08,
      "loss": -0.1033,
      "reward": 1.5769466757774353,
      "reward_std": 0.4214726984500885,
      "reward_std_all": 0.8514030873775482,
      "rewards-max/forensics_sinkhorn_reward": 0.9880838096141815,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.783217042684555,
      "rewards/forensics_strict_edge_reward": 0.19296333193778992,
      "rewards/forensics_tiou_reward": 0.6007662713527679,
      "step": 1242
    },
    {
      "completion_length": 25.59375,
      "epoch": 3.814132104454685,
      "grad_norm": 20.975404739379883,
      "kl": 1.001953125,
      "learning_rate": 4.461538461538461e-08,
      "loss": -0.0656,
      "reward": 1.7214080095291138,
      "reward_std": 0.5281214565038681,
      "reward_std_all": 0.8196294605731964,
      "rewards-max/forensics_sinkhorn_reward": 0.9840653240680695,
      "rewards-max/forensics_strict_edge_reward": 0.8582656681537628,
      "rewards-max/forensics_tiou_reward": 0.9876053929328918,
      "rewards/forensics_sinkhorn_reward": 0.7920029759407043,
      "rewards/forensics_strict_edge_reward": 0.2433033473789692,
      "rewards/forensics_tiou_reward": 0.6861016154289246,
      "step": 1243
    },
    {
      "completion_length": 25.171875,
      "epoch": 3.817204301075269,
      "grad_norm": 28.945484161376953,
      "kl": 1.40625,
      "learning_rate": 4.3846153846153846e-08,
      "loss": -0.0471,
      "reward": 1.7992204427719116,
      "reward_std": 0.5472634434700012,
      "reward_std_all": 0.7841418981552124,
      "rewards-max/forensics_sinkhorn_reward": 0.9894110858440399,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.8219929039478302,
      "rewards/forensics_strict_edge_reward": 0.25665727257728577,
      "rewards/forensics_tiou_reward": 0.7205703556537628,
      "step": 1244
    },
    {
      "completion_length": 31.59375,
      "epoch": 3.8202764976958523,
      "grad_norm": 12.12966537475586,
      "kl": 0.638671875,
      "learning_rate": 4.307692307692307e-08,
      "loss": -0.0423,
      "reward": 1.6007930040359497,
      "reward_std": 0.30600976943969727,
      "reward_std_all": 0.7674245238304138,
      "rewards-max/forensics_sinkhorn_reward": 0.980418860912323,
      "rewards-max/forensics_strict_edge_reward": 0.7482926547527313,
      "rewards-max/forensics_tiou_reward": 0.9802667796611786,
      "rewards/forensics_sinkhorn_reward": 0.7853220701217651,
      "rewards/forensics_strict_edge_reward": 0.14759564027190208,
      "rewards/forensics_tiou_reward": 0.6678752303123474,
      "step": 1245
    },
    {
      "completion_length": 24.453125,
      "epoch": 3.823348694316436,
      "grad_norm": 10.747956275939941,
      "kl": 0.564453125,
      "learning_rate": 4.230769230769231e-08,
      "loss": 0.0255,
      "reward": 1.596056044101715,
      "reward_std": 0.39312273263931274,
      "reward_std_all": 0.8013515174388885,
      "rewards-max/forensics_sinkhorn_reward": 0.9861256182193756,
      "rewards-max/forensics_strict_edge_reward": 0.8351600170135498,
      "rewards-max/forensics_tiou_reward": 0.9910022616386414,
      "rewards/forensics_sinkhorn_reward": 0.7711269855499268,
      "rewards/forensics_strict_edge_reward": 0.15526913851499557,
      "rewards/forensics_tiou_reward": 0.6696598529815674,
      "step": 1246
    },
    {
      "completion_length": 29.53125,
      "epoch": 3.82642089093702,
      "grad_norm": 11.726086616516113,
      "kl": 0.6015625,
      "learning_rate": 4.1538461538461534e-08,
      "loss": -0.0749,
      "reward": 1.422939658164978,
      "reward_std": 0.4229375869035721,
      "reward_std_all": 0.9148435890674591,
      "rewards-max/forensics_sinkhorn_reward": 0.9851272702217102,
      "rewards-max/forensics_strict_edge_reward": 0.905342161655426,
      "rewards-max/forensics_tiou_reward": 0.9917556047439575,
      "rewards/forensics_sinkhorn_reward": 0.7012355029582977,
      "rewards/forensics_strict_edge_reward": 0.11936254426836967,
      "rewards/forensics_tiou_reward": 0.6023416519165039,
      "step": 1247
    },
    {
      "completion_length": 28.953125,
      "epoch": 3.8294930875576036,
      "grad_norm": 11.772804260253906,
      "kl": 0.619140625,
      "learning_rate": 4.076923076923077e-08,
      "loss": -0.0252,
      "reward": 1.8770189881324768,
      "reward_std": 0.4759085774421692,
      "reward_std_all": 0.7041239142417908,
      "rewards-max/forensics_sinkhorn_reward": 0.9893681406974792,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.8323025703430176,
      "rewards/forensics_strict_edge_reward": 0.2845662534236908,
      "rewards/forensics_tiou_reward": 0.7601500749588013,
      "step": 1248
    },
    {
      "completion_length": 23.6875,
      "epoch": 3.8325652841781874,
      "grad_norm": 11.235897064208984,
      "kl": 0.51953125,
      "learning_rate": 4e-08,
      "loss": -0.0963,
      "reward": 1.9100090861320496,
      "reward_std": 0.6454127430915833,
      "reward_std_all": 0.7638944685459137,
      "rewards-max/forensics_sinkhorn_reward": 0.9897510409355164,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.840589314699173,
      "rewards/forensics_strict_edge_reward": 0.337250292301178,
      "rewards/forensics_tiou_reward": 0.7321694195270538,
      "step": 1249
    },
    {
      "completion_length": 25.75,
      "epoch": 3.835637480798771,
      "grad_norm": 14.494762420654297,
      "kl": 0.51171875,
      "learning_rate": 3.923076923076923e-08,
      "loss": -0.1128,
      "reward": 1.6775006651878357,
      "reward_std": 0.27349771559238434,
      "reward_std_all": 0.7260995507240295,
      "rewards-max/forensics_sinkhorn_reward": 0.9899585247039795,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.8216968774795532,
      "rewards/forensics_strict_edge_reward": 0.12298441305756569,
      "rewards/forensics_tiou_reward": 0.7328194677829742,
      "step": 1250
    },
    {
      "completion_length": 25.21875,
      "epoch": 3.838709677419355,
      "grad_norm": 12.937322616577148,
      "kl": 0.537109375,
      "learning_rate": 3.846153846153846e-08,
      "loss": -0.0721,
      "reward": 1.6480374932289124,
      "reward_std": 0.39455830305814743,
      "reward_std_all": 0.7873431146144867,
      "rewards-max/forensics_sinkhorn_reward": 0.9834102690219879,
      "rewards-max/forensics_strict_edge_reward": 0.882964164018631,
      "rewards-max/forensics_tiou_reward": 0.9870569109916687,
      "rewards/forensics_sinkhorn_reward": 0.7737694978713989,
      "rewards/forensics_strict_edge_reward": 0.20623796805739403,
      "rewards/forensics_tiou_reward": 0.6680299639701843,
      "step": 1251
    },
    {
      "completion_length": 24.828125,
      "epoch": 3.8417818740399383,
      "grad_norm": 11.737868309020996,
      "kl": 0.583984375,
      "learning_rate": 3.769230769230769e-08,
      "loss": -0.0627,
      "reward": 1.5733497738838196,
      "reward_std": 0.39786118268966675,
      "reward_std_all": 0.8267669677734375,
      "rewards-max/forensics_sinkhorn_reward": 0.9894868731498718,
      "rewards-max/forensics_strict_edge_reward": 0.9836080372333527,
      "rewards-max/forensics_tiou_reward": 0.9998032748699188,
      "rewards/forensics_sinkhorn_reward": 0.7622936367988586,
      "rewards/forensics_strict_edge_reward": 0.19845084846019745,
      "rewards/forensics_tiou_reward": 0.6126052439212799,
      "step": 1252
    },
    {
      "completion_length": 28.1875,
      "epoch": 3.8448540706605225,
      "grad_norm": 13.89632511138916,
      "kl": 0.6884765625,
      "learning_rate": 3.692307692307692e-08,
      "loss": -0.0033,
      "reward": 1.7858545780181885,
      "reward_std": 0.5195554792881012,
      "reward_std_all": 0.8092029690742493,
      "rewards-max/forensics_sinkhorn_reward": 0.9897888898849487,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.8143295645713806,
      "rewards/forensics_strict_edge_reward": 0.273930087685585,
      "rewards/forensics_tiou_reward": 0.6975949108600616,
      "step": 1253
    },
    {
      "completion_length": 22.875,
      "epoch": 3.847926267281106,
      "grad_norm": 13.383821487426758,
      "kl": 0.5078125,
      "learning_rate": 3.6153846153846155e-08,
      "loss": -0.0481,
      "reward": 1.5324537754058838,
      "reward_std": 0.4275979995727539,
      "reward_std_all": 0.8397700488567352,
      "rewards-max/forensics_sinkhorn_reward": 0.9843883514404297,
      "rewards-max/forensics_strict_edge_reward": 0.8419698476791382,
      "rewards-max/forensics_tiou_reward": 0.9858813881874084,
      "rewards/forensics_sinkhorn_reward": 0.768690675497055,
      "rewards/forensics_strict_edge_reward": 0.16569096967577934,
      "rewards/forensics_tiou_reward": 0.5980721116065979,
      "step": 1254
    },
    {
      "completion_length": 26.1875,
      "epoch": 3.8509984639016897,
      "grad_norm": 11.730456352233887,
      "kl": 0.568359375,
      "learning_rate": 3.538461538461538e-08,
      "loss": -0.0232,
      "reward": 1.4063376188278198,
      "reward_std": 0.39452439546585083,
      "reward_std_all": 0.9017065167427063,
      "rewards-max/forensics_sinkhorn_reward": 0.9843588471412659,
      "rewards-max/forensics_strict_edge_reward": 0.6037485972046852,
      "rewards-max/forensics_tiou_reward": 0.9860668182373047,
      "rewards/forensics_sinkhorn_reward": 0.7502799034118652,
      "rewards/forensics_strict_edge_reward": 0.0999303050339222,
      "rewards/forensics_tiou_reward": 0.5561273992061615,
      "step": 1255
    },
    {
      "completion_length": 24.59375,
      "epoch": 3.8540706605222734,
      "grad_norm": 26.261770248413086,
      "kl": 1.3076171875,
      "learning_rate": 3.4615384615384616e-08,
      "loss": -0.0548,
      "reward": 1.4355856776237488,
      "reward_std": 0.34860897064208984,
      "reward_std_all": 0.8684708476066589,
      "rewards-max/forensics_sinkhorn_reward": 0.988842636346817,
      "rewards-max/forensics_strict_edge_reward": 0.9375866651535034,
      "rewards-max/forensics_tiou_reward": 0.9988231658935547,
      "rewards/forensics_sinkhorn_reward": 0.7043806314468384,
      "rewards/forensics_strict_edge_reward": 0.12698844075202942,
      "rewards/forensics_tiou_reward": 0.6042166650295258,
      "step": 1256
    },
    {
      "completion_length": 26.5625,
      "epoch": 3.857142857142857,
      "grad_norm": 26.730192184448242,
      "kl": 1.259765625,
      "learning_rate": 3.384615384615384e-08,
      "loss": 0.0282,
      "reward": 1.7484791278839111,
      "reward_std": 0.4476809799671173,
      "reward_std_all": 0.7726479768753052,
      "rewards-max/forensics_sinkhorn_reward": 0.9893617928028107,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.7994310557842255,
      "rewards/forensics_strict_edge_reward": 0.24368947744369507,
      "rewards/forensics_tiou_reward": 0.7053585350513458,
      "step": 1257
    },
    {
      "completion_length": 27.765625,
      "epoch": 3.860215053763441,
      "grad_norm": 241.98699951171875,
      "kl": 4.009765625,
      "learning_rate": 3.3076923076923076e-08,
      "loss": 0.1589,
      "reward": 1.6531071066856384,
      "reward_std": 0.36791300773620605,
      "reward_std_all": 0.7802126407623291,
      "rewards-max/forensics_sinkhorn_reward": 0.9851788580417633,
      "rewards-max/forensics_strict_edge_reward": 0.882964164018631,
      "rewards-max/forensics_tiou_reward": 0.9895080029964447,
      "rewards/forensics_sinkhorn_reward": 0.7762988209724426,
      "rewards/forensics_strict_edge_reward": 0.19945801049470901,
      "rewards/forensics_tiou_reward": 0.6773503124713898,
      "step": 1258
    },
    {
      "completion_length": 24.140625,
      "epoch": 3.8632872503840243,
      "grad_norm": 236.38392639160156,
      "kl": 5.6435546875,
      "learning_rate": 3.230769230769231e-08,
      "loss": 0.1523,
      "reward": 1.7349082231521606,
      "reward_std": 0.4463614970445633,
      "reward_std_all": 0.7668578922748566,
      "rewards-max/forensics_sinkhorn_reward": 0.9881570935249329,
      "rewards-max/forensics_strict_edge_reward": 0.9291328191757202,
      "rewards-max/forensics_tiou_reward": 0.9964306950569153,
      "rewards/forensics_sinkhorn_reward": 0.8503859341144562,
      "rewards/forensics_strict_edge_reward": 0.18243017047643661,
      "rewards/forensics_tiou_reward": 0.702092170715332,
      "step": 1259
    },
    {
      "completion_length": 23.109375,
      "epoch": 3.8663594470046085,
      "grad_norm": 12.924217224121094,
      "kl": 0.5,
      "learning_rate": 3.153846153846154e-08,
      "loss": -0.0031,
      "reward": 1.603817641735077,
      "reward_std": 0.41904836893081665,
      "reward_std_all": 0.8448738753795624,
      "rewards-max/forensics_sinkhorn_reward": 0.9872036278247833,
      "rewards-max/forensics_strict_edge_reward": 0.9375866651535034,
      "rewards-max/forensics_tiou_reward": 0.9950099885463715,
      "rewards/forensics_sinkhorn_reward": 0.7998315095901489,
      "rewards/forensics_strict_edge_reward": 0.16660068929195404,
      "rewards/forensics_tiou_reward": 0.6373855471611023,
      "step": 1260
    },
    {
      "completion_length": 26.65625,
      "epoch": 3.869431643625192,
      "grad_norm": 11.29742431640625,
      "kl": 0.54296875,
      "learning_rate": 3.076923076923077e-08,
      "loss": 0.0279,
      "reward": 1.8051802515983582,
      "reward_std": 0.2605019770562649,
      "reward_std_all": 0.6693152189254761,
      "rewards-max/forensics_sinkhorn_reward": 0.989959716796875,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.8522661626338959,
      "rewards/forensics_strict_edge_reward": 0.21696128696203232,
      "rewards/forensics_tiou_reward": 0.7359527945518494,
      "step": 1261
    },
    {
      "completion_length": 25.234375,
      "epoch": 3.8725038402457757,
      "grad_norm": 16.13218879699707,
      "kl": 0.74609375,
      "learning_rate": 3e-08,
      "loss": -0.0253,
      "reward": 1.7884461283683777,
      "reward_std": 0.31274092197418213,
      "reward_std_all": 0.7079493403434753,
      "rewards-max/forensics_sinkhorn_reward": 0.9895022213459015,
      "rewards-max/forensics_strict_edge_reward": 0.9016326665878296,
      "rewards-max/forensics_tiou_reward": 0.9994332194328308,
      "rewards/forensics_sinkhorn_reward": 0.8619344532489777,
      "rewards/forensics_strict_edge_reward": 0.1915382668375969,
      "rewards/forensics_tiou_reward": 0.7349734902381897,
      "step": 1262
    },
    {
      "completion_length": 23.671875,
      "epoch": 3.8755760368663594,
      "grad_norm": 15.601272583007812,
      "kl": 0.662109375,
      "learning_rate": 2.9230769230769228e-08,
      "loss": -0.0349,
      "reward": 1.8069576025009155,
      "reward_std": 0.4853302091360092,
      "reward_std_all": 0.7603846490383148,
      "rewards-max/forensics_sinkhorn_reward": 0.989287406206131,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.8482238352298737,
      "rewards/forensics_strict_edge_reward": 0.2467772513628006,
      "rewards/forensics_tiou_reward": 0.7119566202163696,
      "step": 1263
    },
    {
      "completion_length": 28.09375,
      "epoch": 3.878648233486943,
      "grad_norm": 14.753595352172852,
      "kl": 0.7109375,
      "learning_rate": 2.846153846153846e-08,
      "loss": 0.128,
      "reward": 1.7571059465408325,
      "reward_std": 0.4113054573535919,
      "reward_std_all": 0.7467829585075378,
      "rewards-max/forensics_sinkhorn_reward": 0.9893693625926971,
      "rewards-max/forensics_strict_edge_reward": 0.9375866651535034,
      "rewards-max/forensics_tiou_reward": 0.9992975294589996,
      "rewards/forensics_sinkhorn_reward": 0.8054988384246826,
      "rewards/forensics_strict_edge_reward": 0.2319807931780815,
      "rewards/forensics_tiou_reward": 0.7196262776851654,
      "step": 1264
    },
    {
      "completion_length": 25.625,
      "epoch": 3.881720430107527,
      "grad_norm": 13.161871910095215,
      "kl": 0.607421875,
      "learning_rate": 2.769230769230769e-08,
      "loss": -0.1025,
      "reward": 1.709755539894104,
      "reward_std": 0.4333253502845764,
      "reward_std_all": 0.7710744738578796,
      "rewards-max/forensics_sinkhorn_reward": 0.9876930713653564,
      "rewards-max/forensics_strict_edge_reward": 0.7973326146602631,
      "rewards-max/forensics_tiou_reward": 0.9968572854995728,
      "rewards/forensics_sinkhorn_reward": 0.8073714375495911,
      "rewards/forensics_strict_edge_reward": 0.19292420148849487,
      "rewards/forensics_tiou_reward": 0.7094599008560181,
      "step": 1265
    },
    {
      "completion_length": 23.421875,
      "epoch": 3.8847926267281108,
      "grad_norm": 28.9925537109375,
      "kl": 0.73046875,
      "learning_rate": 2.6923076923076925e-08,
      "loss": -0.1097,
      "reward": 1.4559826254844666,
      "reward_std": 0.242342047393322,
      "reward_std_all": 0.8248766660690308,
      "rewards-max/forensics_sinkhorn_reward": 0.9750259518623352,
      "rewards-max/forensics_strict_edge_reward": 0.6554845571517944,
      "rewards-max/forensics_tiou_reward": 0.9643982350826263,
      "rewards/forensics_sinkhorn_reward": 0.7342789769172668,
      "rewards/forensics_strict_edge_reward": 0.10123488120734692,
      "rewards/forensics_tiou_reward": 0.6204688549041748,
      "step": 1266
    },
    {
      "completion_length": 24.609375,
      "epoch": 3.8878648233486945,
      "grad_norm": 11.767115592956543,
      "kl": 0.60546875,
      "learning_rate": 2.6153846153846152e-08,
      "loss": -0.151,
      "reward": 1.8649728894233704,
      "reward_std": 0.43465128540992737,
      "reward_std_all": 0.7188845276832581,
      "rewards-max/forensics_sinkhorn_reward": 0.9902741312980652,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.8620392680168152,
      "rewards/forensics_strict_edge_reward": 0.2548331022262573,
      "rewards/forensics_tiou_reward": 0.7481004893779755,
      "step": 1267
    },
    {
      "completion_length": 27.15625,
      "epoch": 3.890937019969278,
      "grad_norm": 12.291648864746094,
      "kl": 0.4091796875,
      "learning_rate": 2.5384615384615382e-08,
      "loss": -0.0883,
      "reward": 1.8393176198005676,
      "reward_std": 0.40436048805713654,
      "reward_std_all": 0.7149522304534912,
      "rewards-max/forensics_sinkhorn_reward": 0.9878456890583038,
      "rewards-max/forensics_strict_edge_reward": 0.8783542811870575,
      "rewards-max/forensics_tiou_reward": 0.9968444108963013,
      "rewards/forensics_sinkhorn_reward": 0.871396541595459,
      "rewards/forensics_strict_edge_reward": 0.22029922902584076,
      "rewards/forensics_tiou_reward": 0.7476219236850739,
      "step": 1268
    },
    {
      "completion_length": 24.734375,
      "epoch": 3.8940092165898617,
      "grad_norm": 11.442252159118652,
      "kl": 0.54296875,
      "learning_rate": 2.4615384615384616e-08,
      "loss": -0.1432,
      "reward": 1.7353405356407166,
      "reward_std": 0.47024451196193695,
      "reward_std_all": 0.810890257358551,
      "rewards-max/forensics_sinkhorn_reward": 0.9896320700645447,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.8145990967750549,
      "rewards/forensics_strict_edge_reward": 0.22113485634326935,
      "rewards/forensics_tiou_reward": 0.6996065974235535,
      "step": 1269
    },
    {
      "completion_length": 25.203125,
      "epoch": 3.8970814132104454,
      "grad_norm": 14.790864944458008,
      "kl": 0.591796875,
      "learning_rate": 2.3846153846153846e-08,
      "loss": -0.0905,
      "reward": 1.7021183967590332,
      "reward_std": 0.29558658599853516,
      "reward_std_all": 0.7363360524177551,
      "rewards-max/forensics_sinkhorn_reward": 0.9887199103832245,
      "rewards-max/forensics_strict_edge_reward": 0.8827048540115356,
      "rewards-max/forensics_tiou_reward": 0.9967537820339203,
      "rewards/forensics_sinkhorn_reward": 0.8391863107681274,
      "rewards/forensics_strict_edge_reward": 0.13666844367980957,
      "rewards/forensics_tiou_reward": 0.7262635827064514,
      "step": 1270
    },
    {
      "completion_length": 28.09375,
      "epoch": 3.900153609831029,
      "grad_norm": 10.869563102722168,
      "kl": 0.4287109375,
      "learning_rate": 2.3076923076923076e-08,
      "loss": -0.1252,
      "reward": 1.6491222977638245,
      "reward_std": 0.3816661685705185,
      "reward_std_all": 0.7920848727226257,
      "rewards-max/forensics_sinkhorn_reward": 0.9866542518138885,
      "rewards-max/forensics_strict_edge_reward": 0.8704091012477875,
      "rewards-max/forensics_tiou_reward": 0.9949719607830048,
      "rewards/forensics_sinkhorn_reward": 0.7713395357131958,
      "rewards/forensics_strict_edge_reward": 0.19788622856140137,
      "rewards/forensics_tiou_reward": 0.6798964142799377,
      "step": 1271
    },
    {
      "completion_length": 24.5,
      "epoch": 3.903225806451613,
      "grad_norm": 17.323955535888672,
      "kl": 0.966796875,
      "learning_rate": 2.2307692307692306e-08,
      "loss": -0.0881,
      "reward": 1.6650150418281555,
      "reward_std": 0.4455624967813492,
      "reward_std_all": 0.7973555624485016,
      "rewards-max/forensics_sinkhorn_reward": 0.9888214468955994,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.8136672675609589,
      "rewards/forensics_strict_edge_reward": 0.18605192005634308,
      "rewards/forensics_tiou_reward": 0.6652956902980804,
      "step": 1272
    },
    {
      "completion_length": 27.53125,
      "epoch": 3.9062980030721968,
      "grad_norm": 13.540145874023438,
      "kl": 0.611328125,
      "learning_rate": 2.1538461538461537e-08,
      "loss": -0.0828,
      "reward": 1.836137056350708,
      "reward_std": 0.4471944719552994,
      "reward_std_all": 0.7360112369060516,
      "rewards-max/forensics_sinkhorn_reward": 0.9904079437255859,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.8595080375671387,
      "rewards/forensics_strict_edge_reward": 0.21943630278110504,
      "rewards/forensics_tiou_reward": 0.7571927011013031,
      "step": 1273
    },
    {
      "completion_length": 27.03125,
      "epoch": 3.9093701996927805,
      "grad_norm": 13.43804931640625,
      "kl": 0.576171875,
      "learning_rate": 2.0769230769230767e-08,
      "loss": 0.0391,
      "reward": 1.7159450054168701,
      "reward_std": 0.4438772350549698,
      "reward_std_all": 0.8202894032001495,
      "rewards-max/forensics_sinkhorn_reward": 0.9892876148223877,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.8046803176403046,
      "rewards/forensics_strict_edge_reward": 0.24847817420959473,
      "rewards/forensics_tiou_reward": 0.6627865433692932,
      "step": 1274
    },
    {
      "completion_length": 23.625,
      "epoch": 3.912442396313364,
      "grad_norm": 14.036649703979492,
      "kl": 0.892578125,
      "learning_rate": 2e-08,
      "loss": -0.0457,
      "reward": 1.8193655610084534,
      "reward_std": 0.5136665254831314,
      "reward_std_all": 0.7717899978160858,
      "rewards-max/forensics_sinkhorn_reward": 0.9896764755249023,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.8459870219230652,
      "rewards/forensics_strict_edge_reward": 0.2405387908220291,
      "rewards/forensics_tiou_reward": 0.7328397631645203,
      "step": 1275
    },
    {
      "completion_length": 26.75,
      "epoch": 3.9155145929339477,
      "grad_norm": 408.7628479003906,
      "kl": 6.091796875,
      "learning_rate": 1.923076923076923e-08,
      "loss": 0.2342,
      "reward": 1.82349693775177,
      "reward_std": 0.44843824207782745,
      "reward_std_all": 0.7347189486026764,
      "rewards-max/forensics_sinkhorn_reward": 0.9899720847606659,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.8196179866790771,
      "rewards/forensics_strict_edge_reward": 0.3135419338941574,
      "rewards/forensics_tiou_reward": 0.6903370320796967,
      "step": 1276
    },
    {
      "completion_length": 22.59375,
      "epoch": 3.9185867895545314,
      "grad_norm": 11.00517749786377,
      "kl": 0.4833984375,
      "learning_rate": 1.846153846153846e-08,
      "loss": -0.0866,
      "reward": 1.7261818647384644,
      "reward_std": 0.4312555491924286,
      "reward_std_all": 0.7819935977458954,
      "rewards-max/forensics_sinkhorn_reward": 0.9846229255199432,
      "rewards-max/forensics_strict_edge_reward": 0.8959447741508484,
      "rewards-max/forensics_tiou_reward": 0.9861112534999847,
      "rewards/forensics_sinkhorn_reward": 0.82353475689888,
      "rewards/forensics_strict_edge_reward": 0.20852622389793396,
      "rewards/forensics_tiou_reward": 0.6941209137439728,
      "step": 1277
    },
    {
      "completion_length": 25.921875,
      "epoch": 3.921658986175115,
      "grad_norm": 13.351114273071289,
      "kl": 0.58203125,
      "learning_rate": 1.769230769230769e-08,
      "loss": -0.069,
      "reward": 1.7911851406097412,
      "reward_std": 0.5668556690216064,
      "reward_std_all": 0.8178378343582153,
      "rewards-max/forensics_sinkhorn_reward": 0.9901642501354218,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.8155029118061066,
      "rewards/forensics_strict_edge_reward": 0.2537343204021454,
      "rewards/forensics_tiou_reward": 0.721947968006134,
      "step": 1278
    },
    {
      "completion_length": 23.125,
      "epoch": 3.924731182795699,
      "grad_norm": 10.636469841003418,
      "kl": 0.4892578125,
      "learning_rate": 1.692307692307692e-08,
      "loss": -0.0416,
      "reward": 1.5870813131332397,
      "reward_std": 0.4009993225336075,
      "reward_std_all": 0.8209291994571686,
      "rewards-max/forensics_sinkhorn_reward": 0.9824747145175934,
      "rewards-max/forensics_strict_edge_reward": 0.7506196796894073,
      "rewards-max/forensics_tiou_reward": 0.9860055446624756,
      "rewards/forensics_sinkhorn_reward": 0.7512633502483368,
      "rewards/forensics_strict_edge_reward": 0.19493351876735687,
      "rewards/forensics_tiou_reward": 0.640884518623352,
      "step": 1279
    },
    {
      "completion_length": 27.4375,
      "epoch": 3.9278033794162828,
      "grad_norm": 10.950037002563477,
      "kl": 0.65625,
      "learning_rate": 1.6153846153846155e-08,
      "loss": -0.0679,
      "reward": 2.015050947666168,
      "reward_std": 0.38574211299419403,
      "reward_std_all": 0.6348808109760284,
      "rewards-max/forensics_sinkhorn_reward": 0.9905001223087311,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.8864091336727142,
      "rewards/forensics_strict_edge_reward": 0.3284018337726593,
      "rewards/forensics_tiou_reward": 0.8002400398254395,
      "step": 1280
    },
    {
      "completion_length": 29.640625,
      "epoch": 3.9308755760368665,
      "grad_norm": 14.59674072265625,
      "kl": 0.615234375,
      "learning_rate": 1.5384615384615385e-08,
      "loss": -0.0036,
      "reward": 1.8423824310302734,
      "reward_std": 0.4443907290697098,
      "reward_std_all": 0.7039048075675964,
      "rewards-max/forensics_sinkhorn_reward": 0.9895243644714355,
      "rewards-max/forensics_strict_edge_reward": 0.9414820969104767,
      "rewards-max/forensics_tiou_reward": 0.999061793088913,
      "rewards/forensics_sinkhorn_reward": 0.8412134051322937,
      "rewards/forensics_strict_edge_reward": 0.2489265725016594,
      "rewards/forensics_tiou_reward": 0.752242386341095,
      "step": 1281
    },
    {
      "completion_length": 27.859375,
      "epoch": 3.93394777265745,
      "grad_norm": 12.804311752319336,
      "kl": 0.591796875,
      "learning_rate": 1.4615384615384614e-08,
      "loss": 0.0159,
      "reward": 1.8118246793746948,
      "reward_std": 0.6148938834667206,
      "reward_std_all": 0.8194100558757782,
      "rewards-max/forensics_sinkhorn_reward": 0.9897341430187225,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.8247081339359283,
      "rewards/forensics_strict_edge_reward": 0.25961514562368393,
      "rewards/forensics_tiou_reward": 0.7275012731552124,
      "step": 1282
    },
    {
      "completion_length": 22.25,
      "epoch": 3.9370199692780337,
      "grad_norm": 11.525525093078613,
      "kl": 0.71484375,
      "learning_rate": 1.3846153846153846e-08,
      "loss": -0.0524,
      "reward": 1.752752423286438,
      "reward_std": 0.42782557010650635,
      "reward_std_all": 0.742870420217514,
      "rewards-max/forensics_sinkhorn_reward": 0.982952356338501,
      "rewards-max/forensics_strict_edge_reward": 0.7933230996131897,
      "rewards-max/forensics_tiou_reward": 0.9834766685962677,
      "rewards/forensics_sinkhorn_reward": 0.8596707880496979,
      "rewards/forensics_strict_edge_reward": 0.17536719143390656,
      "rewards/forensics_tiou_reward": 0.7177145481109619,
      "step": 1283
    },
    {
      "completion_length": 26.28125,
      "epoch": 3.9400921658986174,
      "grad_norm": 12.759981155395508,
      "kl": 0.80859375,
      "learning_rate": 1.3076923076923076e-08,
      "loss": 0.0315,
      "reward": 1.8824703097343445,
      "reward_std": 0.40352538228034973,
      "reward_std_all": 0.675883024930954,
      "rewards-max/forensics_sinkhorn_reward": 0.9886520206928253,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.8765160441398621,
      "rewards/forensics_strict_edge_reward": 0.21011298894882202,
      "rewards/forensics_tiou_reward": 0.7958413660526276,
      "step": 1284
    },
    {
      "completion_length": 24.984375,
      "epoch": 3.943164362519201,
      "grad_norm": 11.733210563659668,
      "kl": 0.5546875,
      "learning_rate": 1.2307692307692308e-08,
      "loss": -0.1328,
      "reward": 1.546183466911316,
      "reward_std": 0.5104977786540985,
      "reward_std_all": 0.9034488499164581,
      "rewards-max/forensics_sinkhorn_reward": 0.9898883104324341,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.7745327651500702,
      "rewards/forensics_strict_edge_reward": 0.19169535860419273,
      "rewards/forensics_tiou_reward": 0.5799553096294403,
      "step": 1285
    },
    {
      "completion_length": 26.046875,
      "epoch": 3.946236559139785,
      "grad_norm": 39.586769104003906,
      "kl": 0.908203125,
      "learning_rate": 1.1538461538461538e-08,
      "loss": -0.0532,
      "reward": 1.82859206199646,
      "reward_std": 0.45790165662765503,
      "reward_std_all": 0.7189729511737823,
      "rewards-max/forensics_sinkhorn_reward": 0.9900635182857513,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.8266577422618866,
      "rewards/forensics_strict_edge_reward": 0.24317152798175812,
      "rewards/forensics_tiou_reward": 0.7587628066539764,
      "step": 1286
    },
    {
      "completion_length": 25.109375,
      "epoch": 3.9493087557603688,
      "grad_norm": 14.832757949829102,
      "kl": 0.58203125,
      "learning_rate": 1.0769230769230768e-08,
      "loss": 0.0707,
      "reward": 1.8879910707473755,
      "reward_std": 0.45886316895484924,
      "reward_std_all": 0.7148401141166687,
      "rewards-max/forensics_sinkhorn_reward": 0.9896138310432434,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.8514697253704071,
      "rewards/forensics_strict_edge_reward": 0.2560339719057083,
      "rewards/forensics_tiou_reward": 0.7804874777793884,
      "step": 1287
    },
    {
      "completion_length": 26.046875,
      "epoch": 3.9523809523809526,
      "grad_norm": 13.095833778381348,
      "kl": 0.529296875,
      "learning_rate": 1e-08,
      "loss": -0.087,
      "reward": 1.7956379652023315,
      "reward_std": 0.4324949085712433,
      "reward_std_all": 0.7523831129074097,
      "rewards-max/forensics_sinkhorn_reward": 0.9894715845584869,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.8234241604804993,
      "rewards/forensics_strict_edge_reward": 0.24777637422084808,
      "rewards/forensics_tiou_reward": 0.7244375050067902,
      "step": 1288
    },
    {
      "completion_length": 23.921875,
      "epoch": 3.955453149001536,
      "grad_norm": 10.670540809631348,
      "kl": 0.4228515625,
      "learning_rate": 9.23076923076923e-09,
      "loss": -0.1076,
      "reward": 1.6660825610160828,
      "reward_std": 0.5150482952594757,
      "reward_std_all": 0.8351568579673767,
      "rewards-max/forensics_sinkhorn_reward": 0.9889152646064758,
      "rewards-max/forensics_strict_edge_reward": 0.9677534997463226,
      "rewards-max/forensics_tiou_reward": 0.9983034729957581,
      "rewards/forensics_sinkhorn_reward": 0.7797108292579651,
      "rewards/forensics_strict_edge_reward": 0.20897822082042694,
      "rewards/forensics_tiou_reward": 0.6773934364318848,
      "step": 1289
    },
    {
      "completion_length": 27.375,
      "epoch": 3.9585253456221197,
      "grad_norm": 12.45615005493164,
      "kl": 0.59375,
      "learning_rate": 8.46153846153846e-09,
      "loss": 0.0446,
      "reward": 1.5585106015205383,
      "reward_std": 0.5541020035743713,
      "reward_std_all": 0.8872076869010925,
      "rewards-max/forensics_sinkhorn_reward": 0.9892683923244476,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.7251475751399994,
      "rewards/forensics_strict_edge_reward": 0.20348207652568817,
      "rewards/forensics_tiou_reward": 0.6298809051513672,
      "step": 1290
    },
    {
      "completion_length": 23.71875,
      "epoch": 3.9615975422427034,
      "grad_norm": 17.023387908935547,
      "kl": 0.6123046875,
      "learning_rate": 7.692307692307693e-09,
      "loss": -0.1023,
      "reward": 1.8322492241859436,
      "reward_std": 0.4832625240087509,
      "reward_std_all": 0.7245194911956787,
      "rewards-max/forensics_sinkhorn_reward": 0.9819149971008301,
      "rewards-max/forensics_strict_edge_reward": 0.756708562374115,
      "rewards-max/forensics_tiou_reward": 0.9803550243377686,
      "rewards/forensics_sinkhorn_reward": 0.8667533993721008,
      "rewards/forensics_strict_edge_reward": 0.20867323875427246,
      "rewards/forensics_tiou_reward": 0.7568225562572479,
      "step": 1291
    },
    {
      "completion_length": 25.40625,
      "epoch": 3.964669738863287,
      "grad_norm": 13.348856925964355,
      "kl": 0.5302734375,
      "learning_rate": 6.923076923076923e-09,
      "loss": -0.0916,
      "reward": 1.6290634274482727,
      "reward_std": 0.41453492641448975,
      "reward_std_all": 0.8091398775577545,
      "rewards-max/forensics_sinkhorn_reward": 0.9878363013267517,
      "rewards-max/forensics_strict_edge_reward": 0.870158314704895,
      "rewards-max/forensics_tiou_reward": 0.9939160645008087,
      "rewards/forensics_sinkhorn_reward": 0.8082385659217834,
      "rewards/forensics_strict_edge_reward": 0.1420997902750969,
      "rewards/forensics_tiou_reward": 0.6787249743938446,
      "step": 1292
    },
    {
      "completion_length": 26.546875,
      "epoch": 3.967741935483871,
      "grad_norm": 28.682363510131836,
      "kl": 0.5888671875,
      "learning_rate": 6.153846153846154e-09,
      "loss": -0.1937,
      "reward": 1.4220408201217651,
      "reward_std": 0.3702574446797371,
      "reward_std_all": 0.8955661058425903,
      "rewards-max/forensics_sinkhorn_reward": 0.9885673224925995,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.682708203792572,
      "rewards/forensics_strict_edge_reward": 0.18893352895975113,
      "rewards/forensics_tiou_reward": 0.5503991097211838,
      "step": 1293
    },
    {
      "completion_length": 24.71875,
      "epoch": 3.9708141321044548,
      "grad_norm": 26.230836868286133,
      "kl": 0.7392578125,
      "learning_rate": 5.384615384615384e-09,
      "loss": -0.1115,
      "reward": 1.4311066269874573,
      "reward_std": 0.5971891283988953,
      "reward_std_all": 0.9710775315761566,
      "rewards-max/forensics_sinkhorn_reward": 0.9900251030921936,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.7337995767593384,
      "rewards/forensics_strict_edge_reward": 0.1362369954586029,
      "rewards/forensics_tiou_reward": 0.5610700249671936,
      "step": 1294
    },
    {
      "completion_length": 28.890625,
      "epoch": 3.9738863287250386,
      "grad_norm": 11.348989486694336,
      "kl": 0.8203125,
      "learning_rate": 4.615384615384615e-09,
      "loss": 0.0596,
      "reward": 1.6441694498062134,
      "reward_std": 0.3551916480064392,
      "reward_std_all": 0.7719842195510864,
      "rewards-max/forensics_sinkhorn_reward": 0.9885794520378113,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.7995854020118713,
      "rewards/forensics_strict_edge_reward": 0.18916251510381699,
      "rewards/forensics_tiou_reward": 0.6554215252399445,
      "step": 1295
    },
    {
      "completion_length": 28.515625,
      "epoch": 3.976958525345622,
      "grad_norm": 11.293933868408203,
      "kl": 0.701171875,
      "learning_rate": 3.846153846153846e-09,
      "loss": -0.0165,
      "reward": 1.83852219581604,
      "reward_std": 0.5723601281642914,
      "reward_std_all": 0.7768541276454926,
      "rewards-max/forensics_sinkhorn_reward": 0.9904163777828217,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.805812805891037,
      "rewards/forensics_strict_edge_reward": 0.3104090392589569,
      "rewards/forensics_tiou_reward": 0.7223004996776581,
      "step": 1296
    },
    {
      "completion_length": 29.375,
      "epoch": 3.980030721966206,
      "grad_norm": 11.443639755249023,
      "kl": 0.61328125,
      "learning_rate": 3.076923076923077e-09,
      "loss": -0.0142,
      "reward": 1.7530664205551147,
      "reward_std": 0.3524307608604431,
      "reward_std_all": 0.7438305020332336,
      "rewards-max/forensics_sinkhorn_reward": 0.9899402558803558,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.7943882942199707,
      "rewards/forensics_strict_edge_reward": 0.27204522490501404,
      "rewards/forensics_tiou_reward": 0.6866329312324524,
      "step": 1297
    },
    {
      "completion_length": 25.484375,
      "epoch": 3.9831029185867894,
      "grad_norm": 14.338702201843262,
      "kl": 0.548828125,
      "learning_rate": 2.3076923076923076e-09,
      "loss": 0.1195,
      "reward": 1.6315026879310608,
      "reward_std": 0.44533318281173706,
      "reward_std_all": 0.818045049905777,
      "rewards-max/forensics_sinkhorn_reward": 0.9890919327735901,
      "rewards-max/forensics_strict_edge_reward": 0.9175800085067749,
      "rewards-max/forensics_tiou_reward": 0.9993399679660797,
      "rewards/forensics_sinkhorn_reward": 0.7551975846290588,
      "rewards/forensics_strict_edge_reward": 0.20107398554682732,
      "rewards/forensics_tiou_reward": 0.6752310395240784,
      "step": 1298
    },
    {
      "completion_length": 24.125,
      "epoch": 3.986175115207373,
      "grad_norm": 12.465336799621582,
      "kl": 0.619140625,
      "learning_rate": 1.5384615384615385e-09,
      "loss": -0.1037,
      "reward": 1.8312382698059082,
      "reward_std": 0.5336365699768066,
      "reward_std_all": 0.7849885821342468,
      "rewards-max/forensics_sinkhorn_reward": 0.9903980493545532,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.8271218240261078,
      "rewards/forensics_strict_edge_reward": 0.28457115590572357,
      "rewards/forensics_tiou_reward": 0.7195452749729156,
      "step": 1299
    },
    {
      "completion_length": 27.625,
      "epoch": 3.989247311827957,
      "grad_norm": 13.985198020935059,
      "kl": 0.640625,
      "learning_rate": 7.692307692307692e-10,
      "loss": 0.1439,
      "reward": 1.8901223540306091,
      "reward_std": 0.4190327823162079,
      "reward_std_all": 0.6998014450073242,
      "rewards-max/forensics_sinkhorn_reward": 0.9896668791770935,
      "rewards-max/forensics_strict_edge_reward": 1.0,
      "rewards-max/forensics_tiou_reward": 1.0,
      "rewards/forensics_sinkhorn_reward": 0.8645029664039612,
      "rewards/forensics_strict_edge_reward": 0.2582629770040512,
      "rewards/forensics_tiou_reward": 0.7673564553260803,
      "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
}