{
  "best_global_step": null,
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 3.992670157068063,
  "eval_steps": 500,
  "global_step": 1908,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "completion_length": 25.96875,
      "epoch": 0.0020942408376963353,
      "grad_norm": 14.333218574523926,
      "kl": 0.0,
      "learning_rate": 1e-06,
      "loss": 0.2018,
      "reward": 0.42689773440361023,
      "reward_std": 0.30605557560920715,
      "reward_std_all": 0.3253412991762161,
      "rewards-max/forensics_sinkhorn_reward": 0.8902730941772461,
      "rewards-max/forensics_strict_edge_reward": 0.6951379179954529,
      "rewards-max/forensics_tiou_reward": 0.794002503156662,
      "rewards/forensics_sinkhorn_reward": 0.3201269060373306,
      "rewards/forensics_strict_edge_reward": 0.09414970502257347,
      "rewards/forensics_tiou_reward": 0.1481894925236702,
      "step": 1
    },
    {
      "completion_length": 24.65625,
      "epoch": 0.004188481675392671,
      "grad_norm": 8.595293045043945,
      "kl": 0.0008602142333984375,
      "learning_rate": 9.99475890985325e-07,
      "loss": 0.1031,
      "reward": 0.4677056521177292,
      "reward_std": 0.25022806972265244,
      "reward_std_all": 0.2783888727426529,
      "rewards-max/forensics_sinkhorn_reward": 0.8674305081367493,
      "rewards-max/forensics_strict_edge_reward": 0.5476561039686203,
      "rewards-max/forensics_tiou_reward": 0.7244332581758499,
      "rewards/forensics_sinkhorn_reward": 0.33291587233543396,
      "rewards/forensics_strict_edge_reward": 0.0729209128767252,
      "rewards/forensics_tiou_reward": 0.12080014497041702,
      "step": 2
    },
    {
      "completion_length": 19.53125,
      "epoch": 0.0062827225130890054,
      "grad_norm": 9.198814392089844,
      "kl": 0.00257110595703125,
      "learning_rate": 9.9895178197065e-07,
      "loss": -0.0459,
      "reward": 0.4932297319173813,
      "reward_std": 0.20744877308607101,
      "reward_std_all": 0.2439279407262802,
      "rewards-max/forensics_sinkhorn_reward": 0.9662204682826996,
      "rewards-max/forensics_strict_edge_reward": 0.552458643913269,
      "rewards-max/forensics_tiou_reward": 0.9118770658969879,
      "rewards/forensics_sinkhorn_reward": 0.3446204513311386,
      "rewards/forensics_strict_edge_reward": 0.07415029406547546,
      "rewards/forensics_tiou_reward": 0.11299005523324013,
      "step": 3
    },
    {
      "completion_length": 21.21875,
      "epoch": 0.008376963350785341,
      "grad_norm": 9.565231323242188,
      "kl": 0.0061187744140625,
      "learning_rate": 9.984276729559748e-07,
      "loss": 0.1119,
      "reward": 0.48898573219776154,
      "reward_std": 0.23189018666744232,
      "reward_std_all": 0.25695569813251495,
      "rewards-max/forensics_sinkhorn_reward": 0.9868029952049255,
      "rewards-max/forensics_strict_edge_reward": 0.9375866651535034,
      "rewards-max/forensics_tiou_reward": 0.9914164543151855,
      "rewards/forensics_sinkhorn_reward": 0.399898424744606,
      "rewards/forensics_strict_edge_reward": 0.09478703886270523,
      "rewards/forensics_tiou_reward": 0.17138175666332245,
      "step": 4
    },
    {
      "completion_length": 17.875,
      "epoch": 0.010471204188481676,
      "grad_norm": 10.432210922241211,
      "kl": 0.0074462890625,
      "learning_rate": 9.979035639412998e-07,
      "loss": 0.0284,
      "reward": 0.5223326683044434,
      "reward_std": 0.16908474639058113,
      "reward_std_all": 0.20667129009962082,
      "rewards-max/forensics_sinkhorn_reward": 0.9767535924911499,
      "rewards-max/forensics_strict_edge_reward": 0.5202973932027817,
      "rewards-max/forensics_tiou_reward": 0.9523898363113403,
      "rewards/forensics_sinkhorn_reward": 0.3854602128267288,
      "rewards/forensics_strict_edge_reward": 0.06017846241593361,
      "rewards/forensics_tiou_reward": 0.0978722795844078,
      "step": 5
    },
    {
      "completion_length": 24.296875,
      "epoch": 0.012565445026178011,
      "grad_norm": 8.494117736816406,
      "kl": 0.03350830078125,
      "learning_rate": 9.973794549266246e-07,
      "loss": -0.0306,
      "reward": 0.5473285615444183,
      "reward_std": 0.17800140380859375,
      "reward_std_all": 0.20991145819425583,
      "rewards-max/forensics_sinkhorn_reward": 0.9776652157306671,
      "rewards-max/forensics_strict_edge_reward": 0.7245818972587585,
      "rewards-max/forensics_tiou_reward": 0.9558943808078766,
      "rewards/forensics_sinkhorn_reward": 0.4140594005584717,
      "rewards/forensics_strict_edge_reward": 0.07894312590360641,
      "rewards/forensics_tiou_reward": 0.12791769951581955,
      "step": 6
    },
    {
      "completion_length": 18.5625,
      "epoch": 0.014659685863874346,
      "grad_norm": 9.774253845214844,
      "kl": 0.13720703125,
      "learning_rate": 9.968553459119496e-07,
      "loss": 0.1154,
      "reward": 0.4502907395362854,
      "reward_std": 0.24419735372066498,
      "reward_std_all": 0.2764802724123001,
      "rewards-max/forensics_sinkhorn_reward": 0.8027625381946564,
      "rewards-max/forensics_strict_edge_reward": 0.23637648299336433,
      "rewards-max/forensics_tiou_reward": 0.5569483786821365,
      "rewards/forensics_sinkhorn_reward": 0.3238002359867096,
      "rewards/forensics_strict_edge_reward": 0.02777661825530231,
      "rewards/forensics_tiou_reward": 0.07719171233475208,
      "step": 7
    },
    {
      "completion_length": 26.953125,
      "epoch": 0.016753926701570682,
      "grad_norm": 8.248300552368164,
      "kl": 0.02166748046875,
      "learning_rate": 9.963312368972746e-07,
      "loss": 0.0096,
      "reward": 0.5145599842071533,
      "reward_std": 0.24083347618579865,
      "reward_std_all": 0.2541917413473129,
      "rewards-max/forensics_sinkhorn_reward": 0.6809594035148621,
      "rewards-max/forensics_strict_edge_reward": 0.3021143823862076,
      "rewards-max/forensics_tiou_reward": 0.3323000520467758,
      "rewards/forensics_sinkhorn_reward": 0.34918269515037537,
      "rewards/forensics_strict_edge_reward": 0.03377094678580761,
      "rewards/forensics_tiou_reward": 0.07652347907423973,
      "step": 8
    },
    {
      "completion_length": 17.890625,
      "epoch": 0.018848167539267015,
      "grad_norm": 7.625665664672852,
      "kl": 0.039306640625,
      "learning_rate": 9.958071278825996e-07,
      "loss": 0.0561,
      "reward": 0.5493830740451813,
      "reward_std": 0.20902878791093826,
      "reward_std_all": 0.22043398767709732,
      "rewards-max/forensics_sinkhorn_reward": 0.9784817099571228,
      "rewards-max/forensics_strict_edge_reward": 0.8205901980400085,
      "rewards-max/forensics_tiou_reward": 0.9596747756004333,
      "rewards/forensics_sinkhorn_reward": 0.4114839732646942,
      "rewards/forensics_strict_edge_reward": 0.0970444455742836,
      "rewards/forensics_tiou_reward": 0.16675709187984467,
      "step": 9
    },
    {
      "completion_length": 23.5625,
      "epoch": 0.020942408376963352,
      "grad_norm": 12.70693588256836,
      "kl": 0.2607421875,
      "learning_rate": 9.952830188679244e-07,
      "loss": 0.1457,
      "reward": 0.4744264781475067,
      "reward_std": 0.16745079308748245,
      "reward_std_all": 0.23412670195102692,
      "rewards-max/forensics_sinkhorn_reward": 0.9761435687541962,
      "rewards-max/forensics_strict_edge_reward": 0.5976370573043823,
      "rewards-max/forensics_tiou_reward": 0.9502918124198914,
      "rewards/forensics_sinkhorn_reward": 0.36575934290885925,
      "rewards/forensics_strict_edge_reward": 0.055137749761343,
      "rewards/forensics_tiou_reward": 0.1596624180674553,
      "step": 10
    },
    {
      "completion_length": 22.859375,
      "epoch": 0.023036649214659685,
      "grad_norm": 13.224580764770508,
      "kl": 0.13916015625,
      "learning_rate": 9.947589098532494e-07,
      "loss": 0.079,
      "reward": 0.49177776277065277,
      "reward_std": 0.17425663024187088,
      "reward_std_all": 0.22691109031438828,
      "rewards-max/forensics_sinkhorn_reward": 0.9158868789672852,
      "rewards-max/forensics_strict_edge_reward": 0.5728201940655708,
      "rewards-max/forensics_tiou_reward": 0.8343192338943481,
      "rewards/forensics_sinkhorn_reward": 0.34758029878139496,
      "rewards/forensics_strict_edge_reward": 0.061567931436002254,
      "rewards/forensics_tiou_reward": 0.12372305616736412,
      "step": 11
    },
    {
      "completion_length": 19.546875,
      "epoch": 0.025130890052356022,
      "grad_norm": 289.2701416015625,
      "kl": 5.4814453125,
      "learning_rate": 9.942348008385744e-07,
      "loss": 0.1905,
      "reward": 0.5269018709659576,
      "reward_std": 0.1855892688035965,
      "reward_std_all": 0.21474102139472961,
      "rewards-max/forensics_sinkhorn_reward": 0.9742425680160522,
      "rewards-max/forensics_strict_edge_reward": 0.744240403175354,
      "rewards-max/forensics_tiou_reward": 0.9434200525283813,
      "rewards/forensics_sinkhorn_reward": 0.3749946504831314,
      "rewards/forensics_strict_edge_reward": 0.06960604153573513,
      "rewards/forensics_tiou_reward": 0.13158134371042252,
      "step": 12
    },
    {
      "completion_length": 28.140625,
      "epoch": 0.027225130890052355,
      "grad_norm": 15.453714370727539,
      "kl": 0.22314453125,
      "learning_rate": 9.937106918238992e-07,
      "loss": 0.0675,
      "reward": 0.4976688325405121,
      "reward_std": 0.21703410148620605,
      "reward_std_all": 0.2462759166955948,
      "rewards-max/forensics_sinkhorn_reward": 0.9016018807888031,
      "rewards-max/forensics_strict_edge_reward": 0.48714151978492737,
      "rewards-max/forensics_tiou_reward": 0.7869512736797333,
      "rewards/forensics_sinkhorn_reward": 0.3358522951602936,
      "rewards/forensics_strict_edge_reward": 0.07626041769981384,
      "rewards/forensics_tiou_reward": 0.14197886735200882,
      "step": 13
    },
    {
      "completion_length": 26.875,
      "epoch": 0.02931937172774869,
      "grad_norm": 10.114933013916016,
      "kl": 0.21630859375,
      "learning_rate": 9.931865828092243e-07,
      "loss": -0.0485,
      "reward": 0.5649673342704773,
      "reward_std": 0.1914149522781372,
      "reward_std_all": 0.20955000072717667,
      "rewards-max/forensics_sinkhorn_reward": 0.9782141149044037,
      "rewards-max/forensics_strict_edge_reward": 0.7412298321723938,
      "rewards-max/forensics_tiou_reward": 0.9582109749317169,
      "rewards/forensics_sinkhorn_reward": 0.4451189786195755,
      "rewards/forensics_strict_edge_reward": 0.09188967570662498,
      "rewards/forensics_tiou_reward": 0.20277103781700134,
      "step": 14
    },
    {
      "completion_length": 16.734375,
      "epoch": 0.031413612565445025,
      "grad_norm": 6.457924842834473,
      "kl": 0.0562744140625,
      "learning_rate": 9.926624737945493e-07,
      "loss": 0.0163,
      "reward": 0.5584966838359833,
      "reward_std": 0.11562708392739296,
      "reward_std_all": 0.17405598610639572,
      "rewards-max/forensics_sinkhorn_reward": 0.973569393157959,
      "rewards-max/forensics_strict_edge_reward": 0.7014762461185455,
      "rewards-max/forensics_tiou_reward": 0.9403329193592072,
      "rewards/forensics_sinkhorn_reward": 0.437533363699913,
      "rewards/forensics_strict_edge_reward": 0.03791782446205616,
      "rewards/forensics_tiou_reward": 0.12010636925697327,
      "step": 15
    },
    {
      "completion_length": 21.625,
      "epoch": 0.033507853403141365,
      "grad_norm": 7.312981605529785,
      "kl": 0.0501708984375,
      "learning_rate": 9.92138364779874e-07,
      "loss": -0.03,
      "reward": 0.541819840669632,
      "reward_std": 0.16435690969228745,
      "reward_std_all": 0.1978154331445694,
      "rewards-max/forensics_sinkhorn_reward": 0.96001797914505,
      "rewards-max/forensics_strict_edge_reward": 0.5957017540931702,
      "rewards-max/forensics_tiou_reward": 0.8928579986095428,
      "rewards/forensics_sinkhorn_reward": 0.42477788031101227,
      "rewards/forensics_strict_edge_reward": 0.05167274735867977,
      "rewards/forensics_tiou_reward": 0.14997662231326103,
      "step": 16
    },
    {
      "completion_length": 19.390625,
      "epoch": 0.0356020942408377,
      "grad_norm": 15.326556205749512,
      "kl": 0.466552734375,
      "learning_rate": 9.91614255765199e-07,
      "loss": -0.0656,
      "reward": 0.49376848340034485,
      "reward_std": 0.17868413776159286,
      "reward_std_all": 0.2255372256040573,
      "rewards-max/forensics_sinkhorn_reward": 0.6278069913387299,
      "rewards-max/forensics_strict_edge_reward": 0.21820500493049622,
      "rewards-max/forensics_tiou_reward": 0.35191650688648224,
      "rewards/forensics_sinkhorn_reward": 0.33141301572322845,
      "rewards/forensics_strict_edge_reward": 0.037190718576312065,
      "rewards/forensics_tiou_reward": 0.05756084620952606,
      "step": 17
    },
    {
      "completion_length": 16.71875,
      "epoch": 0.03769633507853403,
      "grad_norm": 8.759509086608887,
      "kl": 0.05126953125,
      "learning_rate": 9.91090146750524e-07,
      "loss": 0.1083,
      "reward": 0.5274200737476349,
      "reward_std": 0.11983940005302429,
      "reward_std_all": 0.18555118888616562,
      "rewards-max/forensics_sinkhorn_reward": 0.7701367437839508,
      "rewards-max/forensics_strict_edge_reward": 0.49217386078089476,
      "rewards-max/forensics_tiou_reward": 0.5709054172039032,
      "rewards/forensics_sinkhorn_reward": 0.41744863986968994,
      "rewards/forensics_strict_edge_reward": 0.09906740178121254,
      "rewards/forensics_tiou_reward": 0.13491839170455933,
      "step": 18
    },
    {
      "completion_length": 15.265625,
      "epoch": 0.039790575916230364,
      "grad_norm": 8.79886531829834,
      "kl": 0.0457763671875,
      "learning_rate": 9.90566037735849e-07,
      "loss": 0.1436,
      "reward": 0.5821963846683502,
      "reward_std": 0.12808215618133545,
      "reward_std_all": 0.16552993655204773,
      "rewards-max/forensics_sinkhorn_reward": 0.7685345411300659,
      "rewards-max/forensics_strict_edge_reward": 0.40417245775461197,
      "rewards-max/forensics_tiou_reward": 0.6132246851921082,
      "rewards/forensics_sinkhorn_reward": 0.4675067663192749,
      "rewards/forensics_strict_edge_reward": 0.07996816746890545,
      "rewards/forensics_tiou_reward": 0.15360993146896362,
      "step": 19
    },
    {
      "completion_length": 24.4375,
      "epoch": 0.041884816753926704,
      "grad_norm": 7.431363105773926,
      "kl": 0.0528564453125,
      "learning_rate": 9.90041928721174e-07,
      "loss": 0.0781,
      "reward": 0.5361181795597076,
      "reward_std": 0.16400451213121414,
      "reward_std_all": 0.20500074326992035,
      "rewards-max/forensics_sinkhorn_reward": 0.9493648409843445,
      "rewards-max/forensics_strict_edge_reward": 0.5801038593053818,
      "rewards-max/forensics_tiou_reward": 0.8504274487495422,
      "rewards/forensics_sinkhorn_reward": 0.4097682386636734,
      "rewards/forensics_strict_edge_reward": 0.06738964468240738,
      "rewards/forensics_tiou_reward": 0.13341761007905006,
      "step": 20
    },
    {
      "completion_length": 17.0,
      "epoch": 0.04397905759162304,
      "grad_norm": 6.405170440673828,
      "kl": 0.058349609375,
      "learning_rate": 9.89517819706499e-07,
      "loss": 0.1,
      "reward": 0.5978019833564758,
      "reward_std": 0.2029353454709053,
      "reward_std_all": 0.20052431523799896,
      "rewards-max/forensics_sinkhorn_reward": 0.9791342914104462,
      "rewards-max/forensics_strict_edge_reward": 0.7923295497894287,
      "rewards-max/forensics_tiou_reward": 0.9618062078952789,
      "rewards/forensics_sinkhorn_reward": 0.4890454113483429,
      "rewards/forensics_strict_edge_reward": 0.10681470483541489,
      "rewards/forensics_tiou_reward": 0.20529936999082565,
      "step": 21
    },
    {
      "completion_length": 23.71875,
      "epoch": 0.04607329842931937,
      "grad_norm": 6.439187526702881,
      "kl": 0.1240234375,
      "learning_rate": 9.889937106918237e-07,
      "loss": -0.0956,
      "reward": 0.5414337515830994,
      "reward_std": 0.16924946755170822,
      "reward_std_all": 0.20801035314798355,
      "rewards-max/forensics_sinkhorn_reward": 0.9644517302513123,
      "rewards-max/forensics_strict_edge_reward": 0.4457087963819504,
      "rewards-max/forensics_tiou_reward": 0.9059503078460693,
      "rewards/forensics_sinkhorn_reward": 0.4245310425758362,
      "rewards/forensics_strict_edge_reward": 0.04118214454501867,
      "rewards/forensics_tiou_reward": 0.13361697271466255,
      "step": 22
    },
    {
      "completion_length": 19.4375,
      "epoch": 0.048167539267015703,
      "grad_norm": 7.269342422485352,
      "kl": 0.081787109375,
      "learning_rate": 9.884696016771487e-07,
      "loss": -0.1102,
      "reward": 0.5599043071269989,
      "reward_std": 0.15295936167240143,
      "reward_std_all": 0.1891302615404129,
      "rewards-max/forensics_sinkhorn_reward": 0.8754719197750092,
      "rewards-max/forensics_strict_edge_reward": 0.6267998367547989,
      "rewards-max/forensics_tiou_reward": 0.7536846101284027,
      "rewards/forensics_sinkhorn_reward": 0.42065297067165375,
      "rewards/forensics_strict_edge_reward": 0.08240972086787224,
      "rewards/forensics_tiou_reward": 0.12996890395879745,
      "step": 23
    },
    {
      "completion_length": 16.765625,
      "epoch": 0.050261780104712044,
      "grad_norm": 7.062565803527832,
      "kl": 0.08154296875,
      "learning_rate": 9.879454926624737e-07,
      "loss": 0.0878,
      "reward": 0.580248236656189,
      "reward_std": 0.13071956485509872,
      "reward_std_all": 0.17203310877084732,
      "rewards-max/forensics_sinkhorn_reward": 0.9879108965396881,
      "rewards-max/forensics_strict_edge_reward": 0.9677534997463226,
      "rewards-max/forensics_tiou_reward": 0.9957446753978729,
      "rewards/forensics_sinkhorn_reward": 0.49911630153656006,
      "rewards/forensics_strict_edge_reward": 0.1279846653342247,
      "rewards/forensics_tiou_reward": 0.21723679453134537,
      "step": 24
    },
    {
      "completion_length": 19.953125,
      "epoch": 0.05235602094240838,
      "grad_norm": 8.34396743774414,
      "kl": 0.10546875,
      "learning_rate": 9.874213836477988e-07,
      "loss": 0.1366,
      "reward": 0.543950542807579,
      "reward_std": 0.1592746526002884,
      "reward_std_all": 0.2019680291414261,
      "rewards-max/forensics_sinkhorn_reward": 0.9802057147026062,
      "rewards-max/forensics_strict_edge_reward": 0.5747895687818527,
      "rewards-max/forensics_tiou_reward": 0.9654661417007446,
      "rewards/forensics_sinkhorn_reward": 0.42259883880615234,
      "rewards/forensics_strict_edge_reward": 0.0765438936650753,
      "rewards/forensics_tiou_reward": 0.16472849249839783,
      "step": 25
    },
    {
      "completion_length": 20.09375,
      "epoch": 0.05445026178010471,
      "grad_norm": 8.551602363586426,
      "kl": 0.07080078125,
      "learning_rate": 9.868972746331238e-07,
      "loss": 0.056,
      "reward": 0.5478739142417908,
      "reward_std": 0.14042585715651512,
      "reward_std_all": 0.18782535940408707,
      "rewards-max/forensics_sinkhorn_reward": 0.9590001702308655,
      "rewards-max/forensics_strict_edge_reward": 0.397090807557106,
      "rewards-max/forensics_tiou_reward": 0.8860040307044983,
      "rewards/forensics_sinkhorn_reward": 0.41715797781944275,
      "rewards/forensics_strict_edge_reward": 0.048177216202020645,
      "rewards/forensics_tiou_reward": 0.1113051287829876,
      "step": 26
    },
    {
      "completion_length": 18.5625,
      "epoch": 0.05654450261780105,
      "grad_norm": 7.4812421798706055,
      "kl": 0.05517578125,
      "learning_rate": 9.863731656184486e-07,
      "loss": -0.0704,
      "reward": 0.6061698496341705,
      "reward_std": 0.1199222207069397,
      "reward_std_all": 0.1568005234003067,
      "rewards-max/forensics_sinkhorn_reward": 0.9707667231559753,
      "rewards-max/forensics_strict_edge_reward": 0.6930834054946899,
      "rewards-max/forensics_tiou_reward": 0.9295561909675598,
      "rewards/forensics_sinkhorn_reward": 0.4793573170900345,
      "rewards/forensics_strict_edge_reward": 0.09639964625239372,
      "rewards/forensics_tiou_reward": 0.18658408522605896,
      "step": 27
    },
    {
      "completion_length": 18.25,
      "epoch": 0.05863874345549738,
      "grad_norm": 10.952295303344727,
      "kl": 0.2939453125,
      "learning_rate": 9.858490566037736e-07,
      "loss": 0.0878,
      "reward": 0.5380826890468597,
      "reward_std": 0.1729043871164322,
      "reward_std_all": 0.20800791680812836,
      "rewards-max/forensics_sinkhorn_reward": 0.9451852738857269,
      "rewards-max/forensics_strict_edge_reward": 0.44064828753471375,
      "rewards-max/forensics_tiou_reward": 0.8293185234069824,
      "rewards/forensics_sinkhorn_reward": 0.43307484686374664,
      "rewards/forensics_strict_edge_reward": 0.044218096882104874,
      "rewards/forensics_tiou_reward": 0.13804905116558075,
      "step": 28
    },
    {
      "completion_length": 19.546875,
      "epoch": 0.060732984293193716,
      "grad_norm": 10.583398818969727,
      "kl": 0.251953125,
      "learning_rate": 9.853249475890984e-07,
      "loss": 0.0465,
      "reward": 0.5680868923664093,
      "reward_std": 0.18650226294994354,
      "reward_std_all": 0.20323707908391953,
      "rewards-max/forensics_sinkhorn_reward": 0.9850790798664093,
      "rewards-max/forensics_strict_edge_reward": 0.7511312067508698,
      "rewards-max/forensics_tiou_reward": 0.984678715467453,
      "rewards/forensics_sinkhorn_reward": 0.45480571687221527,
      "rewards/forensics_strict_edge_reward": 0.10161906853318214,
      "rewards/forensics_tiou_reward": 0.2050769180059433,
      "step": 29
    },
    {
      "completion_length": 20.21875,
      "epoch": 0.06282722513089005,
      "grad_norm": 10.392802238464355,
      "kl": 0.119384765625,
      "learning_rate": 9.848008385744234e-07,
      "loss": -0.0538,
      "reward": 0.5597319900989532,
      "reward_std": 0.13964997977018356,
      "reward_std_all": 0.18292666971683502,
      "rewards-max/forensics_sinkhorn_reward": 0.9829312860965729,
      "rewards-max/forensics_strict_edge_reward": 0.8655564188957214,
      "rewards-max/forensics_tiou_reward": 0.9762244820594788,
      "rewards/forensics_sinkhorn_reward": 0.44567441940307617,
      "rewards/forensics_strict_edge_reward": 0.10956984758377075,
      "rewards/forensics_tiou_reward": 0.17713582515716553,
      "step": 30
    },
    {
      "completion_length": 21.765625,
      "epoch": 0.06492146596858639,
      "grad_norm": 14.079102516174316,
      "kl": 0.357421875,
      "learning_rate": 9.842767295597484e-07,
      "loss": 0.0124,
      "reward": 0.5614169538021088,
      "reward_std": 0.21746256947517395,
      "reward_std_all": 0.22763599455356598,
      "rewards-max/forensics_sinkhorn_reward": 0.9160918593406677,
      "rewards-max/forensics_strict_edge_reward": 0.7328133881092072,
      "rewards-max/forensics_tiou_reward": 0.8370994627475739,
      "rewards/forensics_sinkhorn_reward": 0.4065547436475754,
      "rewards/forensics_strict_edge_reward": 0.10904104262590408,
      "rewards/forensics_tiou_reward": 0.17078252136707306,
      "step": 31
    },
    {
      "completion_length": 20.078125,
      "epoch": 0.06701570680628273,
      "grad_norm": 4.461878299713135,
      "kl": 0.14208984375,
      "learning_rate": 9.837526205450734e-07,
      "loss": 0.0484,
      "reward": 0.6091541051864624,
      "reward_std": 0.16977013647556305,
      "reward_std_all": 0.18284983187913895,
      "rewards-max/forensics_sinkhorn_reward": 0.984937995672226,
      "rewards-max/forensics_strict_edge_reward": 0.758176863193512,
      "rewards-max/forensics_tiou_reward": 0.9841307401657104,
      "rewards/forensics_sinkhorn_reward": 0.5211576819419861,
      "rewards/forensics_strict_edge_reward": 0.08953206613659859,
      "rewards/forensics_tiou_reward": 0.2243952453136444,
      "step": 32
    },
    {
      "completion_length": 17.765625,
      "epoch": 0.06910994764397906,
      "grad_norm": 7.62797212600708,
      "kl": 0.28662109375,
      "learning_rate": 9.832285115303984e-07,
      "loss": -0.0198,
      "reward": 0.5715345740318298,
      "reward_std": 0.16681459173560143,
      "reward_std_all": 0.18995481729507446,
      "rewards-max/forensics_sinkhorn_reward": 0.9672269821166992,
      "rewards-max/forensics_strict_edge_reward": 0.6110665500164032,
      "rewards-max/forensics_tiou_reward": 0.9177483022212982,
      "rewards/forensics_sinkhorn_reward": 0.41287921369075775,
      "rewards/forensics_strict_edge_reward": 0.0830833651125431,
      "rewards/forensics_tiou_reward": 0.13033022731542587,
      "step": 33
    },
    {
      "completion_length": 19.046875,
      "epoch": 0.0712041884816754,
      "grad_norm": 7.505748748779297,
      "kl": 0.100341796875,
      "learning_rate": 9.827044025157232e-07,
      "loss": -0.0941,
      "reward": 0.51111900806427,
      "reward_std": 0.13104921579360962,
      "reward_std_all": 0.19573547691106796,
      "rewards-max/forensics_sinkhorn_reward": 0.6487936079502106,
      "rewards-max/forensics_strict_edge_reward": 0.25531289726495743,
      "rewards-max/forensics_tiou_reward": 0.31342460215091705,
      "rewards/forensics_sinkhorn_reward": 0.35643328726291656,
      "rewards/forensics_strict_edge_reward": 0.022963611409068108,
      "rewards/forensics_tiou_reward": 0.06603450328111649,
      "step": 34
    },
    {
      "completion_length": 18.046875,
      "epoch": 0.07329842931937172,
      "grad_norm": 8.0450439453125,
      "kl": 0.111328125,
      "learning_rate": 9.821802935010482e-07,
      "loss": -0.0655,
      "reward": 0.5623047351837158,
      "reward_std": 0.16703375428915024,
      "reward_std_all": 0.19878746569156647,
      "rewards-max/forensics_sinkhorn_reward": 0.8643887639045715,
      "rewards-max/forensics_strict_edge_reward": 0.6012860685586929,
      "rewards-max/forensics_tiou_reward": 0.722235307097435,
      "rewards/forensics_sinkhorn_reward": 0.4227715879678726,
      "rewards/forensics_strict_edge_reward": 0.0904785804450512,
      "rewards/forensics_tiou_reward": 0.13906511291861534,
      "step": 35
    },
    {
      "completion_length": 26.09375,
      "epoch": 0.07539267015706806,
      "grad_norm": 7.148483753204346,
      "kl": 0.194580078125,
      "learning_rate": 9.81656184486373e-07,
      "loss": 0.1382,
      "reward": 0.5197812020778656,
      "reward_std": 0.227810300886631,
      "reward_std_all": 0.24129044264554977,
      "rewards-max/forensics_sinkhorn_reward": 0.886327475309372,
      "rewards-max/forensics_strict_edge_reward": 0.5534158498048782,
      "rewards-max/forensics_tiou_reward": 0.774616926908493,
      "rewards/forensics_sinkhorn_reward": 0.39081163704395294,
      "rewards/forensics_strict_edge_reward": 0.06117609888315201,
      "rewards/forensics_tiou_reward": 0.13377533107995987,
      "step": 36
    },
    {
      "completion_length": 21.078125,
      "epoch": 0.0774869109947644,
      "grad_norm": 5.3397345542907715,
      "kl": 0.18896484375,
      "learning_rate": 9.81132075471698e-07,
      "loss": -0.027,
      "reward": 0.5681386590003967,
      "reward_std": 0.1619601920247078,
      "reward_std_all": 0.19040708243846893,
      "rewards-max/forensics_sinkhorn_reward": 0.932278960943222,
      "rewards-max/forensics_strict_edge_reward": 0.6440400332212448,
      "rewards-max/forensics_tiou_reward": 0.8684951364994049,
      "rewards/forensics_sinkhorn_reward": 0.45514243841171265,
      "rewards/forensics_strict_edge_reward": 0.06916531175374985,
      "rewards/forensics_tiou_reward": 0.16326303035020828,
      "step": 37
    },
    {
      "completion_length": 19.921875,
      "epoch": 0.07958115183246073,
      "grad_norm": 9.045787811279297,
      "kl": 0.45751953125,
      "learning_rate": 9.80607966457023e-07,
      "loss": -0.0779,
      "reward": 0.570448100566864,
      "reward_std": 0.17412392795085907,
      "reward_std_all": 0.1988677829504013,
      "rewards-max/forensics_sinkhorn_reward": 0.8649998605251312,
      "rewards-max/forensics_strict_edge_reward": 0.57234688103199,
      "rewards-max/forensics_tiou_reward": 0.7251961827278137,
      "rewards/forensics_sinkhorn_reward": 0.43967530131340027,
      "rewards/forensics_strict_edge_reward": 0.08282900881022215,
      "rewards/forensics_tiou_reward": 0.1787845343351364,
      "step": 38
    },
    {
      "completion_length": 15.75,
      "epoch": 0.08167539267015707,
      "grad_norm": 9.684026718139648,
      "kl": 0.14013671875,
      "learning_rate": 9.80083857442348e-07,
      "loss": 0.1304,
      "reward": 0.5602381825447083,
      "reward_std": 0.15565425157546997,
      "reward_std_all": 0.18619989603757858,
      "rewards-max/forensics_sinkhorn_reward": 0.987609326839447,
      "rewards-max/forensics_strict_edge_reward": 0.8681221902370453,
      "rewards-max/forensics_tiou_reward": 0.9945594370365143,
      "rewards/forensics_sinkhorn_reward": 0.4264197051525116,
      "rewards/forensics_strict_edge_reward": 0.05425824597477913,
      "rewards/forensics_tiou_reward": 0.11221450567245483,
      "step": 39
    },
    {
      "completion_length": 15.546875,
      "epoch": 0.08376963350785341,
      "grad_norm": 8.921499252319336,
      "kl": 0.181884765625,
      "learning_rate": 9.795597484276729e-07,
      "loss": 0.024,
      "reward": 0.5199181139469147,
      "reward_std": 0.14326681941747665,
      "reward_std_all": 0.19768837094306946,
      "rewards-max/forensics_sinkhorn_reward": 0.9761850237846375,
      "rewards-max/forensics_strict_edge_reward": 0.4434577636420727,
      "rewards-max/forensics_tiou_reward": 0.9505986869335175,
      "rewards/forensics_sinkhorn_reward": 0.4001358449459076,
      "rewards/forensics_strict_edge_reward": 0.044643142726272345,
      "rewards/forensics_tiou_reward": 0.09657172858715057,
      "step": 40
    },
    {
      "completion_length": 16.484375,
      "epoch": 0.08586387434554973,
      "grad_norm": 8.450931549072266,
      "kl": 0.15771484375,
      "learning_rate": 9.790356394129979e-07,
      "loss": 0.0331,
      "reward": 0.5457338094711304,
      "reward_std": 0.13778870552778244,
      "reward_std_all": 0.18755149096250534,
      "rewards-max/forensics_sinkhorn_reward": 0.9409999549388885,
      "rewards-max/forensics_strict_edge_reward": 0.6848928332328796,
      "rewards-max/forensics_tiou_reward": 0.8684305548667908,
      "rewards/forensics_sinkhorn_reward": 0.4209326058626175,
      "rewards/forensics_strict_edge_reward": 0.06204856000840664,
      "rewards/forensics_tiou_reward": 0.12292483821511269,
      "step": 41
    },
    {
      "completion_length": 18.796875,
      "epoch": 0.08795811518324607,
      "grad_norm": 7.949613094329834,
      "kl": 0.27783203125,
      "learning_rate": 9.785115303983229e-07,
      "loss": -0.0444,
      "reward": 0.5359997749328613,
      "reward_std": 0.12306143343448639,
      "reward_std_all": 0.18030735850334167,
      "rewards-max/forensics_sinkhorn_reward": 0.8498413264751434,
      "rewards-max/forensics_strict_edge_reward": 0.5185607671737671,
      "rewards-max/forensics_tiou_reward": 0.7060125470161438,
      "rewards/forensics_sinkhorn_reward": 0.4014222323894501,
      "rewards/forensics_strict_edge_reward": 0.03859357908368111,
      "rewards/forensics_tiou_reward": 0.0804568137973547,
      "step": 42
    },
    {
      "completion_length": 23.046875,
      "epoch": 0.09005235602094241,
      "grad_norm": 9.21543025970459,
      "kl": 0.1533203125,
      "learning_rate": 9.779874213836477e-07,
      "loss": 0.0516,
      "reward": 0.5149055421352386,
      "reward_std": 0.13152218610048294,
      "reward_std_all": 0.19390593469142914,
      "rewards-max/forensics_sinkhorn_reward": 0.6317548155784607,
      "rewards-max/forensics_strict_edge_reward": 0.21641530841588974,
      "rewards-max/forensics_tiou_reward": 0.36140647530555725,
      "rewards/forensics_sinkhorn_reward": 0.35948994755744934,
      "rewards/forensics_strict_edge_reward": 0.0249488593544811,
      "rewards/forensics_tiou_reward": 0.056404247879981995,
      "step": 43
    },
    {
      "completion_length": 19.71875,
      "epoch": 0.09214659685863874,
      "grad_norm": 4.698977947235107,
      "kl": 0.11572265625,
      "learning_rate": 9.774633123689727e-07,
      "loss": 0.0704,
      "reward": 0.6209640800952911,
      "reward_std": 0.1674826592206955,
      "reward_std_all": 0.17100010812282562,
      "rewards-max/forensics_sinkhorn_reward": 0.9803604781627655,
      "rewards-max/forensics_strict_edge_reward": 0.5482475310564041,
      "rewards-max/forensics_tiou_reward": 0.9663099050521851,
      "rewards/forensics_sinkhorn_reward": 0.5202966928482056,
      "rewards/forensics_strict_edge_reward": 0.09802316129207611,
      "rewards/forensics_tiou_reward": 0.23520218580961227,
      "step": 44
    },
    {
      "completion_length": 18.78125,
      "epoch": 0.09424083769633508,
      "grad_norm": 7.261701583862305,
      "kl": 0.2021484375,
      "learning_rate": 9.769392033542977e-07,
      "loss": 0.0174,
      "reward": 0.5781529247760773,
      "reward_std": 0.1666855588555336,
      "reward_std_all": 0.1887415051460266,
      "rewards-max/forensics_sinkhorn_reward": 0.9515743851661682,
      "rewards-max/forensics_strict_edge_reward": 0.7258487641811371,
      "rewards-max/forensics_tiou_reward": 0.8629984259605408,
      "rewards/forensics_sinkhorn_reward": 0.4587979316711426,
      "rewards/forensics_strict_edge_reward": 0.12021134793758392,
      "rewards/forensics_tiou_reward": 0.20246533304452896,
      "step": 45
    },
    {
      "completion_length": 17.390625,
      "epoch": 0.09633507853403141,
      "grad_norm": 11.405693054199219,
      "kl": 0.4658203125,
      "learning_rate": 9.764150943396225e-07,
      "loss": 0.149,
      "reward": 0.5587800443172455,
      "reward_std": 0.16977554559707642,
      "reward_std_all": 0.19921615719795227,
      "rewards-max/forensics_sinkhorn_reward": 0.89405557513237,
      "rewards-max/forensics_strict_edge_reward": 0.46594104170799255,
      "rewards-max/forensics_tiou_reward": 0.7798915207386017,
      "rewards/forensics_sinkhorn_reward": 0.44653041660785675,
      "rewards/forensics_strict_edge_reward": 0.06150593748316169,
      "rewards/forensics_tiou_reward": 0.14413348212838173,
      "step": 46
    },
    {
      "completion_length": 17.171875,
      "epoch": 0.09842931937172775,
      "grad_norm": 8.212808609008789,
      "kl": 0.5068359375,
      "learning_rate": 9.758909853249475e-07,
      "loss": -0.0453,
      "reward": 0.5973334312438965,
      "reward_std": 0.1604423224925995,
      "reward_std_all": 0.17522171139717102,
      "rewards-max/forensics_sinkhorn_reward": 0.8729328215122223,
      "rewards-max/forensics_strict_edge_reward": 0.7766622006893158,
      "rewards-max/forensics_tiou_reward": 0.7370699942111969,
      "rewards/forensics_sinkhorn_reward": 0.4524310976266861,
      "rewards/forensics_strict_edge_reward": 0.0744248554110527,
      "rewards/forensics_tiou_reward": 0.13266243785619736,
      "step": 47
    },
    {
      "completion_length": 16.484375,
      "epoch": 0.10052356020942409,
      "grad_norm": 5.440723896026611,
      "kl": 0.2578125,
      "learning_rate": 9.753668763102725e-07,
      "loss": 0.0364,
      "reward": 0.5823900997638702,
      "reward_std": 0.17081771045923233,
      "reward_std_all": 0.18750661611557007,
      "rewards-max/forensics_sinkhorn_reward": 0.987975686788559,
      "rewards-max/forensics_strict_edge_reward": 0.9782643616199493,
      "rewards-max/forensics_tiou_reward": 0.9959999918937683,
      "rewards/forensics_sinkhorn_reward": 0.485880583524704,
      "rewards/forensics_strict_edge_reward": 0.13508376851677895,
      "rewards/forensics_tiou_reward": 0.23334138095378876,
      "step": 48
    },
    {
      "completion_length": 19.125,
      "epoch": 0.10261780104712041,
      "grad_norm": 5.91090726852417,
      "kl": 0.3232421875,
      "learning_rate": 9.748427672955975e-07,
      "loss": -0.0477,
      "reward": 0.5782873034477234,
      "reward_std": 0.11114878952503204,
      "reward_std_all": 0.16162586957216263,
      "rewards-max/forensics_sinkhorn_reward": 0.9808835387229919,
      "rewards-max/forensics_strict_edge_reward": 0.8101339638233185,
      "rewards-max/forensics_tiou_reward": 0.9683175981044769,
      "rewards/forensics_sinkhorn_reward": 0.4611712098121643,
      "rewards/forensics_strict_edge_reward": 0.08520396798849106,
      "rewards/forensics_tiou_reward": 0.14999021589756012,
      "step": 49
    },
    {
      "completion_length": 17.921875,
      "epoch": 0.10471204188481675,
      "grad_norm": 6.072912693023682,
      "kl": 0.1748046875,
      "learning_rate": 9.743186582809223e-07,
      "loss": 0.062,
      "reward": 0.6000583171844482,
      "reward_std": 0.158282108604908,
      "reward_std_all": 0.17468855530023575,
      "rewards-max/forensics_sinkhorn_reward": 0.9791588187217712,
      "rewards-max/forensics_strict_edge_reward": 0.8286563754081726,
      "rewards-max/forensics_tiou_reward": 0.961860328912735,
      "rewards/forensics_sinkhorn_reward": 0.49454449117183685,
      "rewards/forensics_strict_edge_reward": 0.1073552593588829,
      "rewards/forensics_tiou_reward": 0.20623983442783356,
      "step": 50
    },
    {
      "completion_length": 15.859375,
      "epoch": 0.1068062827225131,
      "grad_norm": 6.14975118637085,
      "kl": 0.19775390625,
      "learning_rate": 9.737945492662474e-07,
      "loss": 0.0392,
      "reward": 0.5930162966251373,
      "reward_std": 0.15907319635152817,
      "reward_std_all": 0.1812751218676567,
      "rewards-max/forensics_sinkhorn_reward": 0.9830405712127686,
      "rewards-max/forensics_strict_edge_reward": 0.8612285256385803,
      "rewards-max/forensics_tiou_reward": 0.9767602980136871,
      "rewards/forensics_sinkhorn_reward": 0.5005060881376266,
      "rewards/forensics_strict_edge_reward": 0.08872246369719505,
      "rewards/forensics_tiou_reward": 0.19225969165563583,
      "step": 51
    },
    {
      "completion_length": 18.765625,
      "epoch": 0.10890052356020942,
      "grad_norm": 10.964620590209961,
      "kl": 0.75390625,
      "learning_rate": 9.732704402515722e-07,
      "loss": 0.0137,
      "reward": 0.5470761358737946,
      "reward_std": 0.1645297184586525,
      "reward_std_all": 0.20517026633024216,
      "rewards-max/forensics_sinkhorn_reward": 0.8525512516498566,
      "rewards-max/forensics_strict_edge_reward": 0.3797727897763252,
      "rewards-max/forensics_tiou_reward": 0.7058411985635757,
      "rewards/forensics_sinkhorn_reward": 0.4011344611644745,
      "rewards/forensics_strict_edge_reward": 0.07784645818173885,
      "rewards/forensics_tiou_reward": 0.14610759913921356,
      "step": 52
    },
    {
      "completion_length": 17.875,
      "epoch": 0.11099476439790576,
      "grad_norm": 8.09607219696045,
      "kl": 0.24951171875,
      "learning_rate": 9.727463312368972e-07,
      "loss": 0.1475,
      "reward": 0.5310388207435608,
      "reward_std": 0.14970166981220245,
      "reward_std_all": 0.19867583364248276,
      "rewards-max/forensics_sinkhorn_reward": 0.9883804023265839,
      "rewards-max/forensics_strict_edge_reward": 0.9565287232398987,
      "rewards-max/forensics_tiou_reward": 0.9975958168506622,
      "rewards/forensics_sinkhorn_reward": 0.4098014086484909,
      "rewards/forensics_strict_edge_reward": 0.1144927479326725,
      "rewards/forensics_tiou_reward": 0.12761639803647995,
      "step": 53
    },
    {
      "completion_length": 20.484375,
      "epoch": 0.1130890052356021,
      "grad_norm": 11.049537658691406,
      "kl": 0.275390625,
      "learning_rate": 9.722222222222222e-07,
      "loss": -0.0241,
      "reward": 0.5366222858428955,
      "reward_std": 0.1738114356994629,
      "reward_std_all": 0.2121884822845459,
      "rewards-max/forensics_sinkhorn_reward": 0.8248070478439331,
      "rewards-max/forensics_strict_edge_reward": 0.6005616188049316,
      "rewards-max/forensics_tiou_reward": 0.7199755609035492,
      "rewards/forensics_sinkhorn_reward": 0.36458195745944977,
      "rewards/forensics_strict_edge_reward": 0.06804125756025314,
      "rewards/forensics_tiou_reward": 0.13098474219441414,
      "step": 54
    },
    {
      "completion_length": 21.109375,
      "epoch": 0.11518324607329843,
      "grad_norm": 7.352656841278076,
      "kl": 0.271484375,
      "learning_rate": 9.716981132075472e-07,
      "loss": 0.0889,
      "reward": 0.5485919713973999,
      "reward_std": 0.21208396553993225,
      "reward_std_all": 0.2321346253156662,
      "rewards-max/forensics_sinkhorn_reward": 0.9811897575855255,
      "rewards-max/forensics_strict_edge_reward": 0.8056853413581848,
      "rewards-max/forensics_tiou_reward": 0.9694450795650482,
      "rewards/forensics_sinkhorn_reward": 0.43809810280799866,
      "rewards/forensics_strict_edge_reward": 0.167289599776268,
      "rewards/forensics_tiou_reward": 0.2529982030391693,
      "step": 55
    },
    {
      "completion_length": 19.46875,
      "epoch": 0.11727748691099477,
      "grad_norm": 119.5833969116211,
      "kl": 4.119873046875,
      "learning_rate": 9.71174004192872e-07,
      "loss": 0.1762,
      "reward": 0.5876553654670715,
      "reward_std": 0.13376273214817047,
      "reward_std_all": 0.16113139688968658,
      "rewards-max/forensics_sinkhorn_reward": 0.9853425621986389,
      "rewards-max/forensics_strict_edge_reward": 0.9169811904430389,
      "rewards-max/forensics_tiou_reward": 0.9857001602649689,
      "rewards/forensics_sinkhorn_reward": 0.4181206673383713,
      "rewards/forensics_strict_edge_reward": 0.09197552129626274,
      "rewards/forensics_tiou_reward": 0.15857665240764618,
      "step": 56
    },
    {
      "completion_length": 18.203125,
      "epoch": 0.1193717277486911,
      "grad_norm": 6.570766448974609,
      "kl": 0.32421875,
      "learning_rate": 9.70649895178197e-07,
      "loss": -0.0119,
      "reward": 0.5597575306892395,
      "reward_std": 0.15501949191093445,
      "reward_std_all": 0.18674880266189575,
      "rewards-max/forensics_sinkhorn_reward": 0.8947674930095673,
      "rewards-max/forensics_strict_edge_reward": 0.5215517729520798,
      "rewards-max/forensics_tiou_reward": 0.7876205444335938,
      "rewards/forensics_sinkhorn_reward": 0.4153721481561661,
      "rewards/forensics_strict_edge_reward": 0.05692850798368454,
      "rewards/forensics_tiou_reward": 0.11217545717954636,
      "step": 57
    },
    {
      "completion_length": 19.5,
      "epoch": 0.12146596858638743,
      "grad_norm": 17.825668334960938,
      "kl": 0.8935546875,
      "learning_rate": 9.70125786163522e-07,
      "loss": 0.0651,
      "reward": 0.5717510879039764,
      "reward_std": 0.18204624205827713,
      "reward_std_all": 0.19627036154270172,
      "rewards-max/forensics_sinkhorn_reward": 0.8777883350849152,
      "rewards-max/forensics_strict_edge_reward": 0.7597114443778992,
      "rewards-max/forensics_tiou_reward": 0.7725408375263214,
      "rewards/forensics_sinkhorn_reward": 0.39605239033699036,
      "rewards/forensics_strict_edge_reward": 0.1189660131931305,
      "rewards/forensics_tiou_reward": 0.135769285261631,
      "step": 58
    },
    {
      "completion_length": 18.515625,
      "epoch": 0.12356020942408377,
      "grad_norm": 6.420318603515625,
      "kl": 0.30859375,
      "learning_rate": 9.696016771488468e-07,
      "loss": 0.0304,
      "reward": 0.567543089389801,
      "reward_std": 0.18919587135314941,
      "reward_std_all": 0.21219497174024582,
      "rewards-max/forensics_sinkhorn_reward": 0.8486140370368958,
      "rewards-max/forensics_strict_edge_reward": 0.6287326365709305,
      "rewards-max/forensics_tiou_reward": 0.7116474658250809,
      "rewards/forensics_sinkhorn_reward": 0.43817320466041565,
      "rewards/forensics_strict_edge_reward": 0.09128585644066334,
      "rewards/forensics_tiou_reward": 0.19417590647935867,
      "step": 59
    },
    {
      "completion_length": 19.5,
      "epoch": 0.1256544502617801,
      "grad_norm": 12.740870475769043,
      "kl": 0.52490234375,
      "learning_rate": 9.690775681341718e-07,
      "loss": 0.0218,
      "reward": 0.5715497434139252,
      "reward_std": 0.1371840164065361,
      "reward_std_all": 0.18165059387683868,
      "rewards-max/forensics_sinkhorn_reward": 0.9855551421642303,
      "rewards-max/forensics_strict_edge_reward": 0.8365300297737122,
      "rewards-max/forensics_tiou_reward": 0.986541211605072,
      "rewards/forensics_sinkhorn_reward": 0.47062546014785767,
      "rewards/forensics_strict_edge_reward": 0.16312553733587265,
      "rewards/forensics_tiou_reward": 0.25143320858478546,
      "step": 60
    },
    {
      "completion_length": 23.234375,
      "epoch": 0.12774869109947645,
      "grad_norm": 8.482694625854492,
      "kl": 0.34375,
      "learning_rate": 9.685534591194968e-07,
      "loss": 0.1106,
      "reward": 0.5607052743434906,
      "reward_std": 0.11083833873271942,
      "reward_std_all": 0.16991113126277924,
      "rewards-max/forensics_sinkhorn_reward": 0.9795419573783875,
      "rewards-max/forensics_strict_edge_reward": 0.7958523333072662,
      "rewards-max/forensics_tiou_reward": 0.9632185697555542,
      "rewards/forensics_sinkhorn_reward": 0.42102251946926117,
      "rewards/forensics_strict_edge_reward": 0.08992746472358704,
      "rewards/forensics_tiou_reward": 0.15250754356384277,
      "step": 61
    },
    {
      "completion_length": 17.953125,
      "epoch": 0.12984293193717278,
      "grad_norm": 9.916942596435547,
      "kl": 0.3857421875,
      "learning_rate": 9.680293501048219e-07,
      "loss": -0.1414,
      "reward": 0.5695559978485107,
      "reward_std": 0.12144756317138672,
      "reward_std_all": 0.1682542786002159,
      "rewards-max/forensics_sinkhorn_reward": 0.7893884181976318,
      "rewards-max/forensics_strict_edge_reward": 0.44479985535144806,
      "rewards-max/forensics_tiou_reward": 0.5909869596362114,
      "rewards/forensics_sinkhorn_reward": 0.4151027500629425,
      "rewards/forensics_strict_edge_reward": 0.059892299585044384,
      "rewards/forensics_tiou_reward": 0.08640900813043118,
      "step": 62
    },
    {
      "completion_length": 18.9375,
      "epoch": 0.1319371727748691,
      "grad_norm": 7.368887901306152,
      "kl": 0.21240234375,
      "learning_rate": 9.675052410901467e-07,
      "loss": -0.0149,
      "reward": 0.5477419495582581,
      "reward_std": 0.17366386950016022,
      "reward_std_all": 0.20804967731237411,
      "rewards-max/forensics_sinkhorn_reward": 0.9818016290664673,
      "rewards-max/forensics_strict_edge_reward": 0.6342949271202087,
      "rewards-max/forensics_tiou_reward": 0.9717446565628052,
      "rewards/forensics_sinkhorn_reward": 0.3722183406352997,
      "rewards/forensics_strict_edge_reward": 0.08668326213955879,
      "rewards/forensics_tiou_reward": 0.12671668455004692,
      "step": 63
    },
    {
      "completion_length": 18.453125,
      "epoch": 0.13403141361256546,
      "grad_norm": 7.352686405181885,
      "kl": 0.23046875,
      "learning_rate": 9.669811320754717e-07,
      "loss": 0.0108,
      "reward": 0.5168317556381226,
      "reward_std": 0.13517791777849197,
      "reward_std_all": 0.19669809937477112,
      "rewards-max/forensics_sinkhorn_reward": 0.9133476614952087,
      "rewards-max/forensics_strict_edge_reward": 0.4006081074476242,
      "rewards-max/forensics_tiou_reward": 0.800991952419281,
      "rewards/forensics_sinkhorn_reward": 0.3951900452375412,
      "rewards/forensics_strict_edge_reward": 0.05645742826163769,
      "rewards/forensics_tiou_reward": 0.13850241154432297,
      "step": 64
    },
    {
      "completion_length": 18.46875,
      "epoch": 0.13612565445026178,
      "grad_norm": 7.640765190124512,
      "kl": 0.25048828125,
      "learning_rate": 9.664570230607967e-07,
      "loss": 0.0339,
      "reward": 0.5432865619659424,
      "reward_std": 0.18384572118520737,
      "reward_std_all": 0.21756868809461594,
      "rewards-max/forensics_sinkhorn_reward": 0.9492496252059937,
      "rewards-max/forensics_strict_edge_reward": 0.7065972685813904,
      "rewards-max/forensics_tiou_reward": 0.856177806854248,
      "rewards/forensics_sinkhorn_reward": 0.4017072021961212,
      "rewards/forensics_strict_edge_reward": 0.10980472713708878,
      "rewards/forensics_tiou_reward": 0.16554415971040726,
      "step": 65
    },
    {
      "completion_length": 19.71875,
      "epoch": 0.1382198952879581,
      "grad_norm": 9.268746376037598,
      "kl": 0.35498046875,
      "learning_rate": 9.659329140461215e-07,
      "loss": 0.0416,
      "reward": 0.5778852105140686,
      "reward_std": 0.2210601642727852,
      "reward_std_all": 0.22238470613956451,
      "rewards-max/forensics_sinkhorn_reward": 0.9222644567489624,
      "rewards-max/forensics_strict_edge_reward": 0.7133918255567551,
      "rewards-max/forensics_tiou_reward": 0.8574172556400299,
      "rewards/forensics_sinkhorn_reward": 0.438353955745697,
      "rewards/forensics_strict_edge_reward": 0.15019220113754272,
      "rewards/forensics_tiou_reward": 0.21871694922447205,
      "step": 66
    },
    {
      "completion_length": 18.578125,
      "epoch": 0.14031413612565444,
      "grad_norm": 345.5955505371094,
      "kl": 9.0625,
      "learning_rate": 9.654088050314465e-07,
      "loss": 0.5225,
      "reward": 0.48680783808231354,
      "reward_std": 0.1495850794017315,
      "reward_std_all": 0.21851155161857605,
      "rewards-max/forensics_sinkhorn_reward": 0.8156806230545044,
      "rewards-max/forensics_strict_edge_reward": 0.5845613926649094,
      "rewards-max/forensics_tiou_reward": 0.6695508658885956,
      "rewards/forensics_sinkhorn_reward": 0.33896784484386444,
      "rewards/forensics_strict_edge_reward": 0.051489291712641716,
      "rewards/forensics_tiou_reward": 0.08228468149900436,
      "step": 67
    },
    {
      "completion_length": 15.859375,
      "epoch": 0.1424083769633508,
      "grad_norm": 8.639934539794922,
      "kl": 0.25439453125,
      "learning_rate": 9.648846960167715e-07,
      "loss": 0.1025,
      "reward": 0.5860958993434906,
      "reward_std": 0.16126230359077454,
      "reward_std_all": 0.183254636824131,
      "rewards-max/forensics_sinkhorn_reward": 0.9855235517024994,
      "rewards-max/forensics_strict_edge_reward": 0.8371284604072571,
      "rewards-max/forensics_tiou_reward": 0.9863553643226624,
      "rewards/forensics_sinkhorn_reward": 0.450883224606514,
      "rewards/forensics_strict_edge_reward": 0.1282915621995926,
      "rewards/forensics_tiou_reward": 0.17890606820583344,
      "step": 68
    },
    {
      "completion_length": 15.28125,
      "epoch": 0.14450261780104712,
      "grad_norm": 9.613874435424805,
      "kl": 0.24560546875,
      "learning_rate": 9.643605870020965e-07,
      "loss": 0.1342,
      "reward": 0.6094939708709717,
      "reward_std": 0.20681575685739517,
      "reward_std_all": 0.19815723598003387,
      "rewards-max/forensics_sinkhorn_reward": 0.9434049129486084,
      "rewards-max/forensics_strict_edge_reward": 0.6777393221855164,
      "rewards-max/forensics_tiou_reward": 0.83677077293396,
      "rewards/forensics_sinkhorn_reward": 0.5036628097295761,
      "rewards/forensics_strict_edge_reward": 0.13563036546111107,
      "rewards/forensics_tiou_reward": 0.23267383128404617,
      "step": 69
    },
    {
      "completion_length": 16.609375,
      "epoch": 0.14659685863874344,
      "grad_norm": 12.624470710754395,
      "kl": 0.337890625,
      "learning_rate": 9.638364779874213e-07,
      "loss": -0.0033,
      "reward": 0.5416646897792816,
      "reward_std": 0.18159707635641098,
      "reward_std_all": 0.20474886149168015,
      "rewards-max/forensics_sinkhorn_reward": 0.8577893078327179,
      "rewards-max/forensics_strict_edge_reward": 0.6233633011579514,
      "rewards-max/forensics_tiou_reward": 0.7188520282506943,
      "rewards/forensics_sinkhorn_reward": 0.3990073800086975,
      "rewards/forensics_strict_edge_reward": 0.0998157300055027,
      "rewards/forensics_tiou_reward": 0.13154086843132973,
      "step": 70
    },
    {
      "completion_length": 16.109375,
      "epoch": 0.1486910994764398,
      "grad_norm": 10.62109375,
      "kl": 0.7763671875,
      "learning_rate": 9.633123689727463e-07,
      "loss": 0.086,
      "reward": 0.5613532662391663,
      "reward_std": 0.1814444586634636,
      "reward_std_all": 0.20855803042650223,
      "rewards-max/forensics_sinkhorn_reward": 0.9504071474075317,
      "rewards-max/forensics_strict_edge_reward": 0.6941103935241699,
      "rewards-max/forensics_tiou_reward": 0.9042420089244843,
      "rewards/forensics_sinkhorn_reward": 0.41957154870033264,
      "rewards/forensics_strict_edge_reward": 0.13308383151888847,
      "rewards/forensics_tiou_reward": 0.15364820510149002,
      "step": 71
    },
    {
      "completion_length": 18.296875,
      "epoch": 0.15078534031413612,
      "grad_norm": 7.173511505126953,
      "kl": 0.37890625,
      "learning_rate": 9.627882599580711e-07,
      "loss": -0.0012,
      "reward": 0.5903283655643463,
      "reward_std": 0.18521897494792938,
      "reward_std_all": 0.19618768990039825,
      "rewards-max/forensics_sinkhorn_reward": 0.9674198627471924,
      "rewards-max/forensics_strict_edge_reward": 0.7244633734226227,
      "rewards-max/forensics_tiou_reward": 0.9320684671401978,
      "rewards/forensics_sinkhorn_reward": 0.4551491439342499,
      "rewards/forensics_strict_edge_reward": 0.09820066392421722,
      "rewards/forensics_tiou_reward": 0.19487767666578293,
      "step": 72
    },
    {
      "completion_length": 18.015625,
      "epoch": 0.15287958115183245,
      "grad_norm": 7.574920654296875,
      "kl": 0.3291015625,
      "learning_rate": 9.622641509433961e-07,
      "loss": 0.0941,
      "reward": 0.5922363996505737,
      "reward_std": 0.22152404487133026,
      "reward_std_all": 0.21321118623018265,
      "rewards-max/forensics_sinkhorn_reward": 0.9801809787750244,
      "rewards-max/forensics_strict_edge_reward": 0.8205508291721344,
      "rewards-max/forensics_tiou_reward": 0.965608686208725,
      "rewards/forensics_sinkhorn_reward": 0.45716094970703125,
      "rewards/forensics_strict_edge_reward": 0.11996998637914658,
      "rewards/forensics_tiou_reward": 0.21765853464603424,
      "step": 73
    },
    {
      "completion_length": 17.09375,
      "epoch": 0.1549738219895288,
      "grad_norm": 9.238973617553711,
      "kl": 0.3076171875,
      "learning_rate": 9.617400419287211e-07,
      "loss": 0.0805,
      "reward": 0.5486830770969391,
      "reward_std": 0.16116059944033623,
      "reward_std_all": 0.1955752745270729,
      "rewards-max/forensics_sinkhorn_reward": 0.9645492136478424,
      "rewards-max/forensics_strict_edge_reward": 0.658512219786644,
      "rewards-max/forensics_tiou_reward": 0.9086816012859344,
      "rewards/forensics_sinkhorn_reward": 0.424981027841568,
      "rewards/forensics_strict_edge_reward": 0.08112374134361744,
      "rewards/forensics_tiou_reward": 0.14410784095525742,
      "step": 74
    },
    {
      "completion_length": 17.234375,
      "epoch": 0.15706806282722513,
      "grad_norm": 7.886631011962891,
      "kl": 0.5068359375,
      "learning_rate": 9.612159329140462e-07,
      "loss": 0.0559,
      "reward": 0.6016789078712463,
      "reward_std": 0.18353015184402466,
      "reward_std_all": 0.18976137787103653,
      "rewards-max/forensics_sinkhorn_reward": 0.9532087743282318,
      "rewards-max/forensics_strict_edge_reward": 0.7299555838108063,
      "rewards-max/forensics_tiou_reward": 0.9077325165271759,
      "rewards/forensics_sinkhorn_reward": 0.4841351956129074,
      "rewards/forensics_strict_edge_reward": 0.11687701940536499,
      "rewards/forensics_tiou_reward": 0.21896467357873917,
      "step": 75
    },
    {
      "completion_length": 16.1875,
      "epoch": 0.15916230366492146,
      "grad_norm": 7.449006080627441,
      "kl": 0.427734375,
      "learning_rate": 9.60691823899371e-07,
      "loss": 0.1001,
      "reward": 0.582434356212616,
      "reward_std": 0.1933717206120491,
      "reward_std_all": 0.19684581458568573,
      "rewards-max/forensics_sinkhorn_reward": 0.8561492264270782,
      "rewards-max/forensics_strict_edge_reward": 0.6076286435127258,
      "rewards-max/forensics_tiou_reward": 0.7309147864580154,
      "rewards/forensics_sinkhorn_reward": 0.46207717061042786,
      "rewards/forensics_strict_edge_reward": 0.11112282425165176,
      "rewards/forensics_tiou_reward": 0.18428850919008255,
      "step": 76
    },
    {
      "completion_length": 17.0625,
      "epoch": 0.1612565445026178,
      "grad_norm": 7.52979040145874,
      "kl": 0.23779296875,
      "learning_rate": 9.60167714884696e-07,
      "loss": -0.018,
      "reward": 0.5890253782272339,
      "reward_std": 0.19317079335451126,
      "reward_std_all": 0.20404578000307083,
      "rewards-max/forensics_sinkhorn_reward": 0.9886631667613983,
      "rewards-max/forensics_strict_edge_reward": 0.9890114367008209,
      "rewards-max/forensics_tiou_reward": 0.9987096786499023,
      "rewards/forensics_sinkhorn_reward": 0.48262326419353485,
      "rewards/forensics_strict_edge_reward": 0.1421472132205963,
      "rewards/forensics_tiou_reward": 0.23128291219472885,
      "step": 77
    },
    {
      "completion_length": 17.265625,
      "epoch": 0.16335078534031414,
      "grad_norm": 7.7434821128845215,
      "kl": 0.5625,
      "learning_rate": 9.59643605870021e-07,
      "loss": 0.1896,
      "reward": 0.5595799684524536,
      "reward_std": 0.15497423708438873,
      "reward_std_all": 0.19191698729991913,
      "rewards-max/forensics_sinkhorn_reward": 0.8736795783042908,
      "rewards-max/forensics_strict_edge_reward": 0.6521023660898209,
      "rewards-max/forensics_tiou_reward": 0.7470346987247467,
      "rewards/forensics_sinkhorn_reward": 0.4078405797481537,
      "rewards/forensics_strict_edge_reward": 0.08325764164328575,
      "rewards/forensics_tiou_reward": 0.1337125413119793,
      "step": 78
    },
    {
      "completion_length": 17.015625,
      "epoch": 0.16544502617801046,
      "grad_norm": 11.653070449829102,
      "kl": 0.34765625,
      "learning_rate": 9.591194968553458e-07,
      "loss": -0.0414,
      "reward": 0.6164878904819489,
      "reward_std": 0.1631578654050827,
      "reward_std_all": 0.16767232865095139,
      "rewards-max/forensics_sinkhorn_reward": 0.9733112156391144,
      "rewards-max/forensics_strict_edge_reward": 0.758176863193512,
      "rewards-max/forensics_tiou_reward": 0.9395429790019989,
      "rewards/forensics_sinkhorn_reward": 0.48068109154701233,
      "rewards/forensics_strict_edge_reward": 0.0935874804854393,
      "rewards/forensics_tiou_reward": 0.184559166431427,
      "step": 79
    },
    {
      "completion_length": 19.625,
      "epoch": 0.16753926701570682,
      "grad_norm": 6.013023376464844,
      "kl": 0.3076171875,
      "learning_rate": 9.585953878406708e-07,
      "loss": -0.0374,
      "reward": 0.5813277363777161,
      "reward_std": 0.16166915744543076,
      "reward_std_all": 0.18237106502056122,
      "rewards-max/forensics_sinkhorn_reward": 0.791322261095047,
      "rewards-max/forensics_strict_edge_reward": 0.7442108243703842,
      "rewards-max/forensics_tiou_reward": 0.7391054332256317,
      "rewards/forensics_sinkhorn_reward": 0.4202291816473007,
      "rewards/forensics_strict_edge_reward": 0.1014179140329361,
      "rewards/forensics_tiou_reward": 0.15530123561620712,
      "step": 80
    },
    {
      "completion_length": 18.125,
      "epoch": 0.16963350785340314,
      "grad_norm": 9.946266174316406,
      "kl": 0.3388671875,
      "learning_rate": 9.580712788259958e-07,
      "loss": 0.1228,
      "reward": 0.6074149906635284,
      "reward_std": 0.1780976951122284,
      "reward_std_all": 0.18282169103622437,
      "rewards-max/forensics_sinkhorn_reward": 0.8987698554992676,
      "rewards-max/forensics_strict_edge_reward": 0.6816410273313522,
      "rewards-max/forensics_tiou_reward": 0.8086568713188171,
      "rewards/forensics_sinkhorn_reward": 0.4630892127752304,
      "rewards/forensics_strict_edge_reward": 0.09535011649131775,
      "rewards/forensics_tiou_reward": 0.16521327197551727,
      "step": 81
    },
    {
      "completion_length": 17.53125,
      "epoch": 0.17172774869109947,
      "grad_norm": 15.449721336364746,
      "kl": 1.0029296875,
      "learning_rate": 9.575471698113206e-07,
      "loss": -0.0011,
      "reward": 0.5719272494316101,
      "reward_std": 0.143865205347538,
      "reward_std_all": 0.18471118807792664,
      "rewards-max/forensics_sinkhorn_reward": 0.984210878610611,
      "rewards-max/forensics_strict_edge_reward": 0.9075756669044495,
      "rewards-max/forensics_tiou_reward": 0.9813174605369568,
      "rewards/forensics_sinkhorn_reward": 0.43020816147327423,
      "rewards/forensics_strict_edge_reward": 0.08409330248832703,
      "rewards/forensics_tiou_reward": 0.18711590021848679,
      "step": 82
    },
    {
      "completion_length": 18.234375,
      "epoch": 0.17382198952879582,
      "grad_norm": 7.90647554397583,
      "kl": 0.3310546875,
      "learning_rate": 9.570230607966456e-07,
      "loss": 0.0991,
      "reward": 0.5894314348697662,
      "reward_std": 0.1831129938364029,
      "reward_std_all": 0.19514188915491104,
      "rewards-max/forensics_sinkhorn_reward": 0.9700269103050232,
      "rewards-max/forensics_strict_edge_reward": 0.5575894266366959,
      "rewards-max/forensics_tiou_reward": 0.9267630875110626,
      "rewards/forensics_sinkhorn_reward": 0.4431185871362686,
      "rewards/forensics_strict_edge_reward": 0.08916684240102768,
      "rewards/forensics_tiou_reward": 0.19173773378133774,
      "step": 83
    },
    {
      "completion_length": 18.328125,
      "epoch": 0.17591623036649215,
      "grad_norm": 13.687812805175781,
      "kl": 0.830078125,
      "learning_rate": 9.564989517819706e-07,
      "loss": 0.1031,
      "reward": 0.5963675379753113,
      "reward_std": 0.1792677491903305,
      "reward_std_all": 0.1892579421401024,
      "rewards-max/forensics_sinkhorn_reward": 0.9784121215343475,
      "rewards-max/forensics_strict_edge_reward": 0.7485599517822266,
      "rewards-max/forensics_tiou_reward": 0.9583459496498108,
      "rewards/forensics_sinkhorn_reward": 0.4524468332529068,
      "rewards/forensics_strict_edge_reward": 0.11304863169789314,
      "rewards/forensics_tiou_reward": 0.1929517388343811,
      "step": 84
    },
    {
      "completion_length": 18.71875,
      "epoch": 0.17801047120418848,
      "grad_norm": 7.719659805297852,
      "kl": 0.5185546875,
      "learning_rate": 9.559748427672956e-07,
      "loss": 0.1065,
      "reward": 0.5934948027133942,
      "reward_std": 0.15367064625024796,
      "reward_std_all": 0.17270180583000183,
      "rewards-max/forensics_sinkhorn_reward": 0.9878678917884827,
      "rewards-max/forensics_strict_edge_reward": 0.9677534997463226,
      "rewards-max/forensics_tiou_reward": 0.995585024356842,
      "rewards/forensics_sinkhorn_reward": 0.4668109565973282,
      "rewards/forensics_strict_edge_reward": 0.14823134988546371,
      "rewards/forensics_tiou_reward": 0.22651007026433945,
      "step": 85
    },
    {
      "completion_length": 17.375,
      "epoch": 0.18010471204188483,
      "grad_norm": 7.8084235191345215,
      "kl": 0.501953125,
      "learning_rate": 9.554507337526204e-07,
      "loss": 0.1001,
      "reward": 0.6345780491828918,
      "reward_std": 0.14383551850914955,
      "reward_std_all": 0.15202323347330093,
      "rewards-max/forensics_sinkhorn_reward": 0.9837551116943359,
      "rewards-max/forensics_strict_edge_reward": 0.8465339839458466,
      "rewards-max/forensics_tiou_reward": 0.9794284403324127,
      "rewards/forensics_sinkhorn_reward": 0.5179859101772308,
      "rewards/forensics_strict_edge_reward": 0.13078760728240013,
      "rewards/forensics_tiou_reward": 0.22582262009382248,
      "step": 86
    },
    {
      "completion_length": 17.71875,
      "epoch": 0.18219895287958116,
      "grad_norm": 10.964433670043945,
      "kl": 0.875,
      "learning_rate": 9.549266247379455e-07,
      "loss": -0.1373,
      "reward": 0.6012722253799438,
      "reward_std": 0.1380235143005848,
      "reward_std_all": 0.16337230801582336,
      "rewards-max/forensics_sinkhorn_reward": 0.884562760591507,
      "rewards-max/forensics_strict_edge_reward": 0.5740105211734772,
      "rewards-max/forensics_tiou_reward": 0.7690053582191467,
      "rewards/forensics_sinkhorn_reward": 0.4707503765821457,
      "rewards/forensics_strict_edge_reward": 0.07755517214536667,
      "rewards/forensics_tiou_reward": 0.14435339719057083,
      "step": 87
    },
    {
      "completion_length": 18.8125,
      "epoch": 0.18429319371727748,
      "grad_norm": 8.038347244262695,
      "kl": 0.587890625,
      "learning_rate": 9.544025157232703e-07,
      "loss": 0.1012,
      "reward": 0.5985247492790222,
      "reward_std": 0.19403541088104248,
      "reward_std_all": 0.194083072245121,
      "rewards-max/forensics_sinkhorn_reward": 0.9243771433830261,
      "rewards-max/forensics_strict_edge_reward": 0.8635591566562653,
      "rewards-max/forensics_tiou_reward": 0.8604844212532043,
      "rewards/forensics_sinkhorn_reward": 0.46939441561698914,
      "rewards/forensics_strict_edge_reward": 0.15007762610912323,
      "rewards/forensics_tiou_reward": 0.2472762018442154,
      "step": 88
    },
    {
      "completion_length": 16.84375,
      "epoch": 0.18638743455497384,
      "grad_norm": 9.797772407531738,
      "kl": 0.62109375,
      "learning_rate": 9.538784067085953e-07,
      "loss": 0.2718,
      "reward": 0.5724858641624451,
      "reward_std": 0.1805889904499054,
      "reward_std_all": 0.19659440964460373,
      "rewards-max/forensics_sinkhorn_reward": 0.9825617074966431,
      "rewards-max/forensics_strict_edge_reward": 0.9003686904907227,
      "rewards-max/forensics_tiou_reward": 0.9749259352684021,
      "rewards/forensics_sinkhorn_reward": 0.4670276343822479,
      "rewards/forensics_strict_edge_reward": 0.11684128642082214,
      "rewards/forensics_tiou_reward": 0.19894546270370483,
      "step": 89
    },
    {
      "completion_length": 17.828125,
      "epoch": 0.18848167539267016,
      "grad_norm": 8.988774299621582,
      "kl": 0.3603515625,
      "learning_rate": 9.533542976939203e-07,
      "loss": 0.0305,
      "reward": 0.5874480903148651,
      "reward_std": 0.18894121795892715,
      "reward_std_all": 0.19855737686157227,
      "rewards-max/forensics_sinkhorn_reward": 0.9617405235767365,
      "rewards-max/forensics_strict_edge_reward": 0.5402484089136124,
      "rewards-max/forensics_tiou_reward": 0.8967111110687256,
      "rewards/forensics_sinkhorn_reward": 0.43445052206516266,
      "rewards/forensics_strict_edge_reward": 0.08307335898280144,
      "rewards/forensics_tiou_reward": 0.1601545289158821,
      "step": 90
    },
    {
      "completion_length": 19.125,
      "epoch": 0.1905759162303665,
      "grad_norm": 7.266587734222412,
      "kl": 0.48828125,
      "learning_rate": 9.528301886792453e-07,
      "loss": 0.0072,
      "reward": 0.6086283624172211,
      "reward_std": 0.16729876399040222,
      "reward_std_all": 0.18360941112041473,
      "rewards-max/forensics_sinkhorn_reward": 0.9827871918678284,
      "rewards-max/forensics_strict_edge_reward": 0.8771188855171204,
      "rewards-max/forensics_tiou_reward": 0.9757771790027618,
      "rewards/forensics_sinkhorn_reward": 0.49209433794021606,
      "rewards/forensics_strict_edge_reward": 0.15915611386299133,
      "rewards/forensics_tiou_reward": 0.2476056069135666,
      "step": 91
    },
    {
      "completion_length": 18.171875,
      "epoch": 0.19267015706806281,
      "grad_norm": 9.433576583862305,
      "kl": 0.5234375,
      "learning_rate": 9.523060796645702e-07,
      "loss": -0.1156,
      "reward": 0.6423088312149048,
      "reward_std": 0.115151047706604,
      "reward_std_all": 0.1331252008676529,
      "rewards-max/forensics_sinkhorn_reward": 0.8670108020305634,
      "rewards-max/forensics_strict_edge_reward": 0.64028599858284,
      "rewards-max/forensics_tiou_reward": 0.7587502300739288,
      "rewards/forensics_sinkhorn_reward": 0.5000399798154831,
      "rewards/forensics_strict_edge_reward": 0.09847259521484375,
      "rewards/forensics_tiou_reward": 0.1595751903951168,
      "step": 92
    },
    {
      "completion_length": 16.34375,
      "epoch": 0.19476439790575917,
      "grad_norm": 6.797835350036621,
      "kl": 0.3525390625,
      "learning_rate": 9.517819706498952e-07,
      "loss": 0.0095,
      "reward": 0.5866951048374176,
      "reward_std": 0.168817400932312,
      "reward_std_all": 0.19122066348791122,
      "rewards-max/forensics_sinkhorn_reward": 0.9823305606842041,
      "rewards-max/forensics_strict_edge_reward": 0.6875723302364349,
      "rewards-max/forensics_tiou_reward": 0.9740438163280487,
      "rewards/forensics_sinkhorn_reward": 0.45473437011241913,
      "rewards/forensics_strict_edge_reward": 0.12473711743950844,
      "rewards/forensics_tiou_reward": 0.19117574393749237,
      "step": 93
    },
    {
      "completion_length": 17.09375,
      "epoch": 0.1968586387434555,
      "grad_norm": 7.854173183441162,
      "kl": 0.4814453125,
      "learning_rate": 9.5125786163522e-07,
      "loss": 0.1157,
      "reward": 0.6040744483470917,
      "reward_std": 0.12269561737775803,
      "reward_std_all": 0.1611759215593338,
      "rewards-max/forensics_sinkhorn_reward": 0.8777071833610535,
      "rewards-max/forensics_strict_edge_reward": 0.6674256920814514,
      "rewards-max/forensics_tiou_reward": 0.778237909078598,
      "rewards/forensics_sinkhorn_reward": 0.4689861685037613,
      "rewards/forensics_strict_edge_reward": 0.10603944212198257,
      "rewards/forensics_tiou_reward": 0.18897584825754166,
      "step": 94
    },
    {
      "completion_length": 19.0,
      "epoch": 0.19895287958115182,
      "grad_norm": 7.848489284515381,
      "kl": 0.4912109375,
      "learning_rate": 9.50733752620545e-07,
      "loss": 0.0798,
      "reward": 0.5521582663059235,
      "reward_std": 0.20272860676050186,
      "reward_std_all": 0.2172420769929886,
      "rewards-max/forensics_sinkhorn_reward": 0.9687914550304413,
      "rewards-max/forensics_strict_edge_reward": 0.6244729161262512,
      "rewards-max/forensics_tiou_reward": 0.9228803813457489,
      "rewards/forensics_sinkhorn_reward": 0.39557427167892456,
      "rewards/forensics_strict_edge_reward": 0.11910552904009819,
      "rewards/forensics_tiou_reward": 0.18403305113315582,
      "step": 95
    },
    {
      "completion_length": 17.578125,
      "epoch": 0.20104712041884817,
      "grad_norm": 10.435253143310547,
      "kl": 0.546875,
      "learning_rate": 9.502096436058699e-07,
      "loss": 0.0379,
      "reward": 0.5233029574155807,
      "reward_std": 0.20150024443864822,
      "reward_std_all": 0.23211272060871124,
      "rewards-max/forensics_sinkhorn_reward": 0.9177761971950531,
      "rewards-max/forensics_strict_edge_reward": 0.3063247948884964,
      "rewards-max/forensics_tiou_reward": 0.8249912261962891,
      "rewards/forensics_sinkhorn_reward": 0.3843330591917038,
      "rewards/forensics_strict_edge_reward": 0.040310636162757874,
      "rewards/forensics_tiou_reward": 0.11626327782869339,
      "step": 96
    },
    {
      "completion_length": 16.25,
      "epoch": 0.2031413612565445,
      "grad_norm": 10.5584716796875,
      "kl": 0.611328125,
      "learning_rate": 9.496855345911949e-07,
      "loss": 0.206,
      "reward": 0.5791420340538025,
      "reward_std": 0.15349887311458588,
      "reward_std_all": 0.17958635091781616,
      "rewards-max/forensics_sinkhorn_reward": 0.9882758855819702,
      "rewards-max/forensics_strict_edge_reward": 0.9782643616199493,
      "rewards-max/forensics_tiou_reward": 0.997183084487915,
      "rewards/forensics_sinkhorn_reward": 0.4692382663488388,
      "rewards/forensics_strict_edge_reward": 0.15694595873355865,
      "rewards/forensics_tiou_reward": 0.2024979591369629,
      "step": 97
    },
    {
      "completion_length": 20.171875,
      "epoch": 0.20523560209424083,
      "grad_norm": 7.601248741149902,
      "kl": 0.666015625,
      "learning_rate": 9.491614255765199e-07,
      "loss": 0.1233,
      "reward": 0.622769296169281,
      "reward_std": 0.17663567513227463,
      "reward_std_all": 0.18257882446050644,
      "rewards-max/forensics_sinkhorn_reward": 0.9858572483062744,
      "rewards-max/forensics_strict_edge_reward": 0.9375866651535034,
      "rewards-max/forensics_tiou_reward": 0.9877305328845978,
      "rewards/forensics_sinkhorn_reward": 0.4678893983364105,
      "rewards/forensics_strict_edge_reward": 0.1767904981970787,
      "rewards/forensics_tiou_reward": 0.2764357328414917,
      "step": 98
    },
    {
      "completion_length": 17.9375,
      "epoch": 0.20732984293193718,
      "grad_norm": 8.742002487182617,
      "kl": 0.5771484375,
      "learning_rate": 9.486373165618448e-07,
      "loss": 0.0497,
      "reward": 0.6323780417442322,
      "reward_std": 0.15676361322402954,
      "reward_std_all": 0.1622387170791626,
      "rewards-max/forensics_sinkhorn_reward": 0.9812673628330231,
      "rewards-max/forensics_strict_edge_reward": 0.48642148077487946,
      "rewards-max/forensics_tiou_reward": 0.9696118533611298,
      "rewards/forensics_sinkhorn_reward": 0.46672655642032623,
      "rewards/forensics_strict_edge_reward": 0.10409969463944435,
      "rewards/forensics_tiou_reward": 0.19517484307289124,
      "step": 99
    },
    {
      "completion_length": 17.765625,
      "epoch": 0.2094240837696335,
      "grad_norm": 21.96677017211914,
      "kl": 1.2109375,
      "learning_rate": 9.481132075471698e-07,
      "loss": 0.007,
      "reward": 0.5914954245090485,
      "reward_std": 0.16318607330322266,
      "reward_std_all": 0.1784256249666214,
      "rewards-max/forensics_sinkhorn_reward": 0.8060297071933746,
      "rewards-max/forensics_strict_edge_reward": 0.5996330082416534,
      "rewards-max/forensics_tiou_reward": 0.7451362311840057,
      "rewards/forensics_sinkhorn_reward": 0.46682921051979065,
      "rewards/forensics_strict_edge_reward": 0.07933291792869568,
      "rewards/forensics_tiou_reward": 0.15184932202100754,
      "step": 100
    },
    {
      "completion_length": 19.15625,
      "epoch": 0.21151832460732983,
      "grad_norm": 10.77033805847168,
      "kl": 0.9296875,
      "learning_rate": 9.475890985324947e-07,
      "loss": 0.0665,
      "reward": 0.598907470703125,
      "reward_std": 0.16721489280462265,
      "reward_std_all": 0.18174860626459122,
      "rewards-max/forensics_sinkhorn_reward": 0.9243461787700653,
      "rewards-max/forensics_strict_edge_reward": 0.5536971986293793,
      "rewards-max/forensics_tiou_reward": 0.8393867313861847,
      "rewards/forensics_sinkhorn_reward": 0.4420411288738251,
      "rewards/forensics_strict_edge_reward": 0.10424983128905296,
      "rewards/forensics_tiou_reward": 0.1935737431049347,
      "step": 101
    },
    {
      "completion_length": 19.46875,
      "epoch": 0.2136125654450262,
      "grad_norm": 9.94003963470459,
      "kl": 0.705078125,
      "learning_rate": 9.470649895178197e-07,
      "loss": 0.1697,
      "reward": 0.6062203645706177,
      "reward_std": 0.1592223420739174,
      "reward_std_all": 0.17273592948913574,
      "rewards-max/forensics_sinkhorn_reward": 0.9833436012268066,
      "rewards-max/forensics_strict_edge_reward": 0.7988545596599579,
      "rewards-max/forensics_tiou_reward": 0.9777613878250122,
      "rewards/forensics_sinkhorn_reward": 0.43449199199676514,
      "rewards/forensics_strict_edge_reward": 0.13925372064113617,
      "rewards/forensics_tiou_reward": 0.16382312774658203,
      "step": 102
    },
    {
      "completion_length": 17.46875,
      "epoch": 0.2157068062827225,
      "grad_norm": 11.291166305541992,
      "kl": 0.96875,
      "learning_rate": 9.465408805031446e-07,
      "loss": 0.1172,
      "reward": 0.5514489412307739,
      "reward_std": 0.16415461897850037,
      "reward_std_all": 0.1959126740694046,
      "rewards-max/forensics_sinkhorn_reward": 0.9682594239711761,
      "rewards-max/forensics_strict_edge_reward": 0.6961015909910202,
      "rewards-max/forensics_tiou_reward": 0.92144575715065,
      "rewards/forensics_sinkhorn_reward": 0.4160260111093521,
      "rewards/forensics_strict_edge_reward": 0.1126435287296772,
      "rewards/forensics_tiou_reward": 0.15872719138860703,
      "step": 103
    },
    {
      "completion_length": 21.40625,
      "epoch": 0.21780104712041884,
      "grad_norm": 10.607084274291992,
      "kl": 0.595703125,
      "learning_rate": 9.460167714884696e-07,
      "loss": 0.1341,
      "reward": 0.5725248157978058,
      "reward_std": 0.19937710464000702,
      "reward_std_all": 0.21173391491174698,
      "rewards-max/forensics_sinkhorn_reward": 0.9602577984333038,
      "rewards-max/forensics_strict_edge_reward": 0.8404377698898315,
      "rewards-max/forensics_tiou_reward": 0.9347264766693115,
      "rewards/forensics_sinkhorn_reward": 0.44205738604068756,
      "rewards/forensics_strict_edge_reward": 0.1303802728652954,
      "rewards/forensics_tiou_reward": 0.20864629745483398,
      "step": 104
    },
    {
      "completion_length": 21.90625,
      "epoch": 0.2198952879581152,
      "grad_norm": 7.998298645019531,
      "kl": 0.5927734375,
      "learning_rate": 9.454926624737946e-07,
      "loss": 0.1907,
      "reward": 0.5400843322277069,
      "reward_std": 0.17913547158241272,
      "reward_std_all": 0.21405908465385437,
      "rewards-max/forensics_sinkhorn_reward": 0.8390771448612213,
      "rewards-max/forensics_strict_edge_reward": 0.5193275660276413,
      "rewards-max/forensics_tiou_reward": 0.6969046145677567,
      "rewards/forensics_sinkhorn_reward": 0.39506693184375763,
      "rewards/forensics_strict_edge_reward": 0.11330627463757992,
      "rewards/forensics_tiou_reward": 0.20749355480074883,
      "step": 105
    },
    {
      "completion_length": 18.15625,
      "epoch": 0.22198952879581152,
      "grad_norm": 7.341263294219971,
      "kl": 0.4833984375,
      "learning_rate": 9.449685534591194e-07,
      "loss": 0.051,
      "reward": 0.6384311616420746,
      "reward_std": 0.19392330944538116,
      "reward_std_all": 0.18310092389583588,
      "rewards-max/forensics_sinkhorn_reward": 0.963136225938797,
      "rewards-max/forensics_strict_edge_reward": 0.9464850425720215,
      "rewards-max/forensics_tiou_reward": 0.9433788657188416,
      "rewards/forensics_sinkhorn_reward": 0.4920961558818817,
      "rewards/forensics_strict_edge_reward": 0.22838357836008072,
      "rewards/forensics_tiou_reward": 0.28168636560440063,
      "step": 106
    },
    {
      "completion_length": 17.03125,
      "epoch": 0.22408376963350785,
      "grad_norm": 9.669572830200195,
      "kl": 0.86328125,
      "learning_rate": 9.444444444444444e-07,
      "loss": 0.171,
      "reward": 0.5655381381511688,
      "reward_std": 0.19908864796161652,
      "reward_std_all": 0.21156346052885056,
      "rewards-max/forensics_sinkhorn_reward": 0.9691793620586395,
      "rewards-max/forensics_strict_edge_reward": 0.7848804593086243,
      "rewards-max/forensics_tiou_reward": 0.924236536026001,
      "rewards/forensics_sinkhorn_reward": 0.455718532204628,
      "rewards/forensics_strict_edge_reward": 0.13469402492046356,
      "rewards/forensics_tiou_reward": 0.18534605205059052,
      "step": 107
    },
    {
      "completion_length": 17.59375,
      "epoch": 0.2261780104712042,
      "grad_norm": 10.556835174560547,
      "kl": 0.4931640625,
      "learning_rate": 9.439203354297693e-07,
      "loss": 0.1629,
      "reward": 0.5990699231624603,
      "reward_std": 0.18478700518608093,
      "reward_std_all": 0.18699711561203003,
      "rewards-max/forensics_sinkhorn_reward": 0.9626775085926056,
      "rewards-max/forensics_strict_edge_reward": 0.8134243786334991,
      "rewards-max/forensics_tiou_reward": 0.9383604526519775,
      "rewards/forensics_sinkhorn_reward": 0.44310033321380615,
      "rewards/forensics_strict_edge_reward": 0.13944881409406662,
      "rewards/forensics_tiou_reward": 0.1945018246769905,
      "step": 108
    },
    {
      "completion_length": 16.875,
      "epoch": 0.22827225130890053,
      "grad_norm": 8.318401336669922,
      "kl": 0.6484375,
      "learning_rate": 9.433962264150943e-07,
      "loss": 0.0726,
      "reward": 0.6368529200553894,
      "reward_std": 0.1262953318655491,
      "reward_std_all": 0.14284653216600418,
      "rewards-max/forensics_sinkhorn_reward": 0.9819299280643463,
      "rewards-max/forensics_strict_edge_reward": 0.8968285918235779,
      "rewards-max/forensics_tiou_reward": 0.9725334644317627,
      "rewards/forensics_sinkhorn_reward": 0.5201345384120941,
      "rewards/forensics_strict_edge_reward": 0.10490309447050095,
      "rewards/forensics_tiou_reward": 0.223532572388649,
      "step": 109
    },
    {
      "completion_length": 19.59375,
      "epoch": 0.23036649214659685,
      "grad_norm": 8.169112205505371,
      "kl": 0.5859375,
      "learning_rate": 9.428721174004192e-07,
      "loss": 0.2361,
      "reward": 0.5591516494750977,
      "reward_std": 0.18277977406978607,
      "reward_std_all": 0.20795249193906784,
      "rewards-max/forensics_sinkhorn_reward": 0.886284202337265,
      "rewards-max/forensics_strict_edge_reward": 0.6571157872676849,
      "rewards-max/forensics_tiou_reward": 0.7709773778915405,
      "rewards/forensics_sinkhorn_reward": 0.4180222898721695,
      "rewards/forensics_strict_edge_reward": 0.09297327324748039,
      "rewards/forensics_tiou_reward": 0.17377594113349915,
      "step": 110
    },
    {
      "completion_length": 19.640625,
      "epoch": 0.2324607329842932,
      "grad_norm": 7.749205112457275,
      "kl": 0.5283203125,
      "learning_rate": 9.423480083857442e-07,
      "loss": -0.0306,
      "reward": 0.5471068024635315,
      "reward_std": 0.12411847338080406,
      "reward_std_all": 0.17807450145483017,
      "rewards-max/forensics_sinkhorn_reward": 0.8489186763763428,
      "rewards-max/forensics_strict_edge_reward": 0.6057014465332031,
      "rewards-max/forensics_tiou_reward": 0.6749871373176575,
      "rewards/forensics_sinkhorn_reward": 0.39466579258441925,
      "rewards/forensics_strict_edge_reward": 0.07999623566865921,
      "rewards/forensics_tiou_reward": 0.12360842525959015,
      "step": 111
    },
    {
      "completion_length": 17.5,
      "epoch": 0.23455497382198953,
      "grad_norm": 8.393349647521973,
      "kl": 0.64453125,
      "learning_rate": 9.41823899371069e-07,
      "loss": 0.2129,
      "reward": 0.6363856494426727,
      "reward_std": 0.1314365193247795,
      "reward_std_all": 0.14485228061676025,
      "rewards-max/forensics_sinkhorn_reward": 0.9825518131256104,
      "rewards-max/forensics_strict_edge_reward": 0.8205508291721344,
      "rewards-max/forensics_tiou_reward": 0.9749367237091064,
      "rewards/forensics_sinkhorn_reward": 0.49867957830429077,
      "rewards/forensics_strict_edge_reward": 0.11325211822986603,
      "rewards/forensics_tiou_reward": 0.2057536467909813,
      "step": 112
    },
    {
      "completion_length": 17.84375,
      "epoch": 0.23664921465968586,
      "grad_norm": 11.06563949584961,
      "kl": 0.7861328125,
      "learning_rate": 9.412997903563941e-07,
      "loss": 0.1247,
      "reward": 0.6311393976211548,
      "reward_std": 0.1444292888045311,
      "reward_std_all": 0.1554001048207283,
      "rewards-max/forensics_sinkhorn_reward": 0.9815934002399445,
      "rewards-max/forensics_strict_edge_reward": 0.8260191679000854,
      "rewards-max/forensics_tiou_reward": 0.9711412787437439,
      "rewards/forensics_sinkhorn_reward": 0.48166270554065704,
      "rewards/forensics_strict_edge_reward": 0.10069680958986282,
      "rewards/forensics_tiou_reward": 0.2007441446185112,
      "step": 113
    },
    {
      "completion_length": 17.1875,
      "epoch": 0.2387434554973822,
      "grad_norm": 11.310454368591309,
      "kl": 1.3125,
      "learning_rate": 9.407756813417191e-07,
      "loss": 0.0901,
      "reward": 0.6734817028045654,
      "reward_std": 0.10535294190049171,
      "reward_std_all": 0.11686158925294876,
      "rewards-max/forensics_sinkhorn_reward": 0.9865598380565643,
      "rewards-max/forensics_strict_edge_reward": 0.9265981018543243,
      "rewards-max/forensics_tiou_reward": 0.9904498159885406,
      "rewards/forensics_sinkhorn_reward": 0.5926455855369568,
      "rewards/forensics_strict_edge_reward": 0.1781090721487999,
      "rewards/forensics_tiou_reward": 0.2998988702893257,
      "step": 114
    },
    {
      "completion_length": 20.1875,
      "epoch": 0.24083769633507854,
      "grad_norm": 9.507765769958496,
      "kl": 0.720703125,
      "learning_rate": 9.40251572327044e-07,
      "loss": 0.0648,
      "reward": 0.6183537542819977,
      "reward_std": 0.17614112794399261,
      "reward_std_all": 0.17612724006175995,
      "rewards-max/forensics_sinkhorn_reward": 0.9275829792022705,
      "rewards-max/forensics_strict_edge_reward": 0.7653808295726776,
      "rewards-max/forensics_tiou_reward": 0.867822527885437,
      "rewards/forensics_sinkhorn_reward": 0.47219112515449524,
      "rewards/forensics_strict_edge_reward": 0.14331358671188354,
      "rewards/forensics_tiou_reward": 0.2451159507036209,
      "step": 115
    },
    {
      "completion_length": 18.640625,
      "epoch": 0.24293193717277486,
      "grad_norm": 8.020713806152344,
      "kl": 0.7734375,
      "learning_rate": 9.39727463312369e-07,
      "loss": 0.1295,
      "reward": 0.6028166711330414,
      "reward_std": 0.17105261981487274,
      "reward_std_all": 0.17817934602499008,
      "rewards-max/forensics_sinkhorn_reward": 0.8232672810554504,
      "rewards-max/forensics_strict_edge_reward": 0.6020724773406982,
      "rewards-max/forensics_tiou_reward": 0.7185397446155548,
      "rewards/forensics_sinkhorn_reward": 0.44117189943790436,
      "rewards/forensics_strict_edge_reward": 0.10752547904849052,
      "rewards/forensics_tiou_reward": 0.1432231292128563,
      "step": 116
    },
    {
      "completion_length": 20.109375,
      "epoch": 0.2450261780104712,
      "grad_norm": 8.457261085510254,
      "kl": 0.697265625,
      "learning_rate": 9.392033542976939e-07,
      "loss": 0.048,
      "reward": 0.5885985195636749,
      "reward_std": 0.13945657014846802,
      "reward_std_all": 0.17218845337629318,
      "rewards-max/forensics_sinkhorn_reward": 0.9879108965396881,
      "rewards-max/forensics_strict_edge_reward": 0.9782643616199493,
      "rewards-max/forensics_tiou_reward": 0.9957446753978729,
      "rewards/forensics_sinkhorn_reward": 0.462345689535141,
      "rewards/forensics_strict_edge_reward": 0.16483832895755768,
      "rewards/forensics_tiou_reward": 0.22039355337619781,
      "step": 117
    },
    {
      "completion_length": 21.671875,
      "epoch": 0.24712041884816754,
      "grad_norm": 12.781006813049316,
      "kl": 1.333984375,
      "learning_rate": 9.386792452830188e-07,
      "loss": 0.0386,
      "reward": 0.5966211557388306,
      "reward_std": 0.18070293962955475,
      "reward_std_all": 0.1893024444580078,
      "rewards-max/forensics_sinkhorn_reward": 0.9234474897384644,
      "rewards-max/forensics_strict_edge_reward": 0.5971094369888306,
      "rewards-max/forensics_tiou_reward": 0.8493999242782593,
      "rewards/forensics_sinkhorn_reward": 0.41546204686164856,
      "rewards/forensics_strict_edge_reward": 0.10323773697018623,
      "rewards/forensics_tiou_reward": 0.1852574571967125,
      "step": 118
    },
    {
      "completion_length": 17.734375,
      "epoch": 0.24921465968586387,
      "grad_norm": 10.151199340820312,
      "kl": 0.798828125,
      "learning_rate": 9.381551362683437e-07,
      "loss": 0.1482,
      "reward": 0.6072115004062653,
      "reward_std": 0.13867917656898499,
      "reward_std_all": 0.16047328338027,
      "rewards-max/forensics_sinkhorn_reward": 0.9862259328365326,
      "rewards-max/forensics_strict_edge_reward": 0.8968285918235779,
      "rewards-max/forensics_tiou_reward": 0.9891106486320496,
      "rewards/forensics_sinkhorn_reward": 0.5252495408058167,
      "rewards/forensics_strict_edge_reward": 0.2043360248208046,
      "rewards/forensics_tiou_reward": 0.3301563486456871,
      "step": 119
    },
    {
      "completion_length": 18.671875,
      "epoch": 0.2513089005235602,
      "grad_norm": 12.071532249450684,
      "kl": 1.2060546875,
      "learning_rate": 9.376310272536687e-07,
      "loss": 0.1669,
      "reward": 0.5819963812828064,
      "reward_std": 0.1303154155611992,
      "reward_std_all": 0.17427250742912292,
      "rewards-max/forensics_sinkhorn_reward": 0.9453952312469482,
      "rewards-max/forensics_strict_edge_reward": 0.878633052110672,
      "rewards-max/forensics_tiou_reward": 0.9038648307323456,
      "rewards/forensics_sinkhorn_reward": 0.46493685245513916,
      "rewards/forensics_strict_edge_reward": 0.1508997231721878,
      "rewards/forensics_tiou_reward": 0.24025562405586243,
      "step": 120
    },
    {
      "completion_length": 19.328125,
      "epoch": 0.2534031413612565,
      "grad_norm": 9.868488311767578,
      "kl": 0.52734375,
      "learning_rate": 9.371069182389937e-07,
      "loss": 0.0982,
      "reward": 0.5779926776885986,
      "reward_std": 0.16999609023332596,
      "reward_std_all": 0.1910666897892952,
      "rewards-max/forensics_sinkhorn_reward": 0.983395129442215,
      "rewards-max/forensics_strict_edge_reward": 0.7879786789417267,
      "rewards-max/forensics_tiou_reward": 0.977990061044693,
      "rewards/forensics_sinkhorn_reward": 0.44085705280303955,
      "rewards/forensics_strict_edge_reward": 0.11949464678764343,
      "rewards/forensics_tiou_reward": 0.17002076655626297,
      "step": 121
    },
    {
      "completion_length": 20.296875,
      "epoch": 0.2554973821989529,
      "grad_norm": 8.719337463378906,
      "kl": 0.5517578125,
      "learning_rate": 9.365828092243186e-07,
      "loss": 0.0723,
      "reward": 0.6242660582065582,
      "reward_std": 0.18119918555021286,
      "reward_std_all": 0.17710625380277634,
      "rewards-max/forensics_sinkhorn_reward": 0.9608529806137085,
      "rewards-max/forensics_strict_edge_reward": 0.8379553556442261,
      "rewards-max/forensics_tiou_reward": 0.8949689269065857,
      "rewards/forensics_sinkhorn_reward": 0.4793543517589569,
      "rewards/forensics_strict_edge_reward": 0.14149563014507294,
      "rewards/forensics_tiou_reward": 0.23524468392133713,
      "step": 122
    },
    {
      "completion_length": 18.71875,
      "epoch": 0.25759162303664923,
      "grad_norm": 9.951197624206543,
      "kl": 1.013671875,
      "learning_rate": 9.360587002096436e-07,
      "loss": 0.122,
      "reward": 0.597439169883728,
      "reward_std": 0.14349236339330673,
      "reward_std_all": 0.17529775202274323,
      "rewards-max/forensics_sinkhorn_reward": 0.9814422726631165,
      "rewards-max/forensics_strict_edge_reward": 0.8593232333660126,
      "rewards-max/forensics_tiou_reward": 0.9706131517887115,
      "rewards/forensics_sinkhorn_reward": 0.45633018016815186,
      "rewards/forensics_strict_edge_reward": 0.1357072852551937,
      "rewards/forensics_tiou_reward": 0.20609696209430695,
      "step": 123
    },
    {
      "completion_length": 19.453125,
      "epoch": 0.25968586387434556,
      "grad_norm": 11.074886322021484,
      "kl": 0.822265625,
      "learning_rate": 9.355345911949684e-07,
      "loss": 0.1125,
      "reward": 0.5882285237312317,
      "reward_std": 0.18959003686904907,
      "reward_std_all": 0.20078332722187042,
      "rewards-max/forensics_sinkhorn_reward": 0.983075350522995,
      "rewards-max/forensics_strict_edge_reward": 0.8445805907249451,
      "rewards-max/forensics_tiou_reward": 0.9768045246601105,
      "rewards/forensics_sinkhorn_reward": 0.4462936520576477,
      "rewards/forensics_strict_edge_reward": 0.14020083844661713,
      "rewards/forensics_tiou_reward": 0.2373962700366974,
      "step": 124
    },
    {
      "completion_length": 16.828125,
      "epoch": 0.2617801047120419,
      "grad_norm": 8.386302947998047,
      "kl": 0.595703125,
      "learning_rate": 9.350104821802935e-07,
      "loss": 0.1698,
      "reward": 0.6844430565834045,
      "reward_std": 0.11002738773822784,
      "reward_std_all": 0.11344361305236816,
      "rewards-max/forensics_sinkhorn_reward": 0.9861297905445099,
      "rewards-max/forensics_strict_edge_reward": 0.9048932790756226,
      "rewards-max/forensics_tiou_reward": 0.9887248873710632,
      "rewards/forensics_sinkhorn_reward": 0.5603888630867004,
      "rewards/forensics_strict_edge_reward": 0.23826362937688828,
      "rewards/forensics_tiou_reward": 0.32959499955177307,
      "step": 125
    },
    {
      "completion_length": 19.40625,
      "epoch": 0.2638743455497382,
      "grad_norm": 8.632094383239746,
      "kl": 0.521484375,
      "learning_rate": 9.344863731656184e-07,
      "loss": 0.3447,
      "reward": 0.57745361328125,
      "reward_std": 0.13030166923999786,
      "reward_std_all": 0.17250576615333557,
      "rewards-max/forensics_sinkhorn_reward": 0.9846754670143127,
      "rewards-max/forensics_strict_edge_reward": 0.8668389916419983,
      "rewards-max/forensics_tiou_reward": 0.9830564558506012,
      "rewards/forensics_sinkhorn_reward": 0.4566914886236191,
      "rewards/forensics_strict_edge_reward": 0.1268077939748764,
      "rewards/forensics_tiou_reward": 0.2044137492775917,
      "step": 126
    },
    {
      "completion_length": 20.953125,
      "epoch": 0.26596858638743454,
      "grad_norm": 10.380409240722656,
      "kl": 0.779296875,
      "learning_rate": 9.339622641509434e-07,
      "loss": 0.2031,
      "reward": 0.5762774646282196,
      "reward_std": 0.18682565540075302,
      "reward_std_all": 0.198366180062294,
      "rewards-max/forensics_sinkhorn_reward": 0.9581242501735687,
      "rewards-max/forensics_strict_edge_reward": 0.7407414019107819,
      "rewards-max/forensics_tiou_reward": 0.933429092168808,
      "rewards/forensics_sinkhorn_reward": 0.45000937581062317,
      "rewards/forensics_strict_edge_reward": 0.16098745539784431,
      "rewards/forensics_tiou_reward": 0.2748386710882187,
      "step": 127
    },
    {
      "completion_length": 20.78125,
      "epoch": 0.2680628272251309,
      "grad_norm": 9.372124671936035,
      "kl": 0.740234375,
      "learning_rate": 9.334381551362684e-07,
      "loss": 0.1157,
      "reward": 0.5775916576385498,
      "reward_std": 0.16443521529436111,
      "reward_std_all": 0.18556180596351624,
      "rewards-max/forensics_sinkhorn_reward": 0.9181902706623077,
      "rewards-max/forensics_strict_edge_reward": 0.6309370547533035,
      "rewards-max/forensics_tiou_reward": 0.8327869474887848,
      "rewards/forensics_sinkhorn_reward": 0.4006902873516083,
      "rewards/forensics_strict_edge_reward": 0.12905646860599518,
      "rewards/forensics_tiou_reward": 0.17039050161838531,
      "step": 128
    },
    {
      "completion_length": 19.359375,
      "epoch": 0.27015706806282724,
      "grad_norm": 8.797781944274902,
      "kl": 0.830078125,
      "learning_rate": 9.329140461215933e-07,
      "loss": 0.1037,
      "reward": 0.6164916157722473,
      "reward_std": 0.15389414131641388,
      "reward_std_all": 0.16389994323253632,
      "rewards-max/forensics_sinkhorn_reward": 0.9795026481151581,
      "rewards-max/forensics_strict_edge_reward": 0.6869064271450043,
      "rewards-max/forensics_tiou_reward": 0.9626657068729401,
      "rewards/forensics_sinkhorn_reward": 0.42970091104507446,
      "rewards/forensics_strict_edge_reward": 0.11785426735877991,
      "rewards/forensics_tiou_reward": 0.18140383064746857,
      "step": 129
    },
    {
      "completion_length": 16.96875,
      "epoch": 0.27225130890052357,
      "grad_norm": 9.169290542602539,
      "kl": 0.5068359375,
      "learning_rate": 9.323899371069182e-07,
      "loss": -0.0139,
      "reward": 0.6092946231365204,
      "reward_std": 0.1397123634815216,
      "reward_std_all": 0.1616656258702278,
      "rewards-max/forensics_sinkhorn_reward": 0.9873154163360596,
      "rewards-max/forensics_strict_edge_reward": 0.9567649364471436,
      "rewards-max/forensics_tiou_reward": 0.9934000074863434,
      "rewards/forensics_sinkhorn_reward": 0.4697088897228241,
      "rewards/forensics_strict_edge_reward": 0.1336728185415268,
      "rewards/forensics_tiou_reward": 0.18840715289115906,
      "step": 130
    },
    {
      "completion_length": 20.609375,
      "epoch": 0.2743455497382199,
      "grad_norm": 7.5455145835876465,
      "kl": 0.78125,
      "learning_rate": 9.318658280922431e-07,
      "loss": 0.1765,
      "reward": 0.5240360200405121,
      "reward_std": 0.1902187243103981,
      "reward_std_all": 0.22483646869659424,
      "rewards-max/forensics_sinkhorn_reward": 0.9699188470840454,
      "rewards-max/forensics_strict_edge_reward": 0.5230360180139542,
      "rewards-max/forensics_tiou_reward": 0.9269838631153107,
      "rewards/forensics_sinkhorn_reward": 0.35288964211940765,
      "rewards/forensics_strict_edge_reward": 0.07967957109212875,
      "rewards/forensics_tiou_reward": 0.14223427325487137,
      "step": 131
    },
    {
      "completion_length": 18.078125,
      "epoch": 0.2764397905759162,
      "grad_norm": 9.49321460723877,
      "kl": 0.662109375,
      "learning_rate": 9.313417190775681e-07,
      "loss": 0.1517,
      "reward": 0.5963753759860992,
      "reward_std": 0.14386894553899765,
      "reward_std_all": 0.17191891372203827,
      "rewards-max/forensics_sinkhorn_reward": 0.9329905211925507,
      "rewards-max/forensics_strict_edge_reward": 0.7143478095531464,
      "rewards-max/forensics_tiou_reward": 0.870754212141037,
      "rewards/forensics_sinkhorn_reward": 0.45444025099277496,
      "rewards/forensics_strict_edge_reward": 0.14565160125494003,
      "rewards/forensics_tiou_reward": 0.19018231332302094,
      "step": 132
    },
    {
      "completion_length": 21.5625,
      "epoch": 0.27853403141361255,
      "grad_norm": 9.53396987915039,
      "kl": 0.716796875,
      "learning_rate": 9.30817610062893e-07,
      "loss": 0.0698,
      "reward": 0.5795806050300598,
      "reward_std": 0.16961625218391418,
      "reward_std_all": 0.18640349805355072,
      "rewards-max/forensics_sinkhorn_reward": 0.9124003350734711,
      "rewards-max/forensics_strict_edge_reward": 0.6515887677669525,
      "rewards-max/forensics_tiou_reward": 0.814821720123291,
      "rewards/forensics_sinkhorn_reward": 0.4099065065383911,
      "rewards/forensics_strict_edge_reward": 0.1424887552857399,
      "rewards/forensics_tiou_reward": 0.18315713852643967,
      "step": 133
    },
    {
      "completion_length": 18.703125,
      "epoch": 0.2806282722513089,
      "grad_norm": 10.171713829040527,
      "kl": 0.806640625,
      "learning_rate": 9.30293501048218e-07,
      "loss": 0.1822,
      "reward": 0.5398955345153809,
      "reward_std": 0.21921510994434357,
      "reward_std_all": 0.2309124693274498,
      "rewards-max/forensics_sinkhorn_reward": 0.9268893897533417,
      "rewards-max/forensics_strict_edge_reward": 0.8015283942222595,
      "rewards-max/forensics_tiou_reward": 0.8636769354343414,
      "rewards/forensics_sinkhorn_reward": 0.39424222707748413,
      "rewards/forensics_strict_edge_reward": 0.13100221753120422,
      "rewards/forensics_tiou_reward": 0.18188609927892685,
      "step": 134
    },
    {
      "completion_length": 19.421875,
      "epoch": 0.28272251308900526,
      "grad_norm": 8.645074844360352,
      "kl": 0.818359375,
      "learning_rate": 9.297693920335429e-07,
      "loss": 0.0585,
      "reward": 0.5451949238777161,
      "reward_std": 0.142728541046381,
      "reward_std_all": 0.18818094581365585,
      "rewards-max/forensics_sinkhorn_reward": 0.9842140972614288,
      "rewards-max/forensics_strict_edge_reward": 0.855939507484436,
      "rewards-max/forensics_tiou_reward": 0.9811775386333466,
      "rewards/forensics_sinkhorn_reward": 0.4218362271785736,
      "rewards/forensics_strict_edge_reward": 0.14034802466630936,
      "rewards/forensics_tiou_reward": 0.1923118308186531,
      "step": 135
    },
    {
      "completion_length": 20.609375,
      "epoch": 0.2848167539267016,
      "grad_norm": 12.46753215789795,
      "kl": 1.30078125,
      "learning_rate": 9.292452830188678e-07,
      "loss": 0.2151,
      "reward": 0.6306451559066772,
      "reward_std": 0.13448255881667137,
      "reward_std_all": 0.1488155871629715,
      "rewards-max/forensics_sinkhorn_reward": 0.9526635408401489,
      "rewards-max/forensics_strict_edge_reward": 0.7676678597927094,
      "rewards-max/forensics_tiou_reward": 0.916796088218689,
      "rewards/forensics_sinkhorn_reward": 0.4826177954673767,
      "rewards/forensics_strict_edge_reward": 0.12532298639416695,
      "rewards/forensics_tiou_reward": 0.2501226216554642,
      "step": 136
    },
    {
      "completion_length": 19.578125,
      "epoch": 0.2869109947643979,
      "grad_norm": 8.368438720703125,
      "kl": 0.5576171875,
      "learning_rate": 9.287211740041929e-07,
      "loss": 0.195,
      "reward": 0.6047739386558533,
      "reward_std": 0.1607537865638733,
      "reward_std_all": 0.17994622886180878,
      "rewards-max/forensics_sinkhorn_reward": 0.9064387381076813,
      "rewards-max/forensics_strict_edge_reward": 0.5985674411058426,
      "rewards-max/forensics_tiou_reward": 0.8203555047512054,
      "rewards/forensics_sinkhorn_reward": 0.47382666170597076,
      "rewards/forensics_strict_edge_reward": 0.1268300674855709,
      "rewards/forensics_tiou_reward": 0.21395103633403778,
      "step": 137
    },
    {
      "completion_length": 22.125,
      "epoch": 0.28900523560209423,
      "grad_norm": 14.733905792236328,
      "kl": 1.96484375,
      "learning_rate": 9.281970649895178e-07,
      "loss": 0.2311,
      "reward": 0.5969067215919495,
      "reward_std": 0.15016666054725647,
      "reward_std_all": 0.16523876786231995,
      "rewards-max/forensics_sinkhorn_reward": 0.9816936254501343,
      "rewards-max/forensics_strict_edge_reward": 0.8404377698898315,
      "rewards-max/forensics_tiou_reward": 0.9712550640106201,
      "rewards/forensics_sinkhorn_reward": 0.4447943866252899,
      "rewards/forensics_strict_edge_reward": 0.1399136707186699,
      "rewards/forensics_tiou_reward": 0.1927172914147377,
      "step": 138
    },
    {
      "completion_length": 19.890625,
      "epoch": 0.29109947643979056,
      "grad_norm": 8.65511417388916,
      "kl": 1.05078125,
      "learning_rate": 9.276729559748428e-07,
      "loss": 0.1572,
      "reward": 0.5658280253410339,
      "reward_std": 0.20067144930362701,
      "reward_std_all": 0.2133815735578537,
      "rewards-max/forensics_sinkhorn_reward": 0.9854954779148102,
      "rewards-max/forensics_strict_edge_reward": 0.8445805907249451,
      "rewards-max/forensics_tiou_reward": 0.9862264394760132,
      "rewards/forensics_sinkhorn_reward": 0.4358476400375366,
      "rewards/forensics_strict_edge_reward": 0.12391075491905212,
      "rewards/forensics_tiou_reward": 0.20429015159606934,
      "step": 139
    },
    {
      "completion_length": 19.828125,
      "epoch": 0.2931937172774869,
      "grad_norm": 11.01053524017334,
      "kl": 0.7421875,
      "learning_rate": 9.271488469601677e-07,
      "loss": 0.1743,
      "reward": 0.6317234337329865,
      "reward_std": 0.112131517380476,
      "reward_std_all": 0.13626668602228165,
      "rewards-max/forensics_sinkhorn_reward": 0.8165562748908997,
      "rewards-max/forensics_strict_edge_reward": 0.6252883523702621,
      "rewards-max/forensics_tiou_reward": 0.6410407721996307,
      "rewards/forensics_sinkhorn_reward": 0.46636030077934265,
      "rewards/forensics_strict_edge_reward": 0.13824357651174068,
      "rewards/forensics_tiou_reward": 0.18205781280994415,
      "step": 140
    },
    {
      "completion_length": 21.84375,
      "epoch": 0.29528795811518327,
      "grad_norm": 10.752487182617188,
      "kl": 0.958984375,
      "learning_rate": 9.266247379454927e-07,
      "loss": 0.172,
      "reward": 0.5210913866758347,
      "reward_std": 0.1675921231508255,
      "reward_std_all": 0.21749884635210037,
      "rewards-max/forensics_sinkhorn_reward": 0.8850520849227905,
      "rewards-max/forensics_strict_edge_reward": 0.6899373829364777,
      "rewards-max/forensics_tiou_reward": 0.7845514118671417,
      "rewards/forensics_sinkhorn_reward": 0.37836953997612,
      "rewards/forensics_strict_edge_reward": 0.09939144179224968,
      "rewards/forensics_tiou_reward": 0.1714249551296234,
      "step": 141
    },
    {
      "completion_length": 24.40625,
      "epoch": 0.2973821989528796,
      "grad_norm": 12.436666488647461,
      "kl": 1.38671875,
      "learning_rate": 9.261006289308175e-07,
      "loss": 0.1308,
      "reward": 0.6013078093528748,
      "reward_std": 0.23215951025485992,
      "reward_std_all": 0.2183663323521614,
      "rewards-max/forensics_sinkhorn_reward": 0.9443322718143463,
      "rewards-max/forensics_strict_edge_reward": 0.5930130034685135,
      "rewards-max/forensics_tiou_reward": 0.8907090425491333,
      "rewards/forensics_sinkhorn_reward": 0.41255883872509,
      "rewards/forensics_strict_edge_reward": 0.12442930787801743,
      "rewards/forensics_tiou_reward": 0.2053350880742073,
      "step": 142
    },
    {
      "completion_length": 22.578125,
      "epoch": 0.2994764397905759,
      "grad_norm": 10.673911094665527,
      "kl": 0.740234375,
      "learning_rate": 9.255765199161425e-07,
      "loss": 0.181,
      "reward": 0.6549586653709412,
      "reward_std": 0.15125126391649246,
      "reward_std_all": 0.14903942495584488,
      "rewards-max/forensics_sinkhorn_reward": 0.9876273274421692,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9946392476558685,
      "rewards/forensics_sinkhorn_reward": 0.5162429362535477,
      "rewards/forensics_strict_edge_reward": 0.22512956708669662,
      "rewards/forensics_tiou_reward": 0.31242579221725464,
      "step": 143
    },
    {
      "completion_length": 18.5,
      "epoch": 0.30157068062827225,
      "grad_norm": 9.611239433288574,
      "kl": 0.501953125,
      "learning_rate": 9.250524109014674e-07,
      "loss": 0.132,
      "reward": 0.6289951801300049,
      "reward_std": 0.11615698784589767,
      "reward_std_all": 0.1447170451283455,
      "rewards-max/forensics_sinkhorn_reward": 0.9861845076084137,
      "rewards-max/forensics_strict_edge_reward": 0.9048932790756226,
      "rewards-max/forensics_tiou_reward": 0.9889571368694305,
      "rewards/forensics_sinkhorn_reward": 0.5426353514194489,
      "rewards/forensics_strict_edge_reward": 0.20505667477846146,
      "rewards/forensics_tiou_reward": 0.30316874384880066,
      "step": 144
    },
    {
      "completion_length": 17.0625,
      "epoch": 0.3036649214659686,
      "grad_norm": 10.345263481140137,
      "kl": 0.404296875,
      "learning_rate": 9.245283018867924e-07,
      "loss": 0.0313,
      "reward": 0.6544182598590851,
      "reward_std": 0.10644769668579102,
      "reward_std_all": 0.13032565265893936,
      "rewards-max/forensics_sinkhorn_reward": 0.9772094190120697,
      "rewards-max/forensics_strict_edge_reward": 0.7578859627246857,
      "rewards-max/forensics_tiou_reward": 0.9541839361190796,
      "rewards/forensics_sinkhorn_reward": 0.5458329916000366,
      "rewards/forensics_strict_edge_reward": 0.14561517536640167,
      "rewards/forensics_tiou_reward": 0.22770555317401886,
      "step": 145
    },
    {
      "completion_length": 18.09375,
      "epoch": 0.3057591623036649,
      "grad_norm": 9.23371410369873,
      "kl": 0.7373046875,
      "learning_rate": 9.240041928721174e-07,
      "loss": 0.1354,
      "reward": 0.6130673587322235,
      "reward_std": 0.1204785481095314,
      "reward_std_all": 0.15273965150117874,
      "rewards-max/forensics_sinkhorn_reward": 0.9868384301662445,
      "rewards-max/forensics_strict_edge_reward": 0.946017861366272,
      "rewards-max/forensics_tiou_reward": 0.9915247559547424,
      "rewards/forensics_sinkhorn_reward": 0.4990871250629425,
      "rewards/forensics_strict_edge_reward": 0.17503180354833603,
      "rewards/forensics_tiou_reward": 0.25827381014823914,
      "step": 146
    },
    {
      "completion_length": 17.453125,
      "epoch": 0.3078534031413613,
      "grad_norm": 8.505102157592773,
      "kl": 0.703125,
      "learning_rate": 9.234800838574423e-07,
      "loss": 0.0986,
      "reward": 0.6748842895030975,
      "reward_std": 0.12679128721356392,
      "reward_std_all": 0.13232822343707085,
      "rewards-max/forensics_sinkhorn_reward": 0.9880559742450714,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.996317058801651,
      "rewards/forensics_sinkhorn_reward": 0.583326131105423,
      "rewards/forensics_strict_edge_reward": 0.23874127864837646,
      "rewards/forensics_tiou_reward": 0.31419962644577026,
      "step": 147
    },
    {
      "completion_length": 19.53125,
      "epoch": 0.3099476439790576,
      "grad_norm": 7.781407356262207,
      "kl": 0.67822265625,
      "learning_rate": 9.229559748427672e-07,
      "loss": 0.2074,
      "reward": 0.6133051514625549,
      "reward_std": 0.13305331021547318,
      "reward_std_all": 0.16209324449300766,
      "rewards-max/forensics_sinkhorn_reward": 0.9867839515209198,
      "rewards-max/forensics_strict_edge_reward": 0.882964164018631,
      "rewards-max/forensics_tiou_reward": 0.991304337978363,
      "rewards/forensics_sinkhorn_reward": 0.5130292773246765,
      "rewards/forensics_strict_edge_reward": 0.18324856460094452,
      "rewards/forensics_tiou_reward": 0.22678183764219284,
      "step": 148
    },
    {
      "completion_length": 18.03125,
      "epoch": 0.31204188481675393,
      "grad_norm": 9.482017517089844,
      "kl": 0.658203125,
      "learning_rate": 9.224318658280921e-07,
      "loss": 0.0414,
      "reward": 0.6111800670623779,
      "reward_std": 0.13788585364818573,
      "reward_std_all": 0.15753532201051712,
      "rewards-max/forensics_sinkhorn_reward": 0.9616007208824158,
      "rewards-max/forensics_strict_edge_reward": 0.8205901980400085,
      "rewards-max/forensics_tiou_reward": 0.9361807107925415,
      "rewards/forensics_sinkhorn_reward": 0.49708113074302673,
      "rewards/forensics_strict_edge_reward": 0.12726814672350883,
      "rewards/forensics_tiou_reward": 0.22105772793293,
      "step": 149
    },
    {
      "completion_length": 17.015625,
      "epoch": 0.31413612565445026,
      "grad_norm": 16.865877151489258,
      "kl": 1.1484375,
      "learning_rate": 9.219077568134172e-07,
      "loss": 0.135,
      "reward": 0.6607973575592041,
      "reward_std": 0.11869984120130539,
      "reward_std_all": 0.13272378966212273,
      "rewards-max/forensics_sinkhorn_reward": 0.9855398535728455,
      "rewards-max/forensics_strict_edge_reward": 0.9151731729507446,
      "rewards-max/forensics_tiou_reward": 0.9864107072353363,
      "rewards/forensics_sinkhorn_reward": 0.555823802947998,
      "rewards/forensics_strict_edge_reward": 0.20071425288915634,
      "rewards/forensics_tiou_reward": 0.2905631884932518,
      "step": 150
    },
    {
      "completion_length": 16.4375,
      "epoch": 0.3162303664921466,
      "grad_norm": 11.775615692138672,
      "kl": 0.462890625,
      "learning_rate": 9.213836477987421e-07,
      "loss": -0.1214,
      "reward": 0.6289337277412415,
      "reward_std": 0.07950477674603462,
      "reward_std_all": 0.1259106546640396,
      "rewards-max/forensics_sinkhorn_reward": 0.8413887619972229,
      "rewards-max/forensics_strict_edge_reward": 0.6461038440465927,
      "rewards-max/forensics_tiou_reward": 0.7097356021404266,
      "rewards/forensics_sinkhorn_reward": 0.48986147344112396,
      "rewards/forensics_strict_edge_reward": 0.06896733120083809,
      "rewards/forensics_tiou_reward": 0.1185590960085392,
      "step": 151
    },
    {
      "completion_length": 19.203125,
      "epoch": 0.3183246073298429,
      "grad_norm": 7.924731254577637,
      "kl": 0.75390625,
      "learning_rate": 9.208595387840671e-07,
      "loss": -0.0031,
      "reward": 0.6289562582969666,
      "reward_std": 0.107938963919878,
      "reward_std_all": 0.1374085210263729,
      "rewards-max/forensics_sinkhorn_reward": 0.9463672637939453,
      "rewards-max/forensics_strict_edge_reward": 0.8745477199554443,
      "rewards-max/forensics_tiou_reward": 0.9093861877918243,
      "rewards/forensics_sinkhorn_reward": 0.48705074191093445,
      "rewards/forensics_strict_edge_reward": 0.2502259165048599,
      "rewards/forensics_tiou_reward": 0.27147676050662994,
      "step": 152
    },
    {
      "completion_length": 23.140625,
      "epoch": 0.3204188481675393,
      "grad_norm": 9.88139533996582,
      "kl": 0.740234375,
      "learning_rate": 9.203354297693921e-07,
      "loss": 0.1161,
      "reward": 0.5644748508930206,
      "reward_std": 0.20388392359018326,
      "reward_std_all": 0.21630621701478958,
      "rewards-max/forensics_sinkhorn_reward": 0.9243288636207581,
      "rewards-max/forensics_strict_edge_reward": 0.7988545596599579,
      "rewards-max/forensics_tiou_reward": 0.8539114892482758,
      "rewards/forensics_sinkhorn_reward": 0.45306526124477386,
      "rewards/forensics_strict_edge_reward": 0.18319132924079895,
      "rewards/forensics_tiou_reward": 0.23151644319295883,
      "step": 153
    },
    {
      "completion_length": 16.21875,
      "epoch": 0.3225130890052356,
      "grad_norm": 8.18285846710205,
      "kl": 0.5849609375,
      "learning_rate": 9.198113207547169e-07,
      "loss": 0.2192,
      "reward": 0.6727453172206879,
      "reward_std": 0.10072077438235283,
      "reward_std_all": 0.11592096835374832,
      "rewards-max/forensics_sinkhorn_reward": 0.9872389137744904,
      "rewards-max/forensics_strict_edge_reward": 0.9357379674911499,
      "rewards-max/forensics_tiou_reward": 0.9931209087371826,
      "rewards/forensics_sinkhorn_reward": 0.5448403060436249,
      "rewards/forensics_strict_edge_reward": 0.28229179978370667,
      "rewards/forensics_tiou_reward": 0.30605755746364594,
      "step": 154
    },
    {
      "completion_length": 16.59375,
      "epoch": 0.32460732984293195,
      "grad_norm": 13.306401252746582,
      "kl": 0.625,
      "learning_rate": 9.192872117400419e-07,
      "loss": 0.0319,
      "reward": 0.6638397574424744,
      "reward_std": 0.11730729416012764,
      "reward_std_all": 0.1307257078588009,
      "rewards-max/forensics_sinkhorn_reward": 0.9728704988956451,
      "rewards-max/forensics_strict_edge_reward": 0.7823345363140106,
      "rewards-max/forensics_tiou_reward": 0.9366272687911987,
      "rewards/forensics_sinkhorn_reward": 0.5526441335678101,
      "rewards/forensics_strict_edge_reward": 0.24904055148363113,
      "rewards/forensics_tiou_reward": 0.30684809386730194,
      "step": 155
    },
    {
      "completion_length": 16.203125,
      "epoch": 0.3267015706806283,
      "grad_norm": 10.463306427001953,
      "kl": 0.57421875,
      "learning_rate": 9.187631027253668e-07,
      "loss": 0.016,
      "reward": 0.636642575263977,
      "reward_std": 0.15712618082761765,
      "reward_std_all": 0.16228606551885605,
      "rewards-max/forensics_sinkhorn_reward": 0.8889061510562897,
      "rewards-max/forensics_strict_edge_reward": 0.7184170186519623,
      "rewards-max/forensics_tiou_reward": 0.7909715175628662,
      "rewards/forensics_sinkhorn_reward": 0.5003955960273743,
      "rewards/forensics_strict_edge_reward": 0.21376633271574974,
      "rewards/forensics_tiou_reward": 0.2527909129858017,
      "step": 156
    },
    {
      "completion_length": 18.140625,
      "epoch": 0.3287958115183246,
      "grad_norm": 8.086763381958008,
      "kl": 0.501953125,
      "learning_rate": 9.182389937106918e-07,
      "loss": 0.1725,
      "reward": 0.6272859275341034,
      "reward_std": 0.1612076237797737,
      "reward_std_all": 0.16166704893112183,
      "rewards-max/forensics_sinkhorn_reward": 0.9879385828971863,
      "rewards-max/forensics_strict_edge_reward": 0.9464850425720215,
      "rewards-max/forensics_tiou_reward": 0.9958538115024567,
      "rewards/forensics_sinkhorn_reward": 0.526206761598587,
      "rewards/forensics_strict_edge_reward": 0.2181498110294342,
      "rewards/forensics_tiou_reward": 0.3361535668373108,
      "step": 157
    },
    {
      "completion_length": 17.859375,
      "epoch": 0.3308900523560209,
      "grad_norm": 18.729719161987305,
      "kl": 1.265625,
      "learning_rate": 9.177148846960167e-07,
      "loss": 0.1599,
      "reward": 0.5892984569072723,
      "reward_std": 0.16406942903995514,
      "reward_std_all": 0.18295737355947495,
      "rewards-max/forensics_sinkhorn_reward": 0.977961927652359,
      "rewards-max/forensics_strict_edge_reward": 0.8389894068241119,
      "rewards-max/forensics_tiou_reward": 0.9573406279087067,
      "rewards/forensics_sinkhorn_reward": 0.5041118264198303,
      "rewards/forensics_strict_edge_reward": 0.12140719592571259,
      "rewards/forensics_tiou_reward": 0.22310589998960495,
      "step": 158
    },
    {
      "completion_length": 19.03125,
      "epoch": 0.33298429319371725,
      "grad_norm": 8.705206871032715,
      "kl": 0.8046875,
      "learning_rate": 9.171907756813417e-07,
      "loss": 0.2268,
      "reward": 0.566638708114624,
      "reward_std": 0.17279722541570663,
      "reward_std_all": 0.19561578333377838,
      "rewards-max/forensics_sinkhorn_reward": 0.9118331670761108,
      "rewards-max/forensics_strict_edge_reward": 0.7010898292064667,
      "rewards-max/forensics_tiou_reward": 0.8287356793880463,
      "rewards/forensics_sinkhorn_reward": 0.3977857083082199,
      "rewards/forensics_strict_edge_reward": 0.1383293680846691,
      "rewards/forensics_tiou_reward": 0.2007235810160637,
      "step": 159
    },
    {
      "completion_length": 17.703125,
      "epoch": 0.33507853403141363,
      "grad_norm": 9.767001152038574,
      "kl": 0.462890625,
      "learning_rate": 9.166666666666665e-07,
      "loss": 0.072,
      "reward": 0.617378294467926,
      "reward_std": 0.18662354350090027,
      "reward_std_all": 0.18893761187791824,
      "rewards-max/forensics_sinkhorn_reward": 0.9865066707134247,
      "rewards-max/forensics_strict_edge_reward": 0.8241714537143707,
      "rewards-max/forensics_tiou_reward": 0.9902472198009491,
      "rewards/forensics_sinkhorn_reward": 0.4974113553762436,
      "rewards/forensics_strict_edge_reward": 0.19508156925439835,
      "rewards/forensics_tiou_reward": 0.2788073867559433,
      "step": 160
    },
    {
      "completion_length": 18.109375,
      "epoch": 0.33717277486910996,
      "grad_norm": 11.046124458312988,
      "kl": 0.92578125,
      "learning_rate": 9.161425576519915e-07,
      "loss": 0.2049,
      "reward": 0.6071160435676575,
      "reward_std": 0.16546741127967834,
      "reward_std_all": 0.17987412214279175,
      "rewards-max/forensics_sinkhorn_reward": 0.9681143760681152,
      "rewards-max/forensics_strict_edge_reward": 0.8507176637649536,
      "rewards-max/forensics_tiou_reward": 0.9471439719200134,
      "rewards/forensics_sinkhorn_reward": 0.48637135326862335,
      "rewards/forensics_strict_edge_reward": 0.19446956366300583,
      "rewards/forensics_tiou_reward": 0.2601049691438675,
      "step": 161
    },
    {
      "completion_length": 17.140625,
      "epoch": 0.3392670157068063,
      "grad_norm": 9.598858833312988,
      "kl": 0.80078125,
      "learning_rate": 9.156184486373166e-07,
      "loss": 0.2665,
      "reward": 0.5849085748195648,
      "reward_std": 0.1653280332684517,
      "reward_std_all": 0.18436618149280548,
      "rewards-max/forensics_sinkhorn_reward": 0.94355109333992,
      "rewards-max/forensics_strict_edge_reward": 0.6755431294441223,
      "rewards-max/forensics_tiou_reward": 0.8953271806240082,
      "rewards/forensics_sinkhorn_reward": 0.47481897473335266,
      "rewards/forensics_strict_edge_reward": 0.11485162749886513,
      "rewards/forensics_tiou_reward": 0.21313974261283875,
      "step": 162
    },
    {
      "completion_length": 20.53125,
      "epoch": 0.3413612565445026,
      "grad_norm": 7.420440196990967,
      "kl": 0.736328125,
      "learning_rate": 9.150943396226415e-07,
      "loss": 0.0678,
      "reward": 0.6179741621017456,
      "reward_std": 0.19147226959466934,
      "reward_std_all": 0.18470658361911774,
      "rewards-max/forensics_sinkhorn_reward": 0.8629582226276398,
      "rewards-max/forensics_strict_edge_reward": 0.7307599782943726,
      "rewards-max/forensics_tiou_reward": 0.7340010851621628,
      "rewards/forensics_sinkhorn_reward": 0.430791974067688,
      "rewards/forensics_strict_edge_reward": 0.1827091984450817,
      "rewards/forensics_tiou_reward": 0.21899384260177612,
      "step": 163
    },
    {
      "completion_length": 17.9375,
      "epoch": 0.34345549738219894,
      "grad_norm": 8.73521614074707,
      "kl": 1.1171875,
      "learning_rate": 9.145702306079665e-07,
      "loss": 0.0371,
      "reward": 0.5747009217739105,
      "reward_std": 0.20427750796079636,
      "reward_std_all": 0.20843318849802017,
      "rewards-max/forensics_sinkhorn_reward": 0.9237877130508423,
      "rewards-max/forensics_strict_edge_reward": 0.6033799201250076,
      "rewards-max/forensics_tiou_reward": 0.8105555176734924,
      "rewards/forensics_sinkhorn_reward": 0.39630240201950073,
      "rewards/forensics_strict_edge_reward": 0.11987888813018799,
      "rewards/forensics_tiou_reward": 0.1821923702955246,
      "step": 164
    },
    {
      "completion_length": 17.953125,
      "epoch": 0.34554973821989526,
      "grad_norm": 8.592867851257324,
      "kl": 0.69140625,
      "learning_rate": 9.140461215932913e-07,
      "loss": 0.0331,
      "reward": 0.5995545983314514,
      "reward_std": 0.1650599166750908,
      "reward_std_all": 0.17751426994800568,
      "rewards-max/forensics_sinkhorn_reward": 0.9417446255683899,
      "rewards-max/forensics_strict_edge_reward": 0.8854433596134186,
      "rewards-max/forensics_tiou_reward": 0.9002810716629028,
      "rewards/forensics_sinkhorn_reward": 0.4561840742826462,
      "rewards/forensics_strict_edge_reward": 0.13808545097708702,
      "rewards/forensics_tiou_reward": 0.19492115080356598,
      "step": 165
    },
    {
      "completion_length": 19.65625,
      "epoch": 0.34764397905759165,
      "grad_norm": 8.207262992858887,
      "kl": 0.732421875,
      "learning_rate": 9.135220125786163e-07,
      "loss": 0.052,
      "reward": 0.558806449174881,
      "reward_std": 0.1345522664487362,
      "reward_std_all": 0.18196962773799896,
      "rewards-max/forensics_sinkhorn_reward": 0.9467686712741852,
      "rewards-max/forensics_strict_edge_reward": 0.746464729309082,
      "rewards-max/forensics_tiou_reward": 0.866212010383606,
      "rewards/forensics_sinkhorn_reward": 0.43341775238513947,
      "rewards/forensics_strict_edge_reward": 0.09537705779075623,
      "rewards/forensics_tiou_reward": 0.17341331392526627,
      "step": 166
    },
    {
      "completion_length": 17.90625,
      "epoch": 0.34973821989528797,
      "grad_norm": 7.463162899017334,
      "kl": 0.69140625,
      "learning_rate": 9.129979035639412e-07,
      "loss": 0.2205,
      "reward": 0.5693554133176804,
      "reward_std": 0.1458411142230034,
      "reward_std_all": 0.18032611906528473,
      "rewards-max/forensics_sinkhorn_reward": 0.8933024406433105,
      "rewards-max/forensics_strict_edge_reward": 0.7258487641811371,
      "rewards-max/forensics_tiou_reward": 0.7967997789382935,
      "rewards/forensics_sinkhorn_reward": 0.42143887281417847,
      "rewards/forensics_strict_edge_reward": 0.08135611563920975,
      "rewards/forensics_tiou_reward": 0.16178811714053154,
      "step": 167
    },
    {
      "completion_length": 18.59375,
      "epoch": 0.3518324607329843,
      "grad_norm": 8.536108016967773,
      "kl": 0.734375,
      "learning_rate": 9.124737945492662e-07,
      "loss": 0.2792,
      "reward": 0.6421524882316589,
      "reward_std": 0.14527413994073868,
      "reward_std_all": 0.15171798318624496,
      "rewards-max/forensics_sinkhorn_reward": 0.987199068069458,
      "rewards-max/forensics_strict_edge_reward": 0.9158510267734528,
      "rewards-max/forensics_tiou_reward": 0.9929555058479309,
      "rewards/forensics_sinkhorn_reward": 0.4857134222984314,
      "rewards/forensics_strict_edge_reward": 0.17115141451358795,
      "rewards/forensics_tiou_reward": 0.2640385925769806,
      "step": 168
    },
    {
      "completion_length": 17.671875,
      "epoch": 0.3539267015706806,
      "grad_norm": 8.21299934387207,
      "kl": 0.7421875,
      "learning_rate": 9.119496855345912e-07,
      "loss": 0.0205,
      "reward": 0.6107767820358276,
      "reward_std": 0.09513460472226143,
      "reward_std_all": 0.14415302872657776,
      "rewards-max/forensics_sinkhorn_reward": 0.9811064302921295,
      "rewards-max/forensics_strict_edge_reward": 0.8340697288513184,
      "rewards-max/forensics_tiou_reward": 0.9692322313785553,
      "rewards/forensics_sinkhorn_reward": 0.4844350218772888,
      "rewards/forensics_strict_edge_reward": 0.165444478392601,
      "rewards/forensics_tiou_reward": 0.20093795657157898,
      "step": 169
    },
    {
      "completion_length": 23.3125,
      "epoch": 0.35602094240837695,
      "grad_norm": 8.004304885864258,
      "kl": 0.685546875,
      "learning_rate": 9.114255765199161e-07,
      "loss": 0.1898,
      "reward": 0.5933451652526855,
      "reward_std": 0.1548159420490265,
      "reward_std_all": 0.18376466631889343,
      "rewards-max/forensics_sinkhorn_reward": 0.9666195511817932,
      "rewards-max/forensics_strict_edge_reward": 0.7401526570320129,
      "rewards-max/forensics_tiou_reward": 0.9480939507484436,
      "rewards/forensics_sinkhorn_reward": 0.47897981107234955,
      "rewards/forensics_strict_edge_reward": 0.17323888093233109,
      "rewards/forensics_tiou_reward": 0.2973133400082588,
      "step": 170
    },
    {
      "completion_length": 18.5,
      "epoch": 0.3581151832460733,
      "grad_norm": 10.277576446533203,
      "kl": 1.29296875,
      "learning_rate": 9.10901467505241e-07,
      "loss": 0.1554,
      "reward": 0.598002016544342,
      "reward_std": 0.1633664295077324,
      "reward_std_all": 0.17298830300569534,
      "rewards-max/forensics_sinkhorn_reward": 0.9580560624599457,
      "rewards-max/forensics_strict_edge_reward": 0.6991866230964661,
      "rewards-max/forensics_tiou_reward": 0.885978490114212,
      "rewards/forensics_sinkhorn_reward": 0.4646598696708679,
      "rewards/forensics_strict_edge_reward": 0.180337555706501,
      "rewards/forensics_tiou_reward": 0.26354554295539856,
      "step": 171
    },
    {
      "completion_length": 18.28125,
      "epoch": 0.36020942408376966,
      "grad_norm": 8.66808795928955,
      "kl": 0.845703125,
      "learning_rate": 9.103773584905659e-07,
      "loss": 0.2376,
      "reward": 0.6446488499641418,
      "reward_std": 0.13926415145397186,
      "reward_std_all": 0.1477091833949089,
      "rewards-max/forensics_sinkhorn_reward": 0.9876236617565155,
      "rewards-max/forensics_strict_edge_reward": 0.9567649364471436,
      "rewards-max/forensics_tiou_reward": 0.9946145117282867,
      "rewards/forensics_sinkhorn_reward": 0.5493511855602264,
      "rewards/forensics_strict_edge_reward": 0.20454886555671692,
      "rewards/forensics_tiou_reward": 0.3339194655418396,
      "step": 172
    },
    {
      "completion_length": 17.609375,
      "epoch": 0.362303664921466,
      "grad_norm": 9.819930076599121,
      "kl": 0.740234375,
      "learning_rate": 9.09853249475891e-07,
      "loss": -0.0639,
      "reward": 0.6623355448246002,
      "reward_std": 0.12419157847762108,
      "reward_std_all": 0.1286565586924553,
      "rewards-max/forensics_sinkhorn_reward": 0.9712511599063873,
      "rewards-max/forensics_strict_edge_reward": 0.8876297473907471,
      "rewards-max/forensics_tiou_reward": 0.9584318101406097,
      "rewards/forensics_sinkhorn_reward": 0.5520265698432922,
      "rewards/forensics_strict_edge_reward": 0.20294040068984032,
      "rewards/forensics_tiou_reward": 0.2822934687137604,
      "step": 173
    },
    {
      "completion_length": 18.65625,
      "epoch": 0.3643979057591623,
      "grad_norm": 10.808181762695312,
      "kl": 0.6796875,
      "learning_rate": 9.093291404612159e-07,
      "loss": 0.0589,
      "reward": 0.6026942133903503,
      "reward_std": 0.10994680598378181,
      "reward_std_all": 0.14983680099248886,
      "rewards-max/forensics_sinkhorn_reward": 0.9077427685260773,
      "rewards-max/forensics_strict_edge_reward": 0.663029745221138,
      "rewards-max/forensics_tiou_reward": 0.8156323730945587,
      "rewards/forensics_sinkhorn_reward": 0.47436755895614624,
      "rewards/forensics_strict_edge_reward": 0.09750406257808208,
      "rewards/forensics_tiou_reward": 0.1648290976881981,
      "step": 174
    },
    {
      "completion_length": 20.59375,
      "epoch": 0.36649214659685864,
      "grad_norm": 11.18107795715332,
      "kl": 1.2265625,
      "learning_rate": 9.088050314465409e-07,
      "loss": 0.3002,
      "reward": 0.6222840845584869,
      "reward_std": 0.10971655696630478,
      "reward_std_all": 0.1442602202296257,
      "rewards-max/forensics_sinkhorn_reward": 0.90973100066185,
      "rewards-max/forensics_strict_edge_reward": 0.6277818828821182,
      "rewards-max/forensics_tiou_reward": 0.8395801782608032,
      "rewards/forensics_sinkhorn_reward": 0.48114733397960663,
      "rewards/forensics_strict_edge_reward": 0.128538366407156,
      "rewards/forensics_tiou_reward": 0.1890803948044777,
      "step": 175
    },
    {
      "completion_length": 19.78125,
      "epoch": 0.36858638743455496,
      "grad_norm": 9.74940013885498,
      "kl": 1.2265625,
      "learning_rate": 9.082809224318659e-07,
      "loss": 0.1734,
      "reward": 0.6231693625450134,
      "reward_std": 0.16423186659812927,
      "reward_std_all": 0.16999734193086624,
      "rewards-max/forensics_sinkhorn_reward": 0.907945305109024,
      "rewards-max/forensics_strict_edge_reward": 0.6695265024900436,
      "rewards-max/forensics_tiou_reward": 0.8374943435192108,
      "rewards/forensics_sinkhorn_reward": 0.4611458480358124,
      "rewards/forensics_strict_edge_reward": 0.1273398920893669,
      "rewards/forensics_tiou_reward": 0.21521585434675217,
      "step": 176
    },
    {
      "completion_length": 19.25,
      "epoch": 0.3706806282722513,
      "grad_norm": 10.87139892578125,
      "kl": 1.04296875,
      "learning_rate": 9.077568134171907e-07,
      "loss": 0.2742,
      "reward": 0.6426542401313782,
      "reward_std": 0.1269231215119362,
      "reward_std_all": 0.14161694794893265,
      "rewards-max/forensics_sinkhorn_reward": 0.9391684532165527,
      "rewards-max/forensics_strict_edge_reward": 0.8025174736976624,
      "rewards-max/forensics_tiou_reward": 0.8850953280925751,
      "rewards/forensics_sinkhorn_reward": 0.47801314294338226,
      "rewards/forensics_strict_edge_reward": 0.18379263579845428,
      "rewards/forensics_tiou_reward": 0.22841592133045197,
      "step": 177
    },
    {
      "completion_length": 17.46875,
      "epoch": 0.37277486910994767,
      "grad_norm": 9.956706047058105,
      "kl": 0.4970703125,
      "learning_rate": 9.072327044025157e-07,
      "loss": 0.035,
      "reward": 0.6301238536834717,
      "reward_std": 0.12171730026602745,
      "reward_std_all": 0.14206518232822418,
      "rewards-max/forensics_sinkhorn_reward": 0.9071512520313263,
      "rewards-max/forensics_strict_edge_reward": 0.6724614500999451,
      "rewards-max/forensics_tiou_reward": 0.8093510866165161,
      "rewards/forensics_sinkhorn_reward": 0.4975679814815521,
      "rewards/forensics_strict_edge_reward": 0.12767314165830612,
      "rewards/forensics_tiou_reward": 0.19833825528621674,
      "step": 178
    },
    {
      "completion_length": 18.109375,
      "epoch": 0.374869109947644,
      "grad_norm": 10.9401273727417,
      "kl": 0.7890625,
      "learning_rate": 9.067085953878406e-07,
      "loss": 0.1831,
      "reward": 0.6882818341255188,
      "reward_std": 0.09623201191425323,
      "reward_std_all": 0.10110419616103172,
      "rewards-max/forensics_sinkhorn_reward": 0.9880844354629517,
      "rewards-max/forensics_strict_edge_reward": 0.9782643616199493,
      "rewards-max/forensics_tiou_reward": 0.9964285790920258,
      "rewards/forensics_sinkhorn_reward": 0.6026671826839447,
      "rewards/forensics_strict_edge_reward": 0.28005097061395645,
      "rewards/forensics_tiou_reward": 0.40375521779060364,
      "step": 179
    },
    {
      "completion_length": 17.921875,
      "epoch": 0.3769633507853403,
      "grad_norm": 9.162959098815918,
      "kl": 0.4873046875,
      "learning_rate": 9.061844863731656e-07,
      "loss": 0.0552,
      "reward": 0.6219642162322998,
      "reward_std": 0.14752326160669327,
      "reward_std_all": 0.16037575900554657,
      "rewards-max/forensics_sinkhorn_reward": 0.9797694385051727,
      "rewards-max/forensics_strict_edge_reward": 0.8291563987731934,
      "rewards-max/forensics_tiou_reward": 0.9639424085617065,
      "rewards/forensics_sinkhorn_reward": 0.47880449891090393,
      "rewards/forensics_strict_edge_reward": 0.13755852729082108,
      "rewards/forensics_tiou_reward": 0.1985684186220169,
      "step": 180
    },
    {
      "completion_length": 18.015625,
      "epoch": 0.37905759162303665,
      "grad_norm": 11.33039379119873,
      "kl": 1.333984375,
      "learning_rate": 9.056603773584905e-07,
      "loss": 0.0763,
      "reward": 0.6178318858146667,
      "reward_std": 0.18433551490306854,
      "reward_std_all": 0.18438191711902618,
      "rewards-max/forensics_sinkhorn_reward": 0.9771502614021301,
      "rewards-max/forensics_strict_edge_reward": 0.7212928533554077,
      "rewards-max/forensics_tiou_reward": 0.9534251987934113,
      "rewards/forensics_sinkhorn_reward": 0.5053772479295731,
      "rewards/forensics_strict_edge_reward": 0.1274806559085846,
      "rewards/forensics_tiou_reward": 0.23725389689207077,
      "step": 181
    },
    {
      "completion_length": 19.03125,
      "epoch": 0.381151832460733,
      "grad_norm": 10.526856422424316,
      "kl": 0.7265625,
      "learning_rate": 9.051362683438155e-07,
      "loss": 0.2363,
      "reward": 0.6199716925621033,
      "reward_std": 0.1608922928571701,
      "reward_std_all": 0.1700751930475235,
      "rewards-max/forensics_sinkhorn_reward": 0.9870656132698059,
      "rewards-max/forensics_strict_edge_reward": 0.946017861366272,
      "rewards-max/forensics_tiou_reward": 0.9924338161945343,
      "rewards/forensics_sinkhorn_reward": 0.527398094534874,
      "rewards/forensics_strict_edge_reward": 0.2869805321097374,
      "rewards/forensics_tiou_reward": 0.36668479442596436,
      "step": 182
    },
    {
      "completion_length": 17.9375,
      "epoch": 0.3832460732984293,
      "grad_norm": 10.974348068237305,
      "kl": 0.9375,
      "learning_rate": 9.046121593291403e-07,
      "loss": 0.128,
      "reward": 0.6348836421966553,
      "reward_std": 0.10514789074659348,
      "reward_std_all": 0.13323736935853958,
      "rewards-max/forensics_sinkhorn_reward": 0.8683467209339142,
      "rewards-max/forensics_strict_edge_reward": 0.5413009226322174,
      "rewards-max/forensics_tiou_reward": 0.7260300517082214,
      "rewards/forensics_sinkhorn_reward": 0.4548639506101608,
      "rewards/forensics_strict_edge_reward": 0.09944738075137138,
      "rewards/forensics_tiou_reward": 0.18268299847841263,
      "step": 183
    },
    {
      "completion_length": 22.265625,
      "epoch": 0.38534031413612563,
      "grad_norm": 9.537915229797363,
      "kl": 0.9296875,
      "learning_rate": 9.040880503144653e-07,
      "loss": 0.1571,
      "reward": 0.6288592517375946,
      "reward_std": 0.1022612415254116,
      "reward_std_all": 0.13505017757415771,
      "rewards-max/forensics_sinkhorn_reward": 0.9737797975540161,
      "rewards-max/forensics_strict_edge_reward": 0.744240403175354,
      "rewards-max/forensics_tiou_reward": 0.9412475526332855,
      "rewards/forensics_sinkhorn_reward": 0.4479651302099228,
      "rewards/forensics_strict_edge_reward": 0.14892905205488205,
      "rewards/forensics_tiou_reward": 0.19706065952777863,
      "step": 184
    },
    {
      "completion_length": 17.578125,
      "epoch": 0.387434554973822,
      "grad_norm": 11.451875686645508,
      "kl": 1.45703125,
      "learning_rate": 9.035639412997903e-07,
      "loss": 0.039,
      "reward": 0.6594830751419067,
      "reward_std": 0.07049846276640892,
      "reward_std_all": 0.10604196414351463,
      "rewards-max/forensics_sinkhorn_reward": 0.9834267497062683,
      "rewards-max/forensics_strict_edge_reward": 0.8445805907249451,
      "rewards-max/forensics_tiou_reward": 0.9782516360282898,
      "rewards/forensics_sinkhorn_reward": 0.5327247977256775,
      "rewards/forensics_strict_edge_reward": 0.1348271481692791,
      "rewards/forensics_tiou_reward": 0.24268444627523422,
      "step": 185
    },
    {
      "completion_length": 18.53125,
      "epoch": 0.38952879581151834,
      "grad_norm": 10.845983505249023,
      "kl": 1.2734375,
      "learning_rate": 9.030398322851153e-07,
      "loss": 0.0995,
      "reward": 0.6134308278560638,
      "reward_std": 0.15763087570667267,
      "reward_std_all": 0.1679328754544258,
      "rewards-max/forensics_sinkhorn_reward": 0.984023779630661,
      "rewards-max/forensics_strict_edge_reward": 0.8893801271915436,
      "rewards-max/forensics_tiou_reward": 0.980605810880661,
      "rewards/forensics_sinkhorn_reward": 0.4610956907272339,
      "rewards/forensics_strict_edge_reward": 0.18022570759058,
      "rewards/forensics_tiou_reward": 0.21439698338508606,
      "step": 186
    },
    {
      "completion_length": 20.46875,
      "epoch": 0.39162303664921466,
      "grad_norm": 10.975128173828125,
      "kl": 1.380859375,
      "learning_rate": 9.025157232704403e-07,
      "loss": 0.3582,
      "reward": 0.6347965896129608,
      "reward_std": 0.13088324293494225,
      "reward_std_all": 0.14704572409391403,
      "rewards-max/forensics_sinkhorn_reward": 0.9572558999061584,
      "rewards-max/forensics_strict_edge_reward": 0.8291563987731934,
      "rewards-max/forensics_tiou_reward": 0.9235662519931793,
      "rewards/forensics_sinkhorn_reward": 0.4723895788192749,
      "rewards/forensics_strict_edge_reward": 0.15290897712111473,
      "rewards/forensics_tiou_reward": 0.2617165744304657,
      "step": 187
    },
    {
      "completion_length": 19.484375,
      "epoch": 0.393717277486911,
      "grad_norm": 9.157020568847656,
      "kl": 1.0234375,
      "learning_rate": 9.019916142557652e-07,
      "loss": 0.3137,
      "reward": 0.6117814481258392,
      "reward_std": 0.14084792137145996,
      "reward_std_all": 0.1601298525929451,
      "rewards-max/forensics_sinkhorn_reward": 0.9854934513568878,
      "rewards-max/forensics_strict_edge_reward": 0.8760378360748291,
      "rewards-max/forensics_tiou_reward": 0.9862669706344604,
      "rewards/forensics_sinkhorn_reward": 0.4761602431535721,
      "rewards/forensics_strict_edge_reward": 0.19324945658445358,
      "rewards/forensics_tiou_reward": 0.2911766692996025,
      "step": 188
    },
    {
      "completion_length": 20.21875,
      "epoch": 0.3958115183246073,
      "grad_norm": 12.608041763305664,
      "kl": 1.3515625,
      "learning_rate": 9.014675052410901e-07,
      "loss": 0.236,
      "reward": 0.6401252448558807,
      "reward_std": 0.13357801735401154,
      "reward_std_all": 0.14664342999458313,
      "rewards-max/forensics_sinkhorn_reward": 0.9496023058891296,
      "rewards-max/forensics_strict_edge_reward": 0.7932072877883911,
      "rewards-max/forensics_tiou_reward": 0.8597735166549683,
      "rewards/forensics_sinkhorn_reward": 0.5149761140346527,
      "rewards/forensics_strict_edge_reward": 0.20408454537391663,
      "rewards/forensics_tiou_reward": 0.28940411657094955,
      "step": 189
    },
    {
      "completion_length": 17.5,
      "epoch": 0.39790575916230364,
      "grad_norm": 10.546581268310547,
      "kl": 0.79296875,
      "learning_rate": 9.00943396226415e-07,
      "loss": 0.1002,
      "reward": 0.6578803658485413,
      "reward_std": 0.12216107174754143,
      "reward_std_all": 0.13247482478618622,
      "rewards-max/forensics_sinkhorn_reward": 0.9602265655994415,
      "rewards-max/forensics_strict_edge_reward": 0.8771188855171204,
      "rewards-max/forensics_tiou_reward": 0.9330401718616486,
      "rewards/forensics_sinkhorn_reward": 0.5343058109283447,
      "rewards/forensics_strict_edge_reward": 0.1626431867480278,
      "rewards/forensics_tiou_reward": 0.27076925337314606,
      "step": 190
    },
    {
      "completion_length": 19.9375,
      "epoch": 0.4,
      "grad_norm": 9.134513854980469,
      "kl": 1.1875,
      "learning_rate": 9.0041928721174e-07,
      "loss": 0.338,
      "reward": 0.5993331670761108,
      "reward_std": 0.18145322799682617,
      "reward_std_all": 0.18883934617042542,
      "rewards-max/forensics_sinkhorn_reward": 0.9084766209125519,
      "rewards-max/forensics_strict_edge_reward": 0.5625073313713074,
      "rewards-max/forensics_tiou_reward": 0.81744185090065,
      "rewards/forensics_sinkhorn_reward": 0.43235810101032257,
      "rewards/forensics_strict_edge_reward": 0.12087080255150795,
      "rewards/forensics_tiou_reward": 0.20742452144622803,
      "step": 191
    },
    {
      "completion_length": 24.078125,
      "epoch": 0.40209424083769635,
      "grad_norm": 12.603814125061035,
      "kl": 1.39453125,
      "learning_rate": 8.99895178197065e-07,
      "loss": 0.2289,
      "reward": 0.6074932813644409,
      "reward_std": 0.14364003390073776,
      "reward_std_all": 0.1639808490872383,
      "rewards-max/forensics_sinkhorn_reward": 0.9643393456935883,
      "rewards-max/forensics_strict_edge_reward": 0.7184170186519623,
      "rewards-max/forensics_tiou_reward": 0.9077324271202087,
      "rewards/forensics_sinkhorn_reward": 0.4369339048862457,
      "rewards/forensics_strict_edge_reward": 0.18138223886489868,
      "rewards/forensics_tiou_reward": 0.24884232133626938,
      "step": 192
    },
    {
      "completion_length": 21.109375,
      "epoch": 0.4041884816753927,
      "grad_norm": 10.777788162231445,
      "kl": 0.978515625,
      "learning_rate": 8.993710691823899e-07,
      "loss": 0.1434,
      "reward": 0.5880896151065826,
      "reward_std": 0.2044401913881302,
      "reward_std_all": 0.20699328929185867,
      "rewards-max/forensics_sinkhorn_reward": 0.9704753458499908,
      "rewards-max/forensics_strict_edge_reward": 0.7252179682254791,
      "rewards-max/forensics_tiou_reward": 0.9284708201885223,
      "rewards/forensics_sinkhorn_reward": 0.447297140955925,
      "rewards/forensics_strict_edge_reward": 0.13751079887151718,
      "rewards/forensics_tiou_reward": 0.21348582953214645,
      "step": 193
    },
    {
      "completion_length": 26.671875,
      "epoch": 0.406282722513089,
      "grad_norm": 12.027831077575684,
      "kl": 1.36328125,
      "learning_rate": 8.988469601677149e-07,
      "loss": 0.2116,
      "reward": 0.6218685209751129,
      "reward_std": 0.19303740561008453,
      "reward_std_all": 0.18954691290855408,
      "rewards-max/forensics_sinkhorn_reward": 0.9841897785663605,
      "rewards-max/forensics_strict_edge_reward": 0.878633052110672,
      "rewards-max/forensics_tiou_reward": 0.9812250435352325,
      "rewards/forensics_sinkhorn_reward": 0.44156333804130554,
      "rewards/forensics_strict_edge_reward": 0.1938256472349167,
      "rewards/forensics_tiou_reward": 0.2646634802222252,
      "step": 194
    },
    {
      "completion_length": 22.375,
      "epoch": 0.4083769633507853,
      "grad_norm": 13.210715293884277,
      "kl": 1.31640625,
      "learning_rate": 8.983228511530397e-07,
      "loss": 0.291,
      "reward": 0.5452334880828857,
      "reward_std": 0.1717456430196762,
      "reward_std_all": 0.20456937700510025,
      "rewards-max/forensics_sinkhorn_reward": 0.9870094358921051,
      "rewards-max/forensics_strict_edge_reward": 0.9062341153621674,
      "rewards-max/forensics_tiou_reward": 0.9922002553939819,
      "rewards/forensics_sinkhorn_reward": 0.4159773588180542,
      "rewards/forensics_strict_edge_reward": 0.1808074563741684,
      "rewards/forensics_tiou_reward": 0.24930018186569214,
      "step": 195
    },
    {
      "completion_length": 21.375,
      "epoch": 0.41047120418848165,
      "grad_norm": 15.717550277709961,
      "kl": 1.37109375,
      "learning_rate": 8.977987421383647e-07,
      "loss": 0.0984,
      "reward": 0.6292083263397217,
      "reward_std": 0.16963720321655273,
      "reward_std_all": 0.1726069301366806,
      "rewards-max/forensics_sinkhorn_reward": 0.987107902765274,
      "rewards-max/forensics_strict_edge_reward": 0.9464850425720215,
      "rewards-max/forensics_tiou_reward": 0.9925980567932129,
      "rewards/forensics_sinkhorn_reward": 0.4714967906475067,
      "rewards/forensics_strict_edge_reward": 0.21400099992752075,
      "rewards/forensics_tiou_reward": 0.29953067004680634,
      "step": 196
    },
    {
      "completion_length": 22.53125,
      "epoch": 0.41256544502617803,
      "grad_norm": 10.413819313049316,
      "kl": 1.484375,
      "learning_rate": 8.972746331236896e-07,
      "loss": 0.1652,
      "reward": 0.5926506519317627,
      "reward_std": 0.18403460085391998,
      "reward_std_all": 0.1958133801817894,
      "rewards-max/forensics_sinkhorn_reward": 0.8259621858596802,
      "rewards-max/forensics_strict_edge_reward": 0.7038774192333221,
      "rewards-max/forensics_tiou_reward": 0.7787518501281738,
      "rewards/forensics_sinkhorn_reward": 0.45803727209568024,
      "rewards/forensics_strict_edge_reward": 0.135445237159729,
      "rewards/forensics_tiou_reward": 0.20095513761043549,
      "step": 197
    },
    {
      "completion_length": 21.515625,
      "epoch": 0.41465968586387436,
      "grad_norm": 10.629598617553711,
      "kl": 1.4765625,
      "learning_rate": 8.967505241090147e-07,
      "loss": 0.1929,
      "reward": 0.6301287412643433,
      "reward_std": 0.17825506627559662,
      "reward_std_all": 0.17310872673988342,
      "rewards-max/forensics_sinkhorn_reward": 0.9851959049701691,
      "rewards-max/forensics_strict_edge_reward": 0.8983768224716187,
      "rewards-max/forensics_tiou_reward": 0.9851297438144684,
      "rewards/forensics_sinkhorn_reward": 0.47467875480651855,
      "rewards/forensics_strict_edge_reward": 0.20641645044088364,
      "rewards/forensics_tiou_reward": 0.29287189245224,
      "step": 198
    },
    {
      "completion_length": 19.625,
      "epoch": 0.4167539267015707,
      "grad_norm": 11.552783966064453,
      "kl": 1.48828125,
      "learning_rate": 8.962264150943396e-07,
      "loss": 0.1675,
      "reward": 0.6497483253479004,
      "reward_std": 0.15203946828842163,
      "reward_std_all": 0.15847358852624893,
      "rewards-max/forensics_sinkhorn_reward": 0.878698080778122,
      "rewards-max/forensics_strict_edge_reward": 0.7538799047470093,
      "rewards-max/forensics_tiou_reward": 0.790978342294693,
      "rewards/forensics_sinkhorn_reward": 0.4819791615009308,
      "rewards/forensics_strict_edge_reward": 0.18027205765247345,
      "rewards/forensics_tiou_reward": 0.24475117027759552,
      "step": 199
    },
    {
      "completion_length": 19.71875,
      "epoch": 0.418848167539267,
      "grad_norm": 8.212454795837402,
      "kl": 0.814453125,
      "learning_rate": 8.957023060796646e-07,
      "loss": 0.1222,
      "reward": 0.6824156641960144,
      "reward_std": 0.1234564557671547,
      "reward_std_all": 0.12197524681687355,
      "rewards-max/forensics_sinkhorn_reward": 0.9416012167930603,
      "rewards-max/forensics_strict_edge_reward": 0.7303505539894104,
      "rewards-max/forensics_tiou_reward": 0.8874543011188507,
      "rewards/forensics_sinkhorn_reward": 0.5242723822593689,
      "rewards/forensics_strict_edge_reward": 0.18857375532388687,
      "rewards/forensics_tiou_reward": 0.32450515031814575,
      "step": 200
    },
    {
      "completion_length": 26.734375,
      "epoch": 0.42094240837696334,
      "grad_norm": 15.944904327392578,
      "kl": 1.71875,
      "learning_rate": 8.951781970649895e-07,
      "loss": 0.2963,
      "reward": 0.6244710087776184,
      "reward_std": 0.18503447622060776,
      "reward_std_all": 0.17880776524543762,
      "rewards-max/forensics_sinkhorn_reward": 0.9788026809692383,
      "rewards-max/forensics_strict_edge_reward": 0.8719756007194519,
      "rewards-max/forensics_tiou_reward": 0.9724202752113342,
      "rewards/forensics_sinkhorn_reward": 0.44798743724823,
      "rewards/forensics_strict_edge_reward": 0.18948616832494736,
      "rewards/forensics_tiou_reward": 0.24555130302906036,
      "step": 201
    },
    {
      "completion_length": 18.09375,
      "epoch": 0.42303664921465967,
      "grad_norm": 9.899110794067383,
      "kl": 0.580078125,
      "learning_rate": 8.946540880503144e-07,
      "loss": 0.0416,
      "reward": 0.5887658298015594,
      "reward_std": 0.18436544388532639,
      "reward_std_all": 0.19343701004981995,
      "rewards-max/forensics_sinkhorn_reward": 0.9785138070583344,
      "rewards-max/forensics_strict_edge_reward": 0.7939938604831696,
      "rewards-max/forensics_tiou_reward": 0.9587677121162415,
      "rewards/forensics_sinkhorn_reward": 0.4683045595884323,
      "rewards/forensics_strict_edge_reward": 0.18106789141893387,
      "rewards/forensics_tiou_reward": 0.2563960403203964,
      "step": 202
    },
    {
      "completion_length": 20.34375,
      "epoch": 0.42513089005235605,
      "grad_norm": 8.995281219482422,
      "kl": 0.87890625,
      "learning_rate": 8.941299790356394e-07,
      "loss": 0.0004,
      "reward": 0.6504550278186798,
      "reward_std": 0.12576768174767494,
      "reward_std_all": 0.13811925798654556,
      "rewards-max/forensics_sinkhorn_reward": 0.9705483317375183,
      "rewards-max/forensics_strict_edge_reward": 0.7756816148757935,
      "rewards-max/forensics_tiou_reward": 0.9565203487873077,
      "rewards/forensics_sinkhorn_reward": 0.47593964636325836,
      "rewards/forensics_strict_edge_reward": 0.15779831260442734,
      "rewards/forensics_tiou_reward": 0.2396433725953102,
      "step": 203
    },
    {
      "completion_length": 19.40625,
      "epoch": 0.4272251308900524,
      "grad_norm": 10.191656112670898,
      "kl": 0.658203125,
      "learning_rate": 8.936058700209643e-07,
      "loss": 0.1676,
      "reward": 0.6705597043037415,
      "reward_std": 0.12342453747987747,
      "reward_std_all": 0.12603041157126427,
      "rewards-max/forensics_sinkhorn_reward": 0.9868464469909668,
      "rewards-max/forensics_strict_edge_reward": 0.9355069994926453,
      "rewards-max/forensics_tiou_reward": 0.9915668666362762,
      "rewards/forensics_sinkhorn_reward": 0.5705577582120895,
      "rewards/forensics_strict_edge_reward": 0.21674350649118423,
      "rewards/forensics_tiou_reward": 0.33569663763046265,
      "step": 204
    },
    {
      "completion_length": 20.609375,
      "epoch": 0.4293193717277487,
      "grad_norm": 11.116841316223145,
      "kl": 0.8046875,
      "learning_rate": 8.930817610062893e-07,
      "loss": 0.2481,
      "reward": 0.6209653913974762,
      "reward_std": 0.13242827355861664,
      "reward_std_all": 0.15086482465267181,
      "rewards-max/forensics_sinkhorn_reward": 0.9291092753410339,
      "rewards-max/forensics_strict_edge_reward": 0.7557639181613922,
      "rewards-max/forensics_tiou_reward": 0.8619865477085114,
      "rewards/forensics_sinkhorn_reward": 0.4543897956609726,
      "rewards/forensics_strict_edge_reward": 0.12532777339220047,
      "rewards/forensics_tiou_reward": 0.18874046951532364,
      "step": 205
    },
    {
      "completion_length": 19.328125,
      "epoch": 0.431413612565445,
      "grad_norm": 9.921601295471191,
      "kl": 0.998046875,
      "learning_rate": 8.925576519916142e-07,
      "loss": 0.2264,
      "reward": 0.6367908418178558,
      "reward_std": 0.16019775718450546,
      "reward_std_all": 0.16305619478225708,
      "rewards-max/forensics_sinkhorn_reward": 0.9834417402744293,
      "rewards-max/forensics_strict_edge_reward": 0.9075756669044495,
      "rewards-max/forensics_tiou_reward": 0.9783686101436615,
      "rewards/forensics_sinkhorn_reward": 0.49729035794734955,
      "rewards/forensics_strict_edge_reward": 0.17589382082223892,
      "rewards/forensics_tiou_reward": 0.24806836247444153,
      "step": 206
    },
    {
      "completion_length": 20.03125,
      "epoch": 0.43350785340314135,
      "grad_norm": 13.947606086730957,
      "kl": 1.4296875,
      "learning_rate": 8.920335429769391e-07,
      "loss": 0.1201,
      "reward": 0.5972413420677185,
      "reward_std": 0.14160927757620811,
      "reward_std_all": 0.17170073837041855,
      "rewards-max/forensics_sinkhorn_reward": 0.9316500723361969,
      "rewards-max/forensics_strict_edge_reward": 0.7051142752170563,
      "rewards-max/forensics_tiou_reward": 0.8600141406059265,
      "rewards/forensics_sinkhorn_reward": 0.4495324194431305,
      "rewards/forensics_strict_edge_reward": 0.09464028477668762,
      "rewards/forensics_tiou_reward": 0.20346225053071976,
      "step": 207
    },
    {
      "completion_length": 19.046875,
      "epoch": 0.4356020942408377,
      "grad_norm": 10.366974830627441,
      "kl": 0.66796875,
      "learning_rate": 8.915094339622641e-07,
      "loss": 0.0738,
      "reward": 0.6468179523944855,
      "reward_std": 0.10641726851463318,
      "reward_std_all": 0.1300409957766533,
      "rewards-max/forensics_sinkhorn_reward": 0.8267176449298859,
      "rewards-max/forensics_strict_edge_reward": 0.7278552055358887,
      "rewards-max/forensics_tiou_reward": 0.802229642868042,
      "rewards/forensics_sinkhorn_reward": 0.46814677119255066,
      "rewards/forensics_strict_edge_reward": 0.14899854362010956,
      "rewards/forensics_tiou_reward": 0.20075172930955887,
      "step": 208
    },
    {
      "completion_length": 16.796875,
      "epoch": 0.437696335078534,
      "grad_norm": 9.610093116760254,
      "kl": 1.21875,
      "learning_rate": 8.90985324947589e-07,
      "loss": 0.1307,
      "reward": 0.6956358850002289,
      "reward_std": 0.08567466586828232,
      "reward_std_all": 0.09984295815229416,
      "rewards-max/forensics_sinkhorn_reward": 0.9853026568889618,
      "rewards-max/forensics_strict_edge_reward": 0.8237898349761963,
      "rewards-max/forensics_tiou_reward": 0.985506683588028,
      "rewards/forensics_sinkhorn_reward": 0.5353503525257111,
      "rewards/forensics_strict_edge_reward": 0.16902919113636017,
      "rewards/forensics_tiou_reward": 0.2731025964021683,
      "step": 209
    },
    {
      "completion_length": 20.15625,
      "epoch": 0.4397905759162304,
      "grad_norm": 9.912278175354004,
      "kl": 1.4296875,
      "learning_rate": 8.90461215932914e-07,
      "loss": 0.0705,
      "reward": 0.6350713968276978,
      "reward_std": 0.15490558743476868,
      "reward_std_all": 0.16295213252305984,
      "rewards-max/forensics_sinkhorn_reward": 0.9484677910804749,
      "rewards-max/forensics_strict_edge_reward": 0.8097340762615204,
      "rewards-max/forensics_tiou_reward": 0.9038970768451691,
      "rewards/forensics_sinkhorn_reward": 0.5121153444051743,
      "rewards/forensics_strict_edge_reward": 0.19827010482549667,
      "rewards/forensics_tiou_reward": 0.2593724876642227,
      "step": 210
    },
    {
      "completion_length": 17.65625,
      "epoch": 0.4418848167539267,
      "grad_norm": 11.595123291015625,
      "kl": 0.748046875,
      "learning_rate": 8.89937106918239e-07,
      "loss": 0.1302,
      "reward": 0.6243070662021637,
      "reward_std": 0.14167636260390282,
      "reward_std_all": 0.1591150499880314,
      "rewards-max/forensics_sinkhorn_reward": 0.8746742904186249,
      "rewards-max/forensics_strict_edge_reward": 0.6313593983650208,
      "rewards-max/forensics_tiou_reward": 0.7703740894794464,
      "rewards/forensics_sinkhorn_reward": 0.5049678534269333,
      "rewards/forensics_strict_edge_reward": 0.15594472736120224,
      "rewards/forensics_tiou_reward": 0.21857423335313797,
      "step": 211
    },
    {
      "completion_length": 23.171875,
      "epoch": 0.44397905759162304,
      "grad_norm": 12.622501373291016,
      "kl": 1.251953125,
      "learning_rate": 8.89412997903564e-07,
      "loss": 0.1683,
      "reward": 0.618879646062851,
      "reward_std": 0.20569302141666412,
      "reward_std_all": 0.1955024003982544,
      "rewards-max/forensics_sinkhorn_reward": 0.9348609149456024,
      "rewards-max/forensics_strict_edge_reward": 0.7506318390369415,
      "rewards-max/forensics_tiou_reward": 0.8746006786823273,
      "rewards/forensics_sinkhorn_reward": 0.45651574432849884,
      "rewards/forensics_strict_edge_reward": 0.18474052846431732,
      "rewards/forensics_tiou_reward": 0.25953738391399384,
      "step": 212
    },
    {
      "completion_length": 20.203125,
      "epoch": 0.44607329842931936,
      "grad_norm": 23.94741439819336,
      "kl": 1.818359375,
      "learning_rate": 8.888888888888888e-07,
      "loss": 0.1958,
      "reward": 0.6034795641899109,
      "reward_std": 0.15335308015346527,
      "reward_std_all": 0.17327830940485,
      "rewards-max/forensics_sinkhorn_reward": 0.9842705726623535,
      "rewards-max/forensics_strict_edge_reward": 0.8291563987731934,
      "rewards-max/forensics_tiou_reward": 0.9816482365131378,
      "rewards/forensics_sinkhorn_reward": 0.4487856328487396,
      "rewards/forensics_strict_edge_reward": 0.17069876939058304,
      "rewards/forensics_tiou_reward": 0.21309903264045715,
      "step": 213
    },
    {
      "completion_length": 19.03125,
      "epoch": 0.4481675392670157,
      "grad_norm": 11.548115730285645,
      "kl": 1.58984375,
      "learning_rate": 8.883647798742138e-07,
      "loss": 0.2809,
      "reward": 0.6353443264961243,
      "reward_std": 0.11846114322543144,
      "reward_std_all": 0.1384732648730278,
      "rewards-max/forensics_sinkhorn_reward": 0.9183591604232788,
      "rewards-max/forensics_strict_edge_reward": 0.6872453987598419,
      "rewards-max/forensics_tiou_reward": 0.8337136805057526,
      "rewards/forensics_sinkhorn_reward": 0.47452403604984283,
      "rewards/forensics_strict_edge_reward": 0.13780555874109268,
      "rewards/forensics_tiou_reward": 0.19206783920526505,
      "step": 214
    },
    {
      "completion_length": 20.34375,
      "epoch": 0.450261780104712,
      "grad_norm": 11.057548522949219,
      "kl": 1.34765625,
      "learning_rate": 8.878406708595387e-07,
      "loss": 0.1212,
      "reward": 0.6454382836818695,
      "reward_std": 0.14229560643434525,
      "reward_std_all": 0.1498187929391861,
      "rewards-max/forensics_sinkhorn_reward": 0.9166241586208344,
      "rewards-max/forensics_strict_edge_reward": 0.8528120815753937,
      "rewards-max/forensics_tiou_reward": 0.8476732671260834,
      "rewards/forensics_sinkhorn_reward": 0.48279477655887604,
      "rewards/forensics_strict_edge_reward": 0.1858670935034752,
      "rewards/forensics_tiou_reward": 0.2550087794661522,
      "step": 215
    },
    {
      "completion_length": 20.5625,
      "epoch": 0.4523560209424084,
      "grad_norm": 11.94334888458252,
      "kl": 1.71484375,
      "learning_rate": 8.873165618448637e-07,
      "loss": 0.1771,
      "reward": 0.6029662489891052,
      "reward_std": 0.1638682782649994,
      "reward_std_all": 0.17541663348674774,
      "rewards-max/forensics_sinkhorn_reward": 0.8451153934001923,
      "rewards-max/forensics_strict_edge_reward": 0.5951733887195587,
      "rewards-max/forensics_tiou_reward": 0.6880550682544708,
      "rewards/forensics_sinkhorn_reward": 0.39798685908317566,
      "rewards/forensics_strict_edge_reward": 0.12812057882547379,
      "rewards/forensics_tiou_reward": 0.18127574026584625,
      "step": 216
    },
    {
      "completion_length": 21.59375,
      "epoch": 0.4544502617801047,
      "grad_norm": 14.159680366516113,
      "kl": 0.6796875,
      "learning_rate": 8.867924528301887e-07,
      "loss": 0.2159,
      "reward": 0.6069857776165009,
      "reward_std": 0.16513201594352722,
      "reward_std_all": 0.17300346866250038,
      "rewards-max/forensics_sinkhorn_reward": 0.9846606850624084,
      "rewards-max/forensics_strict_edge_reward": 0.9565287232398987,
      "rewards-max/forensics_tiou_reward": 0.9830560684204102,
      "rewards/forensics_sinkhorn_reward": 0.4852840155363083,
      "rewards/forensics_strict_edge_reward": 0.21552613377571106,
      "rewards/forensics_tiou_reward": 0.2796359285712242,
      "step": 217
    },
    {
      "completion_length": 18.265625,
      "epoch": 0.45654450261780105,
      "grad_norm": 12.094685554504395,
      "kl": 0.79296875,
      "learning_rate": 8.862683438155136e-07,
      "loss": 0.2606,
      "reward": 0.6167187988758087,
      "reward_std": 0.18983857333660126,
      "reward_std_all": 0.1855342835187912,
      "rewards-max/forensics_sinkhorn_reward": 0.9877189695835114,
      "rewards-max/forensics_strict_edge_reward": 0.9151731729507446,
      "rewards-max/forensics_tiou_reward": 0.9949954450130463,
      "rewards/forensics_sinkhorn_reward": 0.5037566125392914,
      "rewards/forensics_strict_edge_reward": 0.21014180034399033,
      "rewards/forensics_tiou_reward": 0.3136332333087921,
      "step": 218
    },
    {
      "completion_length": 18.84375,
      "epoch": 0.4586387434554974,
      "grad_norm": 11.930977821350098,
      "kl": 1.39453125,
      "learning_rate": 8.857442348008385e-07,
      "loss": 0.3399,
      "reward": 0.6620303094387054,
      "reward_std": 0.11107823252677917,
      "reward_std_all": 0.1256648451089859,
      "rewards-max/forensics_sinkhorn_reward": 0.9235728979110718,
      "rewards-max/forensics_strict_edge_reward": 0.820955753326416,
      "rewards-max/forensics_tiou_reward": 0.8566561639308929,
      "rewards/forensics_sinkhorn_reward": 0.5058611035346985,
      "rewards/forensics_strict_edge_reward": 0.17156895622611046,
      "rewards/forensics_tiou_reward": 0.2642531394958496,
      "step": 219
    },
    {
      "completion_length": 22.484375,
      "epoch": 0.4607329842931937,
      "grad_norm": 9.989563941955566,
      "kl": 1.51171875,
      "learning_rate": 8.852201257861634e-07,
      "loss": 0.1731,
      "reward": 0.5845200717449188,
      "reward_std": 0.20687153190374374,
      "reward_std_all": 0.21045289933681488,
      "rewards-max/forensics_sinkhorn_reward": 0.8971593081951141,
      "rewards-max/forensics_strict_edge_reward": 0.8237898349761963,
      "rewards-max/forensics_tiou_reward": 0.8119573593139648,
      "rewards/forensics_sinkhorn_reward": 0.4225781261920929,
      "rewards/forensics_strict_edge_reward": 0.1510695219039917,
      "rewards/forensics_tiou_reward": 0.2182348668575287,
      "step": 220
    },
    {
      "completion_length": 19.953125,
      "epoch": 0.46282722513089003,
      "grad_norm": 9.933084487915039,
      "kl": 1.33984375,
      "learning_rate": 8.846960167714884e-07,
      "loss": 0.2524,
      "reward": 0.6113112270832062,
      "reward_std": 0.17218048870563507,
      "reward_std_all": 0.17913232743740082,
      "rewards-max/forensics_sinkhorn_reward": 0.9676103293895721,
      "rewards-max/forensics_strict_edge_reward": 0.8655564188957214,
      "rewards-max/forensics_tiou_reward": 0.9478354454040527,
      "rewards/forensics_sinkhorn_reward": 0.48098771274089813,
      "rewards/forensics_strict_edge_reward": 0.1566062644124031,
      "rewards/forensics_tiou_reward": 0.2552594691514969,
      "step": 221
    },
    {
      "completion_length": 22.4375,
      "epoch": 0.4649214659685864,
      "grad_norm": 11.46503734588623,
      "kl": 1.46484375,
      "learning_rate": 8.841719077568133e-07,
      "loss": 0.2777,
      "reward": 0.6059295535087585,
      "reward_std": 0.13674943149089813,
      "reward_std_all": 0.15885312855243683,
      "rewards-max/forensics_sinkhorn_reward": 0.8833114802837372,
      "rewards-max/forensics_strict_edge_reward": 0.7778737246990204,
      "rewards-max/forensics_tiou_reward": 0.7657861411571503,
      "rewards/forensics_sinkhorn_reward": 0.41657425463199615,
      "rewards/forensics_strict_edge_reward": 0.1477280557155609,
      "rewards/forensics_tiou_reward": 0.19093259423971176,
      "step": 222
    },
    {
      "completion_length": 20.28125,
      "epoch": 0.46701570680628274,
      "grad_norm": 11.161123275756836,
      "kl": 0.810546875,
      "learning_rate": 8.836477987421384e-07,
      "loss": 0.2487,
      "reward": 0.6594053208827972,
      "reward_std": 0.10789573565125465,
      "reward_std_all": 0.1265646144747734,
      "rewards-max/forensics_sinkhorn_reward": 0.9869451820850372,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9919410347938538,
      "rewards/forensics_sinkhorn_reward": 0.5217055082321167,
      "rewards/forensics_strict_edge_reward": 0.23275743424892426,
      "rewards/forensics_tiou_reward": 0.32837677001953125,
      "step": 223
    },
    {
      "completion_length": 19.5,
      "epoch": 0.46910994764397906,
      "grad_norm": 10.481987953186035,
      "kl": 1.26171875,
      "learning_rate": 8.831236897274634e-07,
      "loss": 0.1563,
      "reward": 0.5964893996715546,
      "reward_std": 0.18408669531345367,
      "reward_std_all": 0.1866726502776146,
      "rewards-max/forensics_sinkhorn_reward": 0.9861730635166168,
      "rewards-max/forensics_strict_edge_reward": 0.855939507484436,
      "rewards-max/forensics_tiou_reward": 0.9888996779918671,
      "rewards/forensics_sinkhorn_reward": 0.44590429961681366,
      "rewards/forensics_strict_edge_reward": 0.14781560003757477,
      "rewards/forensics_tiou_reward": 0.22111254185438156,
      "step": 224
    },
    {
      "completion_length": 18.609375,
      "epoch": 0.4712041884816754,
      "grad_norm": 11.822491645812988,
      "kl": 1.765625,
      "learning_rate": 8.825995807127882e-07,
      "loss": 0.2773,
      "reward": 0.6880613267421722,
      "reward_std": 0.11635759472846985,
      "reward_std_all": 0.12115374952554703,
      "rewards-max/forensics_sinkhorn_reward": 0.9858582019805908,
      "rewards-max/forensics_strict_edge_reward": 0.8171346187591553,
      "rewards-max/forensics_tiou_reward": 0.9876964688301086,
      "rewards/forensics_sinkhorn_reward": 0.5708259344100952,
      "rewards/forensics_strict_edge_reward": 0.2687528282403946,
      "rewards/forensics_tiou_reward": 0.36202745139598846,
      "step": 225
    },
    {
      "completion_length": 21.078125,
      "epoch": 0.4732984293193717,
      "grad_norm": 14.255363464355469,
      "kl": 1.3828125,
      "learning_rate": 8.820754716981132e-07,
      "loss": 0.3381,
      "reward": 0.6265283226966858,
      "reward_std": 0.18439937382936478,
      "reward_std_all": 0.18427983671426773,
      "rewards-max/forensics_sinkhorn_reward": 0.9809577465057373,
      "rewards-max/forensics_strict_edge_reward": 0.894839346408844,
      "rewards-max/forensics_tiou_reward": 0.9687182903289795,
      "rewards/forensics_sinkhorn_reward": 0.5177361369132996,
      "rewards/forensics_strict_edge_reward": 0.2583511024713516,
      "rewards/forensics_tiou_reward": 0.33340199291706085,
      "step": 226
    },
    {
      "completion_length": 18.640625,
      "epoch": 0.47539267015706804,
      "grad_norm": 9.652853012084961,
      "kl": 1.640625,
      "learning_rate": 8.815513626834381e-07,
      "loss": 0.281,
      "reward": 0.6611426472663879,
      "reward_std": 0.14213934540748596,
      "reward_std_all": 0.14572670683264732,
      "rewards-max/forensics_sinkhorn_reward": 0.9882901012897491,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9972389936447144,
      "rewards/forensics_sinkhorn_reward": 0.5762556791305542,
      "rewards/forensics_strict_edge_reward": 0.2821539491415024,
      "rewards/forensics_tiou_reward": 0.3353319317102432,
      "step": 227
    },
    {
      "completion_length": 19.140625,
      "epoch": 0.4774869109947644,
      "grad_norm": 15.409743309020996,
      "kl": 1.9375,
      "learning_rate": 8.810272536687631e-07,
      "loss": 0.1841,
      "reward": 0.5525830537080765,
      "reward_std": 0.18670641630887985,
      "reward_std_all": 0.20397336781024933,
      "rewards-max/forensics_sinkhorn_reward": 0.9297569692134857,
      "rewards-max/forensics_strict_edge_reward": 0.756708562374115,
      "rewards-max/forensics_tiou_reward": 0.876034677028656,
      "rewards/forensics_sinkhorn_reward": 0.40268008410930634,
      "rewards/forensics_strict_edge_reward": 0.10970358550548553,
      "rewards/forensics_tiou_reward": 0.14698106050491333,
      "step": 228
    },
    {
      "completion_length": 23.4375,
      "epoch": 0.47958115183246075,
      "grad_norm": 10.98275089263916,
      "kl": 1.23828125,
      "learning_rate": 8.80503144654088e-07,
      "loss": 0.1005,
      "reward": 0.6011937856674194,
      "reward_std": 0.1852237433195114,
      "reward_std_all": 0.19361133128404617,
      "rewards-max/forensics_sinkhorn_reward": 0.9292914867401123,
      "rewards-max/forensics_strict_edge_reward": 0.756708562374115,
      "rewards-max/forensics_tiou_reward": 0.86374831199646,
      "rewards/forensics_sinkhorn_reward": 0.43942247331142426,
      "rewards/forensics_strict_edge_reward": 0.23821015655994415,
      "rewards/forensics_tiou_reward": 0.2720596417784691,
      "step": 229
    },
    {
      "completion_length": 20.375,
      "epoch": 0.4816753926701571,
      "grad_norm": 11.295706748962402,
      "kl": 1.09375,
      "learning_rate": 8.79979035639413e-07,
      "loss": 0.1832,
      "reward": 0.5675439536571503,
      "reward_std": 0.17636221647262573,
      "reward_std_all": 0.20130669325590134,
      "rewards-max/forensics_sinkhorn_reward": 0.8824518620967865,
      "rewards-max/forensics_strict_edge_reward": 0.758176863193512,
      "rewards-max/forensics_tiou_reward": 0.7761743068695068,
      "rewards/forensics_sinkhorn_reward": 0.42871665954589844,
      "rewards/forensics_strict_edge_reward": 0.17765335738658905,
      "rewards/forensics_tiou_reward": 0.20458955317735672,
      "step": 230
    },
    {
      "completion_length": 19.46875,
      "epoch": 0.4837696335078534,
      "grad_norm": 11.190963745117188,
      "kl": 2.4609375,
      "learning_rate": 8.794549266247378e-07,
      "loss": 0.2425,
      "reward": 0.6368997693061829,
      "reward_std": 0.1089475080370903,
      "reward_std_all": 0.1377805694937706,
      "rewards-max/forensics_sinkhorn_reward": 0.9458398818969727,
      "rewards-max/forensics_strict_edge_reward": 0.8121343851089478,
      "rewards-max/forensics_tiou_reward": 0.905963659286499,
      "rewards/forensics_sinkhorn_reward": 0.4642655849456787,
      "rewards/forensics_strict_edge_reward": 0.20543164759874344,
      "rewards/forensics_tiou_reward": 0.3115406483411789,
      "step": 231
    },
    {
      "completion_length": 21.703125,
      "epoch": 0.48586387434554973,
      "grad_norm": 10.724344253540039,
      "kl": 1.7421875,
      "learning_rate": 8.789308176100628e-07,
      "loss": 0.304,
      "reward": 0.599682480096817,
      "reward_std": 0.18585573881864548,
      "reward_std_all": 0.19243714958429337,
      "rewards-max/forensics_sinkhorn_reward": 0.974128931760788,
      "rewards-max/forensics_strict_edge_reward": 0.5682895183563232,
      "rewards-max/forensics_tiou_reward": 0.9426842033863068,
      "rewards/forensics_sinkhorn_reward": 0.47878897190093994,
      "rewards/forensics_strict_edge_reward": 0.10730931535363197,
      "rewards/forensics_tiou_reward": 0.26870255917310715,
      "step": 232
    },
    {
      "completion_length": 17.890625,
      "epoch": 0.48795811518324606,
      "grad_norm": 10.167710304260254,
      "kl": 1.24609375,
      "learning_rate": 8.784067085953878e-07,
      "loss": 0.2057,
      "reward": 0.6095847487449646,
      "reward_std": 0.17061514407396317,
      "reward_std_all": 0.18300533294677734,
      "rewards-max/forensics_sinkhorn_reward": 0.9880791902542114,
      "rewards-max/forensics_strict_edge_reward": 0.9565287232398987,
      "rewards-max/forensics_tiou_reward": 0.9964075982570648,
      "rewards/forensics_sinkhorn_reward": 0.5379672646522522,
      "rewards/forensics_strict_edge_reward": 0.233327217400074,
      "rewards/forensics_tiou_reward": 0.31844523549079895,
      "step": 233
    },
    {
      "completion_length": 19.3125,
      "epoch": 0.4900523560209424,
      "grad_norm": 13.74911880493164,
      "kl": 0.884765625,
      "learning_rate": 8.778825995807127e-07,
      "loss": 0.1712,
      "reward": 0.6139985918998718,
      "reward_std": 0.13486237078905106,
      "reward_std_all": 0.16402269899845123,
      "rewards-max/forensics_sinkhorn_reward": 0.9859665334224701,
      "rewards-max/forensics_strict_edge_reward": 0.9158510267734528,
      "rewards-max/forensics_tiou_reward": 0.9881339371204376,
      "rewards/forensics_sinkhorn_reward": 0.4845893532037735,
      "rewards/forensics_strict_edge_reward": 0.1579417549073696,
      "rewards/forensics_tiou_reward": 0.25601163506507874,
      "step": 234
    },
    {
      "completion_length": 19.0625,
      "epoch": 0.49214659685863876,
      "grad_norm": 12.670844078063965,
      "kl": 1.0546875,
      "learning_rate": 8.773584905660378e-07,
      "loss": 0.1894,
      "reward": 0.6612487435340881,
      "reward_std": 0.0955377109348774,
      "reward_std_all": 0.12126945704221725,
      "rewards-max/forensics_sinkhorn_reward": 0.987452358007431,
      "rewards-max/forensics_strict_edge_reward": 0.946017861366272,
      "rewards-max/forensics_tiou_reward": 0.9939419031143188,
      "rewards/forensics_sinkhorn_reward": 0.5714767277240753,
      "rewards/forensics_strict_edge_reward": 0.28266511857509613,
      "rewards/forensics_tiou_reward": 0.371918648481369,
      "step": 235
    },
    {
      "completion_length": 19.53125,
      "epoch": 0.4942408376963351,
      "grad_norm": 12.13754940032959,
      "kl": 0.8359375,
      "learning_rate": 8.768343815513627e-07,
      "loss": 0.1553,
      "reward": 0.6389447152614594,
      "reward_std": 0.14678896218538284,
      "reward_std_all": 0.15581636875867844,
      "rewards-max/forensics_sinkhorn_reward": 0.9315935671329498,
      "rewards-max/forensics_strict_edge_reward": 0.9217896461486816,
      "rewards-max/forensics_tiou_reward": 0.8830103278160095,
      "rewards/forensics_sinkhorn_reward": 0.534663051366806,
      "rewards/forensics_strict_edge_reward": 0.20105185359716415,
      "rewards/forensics_tiou_reward": 0.26473161578178406,
      "step": 236
    },
    {
      "completion_length": 16.734375,
      "epoch": 0.4963350785340314,
      "grad_norm": 12.183223724365234,
      "kl": 1.10546875,
      "learning_rate": 8.763102725366876e-07,
      "loss": 0.2286,
      "reward": 0.6430891454219818,
      "reward_std": 0.12300734221935272,
      "reward_std_all": 0.1425979845225811,
      "rewards-max/forensics_sinkhorn_reward": 0.9848633110523224,
      "rewards-max/forensics_strict_edge_reward": 0.9252271056175232,
      "rewards-max/forensics_tiou_reward": 0.9879868030548096,
      "rewards/forensics_sinkhorn_reward": 0.5449680387973785,
      "rewards/forensics_strict_edge_reward": 0.1868368498980999,
      "rewards/forensics_tiou_reward": 0.26066286861896515,
      "step": 237
    },
    {
      "completion_length": 20.15625,
      "epoch": 0.49842931937172774,
      "grad_norm": 10.820265769958496,
      "kl": 1.02734375,
      "learning_rate": 8.757861635220125e-07,
      "loss": 0.2091,
      "reward": 0.6707737445831299,
      "reward_std": 0.13159937411546707,
      "reward_std_all": 0.13050998747348785,
      "rewards-max/forensics_sinkhorn_reward": 0.9856510162353516,
      "rewards-max/forensics_strict_edge_reward": 0.9375866651535034,
      "rewards-max/forensics_tiou_reward": 0.9869286715984344,
      "rewards/forensics_sinkhorn_reward": 0.5339072495698929,
      "rewards/forensics_strict_edge_reward": 0.18347687274217606,
      "rewards/forensics_tiou_reward": 0.24701425433158875,
      "step": 238
    },
    {
      "completion_length": 23.421875,
      "epoch": 0.5005235602094241,
      "grad_norm": 9.759842872619629,
      "kl": 1.109375,
      "learning_rate": 8.752620545073375e-07,
      "loss": 0.0094,
      "reward": 0.6204088032245636,
      "reward_std": 0.08629471063613892,
      "reward_std_all": 0.13251905143260956,
      "rewards-max/forensics_sinkhorn_reward": 0.7772186100482941,
      "rewards-max/forensics_strict_edge_reward": 0.5912835896015167,
      "rewards-max/forensics_tiou_reward": 0.6671784818172455,
      "rewards/forensics_sinkhorn_reward": 0.44166897237300873,
      "rewards/forensics_strict_edge_reward": 0.13144567608833313,
      "rewards/forensics_tiou_reward": 0.17085856199264526,
      "step": 239
    },
    {
      "completion_length": 16.375,
      "epoch": 0.5026178010471204,
      "grad_norm": 12.816183090209961,
      "kl": 1.0078125,
      "learning_rate": 8.747379454926625e-07,
      "loss": 0.13,
      "reward": 0.6585984230041504,
      "reward_std": 0.13945884257555008,
      "reward_std_all": 0.1448446735739708,
      "rewards-max/forensics_sinkhorn_reward": 0.9419311881065369,
      "rewards-max/forensics_strict_edge_reward": 0.8096016347408295,
      "rewards-max/forensics_tiou_reward": 0.8941901326179504,
      "rewards/forensics_sinkhorn_reward": 0.5716152787208557,
      "rewards/forensics_strict_edge_reward": 0.25246211886405945,
      "rewards/forensics_tiou_reward": 0.34415262937545776,
      "step": 240
    },
    {
      "completion_length": 28.421875,
      "epoch": 0.5047120418848168,
      "grad_norm": 10.55391788482666,
      "kl": 0.9921875,
      "learning_rate": 8.742138364779874e-07,
      "loss": 0.0798,
      "reward": 0.5494154244661331,
      "reward_std": 0.12822873517870903,
      "reward_std_all": 0.18610879778862,
      "rewards-max/forensics_sinkhorn_reward": 0.9724264144897461,
      "rewards-max/forensics_strict_edge_reward": 0.9158510267734528,
      "rewards-max/forensics_tiou_reward": 0.9613032042980194,
      "rewards/forensics_sinkhorn_reward": 0.40277376770973206,
      "rewards/forensics_strict_edge_reward": 0.14319443702697754,
      "rewards/forensics_tiou_reward": 0.19734331592917442,
      "step": 241
    },
    {
      "completion_length": 19.703125,
      "epoch": 0.506806282722513,
      "grad_norm": 11.022564888000488,
      "kl": 1.88671875,
      "learning_rate": 8.736897274633124e-07,
      "loss": 0.2865,
      "reward": 0.6150659620761871,
      "reward_std": 0.16202395781874657,
      "reward_std_all": 0.1739775910973549,
      "rewards-max/forensics_sinkhorn_reward": 0.9597398042678833,
      "rewards-max/forensics_strict_edge_reward": 0.8968285918235779,
      "rewards-max/forensics_tiou_reward": 0.9344729781150818,
      "rewards/forensics_sinkhorn_reward": 0.4543091952800751,
      "rewards/forensics_strict_edge_reward": 0.2540539726614952,
      "rewards/forensics_tiou_reward": 0.2651613801717758,
      "step": 242
    },
    {
      "completion_length": 21.625,
      "epoch": 0.5089005235602094,
      "grad_norm": 12.044021606445312,
      "kl": 1.3828125,
      "learning_rate": 8.731656184486372e-07,
      "loss": 0.0329,
      "reward": 0.5965654850006104,
      "reward_std": 0.1647581234574318,
      "reward_std_all": 0.17566896975040436,
      "rewards-max/forensics_sinkhorn_reward": 0.8859255313873291,
      "rewards-max/forensics_strict_edge_reward": 0.652410238981247,
      "rewards-max/forensics_tiou_reward": 0.7813787758350372,
      "rewards/forensics_sinkhorn_reward": 0.4604240357875824,
      "rewards/forensics_strict_edge_reward": 0.13381407223641872,
      "rewards/forensics_tiou_reward": 0.21927110105752945,
      "step": 243
    },
    {
      "completion_length": 19.9375,
      "epoch": 0.5109947643979058,
      "grad_norm": 10.614558219909668,
      "kl": 0.916015625,
      "learning_rate": 8.726415094339622e-07,
      "loss": 0.0844,
      "reward": 0.6244606971740723,
      "reward_std": 0.1974266991019249,
      "reward_std_all": 0.1876685842871666,
      "rewards-max/forensics_sinkhorn_reward": 0.9822499454021454,
      "rewards-max/forensics_strict_edge_reward": 0.878633052110672,
      "rewards-max/forensics_tiou_reward": 0.9737050533294678,
      "rewards/forensics_sinkhorn_reward": 0.4262179881334305,
      "rewards/forensics_strict_edge_reward": 0.18423140794038773,
      "rewards/forensics_tiou_reward": 0.24428190290927887,
      "step": 244
    },
    {
      "completion_length": 19.578125,
      "epoch": 0.5130890052356021,
      "grad_norm": 10.91183090209961,
      "kl": 0.8671875,
      "learning_rate": 8.721174004192871e-07,
      "loss": 0.0401,
      "reward": 0.5867103636264801,
      "reward_std": 0.1572207286953926,
      "reward_std_all": 0.1791895478963852,
      "rewards-max/forensics_sinkhorn_reward": 0.9488757848739624,
      "rewards-max/forensics_strict_edge_reward": 0.5468612462282181,
      "rewards-max/forensics_tiou_reward": 0.8518532514572144,
      "rewards/forensics_sinkhorn_reward": 0.4340650141239166,
      "rewards/forensics_strict_edge_reward": 0.11194674298167229,
      "rewards/forensics_tiou_reward": 0.1459752917289734,
      "step": 245
    },
    {
      "completion_length": 26.921875,
      "epoch": 0.5151832460732985,
      "grad_norm": 18.292320251464844,
      "kl": 1.216796875,
      "learning_rate": 8.715932914046121e-07,
      "loss": 0.2518,
      "reward": 0.6472869515419006,
      "reward_std": 0.13096467405557632,
      "reward_std_all": 0.14060847461223602,
      "rewards-max/forensics_sinkhorn_reward": 0.872440755367279,
      "rewards-max/forensics_strict_edge_reward": 0.747575581073761,
      "rewards-max/forensics_tiou_reward": 0.7696933150291443,
      "rewards/forensics_sinkhorn_reward": 0.4611731767654419,
      "rewards/forensics_strict_edge_reward": 0.18814902752637863,
      "rewards/forensics_tiou_reward": 0.23474781215190887,
      "step": 246
    },
    {
      "completion_length": 16.53125,
      "epoch": 0.5172774869109947,
      "grad_norm": 13.34090518951416,
      "kl": 0.826171875,
      "learning_rate": 8.71069182389937e-07,
      "loss": 0.2132,
      "reward": 0.6620666980743408,
      "reward_std": 0.12841062247753143,
      "reward_std_all": 0.13238787651062012,
      "rewards-max/forensics_sinkhorn_reward": 0.9762905240058899,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9715782403945923,
      "rewards/forensics_sinkhorn_reward": 0.5459218621253967,
      "rewards/forensics_strict_edge_reward": 0.2293855845928192,
      "rewards/forensics_tiou_reward": 0.30420008301734924,
      "step": 247
    },
    {
      "completion_length": 18.5,
      "epoch": 0.5193717277486911,
      "grad_norm": 13.607845306396484,
      "kl": 0.962890625,
      "learning_rate": 8.705450733752621e-07,
      "loss": 0.1945,
      "reward": 0.6861264407634735,
      "reward_std": 0.10739583522081375,
      "reward_std_all": 0.11431726813316345,
      "rewards-max/forensics_sinkhorn_reward": 0.972829669713974,
      "rewards-max/forensics_strict_edge_reward": 0.9265981018543243,
      "rewards-max/forensics_tiou_reward": 0.9640879034996033,
      "rewards/forensics_sinkhorn_reward": 0.5597173571586609,
      "rewards/forensics_strict_edge_reward": 0.2622419446706772,
      "rewards/forensics_tiou_reward": 0.3440141975879669,
      "step": 248
    },
    {
      "completion_length": 22.5625,
      "epoch": 0.5214659685863874,
      "grad_norm": 11.34325122833252,
      "kl": 1.54296875,
      "learning_rate": 8.70020964360587e-07,
      "loss": 0.0554,
      "reward": 0.6401096284389496,
      "reward_std": 0.13656936958432198,
      "reward_std_all": 0.14781512320041656,
      "rewards-max/forensics_sinkhorn_reward": 0.9063302576541901,
      "rewards-max/forensics_strict_edge_reward": 0.5864547789096832,
      "rewards-max/forensics_tiou_reward": 0.7749464213848114,
      "rewards/forensics_sinkhorn_reward": 0.4638617932796478,
      "rewards/forensics_strict_edge_reward": 0.15476205199956894,
      "rewards/forensics_tiou_reward": 0.20071542263031006,
      "step": 249
    },
    {
      "completion_length": 24.3125,
      "epoch": 0.5235602094240838,
      "grad_norm": 10.636459350585938,
      "kl": 1.53515625,
      "learning_rate": 8.694968553459119e-07,
      "loss": 0.0439,
      "reward": 0.6142244338989258,
      "reward_std": 0.14919263124465942,
      "reward_std_all": 0.17138206958770752,
      "rewards-max/forensics_sinkhorn_reward": 0.9506672918796539,
      "rewards-max/forensics_strict_edge_reward": 0.7743246257305145,
      "rewards-max/forensics_tiou_reward": 0.9143656194210052,
      "rewards/forensics_sinkhorn_reward": 0.45688436925411224,
      "rewards/forensics_strict_edge_reward": 0.17674055695533752,
      "rewards/forensics_tiou_reward": 0.27037547528743744,
      "step": 250
    },
    {
      "completion_length": 23.34375,
      "epoch": 0.5256544502617801,
      "grad_norm": 588.69677734375,
      "kl": 7.71875,
      "learning_rate": 8.689727463312369e-07,
      "loss": 0.3756,
      "reward": 0.61076420545578,
      "reward_std": 0.18395017087459564,
      "reward_std_all": 0.18424991518259048,
      "rewards-max/forensics_sinkhorn_reward": 0.9281600415706635,
      "rewards-max/forensics_strict_edge_reward": 0.8237898349761963,
      "rewards-max/forensics_tiou_reward": 0.8670226633548737,
      "rewards/forensics_sinkhorn_reward": 0.46986447274684906,
      "rewards/forensics_strict_edge_reward": 0.21140076220035553,
      "rewards/forensics_tiou_reward": 0.2658088579773903,
      "step": 251
    },
    {
      "completion_length": 24.578125,
      "epoch": 0.5277486910994764,
      "grad_norm": 9.535083770751953,
      "kl": 1.6171875,
      "learning_rate": 8.684486373165618e-07,
      "loss": 0.2042,
      "reward": 0.6536811590194702,
      "reward_std": 0.13912680745124817,
      "reward_std_all": 0.14550849050283432,
      "rewards-max/forensics_sinkhorn_reward": 0.9876047074794769,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9945449233055115,
      "rewards/forensics_sinkhorn_reward": 0.48236221075057983,
      "rewards/forensics_strict_edge_reward": 0.2457105740904808,
      "rewards/forensics_tiou_reward": 0.3131924271583557,
      "step": 252
    },
    {
      "completion_length": 27.171875,
      "epoch": 0.5298429319371728,
      "grad_norm": 9.614263534545898,
      "kl": 1.125,
      "learning_rate": 8.679245283018868e-07,
      "loss": 0.1485,
      "reward": 0.6077104210853577,
      "reward_std": 0.16962749511003494,
      "reward_std_all": 0.18072225898504257,
      "rewards-max/forensics_sinkhorn_reward": 0.9857004284858704,
      "rewards-max/forensics_strict_edge_reward": 0.8567850589752197,
      "rewards-max/forensics_tiou_reward": 0.9870471358299255,
      "rewards/forensics_sinkhorn_reward": 0.4937954843044281,
      "rewards/forensics_strict_edge_reward": 0.17384696006774902,
      "rewards/forensics_tiou_reward": 0.33197738230228424,
      "step": 253
    },
    {
      "completion_length": 20.125,
      "epoch": 0.5319371727748691,
      "grad_norm": 12.645286560058594,
      "kl": 0.833984375,
      "learning_rate": 8.674004192872117e-07,
      "loss": -0.0069,
      "reward": 0.6367610394954681,
      "reward_std": 0.12490633130073547,
      "reward_std_all": 0.1404905617237091,
      "rewards-max/forensics_sinkhorn_reward": 0.7745434939861298,
      "rewards-max/forensics_strict_edge_reward": 0.6577740609645844,
      "rewards-max/forensics_tiou_reward": 0.6640009731054306,
      "rewards/forensics_sinkhorn_reward": 0.4954592287540436,
      "rewards/forensics_strict_edge_reward": 0.1942625679075718,
      "rewards/forensics_tiou_reward": 0.22668921761214733,
      "step": 254
    },
    {
      "completion_length": 29.078125,
      "epoch": 0.5340314136125655,
      "grad_norm": 14.659982681274414,
      "kl": 1.8515625,
      "learning_rate": 8.668763102725366e-07,
      "loss": 0.1373,
      "reward": 0.5931526720523834,
      "reward_std": 0.15797576308250427,
      "reward_std_all": 0.17367111146450043,
      "rewards-max/forensics_sinkhorn_reward": 0.9785141050815582,
      "rewards-max/forensics_strict_edge_reward": 0.8951683044433594,
      "rewards-max/forensics_tiou_reward": 0.9756398499011993,
      "rewards/forensics_sinkhorn_reward": 0.40184758603572845,
      "rewards/forensics_strict_edge_reward": 0.19196487218141556,
      "rewards/forensics_tiou_reward": 0.28197090327739716,
      "step": 255
    },
    {
      "completion_length": 19.203125,
      "epoch": 0.5361256544502618,
      "grad_norm": 9.24429702758789,
      "kl": 1.251953125,
      "learning_rate": 8.663522012578616e-07,
      "loss": 0.1895,
      "reward": 0.6450924575328827,
      "reward_std": 0.12409597262740135,
      "reward_std_all": 0.13715672492980957,
      "rewards-max/forensics_sinkhorn_reward": 0.9867846071720123,
      "rewards-max/forensics_strict_edge_reward": 0.9357379674911499,
      "rewards-max/forensics_tiou_reward": 0.9913243055343628,
      "rewards/forensics_sinkhorn_reward": 0.5171486139297485,
      "rewards/forensics_strict_edge_reward": 0.21263807266950607,
      "rewards/forensics_tiou_reward": 0.3038148134946823,
      "step": 256
    },
    {
      "completion_length": 20.90625,
      "epoch": 0.5382198952879581,
      "grad_norm": 11.69151496887207,
      "kl": 0.83203125,
      "learning_rate": 8.658280922431865e-07,
      "loss": 0.1452,
      "reward": 0.6427487730979919,
      "reward_std": 0.15864744037389755,
      "reward_std_all": 0.16416601091623306,
      "rewards-max/forensics_sinkhorn_reward": 0.9848810732364655,
      "rewards-max/forensics_strict_edge_reward": 0.8876297473907471,
      "rewards-max/forensics_tiou_reward": 0.9838070869445801,
      "rewards/forensics_sinkhorn_reward": 0.5401178300380707,
      "rewards/forensics_strict_edge_reward": 0.2085530385375023,
      "rewards/forensics_tiou_reward": 0.3162881135940552,
      "step": 257
    },
    {
      "completion_length": 27.203125,
      "epoch": 0.5403141361256545,
      "grad_norm": 10.526512145996094,
      "kl": 1.01171875,
      "learning_rate": 8.653039832285115e-07,
      "loss": 0.1313,
      "reward": 0.6059607863426208,
      "reward_std": 0.17588359862565994,
      "reward_std_all": 0.1802840605378151,
      "rewards-max/forensics_sinkhorn_reward": 0.9180099964141846,
      "rewards-max/forensics_strict_edge_reward": 0.6388376653194427,
      "rewards-max/forensics_tiou_reward": 0.8368385136127472,
      "rewards/forensics_sinkhorn_reward": 0.4001219868659973,
      "rewards/forensics_strict_edge_reward": 0.16527147963643074,
      "rewards/forensics_tiou_reward": 0.20151855051517487,
      "step": 258
    },
    {
      "completion_length": 17.9375,
      "epoch": 0.5424083769633508,
      "grad_norm": 10.990737915039062,
      "kl": 0.91796875,
      "learning_rate": 8.647798742138364e-07,
      "loss": 0.3066,
      "reward": 0.6097714900970459,
      "reward_std": 0.14390994608402252,
      "reward_std_all": 0.16292963922023773,
      "rewards-max/forensics_sinkhorn_reward": 0.9798287153244019,
      "rewards-max/forensics_strict_edge_reward": 0.8968285918235779,
      "rewards-max/forensics_tiou_reward": 0.9644775390625,
      "rewards/forensics_sinkhorn_reward": 0.5214881151914597,
      "rewards/forensics_strict_edge_reward": 0.20291252434253693,
      "rewards/forensics_tiou_reward": 0.29902738332748413,
      "step": 259
    },
    {
      "completion_length": 25.9375,
      "epoch": 0.5445026178010471,
      "grad_norm": 10.252525329589844,
      "kl": 0.8515625,
      "learning_rate": 8.642557651991615e-07,
      "loss": 0.1679,
      "reward": 0.6441268622875214,
      "reward_std": 0.10627899318933487,
      "reward_std_all": 0.12626713886857033,
      "rewards-max/forensics_sinkhorn_reward": 0.9787415862083435,
      "rewards-max/forensics_strict_edge_reward": 0.9567649364471436,
      "rewards-max/forensics_tiou_reward": 0.9773131906986237,
      "rewards/forensics_sinkhorn_reward": 0.4982130229473114,
      "rewards/forensics_strict_edge_reward": 0.187732495367527,
      "rewards/forensics_tiou_reward": 0.2621312588453293,
      "step": 260
    },
    {
      "completion_length": 22.28125,
      "epoch": 0.5465968586387434,
      "grad_norm": 12.592812538146973,
      "kl": 1.171875,
      "learning_rate": 8.637316561844863e-07,
      "loss": 0.0451,
      "reward": 0.6535472273826599,
      "reward_std": 0.08542453125119209,
      "reward_std_all": 0.1176057867705822,
      "rewards-max/forensics_sinkhorn_reward": 0.9879591763019562,
      "rewards-max/forensics_strict_edge_reward": 0.9574736058712006,
      "rewards-max/forensics_tiou_reward": 0.9959432184696198,
      "rewards/forensics_sinkhorn_reward": 0.5085183680057526,
      "rewards/forensics_strict_edge_reward": 0.21683043241500854,
      "rewards/forensics_tiou_reward": 0.2664042115211487,
      "step": 261
    },
    {
      "completion_length": 17.484375,
      "epoch": 0.5486910994764398,
      "grad_norm": 10.631112098693848,
      "kl": 0.634765625,
      "learning_rate": 8.632075471698113e-07,
      "loss": 0.0436,
      "reward": 0.6450890898704529,
      "reward_std": 0.1243838369846344,
      "reward_std_all": 0.13937664777040482,
      "rewards-max/forensics_sinkhorn_reward": 0.9880818426609039,
      "rewards-max/forensics_strict_edge_reward": 0.9364311099052429,
      "rewards-max/forensics_tiou_reward": 0.9964214861392975,
      "rewards/forensics_sinkhorn_reward": 0.5831611454486847,
      "rewards/forensics_strict_edge_reward": 0.21468501538038254,
      "rewards/forensics_tiou_reward": 0.3113110065460205,
      "step": 262
    },
    {
      "completion_length": 24.890625,
      "epoch": 0.5507853403141362,
      "grad_norm": 9.318801879882812,
      "kl": 0.873046875,
      "learning_rate": 8.626834381551362e-07,
      "loss": 0.137,
      "reward": 0.5627602636814117,
      "reward_std": 0.1852949559688568,
      "reward_std_all": 0.20979664474725723,
      "rewards-max/forensics_sinkhorn_reward": 0.9647673666477203,
      "rewards-max/forensics_strict_edge_reward": 0.7928257584571838,
      "rewards-max/forensics_tiou_reward": 0.9468768239021301,
      "rewards/forensics_sinkhorn_reward": 0.4126291275024414,
      "rewards/forensics_strict_edge_reward": 0.17946043610572815,
      "rewards/forensics_tiou_reward": 0.20724184811115265,
      "step": 263
    },
    {
      "completion_length": 19.28125,
      "epoch": 0.5528795811518324,
      "grad_norm": 10.906588554382324,
      "kl": 0.6171875,
      "learning_rate": 8.621593291404612e-07,
      "loss": 0.056,
      "reward": 0.7051398456096649,
      "reward_std": 0.10952962189912796,
      "reward_std_all": 0.11142835393548012,
      "rewards-max/forensics_sinkhorn_reward": 0.9884411692619324,
      "rewards-max/forensics_strict_edge_reward": 0.9565287232398987,
      "rewards-max/forensics_tiou_reward": 0.9978348910808563,
      "rewards/forensics_sinkhorn_reward": 0.599547803401947,
      "rewards/forensics_strict_edge_reward": 0.3132161498069763,
      "rewards/forensics_tiou_reward": 0.41356562077999115,
      "step": 264
    },
    {
      "completion_length": 19.28125,
      "epoch": 0.5549738219895288,
      "grad_norm": 9.921525955200195,
      "kl": 0.646484375,
      "learning_rate": 8.616352201257862e-07,
      "loss": 0.1731,
      "reward": 0.6543572843074799,
      "reward_std": 0.14682098478078842,
      "reward_std_all": 0.14844144135713577,
      "rewards-max/forensics_sinkhorn_reward": 0.9877201914787292,
      "rewards-max/forensics_strict_edge_reward": 0.9357379674911499,
      "rewards-max/forensics_tiou_reward": 0.9949992895126343,
      "rewards/forensics_sinkhorn_reward": 0.5809644460678101,
      "rewards/forensics_strict_edge_reward": 0.24661199748516083,
      "rewards/forensics_tiou_reward": 0.3554646968841553,
      "step": 265
    },
    {
      "completion_length": 15.953125,
      "epoch": 0.5570680628272251,
      "grad_norm": 10.197257995605469,
      "kl": 0.7734375,
      "learning_rate": 8.611111111111111e-07,
      "loss": -0.0418,
      "reward": 0.6931881308555603,
      "reward_std": 0.08814740926027298,
      "reward_std_all": 0.1037798635661602,
      "rewards-max/forensics_sinkhorn_reward": 0.9866759479045868,
      "rewards-max/forensics_strict_edge_reward": 0.8915697336196899,
      "rewards-max/forensics_tiou_reward": 0.9909205436706543,
      "rewards/forensics_sinkhorn_reward": 0.6120786070823669,
      "rewards/forensics_strict_edge_reward": 0.22432257235050201,
      "rewards/forensics_tiou_reward": 0.35451845824718475,
      "step": 266
    },
    {
      "completion_length": 18.609375,
      "epoch": 0.5591623036649215,
      "grad_norm": 10.276932716369629,
      "kl": 0.83203125,
      "learning_rate": 8.60587002096436e-07,
      "loss": 0.0177,
      "reward": 0.66603022813797,
      "reward_std": 0.1096748411655426,
      "reward_std_all": 0.1274995356798172,
      "rewards-max/forensics_sinkhorn_reward": 0.9879591763019562,
      "rewards-max/forensics_strict_edge_reward": 0.9677534997463226,
      "rewards-max/forensics_tiou_reward": 0.9959432184696198,
      "rewards/forensics_sinkhorn_reward": 0.5642237961292267,
      "rewards/forensics_strict_edge_reward": 0.29445063322782516,
      "rewards/forensics_tiou_reward": 0.3711393475532532,
      "step": 267
    },
    {
      "completion_length": 30.734375,
      "epoch": 0.5612565445026177,
      "grad_norm": 12.227344512939453,
      "kl": 1.07421875,
      "learning_rate": 8.600628930817609e-07,
      "loss": 0.1446,
      "reward": 0.6766237020492554,
      "reward_std": 0.10620668902993202,
      "reward_std_all": 0.11363976448774338,
      "rewards-max/forensics_sinkhorn_reward": 0.9557057023048401,
      "rewards-max/forensics_strict_edge_reward": 0.820955753326416,
      "rewards-max/forensics_tiou_reward": 0.9231069982051849,
      "rewards/forensics_sinkhorn_reward": 0.512598305940628,
      "rewards/forensics_strict_edge_reward": 0.22509253025054932,
      "rewards/forensics_tiou_reward": 0.3014216050505638,
      "step": 268
    },
    {
      "completion_length": 25.484375,
      "epoch": 0.5633507853403141,
      "grad_norm": 9.602425575256348,
      "kl": 1.0625,
      "learning_rate": 8.595387840670859e-07,
      "loss": 0.1123,
      "reward": 0.6095170080661774,
      "reward_std": 0.1380365490913391,
      "reward_std_all": 0.15938229858875275,
      "rewards-max/forensics_sinkhorn_reward": 0.9162319600582123,
      "rewards-max/forensics_strict_edge_reward": 0.6628457307815552,
      "rewards-max/forensics_tiou_reward": 0.8427435159683228,
      "rewards/forensics_sinkhorn_reward": 0.4174303412437439,
      "rewards/forensics_strict_edge_reward": 0.151444923132658,
      "rewards/forensics_tiou_reward": 0.20839598774909973,
      "step": 269
    },
    {
      "completion_length": 21.75,
      "epoch": 0.5654450261780105,
      "grad_norm": 10.014823913574219,
      "kl": 0.966796875,
      "learning_rate": 8.590146750524108e-07,
      "loss": 0.3011,
      "reward": 0.5812174677848816,
      "reward_std": 0.16099389642477036,
      "reward_std_all": 0.18333611637353897,
      "rewards-max/forensics_sinkhorn_reward": 0.980864554643631,
      "rewards-max/forensics_strict_edge_reward": 0.8681221902370453,
      "rewards-max/forensics_tiou_reward": 0.979902982711792,
      "rewards/forensics_sinkhorn_reward": 0.44654659926891327,
      "rewards/forensics_strict_edge_reward": 0.17814738303422928,
      "rewards/forensics_tiou_reward": 0.2310381457209587,
      "step": 270
    },
    {
      "completion_length": 19.671875,
      "epoch": 0.5675392670157068,
      "grad_norm": 10.07303524017334,
      "kl": 1.2109375,
      "learning_rate": 8.584905660377358e-07,
      "loss": 0.2136,
      "reward": 0.6221344172954559,
      "reward_std": 0.13788216561079025,
      "reward_std_all": 0.15511102974414825,
      "rewards-max/forensics_sinkhorn_reward": 0.9629762172698975,
      "rewards-max/forensics_strict_edge_reward": 0.7823345363140106,
      "rewards-max/forensics_tiou_reward": 0.941444605588913,
      "rewards/forensics_sinkhorn_reward": 0.4500209242105484,
      "rewards/forensics_strict_edge_reward": 0.14980252087116241,
      "rewards/forensics_tiou_reward": 0.2608255296945572,
      "step": 271
    },
    {
      "completion_length": 21.46875,
      "epoch": 0.5696335078534032,
      "grad_norm": 11.86811351776123,
      "kl": 1.28515625,
      "learning_rate": 8.579664570230609e-07,
      "loss": 0.1773,
      "reward": 0.5433034002780914,
      "reward_std": 0.1966640055179596,
      "reward_std_all": 0.22171452641487122,
      "rewards-max/forensics_sinkhorn_reward": 0.905055046081543,
      "rewards-max/forensics_strict_edge_reward": 0.662954181432724,
      "rewards-max/forensics_tiou_reward": 0.7772940099239349,
      "rewards/forensics_sinkhorn_reward": 0.38922710716724396,
      "rewards/forensics_strict_edge_reward": 0.1295500546693802,
      "rewards/forensics_tiou_reward": 0.1992422193288803,
      "step": 272
    },
    {
      "completion_length": 24.8125,
      "epoch": 0.5717277486910994,
      "grad_norm": 14.495402336120605,
      "kl": 2.05078125,
      "learning_rate": 8.574423480083857e-07,
      "loss": 0.2337,
      "reward": 0.6215314567089081,
      "reward_std": 0.15407642722129822,
      "reward_std_all": 0.1672782152891159,
      "rewards-max/forensics_sinkhorn_reward": 0.9871731698513031,
      "rewards-max/forensics_strict_edge_reward": 0.9565287232398987,
      "rewards-max/forensics_tiou_reward": 0.992836594581604,
      "rewards/forensics_sinkhorn_reward": 0.47476819157600403,
      "rewards/forensics_strict_edge_reward": 0.2068459317088127,
      "rewards/forensics_tiou_reward": 0.26641765981912613,
      "step": 273
    },
    {
      "completion_length": 25.765625,
      "epoch": 0.5738219895287958,
      "grad_norm": 10.029977798461914,
      "kl": 1.171875,
      "learning_rate": 8.569182389937107e-07,
      "loss": 0.1997,
      "reward": 0.55084627866745,
      "reward_std": 0.16871533542871475,
      "reward_std_all": 0.1975908726453781,
      "rewards-max/forensics_sinkhorn_reward": 0.8021385073661804,
      "rewards-max/forensics_strict_edge_reward": 0.7387099117040634,
      "rewards-max/forensics_tiou_reward": 0.7551293075084686,
      "rewards/forensics_sinkhorn_reward": 0.34922564029693604,
      "rewards/forensics_strict_edge_reward": 0.15102166309952736,
      "rewards/forensics_tiou_reward": 0.19315370544791222,
      "step": 274
    },
    {
      "completion_length": 32.15625,
      "epoch": 0.5759162303664922,
      "grad_norm": 20.807415008544922,
      "kl": 2.7578125,
      "learning_rate": 8.563941299790356e-07,
      "loss": 0.3575,
      "reward": 0.5915515124797821,
      "reward_std": 0.17836470901966095,
      "reward_std_all": 0.19313585758209229,
      "rewards-max/forensics_sinkhorn_reward": 0.9841692745685577,
      "rewards-max/forensics_strict_edge_reward": 0.7768298983573914,
      "rewards-max/forensics_tiou_reward": 0.98100346326828,
      "rewards/forensics_sinkhorn_reward": 0.4385092109441757,
      "rewards/forensics_strict_edge_reward": 0.18197031319141388,
      "rewards/forensics_tiou_reward": 0.2586997449398041,
      "step": 275
    },
    {
      "completion_length": 23.90625,
      "epoch": 0.5780104712041885,
      "grad_norm": 11.126840591430664,
      "kl": 1.8984375,
      "learning_rate": 8.558700209643606e-07,
      "loss": 0.1765,
      "reward": 0.6173188090324402,
      "reward_std": 0.12787365913391113,
      "reward_std_all": 0.15444506704807281,
      "rewards-max/forensics_sinkhorn_reward": 0.9485993981361389,
      "rewards-max/forensics_strict_edge_reward": 0.7528610825538635,
      "rewards-max/forensics_tiou_reward": 0.9129574298858643,
      "rewards/forensics_sinkhorn_reward": 0.4631423205137253,
      "rewards/forensics_strict_edge_reward": 0.20487068593502045,
      "rewards/forensics_tiou_reward": 0.26627594232559204,
      "step": 276
    },
    {
      "completion_length": 21.453125,
      "epoch": 0.5801047120418849,
      "grad_norm": 9.955644607543945,
      "kl": 1.46875,
      "learning_rate": 8.553459119496855e-07,
      "loss": 0.322,
      "reward": 0.5931561589241028,
      "reward_std": 0.1421673260629177,
      "reward_std_all": 0.17465003579854965,
      "rewards-max/forensics_sinkhorn_reward": 0.9857864081859589,
      "rewards-max/forensics_strict_edge_reward": 0.9357379674911499,
      "rewards-max/forensics_tiou_reward": 0.9874196648597717,
      "rewards/forensics_sinkhorn_reward": 0.4490702152252197,
      "rewards/forensics_strict_edge_reward": 0.19204433262348175,
      "rewards/forensics_tiou_reward": 0.261814147233963,
      "step": 277
    },
    {
      "completion_length": 26.5,
      "epoch": 0.5821989528795811,
      "grad_norm": 18.301971435546875,
      "kl": 2.44921875,
      "learning_rate": 8.548218029350105e-07,
      "loss": 0.2812,
      "reward": 0.5860403776168823,
      "reward_std": 0.1591469682753086,
      "reward_std_all": 0.1861787587404251,
      "rewards-max/forensics_sinkhorn_reward": 0.9882503747940063,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9970824718475342,
      "rewards/forensics_sinkhorn_reward": 0.4591779261827469,
      "rewards/forensics_strict_edge_reward": 0.1954377070069313,
      "rewards/forensics_tiou_reward": 0.2786851227283478,
      "step": 278
    },
    {
      "completion_length": 21.234375,
      "epoch": 0.5842931937172775,
      "grad_norm": 10.3997163772583,
      "kl": 1.208984375,
      "learning_rate": 8.542976939203353e-07,
      "loss": 0.2165,
      "reward": 0.6420978307723999,
      "reward_std": 0.15240322053432465,
      "reward_std_all": 0.1566336750984192,
      "rewards-max/forensics_sinkhorn_reward": 0.9868761599063873,
      "rewards-max/forensics_strict_edge_reward": 0.9464850425720215,
      "rewards-max/forensics_tiou_reward": 0.9916676878929138,
      "rewards/forensics_sinkhorn_reward": 0.4798199534416199,
      "rewards/forensics_strict_edge_reward": 0.15470466762781143,
      "rewards/forensics_tiou_reward": 0.22316999733448029,
      "step": 279
    },
    {
      "completion_length": 23.390625,
      "epoch": 0.5863874345549738,
      "grad_norm": 12.564027786254883,
      "kl": 2.5078125,
      "learning_rate": 8.537735849056603e-07,
      "loss": 0.2884,
      "reward": 0.6349433660507202,
      "reward_std": 0.16381293535232544,
      "reward_std_all": 0.16786345094442368,
      "rewards-max/forensics_sinkhorn_reward": 0.9581379294395447,
      "rewards-max/forensics_strict_edge_reward": 0.684741199016571,
      "rewards-max/forensics_tiou_reward": 0.9215342402458191,
      "rewards/forensics_sinkhorn_reward": 0.4846723824739456,
      "rewards/forensics_strict_edge_reward": 0.18088749796152115,
      "rewards/forensics_tiou_reward": 0.2533891499042511,
      "step": 280
    },
    {
      "completion_length": 27.6875,
      "epoch": 0.5884816753926702,
      "grad_norm": 9.144640922546387,
      "kl": 1.4140625,
      "learning_rate": 8.532494758909853e-07,
      "loss": 0.1891,
      "reward": 0.6646487712860107,
      "reward_std": 0.11365281045436859,
      "reward_std_all": 0.125677902251482,
      "rewards-max/forensics_sinkhorn_reward": 0.9865549802780151,
      "rewards-max/forensics_strict_edge_reward": 0.9158510267734528,
      "rewards-max/forensics_tiou_reward": 0.9904007315635681,
      "rewards/forensics_sinkhorn_reward": 0.519242137670517,
      "rewards/forensics_strict_edge_reward": 0.23054340481758118,
      "rewards/forensics_tiou_reward": 0.33500367403030396,
      "step": 281
    },
    {
      "completion_length": 23.359375,
      "epoch": 0.5905759162303665,
      "grad_norm": 12.858111381530762,
      "kl": 0.81640625,
      "learning_rate": 8.527253668763102e-07,
      "loss": 0.1738,
      "reward": 0.6262993216514587,
      "reward_std": 0.14854669570922852,
      "reward_std_all": 0.1562729775905609,
      "rewards-max/forensics_sinkhorn_reward": 0.988310843706131,
      "rewards-max/forensics_strict_edge_reward": 0.9256840348243713,
      "rewards-max/forensics_tiou_reward": 0.997321754693985,
      "rewards/forensics_sinkhorn_reward": 0.48979203402996063,
      "rewards/forensics_strict_edge_reward": 0.2148350030183792,
      "rewards/forensics_tiou_reward": 0.29610659182071686,
      "step": 282
    },
    {
      "completion_length": 22.828125,
      "epoch": 0.5926701570680628,
      "grad_norm": 16.72081756591797,
      "kl": 1.115234375,
      "learning_rate": 8.522012578616352e-07,
      "loss": 0.2399,
      "reward": 0.700709342956543,
      "reward_std": 0.14220046997070312,
      "reward_std_all": 0.12945538014173508,
      "rewards-max/forensics_sinkhorn_reward": 0.9872781336307526,
      "rewards-max/forensics_strict_edge_reward": 0.9565287232398987,
      "rewards-max/forensics_tiou_reward": 0.9932678043842316,
      "rewards/forensics_sinkhorn_reward": 0.5523445010185242,
      "rewards/forensics_strict_edge_reward": 0.25640933215618134,
      "rewards/forensics_tiou_reward": 0.32824525237083435,
      "step": 283
    },
    {
      "completion_length": 29.890625,
      "epoch": 0.5947643979057592,
      "grad_norm": 9.99889850616455,
      "kl": 0.765625,
      "learning_rate": 8.516771488469601e-07,
      "loss": 0.0922,
      "reward": 0.7032071053981781,
      "reward_std": 0.11018265411257744,
      "reward_std_all": 0.10897533223032951,
      "rewards-max/forensics_sinkhorn_reward": 0.9863160848617554,
      "rewards-max/forensics_strict_edge_reward": 0.9357379674911499,
      "rewards-max/forensics_tiou_reward": 0.9894646406173706,
      "rewards/forensics_sinkhorn_reward": 0.6062015891075134,
      "rewards/forensics_strict_edge_reward": 0.27359985560178757,
      "rewards/forensics_tiou_reward": 0.386487677693367,
      "step": 284
    },
    {
      "completion_length": 23.28125,
      "epoch": 0.5968586387434555,
      "grad_norm": 11.318429946899414,
      "kl": 1.072265625,
      "learning_rate": 8.51153039832285e-07,
      "loss": 0.0193,
      "reward": 0.6434085369110107,
      "reward_std": 0.048418331891298294,
      "reward_std_all": 0.10341177508234978,
      "rewards-max/forensics_sinkhorn_reward": 0.9835143387317657,
      "rewards-max/forensics_strict_edge_reward": 0.9093653857707977,
      "rewards-max/forensics_tiou_reward": 0.9794752895832062,
      "rewards/forensics_sinkhorn_reward": 0.544183760881424,
      "rewards/forensics_strict_edge_reward": 0.23208844661712646,
      "rewards/forensics_tiou_reward": 0.3402623385190964,
      "step": 285
    },
    {
      "completion_length": 17.75,
      "epoch": 0.5989528795811518,
      "grad_norm": 12.843972206115723,
      "kl": 1.017578125,
      "learning_rate": 8.5062893081761e-07,
      "loss": 0.2177,
      "reward": 0.6750509142875671,
      "reward_std": 0.08791937120258808,
      "reward_std_all": 0.10615499690175056,
      "rewards-max/forensics_sinkhorn_reward": 0.9885258972644806,
      "rewards-max/forensics_strict_edge_reward": 0.9890114367008209,
      "rewards-max/forensics_tiou_reward": 0.9981701672077179,
      "rewards/forensics_sinkhorn_reward": 0.6141742765903473,
      "rewards/forensics_strict_edge_reward": 0.28264525532722473,
      "rewards/forensics_tiou_reward": 0.3651540279388428,
      "step": 286
    },
    {
      "completion_length": 21.6875,
      "epoch": 0.6010471204188481,
      "grad_norm": 13.82807731628418,
      "kl": 1.341796875,
      "learning_rate": 8.50104821802935e-07,
      "loss": 0.1279,
      "reward": 0.5948160290718079,
      "reward_std": 0.07033568248152733,
      "reward_std_all": 0.13315772637724876,
      "rewards-max/forensics_sinkhorn_reward": 0.9264316558837891,
      "rewards-max/forensics_strict_edge_reward": 0.8208384811878204,
      "rewards-max/forensics_tiou_reward": 0.8645309209823608,
      "rewards/forensics_sinkhorn_reward": 0.4539990574121475,
      "rewards/forensics_strict_edge_reward": 0.1366492137312889,
      "rewards/forensics_tiou_reward": 0.16369221359491348,
      "step": 287
    },
    {
      "completion_length": 21.59375,
      "epoch": 0.6031413612565445,
      "grad_norm": 11.781511306762695,
      "kl": 1.234375,
      "learning_rate": 8.4958071278826e-07,
      "loss": 0.0363,
      "reward": 0.6660945117473602,
      "reward_std": 0.11614571884274483,
      "reward_std_all": 0.12419470399618149,
      "rewards-max/forensics_sinkhorn_reward": 0.9887748062610626,
      "rewards-max/forensics_strict_edge_reward": 0.9890114367008209,
      "rewards-max/forensics_tiou_reward": 0.9991500079631805,
      "rewards/forensics_sinkhorn_reward": 0.5598613619804382,
      "rewards/forensics_strict_edge_reward": 0.3067391663789749,
      "rewards/forensics_tiou_reward": 0.3618961572647095,
      "step": 288
    },
    {
      "completion_length": 22.484375,
      "epoch": 0.6052356020942409,
      "grad_norm": 12.806134223937988,
      "kl": 1.095703125,
      "learning_rate": 8.490566037735849e-07,
      "loss": -0.0774,
      "reward": 0.6502562463283539,
      "reward_std": 0.07627071067690849,
      "reward_std_all": 0.11505085974931717,
      "rewards-max/forensics_sinkhorn_reward": 0.9551281929016113,
      "rewards-max/forensics_strict_edge_reward": 0.7391321808099747,
      "rewards-max/forensics_tiou_reward": 0.9217876195907593,
      "rewards/forensics_sinkhorn_reward": 0.4935089200735092,
      "rewards/forensics_strict_edge_reward": 0.14249609224498272,
      "rewards/forensics_tiou_reward": 0.21073096990585327,
      "step": 289
    },
    {
      "completion_length": 20.0625,
      "epoch": 0.6073298429319371,
      "grad_norm": 9.915245056152344,
      "kl": 0.724609375,
      "learning_rate": 8.485324947589099e-07,
      "loss": 0.0859,
      "reward": 0.6528404653072357,
      "reward_std": 0.10795982554554939,
      "reward_std_all": 0.12669019028544426,
      "rewards-max/forensics_sinkhorn_reward": 0.9875630736351013,
      "rewards-max/forensics_strict_edge_reward": 0.9158510267734528,
      "rewards-max/forensics_tiou_reward": 0.9943811595439911,
      "rewards/forensics_sinkhorn_reward": 0.5689123272895813,
      "rewards/forensics_strict_edge_reward": 0.21999113261699677,
      "rewards/forensics_tiou_reward": 0.3354247808456421,
      "step": 290
    },
    {
      "completion_length": 20.5625,
      "epoch": 0.6094240837696335,
      "grad_norm": 13.346386909484863,
      "kl": 1.98046875,
      "learning_rate": 8.480083857442347e-07,
      "loss": 0.1015,
      "reward": 0.6962237358093262,
      "reward_std": 0.07777205295860767,
      "reward_std_all": 0.09483602456748486,
      "rewards-max/forensics_sinkhorn_reward": 0.9869925081729889,
      "rewards-max/forensics_strict_edge_reward": 0.9890114367008209,
      "rewards-max/forensics_tiou_reward": 0.9921259880065918,
      "rewards/forensics_sinkhorn_reward": 0.5513724088668823,
      "rewards/forensics_strict_edge_reward": 0.26574866473674774,
      "rewards/forensics_tiou_reward": 0.34156471490859985,
      "step": 291
    },
    {
      "completion_length": 38.703125,
      "epoch": 0.6115183246073298,
      "grad_norm": 13.123584747314453,
      "kl": 1.703125,
      "learning_rate": 8.474842767295597e-07,
      "loss": 0.1795,
      "reward": 0.6385513544082642,
      "reward_std": 0.16201703995466232,
      "reward_std_all": 0.16251954436302185,
      "rewards-max/forensics_sinkhorn_reward": 0.9876909852027893,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9948805570602417,
      "rewards/forensics_sinkhorn_reward": 0.4747690260410309,
      "rewards/forensics_strict_edge_reward": 0.23226885497570038,
      "rewards/forensics_tiou_reward": 0.33982083201408386,
      "step": 292
    },
    {
      "completion_length": 36.390625,
      "epoch": 0.6136125654450262,
      "grad_norm": 12.419065475463867,
      "kl": 1.8359375,
      "learning_rate": 8.469601677148846e-07,
      "loss": 0.2644,
      "reward": 0.586426168680191,
      "reward_std": 0.10653099417686462,
      "reward_std_all": 0.15716876089572906,
      "rewards-max/forensics_sinkhorn_reward": 0.9489886164665222,
      "rewards-max/forensics_strict_edge_reward": 0.6567562222480774,
      "rewards-max/forensics_tiou_reward": 0.9135653078556061,
      "rewards/forensics_sinkhorn_reward": 0.3909057080745697,
      "rewards/forensics_strict_edge_reward": 0.19659779220819473,
      "rewards/forensics_tiou_reward": 0.27421835064888,
      "step": 293
    },
    {
      "completion_length": 36.03125,
      "epoch": 0.6157068062827226,
      "grad_norm": 22.694412231445312,
      "kl": 2.3828125,
      "learning_rate": 8.464360587002096e-07,
      "loss": 0.1629,
      "reward": 0.6592994928359985,
      "reward_std": 0.12499300390481949,
      "reward_std_all": 0.13701851293444633,
      "rewards-max/forensics_sinkhorn_reward": 0.9885258972644806,
      "rewards-max/forensics_strict_edge_reward": 0.9782643616199493,
      "rewards-max/forensics_tiou_reward": 0.9981701672077179,
      "rewards/forensics_sinkhorn_reward": 0.4753851592540741,
      "rewards/forensics_strict_edge_reward": 0.2739330381155014,
      "rewards/forensics_tiou_reward": 0.2898665815591812,
      "step": 294
    },
    {
      "completion_length": 28.21875,
      "epoch": 0.6178010471204188,
      "grad_norm": 10.609951972961426,
      "kl": 1.171875,
      "learning_rate": 8.459119496855346e-07,
      "loss": 0.1977,
      "reward": 0.6394460797309875,
      "reward_std": 0.1688079982995987,
      "reward_std_all": 0.16488692164421082,
      "rewards-max/forensics_sinkhorn_reward": 0.9745092391967773,
      "rewards-max/forensics_strict_edge_reward": 0.7999143600463867,
      "rewards-max/forensics_tiou_reward": 0.9446768760681152,
      "rewards/forensics_sinkhorn_reward": 0.5028135776519775,
      "rewards/forensics_strict_edge_reward": 0.1560274101793766,
      "rewards/forensics_tiou_reward": 0.31330378353595734,
      "step": 295
    },
    {
      "completion_length": 28.671875,
      "epoch": 0.6198952879581152,
      "grad_norm": 10.769983291625977,
      "kl": 1.71875,
      "learning_rate": 8.453878406708595e-07,
      "loss": 0.1738,
      "reward": 0.6361522376537323,
      "reward_std": 0.146087396889925,
      "reward_std_all": 0.15023143589496613,
      "rewards-max/forensics_sinkhorn_reward": 0.9463672637939453,
      "rewards-max/forensics_strict_edge_reward": 0.8188429176807404,
      "rewards-max/forensics_tiou_reward": 0.9077328443527222,
      "rewards/forensics_sinkhorn_reward": 0.457528680562973,
      "rewards/forensics_strict_edge_reward": 0.2314934879541397,
      "rewards/forensics_tiou_reward": 0.2853435352444649,
      "step": 296
    },
    {
      "completion_length": 35.9375,
      "epoch": 0.6219895287958115,
      "grad_norm": 13.035636901855469,
      "kl": 1.80859375,
      "learning_rate": 8.448637316561845e-07,
      "loss": 0.3286,
      "reward": 0.6005221009254456,
      "reward_std": 0.15956690162420273,
      "reward_std_all": 0.17239175736904144,
      "rewards-max/forensics_sinkhorn_reward": 0.7756564319133759,
      "rewards-max/forensics_strict_edge_reward": 0.7262495160102844,
      "rewards-max/forensics_tiou_reward": 0.8193676769733429,
      "rewards/forensics_sinkhorn_reward": 0.40421321988105774,
      "rewards/forensics_strict_edge_reward": 0.14691805839538574,
      "rewards/forensics_tiou_reward": 0.2239075005054474,
      "step": 297
    },
    {
      "completion_length": 22.015625,
      "epoch": 0.6240837696335079,
      "grad_norm": 10.5513916015625,
      "kl": 1.65625,
      "learning_rate": 8.443396226415094e-07,
      "loss": 0.3276,
      "reward": 0.6044454574584961,
      "reward_std": 0.12858325242996216,
      "reward_std_all": 0.15711527317762375,
      "rewards-max/forensics_sinkhorn_reward": 0.9162665009498596,
      "rewards-max/forensics_strict_edge_reward": 0.6729699671268463,
      "rewards-max/forensics_tiou_reward": 0.8349570631980896,
      "rewards/forensics_sinkhorn_reward": 0.40859512984752655,
      "rewards/forensics_strict_edge_reward": 0.15961983799934387,
      "rewards/forensics_tiou_reward": 0.23051615804433823,
      "step": 298
    },
    {
      "completion_length": 22.4375,
      "epoch": 0.6261780104712041,
      "grad_norm": 13.682418823242188,
      "kl": 1.88671875,
      "learning_rate": 8.438155136268344e-07,
      "loss": 0.2763,
      "reward": 0.5844536125659943,
      "reward_std": 0.16211523115634918,
      "reward_std_all": 0.18415240198373795,
      "rewards-max/forensics_sinkhorn_reward": 0.9264555871486664,
      "rewards-max/forensics_strict_edge_reward": 0.8320213258266449,
      "rewards-max/forensics_tiou_reward": 0.8638767898082733,
      "rewards/forensics_sinkhorn_reward": 0.3944874554872513,
      "rewards/forensics_strict_edge_reward": 0.22390096634626389,
      "rewards/forensics_tiou_reward": 0.2512965351343155,
      "step": 299
    },
    {
      "completion_length": 39.46875,
      "epoch": 0.6282722513089005,
      "grad_norm": 11.495688438415527,
      "kl": 1.5078125,
      "learning_rate": 8.432914046121593e-07,
      "loss": 0.2439,
      "reward": 0.6237058043479919,
      "reward_std": 0.17360582947731018,
      "reward_std_all": 0.17153603583574295,
      "rewards-max/forensics_sinkhorn_reward": 0.9832939207553864,
      "rewards-max/forensics_strict_edge_reward": 0.8135445415973663,
      "rewards-max/forensics_tiou_reward": 0.9778157770633698,
      "rewards/forensics_sinkhorn_reward": 0.45630139112472534,
      "rewards/forensics_strict_edge_reward": 0.2413168102502823,
      "rewards/forensics_tiou_reward": 0.3071426451206207,
      "step": 300
    },
    {
      "completion_length": 34.6875,
      "epoch": 0.6303664921465969,
      "grad_norm": 11.75849723815918,
      "kl": 1.43359375,
      "learning_rate": 8.427672955974843e-07,
      "loss": 0.3065,
      "reward": 0.6313492655754089,
      "reward_std": 0.15298929810523987,
      "reward_std_all": 0.16575203090906143,
      "rewards-max/forensics_sinkhorn_reward": 0.9859493970870972,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9905534982681274,
      "rewards/forensics_sinkhorn_reward": 0.4296376407146454,
      "rewards/forensics_strict_edge_reward": 0.2182590514421463,
      "rewards/forensics_tiou_reward": 0.29721082746982574,
      "step": 301
    },
    {
      "completion_length": 29.71875,
      "epoch": 0.6324607329842932,
      "grad_norm": 10.242378234863281,
      "kl": 1.400390625,
      "learning_rate": 8.422431865828092e-07,
      "loss": 0.2456,
      "reward": 0.6500298380851746,
      "reward_std": 0.15093037113547325,
      "reward_std_all": 0.15182192251086235,
      "rewards-max/forensics_sinkhorn_reward": 0.9808512926101685,
      "rewards-max/forensics_strict_edge_reward": 0.8001649975776672,
      "rewards-max/forensics_tiou_reward": 0.9682631492614746,
      "rewards/forensics_sinkhorn_reward": 0.5018243789672852,
      "rewards/forensics_strict_edge_reward": 0.2227749079465866,
      "rewards/forensics_tiou_reward": 0.30660609900951385,
      "step": 302
    },
    {
      "completion_length": 24.59375,
      "epoch": 0.6345549738219896,
      "grad_norm": 12.208127975463867,
      "kl": 1.0703125,
      "learning_rate": 8.417190775681341e-07,
      "loss": 0.308,
      "reward": 0.6070585250854492,
      "reward_std": 0.1797003298997879,
      "reward_std_all": 0.18008224666118622,
      "rewards-max/forensics_sinkhorn_reward": 0.9884564280509949,
      "rewards-max/forensics_strict_edge_reward": 0.9574736058712006,
      "rewards-max/forensics_tiou_reward": 0.9978947341442108,
      "rewards/forensics_sinkhorn_reward": 0.4540724456310272,
      "rewards/forensics_strict_edge_reward": 0.23081039637327194,
      "rewards/forensics_tiou_reward": 0.2879806458950043,
      "step": 303
    },
    {
      "completion_length": 21.546875,
      "epoch": 0.6366492146596858,
      "grad_norm": 18.55626106262207,
      "kl": 2.33984375,
      "learning_rate": 8.411949685534591e-07,
      "loss": 0.3417,
      "reward": 0.6566184163093567,
      "reward_std": 0.15055520087480545,
      "reward_std_all": 0.15188030153512955,
      "rewards-max/forensics_sinkhorn_reward": 0.8373124599456787,
      "rewards-max/forensics_strict_edge_reward": 0.6579880714416504,
      "rewards-max/forensics_tiou_reward": 0.6842149943113327,
      "rewards/forensics_sinkhorn_reward": 0.5163719803094864,
      "rewards/forensics_strict_edge_reward": 0.22506630048155785,
      "rewards/forensics_tiou_reward": 0.30117037892341614,
      "step": 304
    },
    {
      "completion_length": 21.84375,
      "epoch": 0.6387434554973822,
      "grad_norm": 11.127751350402832,
      "kl": 1.171875,
      "learning_rate": 8.40670859538784e-07,
      "loss": 0.1018,
      "reward": 0.6478215456008911,
      "reward_std": 0.18582574278116226,
      "reward_std_all": 0.17662687599658966,
      "rewards-max/forensics_sinkhorn_reward": 0.9299340844154358,
      "rewards-max/forensics_strict_edge_reward": 0.791724443435669,
      "rewards-max/forensics_tiou_reward": 0.8660948872566223,
      "rewards/forensics_sinkhorn_reward": 0.4714515209197998,
      "rewards/forensics_strict_edge_reward": 0.23555409163236618,
      "rewards/forensics_tiou_reward": 0.29671522974967957,
      "step": 305
    },
    {
      "completion_length": 34.609375,
      "epoch": 0.6408376963350786,
      "grad_norm": 11.361469268798828,
      "kl": 1.189453125,
      "learning_rate": 8.40146750524109e-07,
      "loss": 0.1681,
      "reward": 0.6157678663730621,
      "reward_std": 0.18355891108512878,
      "reward_std_all": 0.18823711574077606,
      "rewards-max/forensics_sinkhorn_reward": 0.9106490612030029,
      "rewards-max/forensics_strict_edge_reward": 0.8134243786334991,
      "rewards-max/forensics_tiou_reward": 0.8354385793209076,
      "rewards/forensics_sinkhorn_reward": 0.4838225543498993,
      "rewards/forensics_strict_edge_reward": 0.19563017040491104,
      "rewards/forensics_tiou_reward": 0.22248295694589615,
      "step": 306
    },
    {
      "completion_length": 25.125,
      "epoch": 0.6429319371727749,
      "grad_norm": 11.085064888000488,
      "kl": 2.171875,
      "learning_rate": 8.396226415094339e-07,
      "loss": 0.2559,
      "reward": 0.6546761691570282,
      "reward_std": 0.14019610360264778,
      "reward_std_all": 0.14324494823813438,
      "rewards-max/forensics_sinkhorn_reward": 0.967182070016861,
      "rewards-max/forensics_strict_edge_reward": 0.8751733303070068,
      "rewards-max/forensics_tiou_reward": 0.949889212846756,
      "rewards/forensics_sinkhorn_reward": 0.47133585810661316,
      "rewards/forensics_strict_edge_reward": 0.2033293955028057,
      "rewards/forensics_tiou_reward": 0.26768988370895386,
      "step": 307
    },
    {
      "completion_length": 38.96875,
      "epoch": 0.6450261780104712,
      "grad_norm": 10.753460884094238,
      "kl": 1.41796875,
      "learning_rate": 8.390985324947589e-07,
      "loss": 0.1897,
      "reward": 0.6255233883857727,
      "reward_std": 0.1254844330251217,
      "reward_std_all": 0.14623093605041504,
      "rewards-max/forensics_sinkhorn_reward": 0.9350081980228424,
      "rewards-max/forensics_strict_edge_reward": 0.8379553556442261,
      "rewards-max/forensics_tiou_reward": 0.8831852972507477,
      "rewards/forensics_sinkhorn_reward": 0.38802020251750946,
      "rewards/forensics_strict_edge_reward": 0.19054757803678513,
      "rewards/forensics_tiou_reward": 0.20080795139074326,
      "step": 308
    },
    {
      "completion_length": 47.40625,
      "epoch": 0.6471204188481675,
      "grad_norm": 13.026747703552246,
      "kl": 1.6796875,
      "learning_rate": 8.385744234800837e-07,
      "loss": 0.0612,
      "reward": 0.5627466142177582,
      "reward_std": 0.18347804993391037,
      "reward_std_all": 0.20451276749372482,
      "rewards-max/forensics_sinkhorn_reward": 0.9452140629291534,
      "rewards-max/forensics_strict_edge_reward": 0.8286563754081726,
      "rewards-max/forensics_tiou_reward": 0.9085731208324432,
      "rewards/forensics_sinkhorn_reward": 0.4357715994119644,
      "rewards/forensics_strict_edge_reward": 0.19164101034402847,
      "rewards/forensics_tiou_reward": 0.22344226390123367,
      "step": 309
    },
    {
      "completion_length": 24.71875,
      "epoch": 0.6492146596858639,
      "grad_norm": 13.615958213806152,
      "kl": 1.56640625,
      "learning_rate": 8.380503144654088e-07,
      "loss": 0.2978,
      "reward": 0.648348867893219,
      "reward_std": 0.1347706839442253,
      "reward_std_all": 0.14671073853969574,
      "rewards-max/forensics_sinkhorn_reward": 0.9855837225914001,
      "rewards-max/forensics_strict_edge_reward": 0.9677534997463226,
      "rewards-max/forensics_tiou_reward": 0.9871913492679596,
      "rewards/forensics_sinkhorn_reward": 0.5265730023384094,
      "rewards/forensics_strict_edge_reward": 0.2690759003162384,
      "rewards/forensics_tiou_reward": 0.33729904890060425,
      "step": 310
    },
    {
      "completion_length": 34.0,
      "epoch": 0.6513089005235602,
      "grad_norm": 11.213601112365723,
      "kl": 1.052734375,
      "learning_rate": 8.375262054507337e-07,
      "loss": 0.1348,
      "reward": 0.6501199901103973,
      "reward_std": 0.11463744193315506,
      "reward_std_all": 0.13520461693406105,
      "rewards-max/forensics_sinkhorn_reward": 0.8382217884063721,
      "rewards-max/forensics_strict_edge_reward": 0.7279628366231918,
      "rewards-max/forensics_tiou_reward": 0.7491317391395569,
      "rewards/forensics_sinkhorn_reward": 0.48587189614772797,
      "rewards/forensics_strict_edge_reward": 0.1885455921292305,
      "rewards/forensics_tiou_reward": 0.26118776202201843,
      "step": 311
    },
    {
      "completion_length": 34.703125,
      "epoch": 0.6534031413612565,
      "grad_norm": 11.01281452178955,
      "kl": 0.9765625,
      "learning_rate": 8.370020964360587e-07,
      "loss": 0.0667,
      "reward": 0.7090891599655151,
      "reward_std": 0.09398872777819633,
      "reward_std_all": 0.09684497490525246,
      "rewards-max/forensics_sinkhorn_reward": 0.9869766533374786,
      "rewards-max/forensics_strict_edge_reward": 0.9677534997463226,
      "rewards-max/forensics_tiou_reward": 0.9920949935913086,
      "rewards/forensics_sinkhorn_reward": 0.6216415464878082,
      "rewards/forensics_strict_edge_reward": 0.3124738186597824,
      "rewards/forensics_tiou_reward": 0.39212140440940857,
      "step": 312
    },
    {
      "completion_length": 28.03125,
      "epoch": 0.6554973821989529,
      "grad_norm": 9.905401229858398,
      "kl": 1.13671875,
      "learning_rate": 8.364779874213837e-07,
      "loss": 0.0685,
      "reward": 0.6390778124332428,
      "reward_std": 0.16280357539653778,
      "reward_std_all": 0.16493458300828934,
      "rewards-max/forensics_sinkhorn_reward": 0.9885061681270599,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9980913996696472,
      "rewards/forensics_sinkhorn_reward": 0.5259237587451935,
      "rewards/forensics_strict_edge_reward": 0.27337467670440674,
      "rewards/forensics_tiou_reward": 0.35315023362636566,
      "step": 313
    },
    {
      "completion_length": 26.328125,
      "epoch": 0.6575916230366492,
      "grad_norm": 13.805585861206055,
      "kl": 1.33203125,
      "learning_rate": 8.359538784067086e-07,
      "loss": 0.0714,
      "reward": 0.6151461601257324,
      "reward_std": 0.13134340569376945,
      "reward_std_all": 0.15143047273159027,
      "rewards-max/forensics_sinkhorn_reward": 0.9360683560371399,
      "rewards-max/forensics_strict_edge_reward": 0.7022436261177063,
      "rewards-max/forensics_tiou_reward": 0.8726310729980469,
      "rewards/forensics_sinkhorn_reward": 0.43656817078590393,
      "rewards/forensics_strict_edge_reward": 0.14269039779901505,
      "rewards/forensics_tiou_reward": 0.18072039633989334,
      "step": 314
    },
    {
      "completion_length": 50.640625,
      "epoch": 0.6596858638743456,
      "grad_norm": 11.234758377075195,
      "kl": 1.30859375,
      "learning_rate": 8.354297693920335e-07,
      "loss": 0.1731,
      "reward": 0.6719059348106384,
      "reward_std": 0.1559799760580063,
      "reward_std_all": 0.14765460789203644,
      "rewards-max/forensics_sinkhorn_reward": 0.9874365925788879,
      "rewards-max/forensics_strict_edge_reward": 0.9567649364471436,
      "rewards-max/forensics_tiou_reward": 0.9938945472240448,
      "rewards/forensics_sinkhorn_reward": 0.5062593817710876,
      "rewards/forensics_strict_edge_reward": 0.3201345056295395,
      "rewards/forensics_tiou_reward": 0.33601778745651245,
      "step": 315
    },
    {
      "completion_length": 41.078125,
      "epoch": 0.6617801047120419,
      "grad_norm": 14.176000595092773,
      "kl": 1.63671875,
      "learning_rate": 8.349056603773584e-07,
      "loss": 0.1525,
      "reward": 0.6339193880558014,
      "reward_std": 0.147315613925457,
      "reward_std_all": 0.15438074618577957,
      "rewards-max/forensics_sinkhorn_reward": 0.9357033967971802,
      "rewards-max/forensics_strict_edge_reward": 0.8060585260391235,
      "rewards-max/forensics_tiou_reward": 0.8818854987621307,
      "rewards/forensics_sinkhorn_reward": 0.47044341266155243,
      "rewards/forensics_strict_edge_reward": 0.2766024172306061,
      "rewards/forensics_tiou_reward": 0.28773564100265503,
      "step": 316
    },
    {
      "completion_length": 51.140625,
      "epoch": 0.6638743455497382,
      "grad_norm": 9.891611099243164,
      "kl": 1.2265625,
      "learning_rate": 8.343815513626834e-07,
      "loss": 0.1548,
      "reward": 0.5713323354721069,
      "reward_std": 0.1680595576763153,
      "reward_std_all": 0.19377556443214417,
      "rewards-max/forensics_sinkhorn_reward": 0.985872894525528,
      "rewards-max/forensics_strict_edge_reward": 0.9364311099052429,
      "rewards-max/forensics_tiou_reward": 0.9877684116363525,
      "rewards/forensics_sinkhorn_reward": 0.4058777689933777,
      "rewards/forensics_strict_edge_reward": 0.16238753497600555,
      "rewards/forensics_tiou_reward": 0.23026210069656372,
      "step": 317
    },
    {
      "completion_length": 52.609375,
      "epoch": 0.6659685863874345,
      "grad_norm": 12.283357620239258,
      "kl": 1.3125,
      "learning_rate": 8.338574423480083e-07,
      "loss": 0.2308,
      "reward": 0.629082977771759,
      "reward_std": 0.1444995403289795,
      "reward_std_all": 0.15690943598747253,
      "rewards-max/forensics_sinkhorn_reward": 0.9028938710689545,
      "rewards-max/forensics_strict_edge_reward": 0.8205901980400085,
      "rewards-max/forensics_tiou_reward": 0.8202614486217499,
      "rewards/forensics_sinkhorn_reward": 0.4797711819410324,
      "rewards/forensics_strict_edge_reward": 0.21407655999064445,
      "rewards/forensics_tiou_reward": 0.2697145491838455,
      "step": 318
    },
    {
      "completion_length": 62.4375,
      "epoch": 0.6680628272251309,
      "grad_norm": 10.668731689453125,
      "kl": 1.3671875,
      "learning_rate": 8.333333333333333e-07,
      "loss": 0.2384,
      "reward": 0.6684599220752716,
      "reward_std": 0.1565651074051857,
      "reward_std_all": 0.1481100618839264,
      "rewards-max/forensics_sinkhorn_reward": 0.9884341657161713,
      "rewards-max/forensics_strict_edge_reward": 0.9677534997463226,
      "rewards-max/forensics_tiou_reward": 0.9978094100952148,
      "rewards/forensics_sinkhorn_reward": 0.5069746375083923,
      "rewards/forensics_strict_edge_reward": 0.35392361879348755,
      "rewards/forensics_tiou_reward": 0.3841068148612976,
      "step": 319
    },
    {
      "completion_length": 70.578125,
      "epoch": 0.6701570680628273,
      "grad_norm": 11.773828506469727,
      "kl": 1.53125,
      "learning_rate": 8.328092243186582e-07,
      "loss": 0.2799,
      "reward": 0.5666390359401703,
      "reward_std": 0.14299309626221657,
      "reward_std_all": 0.1813809648156166,
      "rewards-max/forensics_sinkhorn_reward": 0.9870813488960266,
      "rewards-max/forensics_strict_edge_reward": 0.878633052110672,
      "rewards-max/forensics_tiou_reward": 0.9924830198287964,
      "rewards/forensics_sinkhorn_reward": 0.370454341173172,
      "rewards/forensics_strict_edge_reward": 0.19910068809986115,
      "rewards/forensics_tiou_reward": 0.2689244747161865,
      "step": 320
    },
    {
      "completion_length": 68.296875,
      "epoch": 0.6722513089005235,
      "grad_norm": 9.911869049072266,
      "kl": 1.3984375,
      "learning_rate": 8.322851153039831e-07,
      "loss": 0.3214,
      "reward": 0.6104362905025482,
      "reward_std": 0.15903203189373016,
      "reward_std_all": 0.16692981868982315,
      "rewards-max/forensics_sinkhorn_reward": 0.9870246350765228,
      "rewards-max/forensics_strict_edge_reward": 0.9364311099052429,
      "rewards-max/forensics_tiou_reward": 0.9922692775726318,
      "rewards/forensics_sinkhorn_reward": 0.416427344083786,
      "rewards/forensics_strict_edge_reward": 0.17043493688106537,
      "rewards/forensics_tiou_reward": 0.2582673504948616,
      "step": 321
    },
    {
      "completion_length": 45.578125,
      "epoch": 0.6743455497382199,
      "grad_norm": 11.494288444519043,
      "kl": 1.52734375,
      "learning_rate": 8.317610062893082e-07,
      "loss": 0.3541,
      "reward": 0.5651996731758118,
      "reward_std": 0.1799856424331665,
      "reward_std_all": 0.20378424227237701,
      "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": 0.4362475126981735,
      "rewards/forensics_strict_edge_reward": 0.2120976522564888,
      "rewards/forensics_tiou_reward": 0.3011540621519089,
      "step": 322
    },
    {
      "completion_length": 68.921875,
      "epoch": 0.6764397905759162,
      "grad_norm": 8.846829414367676,
      "kl": 1.421875,
      "learning_rate": 8.312368972746331e-07,
      "loss": 0.3263,
      "reward": 0.6056540310382843,
      "reward_std": 0.20268651098012924,
      "reward_std_all": 0.19853726774454117,
      "rewards-max/forensics_sinkhorn_reward": 0.9371559023857117,
      "rewards-max/forensics_strict_edge_reward": 0.7922165393829346,
      "rewards-max/forensics_tiou_reward": 0.8201750218868256,
      "rewards/forensics_sinkhorn_reward": 0.39998146891593933,
      "rewards/forensics_strict_edge_reward": 0.23288138210773468,
      "rewards/forensics_tiou_reward": 0.2834782600402832,
      "step": 323
    },
    {
      "completion_length": 50.609375,
      "epoch": 0.6785340314136126,
      "grad_norm": 260.0020751953125,
      "kl": 3.83203125,
      "learning_rate": 8.307127882599581e-07,
      "loss": 0.3608,
      "reward": 0.5562794208526611,
      "reward_std": 0.17796827852725983,
      "reward_std_all": 0.20181817561388016,
      "rewards-max/forensics_sinkhorn_reward": 0.980329304933548,
      "rewards-max/forensics_strict_edge_reward": 0.856150895357132,
      "rewards-max/forensics_tiou_reward": 0.9662842750549316,
      "rewards/forensics_sinkhorn_reward": 0.40313537418842316,
      "rewards/forensics_strict_edge_reward": 0.1753963902592659,
      "rewards/forensics_tiou_reward": 0.26544760167598724,
      "step": 324
    },
    {
      "completion_length": 49.3125,
      "epoch": 0.680628272251309,
      "grad_norm": 8.619097709655762,
      "kl": 1.5078125,
      "learning_rate": 8.30188679245283e-07,
      "loss": 0.1047,
      "reward": 0.6347113251686096,
      "reward_std": 0.15059489011764526,
      "reward_std_all": 0.1597696766257286,
      "rewards-max/forensics_sinkhorn_reward": 0.9866624474525452,
      "rewards-max/forensics_strict_edge_reward": 0.947419673204422,
      "rewards-max/forensics_tiou_reward": 0.9908256828784943,
      "rewards/forensics_sinkhorn_reward": 0.40296047925949097,
      "rewards/forensics_strict_edge_reward": 0.2619842141866684,
      "rewards/forensics_tiou_reward": 0.295825332403183,
      "step": 325
    },
    {
      "completion_length": 55.640625,
      "epoch": 0.6827225130890052,
      "grad_norm": 10.057393074035645,
      "kl": 1.6015625,
      "learning_rate": 8.296645702306079e-07,
      "loss": 0.2825,
      "reward": 0.5926906764507294,
      "reward_std": 0.19028902053833008,
      "reward_std_all": 0.19622642546892166,
      "rewards-max/forensics_sinkhorn_reward": 0.9884204268455505,
      "rewards-max/forensics_strict_edge_reward": 0.9890114367008209,
      "rewards-max/forensics_tiou_reward": 0.9977528154850006,
      "rewards/forensics_sinkhorn_reward": 0.43859605491161346,
      "rewards/forensics_strict_edge_reward": 0.26636429131031036,
      "rewards/forensics_tiou_reward": 0.36081811785697937,
      "step": 326
    },
    {
      "completion_length": 45.0,
      "epoch": 0.6848167539267016,
      "grad_norm": 12.19591999053955,
      "kl": 1.267578125,
      "learning_rate": 8.291404612159328e-07,
      "loss": 0.1687,
      "reward": 0.5867299735546112,
      "reward_std": 0.16641448438167572,
      "reward_std_all": 0.18745188415050507,
      "rewards-max/forensics_sinkhorn_reward": 0.9459587931632996,
      "rewards-max/forensics_strict_edge_reward": 0.8024423122406006,
      "rewards-max/forensics_tiou_reward": 0.904501348733902,
      "rewards/forensics_sinkhorn_reward": 0.40555377304553986,
      "rewards/forensics_strict_edge_reward": 0.24433790147304535,
      "rewards/forensics_tiou_reward": 0.28374814987182617,
      "step": 327
    },
    {
      "completion_length": 49.953125,
      "epoch": 0.6869109947643979,
      "grad_norm": 10.15962028503418,
      "kl": 1.1953125,
      "learning_rate": 8.286163522012578e-07,
      "loss": 0.1822,
      "reward": 0.6508253216743469,
      "reward_std": 0.16944467276334763,
      "reward_std_all": 0.1621062010526657,
      "rewards-max/forensics_sinkhorn_reward": 0.9875766634941101,
      "rewards-max/forensics_strict_edge_reward": 0.9464850425720215,
      "rewards-max/forensics_tiou_reward": 0.9944295287132263,
      "rewards/forensics_sinkhorn_reward": 0.481789767742157,
      "rewards/forensics_strict_edge_reward": 0.2972753643989563,
      "rewards/forensics_tiou_reward": 0.32642681896686554,
      "step": 328
    },
    {
      "completion_length": 30.890625,
      "epoch": 0.6890052356020943,
      "grad_norm": 13.113296508789062,
      "kl": 1.3828125,
      "learning_rate": 8.280922431865828e-07,
      "loss": 0.2561,
      "reward": 0.565260112285614,
      "reward_std": 0.23312091827392578,
      "reward_std_all": 0.23413408547639847,
      "rewards-max/forensics_sinkhorn_reward": 0.9881238043308258,
      "rewards-max/forensics_strict_edge_reward": 0.9565287232398987,
      "rewards-max/forensics_tiou_reward": 0.9965837597846985,
      "rewards/forensics_sinkhorn_reward": 0.43342722952365875,
      "rewards/forensics_strict_edge_reward": 0.16896521300077438,
      "rewards/forensics_tiou_reward": 0.23357029259204865,
      "step": 329
    },
    {
      "completion_length": 29.359375,
      "epoch": 0.6910994764397905,
      "grad_norm": 9.085138320922852,
      "kl": 1.16015625,
      "learning_rate": 8.275681341719077e-07,
      "loss": 0.1188,
      "reward": 0.6775088906288147,
      "reward_std": 0.15469497442245483,
      "reward_std_all": 0.1448800340294838,
      "rewards-max/forensics_sinkhorn_reward": 0.9767748117446899,
      "rewards-max/forensics_strict_edge_reward": 0.9256840348243713,
      "rewards-max/forensics_tiou_reward": 0.9714503884315491,
      "rewards/forensics_sinkhorn_reward": 0.5496282875537872,
      "rewards/forensics_strict_edge_reward": 0.284729927778244,
      "rewards/forensics_tiou_reward": 0.39477644860744476,
      "step": 330
    },
    {
      "completion_length": 34.859375,
      "epoch": 0.6931937172774869,
      "grad_norm": 14.305912971496582,
      "kl": 1.5078125,
      "learning_rate": 8.270440251572327e-07,
      "loss": 0.141,
      "reward": 0.6531115472316742,
      "reward_std": 0.06764721684157848,
      "reward_std_all": 0.10780731961131096,
      "rewards-max/forensics_sinkhorn_reward": 0.9188973009586334,
      "rewards-max/forensics_strict_edge_reward": 0.7572441399097443,
      "rewards-max/forensics_tiou_reward": 0.8486867547035217,
      "rewards/forensics_sinkhorn_reward": 0.45915357768535614,
      "rewards/forensics_strict_edge_reward": 0.2167607918381691,
      "rewards/forensics_tiou_reward": 0.26037999987602234,
      "step": 331
    },
    {
      "completion_length": 51.28125,
      "epoch": 0.6952879581151833,
      "grad_norm": 11.575411796569824,
      "kl": 1.255859375,
      "learning_rate": 8.265199161425575e-07,
      "loss": 0.1525,
      "reward": 0.6912010610103607,
      "reward_std": 0.14460886642336845,
      "reward_std_all": 0.1343661993741989,
      "rewards-max/forensics_sinkhorn_reward": 0.9880426228046417,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.996263712644577,
      "rewards/forensics_sinkhorn_reward": 0.5592432916164398,
      "rewards/forensics_strict_edge_reward": 0.3190319091081619,
      "rewards/forensics_tiou_reward": 0.3895019292831421,
      "step": 332
    },
    {
      "completion_length": 19.515625,
      "epoch": 0.6973821989528796,
      "grad_norm": 9.496267318725586,
      "kl": 0.6572265625,
      "learning_rate": 8.259958071278825e-07,
      "loss": 0.0527,
      "reward": 0.6650302410125732,
      "reward_std": 0.08911880664527416,
      "reward_std_all": 0.1139908954501152,
      "rewards-max/forensics_sinkhorn_reward": 0.9840784072875977,
      "rewards-max/forensics_strict_edge_reward": 0.8957232534885406,
      "rewards-max/forensics_tiou_reward": 0.9807779788970947,
      "rewards/forensics_sinkhorn_reward": 0.5626554191112518,
      "rewards/forensics_strict_edge_reward": 0.15911422669887543,
      "rewards/forensics_tiou_reward": 0.2884412109851837,
      "step": 333
    },
    {
      "completion_length": 61.921875,
      "epoch": 0.6994764397905759,
      "grad_norm": 10.805427551269531,
      "kl": 1.33203125,
      "learning_rate": 8.254716981132074e-07,
      "loss": 0.2432,
      "reward": 0.581198513507843,
      "reward_std": 0.23708177357912064,
      "reward_std_all": 0.22914646565914154,
      "rewards-max/forensics_sinkhorn_reward": 0.9866235554218292,
      "rewards-max/forensics_strict_edge_reward": 0.9256840348243713,
      "rewards-max/forensics_tiou_reward": 0.9906717240810394,
      "rewards/forensics_sinkhorn_reward": 0.44413644075393677,
      "rewards/forensics_strict_edge_reward": 0.22246141731739044,
      "rewards/forensics_tiou_reward": 0.2627488076686859,
      "step": 334
    },
    {
      "completion_length": 59.953125,
      "epoch": 0.7015706806282722,
      "grad_norm": 11.060540199279785,
      "kl": 1.703125,
      "learning_rate": 8.249475890985325e-07,
      "loss": 0.2673,
      "reward": 0.6581838726997375,
      "reward_std": 0.11465683951973915,
      "reward_std_all": 0.1305445246398449,
      "rewards-max/forensics_sinkhorn_reward": 0.9871780872344971,
      "rewards-max/forensics_strict_edge_reward": 0.9375866651535034,
      "rewards-max/forensics_tiou_reward": 0.9928571283817291,
      "rewards/forensics_sinkhorn_reward": 0.5032338052988052,
      "rewards/forensics_strict_edge_reward": 0.21900804340839386,
      "rewards/forensics_tiou_reward": 0.3220863491296768,
      "step": 335
    },
    {
      "completion_length": 26.21875,
      "epoch": 0.7036649214659686,
      "grad_norm": 14.131625175476074,
      "kl": 1.84375,
      "learning_rate": 8.244234800838575e-07,
      "loss": 0.2725,
      "reward": 0.583145946264267,
      "reward_std": 0.20394833385944366,
      "reward_std_all": 0.2094801440834999,
      "rewards-max/forensics_sinkhorn_reward": 0.9045833647251129,
      "rewards-max/forensics_strict_edge_reward": 0.7570423185825348,
      "rewards-max/forensics_tiou_reward": 0.7306034564971924,
      "rewards/forensics_sinkhorn_reward": 0.4570252448320389,
      "rewards/forensics_strict_edge_reward": 0.21467789635062218,
      "rewards/forensics_tiou_reward": 0.2885879836976528,
      "step": 336
    },
    {
      "completion_length": 17.609375,
      "epoch": 0.7057591623036649,
      "grad_norm": 13.265364646911621,
      "kl": 1.17578125,
      "learning_rate": 8.238993710691824e-07,
      "loss": 0.0036,
      "reward": 0.6668352484703064,
      "reward_std": 0.10940390452742577,
      "reward_std_all": 0.12565382942557335,
      "rewards-max/forensics_sinkhorn_reward": 0.9875120520591736,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9941735863685608,
      "rewards/forensics_sinkhorn_reward": 0.5595946311950684,
      "rewards/forensics_strict_edge_reward": 0.20865389704704285,
      "rewards/forensics_tiou_reward": 0.30905769765377045,
      "step": 337
    },
    {
      "completion_length": 32.078125,
      "epoch": 0.7078534031413612,
      "grad_norm": 12.71338939666748,
      "kl": 1.23828125,
      "learning_rate": 8.233752620545073e-07,
      "loss": 0.1386,
      "reward": 0.6521628797054291,
      "reward_std": 0.13382277637720108,
      "reward_std_all": 0.13917624950408936,
      "rewards-max/forensics_sinkhorn_reward": 0.9844887852668762,
      "rewards-max/forensics_strict_edge_reward": 0.9075756669044495,
      "rewards-max/forensics_tiou_reward": 0.9822653830051422,
      "rewards/forensics_sinkhorn_reward": 0.48030611872673035,
      "rewards/forensics_strict_edge_reward": 0.23724062740802765,
      "rewards/forensics_tiou_reward": 0.2885727882385254,
      "step": 338
    },
    {
      "completion_length": 20.703125,
      "epoch": 0.7099476439790576,
      "grad_norm": 10.575904846191406,
      "kl": 1.35546875,
      "learning_rate": 8.228511530398322e-07,
      "loss": 0.0639,
      "reward": 0.6423335671424866,
      "reward_std": 0.12574366107583046,
      "reward_std_all": 0.1384892500936985,
      "rewards-max/forensics_sinkhorn_reward": 0.9873711466789246,
      "rewards-max/forensics_strict_edge_reward": 0.9565287232398987,
      "rewards-max/forensics_tiou_reward": 0.9936269521713257,
      "rewards/forensics_sinkhorn_reward": 0.5337365865707397,
      "rewards/forensics_strict_edge_reward": 0.2431352734565735,
      "rewards/forensics_tiou_reward": 0.30017589032649994,
      "step": 339
    },
    {
      "completion_length": 43.890625,
      "epoch": 0.7120418848167539,
      "grad_norm": 10.34009838104248,
      "kl": 1.115234375,
      "learning_rate": 8.223270440251572e-07,
      "loss": 0.0466,
      "reward": 0.6904700696468353,
      "reward_std": 0.11565479636192322,
      "reward_std_all": 0.11412686854600906,
      "rewards-max/forensics_sinkhorn_reward": 0.9874200820922852,
      "rewards-max/forensics_strict_edge_reward": 0.8854433596134186,
      "rewards-max/forensics_tiou_reward": 0.9938137829303741,
      "rewards/forensics_sinkhorn_reward": 0.49137747287750244,
      "rewards/forensics_strict_edge_reward": 0.21812006831169128,
      "rewards/forensics_tiou_reward": 0.2790921926498413,
      "step": 340
    },
    {
      "completion_length": 87.09375,
      "epoch": 0.7141361256544503,
      "grad_norm": 12.096247673034668,
      "kl": 1.83203125,
      "learning_rate": 8.218029350104821e-07,
      "loss": 0.3794,
      "reward": 0.6767371594905853,
      "reward_std": 0.08775516971945763,
      "reward_std_all": 0.10374559089541435,
      "rewards-max/forensics_sinkhorn_reward": 0.9341912269592285,
      "rewards-max/forensics_strict_edge_reward": 0.87538942694664,
      "rewards-max/forensics_tiou_reward": 0.8723824918270111,
      "rewards/forensics_sinkhorn_reward": 0.49722884595394135,
      "rewards/forensics_strict_edge_reward": 0.19760426878929138,
      "rewards/forensics_tiou_reward": 0.3098302483558655,
      "step": 341
    },
    {
      "completion_length": 64.78125,
      "epoch": 0.7162303664921466,
      "grad_norm": 10.238228797912598,
      "kl": 2.125,
      "learning_rate": 8.212788259958071e-07,
      "loss": 0.1473,
      "reward": 0.609804093837738,
      "reward_std": 0.1565324366092682,
      "reward_std_all": 0.17629382014274597,
      "rewards-max/forensics_sinkhorn_reward": 0.8798550665378571,
      "rewards-max/forensics_strict_edge_reward": 0.7638483941555023,
      "rewards-max/forensics_tiou_reward": 0.7829224765300751,
      "rewards/forensics_sinkhorn_reward": 0.4499097764492035,
      "rewards/forensics_strict_edge_reward": 0.18239471688866615,
      "rewards/forensics_tiou_reward": 0.26424090564250946,
      "step": 342
    },
    {
      "completion_length": 71.5,
      "epoch": 0.7183246073298429,
      "grad_norm": 12.407578468322754,
      "kl": 1.96875,
      "learning_rate": 8.207547169811321e-07,
      "loss": 0.1276,
      "reward": 0.5981168150901794,
      "reward_std": 0.18241192400455475,
      "reward_std_all": 0.1916884407401085,
      "rewards-max/forensics_sinkhorn_reward": 0.9811357855796814,
      "rewards-max/forensics_strict_edge_reward": 0.9252271056175232,
      "rewards-max/forensics_tiou_reward": 0.9809703528881073,
      "rewards/forensics_sinkhorn_reward": 0.4507867693901062,
      "rewards/forensics_strict_edge_reward": 0.27442657947540283,
      "rewards/forensics_tiou_reward": 0.30012422800064087,
      "step": 343
    },
    {
      "completion_length": 84.40625,
      "epoch": 0.7204188481675393,
      "grad_norm": 33.25751495361328,
      "kl": 1.95703125,
      "learning_rate": 8.202306079664569e-07,
      "loss": 0.2025,
      "reward": 0.6189426183700562,
      "reward_std": 0.19074176996946335,
      "reward_std_all": 0.18909632414579391,
      "rewards-max/forensics_sinkhorn_reward": 0.950697273015976,
      "rewards-max/forensics_strict_edge_reward": 0.8032680749893188,
      "rewards-max/forensics_tiou_reward": 0.9192754328250885,
      "rewards/forensics_sinkhorn_reward": 0.43487051129341125,
      "rewards/forensics_strict_edge_reward": 0.24726645648479462,
      "rewards/forensics_tiou_reward": 0.3087089955806732,
      "step": 344
    },
    {
      "completion_length": 43.25,
      "epoch": 0.7225130890052356,
      "grad_norm": 14.576452255249023,
      "kl": 1.6015625,
      "learning_rate": 8.197064989517819e-07,
      "loss": 0.0384,
      "reward": 0.588630199432373,
      "reward_std": 0.16561466455459595,
      "reward_std_all": 0.18240901827812195,
      "rewards-max/forensics_sinkhorn_reward": 0.981472373008728,
      "rewards-max/forensics_strict_edge_reward": 0.8593232333660126,
      "rewards-max/forensics_tiou_reward": 0.9745890498161316,
      "rewards/forensics_sinkhorn_reward": 0.47799909114837646,
      "rewards/forensics_strict_edge_reward": 0.1714608445763588,
      "rewards/forensics_tiou_reward": 0.2821599394083023,
      "step": 345
    },
    {
      "completion_length": 54.4375,
      "epoch": 0.724607329842932,
      "grad_norm": 11.757678985595703,
      "kl": 1.76953125,
      "learning_rate": 8.191823899371068e-07,
      "loss": 0.1753,
      "reward": 0.5942659080028534,
      "reward_std": 0.15799015760421753,
      "reward_std_all": 0.1796722486615181,
      "rewards-max/forensics_sinkhorn_reward": 0.9331445097923279,
      "rewards-max/forensics_strict_edge_reward": 0.7335737943649292,
      "rewards-max/forensics_tiou_reward": 0.8777678608894348,
      "rewards/forensics_sinkhorn_reward": 0.41922469437122345,
      "rewards/forensics_strict_edge_reward": 0.17498427256941795,
      "rewards/forensics_tiou_reward": 0.2505720853805542,
      "step": 346
    },
    {
      "completion_length": 77.6875,
      "epoch": 0.7267015706806282,
      "grad_norm": 13.889348030090332,
      "kl": 1.828125,
      "learning_rate": 8.186582809224319e-07,
      "loss": 0.2924,
      "reward": 0.628637284040451,
      "reward_std": 0.12050345167517662,
      "reward_std_all": 0.15075158327817917,
      "rewards-max/forensics_sinkhorn_reward": 0.9684968590736389,
      "rewards-max/forensics_strict_edge_reward": 0.8472771048545837,
      "rewards-max/forensics_tiou_reward": 0.9516165554523468,
      "rewards/forensics_sinkhorn_reward": 0.473272368311882,
      "rewards/forensics_strict_edge_reward": 0.23037491738796234,
      "rewards/forensics_tiou_reward": 0.3202763646841049,
      "step": 347
    },
    {
      "completion_length": 77.421875,
      "epoch": 0.7287958115183246,
      "grad_norm": 17.61132049560547,
      "kl": 2.25390625,
      "learning_rate": 8.181341719077568e-07,
      "loss": 0.2137,
      "reward": 0.5748191475868225,
      "reward_std": 0.1836763173341751,
      "reward_std_all": 0.2058667242527008,
      "rewards-max/forensics_sinkhorn_reward": 0.9718803763389587,
      "rewards-max/forensics_strict_edge_reward": 0.9256840348243713,
      "rewards-max/forensics_tiou_reward": 0.9615462422370911,
      "rewards/forensics_sinkhorn_reward": 0.4182852506637573,
      "rewards/forensics_strict_edge_reward": 0.16984953731298447,
      "rewards/forensics_tiou_reward": 0.22330951690673828,
      "step": 348
    },
    {
      "completion_length": 82.78125,
      "epoch": 0.7308900523560209,
      "grad_norm": 10.799991607666016,
      "kl": 2.0625,
      "learning_rate": 8.176100628930818e-07,
      "loss": 0.2936,
      "reward": 0.586152970790863,
      "reward_std": 0.18426354974508286,
      "reward_std_all": 0.19748928397893906,
      "rewards-max/forensics_sinkhorn_reward": 0.8953381478786469,
      "rewards-max/forensics_strict_edge_reward": 0.6340309977531433,
      "rewards-max/forensics_tiou_reward": 0.7924914658069611,
      "rewards/forensics_sinkhorn_reward": 0.3887595236301422,
      "rewards/forensics_strict_edge_reward": 0.1607338860630989,
      "rewards/forensics_tiou_reward": 0.22461476922035217,
      "step": 349
    },
    {
      "completion_length": 30.96875,
      "epoch": 0.7329842931937173,
      "grad_norm": 12.614555358886719,
      "kl": 1.6796875,
      "learning_rate": 8.170859538784066e-07,
      "loss": 0.3207,
      "reward": 0.6396270394325256,
      "reward_std": 0.14748891443014145,
      "reward_std_all": 0.1539299637079239,
      "rewards-max/forensics_sinkhorn_reward": 0.9722413718700409,
      "rewards-max/forensics_strict_edge_reward": 0.8078543543815613,
      "rewards-max/forensics_tiou_reward": 0.9362360239028931,
      "rewards/forensics_sinkhorn_reward": 0.4816148281097412,
      "rewards/forensics_strict_edge_reward": 0.19251147657632828,
      "rewards/forensics_tiou_reward": 0.27449987828731537,
      "step": 350
    },
    {
      "completion_length": 48.203125,
      "epoch": 0.7350785340314137,
      "grad_norm": 12.236503601074219,
      "kl": 2.71875,
      "learning_rate": 8.165618448637316e-07,
      "loss": 0.2556,
      "reward": 0.6009339094161987,
      "reward_std": 0.169943705201149,
      "reward_std_all": 0.18027737736701965,
      "rewards-max/forensics_sinkhorn_reward": 0.9852482080459595,
      "rewards-max/forensics_strict_edge_reward": 0.7009570598602295,
      "rewards-max/forensics_tiou_reward": 0.9852834343910217,
      "rewards/forensics_sinkhorn_reward": 0.4636920392513275,
      "rewards/forensics_strict_edge_reward": 0.13324465602636337,
      "rewards/forensics_tiou_reward": 0.2538052275776863,
      "step": 351
    },
    {
      "completion_length": 34.265625,
      "epoch": 0.7371727748691099,
      "grad_norm": 13.776784896850586,
      "kl": 1.5859375,
      "learning_rate": 8.160377358490566e-07,
      "loss": 0.3266,
      "reward": 0.6665847599506378,
      "reward_std": 0.0874926969408989,
      "reward_std_all": 0.11012960970401764,
      "rewards-max/forensics_sinkhorn_reward": 0.9139339327812195,
      "rewards-max/forensics_strict_edge_reward": 0.7309097647666931,
      "rewards-max/forensics_tiou_reward": 0.8361072540283203,
      "rewards/forensics_sinkhorn_reward": 0.4431982785463333,
      "rewards/forensics_strict_edge_reward": 0.22830776870250702,
      "rewards/forensics_tiou_reward": 0.2869989573955536,
      "step": 352
    },
    {
      "completion_length": 42.84375,
      "epoch": 0.7392670157068063,
      "grad_norm": 11.930855751037598,
      "kl": 1.7265625,
      "learning_rate": 8.155136268343815e-07,
      "loss": 0.3084,
      "reward": 0.668004721403122,
      "reward_std": 0.16566256433725357,
      "reward_std_all": 0.15675688162446022,
      "rewards-max/forensics_sinkhorn_reward": 0.9879907965660095,
      "rewards-max/forensics_strict_edge_reward": 0.9464850425720215,
      "rewards-max/forensics_tiou_reward": 0.9960630238056183,
      "rewards/forensics_sinkhorn_reward": 0.507666602730751,
      "rewards/forensics_strict_edge_reward": 0.31966929137706757,
      "rewards/forensics_tiou_reward": 0.3806600570678711,
      "step": 353
    },
    {
      "completion_length": 38.234375,
      "epoch": 0.7413612565445026,
      "grad_norm": 10.906997680664062,
      "kl": 1.21875,
      "learning_rate": 8.149895178197065e-07,
      "loss": 0.0092,
      "reward": 0.5758063197135925,
      "reward_std": 0.17499923706054688,
      "reward_std_all": 0.19631103426218033,
      "rewards-max/forensics_sinkhorn_reward": 0.9872041642665863,
      "rewards-max/forensics_strict_edge_reward": 0.9565287232398987,
      "rewards-max/forensics_tiou_reward": 0.9929845929145813,
      "rewards/forensics_sinkhorn_reward": 0.45889444649219513,
      "rewards/forensics_strict_edge_reward": 0.19153782725334167,
      "rewards/forensics_tiou_reward": 0.2710156738758087,
      "step": 354
    },
    {
      "completion_length": 31.234375,
      "epoch": 0.743455497382199,
      "grad_norm": 12.948569297790527,
      "kl": 1.56640625,
      "learning_rate": 8.144654088050314e-07,
      "loss": 0.166,
      "reward": 0.6176036894321442,
      "reward_std": 0.15971481055021286,
      "reward_std_all": 0.1751282587647438,
      "rewards-max/forensics_sinkhorn_reward": 0.9882758855819702,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9971829652786255,
      "rewards/forensics_sinkhorn_reward": 0.5464436709880829,
      "rewards/forensics_strict_edge_reward": 0.2418707087635994,
      "rewards/forensics_tiou_reward": 0.339821994304657,
      "step": 355
    },
    {
      "completion_length": 48.203125,
      "epoch": 0.7455497382198953,
      "grad_norm": 16.690221786499023,
      "kl": 1.640625,
      "learning_rate": 8.139412997903563e-07,
      "loss": 0.243,
      "reward": 0.5888376533985138,
      "reward_std": 0.13372664153575897,
      "reward_std_all": 0.1710081323981285,
      "rewards-max/forensics_sinkhorn_reward": 0.8672776222229004,
      "rewards-max/forensics_strict_edge_reward": 0.755094051361084,
      "rewards-max/forensics_tiou_reward": 0.7425509691238403,
      "rewards/forensics_sinkhorn_reward": 0.43644213676452637,
      "rewards/forensics_strict_edge_reward": 0.17700960487127304,
      "rewards/forensics_tiou_reward": 0.229353666305542,
      "step": 356
    },
    {
      "completion_length": 41.03125,
      "epoch": 0.7476439790575916,
      "grad_norm": 14.881319046020508,
      "kl": 1.11328125,
      "learning_rate": 8.134171907756812e-07,
      "loss": 0.0528,
      "reward": 0.6245904862880707,
      "reward_std": 0.12718596681952477,
      "reward_std_all": 0.15111911296844482,
      "rewards-max/forensics_sinkhorn_reward": 0.8486663103103638,
      "rewards-max/forensics_strict_edge_reward": 0.7335946708917618,
      "rewards-max/forensics_tiou_reward": 0.7233672142028809,
      "rewards/forensics_sinkhorn_reward": 0.4884377121925354,
      "rewards/forensics_strict_edge_reward": 0.1541886106133461,
      "rewards/forensics_tiou_reward": 0.22484923526644707,
      "step": 357
    },
    {
      "completion_length": 63.125,
      "epoch": 0.749738219895288,
      "grad_norm": 14.010908126831055,
      "kl": 1.68359375,
      "learning_rate": 8.128930817610062e-07,
      "loss": 0.2287,
      "reward": 0.6163650751113892,
      "reward_std": 0.15242037922143936,
      "reward_std_all": 0.17431578785181046,
      "rewards-max/forensics_sinkhorn_reward": 0.9508621692657471,
      "rewards-max/forensics_strict_edge_reward": 0.7689562439918518,
      "rewards-max/forensics_tiou_reward": 0.8958482146263123,
      "rewards/forensics_sinkhorn_reward": 0.4856436401605606,
      "rewards/forensics_strict_edge_reward": 0.16537673771381378,
      "rewards/forensics_tiou_reward": 0.2546505182981491,
      "step": 358
    },
    {
      "completion_length": 35.96875,
      "epoch": 0.7518324607329843,
      "grad_norm": 909.4579467773438,
      "kl": 15.05078125,
      "learning_rate": 8.123689727463313e-07,
      "loss": 0.5931,
      "reward": 0.6604982018470764,
      "reward_std": 0.10411551780998707,
      "reward_std_all": 0.1188453882932663,
      "rewards-max/forensics_sinkhorn_reward": 0.9834103882312775,
      "rewards-max/forensics_strict_edge_reward": 0.878633052110672,
      "rewards-max/forensics_tiou_reward": 0.9780207574367523,
      "rewards/forensics_sinkhorn_reward": 0.5319771319627762,
      "rewards/forensics_strict_edge_reward": 0.1872016340494156,
      "rewards/forensics_tiou_reward": 0.2684275805950165,
      "step": 359
    },
    {
      "completion_length": 93.34375,
      "epoch": 0.7539267015706806,
      "grad_norm": 9.58353042602539,
      "kl": 1.89453125,
      "learning_rate": 8.118448637316562e-07,
      "loss": 0.2527,
      "reward": 0.5638458430767059,
      "reward_std": 0.147149495780468,
      "reward_std_all": 0.1888788640499115,
      "rewards-max/forensics_sinkhorn_reward": 0.9884711802005768,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9979528188705444,
      "rewards/forensics_sinkhorn_reward": 0.47320039570331573,
      "rewards/forensics_strict_edge_reward": 0.2641827091574669,
      "rewards/forensics_tiou_reward": 0.34625355899333954,
      "step": 360
    },
    {
      "completion_length": 75.078125,
      "epoch": 0.7560209424083769,
      "grad_norm": 9.48689079284668,
      "kl": 1.8984375,
      "learning_rate": 8.113207547169812e-07,
      "loss": 0.2706,
      "reward": 0.6172345578670502,
      "reward_std": 0.15785057842731476,
      "reward_std_all": 0.16983821243047714,
      "rewards-max/forensics_sinkhorn_reward": 0.987021803855896,
      "rewards-max/forensics_strict_edge_reward": 0.8771188855171204,
      "rewards-max/forensics_tiou_reward": 0.9922400414943695,
      "rewards/forensics_sinkhorn_reward": 0.4554954171180725,
      "rewards/forensics_strict_edge_reward": 0.2831876873970032,
      "rewards/forensics_tiou_reward": 0.3398389220237732,
      "step": 361
    },
    {
      "completion_length": 31.8125,
      "epoch": 0.7581151832460733,
      "grad_norm": 15.954124450683594,
      "kl": 1.16796875,
      "learning_rate": 8.10796645702306e-07,
      "loss": 0.1184,
      "reward": 0.6757847666740417,
      "reward_std": 0.13507437705993652,
      "reward_std_all": 0.13554973527789116,
      "rewards-max/forensics_sinkhorn_reward": 0.9878574311733246,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9955334365367889,
      "rewards/forensics_sinkhorn_reward": 0.5554863512516022,
      "rewards/forensics_strict_edge_reward": 0.23280444741249084,
      "rewards/forensics_tiou_reward": 0.3276139199733734,
      "step": 362
    },
    {
      "completion_length": 43.015625,
      "epoch": 0.7602094240837697,
      "grad_norm": 10.933754920959473,
      "kl": 1.76171875,
      "learning_rate": 8.10272536687631e-07,
      "loss": 0.1774,
      "reward": 0.638807475566864,
      "reward_std": 0.16403930634260178,
      "reward_std_all": 0.16261859238147736,
      "rewards-max/forensics_sinkhorn_reward": 0.9738484025001526,
      "rewards-max/forensics_strict_edge_reward": 0.830383837223053,
      "rewards-max/forensics_tiou_reward": 0.956670492887497,
      "rewards/forensics_sinkhorn_reward": 0.47867293655872345,
      "rewards/forensics_strict_edge_reward": 0.2124074250459671,
      "rewards/forensics_tiou_reward": 0.2833217829465866,
      "step": 363
    },
    {
      "completion_length": 55.984375,
      "epoch": 0.762303664921466,
      "grad_norm": 12.359119415283203,
      "kl": 1.453125,
      "learning_rate": 8.097484276729559e-07,
      "loss": 0.2019,
      "reward": 0.6805836856365204,
      "reward_std": 0.10405455902218819,
      "reward_std_all": 0.11128552258014679,
      "rewards-max/forensics_sinkhorn_reward": 0.9858456552028656,
      "rewards-max/forensics_strict_edge_reward": 0.9355069994926453,
      "rewards-max/forensics_tiou_reward": 0.9876428842544556,
      "rewards/forensics_sinkhorn_reward": 0.5960907936096191,
      "rewards/forensics_strict_edge_reward": 0.2839723974466324,
      "rewards/forensics_tiou_reward": 0.4161847084760666,
      "step": 364
    },
    {
      "completion_length": 81.4375,
      "epoch": 0.7643979057591623,
      "grad_norm": 11.520055770874023,
      "kl": 1.734375,
      "learning_rate": 8.092243186582809e-07,
      "loss": 0.1187,
      "reward": 0.6398314535617828,
      "reward_std": 0.1374349519610405,
      "reward_std_all": 0.15027379989624023,
      "rewards-max/forensics_sinkhorn_reward": 0.9878842532634735,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9956441521644592,
      "rewards/forensics_sinkhorn_reward": 0.5197647213935852,
      "rewards/forensics_strict_edge_reward": 0.22822795808315277,
      "rewards/forensics_tiou_reward": 0.296373575925827,
      "step": 365
    },
    {
      "completion_length": 44.65625,
      "epoch": 0.7664921465968586,
      "grad_norm": 12.116421699523926,
      "kl": 2.6875,
      "learning_rate": 8.087002096436058e-07,
      "loss": 0.29,
      "reward": 0.6461737155914307,
      "reward_std": 0.15542417019605637,
      "reward_std_all": 0.15810637921094894,
      "rewards-max/forensics_sinkhorn_reward": 0.9879896640777588,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9960553050041199,
      "rewards/forensics_sinkhorn_reward": 0.5047888159751892,
      "rewards/forensics_strict_edge_reward": 0.20472563803195953,
      "rewards/forensics_tiou_reward": 0.28900615870952606,
      "step": 366
    },
    {
      "completion_length": 62.109375,
      "epoch": 0.768586387434555,
      "grad_norm": 17.314891815185547,
      "kl": 1.8203125,
      "learning_rate": 8.081761006289308e-07,
      "loss": 0.189,
      "reward": 0.6333727538585663,
      "reward_std": 0.12739402428269386,
      "reward_std_all": 0.14389222860336304,
      "rewards-max/forensics_sinkhorn_reward": 0.9865863621234894,
      "rewards-max/forensics_strict_edge_reward": 0.895060271024704,
      "rewards-max/forensics_tiou_reward": 0.9905242025852203,
      "rewards/forensics_sinkhorn_reward": 0.5041238069534302,
      "rewards/forensics_strict_edge_reward": 0.17383644729852676,
      "rewards/forensics_tiou_reward": 0.25603166967630386,
      "step": 367
    },
    {
      "completion_length": 73.28125,
      "epoch": 0.7706806282722513,
      "grad_norm": 11.128815650939941,
      "kl": 2.484375,
      "learning_rate": 8.076519916142557e-07,
      "loss": 0.1796,
      "reward": 0.6791796386241913,
      "reward_std": 0.131170816719532,
      "reward_std_all": 0.12807069718837738,
      "rewards-max/forensics_sinkhorn_reward": 0.9835722148418427,
      "rewards-max/forensics_strict_edge_reward": 0.8983768224716187,
      "rewards-max/forensics_tiou_reward": 0.9788882434368134,
      "rewards/forensics_sinkhorn_reward": 0.5509446263313293,
      "rewards/forensics_strict_edge_reward": 0.31466108560562134,
      "rewards/forensics_tiou_reward": 0.35305219888687134,
      "step": 368
    },
    {
      "completion_length": 111.84375,
      "epoch": 0.7727748691099476,
      "grad_norm": 9.744295120239258,
      "kl": 2.43359375,
      "learning_rate": 8.071278825995806e-07,
      "loss": 0.3117,
      "reward": 0.613857239484787,
      "reward_std": 0.20063424110412598,
      "reward_std_all": 0.20429721474647522,
      "rewards-max/forensics_sinkhorn_reward": 0.9870018362998962,
      "rewards-max/forensics_strict_edge_reward": 0.9565287232398987,
      "rewards-max/forensics_tiou_reward": 0.9921696484088898,
      "rewards/forensics_sinkhorn_reward": 0.4883623421192169,
      "rewards/forensics_strict_edge_reward": 0.25966187566518784,
      "rewards/forensics_tiou_reward": 0.36031825840473175,
      "step": 369
    },
    {
      "completion_length": 69.9375,
      "epoch": 0.774869109947644,
      "grad_norm": 11.159025192260742,
      "kl": 1.68359375,
      "learning_rate": 8.066037735849056e-07,
      "loss": 0.2624,
      "reward": 0.6240909695625305,
      "reward_std": 0.17634977400302887,
      "reward_std_all": 0.18131743371486664,
      "rewards-max/forensics_sinkhorn_reward": 0.9872634410858154,
      "rewards-max/forensics_strict_edge_reward": 0.9565287232398987,
      "rewards-max/forensics_tiou_reward": 0.9932172894477844,
      "rewards/forensics_sinkhorn_reward": 0.5401276648044586,
      "rewards/forensics_strict_edge_reward": 0.3015614449977875,
      "rewards/forensics_tiou_reward": 0.40386585891246796,
      "step": 370
    },
    {
      "completion_length": 82.625,
      "epoch": 0.7769633507853403,
      "grad_norm": 10.85241413116455,
      "kl": 2.9296875,
      "learning_rate": 8.060796645702305e-07,
      "loss": 0.2776,
      "reward": 0.616179347038269,
      "reward_std": 0.15108155086636543,
      "reward_std_all": 0.1660265550017357,
      "rewards-max/forensics_sinkhorn_reward": 0.987594723701477,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9961894750595093,
      "rewards/forensics_sinkhorn_reward": 0.4986036419868469,
      "rewards/forensics_strict_edge_reward": 0.2978525459766388,
      "rewards/forensics_tiou_reward": 0.39338624477386475,
      "step": 371
    },
    {
      "completion_length": 57.0,
      "epoch": 0.7790575916230367,
      "grad_norm": 11.088031768798828,
      "kl": 1.447265625,
      "learning_rate": 8.055555555555556e-07,
      "loss": 0.1473,
      "reward": 0.7005992531776428,
      "reward_std": 0.10674995929002762,
      "reward_std_all": 0.10700057819485664,
      "rewards-max/forensics_sinkhorn_reward": 0.9882349669933319,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9970241189002991,
      "rewards/forensics_sinkhorn_reward": 0.597374364733696,
      "rewards/forensics_strict_edge_reward": 0.3640635460615158,
      "rewards/forensics_tiou_reward": 0.44294922053813934,
      "step": 372
    },
    {
      "completion_length": 61.671875,
      "epoch": 0.7811518324607329,
      "grad_norm": 17.260602951049805,
      "kl": 1.65625,
      "learning_rate": 8.050314465408805e-07,
      "loss": 0.0847,
      "reward": 0.6811679303646088,
      "reward_std": 0.12378884851932526,
      "reward_std_all": 0.12550052627921104,
      "rewards-max/forensics_sinkhorn_reward": 0.9883950054645538,
      "rewards-max/forensics_strict_edge_reward": 0.9890114367008209,
      "rewards-max/forensics_tiou_reward": 0.9976553320884705,
      "rewards/forensics_sinkhorn_reward": 0.6019537448883057,
      "rewards/forensics_strict_edge_reward": 0.330498531460762,
      "rewards/forensics_tiou_reward": 0.43000832200050354,
      "step": 373
    },
    {
      "completion_length": 57.5,
      "epoch": 0.7832460732984293,
      "grad_norm": 13.973152160644531,
      "kl": 2.0546875,
      "learning_rate": 8.045073375262054e-07,
      "loss": 0.3423,
      "reward": 0.6531829535961151,
      "reward_std": 0.13038138672709465,
      "reward_std_all": 0.13842306286096573,
      "rewards-max/forensics_sinkhorn_reward": 0.9871909618377686,
      "rewards-max/forensics_strict_edge_reward": 0.9574736058712006,
      "rewards-max/forensics_tiou_reward": 0.9929329454898834,
      "rewards/forensics_sinkhorn_reward": 0.5264302492141724,
      "rewards/forensics_strict_edge_reward": 0.2971761226654053,
      "rewards/forensics_tiou_reward": 0.33810949325561523,
      "step": 374
    },
    {
      "completion_length": 74.15625,
      "epoch": 0.7853403141361257,
      "grad_norm": 16.933921813964844,
      "kl": 2.6484375,
      "learning_rate": 8.039832285115304e-07,
      "loss": 0.2585,
      "reward": 0.6236869096755981,
      "reward_std": 0.1636986956000328,
      "reward_std_all": 0.17266108840703964,
      "rewards-max/forensics_sinkhorn_reward": 0.9856821000576019,
      "rewards-max/forensics_strict_edge_reward": 0.8663162291049957,
      "rewards-max/forensics_tiou_reward": 0.9870495796203613,
      "rewards/forensics_sinkhorn_reward": 0.5138104408979416,
      "rewards/forensics_strict_edge_reward": 0.28367459774017334,
      "rewards/forensics_tiou_reward": 0.3448439985513687,
      "step": 375
    },
    {
      "completion_length": 62.203125,
      "epoch": 0.787434554973822,
      "grad_norm": 26.884782791137695,
      "kl": 2.42578125,
      "learning_rate": 8.034591194968553e-07,
      "loss": 0.1411,
      "reward": 0.6146535575389862,
      "reward_std": 0.12848656624555588,
      "reward_std_all": 0.15186183154582977,
      "rewards-max/forensics_sinkhorn_reward": 0.9788262844085693,
      "rewards-max/forensics_strict_edge_reward": 0.947419673204422,
      "rewards-max/forensics_tiou_reward": 0.976721316576004,
      "rewards/forensics_sinkhorn_reward": 0.5385089814662933,
      "rewards/forensics_strict_edge_reward": 0.29448918998241425,
      "rewards/forensics_tiou_reward": 0.3842790722846985,
      "step": 376
    },
    {
      "completion_length": 70.015625,
      "epoch": 0.7895287958115184,
      "grad_norm": 11.593182563781738,
      "kl": 2.2578125,
      "learning_rate": 8.029350104821803e-07,
      "loss": 0.2795,
      "reward": 0.6974977552890778,
      "reward_std": 0.12010939419269562,
      "reward_std_all": 0.1184532567858696,
      "rewards-max/forensics_sinkhorn_reward": 0.9872660934925079,
      "rewards-max/forensics_strict_edge_reward": 0.9565287232398987,
      "rewards-max/forensics_tiou_reward": 0.9932027757167816,
      "rewards/forensics_sinkhorn_reward": 0.4990675747394562,
      "rewards/forensics_strict_edge_reward": 0.30971337854862213,
      "rewards/forensics_tiou_reward": 0.33629457652568817,
      "step": 377
    },
    {
      "completion_length": 123.359375,
      "epoch": 0.7916230366492146,
      "grad_norm": 29.478458404541016,
      "kl": 3.828125,
      "learning_rate": 8.024109014675052e-07,
      "loss": 0.3427,
      "reward": 0.527700662612915,
      "reward_std": 0.16269393265247345,
      "reward_std_all": 0.216585211455822,
      "rewards-max/forensics_sinkhorn_reward": 0.9878644645214081,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9955626726150513,
      "rewards/forensics_sinkhorn_reward": 0.3823760747909546,
      "rewards/forensics_strict_edge_reward": 0.25827808678150177,
      "rewards/forensics_tiou_reward": 0.2890481799840927,
      "step": 378
    },
    {
      "completion_length": 86.359375,
      "epoch": 0.793717277486911,
      "grad_norm": 15.284066200256348,
      "kl": 2.45703125,
      "learning_rate": 8.018867924528302e-07,
      "loss": 0.3633,
      "reward": 0.5999764502048492,
      "reward_std": 0.21785549819469452,
      "reward_std_all": 0.21008524298667908,
      "rewards-max/forensics_sinkhorn_reward": 0.9023551046848297,
      "rewards-max/forensics_strict_edge_reward": 0.7923639118671417,
      "rewards-max/forensics_tiou_reward": 0.8230381608009338,
      "rewards/forensics_sinkhorn_reward": 0.4749583601951599,
      "rewards/forensics_strict_edge_reward": 0.28309328854084015,
      "rewards/forensics_tiou_reward": 0.33820194005966187,
      "step": 379
    },
    {
      "completion_length": 127.421875,
      "epoch": 0.7958115183246073,
      "grad_norm": 20.35257911682129,
      "kl": 3.8046875,
      "learning_rate": 8.01362683438155e-07,
      "loss": 0.3928,
      "reward": 0.5616852343082428,
      "reward_std": 0.09952060133218765,
      "reward_std_all": 0.16073982417583466,
      "rewards-max/forensics_sinkhorn_reward": 0.939862459897995,
      "rewards-max/forensics_strict_edge_reward": 0.7176906764507294,
      "rewards-max/forensics_tiou_reward": 0.8968012928962708,
      "rewards/forensics_sinkhorn_reward": 0.3553558439016342,
      "rewards/forensics_strict_edge_reward": 0.202090322971344,
      "rewards/forensics_tiou_reward": 0.23910893499851227,
      "step": 380
    },
    {
      "completion_length": 71.375,
      "epoch": 0.7979057591623037,
      "grad_norm": 11.731515884399414,
      "kl": 2.09765625,
      "learning_rate": 8.0083857442348e-07,
      "loss": 0.2017,
      "reward": 0.5440278053283691,
      "reward_std": 0.19696280360221863,
      "reward_std_all": 0.22114164382219315,
      "rewards-max/forensics_sinkhorn_reward": 0.9833482205867767,
      "rewards-max/forensics_strict_edge_reward": 0.7933230996131897,
      "rewards-max/forensics_tiou_reward": 0.9780246913433075,
      "rewards/forensics_sinkhorn_reward": 0.3960641771554947,
      "rewards/forensics_strict_edge_reward": 0.15029443427920341,
      "rewards/forensics_tiou_reward": 0.19461669027805328,
      "step": 381
    },
    {
      "completion_length": 142.640625,
      "epoch": 0.8,
      "grad_norm": 70.4552993774414,
      "kl": 3.1171875,
      "learning_rate": 8.003144654088049e-07,
      "loss": 0.1899,
      "reward": 0.5436204373836517,
      "reward_std": 0.15078365802764893,
      "reward_std_all": 0.19965891540050507,
      "rewards-max/forensics_sinkhorn_reward": 0.8632036745548248,
      "rewards-max/forensics_strict_edge_reward": 0.8078543543815613,
      "rewards-max/forensics_tiou_reward": 0.8185067176818848,
      "rewards/forensics_sinkhorn_reward": 0.3376677930355072,
      "rewards/forensics_strict_edge_reward": 0.1446346715092659,
      "rewards/forensics_tiou_reward": 0.20874658226966858,
      "step": 382
    },
    {
      "completion_length": 125.390625,
      "epoch": 0.8020942408376963,
      "grad_norm": 14.131765365600586,
      "kl": 3.21875,
      "learning_rate": 7.9979035639413e-07,
      "loss": 0.3513,
      "reward": 0.497243732213974,
      "reward_std": 0.22701898217201233,
      "reward_std_all": 0.25690770149230957,
      "rewards-max/forensics_sinkhorn_reward": 0.8643901646137238,
      "rewards-max/forensics_strict_edge_reward": 0.7696853578090668,
      "rewards-max/forensics_tiou_reward": 0.754094660282135,
      "rewards/forensics_sinkhorn_reward": 0.2919945865869522,
      "rewards/forensics_strict_edge_reward": 0.14902885630726814,
      "rewards/forensics_tiou_reward": 0.16229916363954544,
      "step": 383
    },
    {
      "completion_length": 107.734375,
      "epoch": 0.8041884816753927,
      "grad_norm": 12.229015350341797,
      "kl": 2.50390625,
      "learning_rate": 7.99266247379455e-07,
      "loss": 0.2461,
      "reward": 0.5548532903194427,
      "reward_std": 0.2043551653623581,
      "reward_std_all": 0.21773408353328705,
      "rewards-max/forensics_sinkhorn_reward": 0.9546953439712524,
      "rewards-max/forensics_strict_edge_reward": 0.9220033884048462,
      "rewards-max/forensics_tiou_reward": 0.9269488453865051,
      "rewards/forensics_sinkhorn_reward": 0.36277367174625397,
      "rewards/forensics_strict_edge_reward": 0.14448212087154388,
      "rewards/forensics_tiou_reward": 0.19757795333862305,
      "step": 384
    },
    {
      "completion_length": 66.828125,
      "epoch": 0.806282722513089,
      "grad_norm": 11.611054420471191,
      "kl": 1.703125,
      "learning_rate": 7.987421383647799e-07,
      "loss": 0.2507,
      "reward": 0.5828326344490051,
      "reward_std": 0.1610754281282425,
      "reward_std_all": 0.18136221170425415,
      "rewards-max/forensics_sinkhorn_reward": 0.9852785468101501,
      "rewards-max/forensics_strict_edge_reward": 0.7299555838108063,
      "rewards-max/forensics_tiou_reward": 0.9853708148002625,
      "rewards/forensics_sinkhorn_reward": 0.4755006283521652,
      "rewards/forensics_strict_edge_reward": 0.1697278395295143,
      "rewards/forensics_tiou_reward": 0.3334198445081711,
      "step": 385
    },
    {
      "completion_length": 70.296875,
      "epoch": 0.8083769633507853,
      "grad_norm": 10.945001602172852,
      "kl": 1.91015625,
      "learning_rate": 7.982180293501048e-07,
      "loss": 0.205,
      "reward": 0.6203666627407074,
      "reward_std": 0.21164732426404953,
      "reward_std_all": 0.1987750381231308,
      "rewards-max/forensics_sinkhorn_reward": 0.9858415126800537,
      "rewards-max/forensics_strict_edge_reward": 0.9464850425720215,
      "rewards-max/forensics_tiou_reward": 0.9910840094089508,
      "rewards/forensics_sinkhorn_reward": 0.5249990224838257,
      "rewards/forensics_strict_edge_reward": 0.2970542758703232,
      "rewards/forensics_tiou_reward": 0.3975493907928467,
      "step": 386
    },
    {
      "completion_length": 89.34375,
      "epoch": 0.8104712041884817,
      "grad_norm": 9.884316444396973,
      "kl": 2.1796875,
      "learning_rate": 7.976939203354297e-07,
      "loss": 0.2533,
      "reward": 0.5734796226024628,
      "reward_std": 0.1546509936451912,
      "reward_std_all": 0.1874931901693344,
      "rewards-max/forensics_sinkhorn_reward": 0.9841569066047668,
      "rewards-max/forensics_strict_edge_reward": 0.9048932790756226,
      "rewards-max/forensics_tiou_reward": 0.9810456037521362,
      "rewards/forensics_sinkhorn_reward": 0.4448273628950119,
      "rewards/forensics_strict_edge_reward": 0.200165793299675,
      "rewards/forensics_tiou_reward": 0.25793587416410446,
      "step": 387
    },
    {
      "completion_length": 90.171875,
      "epoch": 0.812565445026178,
      "grad_norm": 13.560551643371582,
      "kl": 1.66015625,
      "learning_rate": 7.971698113207547e-07,
      "loss": 0.1317,
      "reward": 0.5621343553066254,
      "reward_std": 0.1527419239282608,
      "reward_std_all": 0.18818213790655136,
      "rewards-max/forensics_sinkhorn_reward": 0.9771097898483276,
      "rewards-max/forensics_strict_edge_reward": 0.744240403175354,
      "rewards-max/forensics_tiou_reward": 0.9692908823490143,
      "rewards/forensics_sinkhorn_reward": 0.42493753135204315,
      "rewards/forensics_strict_edge_reward": 0.13864677399396896,
      "rewards/forensics_tiou_reward": 0.1906595602631569,
      "step": 388
    },
    {
      "completion_length": 60.109375,
      "epoch": 0.8146596858638744,
      "grad_norm": 13.409625053405762,
      "kl": 1.544921875,
      "learning_rate": 7.966457023060796e-07,
      "loss": 0.2289,
      "reward": 0.5614049136638641,
      "reward_std": 0.21321161836385727,
      "reward_std_all": 0.22555771470069885,
      "rewards-max/forensics_sinkhorn_reward": 0.9860903918743134,
      "rewards-max/forensics_strict_edge_reward": 0.8854433596134186,
      "rewards-max/forensics_tiou_reward": 0.988614022731781,
      "rewards/forensics_sinkhorn_reward": 0.44232697784900665,
      "rewards/forensics_strict_edge_reward": 0.16497506946325302,
      "rewards/forensics_tiou_reward": 0.22278673946857452,
      "step": 389
    },
    {
      "completion_length": 55.359375,
      "epoch": 0.8167539267015707,
      "grad_norm": 11.526867866516113,
      "kl": 1.75,
      "learning_rate": 7.961215932914046e-07,
      "loss": 0.2401,
      "reward": 0.6017283201217651,
      "reward_std": 0.18335355818271637,
      "reward_std_all": 0.18900657445192337,
      "rewards-max/forensics_sinkhorn_reward": 0.880865752696991,
      "rewards-max/forensics_strict_edge_reward": 0.7968259453773499,
      "rewards-max/forensics_tiou_reward": 0.7765234410762787,
      "rewards/forensics_sinkhorn_reward": 0.4423629641532898,
      "rewards/forensics_strict_edge_reward": 0.2008807212114334,
      "rewards/forensics_tiou_reward": 0.2338082119822502,
      "step": 390
    },
    {
      "completion_length": 63.390625,
      "epoch": 0.818848167539267,
      "grad_norm": 25.26416015625,
      "kl": 1.94921875,
      "learning_rate": 7.955974842767296e-07,
      "loss": 0.0619,
      "reward": 0.6100369989871979,
      "reward_std": 0.17037254571914673,
      "reward_std_all": 0.1770768240094185,
      "rewards-max/forensics_sinkhorn_reward": 0.8844596743583679,
      "rewards-max/forensics_strict_edge_reward": 0.8049824237823486,
      "rewards-max/forensics_tiou_reward": 0.8288529813289642,
      "rewards/forensics_sinkhorn_reward": 0.44724176824092865,
      "rewards/forensics_strict_edge_reward": 0.19588248431682587,
      "rewards/forensics_tiou_reward": 0.26454252004623413,
      "step": 391
    },
    {
      "completion_length": 57.65625,
      "epoch": 0.8209424083769633,
      "grad_norm": 9.96457576751709,
      "kl": 1.29296875,
      "learning_rate": 7.950733752620544e-07,
      "loss": 0.1483,
      "reward": 0.6337786018848419,
      "reward_std": 0.15878919139504433,
      "reward_std_all": 0.16338606551289558,
      "rewards-max/forensics_sinkhorn_reward": 0.9884005486965179,
      "rewards-max/forensics_strict_edge_reward": 0.9782643616199493,
      "rewards-max/forensics_tiou_reward": 0.9976744055747986,
      "rewards/forensics_sinkhorn_reward": 0.5741629302501678,
      "rewards/forensics_strict_edge_reward": 0.33390963077545166,
      "rewards/forensics_tiou_reward": 0.45079340040683746,
      "step": 392
    },
    {
      "completion_length": 97.015625,
      "epoch": 0.8230366492146597,
      "grad_norm": 13.515162467956543,
      "kl": 2.078125,
      "learning_rate": 7.945492662473794e-07,
      "loss": 0.1811,
      "reward": 0.6405176818370819,
      "reward_std": 0.16243328899145126,
      "reward_std_all": 0.16086919605731964,
      "rewards-max/forensics_sinkhorn_reward": 0.9638344049453735,
      "rewards-max/forensics_strict_edge_reward": 0.7923295497894287,
      "rewards-max/forensics_tiou_reward": 0.9352610111236572,
      "rewards/forensics_sinkhorn_reward": 0.4194979667663574,
      "rewards/forensics_strict_edge_reward": 0.23529259860515594,
      "rewards/forensics_tiou_reward": 0.2543507292866707,
      "step": 393
    },
    {
      "completion_length": 29.0625,
      "epoch": 0.8251308900523561,
      "grad_norm": 14.368531227111816,
      "kl": 1.3125,
      "learning_rate": 7.940251572327043e-07,
      "loss": -0.079,
      "reward": 0.6371451616287231,
      "reward_std": 0.12351946532726288,
      "reward_std_all": 0.14929182827472687,
      "rewards-max/forensics_sinkhorn_reward": 0.9847583472728729,
      "rewards-max/forensics_strict_edge_reward": 0.8968285918235779,
      "rewards-max/forensics_tiou_reward": 0.9833232164382935,
      "rewards/forensics_sinkhorn_reward": 0.4950793981552124,
      "rewards/forensics_strict_edge_reward": 0.18872952088713646,
      "rewards/forensics_tiou_reward": 0.22888176143169403,
      "step": 394
    },
    {
      "completion_length": 37.21875,
      "epoch": 0.8272251308900523,
      "grad_norm": 11.07943344116211,
      "kl": 1.15234375,
      "learning_rate": 7.935010482180293e-07,
      "loss": 0.0862,
      "reward": 0.6629531383514404,
      "reward_std": 0.11788425594568253,
      "reward_std_all": 0.1276939958333969,
      "rewards-max/forensics_sinkhorn_reward": 0.9778182804584503,
      "rewards-max/forensics_strict_edge_reward": 0.7862354218959808,
      "rewards-max/forensics_tiou_reward": 0.9568836092948914,
      "rewards/forensics_sinkhorn_reward": 0.5844452679157257,
      "rewards/forensics_strict_edge_reward": 0.18117403984069824,
      "rewards/forensics_tiou_reward": 0.35162903368473053,
      "step": 395
    },
    {
      "completion_length": 50.203125,
      "epoch": 0.8293193717277487,
      "grad_norm": 9.800939559936523,
      "kl": 1.55859375,
      "learning_rate": 7.929769392033543e-07,
      "loss": 0.1763,
      "reward": 0.6504603028297424,
      "reward_std": 0.14394472539424896,
      "reward_std_all": 0.14866504818201065,
      "rewards-max/forensics_sinkhorn_reward": 0.9190659523010254,
      "rewards-max/forensics_strict_edge_reward": 0.8013064861297607,
      "rewards-max/forensics_tiou_reward": 0.8542194068431854,
      "rewards/forensics_sinkhorn_reward": 0.4828473627567291,
      "rewards/forensics_strict_edge_reward": 0.20529791712760925,
      "rewards/forensics_tiou_reward": 0.29551026225090027,
      "step": 396
    },
    {
      "completion_length": 39.5625,
      "epoch": 0.831413612565445,
      "grad_norm": 12.123475074768066,
      "kl": 1.4921875,
      "learning_rate": 7.924528301886793e-07,
      "loss": 0.217,
      "reward": 0.6418178677558899,
      "reward_std": 0.14790324866771698,
      "reward_std_all": 0.16018139570951462,
      "rewards-max/forensics_sinkhorn_reward": 0.9768860042095184,
      "rewards-max/forensics_strict_edge_reward": 0.9574736058712006,
      "rewards-max/forensics_tiou_reward": 0.9725897908210754,
      "rewards/forensics_sinkhorn_reward": 0.5421286672353745,
      "rewards/forensics_strict_edge_reward": 0.25187472999095917,
      "rewards/forensics_tiou_reward": 0.3522573634982109,
      "step": 397
    },
    {
      "completion_length": 49.25,
      "epoch": 0.8335078534031414,
      "grad_norm": 12.395153999328613,
      "kl": 0.982421875,
      "learning_rate": 7.919287211740041e-07,
      "loss": 0.0392,
      "reward": 0.5886364877223969,
      "reward_std": 0.10485241189599037,
      "reward_std_all": 0.15644297748804092,
      "rewards-max/forensics_sinkhorn_reward": 0.833721399307251,
      "rewards-max/forensics_strict_edge_reward": 0.6762852966785431,
      "rewards-max/forensics_tiou_reward": 0.6939620077610016,
      "rewards/forensics_sinkhorn_reward": 0.4767112135887146,
      "rewards/forensics_strict_edge_reward": 0.13488133996725082,
      "rewards/forensics_tiou_reward": 0.18540336191654205,
      "step": 398
    },
    {
      "completion_length": 70.078125,
      "epoch": 0.8356020942408376,
      "grad_norm": 10.930639266967773,
      "kl": 1.48828125,
      "learning_rate": 7.914046121593291e-07,
      "loss": 0.1344,
      "reward": 0.6223850548267365,
      "reward_std": 0.15344782173633575,
      "reward_std_all": 0.16554705426096916,
      "rewards-max/forensics_sinkhorn_reward": 0.988498866558075,
      "rewards-max/forensics_strict_edge_reward": 0.9677534997463226,
      "rewards-max/forensics_tiou_reward": 0.9980638921260834,
      "rewards/forensics_sinkhorn_reward": 0.49470406770706177,
      "rewards/forensics_strict_edge_reward": 0.1922702193260193,
      "rewards/forensics_tiou_reward": 0.29330237209796906,
      "step": 399
    },
    {
      "completion_length": 56.59375,
      "epoch": 0.837696335078534,
      "grad_norm": 13.32133960723877,
      "kl": 1.80859375,
      "learning_rate": 7.908805031446541e-07,
      "loss": 0.1532,
      "reward": 0.6036877036094666,
      "reward_std": 0.11872747540473938,
      "reward_std_all": 0.15681690722703934,
      "rewards-max/forensics_sinkhorn_reward": 0.9851172864437103,
      "rewards-max/forensics_strict_edge_reward": 0.856150895357132,
      "rewards-max/forensics_tiou_reward": 0.989238053560257,
      "rewards/forensics_sinkhorn_reward": 0.4831975996494293,
      "rewards/forensics_strict_edge_reward": 0.17179244756698608,
      "rewards/forensics_tiou_reward": 0.26162946969270706,
      "step": 400
    },
    {
      "completion_length": 55.578125,
      "epoch": 0.8397905759162304,
      "grad_norm": 11.76836109161377,
      "kl": 2.25,
      "learning_rate": 7.90356394129979e-07,
      "loss": 0.2188,
      "reward": 0.6063149571418762,
      "reward_std": 0.14715921878814697,
      "reward_std_all": 0.17001911252737045,
      "rewards-max/forensics_sinkhorn_reward": 0.9635065495967865,
      "rewards-max/forensics_strict_edge_reward": 0.9375866651535034,
      "rewards-max/forensics_tiou_reward": 0.9416880905628204,
      "rewards/forensics_sinkhorn_reward": 0.4464401453733444,
      "rewards/forensics_strict_edge_reward": 0.19026298075914383,
      "rewards/forensics_tiou_reward": 0.2766178846359253,
      "step": 401
    },
    {
      "completion_length": 52.84375,
      "epoch": 0.8418848167539267,
      "grad_norm": 14.91543197631836,
      "kl": 2.15625,
      "learning_rate": 7.89832285115304e-07,
      "loss": 0.1664,
      "reward": 0.6212252378463745,
      "reward_std": 0.13406309112906456,
      "reward_std_all": 0.15451983362436295,
      "rewards-max/forensics_sinkhorn_reward": 0.9879187941551208,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9957830011844635,
      "rewards/forensics_sinkhorn_reward": 0.48256391286849976,
      "rewards/forensics_strict_edge_reward": 0.20048534125089645,
      "rewards/forensics_tiou_reward": 0.2860342711210251,
      "step": 402
    },
    {
      "completion_length": 72.65625,
      "epoch": 0.8439790575916231,
      "grad_norm": 13.299675941467285,
      "kl": 1.8984375,
      "learning_rate": 7.893081761006289e-07,
      "loss": 0.3953,
      "reward": 0.6504672765731812,
      "reward_std": 0.14938220381736755,
      "reward_std_all": 0.15298476070165634,
      "rewards-max/forensics_sinkhorn_reward": 0.9766904413700104,
      "rewards-max/forensics_strict_edge_reward": 0.8237898349761963,
      "rewards-max/forensics_tiou_reward": 0.9615737497806549,
      "rewards/forensics_sinkhorn_reward": 0.4822758138179779,
      "rewards/forensics_strict_edge_reward": 0.22222352027893066,
      "rewards/forensics_tiou_reward": 0.3912753462791443,
      "step": 403
    },
    {
      "completion_length": 58.71875,
      "epoch": 0.8460732984293193,
      "grad_norm": 14.62213134765625,
      "kl": 2.2890625,
      "learning_rate": 7.887840670859538e-07,
      "loss": 0.2221,
      "reward": 0.6324888169765472,
      "reward_std": 0.12494851648807526,
      "reward_std_all": 0.14545321464538574,
      "rewards-max/forensics_sinkhorn_reward": 0.9885249137878418,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9981691241264343,
      "rewards/forensics_sinkhorn_reward": 0.4903979003429413,
      "rewards/forensics_strict_edge_reward": 0.21597128361463547,
      "rewards/forensics_tiou_reward": 0.2994939014315605,
      "step": 404
    },
    {
      "completion_length": 78.859375,
      "epoch": 0.8481675392670157,
      "grad_norm": 14.40625286102295,
      "kl": 2.0234375,
      "learning_rate": 7.882599580712787e-07,
      "loss": 0.1846,
      "reward": 0.5928353667259216,
      "reward_std": 0.14775271713733673,
      "reward_std_all": 0.17729001492261887,
      "rewards-max/forensics_sinkhorn_reward": 0.9874963164329529,
      "rewards-max/forensics_strict_edge_reward": 0.8968285918235779,
      "rewards-max/forensics_tiou_reward": 0.9941101670265198,
      "rewards/forensics_sinkhorn_reward": 0.44097480177879333,
      "rewards/forensics_strict_edge_reward": 0.22826797515153885,
      "rewards/forensics_tiou_reward": 0.2920066714286804,
      "step": 405
    },
    {
      "completion_length": 92.5625,
      "epoch": 0.8502617801047121,
      "grad_norm": 15.349365234375,
      "kl": 2.109375,
      "learning_rate": 7.877358490566037e-07,
      "loss": 0.2735,
      "reward": 0.560734897851944,
      "reward_std": 0.21677803993225098,
      "reward_std_all": 0.23041219264268875,
      "rewards-max/forensics_sinkhorn_reward": 0.9626942276954651,
      "rewards-max/forensics_strict_edge_reward": 0.856150895357132,
      "rewards-max/forensics_tiou_reward": 0.9379202425479889,
      "rewards/forensics_sinkhorn_reward": 0.39651452004909515,
      "rewards/forensics_strict_edge_reward": 0.21866993606090546,
      "rewards/forensics_tiou_reward": 0.2773212790489197,
      "step": 406
    },
    {
      "completion_length": 129.84375,
      "epoch": 0.8523560209424084,
      "grad_norm": 15.304925918579102,
      "kl": 3.1171875,
      "learning_rate": 7.872117400419287e-07,
      "loss": 0.1917,
      "reward": 0.554117888212204,
      "reward_std": 0.21120433509349823,
      "reward_std_all": 0.22636019438505173,
      "rewards-max/forensics_sinkhorn_reward": 0.9179509878158569,
      "rewards-max/forensics_strict_edge_reward": 0.8612285256385803,
      "rewards-max/forensics_tiou_reward": 0.8515311181545258,
      "rewards/forensics_sinkhorn_reward": 0.3972240388393402,
      "rewards/forensics_strict_edge_reward": 0.2023628056049347,
      "rewards/forensics_tiou_reward": 0.2547265961766243,
      "step": 407
    },
    {
      "completion_length": 95.21875,
      "epoch": 0.8544502617801047,
      "grad_norm": 14.701972961425781,
      "kl": 2.80078125,
      "learning_rate": 7.866876310272537e-07,
      "loss": 0.442,
      "reward": 0.5831924676895142,
      "reward_std": 0.22169587761163712,
      "reward_std_all": 0.21830471605062485,
      "rewards-max/forensics_sinkhorn_reward": 0.9516638517379761,
      "rewards-max/forensics_strict_edge_reward": 0.7559238970279694,
      "rewards-max/forensics_tiou_reward": 0.9167112112045288,
      "rewards/forensics_sinkhorn_reward": 0.43496568500995636,
      "rewards/forensics_strict_edge_reward": 0.22063259780406952,
      "rewards/forensics_tiou_reward": 0.2690858840942383,
      "step": 408
    },
    {
      "completion_length": 77.03125,
      "epoch": 0.856544502617801,
      "grad_norm": 12.737761497497559,
      "kl": 2.39453125,
      "learning_rate": 7.861635220125787e-07,
      "loss": 0.233,
      "reward": 0.6148298978805542,
      "reward_std": 0.1858227252960205,
      "reward_std_all": 0.18289066851139069,
      "rewards-max/forensics_sinkhorn_reward": 0.9344871640205383,
      "rewards-max/forensics_strict_edge_reward": 0.813967227935791,
      "rewards-max/forensics_tiou_reward": 0.8848628401756287,
      "rewards/forensics_sinkhorn_reward": 0.4385828822851181,
      "rewards/forensics_strict_edge_reward": 0.23157928884029388,
      "rewards/forensics_tiou_reward": 0.27162954211235046,
      "step": 409
    },
    {
      "completion_length": 115.828125,
      "epoch": 0.8586387434554974,
      "grad_norm": 14.094083786010742,
      "kl": 2.9765625,
      "learning_rate": 7.856394129979035e-07,
      "loss": 0.2919,
      "reward": 0.6206653118133545,
      "reward_std": 0.19521313160657883,
      "reward_std_all": 0.18966273963451385,
      "rewards-max/forensics_sinkhorn_reward": 0.9860887229442596,
      "rewards-max/forensics_strict_edge_reward": 0.9355069994926453,
      "rewards-max/forensics_tiou_reward": 0.9885775148868561,
      "rewards/forensics_sinkhorn_reward": 0.47772422432899475,
      "rewards/forensics_strict_edge_reward": 0.23825758695602417,
      "rewards/forensics_tiou_reward": 0.3229776546359062,
      "step": 410
    },
    {
      "completion_length": 144.65625,
      "epoch": 0.8607329842931937,
      "grad_norm": 14.07706069946289,
      "kl": 2.90625,
      "learning_rate": 7.851153039832285e-07,
      "loss": 0.2788,
      "reward": 0.5939099490642548,
      "reward_std": 0.2388705462217331,
      "reward_std_all": 0.22499146312475204,
      "rewards-max/forensics_sinkhorn_reward": 0.9853731691837311,
      "rewards-max/forensics_strict_edge_reward": 0.9062341153621674,
      "rewards-max/forensics_tiou_reward": 0.9858105778694153,
      "rewards/forensics_sinkhorn_reward": 0.42681917548179626,
      "rewards/forensics_strict_edge_reward": 0.26046745479106903,
      "rewards/forensics_tiou_reward": 0.3044567406177521,
      "step": 411
    },
    {
      "completion_length": 158.78125,
      "epoch": 0.86282722513089,
      "grad_norm": 14.110298156738281,
      "kl": 3.3515625,
      "learning_rate": 7.845911949685534e-07,
      "loss": 0.3107,
      "reward": 0.5299096703529358,
      "reward_std": 0.21340441703796387,
      "reward_std_all": 0.234683096408844,
      "rewards-max/forensics_sinkhorn_reward": 0.902547150850296,
      "rewards-max/forensics_strict_edge_reward": 0.879088431596756,
      "rewards-max/forensics_tiou_reward": 0.8228042721748352,
      "rewards/forensics_sinkhorn_reward": 0.4104916751384735,
      "rewards/forensics_strict_edge_reward": 0.328441821038723,
      "rewards/forensics_tiou_reward": 0.35313963890075684,
      "step": 412
    },
    {
      "completion_length": 140.28125,
      "epoch": 0.8649214659685864,
      "grad_norm": 15.181442260742188,
      "kl": 3.015625,
      "learning_rate": 7.840670859538784e-07,
      "loss": 0.258,
      "reward": 0.5433684885501862,
      "reward_std": 0.2705175429582596,
      "reward_std_all": 0.26884445548057556,
      "rewards-max/forensics_sinkhorn_reward": 0.9826875030994415,
      "rewards-max/forensics_strict_edge_reward": 0.8272983729839325,
      "rewards-max/forensics_tiou_reward": 0.9752883315086365,
      "rewards/forensics_sinkhorn_reward": 0.39126521348953247,
      "rewards/forensics_strict_edge_reward": 0.20978078991174698,
      "rewards/forensics_tiou_reward": 0.2531125098466873,
      "step": 413
    },
    {
      "completion_length": 102.8125,
      "epoch": 0.8670157068062827,
      "grad_norm": 14.856815338134766,
      "kl": 2.5234375,
      "learning_rate": 7.835429769392033e-07,
      "loss": 0.2905,
      "reward": 0.5606722831726074,
      "reward_std": 0.2432677447795868,
      "reward_std_all": 0.23786145448684692,
      "rewards-max/forensics_sinkhorn_reward": 0.9863535463809967,
      "rewards-max/forensics_strict_edge_reward": 0.9357379674911499,
      "rewards-max/forensics_tiou_reward": 0.9896236956119537,
      "rewards/forensics_sinkhorn_reward": 0.4260269105434418,
      "rewards/forensics_strict_edge_reward": 0.19928161799907684,
      "rewards/forensics_tiou_reward": 0.25950734317302704,
      "step": 414
    },
    {
      "completion_length": 71.6875,
      "epoch": 0.8691099476439791,
      "grad_norm": 13.30706787109375,
      "kl": 2.00390625,
      "learning_rate": 7.830188679245283e-07,
      "loss": 0.1749,
      "reward": 0.6377361714839935,
      "reward_std": 0.18129894882440567,
      "reward_std_all": 0.17701131105422974,
      "rewards-max/forensics_sinkhorn_reward": 0.941081702709198,
      "rewards-max/forensics_strict_edge_reward": 0.905340164899826,
      "rewards-max/forensics_tiou_reward": 0.8991018831729889,
      "rewards/forensics_sinkhorn_reward": 0.49090541899204254,
      "rewards/forensics_strict_edge_reward": 0.23972052335739136,
      "rewards/forensics_tiou_reward": 0.2791629284620285,
      "step": 415
    },
    {
      "completion_length": 136.921875,
      "epoch": 0.8712041884816754,
      "grad_norm": 11.919058799743652,
      "kl": 3.4140625,
      "learning_rate": 7.824947589098532e-07,
      "loss": 0.3597,
      "reward": 0.60076043009758,
      "reward_std": 0.19977588951587677,
      "reward_std_all": 0.1996944397687912,
      "rewards-max/forensics_sinkhorn_reward": 0.9397499561309814,
      "rewards-max/forensics_strict_edge_reward": 0.7133026123046875,
      "rewards-max/forensics_tiou_reward": 0.8875774145126343,
      "rewards/forensics_sinkhorn_reward": 0.39145389199256897,
      "rewards/forensics_strict_edge_reward": 0.20661979541182518,
      "rewards/forensics_tiou_reward": 0.23912303149700165,
      "step": 416
    },
    {
      "completion_length": 98.421875,
      "epoch": 0.8732984293193717,
      "grad_norm": 12.252182960510254,
      "kl": 2.2109375,
      "learning_rate": 7.819706498951781e-07,
      "loss": 0.3275,
      "reward": 0.5513249039649963,
      "reward_std": 0.21504764258861542,
      "reward_std_all": 0.22304680198431015,
      "rewards-max/forensics_sinkhorn_reward": 0.9877166152000427,
      "rewards-max/forensics_strict_edge_reward": 0.9357379674911499,
      "rewards-max/forensics_tiou_reward": 0.9949863851070404,
      "rewards/forensics_sinkhorn_reward": 0.4357858747243881,
      "rewards/forensics_strict_edge_reward": 0.19284583628177643,
      "rewards/forensics_tiou_reward": 0.2575114741921425,
      "step": 417
    },
    {
      "completion_length": 118.671875,
      "epoch": 0.875392670157068,
      "grad_norm": 11.494681358337402,
      "kl": 1.77734375,
      "learning_rate": 7.814465408805031e-07,
      "loss": 0.3863,
      "reward": 0.5809848010540009,
      "reward_std": 0.21497689932584763,
      "reward_std_all": 0.21873065829277039,
      "rewards-max/forensics_sinkhorn_reward": 0.9877805709838867,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9952436685562134,
      "rewards/forensics_sinkhorn_reward": 0.45724938809871674,
      "rewards/forensics_strict_edge_reward": 0.32128477096557617,
      "rewards/forensics_tiou_reward": 0.3363446891307831,
      "step": 418
    },
    {
      "completion_length": 107.359375,
      "epoch": 0.8774869109947644,
      "grad_norm": 10.142457962036133,
      "kl": 2.28125,
      "learning_rate": 7.80922431865828e-07,
      "loss": 0.1179,
      "reward": 0.5777395069599152,
      "reward_std": 0.24832947552204132,
      "reward_std_all": 0.23782728612422943,
      "rewards-max/forensics_sinkhorn_reward": 0.9646746814250946,
      "rewards-max/forensics_strict_edge_reward": 0.8279296159744263,
      "rewards-max/forensics_tiou_reward": 0.9406187832355499,
      "rewards/forensics_sinkhorn_reward": 0.42148667573928833,
      "rewards/forensics_strict_edge_reward": 0.20596761256456375,
      "rewards/forensics_tiou_reward": 0.24669146537780762,
      "step": 419
    },
    {
      "completion_length": 87.53125,
      "epoch": 0.8795811518324608,
      "grad_norm": 16.585960388183594,
      "kl": 2.0703125,
      "learning_rate": 7.80398322851153e-07,
      "loss": 0.2057,
      "reward": 0.6197946667671204,
      "reward_std": 0.1503291204571724,
      "reward_std_all": 0.16635583341121674,
      "rewards-max/forensics_sinkhorn_reward": 0.9883912801742554,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9976393282413483,
      "rewards/forensics_sinkhorn_reward": 0.5435449779033661,
      "rewards/forensics_strict_edge_reward": 0.32245388627052307,
      "rewards/forensics_tiou_reward": 0.3611764758825302,
      "step": 420
    },
    {
      "completion_length": 142.640625,
      "epoch": 0.881675392670157,
      "grad_norm": 14.304481506347656,
      "kl": 2.9296875,
      "learning_rate": 7.79874213836478e-07,
      "loss": 0.1856,
      "reward": 0.5837804675102234,
      "reward_std": 0.20644421875476837,
      "reward_std_all": 0.20915037393569946,
      "rewards-max/forensics_sinkhorn_reward": 0.8820661902427673,
      "rewards-max/forensics_strict_edge_reward": 0.8104448914527893,
      "rewards-max/forensics_tiou_reward": 0.780484139919281,
      "rewards/forensics_sinkhorn_reward": 0.41557949781417847,
      "rewards/forensics_strict_edge_reward": 0.17198379337787628,
      "rewards/forensics_tiou_reward": 0.24108504503965378,
      "step": 421
    },
    {
      "completion_length": 43.375,
      "epoch": 0.8837696335078534,
      "grad_norm": 11.371292114257812,
      "kl": 1.248046875,
      "learning_rate": 7.793501048218029e-07,
      "loss": 0.1324,
      "reward": 0.6457706689834595,
      "reward_std": 0.11506064422428608,
      "reward_std_all": 0.1335364207625389,
      "rewards-max/forensics_sinkhorn_reward": 0.9882758855819702,
      "rewards-max/forensics_strict_edge_reward": 0.9782643616199493,
      "rewards-max/forensics_tiou_reward": 0.997183084487915,
      "rewards/forensics_sinkhorn_reward": 0.6008560955524445,
      "rewards/forensics_strict_edge_reward": 0.19681556522846222,
      "rewards/forensics_tiou_reward": 0.35708408057689667,
      "step": 422
    },
    {
      "completion_length": 70.40625,
      "epoch": 0.8858638743455497,
      "grad_norm": 13.30132007598877,
      "kl": 1.3359375,
      "learning_rate": 7.788259958071279e-07,
      "loss": 0.2682,
      "reward": 0.5744028985500336,
      "reward_std": 0.18829724192619324,
      "reward_std_all": 0.2027176469564438,
      "rewards-max/forensics_sinkhorn_reward": 0.9874448776245117,
      "rewards-max/forensics_strict_edge_reward": 0.9565287232398987,
      "rewards-max/forensics_tiou_reward": 0.9939074516296387,
      "rewards/forensics_sinkhorn_reward": 0.4426903873682022,
      "rewards/forensics_strict_edge_reward": 0.22895631194114685,
      "rewards/forensics_tiou_reward": 0.266868457198143,
      "step": 423
    },
    {
      "completion_length": 47.015625,
      "epoch": 0.8879581151832461,
      "grad_norm": 25.7752628326416,
      "kl": 0.861328125,
      "learning_rate": 7.783018867924528e-07,
      "loss": 0.3088,
      "reward": 0.6447233259677887,
      "reward_std": 0.18179920688271523,
      "reward_std_all": 0.17520327121019363,
      "rewards-max/forensics_sinkhorn_reward": 0.9861707389354706,
      "rewards-max/forensics_strict_edge_reward": 0.947419673204422,
      "rewards-max/forensics_tiou_reward": 0.9889506101608276,
      "rewards/forensics_sinkhorn_reward": 0.5527328252792358,
      "rewards/forensics_strict_edge_reward": 0.30752329528331757,
      "rewards/forensics_tiou_reward": 0.3503987193107605,
      "step": 424
    },
    {
      "completion_length": 108.25,
      "epoch": 0.8900523560209425,
      "grad_norm": 12.130569458007812,
      "kl": 1.77734375,
      "learning_rate": 7.777777777777778e-07,
      "loss": 0.247,
      "reward": 0.5670804381370544,
      "reward_std": 0.20290522277355194,
      "reward_std_all": 0.22268059104681015,
      "rewards-max/forensics_sinkhorn_reward": 0.9748892188072205,
      "rewards-max/forensics_strict_edge_reward": 0.7524020969867706,
      "rewards-max/forensics_tiou_reward": 0.9553068280220032,
      "rewards/forensics_sinkhorn_reward": 0.38103383779525757,
      "rewards/forensics_strict_edge_reward": 0.17220135778188705,
      "rewards/forensics_tiou_reward": 0.20776832103729248,
      "step": 425
    },
    {
      "completion_length": 65.90625,
      "epoch": 0.8921465968586387,
      "grad_norm": 12.190924644470215,
      "kl": 1.083984375,
      "learning_rate": 7.772536687631027e-07,
      "loss": 0.2083,
      "reward": 0.6243508160114288,
      "reward_std": 0.16966508328914642,
      "reward_std_all": 0.1719335839152336,
      "rewards-max/forensics_sinkhorn_reward": 0.9666498601436615,
      "rewards-max/forensics_strict_edge_reward": 0.947419673204422,
      "rewards-max/forensics_tiou_reward": 0.95108562707901,
      "rewards/forensics_sinkhorn_reward": 0.6029802560806274,
      "rewards/forensics_strict_edge_reward": 0.2585233561694622,
      "rewards/forensics_tiou_reward": 0.40731336176395416,
      "step": 426
    },
    {
      "completion_length": 40.890625,
      "epoch": 0.8942408376963351,
      "grad_norm": 14.997766494750977,
      "kl": 0.95703125,
      "learning_rate": 7.767295597484277e-07,
      "loss": 0.0248,
      "reward": 0.6206291615962982,
      "reward_std": 0.13807088136672974,
      "reward_std_all": 0.16779667139053345,
      "rewards-max/forensics_sinkhorn_reward": 0.9877331256866455,
      "rewards-max/forensics_strict_edge_reward": 0.946017861366272,
      "rewards-max/forensics_tiou_reward": 0.9950497448444366,
      "rewards/forensics_sinkhorn_reward": 0.4871816635131836,
      "rewards/forensics_strict_edge_reward": 0.25866667926311493,
      "rewards/forensics_tiou_reward": 0.2819610461592674,
      "step": 427
    },
    {
      "completion_length": 77.296875,
      "epoch": 0.8963350785340314,
      "grad_norm": 15.507702827453613,
      "kl": 1.951171875,
      "learning_rate": 7.762054507337525e-07,
      "loss": 0.392,
      "reward": 0.6123547554016113,
      "reward_std": 0.13262975215911865,
      "reward_std_all": 0.16289466619491577,
      "rewards-max/forensics_sinkhorn_reward": 0.9878730475902557,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9955998361110687,
      "rewards/forensics_sinkhorn_reward": 0.49583807587623596,
      "rewards/forensics_strict_edge_reward": 0.26260754466056824,
      "rewards/forensics_tiou_reward": 0.3240896090865135,
      "step": 428
    },
    {
      "completion_length": 87.25,
      "epoch": 0.8984293193717278,
      "grad_norm": 14.87033462524414,
      "kl": 2.01171875,
      "learning_rate": 7.756813417190775e-07,
      "loss": 0.3037,
      "reward": 0.6139348447322845,
      "reward_std": 0.21250928193330765,
      "reward_std_all": 0.20596018433570862,
      "rewards-max/forensics_sinkhorn_reward": 0.9796075224876404,
      "rewards-max/forensics_strict_edge_reward": 0.9265981018543243,
      "rewards-max/forensics_tiou_reward": 0.9783354997634888,
      "rewards/forensics_sinkhorn_reward": 0.512520432472229,
      "rewards/forensics_strict_edge_reward": 0.33241933584213257,
      "rewards/forensics_tiou_reward": 0.4025065749883652,
      "step": 429
    },
    {
      "completion_length": 177.046875,
      "epoch": 0.900523560209424,
      "grad_norm": 18.864437103271484,
      "kl": 3.8203125,
      "learning_rate": 7.751572327044024e-07,
      "loss": 0.5122,
      "reward": 0.483277827501297,
      "reward_std": 0.18476951867341995,
      "reward_std_all": 0.24151311814785004,
      "rewards-max/forensics_sinkhorn_reward": 0.9632421433925629,
      "rewards-max/forensics_strict_edge_reward": 0.8404377698898315,
      "rewards-max/forensics_tiou_reward": 0.939397931098938,
      "rewards/forensics_sinkhorn_reward": 0.30772799253463745,
      "rewards/forensics_strict_edge_reward": 0.19759443402290344,
      "rewards/forensics_tiou_reward": 0.24071700870990753,
      "step": 430
    },
    {
      "completion_length": 126.125,
      "epoch": 0.9026178010471204,
      "grad_norm": 16.64021110534668,
      "kl": 2.04296875,
      "learning_rate": 7.746331236897274e-07,
      "loss": 0.1604,
      "reward": 0.5319801568984985,
      "reward_std": 0.24385132640600204,
      "reward_std_all": 0.2573131173849106,
      "rewards-max/forensics_sinkhorn_reward": 0.9533178210258484,
      "rewards-max/forensics_strict_edge_reward": 0.8245616853237152,
      "rewards-max/forensics_tiou_reward": 0.9164555966854095,
      "rewards/forensics_sinkhorn_reward": 0.43833136558532715,
      "rewards/forensics_strict_edge_reward": 0.17642192542552948,
      "rewards/forensics_tiou_reward": 0.25390930473804474,
      "step": 431
    },
    {
      "completion_length": 149.109375,
      "epoch": 0.9047120418848168,
      "grad_norm": 11.91620922088623,
      "kl": 2.34375,
      "learning_rate": 7.741090146750524e-07,
      "loss": 0.2847,
      "reward": 0.5436276495456696,
      "reward_std": 0.2676905691623688,
      "reward_std_all": 0.26644524931907654,
      "rewards-max/forensics_sinkhorn_reward": 0.9626915156841278,
      "rewards-max/forensics_strict_edge_reward": 0.8719756007194519,
      "rewards-max/forensics_tiou_reward": 0.9387157261371613,
      "rewards/forensics_sinkhorn_reward": 0.4237435758113861,
      "rewards/forensics_strict_edge_reward": 0.2391504868865013,
      "rewards/forensics_tiou_reward": 0.272327184677124,
      "step": 432
    },
    {
      "completion_length": 137.796875,
      "epoch": 0.9068062827225131,
      "grad_norm": 20.674121856689453,
      "kl": 3.734375,
      "learning_rate": 7.735849056603774e-07,
      "loss": 0.2781,
      "reward": 0.500830352306366,
      "reward_std": 0.24441150575876236,
      "reward_std_all": 0.2682308256626129,
      "rewards-max/forensics_sinkhorn_reward": 0.9837791621685028,
      "rewards-max/forensics_strict_edge_reward": 0.9062341153621674,
      "rewards-max/forensics_tiou_reward": 0.9796079397201538,
      "rewards/forensics_sinkhorn_reward": 0.39041316509246826,
      "rewards/forensics_strict_edge_reward": 0.20041204988956451,
      "rewards/forensics_tiou_reward": 0.2708417773246765,
      "step": 433
    },
    {
      "completion_length": 176.546875,
      "epoch": 0.9089005235602095,
      "grad_norm": 14.913431167602539,
      "kl": 3.7734375,
      "learning_rate": 7.730607966457023e-07,
      "loss": 0.3867,
      "reward": 0.3969699442386627,
      "reward_std": 0.23866062611341476,
      "reward_std_all": 0.30727754533290863,
      "rewards-max/forensics_sinkhorn_reward": 0.9880331456661224,
      "rewards-max/forensics_strict_edge_reward": 0.9782643616199493,
      "rewards-max/forensics_tiou_reward": 0.99622642993927,
      "rewards/forensics_sinkhorn_reward": 0.34593284130096436,
      "rewards/forensics_strict_edge_reward": 0.1935632899403572,
      "rewards/forensics_tiou_reward": 0.25386694073677063,
      "step": 434
    },
    {
      "completion_length": 181.78125,
      "epoch": 0.9109947643979057,
      "grad_norm": 16.343835830688477,
      "kl": 2.6796875,
      "learning_rate": 7.725366876310272e-07,
      "loss": 0.427,
      "reward": 0.4480421245098114,
      "reward_std": 0.21197354048490524,
      "reward_std_all": 0.2759028375148773,
      "rewards-max/forensics_sinkhorn_reward": 0.9813677966594696,
      "rewards-max/forensics_strict_edge_reward": 0.9355069994926453,
      "rewards-max/forensics_tiou_reward": 0.9833773970603943,
      "rewards/forensics_sinkhorn_reward": 0.3361207991838455,
      "rewards/forensics_strict_edge_reward": 0.20464487373828888,
      "rewards/forensics_tiou_reward": 0.25563502311706543,
      "step": 435
    },
    {
      "completion_length": 120.96875,
      "epoch": 0.9130890052356021,
      "grad_norm": 14.232275009155273,
      "kl": 1.8359375,
      "learning_rate": 7.720125786163522e-07,
      "loss": 0.2923,
      "reward": 0.5909488201141357,
      "reward_std": 0.174615278840065,
      "reward_std_all": 0.1899401992559433,
      "rewards-max/forensics_sinkhorn_reward": 0.9871909618377686,
      "rewards-max/forensics_strict_edge_reward": 0.9574736058712006,
      "rewards-max/forensics_tiou_reward": 0.9929329454898834,
      "rewards/forensics_sinkhorn_reward": 0.5350528359413147,
      "rewards/forensics_strict_edge_reward": 0.3057873323559761,
      "rewards/forensics_tiou_reward": 0.4518412798643112,
      "step": 436
    },
    {
      "completion_length": 133.03125,
      "epoch": 0.9151832460732985,
      "grad_norm": 13.240866661071777,
      "kl": 2.6484375,
      "learning_rate": 7.714884696016771e-07,
      "loss": 0.2863,
      "reward": 0.5346865355968475,
      "reward_std": 0.21849191188812256,
      "reward_std_all": 0.236688032746315,
      "rewards-max/forensics_sinkhorn_reward": 0.9882758855819702,
      "rewards-max/forensics_strict_edge_reward": 0.9782643616199493,
      "rewards-max/forensics_tiou_reward": 0.997183084487915,
      "rewards/forensics_sinkhorn_reward": 0.4287244379520416,
      "rewards/forensics_strict_edge_reward": 0.23379945755004883,
      "rewards/forensics_tiou_reward": 0.2702134698629379,
      "step": 437
    },
    {
      "completion_length": 80.546875,
      "epoch": 0.9172774869109948,
      "grad_norm": 14.8170747756958,
      "kl": 1.203125,
      "learning_rate": 7.709643605870021e-07,
      "loss": 0.1079,
      "reward": 0.6286361813545227,
      "reward_std": 0.1852591186761856,
      "reward_std_all": 0.18890568614006042,
      "rewards-max/forensics_sinkhorn_reward": 0.9854259192943573,
      "rewards-max/forensics_strict_edge_reward": 0.9364311099052429,
      "rewards-max/forensics_tiou_reward": 0.9860230088233948,
      "rewards/forensics_sinkhorn_reward": 0.5475422143936157,
      "rewards/forensics_strict_edge_reward": 0.3583173602819443,
      "rewards/forensics_tiou_reward": 0.42501018941402435,
      "step": 438
    },
    {
      "completion_length": 121.921875,
      "epoch": 0.9193717277486911,
      "grad_norm": 18.770959854125977,
      "kl": 1.6953125,
      "learning_rate": 7.704402515723271e-07,
      "loss": 0.0987,
      "reward": 0.4590698331594467,
      "reward_std": 0.203102208673954,
      "reward_std_all": 0.2619708776473999,
      "rewards-max/forensics_sinkhorn_reward": 0.9805993437767029,
      "rewards-max/forensics_strict_edge_reward": 0.8096016347408295,
      "rewards-max/forensics_tiou_reward": 0.9674675464630127,
      "rewards/forensics_sinkhorn_reward": 0.3603529930114746,
      "rewards/forensics_strict_edge_reward": 0.15937282890081406,
      "rewards/forensics_tiou_reward": 0.18985281139612198,
      "step": 439
    },
    {
      "completion_length": 144.015625,
      "epoch": 0.9214659685863874,
      "grad_norm": 16.6231689453125,
      "kl": 2.265625,
      "learning_rate": 7.699161425576519e-07,
      "loss": 0.2637,
      "reward": 0.5105584263801575,
      "reward_std": 0.25489941239356995,
      "reward_std_all": 0.2633126676082611,
      "rewards-max/forensics_sinkhorn_reward": 0.9864113330841064,
      "rewards-max/forensics_strict_edge_reward": 0.9364311099052429,
      "rewards-max/forensics_tiou_reward": 0.9898715615272522,
      "rewards/forensics_sinkhorn_reward": 0.3630937337875366,
      "rewards/forensics_strict_edge_reward": 0.22938795387744904,
      "rewards/forensics_tiou_reward": 0.25616856664419174,
      "step": 440
    },
    {
      "completion_length": 133.90625,
      "epoch": 0.9235602094240838,
      "grad_norm": 14.834274291992188,
      "kl": 1.9375,
      "learning_rate": 7.693920335429769e-07,
      "loss": 0.3811,
      "reward": 0.531988263130188,
      "reward_std": 0.24083011597394943,
      "reward_std_all": 0.2590770572423935,
      "rewards-max/forensics_sinkhorn_reward": 0.9821684956550598,
      "rewards-max/forensics_strict_edge_reward": 0.8320213258266449,
      "rewards-max/forensics_tiou_reward": 0.9734584987163544,
      "rewards/forensics_sinkhorn_reward": 0.4067903459072113,
      "rewards/forensics_strict_edge_reward": 0.23607779294252396,
      "rewards/forensics_tiou_reward": 0.31130731105804443,
      "step": 441
    },
    {
      "completion_length": 66.875,
      "epoch": 0.9256544502617801,
      "grad_norm": 19.951261520385742,
      "kl": 1.400390625,
      "learning_rate": 7.688679245283018e-07,
      "loss": 0.2209,
      "reward": 0.5393068790435791,
      "reward_std": 0.1891845166683197,
      "reward_std_all": 0.21554695069789886,
      "rewards-max/forensics_sinkhorn_reward": 0.9872284531593323,
      "rewards-max/forensics_strict_edge_reward": 0.9890114367008209,
      "rewards-max/forensics_tiou_reward": 0.993079662322998,
      "rewards/forensics_sinkhorn_reward": 0.4219745397567749,
      "rewards/forensics_strict_edge_reward": 0.1921483427286148,
      "rewards/forensics_tiou_reward": 0.24526037275791168,
      "step": 442
    },
    {
      "completion_length": 122.640625,
      "epoch": 0.9277486910994764,
      "grad_norm": 12.63475227355957,
      "kl": 2.359375,
      "learning_rate": 7.683438155136268e-07,
      "loss": 0.2835,
      "reward": 0.5248094201087952,
      "reward_std": 0.16396040469408035,
      "reward_std_all": 0.21559526026248932,
      "rewards-max/forensics_sinkhorn_reward": 0.8465852737426758,
      "rewards-max/forensics_strict_edge_reward": 0.7351652383804321,
      "rewards-max/forensics_tiou_reward": 0.7687425017356873,
      "rewards/forensics_sinkhorn_reward": 0.3521965593099594,
      "rewards/forensics_strict_edge_reward": 0.17719732224941254,
      "rewards/forensics_tiou_reward": 0.2248123288154602,
      "step": 443
    },
    {
      "completion_length": 172.265625,
      "epoch": 0.9298429319371728,
      "grad_norm": 10.94791316986084,
      "kl": 2.3515625,
      "learning_rate": 7.678197064989517e-07,
      "loss": 0.3485,
      "reward": 0.48204857110977173,
      "reward_std": 0.20881588011980057,
      "reward_std_all": 0.265232153236866,
      "rewards-max/forensics_sinkhorn_reward": 0.9881765246391296,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9967943727970123,
      "rewards/forensics_sinkhorn_reward": 0.36055852472782135,
      "rewards/forensics_strict_edge_reward": 0.2477036565542221,
      "rewards/forensics_tiou_reward": 0.26470471918582916,
      "step": 444
    },
    {
      "completion_length": 112.953125,
      "epoch": 0.9319371727748691,
      "grad_norm": 10.158624649047852,
      "kl": 2.28125,
      "learning_rate": 7.672955974842768e-07,
      "loss": 0.2848,
      "reward": 0.5743838846683502,
      "reward_std": 0.16567741334438324,
      "reward_std_all": 0.18298811465501785,
      "rewards-max/forensics_sinkhorn_reward": 0.9859978258609772,
      "rewards-max/forensics_strict_edge_reward": 0.7942859828472137,
      "rewards-max/forensics_tiou_reward": 0.9882112741470337,
      "rewards/forensics_sinkhorn_reward": 0.39143380522727966,
      "rewards/forensics_strict_edge_reward": 0.1662946119904518,
      "rewards/forensics_tiou_reward": 0.2432834357023239,
      "step": 445
    },
    {
      "completion_length": 137.75,
      "epoch": 0.9340314136125655,
      "grad_norm": 14.72571849822998,
      "kl": 2.546875,
      "learning_rate": 7.667714884696016e-07,
      "loss": 0.2757,
      "reward": 0.42219698429107666,
      "reward_std": 0.2937542051076889,
      "reward_std_all": 0.3198237270116806,
      "rewards-max/forensics_sinkhorn_reward": 0.9258266091346741,
      "rewards-max/forensics_strict_edge_reward": 0.7082041949033737,
      "rewards-max/forensics_tiou_reward": 0.8692921102046967,
      "rewards/forensics_sinkhorn_reward": 0.31768885254859924,
      "rewards/forensics_strict_edge_reward": 0.1025194562971592,
      "rewards/forensics_tiou_reward": 0.16657442972064018,
      "step": 446
    },
    {
      "completion_length": 91.46875,
      "epoch": 0.9361256544502617,
      "grad_norm": 12.793394088745117,
      "kl": 1.92578125,
      "learning_rate": 7.662473794549266e-07,
      "loss": 0.2451,
      "reward": 0.5593525767326355,
      "reward_std": 0.2502925544977188,
      "reward_std_all": 0.24260711669921875,
      "rewards-max/forensics_sinkhorn_reward": 0.9475598037242889,
      "rewards-max/forensics_strict_edge_reward": 0.771587461233139,
      "rewards-max/forensics_tiou_reward": 0.9057275354862213,
      "rewards/forensics_sinkhorn_reward": 0.4422003924846649,
      "rewards/forensics_strict_edge_reward": 0.25073087960481644,
      "rewards/forensics_tiou_reward": 0.3083120286464691,
      "step": 447
    },
    {
      "completion_length": 104.765625,
      "epoch": 0.9382198952879581,
      "grad_norm": 12.260875701904297,
      "kl": 2.6015625,
      "learning_rate": 7.657232704402516e-07,
      "loss": 0.1763,
      "reward": 0.4938242882490158,
      "reward_std": 0.24512305110692978,
      "reward_std_all": 0.2672649770975113,
      "rewards-max/forensics_sinkhorn_reward": 0.9780106544494629,
      "rewards-max/forensics_strict_edge_reward": 0.47287894785404205,
      "rewards-max/forensics_tiou_reward": 0.9572742283344269,
      "rewards/forensics_sinkhorn_reward": 0.3446503281593323,
      "rewards/forensics_strict_edge_reward": 0.12957629561424255,
      "rewards/forensics_tiou_reward": 0.1982990950345993,
      "step": 448
    },
    {
      "completion_length": 104.21875,
      "epoch": 0.9403141361256544,
      "grad_norm": 15.580638885498047,
      "kl": 2.04296875,
      "learning_rate": 7.651991614255765e-07,
      "loss": 0.1926,
      "reward": 0.5163750052452087,
      "reward_std": 0.26591066271066666,
      "reward_std_all": 0.2653428763151169,
      "rewards-max/forensics_sinkhorn_reward": 0.9598134458065033,
      "rewards-max/forensics_strict_edge_reward": 0.7833328545093536,
      "rewards-max/forensics_tiou_reward": 0.9304880201816559,
      "rewards/forensics_sinkhorn_reward": 0.3604082018136978,
      "rewards/forensics_strict_edge_reward": 0.1665596291422844,
      "rewards/forensics_tiou_reward": 0.21680450439453125,
      "step": 449
    },
    {
      "completion_length": 120.890625,
      "epoch": 0.9424083769633508,
      "grad_norm": 17.41847038269043,
      "kl": 2.2734375,
      "learning_rate": 7.646750524109015e-07,
      "loss": 0.0978,
      "reward": 0.4973045885562897,
      "reward_std": 0.22935910522937775,
      "reward_std_all": 0.25813955813646317,
      "rewards-max/forensics_sinkhorn_reward": 0.9741223454475403,
      "rewards-max/forensics_strict_edge_reward": 0.8277941644191742,
      "rewards-max/forensics_tiou_reward": 0.9675365686416626,
      "rewards/forensics_sinkhorn_reward": 0.3524785339832306,
      "rewards/forensics_strict_edge_reward": 0.1332421563565731,
      "rewards/forensics_tiou_reward": 0.17393521592020988,
      "step": 450
    },
    {
      "completion_length": 80.03125,
      "epoch": 0.9445026178010472,
      "grad_norm": 14.47159481048584,
      "kl": 2.27734375,
      "learning_rate": 7.641509433962264e-07,
      "loss": 0.2873,
      "reward": 0.5234267115592957,
      "reward_std": 0.2152903899550438,
      "reward_std_all": 0.2427402287721634,
      "rewards-max/forensics_sinkhorn_reward": 0.9624428153038025,
      "rewards-max/forensics_strict_edge_reward": 0.8863177299499512,
      "rewards-max/forensics_tiou_reward": 0.9406877160072327,
      "rewards/forensics_sinkhorn_reward": 0.3449386656284332,
      "rewards/forensics_strict_edge_reward": 0.2286715805530548,
      "rewards/forensics_tiou_reward": 0.24236633628606796,
      "step": 451
    },
    {
      "completion_length": 123.515625,
      "epoch": 0.9465968586387434,
      "grad_norm": 17.94763946533203,
      "kl": 2.9140625,
      "learning_rate": 7.636268343815513e-07,
      "loss": 0.3853,
      "reward": 0.4664240777492523,
      "reward_std": 0.2607395797967911,
      "reward_std_all": 0.2951558530330658,
      "rewards-max/forensics_sinkhorn_reward": 0.9849023520946503,
      "rewards-max/forensics_strict_edge_reward": 0.9357379674911499,
      "rewards-max/forensics_tiou_reward": 0.9839538633823395,
      "rewards/forensics_sinkhorn_reward": 0.321723535656929,
      "rewards/forensics_strict_edge_reward": 0.17962609231472015,
      "rewards/forensics_tiou_reward": 0.20186546444892883,
      "step": 452
    },
    {
      "completion_length": 80.390625,
      "epoch": 0.9486910994764398,
      "grad_norm": 11.47034740447998,
      "kl": 2.1015625,
      "learning_rate": 7.631027253668762e-07,
      "loss": 0.3418,
      "reward": 0.5614511966705322,
      "reward_std": 0.2127850279211998,
      "reward_std_all": 0.2257012575864792,
      "rewards-max/forensics_sinkhorn_reward": 0.9740300476551056,
      "rewards-max/forensics_strict_edge_reward": 0.8582656681537628,
      "rewards-max/forensics_tiou_reward": 0.9682427942752838,
      "rewards/forensics_sinkhorn_reward": 0.4427197426557541,
      "rewards/forensics_strict_edge_reward": 0.23154937475919724,
      "rewards/forensics_tiou_reward": 0.34237444400787354,
      "step": 453
    },
    {
      "completion_length": 79.671875,
      "epoch": 0.9507853403141361,
      "grad_norm": 14.320273399353027,
      "kl": 2.125,
      "learning_rate": 7.625786163522012e-07,
      "loss": 0.2222,
      "reward": 0.5698319673538208,
      "reward_std": 0.22796151041984558,
      "reward_std_all": 0.23407232016324997,
      "rewards-max/forensics_sinkhorn_reward": 0.9841439127922058,
      "rewards-max/forensics_strict_edge_reward": 0.8205508291721344,
      "rewards-max/forensics_tiou_reward": 0.9810520708560944,
      "rewards/forensics_sinkhorn_reward": 0.4214150607585907,
      "rewards/forensics_strict_edge_reward": 0.2496020644903183,
      "rewards/forensics_tiou_reward": 0.27950164675712585,
      "step": 454
    },
    {
      "completion_length": 90.015625,
      "epoch": 0.9528795811518325,
      "grad_norm": 13.00110912322998,
      "kl": 1.68359375,
      "learning_rate": 7.620545073375262e-07,
      "loss": 0.3333,
      "reward": 0.566078782081604,
      "reward_std": 0.22518625855445862,
      "reward_std_all": 0.22681721299886703,
      "rewards-max/forensics_sinkhorn_reward": 0.9856487512588501,
      "rewards-max/forensics_strict_edge_reward": 0.9075756669044495,
      "rewards-max/forensics_tiou_reward": 0.9868952333927155,
      "rewards/forensics_sinkhorn_reward": 0.44829580187797546,
      "rewards/forensics_strict_edge_reward": 0.2458069622516632,
      "rewards/forensics_tiou_reward": 0.30923159420490265,
      "step": 455
    },
    {
      "completion_length": 80.375,
      "epoch": 0.9549738219895288,
      "grad_norm": 13.874382019042969,
      "kl": 1.51953125,
      "learning_rate": 7.615303983228511e-07,
      "loss": 0.2057,
      "reward": 0.5749610066413879,
      "reward_std": 0.2194889336824417,
      "reward_std_all": 0.22306112945079803,
      "rewards-max/forensics_sinkhorn_reward": 0.98213130235672,
      "rewards-max/forensics_strict_edge_reward": 0.8612285256385803,
      "rewards-max/forensics_tiou_reward": 0.973285973072052,
      "rewards/forensics_sinkhorn_reward": 0.45478153228759766,
      "rewards/forensics_strict_edge_reward": 0.2662525102496147,
      "rewards/forensics_tiou_reward": 0.30599892139434814,
      "step": 456
    },
    {
      "completion_length": 92.0,
      "epoch": 0.9570680628272251,
      "grad_norm": 14.861980438232422,
      "kl": 1.5078125,
      "learning_rate": 7.610062893081762e-07,
      "loss": 0.1553,
      "reward": 0.5515971779823303,
      "reward_std": 0.252630814909935,
      "reward_std_all": 0.24885860830545425,
      "rewards-max/forensics_sinkhorn_reward": 0.9496592283248901,
      "rewards-max/forensics_strict_edge_reward": 0.7451611459255219,
      "rewards-max/forensics_tiou_reward": 0.9083175659179688,
      "rewards/forensics_sinkhorn_reward": 0.42147956788539886,
      "rewards/forensics_strict_edge_reward": 0.21137940138578415,
      "rewards/forensics_tiou_reward": 0.27336128801107407,
      "step": 457
    },
    {
      "completion_length": 48.921875,
      "epoch": 0.9591623036649215,
      "grad_norm": 13.936836242675781,
      "kl": 0.849609375,
      "learning_rate": 7.60482180293501e-07,
      "loss": 0.075,
      "reward": 0.6532242894172668,
      "reward_std": 0.1977081149816513,
      "reward_std_all": 0.18117888271808624,
      "rewards-max/forensics_sinkhorn_reward": 0.9880402386188507,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9962567389011383,
      "rewards/forensics_sinkhorn_reward": 0.577132374048233,
      "rewards/forensics_strict_edge_reward": 0.28090064227581024,
      "rewards/forensics_tiou_reward": 0.4165942370891571,
      "step": 458
    },
    {
      "completion_length": 99.03125,
      "epoch": 0.9612565445026178,
      "grad_norm": 14.526921272277832,
      "kl": 1.3046875,
      "learning_rate": 7.59958071278826e-07,
      "loss": 0.0523,
      "reward": 0.5881654620170593,
      "reward_std": 0.20759496092796326,
      "reward_std_all": 0.21083085983991623,
      "rewards-max/forensics_sinkhorn_reward": 0.9861592650413513,
      "rewards-max/forensics_strict_edge_reward": 0.8279296159744263,
      "rewards-max/forensics_tiou_reward": 0.988873690366745,
      "rewards/forensics_sinkhorn_reward": 0.4545868784189224,
      "rewards/forensics_strict_edge_reward": 0.1819036304950714,
      "rewards/forensics_tiou_reward": 0.2589777708053589,
      "step": 459
    },
    {
      "completion_length": 131.015625,
      "epoch": 0.9633507853403142,
      "grad_norm": 15.420987129211426,
      "kl": 1.9140625,
      "learning_rate": 7.594339622641509e-07,
      "loss": 0.1763,
      "reward": 0.6054220199584961,
      "reward_std": 0.2070780247449875,
      "reward_std_all": 0.20312388241291046,
      "rewards-max/forensics_sinkhorn_reward": 0.9803460836410522,
      "rewards-max/forensics_strict_edge_reward": 0.9048932790756226,
      "rewards-max/forensics_tiou_reward": 0.9788707196712494,
      "rewards/forensics_sinkhorn_reward": 0.43186116218566895,
      "rewards/forensics_strict_edge_reward": 0.2215823382139206,
      "rewards/forensics_tiou_reward": 0.2692560479044914,
      "step": 460
    },
    {
      "completion_length": 76.125,
      "epoch": 0.9654450261780104,
      "grad_norm": 12.487428665161133,
      "kl": 1.3671875,
      "learning_rate": 7.589098532494759e-07,
      "loss": 0.3329,
      "reward": 0.5691487491130829,
      "reward_std": 0.17863933742046356,
      "reward_std_all": 0.19557157158851624,
      "rewards-max/forensics_sinkhorn_reward": 0.9872441589832306,
      "rewards-max/forensics_strict_edge_reward": 0.9464850425720215,
      "rewards-max/forensics_tiou_reward": 0.9931344985961914,
      "rewards/forensics_sinkhorn_reward": 0.4661075919866562,
      "rewards/forensics_strict_edge_reward": 0.2668994292616844,
      "rewards/forensics_tiou_reward": 0.3437136337161064,
      "step": 461
    },
    {
      "completion_length": 53.359375,
      "epoch": 0.9675392670157068,
      "grad_norm": 15.351186752319336,
      "kl": 1.83203125,
      "learning_rate": 7.583857442348009e-07,
      "loss": 0.2063,
      "reward": 0.5447511076927185,
      "reward_std": 0.20327218621969223,
      "reward_std_all": 0.22050204873085022,
      "rewards-max/forensics_sinkhorn_reward": 0.979045182466507,
      "rewards-max/forensics_strict_edge_reward": 0.9149472117424011,
      "rewards-max/forensics_tiou_reward": 0.9751504957675934,
      "rewards/forensics_sinkhorn_reward": 0.4354533553123474,
      "rewards/forensics_strict_edge_reward": 0.16472211480140686,
      "rewards/forensics_tiou_reward": 0.2373000755906105,
      "step": 462
    },
    {
      "completion_length": 108.921875,
      "epoch": 0.9696335078534032,
      "grad_norm": 15.1820707321167,
      "kl": 2.3125,
      "learning_rate": 7.578616352201258e-07,
      "loss": 0.2737,
      "reward": 0.5632918775081635,
      "reward_std": 0.19223912805318832,
      "reward_std_all": 0.2101818025112152,
      "rewards-max/forensics_sinkhorn_reward": 0.9871875047683716,
      "rewards-max/forensics_strict_edge_reward": 0.9565287232398987,
      "rewards-max/forensics_tiou_reward": 0.9928986430168152,
      "rewards/forensics_sinkhorn_reward": 0.4805816411972046,
      "rewards/forensics_strict_edge_reward": 0.2671835273504257,
      "rewards/forensics_tiou_reward": 0.34681086242198944,
      "step": 463
    },
    {
      "completion_length": 62.9375,
      "epoch": 0.9717277486910995,
      "grad_norm": 17.49507713317871,
      "kl": 1.65625,
      "learning_rate": 7.573375262054507e-07,
      "loss": 0.2201,
      "reward": 0.6257382035255432,
      "reward_std": 0.1699155643582344,
      "reward_std_all": 0.17849862575531006,
      "rewards-max/forensics_sinkhorn_reward": 0.9758400022983551,
      "rewards-max/forensics_strict_edge_reward": 0.8582656681537628,
      "rewards-max/forensics_tiou_reward": 0.9496135413646698,
      "rewards/forensics_sinkhorn_reward": 0.5252204090356827,
      "rewards/forensics_strict_edge_reward": 0.2736380994319916,
      "rewards/forensics_tiou_reward": 0.3522983491420746,
      "step": 464
    },
    {
      "completion_length": 94.71875,
      "epoch": 0.9738219895287958,
      "grad_norm": 15.135915756225586,
      "kl": 1.83984375,
      "learning_rate": 7.568134171907756e-07,
      "loss": 0.1863,
      "reward": 0.5679442882537842,
      "reward_std": 0.22194059938192368,
      "reward_std_all": 0.2202906310558319,
      "rewards-max/forensics_sinkhorn_reward": 0.9451356828212738,
      "rewards-max/forensics_strict_edge_reward": 0.7879786789417267,
      "rewards-max/forensics_tiou_reward": 0.9009842276573181,
      "rewards/forensics_sinkhorn_reward": 0.40974652767181396,
      "rewards/forensics_strict_edge_reward": 0.14707373082637787,
      "rewards/forensics_tiou_reward": 0.2321183681488037,
      "step": 465
    },
    {
      "completion_length": 42.421875,
      "epoch": 0.9759162303664921,
      "grad_norm": 13.879983901977539,
      "kl": 1.59375,
      "learning_rate": 7.562893081761006e-07,
      "loss": 0.1879,
      "reward": 0.6091578602790833,
      "reward_std": 0.1290205381810665,
      "reward_std_all": 0.15890851616859436,
      "rewards-max/forensics_sinkhorn_reward": 0.9864600598812103,
      "rewards-max/forensics_strict_edge_reward": 0.927969753742218,
      "rewards-max/forensics_tiou_reward": 0.9900782108306885,
      "rewards/forensics_sinkhorn_reward": 0.5369873195886612,
      "rewards/forensics_strict_edge_reward": 0.24063555896282196,
      "rewards/forensics_tiou_reward": 0.34316663444042206,
      "step": 466
    },
    {
      "completion_length": 60.75,
      "epoch": 0.9780104712041885,
      "grad_norm": 16.169992446899414,
      "kl": 1.46875,
      "learning_rate": 7.557651991614255e-07,
      "loss": 0.07,
      "reward": 0.6062722206115723,
      "reward_std": 0.18325234204530716,
      "reward_std_all": 0.18290215730667114,
      "rewards-max/forensics_sinkhorn_reward": 0.983897864818573,
      "rewards-max/forensics_strict_edge_reward": 0.8612285256385803,
      "rewards-max/forensics_tiou_reward": 0.9799432158470154,
      "rewards/forensics_sinkhorn_reward": 0.4601813405752182,
      "rewards/forensics_strict_edge_reward": 0.1641000658273697,
      "rewards/forensics_tiou_reward": 0.20882093161344528,
      "step": 467
    },
    {
      "completion_length": 97.328125,
      "epoch": 0.9801047120418848,
      "grad_norm": 18.052778244018555,
      "kl": 1.52734375,
      "learning_rate": 7.552410901467505e-07,
      "loss": 0.324,
      "reward": 0.6073826551437378,
      "reward_std": 0.16601119190454483,
      "reward_std_all": 0.17812670767307281,
      "rewards-max/forensics_sinkhorn_reward": 0.9327668249607086,
      "rewards-max/forensics_strict_edge_reward": 0.6407220363616943,
      "rewards-max/forensics_tiou_reward": 0.8237121403217316,
      "rewards/forensics_sinkhorn_reward": 0.44055864214897156,
      "rewards/forensics_strict_edge_reward": 0.1427854187786579,
      "rewards/forensics_tiou_reward": 0.2724025771021843,
      "step": 468
    },
    {
      "completion_length": 58.03125,
      "epoch": 0.9821989528795811,
      "grad_norm": 19.878339767456055,
      "kl": 1.484375,
      "learning_rate": 7.547169811320754e-07,
      "loss": 0.1781,
      "reward": 0.5999597907066345,
      "reward_std": 0.1475737988948822,
      "reward_std_all": 0.16939260065555573,
      "rewards-max/forensics_sinkhorn_reward": 0.8868889510631561,
      "rewards-max/forensics_strict_edge_reward": 0.7573071718215942,
      "rewards-max/forensics_tiou_reward": 0.7759947180747986,
      "rewards/forensics_sinkhorn_reward": 0.4512399435043335,
      "rewards/forensics_strict_edge_reward": 0.12475087866187096,
      "rewards/forensics_tiou_reward": 0.20781827718019485,
      "step": 469
    },
    {
      "completion_length": 71.125,
      "epoch": 0.9842931937172775,
      "grad_norm": 26.25996208190918,
      "kl": 2.1484375,
      "learning_rate": 7.541928721174003e-07,
      "loss": 0.2606,
      "reward": 0.607377827167511,
      "reward_std": 0.15535932779312134,
      "reward_std_all": 0.17148490250110626,
      "rewards-max/forensics_sinkhorn_reward": 0.9612360000610352,
      "rewards-max/forensics_strict_edge_reward": 0.8911538422107697,
      "rewards-max/forensics_tiou_reward": 0.9337088167667389,
      "rewards/forensics_sinkhorn_reward": 0.46978968381881714,
      "rewards/forensics_strict_edge_reward": 0.21019890159368515,
      "rewards/forensics_tiou_reward": 0.2894282042980194,
      "step": 470
    },
    {
      "completion_length": 63.671875,
      "epoch": 0.9863874345549738,
      "grad_norm": 19.821361541748047,
      "kl": 1.9375,
      "learning_rate": 7.536687631027254e-07,
      "loss": 0.1901,
      "reward": 0.6187548637390137,
      "reward_std": 0.17062340676784515,
      "reward_std_all": 0.17709749937057495,
      "rewards-max/forensics_sinkhorn_reward": 0.9886393845081329,
      "rewards-max/forensics_strict_edge_reward": 0.9574736058712006,
      "rewards-max/forensics_tiou_reward": 0.9986159205436707,
      "rewards/forensics_sinkhorn_reward": 0.4902709722518921,
      "rewards/forensics_strict_edge_reward": 0.249671071767807,
      "rewards/forensics_tiou_reward": 0.3321286141872406,
      "step": 471
    },
    {
      "completion_length": 71.03125,
      "epoch": 0.9884816753926702,
      "grad_norm": 44.50873565673828,
      "kl": 2.7890625,
      "learning_rate": 7.531446540880503e-07,
      "loss": 0.4312,
      "reward": 0.6253865659236908,
      "reward_std": 0.1640261374413967,
      "reward_std_all": 0.16897239908576012,
      "rewards-max/forensics_sinkhorn_reward": 0.9723258912563324,
      "rewards-max/forensics_strict_edge_reward": 0.8983768224716187,
      "rewards-max/forensics_tiou_reward": 0.9622795581817627,
      "rewards/forensics_sinkhorn_reward": 0.5308051407337189,
      "rewards/forensics_strict_edge_reward": 0.23078199476003647,
      "rewards/forensics_tiou_reward": 0.3891000300645828,
      "step": 472
    },
    {
      "completion_length": 82.78125,
      "epoch": 0.9905759162303664,
      "grad_norm": 14.685425758361816,
      "kl": 1.11328125,
      "learning_rate": 7.526205450733753e-07,
      "loss": 0.108,
      "reward": 0.5810080766677856,
      "reward_std": 0.19129403680562973,
      "reward_std_all": 0.19830747693777084,
      "rewards-max/forensics_sinkhorn_reward": 0.9130255281925201,
      "rewards-max/forensics_strict_edge_reward": 0.8291563987731934,
      "rewards-max/forensics_tiou_reward": 0.8399278223514557,
      "rewards/forensics_sinkhorn_reward": 0.40184326469898224,
      "rewards/forensics_strict_edge_reward": 0.18210864067077637,
      "rewards/forensics_tiou_reward": 0.20341812819242477,
      "step": 473
    },
    {
      "completion_length": 74.96875,
      "epoch": 0.9926701570680628,
      "grad_norm": 15.404865264892578,
      "kl": 1.060546875,
      "learning_rate": 7.520964360587002e-07,
      "loss": 0.2161,
      "reward": 0.5254700183868408,
      "reward_std": 0.24946381151676178,
      "reward_std_all": 0.25410059094429016,
      "rewards-max/forensics_sinkhorn_reward": 0.9836571514606476,
      "rewards-max/forensics_strict_edge_reward": 0.8597765564918518,
      "rewards-max/forensics_tiou_reward": 0.9791692495346069,
      "rewards/forensics_sinkhorn_reward": 0.3829945772886276,
      "rewards/forensics_strict_edge_reward": 0.1872287094593048,
      "rewards/forensics_tiou_reward": 0.22537289559841156,
      "step": 474
    },
    {
      "completion_length": 87.640625,
      "epoch": 0.9947643979057592,
      "grad_norm": 20.012737274169922,
      "kl": 1.3984375,
      "learning_rate": 7.515723270440252e-07,
      "loss": 0.1821,
      "reward": 0.6234315633773804,
      "reward_std": 0.25830285251140594,
      "reward_std_all": 0.23253708332777023,
      "rewards-max/forensics_sinkhorn_reward": 0.9795092046260834,
      "rewards-max/forensics_strict_edge_reward": 0.8983768224716187,
      "rewards-max/forensics_tiou_reward": 0.9782669544219971,
      "rewards/forensics_sinkhorn_reward": 0.49082884192466736,
      "rewards/forensics_strict_edge_reward": 0.28619980067014694,
      "rewards/forensics_tiou_reward": 0.3909042328596115,
      "step": 475
    },
    {
      "completion_length": 112.75,
      "epoch": 0.9968586387434555,
      "grad_norm": 25.049272537231445,
      "kl": 1.765625,
      "learning_rate": 7.5104821802935e-07,
      "loss": 0.1245,
      "reward": 0.5375048518180847,
      "reward_std": 0.2418568730354309,
      "reward_std_all": 0.24942870438098907,
      "rewards-max/forensics_sinkhorn_reward": 0.9826226830482483,
      "rewards-max/forensics_strict_edge_reward": 0.8668389916419983,
      "rewards-max/forensics_tiou_reward": 0.9749108850955963,
      "rewards/forensics_sinkhorn_reward": 0.41203370690345764,
      "rewards/forensics_strict_edge_reward": 0.172962948679924,
      "rewards/forensics_tiou_reward": 0.24053268879652023,
      "step": 476
    },
    {
      "completion_length": 90.359375,
      "epoch": 0.9989528795811519,
      "grad_norm": 16.448680877685547,
      "kl": 1.115234375,
      "learning_rate": 7.50524109014675e-07,
      "loss": 0.2699,
      "reward": 0.5367060303688049,
      "reward_std": 0.226222462952137,
      "reward_std_all": 0.23997463285923004,
      "rewards-max/forensics_sinkhorn_reward": 0.9751295149326324,
      "rewards-max/forensics_strict_edge_reward": 0.8379553556442261,
      "rewards-max/forensics_tiou_reward": 0.9463581144809723,
      "rewards/forensics_sinkhorn_reward": 0.4202313721179962,
      "rewards/forensics_strict_edge_reward": 0.18745631724596024,
      "rewards/forensics_tiou_reward": 0.23805760592222214,
      "step": 477
    },
    {
      "completion_length": 107.75,
      "epoch": 1.0,
      "grad_norm": 12.955926895141602,
      "kl": 1.1875,
      "learning_rate": 7.5e-07,
      "loss": 0.1349,
      "reward": 0.5002918243408203,
      "reward_std": 0.2739506959915161,
      "reward_std_all": 0.27990269660949707,
      "rewards-max/forensics_sinkhorn_reward": 0.9851891994476318,
      "rewards-max/forensics_strict_edge_reward": 0.9565287232398987,
      "rewards-max/forensics_tiou_reward": 0.9850748181343079,
      "rewards/forensics_sinkhorn_reward": 0.3640199899673462,
      "rewards/forensics_strict_edge_reward": 0.16675060987472534,
      "rewards/forensics_tiou_reward": 0.19825224578380585,
      "step": 478
    },
    {
      "completion_length": 91.078125,
      "epoch": 1.0020942408376963,
      "grad_norm": 14.049394607543945,
      "kl": 1.296875,
      "learning_rate": 7.494758909853249e-07,
      "loss": 0.1088,
      "reward": 0.6312054693698883,
      "reward_std": 0.1771041452884674,
      "reward_std_all": 0.17326029390096664,
      "rewards-max/forensics_sinkhorn_reward": 0.9875107705593109,
      "rewards-max/forensics_strict_edge_reward": 0.9357379674911499,
      "rewards-max/forensics_tiou_reward": 0.9941674768924713,
      "rewards/forensics_sinkhorn_reward": 0.5444709062576294,
      "rewards/forensics_strict_edge_reward": 0.35005223751068115,
      "rewards/forensics_tiou_reward": 0.42347289621829987,
      "step": 479
    },
    {
      "completion_length": 139.703125,
      "epoch": 1.0041884816753928,
      "grad_norm": 18.87883949279785,
      "kl": 1.7421875,
      "learning_rate": 7.489517819706499e-07,
      "loss": 0.2916,
      "reward": 0.5623349100351334,
      "reward_std": 0.22964052110910416,
      "reward_std_all": 0.22873633354902267,
      "rewards-max/forensics_sinkhorn_reward": 0.9876399040222168,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9946768581867218,
      "rewards/forensics_sinkhorn_reward": 0.469534307718277,
      "rewards/forensics_strict_edge_reward": 0.26252253353595734,
      "rewards/forensics_tiou_reward": 0.3260047733783722,
      "step": 480
    },
    {
      "completion_length": 84.65625,
      "epoch": 1.006282722513089,
      "grad_norm": 22.551040649414062,
      "kl": 1.51171875,
      "learning_rate": 7.484276729559747e-07,
      "loss": 0.0417,
      "reward": 0.6172755360603333,
      "reward_std": 0.1778905913233757,
      "reward_std_all": 0.1857256516814232,
      "rewards-max/forensics_sinkhorn_reward": 0.9420785307884216,
      "rewards-max/forensics_strict_edge_reward": 0.7955476641654968,
      "rewards-max/forensics_tiou_reward": 0.9004468023777008,
      "rewards/forensics_sinkhorn_reward": 0.48154330253601074,
      "rewards/forensics_strict_edge_reward": 0.19871477782726288,
      "rewards/forensics_tiou_reward": 0.2866392433643341,
      "step": 481
    },
    {
      "completion_length": 138.796875,
      "epoch": 1.0083769633507853,
      "grad_norm": 23.354976654052734,
      "kl": 2.4609375,
      "learning_rate": 7.479035639412997e-07,
      "loss": 0.2071,
      "reward": 0.5090178400278091,
      "reward_std": 0.18730197474360466,
      "reward_std_all": 0.22880914062261581,
      "rewards-max/forensics_sinkhorn_reward": 0.9820910394191742,
      "rewards-max/forensics_strict_edge_reward": 0.8379553556442261,
      "rewards-max/forensics_tiou_reward": 0.9730577766895294,
      "rewards/forensics_sinkhorn_reward": 0.4216884523630142,
      "rewards/forensics_strict_edge_reward": 0.1596815437078476,
      "rewards/forensics_tiou_reward": 0.21654776483774185,
      "step": 482
    },
    {
      "completion_length": 96.140625,
      "epoch": 1.0104712041884816,
      "grad_norm": 12.295806884765625,
      "kl": 1.26953125,
      "learning_rate": 7.473794549266247e-07,
      "loss": 0.0743,
      "reward": 0.6117375493049622,
      "reward_std": 0.17520742118358612,
      "reward_std_all": 0.18526983261108398,
      "rewards-max/forensics_sinkhorn_reward": 0.9874443113803864,
      "rewards-max/forensics_strict_edge_reward": 0.9567649364471436,
      "rewards-max/forensics_tiou_reward": 0.9939264059066772,
      "rewards/forensics_sinkhorn_reward": 0.5471093654632568,
      "rewards/forensics_strict_edge_reward": 0.3063894510269165,
      "rewards/forensics_tiou_reward": 0.38469724357128143,
      "step": 483
    },
    {
      "completion_length": 116.34375,
      "epoch": 1.012565445026178,
      "grad_norm": 16.434906005859375,
      "kl": 1.5234375,
      "learning_rate": 7.468553459119497e-07,
      "loss": 0.2302,
      "reward": 0.5682651698589325,
      "reward_std": 0.18841411173343658,
      "reward_std_all": 0.21064192801713943,
      "rewards-max/forensics_sinkhorn_reward": 0.9874099791049957,
      "rewards-max/forensics_strict_edge_reward": 0.9357379674911499,
      "rewards-max/forensics_tiou_reward": 0.9937761723995209,
      "rewards/forensics_sinkhorn_reward": 0.5403457283973694,
      "rewards/forensics_strict_edge_reward": 0.33507493138313293,
      "rewards/forensics_tiou_reward": 0.4347131550312042,
      "step": 484
    },
    {
      "completion_length": 115.203125,
      "epoch": 1.0146596858638743,
      "grad_norm": 18.000171661376953,
      "kl": 2.046875,
      "learning_rate": 7.463312368972746e-07,
      "loss": 0.2102,
      "reward": 0.5188246071338654,
      "reward_std": 0.17297257855534554,
      "reward_std_all": 0.22171839326620102,
      "rewards-max/forensics_sinkhorn_reward": 0.8622443974018097,
      "rewards-max/forensics_strict_edge_reward": 0.8223447799682617,
      "rewards-max/forensics_tiou_reward": 0.8329994678497314,
      "rewards/forensics_sinkhorn_reward": 0.36179907619953156,
      "rewards/forensics_strict_edge_reward": 0.1262497864663601,
      "rewards/forensics_tiou_reward": 0.17741280049085617,
      "step": 485
    },
    {
      "completion_length": 126.828125,
      "epoch": 1.0167539267015706,
      "grad_norm": 2631499.5,
      "kl": 4258.28125,
      "learning_rate": 7.458071278825996e-07,
      "loss": 170.5845,
      "reward": 0.6333977580070496,
      "reward_std": 0.18792997300624847,
      "reward_std_all": 0.17898652702569962,
      "rewards-max/forensics_sinkhorn_reward": 0.9873742461204529,
      "rewards-max/forensics_strict_edge_reward": 0.9565287232398987,
      "rewards-max/forensics_tiou_reward": 0.9936290383338928,
      "rewards/forensics_sinkhorn_reward": 0.49530769884586334,
      "rewards/forensics_strict_edge_reward": 0.27727124094963074,
      "rewards/forensics_tiou_reward": 0.3681890517473221,
      "step": 486
    },
    {
      "completion_length": 69.84375,
      "epoch": 1.018848167539267,
      "grad_norm": 15.853433609008789,
      "kl": 1.34375,
      "learning_rate": 7.452830188679245e-07,
      "loss": 0.0809,
      "reward": 0.632128894329071,
      "reward_std": 0.1483171433210373,
      "reward_std_all": 0.1587144285440445,
      "rewards-max/forensics_sinkhorn_reward": 0.9876440465450287,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9946939647197723,
      "rewards/forensics_sinkhorn_reward": 0.5259945094585419,
      "rewards/forensics_strict_edge_reward": 0.2344716489315033,
      "rewards/forensics_tiou_reward": 0.37481051683425903,
      "step": 487
    },
    {
      "completion_length": 102.375,
      "epoch": 1.0209424083769634,
      "grad_norm": 18.713275909423828,
      "kl": 1.30078125,
      "learning_rate": 7.447589098532494e-07,
      "loss": 0.2111,
      "reward": 0.5633865296840668,
      "reward_std": 0.15145248919725418,
      "reward_std_all": 0.19069860875606537,
      "rewards-max/forensics_sinkhorn_reward": 0.9813658595085144,
      "rewards-max/forensics_strict_edge_reward": 0.7787675857543945,
      "rewards-max/forensics_tiou_reward": 0.9703608453273773,
      "rewards/forensics_sinkhorn_reward": 0.44274306297302246,
      "rewards/forensics_strict_edge_reward": 0.16534005105495453,
      "rewards/forensics_tiou_reward": 0.2702155113220215,
      "step": 488
    },
    {
      "completion_length": 93.109375,
      "epoch": 1.0230366492146596,
      "grad_norm": 21.870882034301758,
      "kl": 1.27734375,
      "learning_rate": 7.442348008385744e-07,
      "loss": 0.2262,
      "reward": 0.5275819301605225,
      "reward_std": 0.2453763410449028,
      "reward_std_all": 0.26157229393720627,
      "rewards-max/forensics_sinkhorn_reward": 0.9873692393302917,
      "rewards-max/forensics_strict_edge_reward": 0.9565287232398987,
      "rewards-max/forensics_tiou_reward": 0.9936176240444183,
      "rewards/forensics_sinkhorn_reward": 0.41946205496788025,
      "rewards/forensics_strict_edge_reward": 0.1953878477215767,
      "rewards/forensics_tiou_reward": 0.24937580525875092,
      "step": 489
    },
    {
      "completion_length": 70.796875,
      "epoch": 1.025130890052356,
      "grad_norm": 17.10002899169922,
      "kl": 1.0859375,
      "learning_rate": 7.437106918238993e-07,
      "loss": 0.1197,
      "reward": 0.6466534733772278,
      "reward_std": 0.17823785543441772,
      "reward_std_all": 0.1744745895266533,
      "rewards-max/forensics_sinkhorn_reward": 0.9886393845081329,
      "rewards-max/forensics_strict_edge_reward": 0.9890114367008209,
      "rewards-max/forensics_tiou_reward": 0.9986159205436707,
      "rewards/forensics_sinkhorn_reward": 0.5705039501190186,
      "rewards/forensics_strict_edge_reward": 0.3378402292728424,
      "rewards/forensics_tiou_reward": 0.42432253062725067,
      "step": 490
    },
    {
      "completion_length": 103.703125,
      "epoch": 1.0272251308900524,
      "grad_norm": 35.26626968383789,
      "kl": 3.015625,
      "learning_rate": 7.431865828092243e-07,
      "loss": 0.2723,
      "reward": 0.6048932075500488,
      "reward_std": 0.21786823123693466,
      "reward_std_all": 0.2070453241467476,
      "rewards-max/forensics_sinkhorn_reward": 0.9873788356781006,
      "rewards-max/forensics_strict_edge_reward": 0.9158510267734528,
      "rewards-max/forensics_tiou_reward": 0.9936608076095581,
      "rewards/forensics_sinkhorn_reward": 0.5070751011371613,
      "rewards/forensics_strict_edge_reward": 0.34240712225437164,
      "rewards/forensics_tiou_reward": 0.40919825434684753,
      "step": 491
    },
    {
      "completion_length": 121.328125,
      "epoch": 1.0293193717277487,
      "grad_norm": 23.017797470092773,
      "kl": 2.4296875,
      "learning_rate": 7.426624737945492e-07,
      "loss": 0.3393,
      "reward": 0.5811571776866913,
      "reward_std": 0.20259416848421097,
      "reward_std_all": 0.20893565565347672,
      "rewards-max/forensics_sinkhorn_reward": 0.9853453636169434,
      "rewards-max/forensics_strict_edge_reward": 0.9265981018543243,
      "rewards-max/forensics_tiou_reward": 0.9857374131679535,
      "rewards/forensics_sinkhorn_reward": 0.4954771101474762,
      "rewards/forensics_strict_edge_reward": 0.27734240889549255,
      "rewards/forensics_tiou_reward": 0.3610907942056656,
      "step": 492
    },
    {
      "completion_length": 67.171875,
      "epoch": 1.031413612565445,
      "grad_norm": 13.61359691619873,
      "kl": 1.0703125,
      "learning_rate": 7.421383647798741e-07,
      "loss": 0.1726,
      "reward": 0.5869008302688599,
      "reward_std": 0.1699814796447754,
      "reward_std_all": 0.1851014420390129,
      "rewards-max/forensics_sinkhorn_reward": 0.9874564111232758,
      "rewards-max/forensics_strict_edge_reward": 0.9464850425720215,
      "rewards-max/forensics_tiou_reward": 0.9939608573913574,
      "rewards/forensics_sinkhorn_reward": 0.4433785080909729,
      "rewards/forensics_strict_edge_reward": 0.20614664256572723,
      "rewards/forensics_tiou_reward": 0.25967272371053696,
      "step": 493
    },
    {
      "completion_length": 73.109375,
      "epoch": 1.0335078534031414,
      "grad_norm": 16.004980087280273,
      "kl": 1.65234375,
      "learning_rate": 7.41614255765199e-07,
      "loss": 0.1484,
      "reward": 0.6071447730064392,
      "reward_std": 0.23630797117948532,
      "reward_std_all": 0.22336223721504211,
      "rewards-max/forensics_sinkhorn_reward": 0.9013909697532654,
      "rewards-max/forensics_strict_edge_reward": 0.7093792855739594,
      "rewards-max/forensics_tiou_reward": 0.8132624626159668,
      "rewards/forensics_sinkhorn_reward": 0.4425203502178192,
      "rewards/forensics_strict_edge_reward": 0.22591916471719742,
      "rewards/forensics_tiou_reward": 0.2885133922100067,
      "step": 494
    },
    {
      "completion_length": 81.6875,
      "epoch": 1.0356020942408377,
      "grad_norm": 24.021251678466797,
      "kl": 1.15234375,
      "learning_rate": 7.41090146750524e-07,
      "loss": 0.2368,
      "reward": 0.6024344265460968,
      "reward_std": 0.17346178740262985,
      "reward_std_all": 0.18168799579143524,
      "rewards-max/forensics_sinkhorn_reward": 0.9023637175559998,
      "rewards-max/forensics_strict_edge_reward": 0.8553276658058167,
      "rewards-max/forensics_tiou_reward": 0.8217936158180237,
      "rewards/forensics_sinkhorn_reward": 0.4665425717830658,
      "rewards/forensics_strict_edge_reward": 0.2858784943819046,
      "rewards/forensics_tiou_reward": 0.3304591327905655,
      "step": 495
    },
    {
      "completion_length": 71.296875,
      "epoch": 1.037696335078534,
      "grad_norm": 16.34440803527832,
      "kl": 1.03515625,
      "learning_rate": 7.405660377358491e-07,
      "loss": 0.1121,
      "reward": 0.5769129991531372,
      "reward_std": 0.24523751437664032,
      "reward_std_all": 0.23740284144878387,
      "rewards-max/forensics_sinkhorn_reward": 0.9654314815998077,
      "rewards-max/forensics_strict_edge_reward": 0.9357379674911499,
      "rewards-max/forensics_tiou_reward": 0.9456800520420074,
      "rewards/forensics_sinkhorn_reward": 0.4769698679447174,
      "rewards/forensics_strict_edge_reward": 0.1969602108001709,
      "rewards/forensics_tiou_reward": 0.27047666907310486,
      "step": 496
    },
    {
      "completion_length": 58.703125,
      "epoch": 1.0397905759162303,
      "grad_norm": 19.92194938659668,
      "kl": 1.322265625,
      "learning_rate": 7.40041928721174e-07,
      "loss": 0.2314,
      "reward": 0.5608172416687012,
      "reward_std": 0.2244546040892601,
      "reward_std_all": 0.22843100130558014,
      "rewards-max/forensics_sinkhorn_reward": 0.9886357188224792,
      "rewards-max/forensics_strict_edge_reward": 0.9890114367008209,
      "rewards-max/forensics_tiou_reward": 0.9986014068126678,
      "rewards/forensics_sinkhorn_reward": 0.47550877928733826,
      "rewards/forensics_strict_edge_reward": 0.19470055401325226,
      "rewards/forensics_tiou_reward": 0.2479368969798088,
      "step": 497
    },
    {
      "completion_length": 63.859375,
      "epoch": 1.0418848167539267,
      "grad_norm": 15.961243629455566,
      "kl": 1.30078125,
      "learning_rate": 7.39517819706499e-07,
      "loss": 0.2651,
      "reward": 0.5515331029891968,
      "reward_std": 0.20536160469055176,
      "reward_std_all": 0.22255847603082657,
      "rewards-max/forensics_sinkhorn_reward": 0.8307666778564453,
      "rewards-max/forensics_strict_edge_reward": 0.6096038520336151,
      "rewards-max/forensics_tiou_reward": 0.6897245645523071,
      "rewards/forensics_sinkhorn_reward": 0.3464539647102356,
      "rewards/forensics_strict_edge_reward": 0.15292533487081528,
      "rewards/forensics_tiou_reward": 0.18242985010147095,
      "step": 498
    },
    {
      "completion_length": 62.359375,
      "epoch": 1.043979057591623,
      "grad_norm": 16.68674659729004,
      "kl": 0.8720703125,
      "learning_rate": 7.389937106918238e-07,
      "loss": 0.0573,
      "reward": 0.6048471033573151,
      "reward_std": 0.2581226974725723,
      "reward_std_all": 0.23732759803533554,
      "rewards-max/forensics_sinkhorn_reward": 0.9882362186908722,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9970291256904602,
      "rewards/forensics_sinkhorn_reward": 0.4825429618358612,
      "rewards/forensics_strict_edge_reward": 0.2747987061738968,
      "rewards/forensics_tiou_reward": 0.3515544533729553,
      "step": 499
    },
    {
      "completion_length": 58.359375,
      "epoch": 1.0460732984293193,
      "grad_norm": 17.22966766357422,
      "kl": 0.822265625,
      "learning_rate": 7.384696016771488e-07,
      "loss": 0.3598,
      "reward": 0.5924529135227203,
      "reward_std": 0.20475586503744125,
      "reward_std_all": 0.20551985502243042,
      "rewards-max/forensics_sinkhorn_reward": 0.9881068468093872,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9965190887451172,
      "rewards/forensics_sinkhorn_reward": 0.4758411645889282,
      "rewards/forensics_strict_edge_reward": 0.23950999230146408,
      "rewards/forensics_tiou_reward": 0.29654037952423096,
      "step": 500
    },
    {
      "completion_length": 89.78125,
      "epoch": 1.0481675392670158,
      "grad_norm": 14.154056549072266,
      "kl": 1.36328125,
      "learning_rate": 7.379454926624737e-07,
      "loss": 0.3215,
      "reward": 0.5518761575222015,
      "reward_std": 0.21114203333854675,
      "reward_std_all": 0.22113005816936493,
      "rewards-max/forensics_sinkhorn_reward": 0.8959895074367523,
      "rewards-max/forensics_strict_edge_reward": 0.7328133881092072,
      "rewards-max/forensics_tiou_reward": 0.7923597097396851,
      "rewards/forensics_sinkhorn_reward": 0.3614528775215149,
      "rewards/forensics_strict_edge_reward": 0.2028835564851761,
      "rewards/forensics_tiou_reward": 0.20063307881355286,
      "step": 501
    },
    {
      "completion_length": 56.578125,
      "epoch": 1.050261780104712,
      "grad_norm": 15.911866188049316,
      "kl": 1.208984375,
      "learning_rate": 7.374213836477987e-07,
      "loss": 0.3152,
      "reward": 0.6726288497447968,
      "reward_std": 0.14309410750865936,
      "reward_std_all": 0.13919394463300705,
      "rewards-max/forensics_sinkhorn_reward": 0.9876346588134766,
      "rewards-max/forensics_strict_edge_reward": 0.9565287232398987,
      "rewards-max/forensics_tiou_reward": 0.9946580827236176,
      "rewards/forensics_sinkhorn_reward": 0.604799211025238,
      "rewards/forensics_strict_edge_reward": 0.35322602838277817,
      "rewards/forensics_tiou_reward": 0.4701387733221054,
      "step": 502
    },
    {
      "completion_length": 97.734375,
      "epoch": 1.0523560209424083,
      "grad_norm": 16.621793746948242,
      "kl": 1.65234375,
      "learning_rate": 7.368972746331237e-07,
      "loss": 0.3901,
      "reward": 0.5529776811599731,
      "reward_std": 0.20212648063898087,
      "reward_std_all": 0.22788863629102707,
      "rewards-max/forensics_sinkhorn_reward": 0.9884631037712097,
      "rewards-max/forensics_strict_edge_reward": 0.9375866651535034,
      "rewards-max/forensics_tiou_reward": 0.9979231655597687,
      "rewards/forensics_sinkhorn_reward": 0.46796606481075287,
      "rewards/forensics_strict_edge_reward": 0.3065918982028961,
      "rewards/forensics_tiou_reward": 0.34490254521369934,
      "step": 503
    },
    {
      "completion_length": 62.8125,
      "epoch": 1.0544502617801048,
      "grad_norm": 23.835186004638672,
      "kl": 1.875,
      "learning_rate": 7.363731656184486e-07,
      "loss": 0.3656,
      "reward": 0.6214357316493988,
      "reward_std": 0.2188994511961937,
      "reward_std_all": 0.20401781797409058,
      "rewards-max/forensics_sinkhorn_reward": 0.9340793788433075,
      "rewards-max/forensics_strict_edge_reward": 0.8893801271915436,
      "rewards-max/forensics_tiou_reward": 0.8855506777763367,
      "rewards/forensics_sinkhorn_reward": 0.504751667380333,
      "rewards/forensics_strict_edge_reward": 0.30737804621458054,
      "rewards/forensics_tiou_reward": 0.3590254634618759,
      "step": 504
    },
    {
      "completion_length": 101.609375,
      "epoch": 1.056544502617801,
      "grad_norm": 24.413562774658203,
      "kl": 2.34375,
      "learning_rate": 7.358490566037735e-07,
      "loss": 0.2529,
      "reward": 0.5021973699331284,
      "reward_std": 0.24772091954946518,
      "reward_std_all": 0.2627536356449127,
      "rewards-max/forensics_sinkhorn_reward": 0.9877929389476776,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9952799081802368,
      "rewards/forensics_sinkhorn_reward": 0.35389988124370575,
      "rewards/forensics_strict_edge_reward": 0.21208354830741882,
      "rewards/forensics_tiou_reward": 0.2385101094841957,
      "step": 505
    },
    {
      "completion_length": 92.546875,
      "epoch": 1.0586387434554974,
      "grad_norm": 56.0758056640625,
      "kl": 3.7265625,
      "learning_rate": 7.353249475890984e-07,
      "loss": 0.2841,
      "reward": 0.5561835467815399,
      "reward_std": 0.23429866135120392,
      "reward_std_all": 0.23705051094293594,
      "rewards-max/forensics_sinkhorn_reward": 0.9881208837032318,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9965728521347046,
      "rewards/forensics_sinkhorn_reward": 0.4934905916452408,
      "rewards/forensics_strict_edge_reward": 0.340737447142601,
      "rewards/forensics_tiou_reward": 0.4283844828605652,
      "step": 506
    },
    {
      "completion_length": 74.453125,
      "epoch": 1.0607329842931936,
      "grad_norm": 25.131174087524414,
      "kl": 2.5078125,
      "learning_rate": 7.348008385744235e-07,
      "loss": 0.2753,
      "reward": 0.58453568816185,
      "reward_std": 0.22093906998634338,
      "reward_std_all": 0.222455196082592,
      "rewards-max/forensics_sinkhorn_reward": 0.9876264333724976,
      "rewards-max/forensics_strict_edge_reward": 0.9782643616199493,
      "rewards-max/forensics_tiou_reward": 0.9946381151676178,
      "rewards/forensics_sinkhorn_reward": 0.44807687401771545,
      "rewards/forensics_strict_edge_reward": 0.2604019194841385,
      "rewards/forensics_tiou_reward": 0.30481716990470886,
      "step": 507
    },
    {
      "completion_length": 66.890625,
      "epoch": 1.0628272251308901,
      "grad_norm": 23.416231155395508,
      "kl": 2.11328125,
      "learning_rate": 7.342767295597484e-07,
      "loss": 0.1602,
      "reward": 0.6284846365451813,
      "reward_std": 0.1562955416738987,
      "reward_std_all": 0.1594071052968502,
      "rewards-max/forensics_sinkhorn_reward": 0.9875780642032623,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9944413006305695,
      "rewards/forensics_sinkhorn_reward": 0.4790893495082855,
      "rewards/forensics_strict_edge_reward": 0.154500350356102,
      "rewards/forensics_tiou_reward": 0.2684135288000107,
      "step": 508
    },
    {
      "completion_length": 74.46875,
      "epoch": 1.0649214659685864,
      "grad_norm": 16.295452117919922,
      "kl": 2.234375,
      "learning_rate": 7.337526205450734e-07,
      "loss": 0.2367,
      "reward": 0.6462744474411011,
      "reward_std": 0.12925580888986588,
      "reward_std_all": 0.1431896835565567,
      "rewards-max/forensics_sinkhorn_reward": 0.9882093966007233,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9969207346439362,
      "rewards/forensics_sinkhorn_reward": 0.5639992654323578,
      "rewards/forensics_strict_edge_reward": 0.2735386937856674,
      "rewards/forensics_tiou_reward": 0.3802160918712616,
      "step": 509
    },
    {
      "completion_length": 56.984375,
      "epoch": 1.0670157068062827,
      "grad_norm": 15.483229637145996,
      "kl": 1.453125,
      "learning_rate": 7.332285115303984e-07,
      "loss": 0.1869,
      "reward": 0.6091759502887726,
      "reward_std": 0.16483874805271626,
      "reward_std_all": 0.1777092181146145,
      "rewards-max/forensics_sinkhorn_reward": 0.9878043234348297,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9953339099884033,
      "rewards/forensics_sinkhorn_reward": 0.5101052820682526,
      "rewards/forensics_strict_edge_reward": 0.26473384350538254,
      "rewards/forensics_tiou_reward": 0.34738753736019135,
      "step": 510
    },
    {
      "completion_length": 87.125,
      "epoch": 1.0691099476439792,
      "grad_norm": 22.370695114135742,
      "kl": 2.33984375,
      "learning_rate": 7.327044025157232e-07,
      "loss": 0.2314,
      "reward": 0.5889894068241119,
      "reward_std": 0.19216328114271164,
      "reward_std_all": 0.2058073729276657,
      "rewards-max/forensics_sinkhorn_reward": 0.9879955947399139,
      "rewards-max/forensics_strict_edge_reward": 0.9677534997463226,
      "rewards-max/forensics_tiou_reward": 0.9960784316062927,
      "rewards/forensics_sinkhorn_reward": 0.5049192607402802,
      "rewards/forensics_strict_edge_reward": 0.3227221295237541,
      "rewards/forensics_tiou_reward": 0.36073847115039825,
      "step": 511
    },
    {
      "completion_length": 82.234375,
      "epoch": 1.0712041884816754,
      "grad_norm": 19.83217430114746,
      "kl": 1.8515625,
      "learning_rate": 7.321802935010482e-07,
      "loss": -0.0027,
      "reward": 0.5832606256008148,
      "reward_std": 0.23635698109865189,
      "reward_std_all": 0.2303152158856392,
      "rewards-max/forensics_sinkhorn_reward": 0.9748813509941101,
      "rewards-max/forensics_strict_edge_reward": 0.8863177299499512,
      "rewards-max/forensics_tiou_reward": 0.9567746818065643,
      "rewards/forensics_sinkhorn_reward": 0.46278800070285797,
      "rewards/forensics_strict_edge_reward": 0.2192613184452057,
      "rewards/forensics_tiou_reward": 0.26120559871196747,
      "step": 512
    },
    {
      "completion_length": 69.125,
      "epoch": 1.0732984293193717,
      "grad_norm": 17.220733642578125,
      "kl": 1.701171875,
      "learning_rate": 7.316561844863731e-07,
      "loss": 0.1538,
      "reward": 0.6108542680740356,
      "reward_std": 0.14365491643548012,
      "reward_std_all": 0.16852950304746628,
      "rewards-max/forensics_sinkhorn_reward": 0.988754540681839,
      "rewards-max/forensics_strict_edge_reward": 0.9890114367008209,
      "rewards-max/forensics_tiou_reward": 0.9990697801113129,
      "rewards/forensics_sinkhorn_reward": 0.5234821885824203,
      "rewards/forensics_strict_edge_reward": 0.3164670020341873,
      "rewards/forensics_tiou_reward": 0.3749896138906479,
      "step": 513
    },
    {
      "completion_length": 78.1875,
      "epoch": 1.075392670157068,
      "grad_norm": 18.41690444946289,
      "kl": 1.76953125,
      "learning_rate": 7.311320754716981e-07,
      "loss": 0.2227,
      "reward": 0.600681722164154,
      "reward_std": 0.1813730150461197,
      "reward_std_all": 0.1966528743505478,
      "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": 0.4939858615398407,
      "rewards/forensics_strict_edge_reward": 0.25872498750686646,
      "rewards/forensics_tiou_reward": 0.3133735954761505,
      "step": 514
    },
    {
      "completion_length": 94.453125,
      "epoch": 1.0774869109947645,
      "grad_norm": 21.907541275024414,
      "kl": 1.9375,
      "learning_rate": 7.30607966457023e-07,
      "loss": 0.2658,
      "reward": 0.5861891210079193,
      "reward_std": 0.1943235993385315,
      "reward_std_all": 0.2111702561378479,
      "rewards-max/forensics_sinkhorn_reward": 0.9866428673267365,
      "rewards-max/forensics_strict_edge_reward": 0.946017861366272,
      "rewards-max/forensics_tiou_reward": 0.9907628893852234,
      "rewards/forensics_sinkhorn_reward": 0.4662705957889557,
      "rewards/forensics_strict_edge_reward": 0.2536363750696182,
      "rewards/forensics_tiou_reward": 0.3218904137611389,
      "step": 515
    },
    {
      "completion_length": 98.265625,
      "epoch": 1.0795811518324607,
      "grad_norm": 24.102283477783203,
      "kl": 2.296875,
      "learning_rate": 7.30083857442348e-07,
      "loss": 0.1866,
      "reward": 0.571931391954422,
      "reward_std": 0.17590603232383728,
      "reward_std_all": 0.20362502336502075,
      "rewards-max/forensics_sinkhorn_reward": 0.7648443579673767,
      "rewards-max/forensics_strict_edge_reward": 0.6739518940448761,
      "rewards-max/forensics_tiou_reward": 0.6858799308538437,
      "rewards/forensics_sinkhorn_reward": 0.41710880398750305,
      "rewards/forensics_strict_edge_reward": 0.20580115914344788,
      "rewards/forensics_tiou_reward": 0.22663873434066772,
      "step": 516
    },
    {
      "completion_length": 120.734375,
      "epoch": 1.081675392670157,
      "grad_norm": 48.06169509887695,
      "kl": 3.53125,
      "learning_rate": 7.295597484276728e-07,
      "loss": 0.2831,
      "reward": 0.5750539898872375,
      "reward_std": 0.2000800520181656,
      "reward_std_all": 0.21308576315641403,
      "rewards-max/forensics_sinkhorn_reward": 0.9357297122478485,
      "rewards-max/forensics_strict_edge_reward": 0.6524847000837326,
      "rewards-max/forensics_tiou_reward": 0.8813618421554565,
      "rewards/forensics_sinkhorn_reward": 0.450186088681221,
      "rewards/forensics_strict_edge_reward": 0.1856793761253357,
      "rewards/forensics_tiou_reward": 0.30526261031627655,
      "step": 517
    },
    {
      "completion_length": 81.40625,
      "epoch": 1.0837696335078535,
      "grad_norm": 20.29426383972168,
      "kl": 2.296875,
      "learning_rate": 7.290356394129978e-07,
      "loss": 0.2134,
      "reward": 0.5719835758209229,
      "reward_std": 0.21256975829601288,
      "reward_std_all": 0.21943044662475586,
      "rewards-max/forensics_sinkhorn_reward": 0.9882877469062805,
      "rewards-max/forensics_strict_edge_reward": 0.9375866651535034,
      "rewards-max/forensics_tiou_reward": 0.9972337484359741,
      "rewards/forensics_sinkhorn_reward": 0.4733058512210846,
      "rewards/forensics_strict_edge_reward": 0.28433550149202347,
      "rewards/forensics_tiou_reward": 0.33446434140205383,
      "step": 518
    },
    {
      "completion_length": 70.6875,
      "epoch": 1.0858638743455498,
      "grad_norm": 17.251527786254883,
      "kl": 2.46875,
      "learning_rate": 7.285115303983228e-07,
      "loss": 0.2615,
      "reward": 0.6462964415550232,
      "reward_std": 0.14295321702957153,
      "reward_std_all": 0.1545681431889534,
      "rewards-max/forensics_sinkhorn_reward": 0.9851558208465576,
      "rewards-max/forensics_strict_edge_reward": 0.87538942694664,
      "rewards-max/forensics_tiou_reward": 0.9848881959915161,
      "rewards/forensics_sinkhorn_reward": 0.5295253098011017,
      "rewards/forensics_strict_edge_reward": 0.2919923812150955,
      "rewards/forensics_tiou_reward": 0.40298134088516235,
      "step": 519
    },
    {
      "completion_length": 83.328125,
      "epoch": 1.087958115183246,
      "grad_norm": 21.63471794128418,
      "kl": 2.328125,
      "learning_rate": 7.279874213836478e-07,
      "loss": 0.2711,
      "reward": 0.5272080600261688,
      "reward_std": 0.15639668703079224,
      "reward_std_all": 0.2146027684211731,
      "rewards-max/forensics_sinkhorn_reward": 0.981073796749115,
      "rewards-max/forensics_strict_edge_reward": 0.8237898349761963,
      "rewards-max/forensics_tiou_reward": 0.9688286781311035,
      "rewards/forensics_sinkhorn_reward": 0.4013645350933075,
      "rewards/forensics_strict_edge_reward": 0.16954537481069565,
      "rewards/forensics_tiou_reward": 0.247541144490242,
      "step": 520
    },
    {
      "completion_length": 80.453125,
      "epoch": 1.0900523560209425,
      "grad_norm": 37.19434356689453,
      "kl": 3.6328125,
      "learning_rate": 7.274633123689728e-07,
      "loss": 0.456,
      "reward": 0.6176568865776062,
      "reward_std": 0.19933338463306427,
      "reward_std_all": 0.19170677661895752,
      "rewards-max/forensics_sinkhorn_reward": 0.9879108965396881,
      "rewards-max/forensics_strict_edge_reward": 0.9782643616199493,
      "rewards-max/forensics_tiou_reward": 0.9957446753978729,
      "rewards/forensics_sinkhorn_reward": 0.5208576619625092,
      "rewards/forensics_strict_edge_reward": 0.3434048444032669,
      "rewards/forensics_tiou_reward": 0.4057596027851105,
      "step": 521
    },
    {
      "completion_length": 70.9375,
      "epoch": 1.0921465968586388,
      "grad_norm": 20.822349548339844,
      "kl": 2.4453125,
      "learning_rate": 7.269392033542977e-07,
      "loss": 0.2488,
      "reward": 0.638585090637207,
      "reward_std": 0.1528386026620865,
      "reward_std_all": 0.16164277493953705,
      "rewards-max/forensics_sinkhorn_reward": 0.9885258972644806,
      "rewards-max/forensics_strict_edge_reward": 0.9782643616199493,
      "rewards-max/forensics_tiou_reward": 0.9981701672077179,
      "rewards/forensics_sinkhorn_reward": 0.5419031530618668,
      "rewards/forensics_strict_edge_reward": 0.2599894553422928,
      "rewards/forensics_tiou_reward": 0.3061997890472412,
      "step": 522
    },
    {
      "completion_length": 75.984375,
      "epoch": 1.094240837696335,
      "grad_norm": 17.159053802490234,
      "kl": 3.4296875,
      "learning_rate": 7.264150943396226e-07,
      "loss": 0.2529,
      "reward": 0.5677027404308319,
      "reward_std": 0.1644466668367386,
      "reward_std_all": 0.19787226617336273,
      "rewards-max/forensics_sinkhorn_reward": 0.8726876974105835,
      "rewards-max/forensics_strict_edge_reward": 0.8062867522239685,
      "rewards-max/forensics_tiou_reward": 0.7725595831871033,
      "rewards/forensics_sinkhorn_reward": 0.41738566756248474,
      "rewards/forensics_strict_edge_reward": 0.1683334931731224,
      "rewards/forensics_tiou_reward": 0.2176692709326744,
      "step": 523
    },
    {
      "completion_length": 72.546875,
      "epoch": 1.0963350785340313,
      "grad_norm": 16.43071746826172,
      "kl": 3.0078125,
      "learning_rate": 7.258909853249475e-07,
      "loss": 0.4534,
      "reward": 0.6315715909004211,
      "reward_std": 0.16339369863271713,
      "reward_std_all": 0.1638697162270546,
      "rewards-max/forensics_sinkhorn_reward": 0.9863370954990387,
      "rewards-max/forensics_strict_edge_reward": 0.87538942694664,
      "rewards-max/forensics_tiou_reward": 0.9928492903709412,
      "rewards/forensics_sinkhorn_reward": 0.4494209587574005,
      "rewards/forensics_strict_edge_reward": 0.25746244192123413,
      "rewards/forensics_tiou_reward": 0.3405311554670334,
      "step": 524
    },
    {
      "completion_length": 56.015625,
      "epoch": 1.0984293193717278,
      "grad_norm": 17.355045318603516,
      "kl": 1.88671875,
      "learning_rate": 7.253668763102725e-07,
      "loss": 0.1905,
      "reward": 0.6453288793563843,
      "reward_std": 0.12216811627149582,
      "reward_std_all": 0.13876605778932571,
      "rewards-max/forensics_sinkhorn_reward": 0.9874229729175568,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9938312470912933,
      "rewards/forensics_sinkhorn_reward": 0.5259588211774826,
      "rewards/forensics_strict_edge_reward": 0.2962717264890671,
      "rewards/forensics_tiou_reward": 0.37872301787137985,
      "step": 525
    },
    {
      "completion_length": 75.671875,
      "epoch": 1.100523560209424,
      "grad_norm": 111.56481170654297,
      "kl": 4.84375,
      "learning_rate": 7.248427672955975e-07,
      "loss": 0.4788,
      "reward": 0.5784626603126526,
      "reward_std": 0.17734461277723312,
      "reward_std_all": 0.18959414958953857,
      "rewards-max/forensics_sinkhorn_reward": 0.9606868028640747,
      "rewards-max/forensics_strict_edge_reward": 0.8219673931598663,
      "rewards-max/forensics_tiou_reward": 0.9329863786697388,
      "rewards/forensics_sinkhorn_reward": 0.4333551228046417,
      "rewards/forensics_strict_edge_reward": 0.1675555482506752,
      "rewards/forensics_tiou_reward": 0.25557805597782135,
      "step": 526
    },
    {
      "completion_length": 82.8125,
      "epoch": 1.1026178010471204,
      "grad_norm": 343.89031982421875,
      "kl": 6.015625,
      "learning_rate": 7.243186582809224e-07,
      "loss": 0.3371,
      "reward": 0.5791169106960297,
      "reward_std": 0.1781386137008667,
      "reward_std_all": 0.19434276223182678,
      "rewards-max/forensics_sinkhorn_reward": 0.9185917377471924,
      "rewards-max/forensics_strict_edge_reward": 0.8528120815753937,
      "rewards-max/forensics_tiou_reward": 0.8523109257221222,
      "rewards/forensics_sinkhorn_reward": 0.4120207726955414,
      "rewards/forensics_strict_edge_reward": 0.18247142434120178,
      "rewards/forensics_tiou_reward": 0.21400446444749832,
      "step": 527
    },
    {
      "completion_length": 69.546875,
      "epoch": 1.1047120418848166,
      "grad_norm": 22.444053649902344,
      "kl": 4.65625,
      "learning_rate": 7.237945492662474e-07,
      "loss": 0.3093,
      "reward": 0.594334751367569,
      "reward_std": 0.17746754735708237,
      "reward_std_all": 0.19586240500211716,
      "rewards-max/forensics_sinkhorn_reward": 0.9700835943222046,
      "rewards-max/forensics_strict_edge_reward": 0.8745477199554443,
      "rewards-max/forensics_tiou_reward": 0.9564444124698639,
      "rewards/forensics_sinkhorn_reward": 0.4417133331298828,
      "rewards/forensics_strict_edge_reward": 0.2163708508014679,
      "rewards/forensics_tiou_reward": 0.2815944328904152,
      "step": 528
    },
    {
      "completion_length": 58.515625,
      "epoch": 1.1068062827225131,
      "grad_norm": 20.134212493896484,
      "kl": 4.09375,
      "learning_rate": 7.232704402515722e-07,
      "loss": 0.4047,
      "reward": 0.6502615213394165,
      "reward_std": 0.17847851663827896,
      "reward_std_all": 0.16985740512609482,
      "rewards-max/forensics_sinkhorn_reward": 0.9882286787033081,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.996999591588974,
      "rewards/forensics_sinkhorn_reward": 0.49175310134887695,
      "rewards/forensics_strict_edge_reward": 0.2983477860689163,
      "rewards/forensics_tiou_reward": 0.33431100845336914,
      "step": 529
    },
    {
      "completion_length": 39.53125,
      "epoch": 1.1089005235602094,
      "grad_norm": 18.879507064819336,
      "kl": 2.6875,
      "learning_rate": 7.227463312368972e-07,
      "loss": 0.3017,
      "reward": 0.6215242743492126,
      "reward_std": 0.18079322576522827,
      "reward_std_all": 0.1804647147655487,
      "rewards-max/forensics_sinkhorn_reward": 0.9855463802814484,
      "rewards-max/forensics_strict_edge_reward": 0.9464850425720215,
      "rewards-max/forensics_tiou_reward": 0.9865058064460754,
      "rewards/forensics_sinkhorn_reward": 0.5173062086105347,
      "rewards/forensics_strict_edge_reward": 0.25077632814645767,
      "rewards/forensics_tiou_reward": 0.3457854837179184,
      "step": 530
    },
    {
      "completion_length": 53.75,
      "epoch": 1.1109947643979057,
      "grad_norm": 20.075054168701172,
      "kl": 3.8828125,
      "learning_rate": 7.222222222222221e-07,
      "loss": 0.3585,
      "reward": 0.6147044599056244,
      "reward_std": 0.19459488987922668,
      "reward_std_all": 0.1977100595831871,
      "rewards-max/forensics_sinkhorn_reward": 0.9530780017375946,
      "rewards-max/forensics_strict_edge_reward": 0.8503920137882233,
      "rewards-max/forensics_tiou_reward": 0.9199245870113373,
      "rewards/forensics_sinkhorn_reward": 0.45870520174503326,
      "rewards/forensics_strict_edge_reward": 0.28552989661693573,
      "rewards/forensics_tiou_reward": 0.36466702818870544,
      "step": 531
    },
    {
      "completion_length": 40.765625,
      "epoch": 1.1130890052356022,
      "grad_norm": 17.382671356201172,
      "kl": 2.625,
      "learning_rate": 7.216981132075472e-07,
      "loss": 0.3202,
      "reward": 0.6169088780879974,
      "reward_std": 0.14066673815250397,
      "reward_std_all": 0.16681994497776031,
      "rewards-max/forensics_sinkhorn_reward": 0.9414360821247101,
      "rewards-max/forensics_strict_edge_reward": 0.7401526570320129,
      "rewards-max/forensics_tiou_reward": 0.8952431976795197,
      "rewards/forensics_sinkhorn_reward": 0.44509342312812805,
      "rewards/forensics_strict_edge_reward": 0.21271584182977676,
      "rewards/forensics_tiou_reward": 0.2786487862467766,
      "step": 532
    },
    {
      "completion_length": 77.765625,
      "epoch": 1.1151832460732984,
      "grad_norm": 21.448915481567383,
      "kl": 4.9375,
      "learning_rate": 7.211740041928721e-07,
      "loss": 0.3351,
      "reward": 0.5580929219722748,
      "reward_std": 0.13746213912963867,
      "reward_std_all": 0.18998516350984573,
      "rewards-max/forensics_sinkhorn_reward": 0.7971498966217041,
      "rewards-max/forensics_strict_edge_reward": 0.6710749566555023,
      "rewards-max/forensics_tiou_reward": 0.6060540676116943,
      "rewards/forensics_sinkhorn_reward": 0.36215628683567047,
      "rewards/forensics_strict_edge_reward": 0.1741843745112419,
      "rewards/forensics_tiou_reward": 0.18437089025974274,
      "step": 533
    },
    {
      "completion_length": 57.03125,
      "epoch": 1.1172774869109947,
      "grad_norm": 17.207162857055664,
      "kl": 3.3984375,
      "learning_rate": 7.206498951781971e-07,
      "loss": 0.2364,
      "reward": 0.6099632978439331,
      "reward_std": 0.17309866845607758,
      "reward_std_all": 0.1833997145295143,
      "rewards-max/forensics_sinkhorn_reward": 0.9875343441963196,
      "rewards-max/forensics_strict_edge_reward": 0.9062341153621674,
      "rewards-max/forensics_tiou_reward": 0.9942683577537537,
      "rewards/forensics_sinkhorn_reward": 0.48406243324279785,
      "rewards/forensics_strict_edge_reward": 0.27410030364990234,
      "rewards/forensics_tiou_reward": 0.31846241652965546,
      "step": 534
    },
    {
      "completion_length": 43.515625,
      "epoch": 1.1193717277486912,
      "grad_norm": 17.530162811279297,
      "kl": 2.80859375,
      "learning_rate": 7.20125786163522e-07,
      "loss": 0.3378,
      "reward": 0.5866970419883728,
      "reward_std": 0.1497405394911766,
      "reward_std_all": 0.17764656990766525,
      "rewards-max/forensics_sinkhorn_reward": 0.9864841103553772,
      "rewards-max/forensics_strict_edge_reward": 0.9464850425720215,
      "rewards-max/forensics_tiou_reward": 0.9901230037212372,
      "rewards/forensics_sinkhorn_reward": 0.45993684232234955,
      "rewards/forensics_strict_edge_reward": 0.21432123333215714,
      "rewards/forensics_tiou_reward": 0.27632106840610504,
      "step": 535
    },
    {
      "completion_length": 51.46875,
      "epoch": 1.1214659685863875,
      "grad_norm": 32.66473388671875,
      "kl": 4.46875,
      "learning_rate": 7.196016771488469e-07,
      "loss": 0.4493,
      "reward": 0.5965942144393921,
      "reward_std": 0.20477820932865143,
      "reward_std_all": 0.2107321321964264,
      "rewards-max/forensics_sinkhorn_reward": 0.9423788189888,
      "rewards-max/forensics_strict_edge_reward": 0.8797933161258698,
      "rewards-max/forensics_tiou_reward": 0.900881439447403,
      "rewards/forensics_sinkhorn_reward": 0.4374884366989136,
      "rewards/forensics_strict_edge_reward": 0.19479665160179138,
      "rewards/forensics_tiou_reward": 0.28900669515132904,
      "step": 536
    },
    {
      "completion_length": 62.34375,
      "epoch": 1.1235602094240837,
      "grad_norm": 19.263526916503906,
      "kl": 3.1953125,
      "learning_rate": 7.190775681341719e-07,
      "loss": 0.1729,
      "reward": 0.6346336901187897,
      "reward_std": 0.20603370666503906,
      "reward_std_all": 0.19248012453317642,
      "rewards-max/forensics_sinkhorn_reward": 0.9763981699943542,
      "rewards-max/forensics_strict_edge_reward": 0.9357379674911499,
      "rewards-max/forensics_tiou_reward": 0.9713259339332581,
      "rewards/forensics_sinkhorn_reward": 0.4729301631450653,
      "rewards/forensics_strict_edge_reward": 0.29485005140304565,
      "rewards/forensics_tiou_reward": 0.3512546122074127,
      "step": 537
    },
    {
      "completion_length": 28.515625,
      "epoch": 1.12565445026178,
      "grad_norm": 17.990312576293945,
      "kl": 1.72265625,
      "learning_rate": 7.185534591194968e-07,
      "loss": 0.2502,
      "reward": 0.676074206829071,
      "reward_std": 0.12011812999844551,
      "reward_std_all": 0.12261595390737057,
      "rewards-max/forensics_sinkhorn_reward": 0.9815551042556763,
      "rewards-max/forensics_strict_edge_reward": 0.9256840348243713,
      "rewards-max/forensics_tiou_reward": 0.9818349480628967,
      "rewards/forensics_sinkhorn_reward": 0.5450222194194794,
      "rewards/forensics_strict_edge_reward": 0.3207526355981827,
      "rewards/forensics_tiou_reward": 0.3961624205112457,
      "step": 538
    },
    {
      "completion_length": 44.59375,
      "epoch": 1.1277486910994765,
      "grad_norm": 20.571916580200195,
      "kl": 2.9453125,
      "learning_rate": 7.180293501048218e-07,
      "loss": 0.3726,
      "reward": 0.6450984477996826,
      "reward_std": 0.14711740612983704,
      "reward_std_all": 0.15249016880989075,
      "rewards-max/forensics_sinkhorn_reward": 0.9421635866165161,
      "rewards-max/forensics_strict_edge_reward": 0.8096016347408295,
      "rewards-max/forensics_tiou_reward": 0.8997991979122162,
      "rewards/forensics_sinkhorn_reward": 0.4946257770061493,
      "rewards/forensics_strict_edge_reward": 0.2652013450860977,
      "rewards/forensics_tiou_reward": 0.30138230323791504,
      "step": 539
    },
    {
      "completion_length": 44.53125,
      "epoch": 1.1298429319371728,
      "grad_norm": 18.37093162536621,
      "kl": 2.9140625,
      "learning_rate": 7.175052410901467e-07,
      "loss": 0.2444,
      "reward": 0.6102450788021088,
      "reward_std": 0.17100008577108383,
      "reward_std_all": 0.17872972041368484,
      "rewards-max/forensics_sinkhorn_reward": 0.9591811299324036,
      "rewards-max/forensics_strict_edge_reward": 0.7443503141403198,
      "rewards-max/forensics_tiou_reward": 0.9315334856510162,
      "rewards/forensics_sinkhorn_reward": 0.4487670511007309,
      "rewards/forensics_strict_edge_reward": 0.2086046300828457,
      "rewards/forensics_tiou_reward": 0.2501577213406563,
      "step": 540
    },
    {
      "completion_length": 34.359375,
      "epoch": 1.131937172774869,
      "grad_norm": 21.05216407775879,
      "kl": 2.9609375,
      "learning_rate": 7.169811320754716e-07,
      "loss": 0.2508,
      "reward": 0.6663405299186707,
      "reward_std": 0.12725649774074554,
      "reward_std_all": 0.1316004917025566,
      "rewards-max/forensics_sinkhorn_reward": 0.9830059111118317,
      "rewards-max/forensics_strict_edge_reward": 0.8663162291049957,
      "rewards-max/forensics_tiou_reward": 0.9767087399959564,
      "rewards/forensics_sinkhorn_reward": 0.5695444643497467,
      "rewards/forensics_strict_edge_reward": 0.2984345331788063,
      "rewards/forensics_tiou_reward": 0.3521779477596283,
      "step": 541
    },
    {
      "completion_length": 65.25,
      "epoch": 1.1340314136125655,
      "grad_norm": 16.773069381713867,
      "kl": 2.6015625,
      "learning_rate": 7.164570230607966e-07,
      "loss": 0.2985,
      "reward": 0.64109668135643,
      "reward_std": 0.12488401681184769,
      "reward_std_all": 0.14073511958122253,
      "rewards-max/forensics_sinkhorn_reward": 0.9607572555541992,
      "rewards-max/forensics_strict_edge_reward": 0.878633052110672,
      "rewards-max/forensics_tiou_reward": 0.9381617605686188,
      "rewards/forensics_sinkhorn_reward": 0.48189307749271393,
      "rewards/forensics_strict_edge_reward": 0.2760924696922302,
      "rewards/forensics_tiou_reward": 0.32383111119270325,
      "step": 542
    },
    {
      "completion_length": 66.625,
      "epoch": 1.1361256544502618,
      "grad_norm": 15.735708236694336,
      "kl": 2.7265625,
      "learning_rate": 7.159329140461215e-07,
      "loss": 0.2679,
      "reward": 0.5548494756221771,
      "reward_std": 0.162759430706501,
      "reward_std_all": 0.19895009696483612,
      "rewards-max/forensics_sinkhorn_reward": 0.9877512156963348,
      "rewards-max/forensics_strict_edge_reward": 0.9565287232398987,
      "rewards-max/forensics_tiou_reward": 0.9951196908950806,
      "rewards/forensics_sinkhorn_reward": 0.46368131041526794,
      "rewards/forensics_strict_edge_reward": 0.2378263622522354,
      "rewards/forensics_tiou_reward": 0.3057974874973297,
      "step": 543
    },
    {
      "completion_length": 35.875,
      "epoch": 1.138219895287958,
      "grad_norm": 20.112546920776367,
      "kl": 2.16015625,
      "learning_rate": 7.154088050314466e-07,
      "loss": 0.2403,
      "reward": 0.6881783306598663,
      "reward_std": 0.11351479217410088,
      "reward_std_all": 0.11663392186164856,
      "rewards-max/forensics_sinkhorn_reward": 0.9886906445026398,
      "rewards-max/forensics_strict_edge_reward": 0.9890114367008209,
      "rewards-max/forensics_tiou_reward": 0.9988186657428741,
      "rewards/forensics_sinkhorn_reward": 0.6122770607471466,
      "rewards/forensics_strict_edge_reward": 0.3287047743797302,
      "rewards/forensics_tiou_reward": 0.410149410367012,
      "step": 544
    },
    {
      "completion_length": 63.078125,
      "epoch": 1.1403141361256544,
      "grad_norm": 14.670991897583008,
      "kl": 2.28515625,
      "learning_rate": 7.148846960167715e-07,
      "loss": 0.237,
      "reward": 0.6155953705310822,
      "reward_std": 0.14380812644958496,
      "reward_std_all": 0.15959220379590988,
      "rewards-max/forensics_sinkhorn_reward": 0.9187544882297516,
      "rewards-max/forensics_strict_edge_reward": 0.6845352947711945,
      "rewards-max/forensics_tiou_reward": 0.8488618433475494,
      "rewards/forensics_sinkhorn_reward": 0.44645553827285767,
      "rewards/forensics_strict_edge_reward": 0.1518145650625229,
      "rewards/forensics_tiou_reward": 0.20954658091068268,
      "step": 545
    },
    {
      "completion_length": 83.421875,
      "epoch": 1.1424083769633508,
      "grad_norm": 25.763029098510742,
      "kl": 3.21875,
      "learning_rate": 7.143605870020965e-07,
      "loss": 0.4489,
      "reward": 0.6152375936508179,
      "reward_std": 0.16789327561855316,
      "reward_std_all": 0.17878465354442596,
      "rewards-max/forensics_sinkhorn_reward": 0.9818741977214813,
      "rewards-max/forensics_strict_edge_reward": 0.8850063383579254,
      "rewards-max/forensics_tiou_reward": 0.9720970988273621,
      "rewards/forensics_sinkhorn_reward": 0.49744588136672974,
      "rewards/forensics_strict_edge_reward": 0.3074953183531761,
      "rewards/forensics_tiou_reward": 0.39565202593803406,
      "step": 546
    },
    {
      "completion_length": 96.9375,
      "epoch": 1.1445026178010471,
      "grad_norm": 16.20183563232422,
      "kl": 3.75,
      "learning_rate": 7.138364779874213e-07,
      "loss": 0.4133,
      "reward": 0.55503910779953,
      "reward_std": 0.15106049180030823,
      "reward_std_all": 0.19426724314689636,
      "rewards-max/forensics_sinkhorn_reward": 0.9865965247154236,
      "rewards-max/forensics_strict_edge_reward": 0.9375866651535034,
      "rewards-max/forensics_tiou_reward": 0.9906105399131775,
      "rewards/forensics_sinkhorn_reward": 0.4328407198190689,
      "rewards/forensics_strict_edge_reward": 0.20595046132802963,
      "rewards/forensics_tiou_reward": 0.29242032021284103,
      "step": 547
    },
    {
      "completion_length": 54.6875,
      "epoch": 1.1465968586387434,
      "grad_norm": 17.862943649291992,
      "kl": 3.2578125,
      "learning_rate": 7.133123689727463e-07,
      "loss": 0.2632,
      "reward": 0.6406417489051819,
      "reward_std": 0.16605371236801147,
      "reward_std_all": 0.16526567190885544,
      "rewards-max/forensics_sinkhorn_reward": 0.9779489934444427,
      "rewards-max/forensics_strict_edge_reward": 0.7745338976383209,
      "rewards-max/forensics_tiou_reward": 0.9648241996765137,
      "rewards/forensics_sinkhorn_reward": 0.4728260785341263,
      "rewards/forensics_strict_edge_reward": 0.20412252098321915,
      "rewards/forensics_tiou_reward": 0.31585003435611725,
      "step": 548
    },
    {
      "completion_length": 60.5,
      "epoch": 1.1486910994764399,
      "grad_norm": 13.282940864562988,
      "kl": 2.76953125,
      "learning_rate": 7.127882599580712e-07,
      "loss": 0.4105,
      "reward": 0.6806958913803101,
      "reward_std": 0.16048834472894669,
      "reward_std_all": 0.14778216183185577,
      "rewards-max/forensics_sinkhorn_reward": 0.988188773393631,
      "rewards-max/forensics_strict_edge_reward": 0.927969753742218,
      "rewards-max/forensics_tiou_reward": 0.9968397319316864,
      "rewards/forensics_sinkhorn_reward": 0.5248627364635468,
      "rewards/forensics_strict_edge_reward": 0.37050169706344604,
      "rewards/forensics_tiou_reward": 0.4205700159072876,
      "step": 549
    },
    {
      "completion_length": 54.765625,
      "epoch": 1.1507853403141362,
      "grad_norm": 17.549663543701172,
      "kl": 2.6328125,
      "learning_rate": 7.122641509433962e-07,
      "loss": 0.1827,
      "reward": 0.692865401506424,
      "reward_std": 0.11839696019887924,
      "reward_std_all": 0.12152637913823128,
      "rewards-max/forensics_sinkhorn_reward": 0.9872001111507416,
      "rewards-max/forensics_strict_edge_reward": 0.9565287232398987,
      "rewards-max/forensics_tiou_reward": 0.9929501116275787,
      "rewards/forensics_sinkhorn_reward": 0.5570747256278992,
      "rewards/forensics_strict_edge_reward": 0.2650385797023773,
      "rewards/forensics_tiou_reward": 0.3572184890508652,
      "step": 550
    },
    {
      "completion_length": 44.71875,
      "epoch": 1.1528795811518324,
      "grad_norm": 23.291484832763672,
      "kl": 3.3828125,
      "learning_rate": 7.117400419287212e-07,
      "loss": 0.1428,
      "reward": 0.6465769708156586,
      "reward_std": 0.1571587473154068,
      "reward_std_all": 0.1652611866593361,
      "rewards-max/forensics_sinkhorn_reward": 0.9861707389354706,
      "rewards-max/forensics_strict_edge_reward": 0.947419673204422,
      "rewards-max/forensics_tiou_reward": 0.9889506101608276,
      "rewards/forensics_sinkhorn_reward": 0.5679609775543213,
      "rewards/forensics_strict_edge_reward": 0.28569769859313965,
      "rewards/forensics_tiou_reward": 0.389731764793396,
      "step": 551
    },
    {
      "completion_length": 61.171875,
      "epoch": 1.154973821989529,
      "grad_norm": 23.232709884643555,
      "kl": 3.109375,
      "learning_rate": 7.112159329140461e-07,
      "loss": 0.2033,
      "reward": 0.6066001057624817,
      "reward_std": 0.13776062428951263,
      "reward_std_all": 0.1638009324669838,
      "rewards-max/forensics_sinkhorn_reward": 0.9759587049484253,
      "rewards-max/forensics_strict_edge_reward": 0.8612285256385803,
      "rewards-max/forensics_tiou_reward": 0.9665421843528748,
      "rewards/forensics_sinkhorn_reward": 0.4541338086128235,
      "rewards/forensics_strict_edge_reward": 0.2355777472257614,
      "rewards/forensics_tiou_reward": 0.2976192981004715,
      "step": 552
    },
    {
      "completion_length": 59.1875,
      "epoch": 1.1570680628272252,
      "grad_norm": 20.257442474365234,
      "kl": 3.8046875,
      "learning_rate": 7.10691823899371e-07,
      "loss": 0.294,
      "reward": 0.6699214279651642,
      "reward_std": 0.14063488692045212,
      "reward_std_all": 0.1402226835489273,
      "rewards-max/forensics_sinkhorn_reward": 0.9872119724750519,
      "rewards-max/forensics_strict_edge_reward": 0.9185642302036285,
      "rewards-max/forensics_tiou_reward": 0.9930152297019958,
      "rewards/forensics_sinkhorn_reward": 0.5668859779834747,
      "rewards/forensics_strict_edge_reward": 0.37781406939029694,
      "rewards/forensics_tiou_reward": 0.4600178003311157,
      "step": 553
    },
    {
      "completion_length": 37.9375,
      "epoch": 1.1591623036649215,
      "grad_norm": 17.274702072143555,
      "kl": 2.40625,
      "learning_rate": 7.101677148846959e-07,
      "loss": 0.0816,
      "reward": 0.6731234192848206,
      "reward_std": 0.10515929758548737,
      "reward_std_all": 0.1185045912861824,
      "rewards-max/forensics_sinkhorn_reward": 0.9721862077713013,
      "rewards-max/forensics_strict_edge_reward": 0.9158510267734528,
      "rewards-max/forensics_tiou_reward": 0.9591796696186066,
      "rewards/forensics_sinkhorn_reward": 0.5233640670776367,
      "rewards/forensics_strict_edge_reward": 0.2544364184141159,
      "rewards/forensics_tiou_reward": 0.30568739771842957,
      "step": 554
    },
    {
      "completion_length": 42.21875,
      "epoch": 1.1612565445026177,
      "grad_norm": 18.86699867248535,
      "kl": 2.619140625,
      "learning_rate": 7.096436058700209e-07,
      "loss": 0.1056,
      "reward": 0.629641205072403,
      "reward_std": 0.09514028578996658,
      "reward_std_all": 0.13021773472428322,
      "rewards-max/forensics_sinkhorn_reward": 0.9886906445026398,
      "rewards-max/forensics_strict_edge_reward": 0.9890114367008209,
      "rewards-max/forensics_tiou_reward": 0.9988186657428741,
      "rewards/forensics_sinkhorn_reward": 0.5291589498519897,
      "rewards/forensics_strict_edge_reward": 0.27860434353351593,
      "rewards/forensics_tiou_reward": 0.3243657797574997,
      "step": 555
    },
    {
      "completion_length": 61.046875,
      "epoch": 1.1633507853403142,
      "grad_norm": 16.802316665649414,
      "kl": 3.125,
      "learning_rate": 7.091194968553458e-07,
      "loss": 0.3215,
      "reward": 0.6467269361019135,
      "reward_std": 0.16824446991086006,
      "reward_std_all": 0.16154425963759422,
      "rewards-max/forensics_sinkhorn_reward": 0.9875765740871429,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9944568872451782,
      "rewards/forensics_sinkhorn_reward": 0.5213627219200134,
      "rewards/forensics_strict_edge_reward": 0.2908647507429123,
      "rewards/forensics_tiou_reward": 0.3591051250696182,
      "step": 556
    },
    {
      "completion_length": 45.171875,
      "epoch": 1.1654450261780105,
      "grad_norm": 23.65601348876953,
      "kl": 3.421875,
      "learning_rate": 7.085953878406709e-07,
      "loss": 0.3193,
      "reward": 0.6060139834880829,
      "reward_std": 0.1369483843445778,
      "reward_std_all": 0.16313155740499496,
      "rewards-max/forensics_sinkhorn_reward": 0.9752991199493408,
      "rewards-max/forensics_strict_edge_reward": 0.8663162291049957,
      "rewards-max/forensics_tiou_reward": 0.9692237377166748,
      "rewards/forensics_sinkhorn_reward": 0.47162285447120667,
      "rewards/forensics_strict_edge_reward": 0.26591673493385315,
      "rewards/forensics_tiou_reward": 0.3245232105255127,
      "step": 557
    },
    {
      "completion_length": 52.359375,
      "epoch": 1.1675392670157068,
      "grad_norm": 16.771696090698242,
      "kl": 3.25,
      "learning_rate": 7.080712788259959e-07,
      "loss": 0.2929,
      "reward": 0.684045135974884,
      "reward_std": 0.13603021204471588,
      "reward_std_all": 0.13415087014436722,
      "rewards-max/forensics_sinkhorn_reward": 0.9871909618377686,
      "rewards-max/forensics_strict_edge_reward": 0.9574736058712006,
      "rewards-max/forensics_tiou_reward": 0.9929329454898834,
      "rewards/forensics_sinkhorn_reward": 0.5958112478256226,
      "rewards/forensics_strict_edge_reward": 0.39560043811798096,
      "rewards/forensics_tiou_reward": 0.4155190885066986,
      "step": 558
    },
    {
      "completion_length": 67.421875,
      "epoch": 1.169633507853403,
      "grad_norm": 17.57123374938965,
      "kl": 3.4453125,
      "learning_rate": 7.075471698113207e-07,
      "loss": 0.2686,
      "reward": 0.6605044603347778,
      "reward_std": 0.13613655418157578,
      "reward_std_all": 0.13651373237371445,
      "rewards-max/forensics_sinkhorn_reward": 0.9879686832427979,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9959746599197388,
      "rewards/forensics_sinkhorn_reward": 0.525356113910675,
      "rewards/forensics_strict_edge_reward": 0.31552577018737793,
      "rewards/forensics_tiou_reward": 0.37385955452919006,
      "step": 559
    },
    {
      "completion_length": 89.515625,
      "epoch": 1.1717277486910995,
      "grad_norm": 21.410078048706055,
      "kl": 5.8125,
      "learning_rate": 7.070230607966457e-07,
      "loss": 0.4134,
      "reward": 0.6262194216251373,
      "reward_std": 0.16561391949653625,
      "reward_std_all": 0.17078930139541626,
      "rewards-max/forensics_sinkhorn_reward": 0.9880367815494537,
      "rewards-max/forensics_strict_edge_reward": 0.9265981018543243,
      "rewards-max/forensics_tiou_reward": 0.996242493391037,
      "rewards/forensics_sinkhorn_reward": 0.4730255901813507,
      "rewards/forensics_strict_edge_reward": 0.23407060652971268,
      "rewards/forensics_tiou_reward": 0.2854003310203552,
      "step": 560
    },
    {
      "completion_length": 54.59375,
      "epoch": 1.1738219895287958,
      "grad_norm": 16.850170135498047,
      "kl": 3.2734375,
      "learning_rate": 7.064989517819706e-07,
      "loss": 0.3938,
      "reward": 0.6317216753959656,
      "reward_std": 0.15443499386310577,
      "reward_std_all": 0.16017769277095795,
      "rewards-max/forensics_sinkhorn_reward": 0.9883950054645538,
      "rewards-max/forensics_strict_edge_reward": 0.9890114367008209,
      "rewards-max/forensics_tiou_reward": 0.9976553320884705,
      "rewards/forensics_sinkhorn_reward": 0.521816223859787,
      "rewards/forensics_strict_edge_reward": 0.28927579522132874,
      "rewards/forensics_tiou_reward": 0.3623223155736923,
      "step": 561
    },
    {
      "completion_length": 62.703125,
      "epoch": 1.175916230366492,
      "grad_norm": 20.55358123779297,
      "kl": 4.375,
      "learning_rate": 7.059748427672956e-07,
      "loss": 0.4299,
      "reward": 0.5662661194801331,
      "reward_std": 0.24402376264333725,
      "reward_std_all": 0.2389684021472931,
      "rewards-max/forensics_sinkhorn_reward": 0.9881008863449097,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9964931011199951,
      "rewards/forensics_sinkhorn_reward": 0.4728384464979172,
      "rewards/forensics_strict_edge_reward": 0.26940417289733887,
      "rewards/forensics_tiou_reward": 0.350188747048378,
      "step": 562
    },
    {
      "completion_length": 79.0,
      "epoch": 1.1780104712041886,
      "grad_norm": 20.882068634033203,
      "kl": 4.375,
      "learning_rate": 7.054507337526205e-07,
      "loss": 0.2711,
      "reward": 0.5377094596624374,
      "reward_std": 0.15510165691375732,
      "reward_std_all": 0.2025645226240158,
      "rewards-max/forensics_sinkhorn_reward": 0.9480679929256439,
      "rewards-max/forensics_strict_edge_reward": 0.7534597218036652,
      "rewards-max/forensics_tiou_reward": 0.9058429896831512,
      "rewards/forensics_sinkhorn_reward": 0.3714349716901779,
      "rewards/forensics_strict_edge_reward": 0.15960855036973953,
      "rewards/forensics_tiou_reward": 0.19983533769845963,
      "step": 563
    },
    {
      "completion_length": 75.765625,
      "epoch": 1.1801047120418848,
      "grad_norm": 20.250822067260742,
      "kl": 5.5625,
      "learning_rate": 7.049266247379455e-07,
      "loss": 0.4751,
      "reward": 0.6004819273948669,
      "reward_std": 0.2362009584903717,
      "reward_std_all": 0.2235044687986374,
      "rewards-max/forensics_sinkhorn_reward": 0.9669939279556274,
      "rewards-max/forensics_strict_edge_reward": 0.7676678597927094,
      "rewards-max/forensics_tiou_reward": 0.9433187246322632,
      "rewards/forensics_sinkhorn_reward": 0.4556818902492523,
      "rewards/forensics_strict_edge_reward": 0.2498968094587326,
      "rewards/forensics_tiou_reward": 0.29861344397068024,
      "step": 564
    },
    {
      "completion_length": 61.40625,
      "epoch": 1.182198952879581,
      "grad_norm": 20.574417114257812,
      "kl": 3.3359375,
      "learning_rate": 7.044025157232703e-07,
      "loss": 0.3152,
      "reward": 0.6310699582099915,
      "reward_std": 0.16392941027879715,
      "reward_std_all": 0.1668136566877365,
      "rewards-max/forensics_sinkhorn_reward": 0.9875226616859436,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.994225025177002,
      "rewards/forensics_sinkhorn_reward": 0.5222769230604172,
      "rewards/forensics_strict_edge_reward": 0.31687110662460327,
      "rewards/forensics_tiou_reward": 0.34840531647205353,
      "step": 565
    },
    {
      "completion_length": 64.03125,
      "epoch": 1.1842931937172776,
      "grad_norm": 20.083492279052734,
      "kl": 3.296875,
      "learning_rate": 7.038784067085953e-07,
      "loss": 0.1742,
      "reward": 0.6102859377861023,
      "reward_std": 0.18788541853427887,
      "reward_std_all": 0.19352395087480545,
      "rewards-max/forensics_sinkhorn_reward": 0.9868795871734619,
      "rewards-max/forensics_strict_edge_reward": 0.9565287232398987,
      "rewards-max/forensics_tiou_reward": 0.9916878938674927,
      "rewards/forensics_sinkhorn_reward": 0.4987547844648361,
      "rewards/forensics_strict_edge_reward": 0.25531867146492004,
      "rewards/forensics_tiou_reward": 0.3499922454357147,
      "step": 566
    },
    {
      "completion_length": 78.265625,
      "epoch": 1.1863874345549739,
      "grad_norm": 16.036184310913086,
      "kl": 3.421875,
      "learning_rate": 7.033542976939203e-07,
      "loss": 0.1808,
      "reward": 0.6635226011276245,
      "reward_std": 0.1432429924607277,
      "reward_std_all": 0.1419905424118042,
      "rewards-max/forensics_sinkhorn_reward": 0.9721633195877075,
      "rewards-max/forensics_strict_edge_reward": 0.8751733303070068,
      "rewards-max/forensics_tiou_reward": 0.9503344297409058,
      "rewards/forensics_sinkhorn_reward": 0.4984682649374008,
      "rewards/forensics_strict_edge_reward": 0.28994984924793243,
      "rewards/forensics_tiou_reward": 0.3579219877719879,
      "step": 567
    },
    {
      "completion_length": 47.734375,
      "epoch": 1.1884816753926701,
      "grad_norm": 19.029415130615234,
      "kl": 3.6171875,
      "learning_rate": 7.028301886792452e-07,
      "loss": 0.2776,
      "reward": 0.6539453268051147,
      "reward_std": 0.11299245432019234,
      "reward_std_all": 0.13075942546129227,
      "rewards-max/forensics_sinkhorn_reward": 0.9855373501777649,
      "rewards-max/forensics_strict_edge_reward": 0.882964164018631,
      "rewards-max/forensics_tiou_reward": 0.9864871203899384,
      "rewards/forensics_sinkhorn_reward": 0.5594481527805328,
      "rewards/forensics_strict_edge_reward": 0.2930300161242485,
      "rewards/forensics_tiou_reward": 0.3546089380979538,
      "step": 568
    },
    {
      "completion_length": 52.34375,
      "epoch": 1.1905759162303664,
      "grad_norm": 18.96901512145996,
      "kl": 3.171875,
      "learning_rate": 7.023060796645703e-07,
      "loss": 0.3318,
      "reward": 0.6186900734901428,
      "reward_std": 0.17300888150930405,
      "reward_std_all": 0.17871225625276566,
      "rewards-max/forensics_sinkhorn_reward": 0.987417072057724,
      "rewards-max/forensics_strict_edge_reward": 0.9567649364471436,
      "rewards-max/forensics_tiou_reward": 0.9938099980354309,
      "rewards/forensics_sinkhorn_reward": 0.4394318610429764,
      "rewards/forensics_strict_edge_reward": 0.29229308664798737,
      "rewards/forensics_tiou_reward": 0.3006819188594818,
      "step": 569
    },
    {
      "completion_length": 41.890625,
      "epoch": 1.192670157068063,
      "grad_norm": 16.574678421020508,
      "kl": 2.4453125,
      "learning_rate": 7.017819706498952e-07,
      "loss": 0.1437,
      "reward": 0.6258413195610046,
      "reward_std": 0.16814392805099487,
      "reward_std_all": 0.1682438626885414,
      "rewards-max/forensics_sinkhorn_reward": 0.9870999455451965,
      "rewards-max/forensics_strict_edge_reward": 0.9567649364471436,
      "rewards-max/forensics_tiou_reward": 0.9925644099712372,
      "rewards/forensics_sinkhorn_reward": 0.5060243904590607,
      "rewards/forensics_strict_edge_reward": 0.23227562010288239,
      "rewards/forensics_tiou_reward": 0.34809477627277374,
      "step": 570
    },
    {
      "completion_length": 54.8125,
      "epoch": 1.1947643979057592,
      "grad_norm": 18.28180503845215,
      "kl": 3.6953125,
      "learning_rate": 7.012578616352201e-07,
      "loss": 0.4619,
      "reward": 0.5671627223491669,
      "reward_std": 0.2361893653869629,
      "reward_std_all": 0.23337381333112717,
      "rewards-max/forensics_sinkhorn_reward": 0.9871013164520264,
      "rewards-max/forensics_strict_edge_reward": 0.9565287232398987,
      "rewards-max/forensics_tiou_reward": 0.9925641715526581,
      "rewards/forensics_sinkhorn_reward": 0.46177130937576294,
      "rewards/forensics_strict_edge_reward": 0.25508376955986023,
      "rewards/forensics_tiou_reward": 0.3097843527793884,
      "step": 571
    },
    {
      "completion_length": 43.171875,
      "epoch": 1.1968586387434554,
      "grad_norm": 21.94469451904297,
      "kl": 2.5703125,
      "learning_rate": 7.00733752620545e-07,
      "loss": 0.3834,
      "reward": 0.6920926570892334,
      "reward_std": 0.13812628760933876,
      "reward_std_all": 0.13366016373038292,
      "rewards-max/forensics_sinkhorn_reward": 0.9884631037712097,
      "rewards-max/forensics_strict_edge_reward": 0.9782643616199493,
      "rewards-max/forensics_tiou_reward": 0.9979231655597687,
      "rewards/forensics_sinkhorn_reward": 0.603813648223877,
      "rewards/forensics_strict_edge_reward": 0.40145663917064667,
      "rewards/forensics_tiou_reward": 0.46811527013778687,
      "step": 572
    },
    {
      "completion_length": 50.859375,
      "epoch": 1.1989528795811517,
      "grad_norm": 14.901327133178711,
      "kl": 2.6796875,
      "learning_rate": 7.0020964360587e-07,
      "loss": 0.2555,
      "reward": 0.6145595908164978,
      "reward_std": 0.13912352919578552,
      "reward_std_all": 0.15916795656085014,
      "rewards-max/forensics_sinkhorn_reward": 0.9171607792377472,
      "rewards-max/forensics_strict_edge_reward": 0.7918089032173157,
      "rewards-max/forensics_tiou_reward": 0.8537468910217285,
      "rewards/forensics_sinkhorn_reward": 0.4631106108427048,
      "rewards/forensics_strict_edge_reward": 0.2537044882774353,
      "rewards/forensics_tiou_reward": 0.337995283305645,
      "step": 573
    },
    {
      "completion_length": 44.03125,
      "epoch": 1.2010471204188482,
      "grad_norm": 17.544099807739258,
      "kl": 3.1171875,
      "learning_rate": 6.99685534591195e-07,
      "loss": 0.4,
      "reward": 0.5911522507667542,
      "reward_std": 0.1983940750360489,
      "reward_std_all": 0.20377495139837265,
      "rewards-max/forensics_sinkhorn_reward": 0.98691126704216,
      "rewards-max/forensics_strict_edge_reward": 0.9464850425720215,
      "rewards-max/forensics_tiou_reward": 0.9918244779109955,
      "rewards/forensics_sinkhorn_reward": 0.4365100562572479,
      "rewards/forensics_strict_edge_reward": 0.18767789751291275,
      "rewards/forensics_tiou_reward": 0.2812056243419647,
      "step": 574
    },
    {
      "completion_length": 31.96875,
      "epoch": 1.2031413612565445,
      "grad_norm": 17.38408660888672,
      "kl": 2.09765625,
      "learning_rate": 6.991614255765199e-07,
      "loss": 0.2852,
      "reward": 0.647252768278122,
      "reward_std": 0.12677733600139618,
      "reward_std_all": 0.14605145156383514,
      "rewards-max/forensics_sinkhorn_reward": 0.987781822681427,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9952354431152344,
      "rewards/forensics_sinkhorn_reward": 0.6177314519882202,
      "rewards/forensics_strict_edge_reward": 0.31263838708400726,
      "rewards/forensics_tiou_reward": 0.39359718561172485,
      "step": 575
    },
    {
      "completion_length": 38.3125,
      "epoch": 1.2052356020942407,
      "grad_norm": 16.725690841674805,
      "kl": 2.1953125,
      "learning_rate": 6.986373165618449e-07,
      "loss": 0.2736,
      "reward": 0.6399997472763062,
      "reward_std": 0.12453746795654297,
      "reward_std_all": 0.1455679088830948,
      "rewards-max/forensics_sinkhorn_reward": 0.9880594909191132,
      "rewards-max/forensics_strict_edge_reward": 0.9574736058712006,
      "rewards-max/forensics_tiou_reward": 0.9963302612304688,
      "rewards/forensics_sinkhorn_reward": 0.5232207179069519,
      "rewards/forensics_strict_edge_reward": 0.30365848541259766,
      "rewards/forensics_tiou_reward": 0.34390875697135925,
      "step": 576
    },
    {
      "completion_length": 35.265625,
      "epoch": 1.2073298429319372,
      "grad_norm": 14.872918128967285,
      "kl": 2.546875,
      "learning_rate": 6.981132075471697e-07,
      "loss": 0.2834,
      "reward": 0.663373202085495,
      "reward_std": 0.12036080285906792,
      "reward_std_all": 0.12951237335801125,
      "rewards-max/forensics_sinkhorn_reward": 0.951621800661087,
      "rewards-max/forensics_strict_edge_reward": 0.8562542200088501,
      "rewards-max/forensics_tiou_reward": 0.915681004524231,
      "rewards/forensics_sinkhorn_reward": 0.5220522880554199,
      "rewards/forensics_strict_edge_reward": 0.26266953349113464,
      "rewards/forensics_tiou_reward": 0.36082689464092255,
      "step": 577
    },
    {
      "completion_length": 41.9375,
      "epoch": 1.2094240837696335,
      "grad_norm": 18.710134506225586,
      "kl": 3.25,
      "learning_rate": 6.975890985324947e-07,
      "loss": 0.2818,
      "reward": 0.6455201506614685,
      "reward_std": 0.16870394349098206,
      "reward_std_all": 0.16274286061525345,
      "rewards-max/forensics_sinkhorn_reward": 0.98871049284935,
      "rewards-max/forensics_strict_edge_reward": 0.9782643616199493,
      "rewards-max/forensics_tiou_reward": 0.9988968670368195,
      "rewards/forensics_sinkhorn_reward": 0.5295265018939972,
      "rewards/forensics_strict_edge_reward": 0.3301871716976166,
      "rewards/forensics_tiou_reward": 0.35841265320777893,
      "step": 578
    },
    {
      "completion_length": 49.484375,
      "epoch": 1.2115183246073298,
      "grad_norm": 21.23697280883789,
      "kl": 3.7734375,
      "learning_rate": 6.970649895178196e-07,
      "loss": 0.3766,
      "reward": 0.5958080589771271,
      "reward_std": 0.16055088490247726,
      "reward_std_all": 0.17777476459741592,
      "rewards-max/forensics_sinkhorn_reward": 0.9884815514087677,
      "rewards-max/forensics_strict_edge_reward": 0.9567649364471436,
      "rewards-max/forensics_tiou_reward": 0.9979939460754395,
      "rewards/forensics_sinkhorn_reward": 0.46190980076789856,
      "rewards/forensics_strict_edge_reward": 0.265601247549057,
      "rewards/forensics_tiou_reward": 0.31001825630664825,
      "step": 579
    },
    {
      "completion_length": 69.1875,
      "epoch": 1.2136125654450263,
      "grad_norm": 18.513206481933594,
      "kl": 4.578125,
      "learning_rate": 6.965408805031446e-07,
      "loss": 0.1901,
      "reward": 0.630945473909378,
      "reward_std": 0.21257217973470688,
      "reward_std_all": 0.2011997103691101,
      "rewards-max/forensics_sinkhorn_reward": 0.9430146217346191,
      "rewards-max/forensics_strict_edge_reward": 0.7926072478294373,
      "rewards-max/forensics_tiou_reward": 0.9006243646144867,
      "rewards/forensics_sinkhorn_reward": 0.472220242023468,
      "rewards/forensics_strict_edge_reward": 0.2500711604952812,
      "rewards/forensics_tiou_reward": 0.27598871290683746,
      "step": 580
    },
    {
      "completion_length": 67.75,
      "epoch": 1.2157068062827225,
      "grad_norm": 21.14455795288086,
      "kl": 4.609375,
      "learning_rate": 6.960167714884695e-07,
      "loss": 0.3976,
      "reward": 0.649256557226181,
      "reward_std": 0.09390716999769211,
      "reward_std_all": 0.12505920976400375,
      "rewards-max/forensics_sinkhorn_reward": 0.9348048865795135,
      "rewards-max/forensics_strict_edge_reward": 0.835968405008316,
      "rewards-max/forensics_tiou_reward": 0.877520889043808,
      "rewards/forensics_sinkhorn_reward": 0.477204367518425,
      "rewards/forensics_strict_edge_reward": 0.2622798681259155,
      "rewards/forensics_tiou_reward": 0.3170240521430969,
      "step": 581
    },
    {
      "completion_length": 78.609375,
      "epoch": 1.2178010471204188,
      "grad_norm": 25.633834838867188,
      "kl": 4.8125,
      "learning_rate": 6.954926624737946e-07,
      "loss": 0.3299,
      "reward": 0.5631152987480164,
      "reward_std": 0.21244611591100693,
      "reward_std_all": 0.2248181328177452,
      "rewards-max/forensics_sinkhorn_reward": 0.988304853439331,
      "rewards-max/forensics_strict_edge_reward": 0.9782643616199493,
      "rewards-max/forensics_tiou_reward": 0.9972972869873047,
      "rewards/forensics_sinkhorn_reward": 0.4605614095926285,
      "rewards/forensics_strict_edge_reward": 0.2516847401857376,
      "rewards/forensics_tiou_reward": 0.34755199402570724,
      "step": 582
    },
    {
      "completion_length": 78.09375,
      "epoch": 1.2198952879581153,
      "grad_norm": 20.616601943969727,
      "kl": 5.0,
      "learning_rate": 6.949685534591195e-07,
      "loss": 0.251,
      "reward": 0.6131627261638641,
      "reward_std": 0.1268831081688404,
      "reward_std_all": 0.16016462445259094,
      "rewards-max/forensics_sinkhorn_reward": 0.9880367815494537,
      "rewards-max/forensics_strict_edge_reward": 0.9158510267734528,
      "rewards-max/forensics_tiou_reward": 0.996242493391037,
      "rewards/forensics_sinkhorn_reward": 0.5093908905982971,
      "rewards/forensics_strict_edge_reward": 0.2342163696885109,
      "rewards/forensics_tiou_reward": 0.30651649832725525,
      "step": 583
    },
    {
      "completion_length": 78.90625,
      "epoch": 1.2219895287958116,
      "grad_norm": 22.729698181152344,
      "kl": 5.25,
      "learning_rate": 6.944444444444444e-07,
      "loss": 0.3745,
      "reward": 0.6598003804683685,
      "reward_std": 0.1390436477959156,
      "reward_std_all": 0.13983823359012604,
      "rewards-max/forensics_sinkhorn_reward": 0.988023579120636,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9961899816989899,
      "rewards/forensics_sinkhorn_reward": 0.5388108491897583,
      "rewards/forensics_strict_edge_reward": 0.3056163787841797,
      "rewards/forensics_tiou_reward": 0.3560701906681061,
      "step": 584
    },
    {
      "completion_length": 75.796875,
      "epoch": 1.2240837696335078,
      "grad_norm": 30.01637077331543,
      "kl": 5.78125,
      "learning_rate": 6.939203354297694e-07,
      "loss": 0.3206,
      "reward": 0.5686949193477631,
      "reward_std": 0.22953256964683533,
      "reward_std_all": 0.22678089886903763,
      "rewards-max/forensics_sinkhorn_reward": 0.953315258026123,
      "rewards-max/forensics_strict_edge_reward": 0.9217896461486816,
      "rewards-max/forensics_tiou_reward": 0.9261489510536194,
      "rewards/forensics_sinkhorn_reward": 0.4249603599309921,
      "rewards/forensics_strict_edge_reward": 0.18164243549108505,
      "rewards/forensics_tiou_reward": 0.2370169311761856,
      "step": 585
    },
    {
      "completion_length": 95.53125,
      "epoch": 1.2261780104712041,
      "grad_norm": 24.212711334228516,
      "kl": 6.390625,
      "learning_rate": 6.933962264150943e-07,
      "loss": 0.3709,
      "reward": 0.5926173031330109,
      "reward_std": 0.16562245786190033,
      "reward_std_all": 0.1903555542230606,
      "rewards-max/forensics_sinkhorn_reward": 0.9875659644603729,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9943926930427551,
      "rewards/forensics_sinkhorn_reward": 0.47323213517665863,
      "rewards/forensics_strict_edge_reward": 0.22874954342842102,
      "rewards/forensics_tiou_reward": 0.3269018605351448,
      "step": 586
    },
    {
      "completion_length": 132.0,
      "epoch": 1.2282722513089006,
      "grad_norm": 24.30720329284668,
      "kl": 6.46875,
      "learning_rate": 6.928721174004193e-07,
      "loss": 0.535,
      "reward": 0.636871725320816,
      "reward_std": 0.1771121472120285,
      "reward_std_all": 0.1749364584684372,
      "rewards-max/forensics_sinkhorn_reward": 0.9874401390552521,
      "rewards-max/forensics_strict_edge_reward": 0.946017861366272,
      "rewards-max/forensics_tiou_reward": 0.9938889741897583,
      "rewards/forensics_sinkhorn_reward": 0.4747569411993027,
      "rewards/forensics_strict_edge_reward": 0.31376057118177414,
      "rewards/forensics_tiou_reward": 0.38875506818294525,
      "step": 587
    },
    {
      "completion_length": 80.078125,
      "epoch": 1.2303664921465969,
      "grad_norm": 19.975467681884766,
      "kl": 4.453125,
      "learning_rate": 6.923480083857442e-07,
      "loss": 0.4213,
      "reward": 0.6110675632953644,
      "reward_std": 0.14113274589180946,
      "reward_std_all": 0.16132321953773499,
      "rewards-max/forensics_sinkhorn_reward": 0.9499851167201996,
      "rewards-max/forensics_strict_edge_reward": 0.915417492389679,
      "rewards-max/forensics_tiou_reward": 0.9122388362884521,
      "rewards/forensics_sinkhorn_reward": 0.4675663411617279,
      "rewards/forensics_strict_edge_reward": 0.2189214825630188,
      "rewards/forensics_tiou_reward": 0.3162681311368942,
      "step": 588
    },
    {
      "completion_length": 49.359375,
      "epoch": 1.2324607329842932,
      "grad_norm": 30.638736724853516,
      "kl": 3.5234375,
      "learning_rate": 6.918238993710691e-07,
      "loss": 0.371,
      "reward": 0.6681190729141235,
      "reward_std": 0.18379200994968414,
      "reward_std_all": 0.1676800772547722,
      "rewards-max/forensics_sinkhorn_reward": 0.9883087277412415,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.997314065694809,
      "rewards/forensics_sinkhorn_reward": 0.5758337378501892,
      "rewards/forensics_strict_edge_reward": 0.36133089661598206,
      "rewards/forensics_tiou_reward": 0.4221174865961075,
      "step": 589
    },
    {
      "completion_length": 70.171875,
      "epoch": 1.2345549738219894,
      "grad_norm": 20.572982788085938,
      "kl": 4.71875,
      "learning_rate": 6.912997903563941e-07,
      "loss": 0.3817,
      "reward": 0.6251569092273712,
      "reward_std": 0.18570983409881592,
      "reward_std_all": 0.18249809741973877,
      "rewards-max/forensics_sinkhorn_reward": 0.9585427939891815,
      "rewards-max/forensics_strict_edge_reward": 0.8445805907249451,
      "rewards-max/forensics_tiou_reward": 0.9321869015693665,
      "rewards/forensics_sinkhorn_reward": 0.4886086881160736,
      "rewards/forensics_strict_edge_reward": 0.20531349629163742,
      "rewards/forensics_tiou_reward": 0.2965753674507141,
      "step": 590
    },
    {
      "completion_length": 46.65625,
      "epoch": 1.236649214659686,
      "grad_norm": 19.320030212402344,
      "kl": 3.6875,
      "learning_rate": 6.90775681341719e-07,
      "loss": 0.3512,
      "reward": 0.6719376444816589,
      "reward_std": 0.1459192931652069,
      "reward_std_all": 0.13981686159968376,
      "rewards-max/forensics_sinkhorn_reward": 0.9720290005207062,
      "rewards-max/forensics_strict_edge_reward": 0.9677534997463226,
      "rewards-max/forensics_tiou_reward": 0.9629629552364349,
      "rewards/forensics_sinkhorn_reward": 0.566482663154602,
      "rewards/forensics_strict_edge_reward": 0.34823331236839294,
      "rewards/forensics_tiou_reward": 0.4249439090490341,
      "step": 591
    },
    {
      "completion_length": 28.921875,
      "epoch": 1.2387434554973822,
      "grad_norm": 16.107995986938477,
      "kl": 3.265625,
      "learning_rate": 6.90251572327044e-07,
      "loss": 0.2301,
      "reward": 0.6887480914592743,
      "reward_std": 0.09758677333593369,
      "reward_std_all": 0.11107420176267624,
      "rewards-max/forensics_sinkhorn_reward": 0.9586299359798431,
      "rewards-max/forensics_strict_edge_reward": 0.8823045492172241,
      "rewards-max/forensics_tiou_reward": 0.9335608184337616,
      "rewards/forensics_sinkhorn_reward": 0.556813657283783,
      "rewards/forensics_strict_edge_reward": 0.2898898646235466,
      "rewards/forensics_tiou_reward": 0.3509506359696388,
      "step": 592
    },
    {
      "completion_length": 48.75,
      "epoch": 1.2408376963350785,
      "grad_norm": 32.42523956298828,
      "kl": 3.578125,
      "learning_rate": 6.89727463312369e-07,
      "loss": 0.2403,
      "reward": 0.6537601947784424,
      "reward_std": 0.17165300250053406,
      "reward_std_all": 0.1649547815322876,
      "rewards-max/forensics_sinkhorn_reward": 0.9828588962554932,
      "rewards-max/forensics_strict_edge_reward": 0.8347242176532745,
      "rewards-max/forensics_tiou_reward": 0.9760220348834991,
      "rewards/forensics_sinkhorn_reward": 0.5184071063995361,
      "rewards/forensics_strict_edge_reward": 0.21799171715974808,
      "rewards/forensics_tiou_reward": 0.32635027170181274,
      "step": 593
    },
    {
      "completion_length": 48.15625,
      "epoch": 1.242931937172775,
      "grad_norm": 24.018640518188477,
      "kl": 5.0234375,
      "learning_rate": 6.89203354297694e-07,
      "loss": 0.337,
      "reward": 0.6332803070545197,
      "reward_std": 0.12808115407824516,
      "reward_std_all": 0.1450929306447506,
      "rewards-max/forensics_sinkhorn_reward": 0.9849994480609894,
      "rewards-max/forensics_strict_edge_reward": 0.7873914241790771,
      "rewards-max/forensics_tiou_reward": 0.9844003319740295,
      "rewards/forensics_sinkhorn_reward": 0.4598941206932068,
      "rewards/forensics_strict_edge_reward": 0.2569776102900505,
      "rewards/forensics_tiou_reward": 0.28243082761764526,
      "step": 594
    },
    {
      "completion_length": 26.015625,
      "epoch": 1.2450261780104712,
      "grad_norm": 19.55642318725586,
      "kl": 2.5390625,
      "learning_rate": 6.886792452830188e-07,
      "loss": 0.1766,
      "reward": 0.641263484954834,
      "reward_std": 0.1855187565088272,
      "reward_std_all": 0.17883770167827606,
      "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": 0.541014552116394,
      "rewards/forensics_strict_edge_reward": 0.24729209393262863,
      "rewards/forensics_tiou_reward": 0.3056720197200775,
      "step": 595
    },
    {
      "completion_length": 42.046875,
      "epoch": 1.2471204188481675,
      "grad_norm": 26.213293075561523,
      "kl": 3.1796875,
      "learning_rate": 6.881551362683438e-07,
      "loss": 0.2605,
      "reward": 0.6514906585216522,
      "reward_std": 0.12574425339698792,
      "reward_std_all": 0.1371607445180416,
      "rewards-max/forensics_sinkhorn_reward": 0.9852094352245331,
      "rewards-max/forensics_strict_edge_reward": 0.9567649364471436,
      "rewards-max/forensics_tiou_reward": 0.9913322627544403,
      "rewards/forensics_sinkhorn_reward": 0.49908873438835144,
      "rewards/forensics_strict_edge_reward": 0.21819330006837845,
      "rewards/forensics_tiou_reward": 0.27036987245082855,
      "step": 596
    },
    {
      "completion_length": 29.65625,
      "epoch": 1.249214659685864,
      "grad_norm": 21.87642478942871,
      "kl": 3.5,
      "learning_rate": 6.876310272536687e-07,
      "loss": 0.3511,
      "reward": 0.6282182931900024,
      "reward_std": 0.12069518491625786,
      "reward_std_all": 0.14653868228197098,
      "rewards-max/forensics_sinkhorn_reward": 0.9613209664821625,
      "rewards-max/forensics_strict_edge_reward": 0.9075756669044495,
      "rewards-max/forensics_tiou_reward": 0.9396018087863922,
      "rewards/forensics_sinkhorn_reward": 0.4783800542354584,
      "rewards/forensics_strict_edge_reward": 0.28003574907779694,
      "rewards/forensics_tiou_reward": 0.3382870703935623,
      "step": 597
    },
    {
      "completion_length": 33.9375,
      "epoch": 1.2513089005235603,
      "grad_norm": 25.283308029174805,
      "kl": 3.6328125,
      "learning_rate": 6.871069182389937e-07,
      "loss": 0.2434,
      "reward": 0.604846328496933,
      "reward_std": 0.10788432508707047,
      "reward_std_all": 0.14503175020217896,
      "rewards-max/forensics_sinkhorn_reward": 0.9879895448684692,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9960554540157318,
      "rewards/forensics_sinkhorn_reward": 0.46548232436180115,
      "rewards/forensics_strict_edge_reward": 0.18494533747434616,
      "rewards/forensics_tiou_reward": 0.22207928448915482,
      "step": 598
    },
    {
      "completion_length": 56.5625,
      "epoch": 1.2534031413612565,
      "grad_norm": 21.197669982910156,
      "kl": 6.25,
      "learning_rate": 6.865828092243187e-07,
      "loss": 0.3872,
      "reward": 0.6642890572547913,
      "reward_std": 0.1606880947947502,
      "reward_std_all": 0.1523211970925331,
      "rewards-max/forensics_sinkhorn_reward": 0.9706617295742035,
      "rewards-max/forensics_strict_edge_reward": 0.882964164018631,
      "rewards-max/forensics_tiou_reward": 0.9580255150794983,
      "rewards/forensics_sinkhorn_reward": 0.4817471355199814,
      "rewards/forensics_strict_edge_reward": 0.29896190762519836,
      "rewards/forensics_tiou_reward": 0.32015347480773926,
      "step": 599
    },
    {
      "completion_length": 35.84375,
      "epoch": 1.255497382198953,
      "grad_norm": 28.164331436157227,
      "kl": 4.515625,
      "learning_rate": 6.860587002096436e-07,
      "loss": 0.2882,
      "reward": 0.6300339698791504,
      "reward_std": 0.18433037400245667,
      "reward_std_all": 0.18215418606996536,
      "rewards-max/forensics_sinkhorn_reward": 0.9848147332668304,
      "rewards-max/forensics_strict_edge_reward": 0.9062341153621674,
      "rewards-max/forensics_tiou_reward": 0.9836256802082062,
      "rewards/forensics_sinkhorn_reward": 0.5054740905761719,
      "rewards/forensics_strict_edge_reward": 0.24401016533374786,
      "rewards/forensics_tiou_reward": 0.3104161322116852,
      "step": 600
    },
    {
      "completion_length": 34.0,
      "epoch": 1.2575916230366493,
      "grad_norm": 20.2867488861084,
      "kl": 3.578125,
      "learning_rate": 6.855345911949685e-07,
      "loss": 0.212,
      "reward": 0.6646488606929779,
      "reward_std": 0.16959497705101967,
      "reward_std_all": 0.16131852567195892,
      "rewards-max/forensics_sinkhorn_reward": 0.9880402386188507,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9964445531368256,
      "rewards/forensics_sinkhorn_reward": 0.5878801047801971,
      "rewards/forensics_strict_edge_reward": 0.379361167550087,
      "rewards/forensics_tiou_reward": 0.4176354855298996,
      "step": 601
    },
    {
      "completion_length": 37.796875,
      "epoch": 1.2596858638743456,
      "grad_norm": 22.66136932373047,
      "kl": 4.3046875,
      "learning_rate": 6.850104821802934e-07,
      "loss": 0.3693,
      "reward": 0.6354508996009827,
      "reward_std": 0.12796089053153992,
      "reward_std_all": 0.1508554369211197,
      "rewards-max/forensics_sinkhorn_reward": 0.9390729665756226,
      "rewards-max/forensics_strict_edge_reward": 0.7901110649108887,
      "rewards-max/forensics_tiou_reward": 0.8861020505428314,
      "rewards/forensics_sinkhorn_reward": 0.45702457427978516,
      "rewards/forensics_strict_edge_reward": 0.2523859068751335,
      "rewards/forensics_tiou_reward": 0.3011096715927124,
      "step": 602
    },
    {
      "completion_length": 30.421875,
      "epoch": 1.2617801047120418,
      "grad_norm": 15.697944641113281,
      "kl": 2.05078125,
      "learning_rate": 6.844863731656184e-07,
      "loss": 0.2703,
      "reward": 0.6679590046405792,
      "reward_std": 0.13804228603839874,
      "reward_std_all": 0.13796672597527504,
      "rewards-max/forensics_sinkhorn_reward": 0.9882205426692963,
      "rewards-max/forensics_strict_edge_reward": 0.9565287232398987,
      "rewards-max/forensics_tiou_reward": 0.9969651401042938,
      "rewards/forensics_sinkhorn_reward": 0.5597814619541168,
      "rewards/forensics_strict_edge_reward": 0.3594793677330017,
      "rewards/forensics_tiou_reward": 0.41006679832935333,
      "step": 603
    },
    {
      "completion_length": 55.375,
      "epoch": 1.263874345549738,
      "grad_norm": 34.0301513671875,
      "kl": 5.609375,
      "learning_rate": 6.839622641509433e-07,
      "loss": 0.466,
      "reward": 0.6452687382698059,
      "reward_std": 0.14644968509674072,
      "reward_std_all": 0.1570095531642437,
      "rewards-max/forensics_sinkhorn_reward": 0.9884606897830963,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9979122281074524,
      "rewards/forensics_sinkhorn_reward": 0.539142370223999,
      "rewards/forensics_strict_edge_reward": 0.31134840846061707,
      "rewards/forensics_tiou_reward": 0.3721619248390198,
      "step": 604
    },
    {
      "completion_length": 44.9375,
      "epoch": 1.2659685863874346,
      "grad_norm": 18.17864418029785,
      "kl": 3.4765625,
      "learning_rate": 6.834381551362683e-07,
      "loss": 0.198,
      "reward": 0.668517529964447,
      "reward_std": 0.13036265596747398,
      "reward_std_all": 0.13517580553889275,
      "rewards-max/forensics_sinkhorn_reward": 0.9883562922477722,
      "rewards-max/forensics_strict_edge_reward": 0.9782643616199493,
      "rewards-max/forensics_tiou_reward": 0.9975000023841858,
      "rewards/forensics_sinkhorn_reward": 0.5575858354568481,
      "rewards/forensics_strict_edge_reward": 0.2958351820707321,
      "rewards/forensics_tiou_reward": 0.3617328405380249,
      "step": 605
    },
    {
      "completion_length": 32.328125,
      "epoch": 1.2680628272251309,
      "grad_norm": 24.693370819091797,
      "kl": 3.171875,
      "learning_rate": 6.829140461215934e-07,
      "loss": 0.2165,
      "reward": 0.5942095220088959,
      "reward_std": 0.1691654771566391,
      "reward_std_all": 0.18603092432022095,
      "rewards-max/forensics_sinkhorn_reward": 0.9861707389354706,
      "rewards-max/forensics_strict_edge_reward": 0.947419673204422,
      "rewards-max/forensics_tiou_reward": 0.9889506101608276,
      "rewards/forensics_sinkhorn_reward": 0.4555942714214325,
      "rewards/forensics_strict_edge_reward": 0.23730988800525665,
      "rewards/forensics_tiou_reward": 0.2894982844591141,
      "step": 606
    },
    {
      "completion_length": 40.859375,
      "epoch": 1.2701570680628271,
      "grad_norm": 19.612577438354492,
      "kl": 3.34375,
      "learning_rate": 6.823899371069182e-07,
      "loss": 0.4009,
      "reward": 0.6592204868793488,
      "reward_std": 0.14261560142040253,
      "reward_std_all": 0.14888111501932144,
      "rewards-max/forensics_sinkhorn_reward": 0.9873464405536652,
      "rewards-max/forensics_strict_edge_reward": 0.9567649364471436,
      "rewards-max/forensics_tiou_reward": 0.9935349822044373,
      "rewards/forensics_sinkhorn_reward": 0.5209690630435944,
      "rewards/forensics_strict_edge_reward": 0.31405480206012726,
      "rewards/forensics_tiou_reward": 0.3679800480604172,
      "step": 607
    },
    {
      "completion_length": 52.875,
      "epoch": 1.2722513089005236,
      "grad_norm": 32.09392547607422,
      "kl": 5.4453125,
      "learning_rate": 6.818658280922432e-07,
      "loss": 0.4737,
      "reward": 0.5958593785762787,
      "reward_std": 0.22150391340255737,
      "reward_std_all": 0.2133317068219185,
      "rewards-max/forensics_sinkhorn_reward": 0.9678730964660645,
      "rewards-max/forensics_strict_edge_reward": 0.9567649364471436,
      "rewards-max/forensics_tiou_reward": 0.955272376537323,
      "rewards/forensics_sinkhorn_reward": 0.48453694581985474,
      "rewards/forensics_strict_edge_reward": 0.33214806020259857,
      "rewards/forensics_tiou_reward": 0.3639014810323715,
      "step": 608
    },
    {
      "completion_length": 39.640625,
      "epoch": 1.27434554973822,
      "grad_norm": 1960.241455078125,
      "kl": 6.1875,
      "learning_rate": 6.813417190775681e-07,
      "loss": 0.3803,
      "reward": 0.6004193425178528,
      "reward_std": 0.1817716471850872,
      "reward_std_all": 0.1845589429140091,
      "rewards-max/forensics_sinkhorn_reward": 0.9884149432182312,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9977313876152039,
      "rewards/forensics_sinkhorn_reward": 0.5096827298402786,
      "rewards/forensics_strict_edge_reward": 0.2541753724217415,
      "rewards/forensics_tiou_reward": 0.32598528265953064,
      "step": 609
    },
    {
      "completion_length": 49.15625,
      "epoch": 1.2764397905759162,
      "grad_norm": 24.982450485229492,
      "kl": 5.453125,
      "learning_rate": 6.808176100628931e-07,
      "loss": 0.5048,
      "reward": 0.6015301942825317,
      "reward_std": 0.25589292496442795,
      "reward_std_all": 0.23598362505435944,
      "rewards-max/forensics_sinkhorn_reward": 0.9870907366275787,
      "rewards-max/forensics_strict_edge_reward": 0.9256840348243713,
      "rewards-max/forensics_tiou_reward": 0.9925148487091064,
      "rewards/forensics_sinkhorn_reward": 0.5121811926364899,
      "rewards/forensics_strict_edge_reward": 0.2914693057537079,
      "rewards/forensics_tiou_reward": 0.3865080624818802,
      "step": 610
    },
    {
      "completion_length": 51.875,
      "epoch": 1.2785340314136127,
      "grad_norm": 25.779550552368164,
      "kl": 4.78125,
      "learning_rate": 6.80293501048218e-07,
      "loss": 0.457,
      "reward": 0.6068288683891296,
      "reward_std": 0.18274608254432678,
      "reward_std_all": 0.18972307443618774,
      "rewards-max/forensics_sinkhorn_reward": 0.9690039753913879,
      "rewards-max/forensics_strict_edge_reward": 0.8893801271915436,
      "rewards-max/forensics_tiou_reward": 0.9533863067626953,
      "rewards/forensics_sinkhorn_reward": 0.46821314096450806,
      "rewards/forensics_strict_edge_reward": 0.31100381165742874,
      "rewards/forensics_tiou_reward": 0.3462091088294983,
      "step": 611
    },
    {
      "completion_length": 77.3125,
      "epoch": 1.280628272251309,
      "grad_norm": 28.122926712036133,
      "kl": 7.1875,
      "learning_rate": 6.79769392033543e-07,
      "loss": 0.515,
      "reward": 0.6069575846195221,
      "reward_std": 0.1778281033039093,
      "reward_std_all": 0.18330645561218262,
      "rewards-max/forensics_sinkhorn_reward": 0.988244354724884,
      "rewards-max/forensics_strict_edge_reward": 0.9782643616199493,
      "rewards-max/forensics_tiou_reward": 0.9970588088035583,
      "rewards/forensics_sinkhorn_reward": 0.47478654980659485,
      "rewards/forensics_strict_edge_reward": 0.27948257327079773,
      "rewards/forensics_tiou_reward": 0.3513248860836029,
      "step": 612
    },
    {
      "completion_length": 58.859375,
      "epoch": 1.2827225130890052,
      "grad_norm": 29.277925491333008,
      "kl": 6.203125,
      "learning_rate": 6.792452830188678e-07,
      "loss": 0.3638,
      "reward": 0.6255092918872833,
      "reward_std": 0.21674524247646332,
      "reward_std_all": 0.210982047021389,
      "rewards-max/forensics_sinkhorn_reward": 0.9841502606868744,
      "rewards-max/forensics_strict_edge_reward": 0.8445805907249451,
      "rewards-max/forensics_tiou_reward": 0.9810598492622375,
      "rewards/forensics_sinkhorn_reward": 0.5086531937122345,
      "rewards/forensics_strict_edge_reward": 0.27886079251766205,
      "rewards/forensics_tiou_reward": 0.3736202269792557,
      "step": 613
    },
    {
      "completion_length": 50.046875,
      "epoch": 1.2848167539267017,
      "grad_norm": 87.14837646484375,
      "kl": 6.171875,
      "learning_rate": 6.787211740041928e-07,
      "loss": 0.504,
      "reward": 0.5952584445476532,
      "reward_std": 0.24793099611997604,
      "reward_std_all": 0.23461759090423584,
      "rewards-max/forensics_sinkhorn_reward": 0.9868759214878082,
      "rewards-max/forensics_strict_edge_reward": 0.9375866651535034,
      "rewards-max/forensics_tiou_reward": 0.9916666746139526,
      "rewards/forensics_sinkhorn_reward": 0.4562491476535797,
      "rewards/forensics_strict_edge_reward": 0.29278434813022614,
      "rewards/forensics_tiou_reward": 0.3463045358657837,
      "step": 614
    },
    {
      "completion_length": 61.203125,
      "epoch": 1.286910994764398,
      "grad_norm": 24.471406936645508,
      "kl": 5.6796875,
      "learning_rate": 6.781970649895178e-07,
      "loss": 0.3291,
      "reward": 0.6481616199016571,
      "reward_std": 0.1220625787973404,
      "reward_std_all": 0.13734888285398483,
      "rewards-max/forensics_sinkhorn_reward": 0.9862675964832306,
      "rewards-max/forensics_strict_edge_reward": 0.947419673204422,
      "rewards-max/forensics_tiou_reward": 0.9937855303287506,
      "rewards/forensics_sinkhorn_reward": 0.519564151763916,
      "rewards/forensics_strict_edge_reward": 0.272492378950119,
      "rewards/forensics_tiou_reward": 0.3657718002796173,
      "step": 615
    },
    {
      "completion_length": 72.0,
      "epoch": 1.2890052356020942,
      "grad_norm": 28.3098201751709,
      "kl": 6.3359375,
      "learning_rate": 6.776729559748427e-07,
      "loss": 0.5064,
      "reward": 0.5861680209636688,
      "reward_std": 0.18761256337165833,
      "reward_std_all": 0.20033235847949982,
      "rewards-max/forensics_sinkhorn_reward": 0.825837105512619,
      "rewards-max/forensics_strict_edge_reward": 0.7619216740131378,
      "rewards-max/forensics_tiou_reward": 0.7609353363513947,
      "rewards/forensics_sinkhorn_reward": 0.4149342030286789,
      "rewards/forensics_strict_edge_reward": 0.19031250476837158,
      "rewards/forensics_tiou_reward": 0.2334190011024475,
      "step": 616
    },
    {
      "completion_length": 70.203125,
      "epoch": 1.2910994764397905,
      "grad_norm": 1294889.375,
      "kl": 6235.0,
      "learning_rate": 6.771488469601677e-07,
      "loss": 249.7973,
      "reward": 0.6369753181934357,
      "reward_std": 0.18531867116689682,
      "reward_std_all": 0.17614777386188507,
      "rewards-max/forensics_sinkhorn_reward": 0.9833683669567108,
      "rewards-max/forensics_strict_edge_reward": 0.8108904361724854,
      "rewards-max/forensics_tiou_reward": 0.9780564308166504,
      "rewards/forensics_sinkhorn_reward": 0.48759834468364716,
      "rewards/forensics_strict_edge_reward": 0.2619227468967438,
      "rewards/forensics_tiou_reward": 0.33246082067489624,
      "step": 617
    },
    {
      "completion_length": 61.25,
      "epoch": 1.2931937172774868,
      "grad_norm": 20.031932830810547,
      "kl": 5.0390625,
      "learning_rate": 6.766247379454927e-07,
      "loss": 0.4469,
      "reward": 0.6180500388145447,
      "reward_std": 0.16507313400506973,
      "reward_std_all": 0.1750004142522812,
      "rewards-max/forensics_sinkhorn_reward": 0.9879986345767975,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9960914850234985,
      "rewards/forensics_sinkhorn_reward": 0.5358611941337585,
      "rewards/forensics_strict_edge_reward": 0.30263976752758026,
      "rewards/forensics_tiou_reward": 0.3746345639228821,
      "step": 618
    },
    {
      "completion_length": 49.390625,
      "epoch": 1.2952879581151833,
      "grad_norm": 178.6129608154297,
      "kl": 7.0,
      "learning_rate": 6.761006289308176e-07,
      "loss": 0.3073,
      "reward": 0.6566954255104065,
      "reward_std": 0.1372717320919037,
      "reward_std_all": 0.14405760914087296,
      "rewards-max/forensics_sinkhorn_reward": 0.9880331456661224,
      "rewards-max/forensics_strict_edge_reward": 0.9782643616199493,
      "rewards-max/forensics_tiou_reward": 0.99622642993927,
      "rewards/forensics_sinkhorn_reward": 0.549290269613266,
      "rewards/forensics_strict_edge_reward": 0.2779313400387764,
      "rewards/forensics_tiou_reward": 0.34967009723186493,
      "step": 619
    },
    {
      "completion_length": 65.0,
      "epoch": 1.2973821989528795,
      "grad_norm": 322.612060546875,
      "kl": 8.046875,
      "learning_rate": 6.755765199161425e-07,
      "loss": 0.4892,
      "reward": 0.6048753261566162,
      "reward_std": 0.1765112206339836,
      "reward_std_all": 0.18383240699768066,
      "rewards-max/forensics_sinkhorn_reward": 0.944654792547226,
      "rewards-max/forensics_strict_edge_reward": 0.8062311410903931,
      "rewards-max/forensics_tiou_reward": 0.8937498927116394,
      "rewards/forensics_sinkhorn_reward": 0.4513290077447891,
      "rewards/forensics_strict_edge_reward": 0.22374046593904495,
      "rewards/forensics_tiou_reward": 0.26235469430685043,
      "step": 620
    },
    {
      "completion_length": 48.640625,
      "epoch": 1.2994764397905758,
      "grad_norm": 7299.38232421875,
      "kl": 24.15625,
      "learning_rate": 6.750524109014675e-07,
      "loss": 1.1241,
      "reward": 0.6674017906188965,
      "reward_std": 0.13863226771354675,
      "reward_std_all": 0.13834740221500397,
      "rewards-max/forensics_sinkhorn_reward": 0.9873640239238739,
      "rewards-max/forensics_strict_edge_reward": 0.9677534997463226,
      "rewards-max/forensics_tiou_reward": 0.993610292673111,
      "rewards/forensics_sinkhorn_reward": 0.49837951362133026,
      "rewards/forensics_strict_edge_reward": 0.27913224697113037,
      "rewards/forensics_tiou_reward": 0.3653292655944824,
      "step": 621
    },
    {
      "completion_length": 68.1875,
      "epoch": 1.3015706806282723,
      "grad_norm": 796.1355590820312,
      "kl": 9.125,
      "learning_rate": 6.745283018867925e-07,
      "loss": 0.6435,
      "reward": 0.6361902356147766,
      "reward_std": 0.20442545413970947,
      "reward_std_all": 0.1899527758359909,
      "rewards-max/forensics_sinkhorn_reward": 0.9870256185531616,
      "rewards-max/forensics_strict_edge_reward": 0.946017861366272,
      "rewards-max/forensics_tiou_reward": 0.9922706782817841,
      "rewards/forensics_sinkhorn_reward": 0.5291369259357452,
      "rewards/forensics_strict_edge_reward": 0.34036974608898163,
      "rewards/forensics_tiou_reward": 0.48782317340373993,
      "step": 622
    },
    {
      "completion_length": 39.59375,
      "epoch": 1.3036649214659686,
      "grad_norm": 1929.099609375,
      "kl": 39.59375,
      "learning_rate": 6.740041928721174e-07,
      "loss": 1.7828,
      "reward": 0.6358546018600464,
      "reward_std": 0.140821672976017,
      "reward_std_all": 0.1516585871577263,
      "rewards-max/forensics_sinkhorn_reward": 0.9882179200649261,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9969543516635895,
      "rewards/forensics_sinkhorn_reward": 0.5529657602310181,
      "rewards/forensics_strict_edge_reward": 0.34958599507808685,
      "rewards/forensics_tiou_reward": 0.3823027163743973,
      "step": 623
    },
    {
      "completion_length": 26.625,
      "epoch": 1.3057591623036648,
      "grad_norm": 32.9283332824707,
      "kl": 2.9921875,
      "learning_rate": 6.734800838574424e-07,
      "loss": 0.248,
      "reward": 0.6629895269870758,
      "reward_std": 0.1511375606060028,
      "reward_std_all": 0.14741478860378265,
      "rewards-max/forensics_sinkhorn_reward": 0.9383763670921326,
      "rewards-max/forensics_strict_edge_reward": 0.8681221902370453,
      "rewards-max/forensics_tiou_reward": 0.8863385319709778,
      "rewards/forensics_sinkhorn_reward": 0.5367567241191864,
      "rewards/forensics_strict_edge_reward": 0.26500403881073,
      "rewards/forensics_tiou_reward": 0.3566824793815613,
      "step": 624
    },
    {
      "completion_length": 24.390625,
      "epoch": 1.3078534031413613,
      "grad_norm": 14297.6552734375,
      "kl": 63.0,
      "learning_rate": 6.729559748427672e-07,
      "loss": 2.6942,
      "reward": 0.681751549243927,
      "reward_std": 0.11004515737295151,
      "reward_std_all": 0.11798293516039848,
      "rewards-max/forensics_sinkhorn_reward": 0.9873878359794617,
      "rewards-max/forensics_strict_edge_reward": 0.9567649364471436,
      "rewards-max/forensics_tiou_reward": 0.9937100410461426,
      "rewards/forensics_sinkhorn_reward": 0.5632267892360687,
      "rewards/forensics_strict_edge_reward": 0.27804385125637054,
      "rewards/forensics_tiou_reward": 0.35926297307014465,
      "step": 625
    },
    {
      "completion_length": 46.0625,
      "epoch": 1.3099476439790576,
      "grad_norm": 32879.12109375,
      "kl": 237.9375,
      "learning_rate": 6.724318658280922e-07,
      "loss": 9.5499,
      "reward": 0.5957923233509064,
      "reward_std": 0.11158519238233566,
      "reward_std_all": 0.1530364602804184,
      "rewards-max/forensics_sinkhorn_reward": 0.9870123565196991,
      "rewards-max/forensics_strict_edge_reward": 0.9169811904430389,
      "rewards-max/forensics_tiou_reward": 0.9922449588775635,
      "rewards/forensics_sinkhorn_reward": 0.47797921299934387,
      "rewards/forensics_strict_edge_reward": 0.1958448551595211,
      "rewards/forensics_tiou_reward": 0.287371963262558,
      "step": 626
    },
    {
      "completion_length": 53.109375,
      "epoch": 1.3120418848167539,
      "grad_norm": 365.20147705078125,
      "kl": 5.78125,
      "learning_rate": 6.719077568134171e-07,
      "loss": 0.2857,
      "reward": 0.6126241683959961,
      "reward_std": 0.14846621453762054,
      "reward_std_all": 0.16663478314876556,
      "rewards-max/forensics_sinkhorn_reward": 0.8896114528179169,
      "rewards-max/forensics_strict_edge_reward": 0.872120201587677,
      "rewards-max/forensics_tiou_reward": 0.8025760650634766,
      "rewards/forensics_sinkhorn_reward": 0.44155940413475037,
      "rewards/forensics_strict_edge_reward": 0.187803253531456,
      "rewards/forensics_tiou_reward": 0.2084944173693657,
      "step": 627
    },
    {
      "completion_length": 25.28125,
      "epoch": 1.3141361256544504,
      "grad_norm": 56.85527801513672,
      "kl": 2.6796875,
      "learning_rate": 6.713836477987421e-07,
      "loss": 0.0951,
      "reward": 0.678017258644104,
      "reward_std": 0.09298442117869854,
      "reward_std_all": 0.10852766782045364,
      "rewards-max/forensics_sinkhorn_reward": 0.9885187447071075,
      "rewards-max/forensics_strict_edge_reward": 0.9567649364471436,
      "rewards-max/forensics_tiou_reward": 0.9981406331062317,
      "rewards/forensics_sinkhorn_reward": 0.6361528933048248,
      "rewards/forensics_strict_edge_reward": 0.2781693860888481,
      "rewards/forensics_tiou_reward": 0.3889060765504837,
      "step": 628
    },
    {
      "completion_length": 36.65625,
      "epoch": 1.3162303664921466,
      "grad_norm": 1524.547607421875,
      "kl": 8.28125,
      "learning_rate": 6.708595387840671e-07,
      "loss": 0.5028,
      "reward": 0.6918996870517731,
      "reward_std": 0.12856069207191467,
      "reward_std_all": 0.12510783597826958,
      "rewards-max/forensics_sinkhorn_reward": 0.9882001876831055,
      "rewards-max/forensics_strict_edge_reward": 0.9782643616199493,
      "rewards-max/forensics_tiou_reward": 0.996889591217041,
      "rewards/forensics_sinkhorn_reward": 0.5930947065353394,
      "rewards/forensics_strict_edge_reward": 0.30269764363765717,
      "rewards/forensics_tiou_reward": 0.38727033138275146,
      "step": 629
    },
    {
      "completion_length": 36.734375,
      "epoch": 1.318324607329843,
      "grad_norm": 225.60580444335938,
      "kl": 4.328125,
      "learning_rate": 6.70335429769392e-07,
      "loss": 0.3569,
      "reward": 0.6845586597919464,
      "reward_std": 0.10335208475589752,
      "reward_std_all": 0.1108105257153511,
      "rewards-max/forensics_sinkhorn_reward": 0.9854055643081665,
      "rewards-max/forensics_strict_edge_reward": 0.8146511614322662,
      "rewards-max/forensics_tiou_reward": 0.985975444316864,
      "rewards/forensics_sinkhorn_reward": 0.6022590696811676,
      "rewards/forensics_strict_edge_reward": 0.29499848186969757,
      "rewards/forensics_tiou_reward": 0.4630134403705597,
      "step": 630
    },
    {
      "completion_length": 32.046875,
      "epoch": 1.3204188481675394,
      "grad_norm": 185.39730834960938,
      "kl": 2.859375,
      "learning_rate": 6.69811320754717e-07,
      "loss": 0.2984,
      "reward": 0.6899937689304352,
      "reward_std": 0.1030789241194725,
      "reward_std_all": 0.11118648573756218,
      "rewards-max/forensics_sinkhorn_reward": 0.9759286642074585,
      "rewards-max/forensics_strict_edge_reward": 0.9357379674911499,
      "rewards-max/forensics_tiou_reward": 0.9680681228637695,
      "rewards/forensics_sinkhorn_reward": 0.6199555099010468,
      "rewards/forensics_strict_edge_reward": 0.27147629857063293,
      "rewards/forensics_tiou_reward": 0.4155588895082474,
      "step": 631
    },
    {
      "completion_length": 30.0625,
      "epoch": 1.3225130890052357,
      "grad_norm": 289598.625,
      "kl": 45.640625,
      "learning_rate": 6.692872117400419e-07,
      "loss": 1.9088,
      "reward": 0.6625252366065979,
      "reward_std": 0.15420660376548767,
      "reward_std_all": 0.15435339510440826,
      "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": 0.6114405393600464,
      "rewards/forensics_strict_edge_reward": 0.3375023901462555,
      "rewards/forensics_tiou_reward": 0.4308517277240753,
      "step": 632
    },
    {
      "completion_length": 74.78125,
      "epoch": 1.324607329842932,
      "grad_norm": 1496224759808.0,
      "kl": 96469002.5,
      "learning_rate": 6.687631027253669e-07,
      "loss": 3845286.25,
      "reward": 0.6018253862857819,
      "reward_std": 0.18122870475053787,
      "reward_std_all": 0.19255982339382172,
      "rewards-max/forensics_sinkhorn_reward": 0.9592754244804382,
      "rewards-max/forensics_strict_edge_reward": 0.9003686904907227,
      "rewards-max/forensics_tiou_reward": 0.9339939653873444,
      "rewards/forensics_sinkhorn_reward": 0.4181234538555145,
      "rewards/forensics_strict_edge_reward": 0.27865585684776306,
      "rewards/forensics_tiou_reward": 0.30896155536174774,
      "step": 633
    },
    {
      "completion_length": 45.171875,
      "epoch": 1.3267015706806282,
      "grad_norm": 1.4827529554624512e+16,
      "kl": 1614907703297.6797,
      "learning_rate": 6.682389937106918e-07,
      "loss": 64412712960.0,
      "reward": 0.6344020962715149,
      "reward_std": 0.14356382936239243,
      "reward_std_all": 0.14806262776255608,
      "rewards-max/forensics_sinkhorn_reward": 0.9742290675640106,
      "rewards-max/forensics_strict_edge_reward": 0.9464850425720215,
      "rewards-max/forensics_tiou_reward": 0.96634241938591,
      "rewards/forensics_sinkhorn_reward": 0.4981706738471985,
      "rewards/forensics_strict_edge_reward": 0.3008653223514557,
      "rewards/forensics_tiou_reward": 0.32425549626350403,
      "step": 634
    },
    {
      "completion_length": 28.4375,
      "epoch": 1.3287958115183245,
      "grad_norm": 915448.8125,
      "kl": 203.0546875,
      "learning_rate": 6.677148846960168e-07,
      "loss": 8.1569,
      "reward": 0.6403464078903198,
      "reward_std": 0.1281743310391903,
      "reward_std_all": 0.14155516400933266,
      "rewards-max/forensics_sinkhorn_reward": 0.9314765930175781,
      "rewards-max/forensics_strict_edge_reward": 0.8168056011199951,
      "rewards-max/forensics_tiou_reward": 0.8734104633331299,
      "rewards/forensics_sinkhorn_reward": 0.5180447101593018,
      "rewards/forensics_strict_edge_reward": 0.21210186183452606,
      "rewards/forensics_tiou_reward": 0.28259633481502533,
      "step": 635
    },
    {
      "completion_length": 36.234375,
      "epoch": 1.330890052356021,
      "grad_norm": 628.457763671875,
      "kl": 3.2734375,
      "learning_rate": 6.671907756813417e-07,
      "loss": 0.1531,
      "reward": 0.7122132778167725,
      "reward_std": 0.10135439783334732,
      "reward_std_all": 0.0975937657058239,
      "rewards-max/forensics_sinkhorn_reward": 0.9883318543434143,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9974035620689392,
      "rewards/forensics_sinkhorn_reward": 0.612183690071106,
      "rewards/forensics_strict_edge_reward": 0.3753073513507843,
      "rewards/forensics_tiou_reward": 0.4166148751974106,
      "step": 636
    },
    {
      "completion_length": 25.890625,
      "epoch": 1.3329842931937173,
      "grad_norm": 1.188969354100736e+17,
      "kl": 19791209299969.4,
      "learning_rate": 6.666666666666666e-07,
      "loss": 788221919232.0,
      "reward": 0.69837886095047,
      "reward_std": 0.09614981524646282,
      "reward_std_all": 0.10436351597309113,
      "rewards-max/forensics_sinkhorn_reward": 0.9743883013725281,
      "rewards-max/forensics_strict_edge_reward": 0.9464850425720215,
      "rewards-max/forensics_tiou_reward": 0.9665964245796204,
      "rewards/forensics_sinkhorn_reward": 0.6267287135124207,
      "rewards/forensics_strict_edge_reward": 0.3298185467720032,
      "rewards/forensics_tiou_reward": 0.43108780682086945,
      "step": 637
    },
    {
      "completion_length": 55.453125,
      "epoch": 1.3350785340314135,
      "grad_norm": 141.6264190673828,
      "kl": 4.09375,
      "learning_rate": 6.661425576519916e-07,
      "loss": 0.3111,
      "reward": 0.6399264931678772,
      "reward_std": 0.172512736171484,
      "reward_std_all": 0.16651087254285812,
      "rewards-max/forensics_sinkhorn_reward": 0.988130509853363,
      "rewards-max/forensics_strict_edge_reward": 0.9782643616199493,
      "rewards-max/forensics_tiou_reward": 0.996610164642334,
      "rewards/forensics_sinkhorn_reward": 0.5755304992198944,
      "rewards/forensics_strict_edge_reward": 0.3279581591486931,
      "rewards/forensics_tiou_reward": 0.3846091479063034,
      "step": 638
    },
    {
      "completion_length": 41.40625,
      "epoch": 1.33717277486911,
      "grad_norm": 55.40884017944336,
      "kl": 4.4375,
      "learning_rate": 6.656184486373165e-07,
      "loss": 0.3292,
      "reward": 0.6602936387062073,
      "reward_std": 0.13901333138346672,
      "reward_std_all": 0.1409263275563717,
      "rewards-max/forensics_sinkhorn_reward": 0.9877056777477264,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.994935929775238,
      "rewards/forensics_sinkhorn_reward": 0.5313213467597961,
      "rewards/forensics_strict_edge_reward": 0.29568421840667725,
      "rewards/forensics_tiou_reward": 0.3667050153017044,
      "step": 639
    },
    {
      "completion_length": 38.59375,
      "epoch": 1.3392670157068063,
      "grad_norm": 77.54600524902344,
      "kl": 3.8671875,
      "learning_rate": 6.650943396226415e-07,
      "loss": 0.39,
      "reward": 0.7137519121170044,
      "reward_std": 0.10486667230725288,
      "reward_std_all": 0.09951753541827202,
      "rewards-max/forensics_sinkhorn_reward": 0.9873704314231873,
      "rewards-max/forensics_strict_edge_reward": 0.946017861366272,
      "rewards-max/forensics_tiou_reward": 0.9936253428459167,
      "rewards/forensics_sinkhorn_reward": 0.6329918801784515,
      "rewards/forensics_strict_edge_reward": 0.37035098671913147,
      "rewards/forensics_tiou_reward": 0.46290141344070435,
      "step": 640
    },
    {
      "completion_length": 38.875,
      "epoch": 1.3413612565445026,
      "grad_norm": 150.89630126953125,
      "kl": 4.34375,
      "learning_rate": 6.645702306079664e-07,
      "loss": 0.2347,
      "reward": 0.6251911520957947,
      "reward_std": 0.17436525970697403,
      "reward_std_all": 0.17860563099384308,
      "rewards-max/forensics_sinkhorn_reward": 0.987617552280426,
      "rewards-max/forensics_strict_edge_reward": 0.9256840348243713,
      "rewards-max/forensics_tiou_reward": 0.9945927560329437,
      "rewards/forensics_sinkhorn_reward": 0.5418424010276794,
      "rewards/forensics_strict_edge_reward": 0.20515966415405273,
      "rewards/forensics_tiou_reward": 0.318294495344162,
      "step": 641
    },
    {
      "completion_length": 49.84375,
      "epoch": 1.343455497382199,
      "grad_norm": 71.22740936279297,
      "kl": 3.7578125,
      "learning_rate": 6.640461215932913e-07,
      "loss": 0.233,
      "reward": 0.6002466976642609,
      "reward_std": 0.1179259791970253,
      "reward_std_all": 0.15431656688451767,
      "rewards-max/forensics_sinkhorn_reward": 0.9876988530158997,
      "rewards-max/forensics_strict_edge_reward": 0.927969753742218,
      "rewards-max/forensics_tiou_reward": 0.9949090778827667,
      "rewards/forensics_sinkhorn_reward": 0.5048309862613678,
      "rewards/forensics_strict_edge_reward": 0.21517469733953476,
      "rewards/forensics_tiou_reward": 0.2859988138079643,
      "step": 642
    },
    {
      "completion_length": 27.921875,
      "epoch": 1.3455497382198953,
      "grad_norm": 112.0392074584961,
      "kl": 2.5625,
      "learning_rate": 6.635220125786162e-07,
      "loss": 0.2071,
      "reward": 0.6721024513244629,
      "reward_std": 0.11911091953516006,
      "reward_std_all": 0.12792592495679855,
      "rewards-max/forensics_sinkhorn_reward": 0.9884005486965179,
      "rewards-max/forensics_strict_edge_reward": 0.9782643616199493,
      "rewards-max/forensics_tiou_reward": 0.9976744055747986,
      "rewards/forensics_sinkhorn_reward": 0.5998244881629944,
      "rewards/forensics_strict_edge_reward": 0.33261658251285553,
      "rewards/forensics_tiou_reward": 0.3850151300430298,
      "step": 643
    },
    {
      "completion_length": 18.234375,
      "epoch": 1.3476439790575916,
      "grad_norm": 1253.7418212890625,
      "kl": 2.33203125,
      "learning_rate": 6.629979035639413e-07,
      "loss": 0.1994,
      "reward": 0.6681438386440277,
      "reward_std": 0.07060827687382698,
      "reward_std_all": 0.10766427591443062,
      "rewards-max/forensics_sinkhorn_reward": 0.988091230392456,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9964577257633209,
      "rewards/forensics_sinkhorn_reward": 0.5382572114467621,
      "rewards/forensics_strict_edge_reward": 0.269245870411396,
      "rewards/forensics_tiou_reward": 0.2994280159473419,
      "step": 644
    },
    {
      "completion_length": 42.84375,
      "epoch": 1.349738219895288,
      "grad_norm": 202988.9375,
      "kl": 121.671875,
      "learning_rate": 6.624737945492663e-07,
      "loss": 5.0993,
      "reward": 0.6069838404655457,
      "reward_std": 0.13877535611391068,
      "reward_std_all": 0.1602507382631302,
      "rewards-max/forensics_sinkhorn_reward": 0.9764338135719299,
      "rewards-max/forensics_strict_edge_reward": 0.941106379032135,
      "rewards-max/forensics_tiou_reward": 0.9638443887233734,
      "rewards/forensics_sinkhorn_reward": 0.5271150171756744,
      "rewards/forensics_strict_edge_reward": 0.30451202392578125,
      "rewards/forensics_tiou_reward": 0.3661269247531891,
      "step": 645
    },
    {
      "completion_length": 60.75,
      "epoch": 1.3518324607329844,
      "grad_norm": 30.23000144958496,
      "kl": 5.03125,
      "learning_rate": 6.619496855345912e-07,
      "loss": 0.3741,
      "reward": 0.6126999258995056,
      "reward_std": 0.1643676981329918,
      "reward_std_all": 0.1803848072886467,
      "rewards-max/forensics_sinkhorn_reward": 0.9806618988513947,
      "rewards-max/forensics_strict_edge_reward": 0.9185642302036285,
      "rewards-max/forensics_tiou_reward": 0.9819990992546082,
      "rewards/forensics_sinkhorn_reward": 0.4832116812467575,
      "rewards/forensics_strict_edge_reward": 0.2738970071077347,
      "rewards/forensics_tiou_reward": 0.33530427515506744,
      "step": 646
    },
    {
      "completion_length": 38.390625,
      "epoch": 1.3539267015706806,
      "grad_norm": 29.186208724975586,
      "kl": 2.59375,
      "learning_rate": 6.614255765199162e-07,
      "loss": 0.2227,
      "reward": 0.6374822556972504,
      "reward_std": 0.14003697037696838,
      "reward_std_all": 0.1572999209165573,
      "rewards-max/forensics_sinkhorn_reward": 0.9500982761383057,
      "rewards-max/forensics_strict_edge_reward": 0.8578422963619232,
      "rewards-max/forensics_tiou_reward": 0.9134242832660675,
      "rewards/forensics_sinkhorn_reward": 0.5295826196670532,
      "rewards/forensics_strict_edge_reward": 0.29393815994262695,
      "rewards/forensics_tiou_reward": 0.34743861854076385,
      "step": 647
    },
    {
      "completion_length": 26.75,
      "epoch": 1.356020942408377,
      "grad_norm": 24.514041900634766,
      "kl": 2.3984375,
      "learning_rate": 6.60901467505241e-07,
      "loss": 0.1976,
      "reward": 0.674822211265564,
      "reward_std": 0.11295288056135178,
      "reward_std_all": 0.12149397656321526,
      "rewards-max/forensics_sinkhorn_reward": 0.9871002733707428,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9925510883331299,
      "rewards/forensics_sinkhorn_reward": 0.5636391341686249,
      "rewards/forensics_strict_edge_reward": 0.24785755574703217,
      "rewards/forensics_tiou_reward": 0.30803607404232025,
      "step": 648
    },
    {
      "completion_length": 39.75,
      "epoch": 1.3581151832460732,
      "grad_norm": 27.190645217895508,
      "kl": 3.8828125,
      "learning_rate": 6.60377358490566e-07,
      "loss": 0.28,
      "reward": 0.63661789894104,
      "reward_std": 0.20228972285985947,
      "reward_std_all": 0.19116811454296112,
      "rewards-max/forensics_sinkhorn_reward": 0.939428061246872,
      "rewards-max/forensics_strict_edge_reward": 0.9289211630821228,
      "rewards-max/forensics_tiou_reward": 0.8969413936138153,
      "rewards/forensics_sinkhorn_reward": 0.490271657705307,
      "rewards/forensics_strict_edge_reward": 0.22226373851299286,
      "rewards/forensics_tiou_reward": 0.2764655500650406,
      "step": 649
    },
    {
      "completion_length": 37.53125,
      "epoch": 1.3602094240837697,
      "grad_norm": 41.47894287109375,
      "kl": 4.015625,
      "learning_rate": 6.598532494758909e-07,
      "loss": 0.2425,
      "reward": 0.6976262331008911,
      "reward_std": 0.11383871361613274,
      "reward_std_all": 0.115411426872015,
      "rewards-max/forensics_sinkhorn_reward": 0.9850359857082367,
      "rewards-max/forensics_strict_edge_reward": 0.8578422963619232,
      "rewards-max/forensics_tiou_reward": 0.9844359755516052,
      "rewards/forensics_sinkhorn_reward": 0.5338510274887085,
      "rewards/forensics_strict_edge_reward": 0.30279386043548584,
      "rewards/forensics_tiou_reward": 0.37318387627601624,
      "step": 650
    },
    {
      "completion_length": 34.046875,
      "epoch": 1.362303664921466,
      "grad_norm": 23823.083984375,
      "kl": 121.3203125,
      "learning_rate": 6.593291404612159e-07,
      "loss": 4.8841,
      "reward": 0.6651599407196045,
      "reward_std": 0.16154295951128006,
      "reward_std_all": 0.15219425410032272,
      "rewards-max/forensics_sinkhorn_reward": 0.9756087064743042,
      "rewards-max/forensics_strict_edge_reward": 0.9574736058712006,
      "rewards-max/forensics_tiou_reward": 0.9697368443012238,
      "rewards/forensics_sinkhorn_reward": 0.5314939618110657,
      "rewards/forensics_strict_edge_reward": 0.32345664501190186,
      "rewards/forensics_tiou_reward": 0.36529630422592163,
      "step": 651
    },
    {
      "completion_length": 42.34375,
      "epoch": 1.3643979057591622,
      "grad_norm": 50.06023406982422,
      "kl": 3.2265625,
      "learning_rate": 6.588050314465408e-07,
      "loss": 0.3367,
      "reward": 0.6322231590747833,
      "reward_std": 0.19091639667749405,
      "reward_std_all": 0.18597443401813507,
      "rewards-max/forensics_sinkhorn_reward": 0.9867116212844849,
      "rewards-max/forensics_strict_edge_reward": 0.946017861366272,
      "rewards-max/forensics_tiou_reward": 0.991023063659668,
      "rewards/forensics_sinkhorn_reward": 0.5262667536735535,
      "rewards/forensics_strict_edge_reward": 0.30672378838062286,
      "rewards/forensics_tiou_reward": 0.36371442675590515,
      "step": 652
    },
    {
      "completion_length": 61.125,
      "epoch": 1.3664921465968587,
      "grad_norm": 23.965721130371094,
      "kl": 4.53125,
      "learning_rate": 6.582809224318658e-07,
      "loss": 0.3522,
      "reward": 0.6305320262908936,
      "reward_std": 0.14640244841575623,
      "reward_std_all": 0.15540951490402222,
      "rewards-max/forensics_sinkhorn_reward": 0.9880248010158539,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9961974322795868,
      "rewards/forensics_sinkhorn_reward": 0.5038793683052063,
      "rewards/forensics_strict_edge_reward": 0.21257027983665466,
      "rewards/forensics_tiou_reward": 0.29680803418159485,
      "step": 653
    },
    {
      "completion_length": 48.4375,
      "epoch": 1.368586387434555,
      "grad_norm": 55.00540542602539,
      "kl": 5.0,
      "learning_rate": 6.577568134171907e-07,
      "loss": 0.2607,
      "reward": 0.6782575249671936,
      "reward_std": 0.14057716727256775,
      "reward_std_all": 0.13379914313554764,
      "rewards-max/forensics_sinkhorn_reward": 0.9476121366024017,
      "rewards-max/forensics_strict_edge_reward": 0.8659839034080505,
      "rewards-max/forensics_tiou_reward": 0.9109496176242828,
      "rewards/forensics_sinkhorn_reward": 0.5045507848262787,
      "rewards/forensics_strict_edge_reward": 0.2546602487564087,
      "rewards/forensics_tiou_reward": 0.3205821216106415,
      "step": 654
    },
    {
      "completion_length": 35.859375,
      "epoch": 1.3706806282722512,
      "grad_norm": 33.02715301513672,
      "kl": 3.3203125,
      "learning_rate": 6.572327044025156e-07,
      "loss": 0.373,
      "reward": 0.6111722886562347,
      "reward_std": 0.20225799828767776,
      "reward_std_all": 0.19528090953826904,
      "rewards-max/forensics_sinkhorn_reward": 0.9838645160198212,
      "rewards-max/forensics_strict_edge_reward": 0.8719756007194519,
      "rewards-max/forensics_tiou_reward": 0.9799894094467163,
      "rewards/forensics_sinkhorn_reward": 0.47460509836673737,
      "rewards/forensics_strict_edge_reward": 0.2004498988389969,
      "rewards/forensics_tiou_reward": 0.24289237707853317,
      "step": 655
    },
    {
      "completion_length": 27.5625,
      "epoch": 1.3727748691099477,
      "grad_norm": 33.23225784301758,
      "kl": 2.10546875,
      "learning_rate": 6.567085953878407e-07,
      "loss": 0.3139,
      "reward": 0.6851203441619873,
      "reward_std": 0.1339336335659027,
      "reward_std_all": 0.1295628920197487,
      "rewards-max/forensics_sinkhorn_reward": 0.9880331456661224,
      "rewards-max/forensics_strict_edge_reward": 0.9782643616199493,
      "rewards-max/forensics_tiou_reward": 0.99622642993927,
      "rewards/forensics_sinkhorn_reward": 0.6385070085525513,
      "rewards/forensics_strict_edge_reward": 0.35763292014598846,
      "rewards/forensics_tiou_reward": 0.4979260414838791,
      "step": 656
    },
    {
      "completion_length": 46.578125,
      "epoch": 1.374869109947644,
      "grad_norm": 46.4047737121582,
      "kl": 4.8125,
      "learning_rate": 6.561844863731656e-07,
      "loss": 0.3898,
      "reward": 0.6588613986968994,
      "reward_std": 0.10596012324094772,
      "reward_std_all": 0.12126617878675461,
      "rewards-max/forensics_sinkhorn_reward": 0.9825639128684998,
      "rewards-max/forensics_strict_edge_reward": 0.9364311099052429,
      "rewards-max/forensics_tiou_reward": 0.9847196340560913,
      "rewards/forensics_sinkhorn_reward": 0.5113358497619629,
      "rewards/forensics_strict_edge_reward": 0.271371953189373,
      "rewards/forensics_tiou_reward": 0.302914559841156,
      "step": 657
    },
    {
      "completion_length": 67.4375,
      "epoch": 1.3769633507853403,
      "grad_norm": 35.215057373046875,
      "kl": 5.1875,
      "learning_rate": 6.556603773584906e-07,
      "loss": 0.4845,
      "reward": 0.6678714156150818,
      "reward_std": 0.177727110683918,
      "reward_std_all": 0.16315676271915436,
      "rewards-max/forensics_sinkhorn_reward": 0.9878128170967102,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9953580498695374,
      "rewards/forensics_sinkhorn_reward": 0.572762668132782,
      "rewards/forensics_strict_edge_reward": 0.3258103281259537,
      "rewards/forensics_tiou_reward": 0.4186212122440338,
      "step": 658
    },
    {
      "completion_length": 40.34375,
      "epoch": 1.3790575916230368,
      "grad_norm": 49.660552978515625,
      "kl": 4.328125,
      "learning_rate": 6.551362683438155e-07,
      "loss": 0.3361,
      "reward": 0.6443052291870117,
      "reward_std": 0.17108815908432007,
      "reward_std_all": 0.17086955159902573,
      "rewards-max/forensics_sinkhorn_reward": 0.9881126582622528,
      "rewards-max/forensics_strict_edge_reward": 0.9574736058712006,
      "rewards-max/forensics_tiou_reward": 0.9965398013591766,
      "rewards/forensics_sinkhorn_reward": 0.5968375205993652,
      "rewards/forensics_strict_edge_reward": 0.3636243939399719,
      "rewards/forensics_tiou_reward": 0.4671599566936493,
      "step": 659
    },
    {
      "completion_length": 59.03125,
      "epoch": 1.381151832460733,
      "grad_norm": 152.08331298828125,
      "kl": 8.578125,
      "learning_rate": 6.546121593291404e-07,
      "loss": 0.4614,
      "reward": 0.6375198066234589,
      "reward_std": 0.19119223207235336,
      "reward_std_all": 0.1816055253148079,
      "rewards-max/forensics_sinkhorn_reward": 0.9880926012992859,
      "rewards-max/forensics_strict_edge_reward": 0.946017861366272,
      "rewards-max/forensics_tiou_reward": 0.9964636862277985,
      "rewards/forensics_sinkhorn_reward": 0.5031506717205048,
      "rewards/forensics_strict_edge_reward": 0.27357129007577896,
      "rewards/forensics_tiou_reward": 0.3429281413555145,
      "step": 660
    },
    {
      "completion_length": 51.015625,
      "epoch": 1.3832460732984293,
      "grad_norm": 61.936920166015625,
      "kl": 5.546875,
      "learning_rate": 6.540880503144653e-07,
      "loss": 0.2909,
      "reward": 0.6432760655879974,
      "reward_std": 0.10452768951654434,
      "reward_std_all": 0.13697730749845505,
      "rewards-max/forensics_sinkhorn_reward": 0.9801903367042542,
      "rewards-max/forensics_strict_edge_reward": 0.9357379674911499,
      "rewards-max/forensics_tiou_reward": 0.9765743613243103,
      "rewards/forensics_sinkhorn_reward": 0.47604309022426605,
      "rewards/forensics_strict_edge_reward": 0.22267670184373856,
      "rewards/forensics_tiou_reward": 0.2874780148267746,
      "step": 661
    },
    {
      "completion_length": 54.40625,
      "epoch": 1.3853403141361256,
      "grad_norm": 31.358545303344727,
      "kl": 7.265625,
      "learning_rate": 6.535639412997903e-07,
      "loss": 0.4455,
      "reward": 0.7150108218193054,
      "reward_std": 0.10092630982398987,
      "reward_std_all": 0.09611629694700241,
      "rewards-max/forensics_sinkhorn_reward": 0.9876894056797028,
      "rewards-max/forensics_strict_edge_reward": 0.9677534997463226,
      "rewards-max/forensics_tiou_reward": 0.9948717951774597,
      "rewards/forensics_sinkhorn_reward": 0.5639482140541077,
      "rewards/forensics_strict_edge_reward": 0.3271002173423767,
      "rewards/forensics_tiou_reward": 0.3959490805864334,
      "step": 662
    },
    {
      "completion_length": 56.859375,
      "epoch": 1.387434554973822,
      "grad_norm": 32.085784912109375,
      "kl": 6.515625,
      "learning_rate": 6.530398322851153e-07,
      "loss": 0.401,
      "reward": 0.628879189491272,
      "reward_std": 0.1273992732167244,
      "reward_std_all": 0.1511199176311493,
      "rewards-max/forensics_sinkhorn_reward": 0.9851574897766113,
      "rewards-max/forensics_strict_edge_reward": 0.946017861366272,
      "rewards-max/forensics_tiou_reward": 0.9886869490146637,
      "rewards/forensics_sinkhorn_reward": 0.46057215332984924,
      "rewards/forensics_strict_edge_reward": 0.26092498004436493,
      "rewards/forensics_tiou_reward": 0.3146958649158478,
      "step": 663
    },
    {
      "completion_length": 56.3125,
      "epoch": 1.3895287958115183,
      "grad_norm": 42.100521087646484,
      "kl": 6.890625,
      "learning_rate": 6.525157232704402e-07,
      "loss": 0.5126,
      "reward": 0.6742377877235413,
      "reward_std": 0.14881528168916702,
      "reward_std_all": 0.14747430384159088,
      "rewards-max/forensics_sinkhorn_reward": 0.9099392592906952,
      "rewards-max/forensics_strict_edge_reward": 0.884223610162735,
      "rewards-max/forensics_tiou_reward": 0.8329884707927704,
      "rewards/forensics_sinkhorn_reward": 0.5061140060424805,
      "rewards/forensics_strict_edge_reward": 0.2720901668071747,
      "rewards/forensics_tiou_reward": 0.3126923441886902,
      "step": 664
    },
    {
      "completion_length": 72.5,
      "epoch": 1.3916230366492146,
      "grad_norm": 2665.868408203125,
      "kl": 10.0625,
      "learning_rate": 6.519916142557652e-07,
      "loss": 0.5848,
      "reward": 0.6518782377243042,
      "reward_std": 0.1304151602089405,
      "reward_std_all": 0.14067376032471657,
      "rewards-max/forensics_sinkhorn_reward": 0.988856703042984,
      "rewards-max/forensics_strict_edge_reward": 0.9890114367008209,
      "rewards-max/forensics_tiou_reward": 0.9994727075099945,
      "rewards/forensics_sinkhorn_reward": 0.5078230202198029,
      "rewards/forensics_strict_edge_reward": 0.3251779302954674,
      "rewards/forensics_tiou_reward": 0.35491321980953217,
      "step": 665
    },
    {
      "completion_length": 58.84375,
      "epoch": 1.3937172774869109,
      "grad_norm": 180.99932861328125,
      "kl": 4.8125,
      "learning_rate": 6.5146750524109e-07,
      "loss": 0.454,
      "reward": 0.6241802871227264,
      "reward_std": 0.14613399282097816,
      "reward_std_all": 0.16118068993091583,
      "rewards-max/forensics_sinkhorn_reward": 0.9851446747779846,
      "rewards-max/forensics_strict_edge_reward": 0.9677534997463226,
      "rewards-max/forensics_tiou_reward": 0.9911351799964905,
      "rewards/forensics_sinkhorn_reward": 0.4909314811229706,
      "rewards/forensics_strict_edge_reward": 0.26710090041160583,
      "rewards/forensics_tiou_reward": 0.33512571454048157,
      "step": 666
    },
    {
      "completion_length": 56.90625,
      "epoch": 1.3958115183246074,
      "grad_norm": 939206.875,
      "kl": 830.59375,
      "learning_rate": 6.50943396226415e-07,
      "loss": 33.3425,
      "reward": 0.6397081315517426,
      "reward_std": 0.1590806543827057,
      "reward_std_all": 0.16412337869405746,
      "rewards-max/forensics_sinkhorn_reward": 0.9860130548477173,
      "rewards-max/forensics_strict_edge_reward": 0.9357379674911499,
      "rewards-max/forensics_tiou_reward": 0.9882653057575226,
      "rewards/forensics_sinkhorn_reward": 0.5840932428836823,
      "rewards/forensics_strict_edge_reward": 0.3450586348772049,
      "rewards/forensics_tiou_reward": 0.4168233424425125,
      "step": 667
    },
    {
      "completion_length": 95.546875,
      "epoch": 1.3979057591623036,
      "grad_norm": 59.609535217285156,
      "kl": 6.453125,
      "learning_rate": 6.5041928721174e-07,
      "loss": 0.5248,
      "reward": 0.6638678014278412,
      "reward_std": 0.13799858465790749,
      "reward_std_all": 0.140681691467762,
      "rewards-max/forensics_sinkhorn_reward": 0.9583784341812134,
      "rewards-max/forensics_strict_edge_reward": 0.9093653857707977,
      "rewards-max/forensics_tiou_reward": 0.9308804869651794,
      "rewards/forensics_sinkhorn_reward": 0.4909600466489792,
      "rewards/forensics_strict_edge_reward": 0.2136596366763115,
      "rewards/forensics_tiou_reward": 0.31919270753860474,
      "step": 668
    },
    {
      "completion_length": 53.328125,
      "epoch": 1.4,
      "grad_norm": 82.13439178466797,
      "kl": 5.5,
      "learning_rate": 6.49895178197065e-07,
      "loss": 0.2929,
      "reward": 0.6249657571315765,
      "reward_std": 0.16068852692842484,
      "reward_std_all": 0.17100118100643158,
      "rewards-max/forensics_sinkhorn_reward": 0.9658338129520416,
      "rewards-max/forensics_strict_edge_reward": 0.9375866651535034,
      "rewards-max/forensics_tiou_reward": 0.9507575631141663,
      "rewards/forensics_sinkhorn_reward": 0.5229489803314209,
      "rewards/forensics_strict_edge_reward": 0.262879878282547,
      "rewards/forensics_tiou_reward": 0.333856537938118,
      "step": 669
    },
    {
      "completion_length": 80.390625,
      "epoch": 1.4020942408376964,
      "grad_norm": 55.529659271240234,
      "kl": 5.7421875,
      "learning_rate": 6.4937106918239e-07,
      "loss": 0.5329,
      "reward": 0.6079926490783691,
      "reward_std": 0.21407188475131989,
      "reward_std_all": 0.20718201994895935,
      "rewards-max/forensics_sinkhorn_reward": 0.9463672637939453,
      "rewards-max/forensics_strict_edge_reward": 0.7284609377384186,
      "rewards-max/forensics_tiou_reward": 0.9083564877510071,
      "rewards/forensics_sinkhorn_reward": 0.45223748683929443,
      "rewards/forensics_strict_edge_reward": 0.26365751028060913,
      "rewards/forensics_tiou_reward": 0.313664548099041,
      "step": 670
    },
    {
      "completion_length": 79.40625,
      "epoch": 1.4041884816753927,
      "grad_norm": 27.934978485107422,
      "kl": 5.875,
      "learning_rate": 6.488469601677149e-07,
      "loss": 0.4476,
      "reward": 0.6746438145637512,
      "reward_std": 0.17298530787229538,
      "reward_std_all": 0.15828058123588562,
      "rewards-max/forensics_sinkhorn_reward": 0.9889025986194611,
      "rewards-max/forensics_strict_edge_reward": 0.9890114367008209,
      "rewards-max/forensics_tiou_reward": 0.999653548002243,
      "rewards/forensics_sinkhorn_reward": 0.577530175447464,
      "rewards/forensics_strict_edge_reward": 0.3881634920835495,
      "rewards/forensics_tiou_reward": 0.46554234623908997,
      "step": 671
    },
    {
      "completion_length": 59.71875,
      "epoch": 1.406282722513089,
      "grad_norm": 89.1027603149414,
      "kl": 4.71875,
      "learning_rate": 6.483228511530398e-07,
      "loss": 0.2041,
      "reward": 0.6962752044200897,
      "reward_std": 0.1592731699347496,
      "reward_std_all": 0.14661070704460144,
      "rewards-max/forensics_sinkhorn_reward": 0.9885163307189941,
      "rewards-max/forensics_strict_edge_reward": 0.9890114367008209,
      "rewards-max/forensics_tiou_reward": 0.9981308281421661,
      "rewards/forensics_sinkhorn_reward": 0.5461770296096802,
      "rewards/forensics_strict_edge_reward": 0.3549957275390625,
      "rewards/forensics_tiou_reward": 0.4071410298347473,
      "step": 672
    },
    {
      "completion_length": 73.859375,
      "epoch": 1.4083769633507854,
      "grad_norm": 130.81558227539062,
      "kl": 5.09375,
      "learning_rate": 6.477987421383647e-07,
      "loss": 0.3326,
      "reward": 0.6469121277332306,
      "reward_std": 0.1560375988483429,
      "reward_std_all": 0.1618332415819168,
      "rewards-max/forensics_sinkhorn_reward": 0.9881324172019958,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9966206848621368,
      "rewards/forensics_sinkhorn_reward": 0.581626832485199,
      "rewards/forensics_strict_edge_reward": 0.3804406374692917,
      "rewards/forensics_tiou_reward": 0.48146775364875793,
      "step": 673
    },
    {
      "completion_length": 57.375,
      "epoch": 1.4104712041884817,
      "grad_norm": 27.910221099853516,
      "kl": 3.375,
      "learning_rate": 6.472746331236897e-07,
      "loss": 0.2539,
      "reward": 0.6557599306106567,
      "reward_std": 0.15054070204496384,
      "reward_std_all": 0.153249591588974,
      "rewards-max/forensics_sinkhorn_reward": 0.9842386245727539,
      "rewards-max/forensics_strict_edge_reward": 0.9565287232398987,
      "rewards-max/forensics_tiou_reward": 0.9879409372806549,
      "rewards/forensics_sinkhorn_reward": 0.5760226249694824,
      "rewards/forensics_strict_edge_reward": 0.23718051612377167,
      "rewards/forensics_tiou_reward": 0.43172064423561096,
      "step": 674
    },
    {
      "completion_length": 46.421875,
      "epoch": 1.412565445026178,
      "grad_norm": 32.936187744140625,
      "kl": 3.453125,
      "learning_rate": 6.467505241090146e-07,
      "loss": 0.2984,
      "reward": 0.6509928107261658,
      "reward_std": 0.10444313287734985,
      "reward_std_all": 0.12821882218122482,
      "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": 0.5133018791675568,
      "rewards/forensics_strict_edge_reward": 0.3323643058538437,
      "rewards/forensics_tiou_reward": 0.35150784254074097,
      "step": 675
    },
    {
      "completion_length": 55.765625,
      "epoch": 1.4146596858638745,
      "grad_norm": 35.48429489135742,
      "kl": 3.6953125,
      "learning_rate": 6.462264150943396e-07,
      "loss": 0.2607,
      "reward": 0.667899489402771,
      "reward_std": 0.15234696120023727,
      "reward_std_all": 0.14809562265872955,
      "rewards-max/forensics_sinkhorn_reward": 0.9880844354629517,
      "rewards-max/forensics_strict_edge_reward": 0.9782643616199493,
      "rewards-max/forensics_tiou_reward": 0.9964285790920258,
      "rewards/forensics_sinkhorn_reward": 0.5644010305404663,
      "rewards/forensics_strict_edge_reward": 0.2567705661058426,
      "rewards/forensics_tiou_reward": 0.360524982213974,
      "step": 676
    },
    {
      "completion_length": 86.953125,
      "epoch": 1.4167539267015707,
      "grad_norm": 49.96458435058594,
      "kl": 6.03125,
      "learning_rate": 6.457023060796646e-07,
      "loss": 0.3967,
      "reward": 0.5855772793292999,
      "reward_std": 0.22020895779132843,
      "reward_std_all": 0.2197769582271576,
      "rewards-max/forensics_sinkhorn_reward": 0.9443534314632416,
      "rewards-max/forensics_strict_edge_reward": 0.8681221902370453,
      "rewards-max/forensics_tiou_reward": 0.9037980735301971,
      "rewards/forensics_sinkhorn_reward": 0.43501847982406616,
      "rewards/forensics_strict_edge_reward": 0.16630961000919342,
      "rewards/forensics_tiou_reward": 0.23299355804920197,
      "step": 677
    },
    {
      "completion_length": 28.4375,
      "epoch": 1.418848167539267,
      "grad_norm": 38.03727340698242,
      "kl": 2.328125,
      "learning_rate": 6.451781970649894e-07,
      "loss": 0.1532,
      "reward": 0.6418489217758179,
      "reward_std": 0.1281881369650364,
      "reward_std_all": 0.14174798876047134,
      "rewards-max/forensics_sinkhorn_reward": 0.988698422908783,
      "rewards-max/forensics_strict_edge_reward": 0.9782643616199493,
      "rewards-max/forensics_tiou_reward": 0.9988492429256439,
      "rewards/forensics_sinkhorn_reward": 0.584352046251297,
      "rewards/forensics_strict_edge_reward": 0.341693639755249,
      "rewards/forensics_tiou_reward": 0.4176921546459198,
      "step": 678
    },
    {
      "completion_length": 72.984375,
      "epoch": 1.4209424083769633,
      "grad_norm": 23.51864242553711,
      "kl": 5.53125,
      "learning_rate": 6.446540880503144e-07,
      "loss": 0.4145,
      "reward": 0.6063327789306641,
      "reward_std": 0.18113641440868378,
      "reward_std_all": 0.18155406415462494,
      "rewards-max/forensics_sinkhorn_reward": 0.8888011872768402,
      "rewards-max/forensics_strict_edge_reward": 0.8078543543815613,
      "rewards-max/forensics_tiou_reward": 0.8152577579021454,
      "rewards/forensics_sinkhorn_reward": 0.4618302434682846,
      "rewards/forensics_strict_edge_reward": 0.22223202139139175,
      "rewards/forensics_tiou_reward": 0.2898636758327484,
      "step": 679
    },
    {
      "completion_length": 40.75,
      "epoch": 1.4230366492146596,
      "grad_norm": 41.08524703979492,
      "kl": 3.90625,
      "learning_rate": 6.441299790356393e-07,
      "loss": 0.2748,
      "reward": 0.5915248095989227,
      "reward_std": 0.21703088283538818,
      "reward_std_all": 0.21124311536550522,
      "rewards-max/forensics_sinkhorn_reward": 0.9862475991249084,
      "rewards-max/forensics_strict_edge_reward": 0.9265981018543243,
      "rewards-max/forensics_tiou_reward": 0.9891912639141083,
      "rewards/forensics_sinkhorn_reward": 0.4821361154317856,
      "rewards/forensics_strict_edge_reward": 0.24010051041841507,
      "rewards/forensics_tiou_reward": 0.30557380616664886,
      "step": 680
    },
    {
      "completion_length": 55.03125,
      "epoch": 1.425130890052356,
      "grad_norm": 51.21163558959961,
      "kl": 5.1953125,
      "learning_rate": 6.436058700209644e-07,
      "loss": 0.3956,
      "reward": 0.61529541015625,
      "reward_std": 0.16110998392105103,
      "reward_std_all": 0.18249817937612534,
      "rewards-max/forensics_sinkhorn_reward": 0.9873513281345367,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9935512542724609,
      "rewards/forensics_sinkhorn_reward": 0.485066294670105,
      "rewards/forensics_strict_edge_reward": 0.26245512068271637,
      "rewards/forensics_tiou_reward": 0.3549579828977585,
      "step": 681
    },
    {
      "completion_length": 37.40625,
      "epoch": 1.4272251308900523,
      "grad_norm": 29.88862419128418,
      "kl": 3.71875,
      "learning_rate": 6.430817610062893e-07,
      "loss": 0.2765,
      "reward": 0.6897804439067841,
      "reward_std": 0.13059116899967194,
      "reward_std_all": 0.12939544022083282,
      "rewards-max/forensics_sinkhorn_reward": 0.9880844354629517,
      "rewards-max/forensics_strict_edge_reward": 0.9782643616199493,
      "rewards-max/forensics_tiou_reward": 0.9964285790920258,
      "rewards/forensics_sinkhorn_reward": 0.5773542523384094,
      "rewards/forensics_strict_edge_reward": 0.36219552159309387,
      "rewards/forensics_tiou_reward": 0.40508534014225006,
      "step": 682
    },
    {
      "completion_length": 55.828125,
      "epoch": 1.4293193717277486,
      "grad_norm": 66.6314468383789,
      "kl": 4.84375,
      "learning_rate": 6.425576519916143e-07,
      "loss": 0.1921,
      "reward": 0.6273126900196075,
      "reward_std": 0.16777485609054565,
      "reward_std_all": 0.1680881604552269,
      "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": 0.4530039429664612,
      "rewards/forensics_strict_edge_reward": 0.22651034593582153,
      "rewards/forensics_tiou_reward": 0.2498575672507286,
      "step": 683
    },
    {
      "completion_length": 59.171875,
      "epoch": 1.431413612565445,
      "grad_norm": 46.9109992980957,
      "kl": 5.125,
      "learning_rate": 6.420335429769391e-07,
      "loss": 0.4035,
      "reward": 0.612569272518158,
      "reward_std": 0.1856733113527298,
      "reward_std_all": 0.18593017011880875,
      "rewards-max/forensics_sinkhorn_reward": 0.9736596345901489,
      "rewards-max/forensics_strict_edge_reward": 0.9158510267734528,
      "rewards-max/forensics_tiou_reward": 0.9628950953483582,
      "rewards/forensics_sinkhorn_reward": 0.4616287648677826,
      "rewards/forensics_strict_edge_reward": 0.3132481127977371,
      "rewards/forensics_tiou_reward": 0.35011737048625946,
      "step": 684
    },
    {
      "completion_length": 38.21875,
      "epoch": 1.4335078534031414,
      "grad_norm": 24.285921096801758,
      "kl": 4.0390625,
      "learning_rate": 6.415094339622641e-07,
      "loss": 0.294,
      "reward": 0.6868141293525696,
      "reward_std": 0.09387017041444778,
      "reward_std_all": 0.10559986904263496,
      "rewards-max/forensics_sinkhorn_reward": 0.9863470494747162,
      "rewards-max/forensics_strict_edge_reward": 0.9677534997463226,
      "rewards-max/forensics_tiou_reward": 0.994374543428421,
      "rewards/forensics_sinkhorn_reward": 0.6136327385902405,
      "rewards/forensics_strict_edge_reward": 0.38167591392993927,
      "rewards/forensics_tiou_reward": 0.4618613123893738,
      "step": 685
    },
    {
      "completion_length": 41.640625,
      "epoch": 1.4356020942408376,
      "grad_norm": 27.906705856323242,
      "kl": 5.03125,
      "learning_rate": 6.409853249475891e-07,
      "loss": 0.4993,
      "reward": 0.6219124495983124,
      "reward_std": 0.13953818380832672,
      "reward_std_all": 0.16187674924731255,
      "rewards-max/forensics_sinkhorn_reward": 0.9871806204319,
      "rewards-max/forensics_strict_edge_reward": 0.9357379674911499,
      "rewards-max/forensics_tiou_reward": 0.9928747415542603,
      "rewards/forensics_sinkhorn_reward": 0.4702446907758713,
      "rewards/forensics_strict_edge_reward": 0.23863685131072998,
      "rewards/forensics_tiou_reward": 0.2655583322048187,
      "step": 686
    },
    {
      "completion_length": 50.921875,
      "epoch": 1.4376963350785341,
      "grad_norm": 50.897117614746094,
      "kl": 4.5390625,
      "learning_rate": 6.40461215932914e-07,
      "loss": 0.4542,
      "reward": 0.6493012309074402,
      "reward_std": 0.13564006239175797,
      "reward_std_all": 0.14653217419981956,
      "rewards-max/forensics_sinkhorn_reward": 0.9868140816688538,
      "rewards-max/forensics_strict_edge_reward": 0.946017861366272,
      "rewards-max/forensics_tiou_reward": 0.9914548397064209,
      "rewards/forensics_sinkhorn_reward": 0.5155063271522522,
      "rewards/forensics_strict_edge_reward": 0.3362481892108917,
      "rewards/forensics_tiou_reward": 0.3571451008319855,
      "step": 687
    },
    {
      "completion_length": 54.109375,
      "epoch": 1.4397905759162304,
      "grad_norm": 24.6798038482666,
      "kl": 5.890625,
      "learning_rate": 6.39937106918239e-07,
      "loss": 0.4548,
      "reward": 0.5998222231864929,
      "reward_std": 0.16783790290355682,
      "reward_std_all": 0.1799587532877922,
      "rewards-max/forensics_sinkhorn_reward": 0.945705384016037,
      "rewards-max/forensics_strict_edge_reward": 0.8915697336196899,
      "rewards-max/forensics_tiou_reward": 0.9065763056278229,
      "rewards/forensics_sinkhorn_reward": 0.4549129158258438,
      "rewards/forensics_strict_edge_reward": 0.18560056388378143,
      "rewards/forensics_tiou_reward": 0.2592836171388626,
      "step": 688
    },
    {
      "completion_length": 51.921875,
      "epoch": 1.4418848167539267,
      "grad_norm": 25.35460090637207,
      "kl": 6.65625,
      "learning_rate": 6.394129979035639e-07,
      "loss": 0.4284,
      "reward": 0.6602393686771393,
      "reward_std": 0.13566774874925613,
      "reward_std_all": 0.14067239314317703,
      "rewards-max/forensics_sinkhorn_reward": 0.9885666370391846,
      "rewards-max/forensics_strict_edge_reward": 0.9574736058712006,
      "rewards-max/forensics_tiou_reward": 0.9983305633068085,
      "rewards/forensics_sinkhorn_reward": 0.562867134809494,
      "rewards/forensics_strict_edge_reward": 0.32889091968536377,
      "rewards/forensics_tiou_reward": 0.44891875982284546,
      "step": 689
    },
    {
      "completion_length": 47.25,
      "epoch": 1.4439790575916231,
      "grad_norm": 25.942739486694336,
      "kl": 4.734375,
      "learning_rate": 6.388888888888888e-07,
      "loss": 0.4119,
      "reward": 0.6829028129577637,
      "reward_std": 0.1453668251633644,
      "reward_std_all": 0.13946601003408432,
      "rewards-max/forensics_sinkhorn_reward": 0.9862017333507538,
      "rewards-max/forensics_strict_edge_reward": 0.947419673204422,
      "rewards-max/forensics_tiou_reward": 0.9890109896659851,
      "rewards/forensics_sinkhorn_reward": 0.5078350007534027,
      "rewards/forensics_strict_edge_reward": 0.35046765208244324,
      "rewards/forensics_tiou_reward": 0.3940321058034897,
      "step": 690
    },
    {
      "completion_length": 47.28125,
      "epoch": 1.4460732984293194,
      "grad_norm": 32.75910568237305,
      "kl": 6.046875,
      "learning_rate": 6.383647798742137e-07,
      "loss": 0.502,
      "reward": 0.6360175907611847,
      "reward_std": 0.1415577456355095,
      "reward_std_all": 0.15370426326990128,
      "rewards-max/forensics_sinkhorn_reward": 0.9848338067531586,
      "rewards-max/forensics_strict_edge_reward": 0.9169811904430389,
      "rewards-max/forensics_tiou_reward": 0.9837385416030884,
      "rewards/forensics_sinkhorn_reward": 0.49323758482933044,
      "rewards/forensics_strict_edge_reward": 0.25588129460811615,
      "rewards/forensics_tiou_reward": 0.3483400046825409,
      "step": 691
    },
    {
      "completion_length": 49.9375,
      "epoch": 1.4481675392670157,
      "grad_norm": 52.77130126953125,
      "kl": 7.0,
      "learning_rate": 6.378406708595387e-07,
      "loss": 0.5126,
      "reward": 0.5597179532051086,
      "reward_std": 0.17175192385911942,
      "reward_std_all": 0.19855375587940216,
      "rewards-max/forensics_sinkhorn_reward": 0.978159636259079,
      "rewards-max/forensics_strict_edge_reward": 0.6076286435127258,
      "rewards-max/forensics_tiou_reward": 0.9578195810317993,
      "rewards/forensics_sinkhorn_reward": 0.37715983390808105,
      "rewards/forensics_strict_edge_reward": 0.14490390941500664,
      "rewards/forensics_tiou_reward": 0.19392987340688705,
      "step": 692
    },
    {
      "completion_length": 52.578125,
      "epoch": 1.450261780104712,
      "grad_norm": 57.02268981933594,
      "kl": 6.953125,
      "learning_rate": 6.373165618448638e-07,
      "loss": 0.4296,
      "reward": 0.5862086713314056,
      "reward_std": 0.19707722961902618,
      "reward_std_all": 0.20595889538526535,
      "rewards-max/forensics_sinkhorn_reward": 0.9776380956172943,
      "rewards-max/forensics_strict_edge_reward": 0.8477883636951447,
      "rewards-max/forensics_tiou_reward": 0.9555799067020416,
      "rewards/forensics_sinkhorn_reward": 0.4161462336778641,
      "rewards/forensics_strict_edge_reward": 0.1717499941587448,
      "rewards/forensics_tiou_reward": 0.2234511449933052,
      "step": 693
    },
    {
      "completion_length": 67.78125,
      "epoch": 1.4523560209424085,
      "grad_norm": 46.656829833984375,
      "kl": 8.125,
      "learning_rate": 6.367924528301887e-07,
      "loss": 0.5446,
      "reward": 0.6113086640834808,
      "reward_std": 0.21365363895893097,
      "reward_std_all": 0.20291268080472946,
      "rewards-max/forensics_sinkhorn_reward": 0.9837563931941986,
      "rewards-max/forensics_strict_edge_reward": 0.8983768224716187,
      "rewards-max/forensics_tiou_reward": 0.9795887172222137,
      "rewards/forensics_sinkhorn_reward": 0.5070337951183319,
      "rewards/forensics_strict_edge_reward": 0.27357110381126404,
      "rewards/forensics_tiou_reward": 0.3691130578517914,
      "step": 694
    },
    {
      "completion_length": 49.765625,
      "epoch": 1.4544502617801047,
      "grad_norm": 41.5693244934082,
      "kl": 5.71875,
      "learning_rate": 6.362683438155137e-07,
      "loss": 0.4599,
      "reward": 0.6231542527675629,
      "reward_std": 0.15933754295110703,
      "reward_std_all": 0.16423477232456207,
      "rewards-max/forensics_sinkhorn_reward": 0.9872148036956787,
      "rewards-max/forensics_strict_edge_reward": 0.946017861366272,
      "rewards-max/forensics_tiou_reward": 0.9930198788642883,
      "rewards/forensics_sinkhorn_reward": 0.4328705817461014,
      "rewards/forensics_strict_edge_reward": 0.22700370848178864,
      "rewards/forensics_tiou_reward": 0.2749551311135292,
      "step": 695
    },
    {
      "completion_length": 38.65625,
      "epoch": 1.456544502617801,
      "grad_norm": 28.195117950439453,
      "kl": 5.78125,
      "learning_rate": 6.357442348008385e-07,
      "loss": 0.4091,
      "reward": 0.6099539995193481,
      "reward_std": 0.21313218772411346,
      "reward_std_all": 0.2046290785074234,
      "rewards-max/forensics_sinkhorn_reward": 0.988498866558075,
      "rewards-max/forensics_strict_edge_reward": 0.9890114367008209,
      "rewards-max/forensics_tiou_reward": 0.9980638921260834,
      "rewards/forensics_sinkhorn_reward": 0.5158957690000534,
      "rewards/forensics_strict_edge_reward": 0.32180555909872055,
      "rewards/forensics_tiou_reward": 0.3745407164096832,
      "step": 696
    },
    {
      "completion_length": 61.5,
      "epoch": 1.4586387434554973,
      "grad_norm": 55.55792999267578,
      "kl": 7.984375,
      "learning_rate": 6.352201257861635e-07,
      "loss": 0.5727,
      "reward": 0.6379373371601105,
      "reward_std": 0.20356105268001556,
      "reward_std_all": 0.18951945006847382,
      "rewards-max/forensics_sinkhorn_reward": 0.9868706166744232,
      "rewards-max/forensics_strict_edge_reward": 0.9169811904430389,
      "rewards-max/forensics_tiou_reward": 0.9916689693927765,
      "rewards/forensics_sinkhorn_reward": 0.4486064016819,
      "rewards/forensics_strict_edge_reward": 0.29536980390548706,
      "rewards/forensics_tiou_reward": 0.3003832846879959,
      "step": 697
    },
    {
      "completion_length": 63.125,
      "epoch": 1.4607329842931938,
      "grad_norm": 41.11799240112305,
      "kl": 4.375,
      "learning_rate": 6.346960167714884e-07,
      "loss": 0.3355,
      "reward": 0.5849645137786865,
      "reward_std": 0.20256179571151733,
      "reward_std_all": 0.2131635770201683,
      "rewards-max/forensics_sinkhorn_reward": 0.9853219985961914,
      "rewards-max/forensics_strict_edge_reward": 0.9062341153621674,
      "rewards-max/forensics_tiou_reward": 0.9856325089931488,
      "rewards/forensics_sinkhorn_reward": 0.5197432339191437,
      "rewards/forensics_strict_edge_reward": 0.2823491096496582,
      "rewards/forensics_tiou_reward": 0.36635367572307587,
      "step": 698
    },
    {
      "completion_length": 41.015625,
      "epoch": 1.46282722513089,
      "grad_norm": 45.025123596191406,
      "kl": 4.3046875,
      "learning_rate": 6.341719077568134e-07,
      "loss": 0.3974,
      "reward": 0.6491582691669464,
      "reward_std": 0.11924906820058823,
      "reward_std_all": 0.13535142689943314,
      "rewards-max/forensics_sinkhorn_reward": 0.9885256886482239,
      "rewards-max/forensics_strict_edge_reward": 0.9364311099052429,
      "rewards-max/forensics_tiou_reward": 0.9981676042079926,
      "rewards/forensics_sinkhorn_reward": 0.5226916372776031,
      "rewards/forensics_strict_edge_reward": 0.2903807461261749,
      "rewards/forensics_tiou_reward": 0.35426436364650726,
      "step": 699
    },
    {
      "completion_length": 49.484375,
      "epoch": 1.4649214659685863,
      "grad_norm": 33.466121673583984,
      "kl": 3.4921875,
      "learning_rate": 6.336477987421383e-07,
      "loss": 0.3886,
      "reward": 0.6466743350028992,
      "reward_std": 0.10033049434423447,
      "reward_std_all": 0.12623273953795433,
      "rewards-max/forensics_sinkhorn_reward": 0.9772405028343201,
      "rewards-max/forensics_strict_edge_reward": 0.9565287232398987,
      "rewards-max/forensics_tiou_reward": 0.9702531397342682,
      "rewards/forensics_sinkhorn_reward": 0.5076686888933182,
      "rewards/forensics_strict_edge_reward": 0.3068460077047348,
      "rewards/forensics_tiou_reward": 0.3737930804491043,
      "step": 700
    },
    {
      "completion_length": 45.421875,
      "epoch": 1.4670157068062828,
      "grad_norm": 29.17562484741211,
      "kl": 5.09375,
      "learning_rate": 6.331236897274633e-07,
      "loss": 0.3953,
      "reward": 0.6230969727039337,
      "reward_std": 0.1599595993757248,
      "reward_std_all": 0.16676601767539978,
      "rewards-max/forensics_sinkhorn_reward": 0.834141880273819,
      "rewards-max/forensics_strict_edge_reward": 0.7868781685829163,
      "rewards-max/forensics_tiou_reward": 0.7439420819282532,
      "rewards/forensics_sinkhorn_reward": 0.461631715297699,
      "rewards/forensics_strict_edge_reward": 0.21363919228315353,
      "rewards/forensics_tiou_reward": 0.271624356508255,
      "step": 701
    },
    {
      "completion_length": 37.265625,
      "epoch": 1.469109947643979,
      "grad_norm": 24.431821823120117,
      "kl": 3.984375,
      "learning_rate": 6.325995807127882e-07,
      "loss": 0.39,
      "reward": 0.649905264377594,
      "reward_std": 0.13346567749977112,
      "reward_std_all": 0.14815960824489594,
      "rewards-max/forensics_sinkhorn_reward": 0.9506994187831879,
      "rewards-max/forensics_strict_edge_reward": 0.8104448914527893,
      "rewards-max/forensics_tiou_reward": 0.908122718334198,
      "rewards/forensics_sinkhorn_reward": 0.5364058911800385,
      "rewards/forensics_strict_edge_reward": 0.24428606778383255,
      "rewards/forensics_tiou_reward": 0.2808602973818779,
      "step": 702
    },
    {
      "completion_length": 35.734375,
      "epoch": 1.4712041884816753,
      "grad_norm": 29.700725555419922,
      "kl": 3.421875,
      "learning_rate": 6.320754716981131e-07,
      "loss": 0.336,
      "reward": 0.6093720197677612,
      "reward_std": 0.18947208672761917,
      "reward_std_all": 0.18993310630321503,
      "rewards-max/forensics_sinkhorn_reward": 0.9844571053981781,
      "rewards-max/forensics_strict_edge_reward": 0.905340164899826,
      "rewards-max/forensics_tiou_reward": 0.9822227954864502,
      "rewards/forensics_sinkhorn_reward": 0.4771719425916672,
      "rewards/forensics_strict_edge_reward": 0.27234312891960144,
      "rewards/forensics_tiou_reward": 0.29463743418455124,
      "step": 703
    },
    {
      "completion_length": 49.671875,
      "epoch": 1.4732984293193718,
      "grad_norm": 23.513498306274414,
      "kl": 4.765625,
      "learning_rate": 6.315513626834381e-07,
      "loss": 0.3287,
      "reward": 0.6225029528141022,
      "reward_std": 0.16328265517950058,
      "reward_std_all": 0.16811127215623856,
      "rewards-max/forensics_sinkhorn_reward": 0.8520194292068481,
      "rewards-max/forensics_strict_edge_reward": 0.6570772528648376,
      "rewards-max/forensics_tiou_reward": 0.7220498621463776,
      "rewards/forensics_sinkhorn_reward": 0.4362276643514633,
      "rewards/forensics_strict_edge_reward": 0.19249313324689865,
      "rewards/forensics_tiou_reward": 0.20905159413814545,
      "step": 704
    },
    {
      "completion_length": 34.140625,
      "epoch": 1.475392670157068,
      "grad_norm": 26.201786041259766,
      "kl": 3.2578125,
      "learning_rate": 6.31027253668763e-07,
      "loss": 0.3251,
      "reward": 0.6867436170578003,
      "reward_std": 0.10937213897705078,
      "reward_std_all": 0.10931302979588509,
      "rewards-max/forensics_sinkhorn_reward": 0.9850890040397644,
      "rewards-max/forensics_strict_edge_reward": 0.828941822052002,
      "rewards-max/forensics_tiou_reward": 0.9847141206264496,
      "rewards/forensics_sinkhorn_reward": 0.5615752935409546,
      "rewards/forensics_strict_edge_reward": 0.31166112422943115,
      "rewards/forensics_tiou_reward": 0.37638533115386963,
      "step": 705
    },
    {
      "completion_length": 64.296875,
      "epoch": 1.4774869109947644,
      "grad_norm": 29.139511108398438,
      "kl": 4.84375,
      "learning_rate": 6.305031446540881e-07,
      "loss": 0.3516,
      "reward": 0.6796714663505554,
      "reward_std": 0.13161303102970123,
      "reward_std_all": 0.13688930496573448,
      "rewards-max/forensics_sinkhorn_reward": 0.9842815697193146,
      "rewards-max/forensics_strict_edge_reward": 0.9355069994926453,
      "rewards-max/forensics_tiou_reward": 0.9865257441997528,
      "rewards/forensics_sinkhorn_reward": 0.5684594810009003,
      "rewards/forensics_strict_edge_reward": 0.3144069015979767,
      "rewards/forensics_tiou_reward": 0.4262825548648834,
      "step": 706
    },
    {
      "completion_length": 35.4375,
      "epoch": 1.4795811518324609,
      "grad_norm": 30.156105041503906,
      "kl": 3.09375,
      "learning_rate": 6.29979035639413e-07,
      "loss": 0.3522,
      "reward": 0.6738090515136719,
      "reward_std": 0.148616973310709,
      "reward_std_all": 0.141844742000103,
      "rewards-max/forensics_sinkhorn_reward": 0.9853614568710327,
      "rewards-max/forensics_strict_edge_reward": 0.8983768224716187,
      "rewards-max/forensics_tiou_reward": 0.9857877790927887,
      "rewards/forensics_sinkhorn_reward": 0.5781141221523285,
      "rewards/forensics_strict_edge_reward": 0.3483920246362686,
      "rewards/forensics_tiou_reward": 0.437408983707428,
      "step": 707
    },
    {
      "completion_length": 62.796875,
      "epoch": 1.4816753926701571,
      "grad_norm": 55.20499801635742,
      "kl": 3.453125,
      "learning_rate": 6.294549266247379e-07,
      "loss": 0.2943,
      "reward": 0.6250321269035339,
      "reward_std": 0.17203880846500397,
      "reward_std_all": 0.17910929024219513,
      "rewards-max/forensics_sinkhorn_reward": 0.8460542857646942,
      "rewards-max/forensics_strict_edge_reward": 0.803118109703064,
      "rewards-max/forensics_tiou_reward": 0.7892215847969055,
      "rewards/forensics_sinkhorn_reward": 0.4855532795190811,
      "rewards/forensics_strict_edge_reward": 0.22081471979618073,
      "rewards/forensics_tiou_reward": 0.3081066533923149,
      "step": 708
    },
    {
      "completion_length": 54.671875,
      "epoch": 1.4837696335078534,
      "grad_norm": 29.15141487121582,
      "kl": 5.875,
      "learning_rate": 6.289308176100629e-07,
      "loss": 0.4168,
      "reward": 0.668212354183197,
      "reward_std": 0.11594247445464134,
      "reward_std_all": 0.1257864125072956,
      "rewards-max/forensics_sinkhorn_reward": 0.9725027978420258,
      "rewards-max/forensics_strict_edge_reward": 0.9048932790756226,
      "rewards-max/forensics_tiou_reward": 0.9587394893169403,
      "rewards/forensics_sinkhorn_reward": 0.480813667178154,
      "rewards/forensics_strict_edge_reward": 0.29646679013967514,
      "rewards/forensics_tiou_reward": 0.3051028698682785,
      "step": 709
    },
    {
      "completion_length": 54.15625,
      "epoch": 1.4858638743455497,
      "grad_norm": 43.931053161621094,
      "kl": 5.84375,
      "learning_rate": 6.284067085953878e-07,
      "loss": 0.4612,
      "reward": 0.6738070547580719,
      "reward_std": 0.12167604640126228,
      "reward_std_all": 0.12491209805011749,
      "rewards-max/forensics_sinkhorn_reward": 0.9867846071720123,
      "rewards-max/forensics_strict_edge_reward": 0.9357379674911499,
      "rewards-max/forensics_tiou_reward": 0.9913243055343628,
      "rewards/forensics_sinkhorn_reward": 0.5203032195568085,
      "rewards/forensics_strict_edge_reward": 0.31531578302383423,
      "rewards/forensics_tiou_reward": 0.3726181983947754,
      "step": 710
    },
    {
      "completion_length": 24.46875,
      "epoch": 1.487958115183246,
      "grad_norm": 18.8195743560791,
      "kl": 2.234375,
      "learning_rate": 6.278825995807128e-07,
      "loss": 0.2036,
      "reward": 0.6806367337703705,
      "reward_std": 0.12260957062244415,
      "reward_std_all": 0.12199937552213669,
      "rewards-max/forensics_sinkhorn_reward": 0.9886867105960846,
      "rewards-max/forensics_strict_edge_reward": 0.9782643616199493,
      "rewards-max/forensics_tiou_reward": 0.9988023936748505,
      "rewards/forensics_sinkhorn_reward": 0.6149644255638123,
      "rewards/forensics_strict_edge_reward": 0.3814220726490021,
      "rewards/forensics_tiou_reward": 0.47379744052886963,
      "step": 711
    },
    {
      "completion_length": 36.125,
      "epoch": 1.4900523560209424,
      "grad_norm": 32.157386779785156,
      "kl": 3.2421875,
      "learning_rate": 6.273584905660377e-07,
      "loss": 0.1919,
      "reward": 0.6426628232002258,
      "reward_std": 0.11548373103141785,
      "reward_std_all": 0.13759224116802216,
      "rewards-max/forensics_sinkhorn_reward": 0.8986881077289581,
      "rewards-max/forensics_strict_edge_reward": 0.6945675760507584,
      "rewards-max/forensics_tiou_reward": 0.8022967875003815,
      "rewards/forensics_sinkhorn_reward": 0.5388667285442352,
      "rewards/forensics_strict_edge_reward": 0.16956694424152374,
      "rewards/forensics_tiou_reward": 0.2954955920577049,
      "step": 712
    },
    {
      "completion_length": 36.296875,
      "epoch": 1.4921465968586387,
      "grad_norm": 21.66863441467285,
      "kl": 4.140625,
      "learning_rate": 6.268343815513627e-07,
      "loss": 0.2808,
      "reward": 0.698405921459198,
      "reward_std": 0.09902209788560867,
      "reward_std_all": 0.1054789088666439,
      "rewards-max/forensics_sinkhorn_reward": 0.9871390163898468,
      "rewards-max/forensics_strict_edge_reward": 0.9149472117424011,
      "rewards-max/forensics_tiou_reward": 0.9927152991294861,
      "rewards/forensics_sinkhorn_reward": 0.5758073925971985,
      "rewards/forensics_strict_edge_reward": 0.3222087025642395,
      "rewards/forensics_tiou_reward": 0.4175049215555191,
      "step": 713
    },
    {
      "completion_length": 38.96875,
      "epoch": 1.494240837696335,
      "grad_norm": 30.444150924682617,
      "kl": 3.6484375,
      "learning_rate": 6.263102725366875e-07,
      "loss": 0.382,
      "reward": 0.6798550188541412,
      "reward_std": 0.1373695321381092,
      "reward_std_all": 0.13330600410699844,
      "rewards-max/forensics_sinkhorn_reward": 0.9853300750255585,
      "rewards-max/forensics_strict_edge_reward": 0.9158510267734528,
      "rewards-max/forensics_tiou_reward": 0.9856313169002533,
      "rewards/forensics_sinkhorn_reward": 0.5721279978752136,
      "rewards/forensics_strict_edge_reward": 0.3152821362018585,
      "rewards/forensics_tiou_reward": 0.4042094200849533,
      "step": 714
    },
    {
      "completion_length": 44.84375,
      "epoch": 1.4963350785340315,
      "grad_norm": 28.40567398071289,
      "kl": 4.9375,
      "learning_rate": 6.257861635220125e-07,
      "loss": 0.3884,
      "reward": 0.7031184136867523,
      "reward_std": 0.10862158611416817,
      "reward_std_all": 0.10615911334753036,
      "rewards-max/forensics_sinkhorn_reward": 0.9877695143222809,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9951884746551514,
      "rewards/forensics_sinkhorn_reward": 0.5809151828289032,
      "rewards/forensics_strict_edge_reward": 0.3597126007080078,
      "rewards/forensics_tiou_reward": 0.40040963888168335,
      "step": 715
    },
    {
      "completion_length": 49.03125,
      "epoch": 1.4984293193717277,
      "grad_norm": 30.525390625,
      "kl": 4.53125,
      "learning_rate": 6.252620545073374e-07,
      "loss": 0.3988,
      "reward": 0.6441512703895569,
      "reward_std": 0.19537128880620003,
      "reward_std_all": 0.18272123858332634,
      "rewards-max/forensics_sinkhorn_reward": 0.9865707755088806,
      "rewards-max/forensics_strict_edge_reward": 0.9464850425720215,
      "rewards-max/forensics_tiou_reward": 0.9904949963092804,
      "rewards/forensics_sinkhorn_reward": 0.5310342758893967,
      "rewards/forensics_strict_edge_reward": 0.28561000525951385,
      "rewards/forensics_tiou_reward": 0.37527190148830414,
      "step": 716
    },
    {
      "completion_length": 45.109375,
      "epoch": 1.500523560209424,
      "grad_norm": 31.279739379882812,
      "kl": 4.2421875,
      "learning_rate": 6.247379454926625e-07,
      "loss": 0.3978,
      "reward": 0.5782870650291443,
      "reward_std": 0.15854917466640472,
      "reward_std_all": 0.1876213476061821,
      "rewards-max/forensics_sinkhorn_reward": 0.9869681596755981,
      "rewards-max/forensics_strict_edge_reward": 0.9151731729507446,
      "rewards-max/forensics_tiou_reward": 0.992044597864151,
      "rewards/forensics_sinkhorn_reward": 0.44799308478832245,
      "rewards/forensics_strict_edge_reward": 0.24000731855630875,
      "rewards/forensics_tiou_reward": 0.29229943454265594,
      "step": 717
    },
    {
      "completion_length": 44.234375,
      "epoch": 1.5026178010471205,
      "grad_norm": 32.54164505004883,
      "kl": 5.3515625,
      "learning_rate": 6.242138364779875e-07,
      "loss": 0.3378,
      "reward": 0.6735872328281403,
      "reward_std": 0.1603207141160965,
      "reward_std_all": 0.14934422820806503,
      "rewards-max/forensics_sinkhorn_reward": 0.9855024516582489,
      "rewards-max/forensics_strict_edge_reward": 0.9567649364471436,
      "rewards-max/forensics_tiou_reward": 0.9912121593952179,
      "rewards/forensics_sinkhorn_reward": 0.5370742380619049,
      "rewards/forensics_strict_edge_reward": 0.3287600725889206,
      "rewards/forensics_tiou_reward": 0.4020886719226837,
      "step": 718
    },
    {
      "completion_length": 71.4375,
      "epoch": 1.5047120418848168,
      "grad_norm": 39.093414306640625,
      "kl": 6.96875,
      "learning_rate": 6.236897274633124e-07,
      "loss": 0.4633,
      "reward": 0.6096629500389099,
      "reward_std": 0.15421109646558762,
      "reward_std_all": 0.1684885174036026,
      "rewards-max/forensics_sinkhorn_reward": 0.9870235621929169,
      "rewards-max/forensics_strict_edge_reward": 0.9567649364471436,
      "rewards-max/forensics_tiou_reward": 0.9922499358654022,
      "rewards/forensics_sinkhorn_reward": 0.47044500708580017,
      "rewards/forensics_strict_edge_reward": 0.35388994216918945,
      "rewards/forensics_tiou_reward": 0.35494595766067505,
      "step": 719
    },
    {
      "completion_length": 58.046875,
      "epoch": 1.506806282722513,
      "grad_norm": 31.21193504333496,
      "kl": 5.875,
      "learning_rate": 6.231656184486373e-07,
      "loss": 0.3209,
      "reward": 0.6041970551013947,
      "reward_std": 0.15096522867679596,
      "reward_std_all": 0.17018447816371918,
      "rewards-max/forensics_sinkhorn_reward": 0.9612901210784912,
      "rewards-max/forensics_strict_edge_reward": 0.7118590325117111,
      "rewards-max/forensics_tiou_reward": 0.9001071155071259,
      "rewards/forensics_sinkhorn_reward": 0.44127707183361053,
      "rewards/forensics_strict_edge_reward": 0.12366623803973198,
      "rewards/forensics_tiou_reward": 0.19570837914943695,
      "step": 720
    },
    {
      "completion_length": 34.484375,
      "epoch": 1.5089005235602095,
      "grad_norm": 30.610502243041992,
      "kl": 3.0,
      "learning_rate": 6.226415094339622e-07,
      "loss": 0.0934,
      "reward": 0.5979203283786774,
      "reward_std": 0.16924619302153587,
      "reward_std_all": 0.18637053295969963,
      "rewards-max/forensics_sinkhorn_reward": 0.9453800916671753,
      "rewards-max/forensics_strict_edge_reward": 0.8850063383579254,
      "rewards-max/forensics_tiou_reward": 0.9065219759941101,
      "rewards/forensics_sinkhorn_reward": 0.44388099014759064,
      "rewards/forensics_strict_edge_reward": 0.20776579529047012,
      "rewards/forensics_tiou_reward": 0.21694883704185486,
      "step": 721
    },
    {
      "completion_length": 78.75,
      "epoch": 1.5109947643979058,
      "grad_norm": 27.30595588684082,
      "kl": 6.75,
      "learning_rate": 6.221174004192872e-07,
      "loss": 0.5241,
      "reward": 0.580516129732132,
      "reward_std": 0.15782087296247482,
      "reward_std_all": 0.18370672315359116,
      "rewards-max/forensics_sinkhorn_reward": 0.8625302016735077,
      "rewards-max/forensics_strict_edge_reward": 0.7041463553905487,
      "rewards-max/forensics_tiou_reward": 0.7368100583553314,
      "rewards/forensics_sinkhorn_reward": 0.39925020933151245,
      "rewards/forensics_strict_edge_reward": 0.18584545701742172,
      "rewards/forensics_tiou_reward": 0.19699688255786896,
      "step": 722
    },
    {
      "completion_length": 82.015625,
      "epoch": 1.513089005235602,
      "grad_norm": 56.25462341308594,
      "kl": 5.5,
      "learning_rate": 6.215932914046121e-07,
      "loss": 0.4679,
      "reward": 0.6391095817089081,
      "reward_std": 0.1959865391254425,
      "reward_std_all": 0.1860499158501625,
      "rewards-max/forensics_sinkhorn_reward": 0.9873419404029846,
      "rewards-max/forensics_strict_edge_reward": 0.9567649364471436,
      "rewards-max/forensics_tiou_reward": 0.9935023188591003,
      "rewards/forensics_sinkhorn_reward": 0.5288574993610382,
      "rewards/forensics_strict_edge_reward": 0.33255384117364883,
      "rewards/forensics_tiou_reward": 0.39866746962070465,
      "step": 723
    },
    {
      "completion_length": 60.71875,
      "epoch": 1.5151832460732986,
      "grad_norm": 32.65022659301758,
      "kl": 4.7734375,
      "learning_rate": 6.210691823899371e-07,
      "loss": 0.3213,
      "reward": 0.6229605376720428,
      "reward_std": 0.11190588772296906,
      "reward_std_all": 0.13857350498437881,
      "rewards-max/forensics_sinkhorn_reward": 0.9877111613750458,
      "rewards-max/forensics_strict_edge_reward": 0.9565287232398987,
      "rewards-max/forensics_tiou_reward": 0.9949668347835541,
      "rewards/forensics_sinkhorn_reward": 0.4825257658958435,
      "rewards/forensics_strict_edge_reward": 0.22119811177253723,
      "rewards/forensics_tiou_reward": 0.25667862594127655,
      "step": 724
    },
    {
      "completion_length": 58.0625,
      "epoch": 1.5172774869109946,
      "grad_norm": 22.121639251708984,
      "kl": 5.1875,
      "learning_rate": 6.205450733752621e-07,
      "loss": 0.4983,
      "reward": 0.6722906231880188,
      "reward_std": 0.17113515734672546,
      "reward_std_all": 0.16053515672683716,
      "rewards-max/forensics_sinkhorn_reward": 0.9884056448936462,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9976944625377655,
      "rewards/forensics_sinkhorn_reward": 0.5381393730640411,
      "rewards/forensics_strict_edge_reward": 0.3446955233812332,
      "rewards/forensics_tiou_reward": 0.40429021418094635,
      "step": 725
    },
    {
      "completion_length": 66.375,
      "epoch": 1.5193717277486911,
      "grad_norm": 38.75007629394531,
      "kl": 4.5625,
      "learning_rate": 6.200209643605869e-07,
      "loss": 0.3909,
      "reward": 0.654413253068924,
      "reward_std": 0.15400130301713943,
      "reward_std_all": 0.15796931833028793,
      "rewards-max/forensics_sinkhorn_reward": 0.9874768257141113,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9940358400344849,
      "rewards/forensics_sinkhorn_reward": 0.5089257210493088,
      "rewards/forensics_strict_edge_reward": 0.3312351405620575,
      "rewards/forensics_tiou_reward": 0.34022341668605804,
      "step": 726
    },
    {
      "completion_length": 39.78125,
      "epoch": 1.5214659685863874,
      "grad_norm": 32.38941192626953,
      "kl": 4.3125,
      "learning_rate": 6.194968553459119e-07,
      "loss": 0.2471,
      "reward": 0.6561766862869263,
      "reward_std": 0.14476362243294716,
      "reward_std_all": 0.14723793417215347,
      "rewards-max/forensics_sinkhorn_reward": 0.9666426777839661,
      "rewards-max/forensics_strict_edge_reward": 0.8137359023094177,
      "rewards-max/forensics_tiou_reward": 0.9462071061134338,
      "rewards/forensics_sinkhorn_reward": 0.5223706066608429,
      "rewards/forensics_strict_edge_reward": 0.26955805718898773,
      "rewards/forensics_tiou_reward": 0.30955080687999725,
      "step": 727
    },
    {
      "completion_length": 61.984375,
      "epoch": 1.5235602094240837,
      "grad_norm": 41.21061706542969,
      "kl": 4.9375,
      "learning_rate": 6.189727463312368e-07,
      "loss": 0.3451,
      "reward": 0.6517715156078339,
      "reward_std": 0.16956951469182968,
      "reward_std_all": 0.16756601631641388,
      "rewards-max/forensics_sinkhorn_reward": 0.9885163307189941,
      "rewards-max/forensics_strict_edge_reward": 0.9890114367008209,
      "rewards-max/forensics_tiou_reward": 0.9981308281421661,
      "rewards/forensics_sinkhorn_reward": 0.5790371298789978,
      "rewards/forensics_strict_edge_reward": 0.377121701836586,
      "rewards/forensics_tiou_reward": 0.465580016374588,
      "step": 728
    },
    {
      "completion_length": 72.90625,
      "epoch": 1.5256544502617801,
      "grad_norm": 27.88237190246582,
      "kl": 5.765625,
      "learning_rate": 6.184486373165618e-07,
      "loss": 0.4151,
      "reward": 0.5327071845531464,
      "reward_std": 0.21065238118171692,
      "reward_std_all": 0.2329377681016922,
      "rewards-max/forensics_sinkhorn_reward": 0.9601370394229889,
      "rewards-max/forensics_strict_edge_reward": 0.8893801271915436,
      "rewards-max/forensics_tiou_reward": 0.935427188873291,
      "rewards/forensics_sinkhorn_reward": 0.3857044577598572,
      "rewards/forensics_strict_edge_reward": 0.1429104432463646,
      "rewards/forensics_tiou_reward": 0.19974631816148758,
      "step": 729
    },
    {
      "completion_length": 41.828125,
      "epoch": 1.5277486910994764,
      "grad_norm": 29.19942855834961,
      "kl": 3.3359375,
      "learning_rate": 6.179245283018868e-07,
      "loss": 0.1465,
      "reward": 0.6562177240848541,
      "reward_std": 0.10134021937847137,
      "reward_std_all": 0.11774925142526627,
      "rewards-max/forensics_sinkhorn_reward": 0.9874487817287445,
      "rewards-max/forensics_strict_edge_reward": 0.947419673204422,
      "rewards-max/forensics_tiou_reward": 0.9939420819282532,
      "rewards/forensics_sinkhorn_reward": 0.6001763939857483,
      "rewards/forensics_strict_edge_reward": 0.3212593048810959,
      "rewards/forensics_tiou_reward": 0.4301016181707382,
      "step": 730
    },
    {
      "completion_length": 43.609375,
      "epoch": 1.5298429319371727,
      "grad_norm": 28.44135856628418,
      "kl": 4.46875,
      "learning_rate": 6.174004192872118e-07,
      "loss": 0.3745,
      "reward": 0.5946845710277557,
      "reward_std": 0.14343231916427612,
      "reward_std_all": 0.17076639086008072,
      "rewards-max/forensics_sinkhorn_reward": 0.9877195656299591,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9949924945831299,
      "rewards/forensics_sinkhorn_reward": 0.4941297769546509,
      "rewards/forensics_strict_edge_reward": 0.26020167768001556,
      "rewards/forensics_tiou_reward": 0.28670625388622284,
      "step": 731
    },
    {
      "completion_length": 38.25,
      "epoch": 1.5319371727748692,
      "grad_norm": 30.201210021972656,
      "kl": 3.25,
      "learning_rate": 6.168763102725366e-07,
      "loss": 0.2827,
      "reward": 0.6338068842887878,
      "reward_std": 0.2016729861497879,
      "reward_std_all": 0.20052117109298706,
      "rewards-max/forensics_sinkhorn_reward": 0.988442599773407,
      "rewards-max/forensics_strict_edge_reward": 0.947419673204422,
      "rewards-max/forensics_tiou_reward": 0.997840166091919,
      "rewards/forensics_sinkhorn_reward": 0.5517094731330872,
      "rewards/forensics_strict_edge_reward": 0.3595913201570511,
      "rewards/forensics_tiou_reward": 0.4119182378053665,
      "step": 732
    },
    {
      "completion_length": 46.703125,
      "epoch": 1.5340314136125655,
      "grad_norm": 39.81776428222656,
      "kl": 4.9375,
      "learning_rate": 6.163522012578616e-07,
      "loss": 0.4433,
      "reward": 0.643484890460968,
      "reward_std": 0.1891418620944023,
      "reward_std_all": 0.17914917320013046,
      "rewards-max/forensics_sinkhorn_reward": 0.988323450088501,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.997372567653656,
      "rewards/forensics_sinkhorn_reward": 0.5703508853912354,
      "rewards/forensics_strict_edge_reward": 0.4373525530099869,
      "rewards/forensics_tiou_reward": 0.47242632508277893,
      "step": 733
    },
    {
      "completion_length": 27.03125,
      "epoch": 1.5361256544502617,
      "grad_norm": 22.844970703125,
      "kl": 2.5703125,
      "learning_rate": 6.158280922431866e-07,
      "loss": 0.165,
      "reward": 0.6430263221263885,
      "reward_std": 0.12567421793937683,
      "reward_std_all": 0.1476331576704979,
      "rewards-max/forensics_sinkhorn_reward": 0.9881832897663116,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.996822327375412,
      "rewards/forensics_sinkhorn_reward": 0.5497598052024841,
      "rewards/forensics_strict_edge_reward": 0.2626817226409912,
      "rewards/forensics_tiou_reward": 0.34078484773635864,
      "step": 734
    },
    {
      "completion_length": 39.3125,
      "epoch": 1.5382198952879582,
      "grad_norm": 22.915674209594727,
      "kl": 3.4140625,
      "learning_rate": 6.153039832285115e-07,
      "loss": 0.2679,
      "reward": 0.6572907269001007,
      "reward_std": 0.12840693444013596,
      "reward_std_all": 0.13936152681708336,
      "rewards-max/forensics_sinkhorn_reward": 0.9879961609840393,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9960842430591583,
      "rewards/forensics_sinkhorn_reward": 0.5820769667625427,
      "rewards/forensics_strict_edge_reward": 0.3632706552743912,
      "rewards/forensics_tiou_reward": 0.41957589983940125,
      "step": 735
    },
    {
      "completion_length": 27.609375,
      "epoch": 1.5403141361256545,
      "grad_norm": 19.336109161376953,
      "kl": 1.83984375,
      "learning_rate": 6.147798742138365e-07,
      "loss": 0.0874,
      "reward": 0.6455114483833313,
      "reward_std": 0.10363774746656418,
      "reward_std_all": 0.1319294571876526,
      "rewards-max/forensics_sinkhorn_reward": 0.9762905240058899,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.972042053937912,
      "rewards/forensics_sinkhorn_reward": 0.48671315610408783,
      "rewards/forensics_strict_edge_reward": 0.16252786666154861,
      "rewards/forensics_tiou_reward": 0.20426657795906067,
      "step": 736
    },
    {
      "completion_length": 38.25,
      "epoch": 1.5424083769633508,
      "grad_norm": 36.971744537353516,
      "kl": 4.828125,
      "learning_rate": 6.142557651991614e-07,
      "loss": 0.4197,
      "reward": 0.6445138454437256,
      "reward_std": 0.14169345796108246,
      "reward_std_all": 0.1481519341468811,
      "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": 0.46910615265369415,
      "rewards/forensics_strict_edge_reward": 0.27430306375026703,
      "rewards/forensics_tiou_reward": 0.3024500608444214,
      "step": 737
    },
    {
      "completion_length": 30.59375,
      "epoch": 1.5445026178010473,
      "grad_norm": 23.42348289489746,
      "kl": 4.21875,
      "learning_rate": 6.137316561844863e-07,
      "loss": 0.2984,
      "reward": 0.6997696757316589,
      "reward_std": 0.15017062425613403,
      "reward_std_all": 0.13531583547592163,
      "rewards-max/forensics_sinkhorn_reward": 0.9871909618377686,
      "rewards-max/forensics_strict_edge_reward": 0.9574736058712006,
      "rewards-max/forensics_tiou_reward": 0.9929329454898834,
      "rewards/forensics_sinkhorn_reward": 0.6447956264019012,
      "rewards/forensics_strict_edge_reward": 0.4556438773870468,
      "rewards/forensics_tiou_reward": 0.5116007626056671,
      "step": 738
    },
    {
      "completion_length": 34.828125,
      "epoch": 1.5465968586387433,
      "grad_norm": 30.29391098022461,
      "kl": 4.9609375,
      "learning_rate": 6.132075471698112e-07,
      "loss": 0.3114,
      "reward": 0.6217500567436218,
      "reward_std": 0.14025018364191055,
      "reward_std_all": 0.16023299098014832,
      "rewards-max/forensics_sinkhorn_reward": 0.9787529706954956,
      "rewards-max/forensics_strict_edge_reward": 0.6074669063091278,
      "rewards-max/forensics_tiou_reward": 0.9601705372333527,
      "rewards/forensics_sinkhorn_reward": 0.4862196743488312,
      "rewards/forensics_strict_edge_reward": 0.17331186681985855,
      "rewards/forensics_tiou_reward": 0.27909715473651886,
      "step": 739
    },
    {
      "completion_length": 34.671875,
      "epoch": 1.5486910994764398,
      "grad_norm": 20.96885108947754,
      "kl": 3.5859375,
      "learning_rate": 6.126834381551362e-07,
      "loss": 0.3642,
      "reward": 0.6172292828559875,
      "reward_std": 0.17263668030500412,
      "reward_std_all": 0.18357933312654495,
      "rewards-max/forensics_sinkhorn_reward": 0.9870151579380035,
      "rewards-max/forensics_strict_edge_reward": 0.946017861366272,
      "rewards-max/forensics_tiou_reward": 0.9922284781932831,
      "rewards/forensics_sinkhorn_reward": 0.5191158056259155,
      "rewards/forensics_strict_edge_reward": 0.2869107276201248,
      "rewards/forensics_tiou_reward": 0.3252789080142975,
      "step": 740
    },
    {
      "completion_length": 43.046875,
      "epoch": 1.5507853403141363,
      "grad_norm": 30.445329666137695,
      "kl": 4.7265625,
      "learning_rate": 6.121593291404612e-07,
      "loss": 0.2912,
      "reward": 0.6501241028308868,
      "reward_std": 0.16536549478769302,
      "reward_std_all": 0.15856260061264038,
      "rewards-max/forensics_sinkhorn_reward": 0.9881715178489685,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.996774286031723,
      "rewards/forensics_sinkhorn_reward": 0.5080281347036362,
      "rewards/forensics_strict_edge_reward": 0.3060755133628845,
      "rewards/forensics_tiou_reward": 0.35688355565071106,
      "step": 741
    },
    {
      "completion_length": 35.96875,
      "epoch": 1.5528795811518323,
      "grad_norm": 20.839244842529297,
      "kl": 4.484375,
      "learning_rate": 6.116352201257862e-07,
      "loss": 0.2932,
      "reward": 0.633119136095047,
      "reward_std": 0.15204257518053055,
      "reward_std_all": 0.16038762032985687,
      "rewards-max/forensics_sinkhorn_reward": 0.9415823519229889,
      "rewards-max/forensics_strict_edge_reward": 0.9003686904907227,
      "rewards-max/forensics_tiou_reward": 0.9007111489772797,
      "rewards/forensics_sinkhorn_reward": 0.4613586664199829,
      "rewards/forensics_strict_edge_reward": 0.26661092787981033,
      "rewards/forensics_tiou_reward": 0.3057781308889389,
      "step": 742
    },
    {
      "completion_length": 40.234375,
      "epoch": 1.5549738219895288,
      "grad_norm": 42.349918365478516,
      "kl": 5.3125,
      "learning_rate": 6.111111111111112e-07,
      "loss": 0.4383,
      "reward": 0.5938127040863037,
      "reward_std": 0.20491333305835724,
      "reward_std_all": 0.2090834230184555,
      "rewards-max/forensics_sinkhorn_reward": 0.9882853627204895,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9972234070301056,
      "rewards/forensics_sinkhorn_reward": 0.4437977075576782,
      "rewards/forensics_strict_edge_reward": 0.28163282573223114,
      "rewards/forensics_tiou_reward": 0.32001379132270813,
      "step": 743
    },
    {
      "completion_length": 38.28125,
      "epoch": 1.557068062827225,
      "grad_norm": 28.390710830688477,
      "kl": 4.546875,
      "learning_rate": 6.10587002096436e-07,
      "loss": 0.3724,
      "reward": 0.6753325760364532,
      "reward_std": 0.15064097940921783,
      "reward_std_all": 0.14517010003328323,
      "rewards-max/forensics_sinkhorn_reward": 0.9789911508560181,
      "rewards-max/forensics_strict_edge_reward": 0.9464850425720215,
      "rewards-max/forensics_tiou_reward": 0.9762549996376038,
      "rewards/forensics_sinkhorn_reward": 0.5486673712730408,
      "rewards/forensics_strict_edge_reward": 0.34993113577365875,
      "rewards/forensics_tiou_reward": 0.4547799527645111,
      "step": 744
    },
    {
      "completion_length": 30.140625,
      "epoch": 1.5591623036649214,
      "grad_norm": 21.085800170898438,
      "kl": 3.484375,
      "learning_rate": 6.10062893081761e-07,
      "loss": 0.3315,
      "reward": 0.6598502397537231,
      "reward_std": 0.14292467385530472,
      "reward_std_all": 0.1462918221950531,
      "rewards-max/forensics_sinkhorn_reward": 0.968094527721405,
      "rewards-max/forensics_strict_edge_reward": 0.8217175900936127,
      "rewards-max/forensics_tiou_reward": 0.9533535242080688,
      "rewards/forensics_sinkhorn_reward": 0.5176790505647659,
      "rewards/forensics_strict_edge_reward": 0.23792918026447296,
      "rewards/forensics_tiou_reward": 0.3336936980485916,
      "step": 745
    },
    {
      "completion_length": 30.90625,
      "epoch": 1.5612565445026179,
      "grad_norm": 21.886571884155273,
      "kl": 4.34375,
      "learning_rate": 6.095387840670859e-07,
      "loss": 0.2637,
      "reward": 0.6461512744426727,
      "reward_std": 0.17179978638887405,
      "reward_std_all": 0.16877897083759308,
      "rewards-max/forensics_sinkhorn_reward": 0.969409853219986,
      "rewards-max/forensics_strict_edge_reward": 0.947419673204422,
      "rewards-max/forensics_tiou_reward": 0.9566822350025177,
      "rewards/forensics_sinkhorn_reward": 0.5236051231622696,
      "rewards/forensics_strict_edge_reward": 0.31990325450897217,
      "rewards/forensics_tiou_reward": 0.315861776471138,
      "step": 746
    },
    {
      "completion_length": 36.0,
      "epoch": 1.5633507853403141,
      "grad_norm": 23.24614906311035,
      "kl": 4.1328125,
      "learning_rate": 6.090146750524109e-07,
      "loss": 0.3645,
      "reward": 0.6457470953464508,
      "reward_std": 0.160905122756958,
      "reward_std_all": 0.15677310526371002,
      "rewards-max/forensics_sinkhorn_reward": 0.8575385212898254,
      "rewards-max/forensics_strict_edge_reward": 0.8168272376060486,
      "rewards-max/forensics_tiou_reward": 0.8158701062202454,
      "rewards/forensics_sinkhorn_reward": 0.4545510411262512,
      "rewards/forensics_strict_edge_reward": 0.24001531302928925,
      "rewards/forensics_tiou_reward": 0.28342224657535553,
      "step": 747
    },
    {
      "completion_length": 27.046875,
      "epoch": 1.5654450261780104,
      "grad_norm": 36.4339485168457,
      "kl": 3.234375,
      "learning_rate": 6.084905660377358e-07,
      "loss": 0.3562,
      "reward": 0.6238115429878235,
      "reward_std": 0.17606844007968903,
      "reward_std_all": 0.1789821982383728,
      "rewards-max/forensics_sinkhorn_reward": 0.987462043762207,
      "rewards-max/forensics_strict_edge_reward": 0.9677534997463226,
      "rewards-max/forensics_tiou_reward": 0.9939940571784973,
      "rewards/forensics_sinkhorn_reward": 0.5317098945379257,
      "rewards/forensics_strict_edge_reward": 0.3074716851115227,
      "rewards/forensics_tiou_reward": 0.36799901723861694,
      "step": 748
    },
    {
      "completion_length": 28.96875,
      "epoch": 1.567539267015707,
      "grad_norm": 24.5753116607666,
      "kl": 3.1328125,
      "learning_rate": 6.079664570230608e-07,
      "loss": 0.2293,
      "reward": 0.6396543383598328,
      "reward_std": 0.16385327279567719,
      "reward_std_all": 0.1627293974161148,
      "rewards-max/forensics_sinkhorn_reward": 0.9812990427017212,
      "rewards-max/forensics_strict_edge_reward": 0.9048932790756226,
      "rewards-max/forensics_tiou_reward": 0.9800885021686554,
      "rewards/forensics_sinkhorn_reward": 0.48866812884807587,
      "rewards/forensics_strict_edge_reward": 0.29231712222099304,
      "rewards/forensics_tiou_reward": 0.32964494824409485,
      "step": 749
    },
    {
      "completion_length": 52.8125,
      "epoch": 1.5696335078534032,
      "grad_norm": 23.580957412719727,
      "kl": 6.359375,
      "learning_rate": 6.074423480083857e-07,
      "loss": 0.4586,
      "reward": 0.6253271400928497,
      "reward_std": 0.14608263969421387,
      "reward_std_all": 0.15978778898715973,
      "rewards-max/forensics_sinkhorn_reward": 0.9880504012107849,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.996297687292099,
      "rewards/forensics_sinkhorn_reward": 0.5536990016698837,
      "rewards/forensics_strict_edge_reward": 0.3431321233510971,
      "rewards/forensics_tiou_reward": 0.4123115688562393,
      "step": 750
    },
    {
      "completion_length": 37.078125,
      "epoch": 1.5717277486910994,
      "grad_norm": 37.82460403442383,
      "kl": 5.4375,
      "learning_rate": 6.069182389937106e-07,
      "loss": 0.4513,
      "reward": 0.5691342055797577,
      "reward_std": 0.18101206421852112,
      "reward_std_all": 0.19935310631990433,
      "rewards-max/forensics_sinkhorn_reward": 0.9460267722606659,
      "rewards-max/forensics_strict_edge_reward": 0.8635591566562653,
      "rewards-max/forensics_tiou_reward": 0.9093910455703735,
      "rewards/forensics_sinkhorn_reward": 0.4412032514810562,
      "rewards/forensics_strict_edge_reward": 0.192766435444355,
      "rewards/forensics_tiou_reward": 0.22437234222888947,
      "step": 751
    },
    {
      "completion_length": 31.4375,
      "epoch": 1.573821989528796,
      "grad_norm": 25.91728401184082,
      "kl": 5.1015625,
      "learning_rate": 6.063941299790356e-07,
      "loss": 0.4151,
      "reward": 0.660359114408493,
      "reward_std": 0.13592873886227608,
      "reward_std_all": 0.14539729058742523,
      "rewards-max/forensics_sinkhorn_reward": 0.9874483942985535,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9939287304878235,
      "rewards/forensics_sinkhorn_reward": 0.5947471261024475,
      "rewards/forensics_strict_edge_reward": 0.3294019550085068,
      "rewards/forensics_tiou_reward": 0.4241979569196701,
      "step": 752
    },
    {
      "completion_length": 33.09375,
      "epoch": 1.5759162303664922,
      "grad_norm": 38.99451446533203,
      "kl": 5.0703125,
      "learning_rate": 6.058700209643605e-07,
      "loss": 0.3925,
      "reward": 0.5817466676235199,
      "reward_std": 0.18482592701911926,
      "reward_std_all": 0.20335006713867188,
      "rewards-max/forensics_sinkhorn_reward": 0.9880759418010712,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9963962137699127,
      "rewards/forensics_sinkhorn_reward": 0.46516774594783783,
      "rewards/forensics_strict_edge_reward": 0.19590217620134354,
      "rewards/forensics_tiou_reward": 0.2407642975449562,
      "step": 753
    },
    {
      "completion_length": 47.9375,
      "epoch": 1.5780104712041885,
      "grad_norm": 29.610904693603516,
      "kl": 6.90625,
      "learning_rate": 6.053459119496856e-07,
      "loss": 0.3844,
      "reward": 0.5825175642967224,
      "reward_std": 0.21363380551338196,
      "reward_std_all": 0.21120186150074005,
      "rewards-max/forensics_sinkhorn_reward": 0.972097635269165,
      "rewards-max/forensics_strict_edge_reward": 0.878633052110672,
      "rewards-max/forensics_tiou_reward": 0.9600389301776886,
      "rewards/forensics_sinkhorn_reward": 0.4788602441549301,
      "rewards/forensics_strict_edge_reward": 0.24185822159051895,
      "rewards/forensics_tiou_reward": 0.33771124482154846,
      "step": 754
    },
    {
      "completion_length": 34.65625,
      "epoch": 1.580104712041885,
      "grad_norm": 25.67333221435547,
      "kl": 5.484375,
      "learning_rate": 6.048218029350105e-07,
      "loss": 0.5051,
      "reward": 0.6310980319976807,
      "reward_std": 0.17362699657678604,
      "reward_std_all": 0.17444316297769547,
      "rewards-max/forensics_sinkhorn_reward": 0.9835250079631805,
      "rewards-max/forensics_strict_edge_reward": 0.8237898349761963,
      "rewards-max/forensics_tiou_reward": 0.9786132574081421,
      "rewards/forensics_sinkhorn_reward": 0.4808633029460907,
      "rewards/forensics_strict_edge_reward": 0.24619781225919724,
      "rewards/forensics_tiou_reward": 0.3148740082979202,
      "step": 755
    },
    {
      "completion_length": 36.09375,
      "epoch": 1.582198952879581,
      "grad_norm": 32.670108795166016,
      "kl": 4.78125,
      "learning_rate": 6.042976939203354e-07,
      "loss": 0.2467,
      "reward": 0.6261565089225769,
      "reward_std": 0.12024622410535812,
      "reward_std_all": 0.14895520359277725,
      "rewards-max/forensics_sinkhorn_reward": 0.9645561277866364,
      "rewards-max/forensics_strict_edge_reward": 0.6896179020404816,
      "rewards-max/forensics_tiou_reward": 0.9419783651828766,
      "rewards/forensics_sinkhorn_reward": 0.44480176270008087,
      "rewards/forensics_strict_edge_reward": 0.16490431874990463,
      "rewards/forensics_tiou_reward": 0.24207797646522522,
      "step": 756
    },
    {
      "completion_length": 34.296875,
      "epoch": 1.5842931937172775,
      "grad_norm": 24.98586654663086,
      "kl": 4.375,
      "learning_rate": 6.037735849056604e-07,
      "loss": 0.4022,
      "reward": 0.6296873986721039,
      "reward_std": 0.15101439505815506,
      "reward_std_all": 0.16153176873922348,
      "rewards-max/forensics_sinkhorn_reward": 0.9874376654624939,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9938817620277405,
      "rewards/forensics_sinkhorn_reward": 0.5146733820438385,
      "rewards/forensics_strict_edge_reward": 0.22512472420930862,
      "rewards/forensics_tiou_reward": 0.3792361915111542,
      "step": 757
    },
    {
      "completion_length": 40.78125,
      "epoch": 1.5863874345549738,
      "grad_norm": 22.257902145385742,
      "kl": 4.671875,
      "learning_rate": 6.032494758909853e-07,
      "loss": 0.398,
      "reward": 0.6662065386772156,
      "reward_std": 0.1634022817015648,
      "reward_std_all": 0.15304942429065704,
      "rewards-max/forensics_sinkhorn_reward": 0.9881921112537384,
      "rewards-max/forensics_strict_edge_reward": 0.9565287232398987,
      "rewards-max/forensics_tiou_reward": 0.9968553483486176,
      "rewards/forensics_sinkhorn_reward": 0.5878807306289673,
      "rewards/forensics_strict_edge_reward": 0.3764660209417343,
      "rewards/forensics_tiou_reward": 0.43824437260627747,
      "step": 758
    },
    {
      "completion_length": 35.1875,
      "epoch": 1.58848167539267,
      "grad_norm": 21.361635208129883,
      "kl": 3.7734375,
      "learning_rate": 6.027253668763103e-07,
      "loss": 0.3786,
      "reward": 0.6586382389068604,
      "reward_std": 0.1257011741399765,
      "reward_std_all": 0.13791534304618835,
      "rewards-max/forensics_sinkhorn_reward": 0.9886179864406586,
      "rewards-max/forensics_strict_edge_reward": 0.9782643616199493,
      "rewards-max/forensics_tiou_reward": 0.9985326528549194,
      "rewards/forensics_sinkhorn_reward": 0.6334372162818909,
      "rewards/forensics_strict_edge_reward": 0.3786895424127579,
      "rewards/forensics_tiou_reward": 0.47071702778339386,
      "step": 759
    },
    {
      "completion_length": 35.15625,
      "epoch": 1.5905759162303665,
      "grad_norm": 24.448577880859375,
      "kl": 3.109375,
      "learning_rate": 6.022012578616352e-07,
      "loss": 0.27,
      "reward": 0.6697236895561218,
      "reward_std": 0.14253748953342438,
      "reward_std_all": 0.14537647366523743,
      "rewards-max/forensics_sinkhorn_reward": 0.9878795146942139,
      "rewards-max/forensics_strict_edge_reward": 0.9565287232398987,
      "rewards-max/forensics_tiou_reward": 0.9956205785274506,
      "rewards/forensics_sinkhorn_reward": 0.6549778282642365,
      "rewards/forensics_strict_edge_reward": 0.40852999687194824,
      "rewards/forensics_tiou_reward": 0.5380854904651642,
      "step": 760
    },
    {
      "completion_length": 36.328125,
      "epoch": 1.5926701570680628,
      "grad_norm": 29.87828254699707,
      "kl": 5.0625,
      "learning_rate": 6.016771488469602e-07,
      "loss": 0.3669,
      "reward": 0.6354582905769348,
      "reward_std": 0.14701619744300842,
      "reward_std_all": 0.15627487748861313,
      "rewards-max/forensics_sinkhorn_reward": 0.9881016612052917,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9964961409568787,
      "rewards/forensics_sinkhorn_reward": 0.5210131406784058,
      "rewards/forensics_strict_edge_reward": 0.290124848484993,
      "rewards/forensics_tiou_reward": 0.3235928863286972,
      "step": 761
    },
    {
      "completion_length": 34.421875,
      "epoch": 1.594764397905759,
      "grad_norm": 72.66444396972656,
      "kl": 7.5,
      "learning_rate": 6.01153039832285e-07,
      "loss": 0.3845,
      "reward": 0.6569921672344208,
      "reward_std": 0.1506268009543419,
      "reward_std_all": 0.1513695865869522,
      "rewards-max/forensics_sinkhorn_reward": 0.9709664583206177,
      "rewards-max/forensics_strict_edge_reward": 0.8944784104824066,
      "rewards-max/forensics_tiou_reward": 0.957120805978775,
      "rewards/forensics_sinkhorn_reward": 0.5178192257881165,
      "rewards/forensics_strict_edge_reward": 0.2895311564207077,
      "rewards/forensics_tiou_reward": 0.3127017170190811,
      "step": 762
    },
    {
      "completion_length": 27.953125,
      "epoch": 1.5968586387434556,
      "grad_norm": 30.246597290039062,
      "kl": 3.3359375,
      "learning_rate": 6.0062893081761e-07,
      "loss": 0.189,
      "reward": 0.7221558094024658,
      "reward_std": 0.09855394065380096,
      "reward_std_all": 0.09420850686728954,
      "rewards-max/forensics_sinkhorn_reward": 0.9757943749427795,
      "rewards-max/forensics_strict_edge_reward": 0.9375866651535034,
      "rewards-max/forensics_tiou_reward": 0.9691543877124786,
      "rewards/forensics_sinkhorn_reward": 0.5796151459217072,
      "rewards/forensics_strict_edge_reward": 0.3118167966604233,
      "rewards/forensics_tiou_reward": 0.428374707698822,
      "step": 763
    },
    {
      "completion_length": 26.734375,
      "epoch": 1.5989528795811518,
      "grad_norm": 37.011924743652344,
      "kl": 4.6953125,
      "learning_rate": 6.001048218029349e-07,
      "loss": 0.3139,
      "reward": 0.654941201210022,
      "reward_std": 0.16254188120365143,
      "reward_std_all": 0.16477792710065842,
      "rewards-max/forensics_sinkhorn_reward": 0.9875067472457886,
      "rewards-max/forensics_strict_edge_reward": 0.9574736058712006,
      "rewards-max/forensics_tiou_reward": 0.994169145822525,
      "rewards/forensics_sinkhorn_reward": 0.5525612235069275,
      "rewards/forensics_strict_edge_reward": 0.32552629709243774,
      "rewards/forensics_tiou_reward": 0.394923597574234,
      "step": 764
    },
    {
      "completion_length": 38.9375,
      "epoch": 1.6010471204188481,
      "grad_norm": 30.09810447692871,
      "kl": 5.203125,
      "learning_rate": 5.995807127882599e-07,
      "loss": 0.2602,
      "reward": 0.6055014133453369,
      "reward_std": 0.20153699815273285,
      "reward_std_all": 0.19821998476982117,
      "rewards-max/forensics_sinkhorn_reward": 0.9127197265625,
      "rewards-max/forensics_strict_edge_reward": 0.7901000380516052,
      "rewards-max/forensics_tiou_reward": 0.8376535475254059,
      "rewards/forensics_sinkhorn_reward": 0.44506166875362396,
      "rewards/forensics_strict_edge_reward": 0.22326675802469254,
      "rewards/forensics_tiou_reward": 0.257423534989357,
      "step": 765
    },
    {
      "completion_length": 46.046875,
      "epoch": 1.6031413612565446,
      "grad_norm": 37.239471435546875,
      "kl": 5.0,
      "learning_rate": 5.99056603773585e-07,
      "loss": 0.4467,
      "reward": 0.6547594666481018,
      "reward_std": 0.12037017568945885,
      "reward_std_all": 0.13316159695386887,
      "rewards-max/forensics_sinkhorn_reward": 0.986309826374054,
      "rewards-max/forensics_strict_edge_reward": 0.9357379674911499,
      "rewards-max/forensics_tiou_reward": 0.9894645810127258,
      "rewards/forensics_sinkhorn_reward": 0.501752182841301,
      "rewards/forensics_strict_edge_reward": 0.24645745754241943,
      "rewards/forensics_tiou_reward": 0.299888551235199,
      "step": 766
    },
    {
      "completion_length": 29.78125,
      "epoch": 1.6052356020942409,
      "grad_norm": 27.561607360839844,
      "kl": 3.3125,
      "learning_rate": 5.985324947589099e-07,
      "loss": 0.2109,
      "reward": 0.6674756705760956,
      "reward_std": 0.15214323997497559,
      "reward_std_all": 0.1492176353931427,
      "rewards-max/forensics_sinkhorn_reward": 0.9876145422458649,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9945776164531708,
      "rewards/forensics_sinkhorn_reward": 0.5934245586395264,
      "rewards/forensics_strict_edge_reward": 0.3734452426433563,
      "rewards/forensics_tiou_reward": 0.4406086504459381,
      "step": 767
    },
    {
      "completion_length": 32.640625,
      "epoch": 1.6073298429319371,
      "grad_norm": 27.31714630126953,
      "kl": 5.171875,
      "learning_rate": 5.980083857442348e-07,
      "loss": 0.3492,
      "reward": 0.6543519496917725,
      "reward_std": 0.14041262865066528,
      "reward_std_all": 0.14566075801849365,
      "rewards-max/forensics_sinkhorn_reward": 0.9410297274589539,
      "rewards-max/forensics_strict_edge_reward": 0.842301219701767,
      "rewards-max/forensics_tiou_reward": 0.8905882835388184,
      "rewards/forensics_sinkhorn_reward": 0.5211040079593658,
      "rewards/forensics_strict_edge_reward": 0.192767933011055,
      "rewards/forensics_tiou_reward": 0.23633234202861786,
      "step": 768
    },
    {
      "completion_length": 28.1875,
      "epoch": 1.6094240837696336,
      "grad_norm": 18.230772018432617,
      "kl": 3.515625,
      "learning_rate": 5.974842767295597e-07,
      "loss": 0.2213,
      "reward": 0.7095624208450317,
      "reward_std": 0.11967406421899796,
      "reward_std_all": 0.11443822830915451,
      "rewards-max/forensics_sinkhorn_reward": 0.9882461130619049,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9970659017562866,
      "rewards/forensics_sinkhorn_reward": 0.6157732009887695,
      "rewards/forensics_strict_edge_reward": 0.37335897982120514,
      "rewards/forensics_tiou_reward": 0.4187934100627899,
      "step": 769
    },
    {
      "completion_length": 30.40625,
      "epoch": 1.6115183246073297,
      "grad_norm": 18.321422576904297,
      "kl": 3.234375,
      "learning_rate": 5.969601677148847e-07,
      "loss": 0.1478,
      "reward": 0.6112714111804962,
      "reward_std": 0.1831095963716507,
      "reward_std_all": 0.18770389258861542,
      "rewards-max/forensics_sinkhorn_reward": 0.9387769401073456,
      "rewards-max/forensics_strict_edge_reward": 0.8241345286369324,
      "rewards-max/forensics_tiou_reward": 0.8807839751243591,
      "rewards/forensics_sinkhorn_reward": 0.48155684769153595,
      "rewards/forensics_strict_edge_reward": 0.17451678216457367,
      "rewards/forensics_tiou_reward": 0.20995156466960907,
      "step": 770
    },
    {
      "completion_length": 29.96875,
      "epoch": 1.6136125654450262,
      "grad_norm": 43.1091423034668,
      "kl": 4.34375,
      "learning_rate": 5.964360587002096e-07,
      "loss": 0.4812,
      "reward": 0.6635499894618988,
      "reward_std": 0.1394205540418625,
      "reward_std_all": 0.14185039699077606,
      "rewards-max/forensics_sinkhorn_reward": 0.9618923962116241,
      "rewards-max/forensics_strict_edge_reward": 0.8968285918235779,
      "rewards-max/forensics_tiou_reward": 0.9398371577262878,
      "rewards/forensics_sinkhorn_reward": 0.5615809559822083,
      "rewards/forensics_strict_edge_reward": 0.33898888528347015,
      "rewards/forensics_tiou_reward": 0.42476315796375275,
      "step": 771
    },
    {
      "completion_length": 30.28125,
      "epoch": 1.6157068062827227,
      "grad_norm": 29.555383682250977,
      "kl": 2.640625,
      "learning_rate": 5.959119496855346e-07,
      "loss": 0.3383,
      "reward": 0.639605700969696,
      "reward_std": 0.15758450329303741,
      "reward_std_all": 0.17022079229354858,
      "rewards-max/forensics_sinkhorn_reward": 0.9883168041706085,
      "rewards-max/forensics_strict_edge_reward": 0.9256840348243713,
      "rewards-max/forensics_tiou_reward": 0.9973464012145996,
      "rewards/forensics_sinkhorn_reward": 0.5752535760402679,
      "rewards/forensics_strict_edge_reward": 0.2741789072751999,
      "rewards/forensics_tiou_reward": 0.3847055733203888,
      "step": 772
    },
    {
      "completion_length": 19.9375,
      "epoch": 1.6178010471204187,
      "grad_norm": 38.26924514770508,
      "kl": 1.8046875,
      "learning_rate": 5.953878406708596e-07,
      "loss": 0.2412,
      "reward": 0.7112389504909515,
      "reward_std": 0.07492753863334656,
      "reward_std_all": 0.08739673718810081,
      "rewards-max/forensics_sinkhorn_reward": 0.9748876988887787,
      "rewards-max/forensics_strict_edge_reward": 0.9464850425720215,
      "rewards-max/forensics_tiou_reward": 0.9677963554859161,
      "rewards/forensics_sinkhorn_reward": 0.6301042139530182,
      "rewards/forensics_strict_edge_reward": 0.3222511410713196,
      "rewards/forensics_tiou_reward": 0.42123666405677795,
      "step": 773
    },
    {
      "completion_length": 35.0625,
      "epoch": 1.6198952879581152,
      "grad_norm": 21.540496826171875,
      "kl": 3.34375,
      "learning_rate": 5.948637316561844e-07,
      "loss": 0.2372,
      "reward": 0.6320281624794006,
      "reward_std": 0.16731205582618713,
      "reward_std_all": 0.1680442988872528,
      "rewards-max/forensics_sinkhorn_reward": 0.9572674632072449,
      "rewards-max/forensics_strict_edge_reward": 0.6496154218912125,
      "rewards-max/forensics_tiou_reward": 0.9289334416389465,
      "rewards/forensics_sinkhorn_reward": 0.4357094466686249,
      "rewards/forensics_strict_edge_reward": 0.21614714711904526,
      "rewards/forensics_tiou_reward": 0.23756076395511627,
      "step": 774
    },
    {
      "completion_length": 24.8125,
      "epoch": 1.6219895287958115,
      "grad_norm": 30.49280548095703,
      "kl": 2.5859375,
      "learning_rate": 5.943396226415094e-07,
      "loss": 0.2184,
      "reward": 0.6751098036766052,
      "reward_std": 0.14749164879322052,
      "reward_std_all": 0.14273419231176376,
      "rewards-max/forensics_sinkhorn_reward": 0.983616977930069,
      "rewards-max/forensics_strict_edge_reward": 0.8983768224716187,
      "rewards-max/forensics_tiou_reward": 0.9790395200252533,
      "rewards/forensics_sinkhorn_reward": 0.5861448645591736,
      "rewards/forensics_strict_edge_reward": 0.330244243144989,
      "rewards/forensics_tiou_reward": 0.39479750394821167,
      "step": 775
    },
    {
      "completion_length": 30.40625,
      "epoch": 1.6240837696335078,
      "grad_norm": 20.561065673828125,
      "kl": 3.640625,
      "learning_rate": 5.938155136268343e-07,
      "loss": 0.3851,
      "reward": 0.6930391192436218,
      "reward_std": 0.10813611373305321,
      "reward_std_all": 0.1109968014061451,
      "rewards-max/forensics_sinkhorn_reward": 0.9873959422111511,
      "rewards-max/forensics_strict_edge_reward": 0.9565287232398987,
      "rewards-max/forensics_tiou_reward": 0.9937245547771454,
      "rewards/forensics_sinkhorn_reward": 0.589040219783783,
      "rewards/forensics_strict_edge_reward": 0.3338116407394409,
      "rewards/forensics_tiou_reward": 0.3903975188732147,
      "step": 776
    },
    {
      "completion_length": 39.609375,
      "epoch": 1.6261780104712042,
      "grad_norm": 21.690608978271484,
      "kl": 5.0859375,
      "learning_rate": 5.932914046121593e-07,
      "loss": 0.3971,
      "reward": 0.6182042956352234,
      "reward_std": 0.165343277156353,
      "reward_std_all": 0.17235764861106873,
      "rewards-max/forensics_sinkhorn_reward": 0.9875809550285339,
      "rewards-max/forensics_strict_edge_reward": 0.9574736058712006,
      "rewards-max/forensics_tiou_reward": 0.994459867477417,
      "rewards/forensics_sinkhorn_reward": 0.46599917113780975,
      "rewards/forensics_strict_edge_reward": 0.21899351477622986,
      "rewards/forensics_tiou_reward": 0.25672387331724167,
      "step": 777
    },
    {
      "completion_length": 27.109375,
      "epoch": 1.6282722513089005,
      "grad_norm": 21.188697814941406,
      "kl": 3.1015625,
      "learning_rate": 5.927672955974842e-07,
      "loss": 0.2141,
      "reward": 0.5967498123645782,
      "reward_std": 0.1596216894686222,
      "reward_std_all": 0.17944984883069992,
      "rewards-max/forensics_sinkhorn_reward": 0.9729367196559906,
      "rewards-max/forensics_strict_edge_reward": 0.8918541073799133,
      "rewards-max/forensics_tiou_reward": 0.9627667367458344,
      "rewards/forensics_sinkhorn_reward": 0.4932311624288559,
      "rewards/forensics_strict_edge_reward": 0.24715574085712433,
      "rewards/forensics_tiou_reward": 0.2986658737063408,
      "step": 778
    },
    {
      "completion_length": 35.46875,
      "epoch": 1.6303664921465968,
      "grad_norm": 24.764986038208008,
      "kl": 4.68359375,
      "learning_rate": 5.922431865828093e-07,
      "loss": 0.3411,
      "reward": 0.6512329876422882,
      "reward_std": 0.09824598580598831,
      "reward_std_all": 0.12309402599930763,
      "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": 0.5691537261009216,
      "rewards/forensics_strict_edge_reward": 0.3120700567960739,
      "rewards/forensics_tiou_reward": 0.37924520671367645,
      "step": 779
    },
    {
      "completion_length": 29.15625,
      "epoch": 1.6324607329842933,
      "grad_norm": 24.967741012573242,
      "kl": 3.9609375,
      "learning_rate": 5.917190775681341e-07,
      "loss": 0.249,
      "reward": 0.6133428812026978,
      "reward_std": 0.15207809209823608,
      "reward_std_all": 0.16870474815368652,
      "rewards-max/forensics_sinkhorn_reward": 0.9866122007369995,
      "rewards-max/forensics_strict_edge_reward": 0.9464850425720215,
      "rewards-max/forensics_tiou_reward": 0.990659236907959,
      "rewards/forensics_sinkhorn_reward": 0.5492790043354034,
      "rewards/forensics_strict_edge_reward": 0.3025241941213608,
      "rewards/forensics_tiou_reward": 0.35345979034900665,
      "step": 780
    },
    {
      "completion_length": 37.09375,
      "epoch": 1.6345549738219896,
      "grad_norm": 29.305809020996094,
      "kl": 5.71875,
      "learning_rate": 5.911949685534591e-07,
      "loss": 0.3886,
      "reward": 0.648542046546936,
      "reward_std": 0.19571006298065186,
      "reward_std_all": 0.18115083873271942,
      "rewards-max/forensics_sinkhorn_reward": 0.978526771068573,
      "rewards-max/forensics_strict_edge_reward": 0.8663162291049957,
      "rewards-max/forensics_tiou_reward": 0.9760573208332062,
      "rewards/forensics_sinkhorn_reward": 0.5181841850280762,
      "rewards/forensics_strict_edge_reward": 0.3092679977416992,
      "rewards/forensics_tiou_reward": 0.3622045964002609,
      "step": 781
    },
    {
      "completion_length": 47.609375,
      "epoch": 1.6366492146596858,
      "grad_norm": 34.385555267333984,
      "kl": 8.234375,
      "learning_rate": 5.906708595387841e-07,
      "loss": 0.4851,
      "reward": 0.5937764048576355,
      "reward_std": 0.19477389007806778,
      "reward_std_all": 0.20776593685150146,
      "rewards-max/forensics_sinkhorn_reward": 0.8948068022727966,
      "rewards-max/forensics_strict_edge_reward": 0.7449297606945038,
      "rewards-max/forensics_tiou_reward": 0.802815854549408,
      "rewards/forensics_sinkhorn_reward": 0.44323769211769104,
      "rewards/forensics_strict_edge_reward": 0.23487430065870285,
      "rewards/forensics_tiou_reward": 0.2756159156560898,
      "step": 782
    },
    {
      "completion_length": 39.453125,
      "epoch": 1.6387434554973823,
      "grad_norm": 28.02089500427246,
      "kl": 5.4453125,
      "learning_rate": 5.90146750524109e-07,
      "loss": 0.3378,
      "reward": 0.6211130619049072,
      "reward_std": 0.17538489401340485,
      "reward_std_all": 0.18475685268640518,
      "rewards-max/forensics_sinkhorn_reward": 0.9830209016799927,
      "rewards-max/forensics_strict_edge_reward": 0.9567649364471436,
      "rewards-max/forensics_tiou_reward": 0.9855949282646179,
      "rewards/forensics_sinkhorn_reward": 0.5284520387649536,
      "rewards/forensics_strict_edge_reward": 0.22651223465800285,
      "rewards/forensics_tiou_reward": 0.3578581064939499,
      "step": 783
    },
    {
      "completion_length": 43.53125,
      "epoch": 1.6408376963350786,
      "grad_norm": 24.954496383666992,
      "kl": 4.015625,
      "learning_rate": 5.89622641509434e-07,
      "loss": 0.1445,
      "reward": 0.6265533566474915,
      "reward_std": 0.13293840736150742,
      "reward_std_all": 0.14898681640625,
      "rewards-max/forensics_sinkhorn_reward": 0.9835985898971558,
      "rewards-max/forensics_strict_edge_reward": 0.8137359023094177,
      "rewards-max/forensics_tiou_reward": 0.9788898229598999,
      "rewards/forensics_sinkhorn_reward": 0.5289206504821777,
      "rewards/forensics_strict_edge_reward": 0.16118347644805908,
      "rewards/forensics_tiou_reward": 0.2766062468290329,
      "step": 784
    },
    {
      "completion_length": 29.5,
      "epoch": 1.6429319371727749,
      "grad_norm": 24.168331146240234,
      "kl": 5.25,
      "learning_rate": 5.890985324947589e-07,
      "loss": 0.3685,
      "reward": 0.6314634084701538,
      "reward_std": 0.17699525505304337,
      "reward_std_all": 0.17698224633932114,
      "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": 0.5738109052181244,
      "rewards/forensics_strict_edge_reward": 0.2995377480983734,
      "rewards/forensics_tiou_reward": 0.3854622542858124,
      "step": 785
    },
    {
      "completion_length": 32.25,
      "epoch": 1.6450261780104714,
      "grad_norm": 28.033443450927734,
      "kl": 3.3515625,
      "learning_rate": 5.885744234800838e-07,
      "loss": 0.1987,
      "reward": 0.6810809671878815,
      "reward_std": 0.1251366250216961,
      "reward_std_all": 0.12878679111599922,
      "rewards-max/forensics_sinkhorn_reward": 0.9877981245517731,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9953023791313171,
      "rewards/forensics_sinkhorn_reward": 0.6157416701316833,
      "rewards/forensics_strict_edge_reward": 0.31757573783397675,
      "rewards/forensics_tiou_reward": 0.42260900139808655,
      "step": 786
    },
    {
      "completion_length": 43.984375,
      "epoch": 1.6471204188481674,
      "grad_norm": 28.862852096557617,
      "kl": 6.40625,
      "learning_rate": 5.880503144654087e-07,
      "loss": 0.3828,
      "reward": 0.578849732875824,
      "reward_std": 0.18861304223537445,
      "reward_std_all": 0.20624907314777374,
      "rewards-max/forensics_sinkhorn_reward": 0.8303841650485992,
      "rewards-max/forensics_strict_edge_reward": 0.787904292345047,
      "rewards-max/forensics_tiou_reward": 0.7240915894508362,
      "rewards/forensics_sinkhorn_reward": 0.3940947502851486,
      "rewards/forensics_strict_edge_reward": 0.19701150059700012,
      "rewards/forensics_tiou_reward": 0.19725903868675232,
      "step": 787
    },
    {
      "completion_length": 45.359375,
      "epoch": 1.649214659685864,
      "grad_norm": 40.73435974121094,
      "kl": 6.90625,
      "learning_rate": 5.875262054507337e-07,
      "loss": 0.4386,
      "reward": 0.5857326984405518,
      "reward_std": 0.21137945353984833,
      "reward_std_all": 0.21509233117103577,
      "rewards-max/forensics_sinkhorn_reward": 0.9864841103553772,
      "rewards-max/forensics_strict_edge_reward": 0.9364311099052429,
      "rewards-max/forensics_tiou_reward": 0.9901230037212372,
      "rewards/forensics_sinkhorn_reward": 0.4613238722085953,
      "rewards/forensics_strict_edge_reward": 0.2939680367708206,
      "rewards/forensics_tiou_reward": 0.3299078792333603,
      "step": 788
    },
    {
      "completion_length": 41.1875,
      "epoch": 1.6513089005235602,
      "grad_norm": 20.34398078918457,
      "kl": 3.859375,
      "learning_rate": 5.870020964360587e-07,
      "loss": 0.2576,
      "reward": 0.7043662369251251,
      "reward_std": 0.11449456214904785,
      "reward_std_all": 0.11130944266915321,
      "rewards-max/forensics_sinkhorn_reward": 0.9885258972644806,
      "rewards-max/forensics_strict_edge_reward": 0.9890114367008209,
      "rewards-max/forensics_tiou_reward": 0.9981701672077179,
      "rewards/forensics_sinkhorn_reward": 0.6442719101905823,
      "rewards/forensics_strict_edge_reward": 0.4163445383310318,
      "rewards/forensics_tiou_reward": 0.519516795873642,
      "step": 789
    },
    {
      "completion_length": 50.953125,
      "epoch": 1.6534031413612564,
      "grad_norm": 31.523269653320312,
      "kl": 4.7421875,
      "learning_rate": 5.864779874213836e-07,
      "loss": 0.4023,
      "reward": 0.6086545288562775,
      "reward_std": 0.212920181453228,
      "reward_std_all": 0.209384523332119,
      "rewards-max/forensics_sinkhorn_reward": 0.9884852766990662,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9980084598064423,
      "rewards/forensics_sinkhorn_reward": 0.5730360597372055,
      "rewards/forensics_strict_edge_reward": 0.3441467136144638,
      "rewards/forensics_tiou_reward": 0.41882088780403137,
      "step": 790
    },
    {
      "completion_length": 35.328125,
      "epoch": 1.655497382198953,
      "grad_norm": 26.44875144958496,
      "kl": 3.7734375,
      "learning_rate": 5.859538784067087e-07,
      "loss": 0.2325,
      "reward": 0.6279251575469971,
      "reward_std": 0.13721344992518425,
      "reward_std_all": 0.15271474421024323,
      "rewards-max/forensics_sinkhorn_reward": 0.9806980788707733,
      "rewards-max/forensics_strict_edge_reward": 0.8472771048545837,
      "rewards-max/forensics_tiou_reward": 0.9805607497692108,
      "rewards/forensics_sinkhorn_reward": 0.5119377672672272,
      "rewards/forensics_strict_edge_reward": 0.25135011970996857,
      "rewards/forensics_tiou_reward": 0.30593380331993103,
      "step": 791
    },
    {
      "completion_length": 40.796875,
      "epoch": 1.6575916230366492,
      "grad_norm": 25.086078643798828,
      "kl": 3.0078125,
      "learning_rate": 5.854297693920335e-07,
      "loss": 0.2676,
      "reward": 0.5902216136455536,
      "reward_std": 0.1597120426595211,
      "reward_std_all": 0.17700336128473282,
      "rewards-max/forensics_sinkhorn_reward": 0.923370748758316,
      "rewards-max/forensics_strict_edge_reward": 0.8078656196594238,
      "rewards-max/forensics_tiou_reward": 0.8493809103965759,
      "rewards/forensics_sinkhorn_reward": 0.43955834209918976,
      "rewards/forensics_strict_edge_reward": 0.24445822089910507,
      "rewards/forensics_tiou_reward": 0.25267135724425316,
      "step": 792
    },
    {
      "completion_length": 42.96875,
      "epoch": 1.6596858638743455,
      "grad_norm": 28.132904052734375,
      "kl": 4.828125,
      "learning_rate": 5.849056603773585e-07,
      "loss": 0.3079,
      "reward": 0.6247266530990601,
      "reward_std": 0.16797290742397308,
      "reward_std_all": 0.17352233827114105,
      "rewards-max/forensics_sinkhorn_reward": 0.8899435102939606,
      "rewards-max/forensics_strict_edge_reward": 0.7657788097858429,
      "rewards-max/forensics_tiou_reward": 0.7994090616703033,
      "rewards/forensics_sinkhorn_reward": 0.5175371468067169,
      "rewards/forensics_strict_edge_reward": 0.22082903236150742,
      "rewards/forensics_tiou_reward": 0.326199971139431,
      "step": 793
    },
    {
      "completion_length": 30.515625,
      "epoch": 1.661780104712042,
      "grad_norm": 22.147226333618164,
      "kl": 3.703125,
      "learning_rate": 5.843815513626834e-07,
      "loss": 0.3495,
      "reward": 0.6382522881031036,
      "reward_std": 0.15584773942828178,
      "reward_std_all": 0.16418632119894028,
      "rewards-max/forensics_sinkhorn_reward": 0.9594854712486267,
      "rewards-max/forensics_strict_edge_reward": 0.8751733303070068,
      "rewards-max/forensics_tiou_reward": 0.9352993667125702,
      "rewards/forensics_sinkhorn_reward": 0.5066320300102234,
      "rewards/forensics_strict_edge_reward": 0.24665986001491547,
      "rewards/forensics_tiou_reward": 0.3196435272693634,
      "step": 794
    },
    {
      "completion_length": 30.59375,
      "epoch": 1.6638743455497382,
      "grad_norm": 30.201278686523438,
      "kl": 5.0,
      "learning_rate": 5.838574423480084e-07,
      "loss": 0.3582,
      "reward": 0.578123152256012,
      "reward_std": 0.20812192559242249,
      "reward_std_all": 0.21322280168533325,
      "rewards-max/forensics_sinkhorn_reward": 0.855248898267746,
      "rewards-max/forensics_strict_edge_reward": 0.6874399483203888,
      "rewards-max/forensics_tiou_reward": 0.7184737324714661,
      "rewards/forensics_sinkhorn_reward": 0.4376166760921478,
      "rewards/forensics_strict_edge_reward": 0.13732050359249115,
      "rewards/forensics_tiou_reward": 0.21859540045261383,
      "step": 795
    },
    {
      "completion_length": 32.21875,
      "epoch": 1.6659685863874345,
      "grad_norm": 14.22161865234375,
      "kl": 3.2109375,
      "learning_rate": 5.833333333333334e-07,
      "loss": 0.2584,
      "reward": 0.7231225371360779,
      "reward_std": 0.08892474323511124,
      "reward_std_all": 0.0863165482878685,
      "rewards-max/forensics_sinkhorn_reward": 0.988743245601654,
      "rewards-max/forensics_strict_edge_reward": 0.9782643616199493,
      "rewards-max/forensics_tiou_reward": 0.999025821685791,
      "rewards/forensics_sinkhorn_reward": 0.6623509228229523,
      "rewards/forensics_strict_edge_reward": 0.46342289447784424,
      "rewards/forensics_tiou_reward": 0.5210943073034286,
      "step": 796
    },
    {
      "completion_length": 26.046875,
      "epoch": 1.668062827225131,
      "grad_norm": 29.421527862548828,
      "kl": 3.2109375,
      "learning_rate": 5.828092243186582e-07,
      "loss": 0.3228,
      "reward": 0.6608780324459076,
      "reward_std": 0.17011215537786484,
      "reward_std_all": 0.1664063185453415,
      "rewards-max/forensics_sinkhorn_reward": 0.9884390234947205,
      "rewards-max/forensics_strict_edge_reward": 0.9782643616199493,
      "rewards-max/forensics_tiou_reward": 0.997826099395752,
      "rewards/forensics_sinkhorn_reward": 0.5717744529247284,
      "rewards/forensics_strict_edge_reward": 0.34708237648010254,
      "rewards/forensics_tiou_reward": 0.39245641231536865,
      "step": 797
    },
    {
      "completion_length": 29.140625,
      "epoch": 1.6701570680628273,
      "grad_norm": 19.03471565246582,
      "kl": 3.2890625,
      "learning_rate": 5.822851153039832e-07,
      "loss": 0.1334,
      "reward": 0.6888573467731476,
      "reward_std": 0.1341002732515335,
      "reward_std_all": 0.12770496308803558,
      "rewards-max/forensics_sinkhorn_reward": 0.9878038465976715,
      "rewards-max/forensics_strict_edge_reward": 0.946017861366272,
      "rewards-max/forensics_tiou_reward": 0.9953282177448273,
      "rewards/forensics_sinkhorn_reward": 0.6108681261539459,
      "rewards/forensics_strict_edge_reward": 0.3650358319282532,
      "rewards/forensics_tiou_reward": 0.41589125990867615,
      "step": 798
    },
    {
      "completion_length": 34.8125,
      "epoch": 1.6722513089005235,
      "grad_norm": 25.436580657958984,
      "kl": 4.9140625,
      "learning_rate": 5.817610062893081e-07,
      "loss": 0.3299,
      "reward": 0.6565360128879547,
      "reward_std": 0.14900464192032814,
      "reward_std_all": 0.15439357236027718,
      "rewards-max/forensics_sinkhorn_reward": 0.9835855066776276,
      "rewards-max/forensics_strict_edge_reward": 0.9149472117424011,
      "rewards-max/forensics_tiou_reward": 0.984502524137497,
      "rewards/forensics_sinkhorn_reward": 0.5539811849594116,
      "rewards/forensics_strict_edge_reward": 0.27535563707351685,
      "rewards/forensics_tiou_reward": 0.3495451509952545,
      "step": 799
    },
    {
      "completion_length": 40.515625,
      "epoch": 1.67434554973822,
      "grad_norm": 26.188486099243164,
      "kl": 6.5625,
      "learning_rate": 5.812368972746331e-07,
      "loss": 0.4522,
      "reward": 0.641694188117981,
      "reward_std": 0.12822487950325012,
      "reward_std_all": 0.1400880292057991,
      "rewards-max/forensics_sinkhorn_reward": 0.9652774930000305,
      "rewards-max/forensics_strict_edge_reward": 0.7919523417949677,
      "rewards-max/forensics_tiou_reward": 0.945432722568512,
      "rewards/forensics_sinkhorn_reward": 0.5320059359073639,
      "rewards/forensics_strict_edge_reward": 0.2693034037947655,
      "rewards/forensics_tiou_reward": 0.3415272980928421,
      "step": 800
    },
    {
      "completion_length": 35.03125,
      "epoch": 1.676439790575916,
      "grad_norm": 26.211336135864258,
      "kl": 4.7421875,
      "learning_rate": 5.80712788259958e-07,
      "loss": 0.2598,
      "reward": 0.5810415744781494,
      "reward_std": 0.15068981051445007,
      "reward_std_all": 0.1827547326683998,
      "rewards-max/forensics_sinkhorn_reward": 0.9866982996463776,
      "rewards-max/forensics_strict_edge_reward": 0.9357379674911499,
      "rewards-max/forensics_tiou_reward": 0.990974485874176,
      "rewards/forensics_sinkhorn_reward": 0.5272149443626404,
      "rewards/forensics_strict_edge_reward": 0.2707747370004654,
      "rewards/forensics_tiou_reward": 0.33237241208553314,
      "step": 801
    },
    {
      "completion_length": 37.421875,
      "epoch": 1.6785340314136126,
      "grad_norm": 25.69233512878418,
      "kl": 3.859375,
      "learning_rate": 5.80188679245283e-07,
      "loss": 0.3008,
      "reward": 0.6750258505344391,
      "reward_std": 0.09162227436900139,
      "reward_std_all": 0.11584089696407318,
      "rewards-max/forensics_sinkhorn_reward": 0.8880696892738342,
      "rewards-max/forensics_strict_edge_reward": 0.8184211552143097,
      "rewards-max/forensics_tiou_reward": 0.8569404482841492,
      "rewards/forensics_sinkhorn_reward": 0.534759908914566,
      "rewards/forensics_strict_edge_reward": 0.28157536685466766,
      "rewards/forensics_tiou_reward": 0.3268285095691681,
      "step": 802
    },
    {
      "completion_length": 51.625,
      "epoch": 1.680628272251309,
      "grad_norm": 74.89118957519531,
      "kl": 7.0,
      "learning_rate": 5.796645702306078e-07,
      "loss": 0.3727,
      "reward": 0.5694229602813721,
      "reward_std": 0.15262073278427124,
      "reward_std_all": 0.18636822700500488,
      "rewards-max/forensics_sinkhorn_reward": 0.9877617061138153,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9951571226119995,
      "rewards/forensics_sinkhorn_reward": 0.42496080696582794,
      "rewards/forensics_strict_edge_reward": 0.10290485620498657,
      "rewards/forensics_tiou_reward": 0.17589056491851807,
      "step": 803
    },
    {
      "completion_length": 54.859375,
      "epoch": 1.6827225130890051,
      "grad_norm": 33.51113510131836,
      "kl": 7.046875,
      "learning_rate": 5.791404612159329e-07,
      "loss": 0.5981,
      "reward": 0.600910872220993,
      "reward_std": 0.13163954764604568,
      "reward_std_all": 0.15961919724941254,
      "rewards-max/forensics_sinkhorn_reward": 0.9806618988513947,
      "rewards-max/forensics_strict_edge_reward": 0.9782643616199493,
      "rewards-max/forensics_tiou_reward": 0.9815152883529663,
      "rewards/forensics_sinkhorn_reward": 0.4846518635749817,
      "rewards/forensics_strict_edge_reward": 0.21390970051288605,
      "rewards/forensics_tiou_reward": 0.2556379437446594,
      "step": 804
    },
    {
      "completion_length": 38.765625,
      "epoch": 1.6848167539267016,
      "grad_norm": 20.35226058959961,
      "kl": 4.15625,
      "learning_rate": 5.786163522012579e-07,
      "loss": 0.3722,
      "reward": 0.5847859382629395,
      "reward_std": 0.19127333909273148,
      "reward_std_all": 0.20357467979192734,
      "rewards-max/forensics_sinkhorn_reward": 0.9258936643600464,
      "rewards-max/forensics_strict_edge_reward": 0.7727466821670532,
      "rewards-max/forensics_tiou_reward": 0.8611004054546356,
      "rewards/forensics_sinkhorn_reward": 0.4538683444261551,
      "rewards/forensics_strict_edge_reward": 0.2385043352842331,
      "rewards/forensics_tiou_reward": 0.2559794485569,
      "step": 805
    },
    {
      "completion_length": 45.6875,
      "epoch": 1.6869109947643979,
      "grad_norm": 35.634952545166016,
      "kl": 7.3125,
      "learning_rate": 5.780922431865828e-07,
      "loss": 0.4616,
      "reward": 0.6000767946243286,
      "reward_std": 0.168451726436615,
      "reward_std_all": 0.18584153056144714,
      "rewards-max/forensics_sinkhorn_reward": 0.9028123617172241,
      "rewards-max/forensics_strict_edge_reward": 0.8610626757144928,
      "rewards-max/forensics_tiou_reward": 0.8219926059246063,
      "rewards/forensics_sinkhorn_reward": 0.4533984363079071,
      "rewards/forensics_strict_edge_reward": 0.22335078567266464,
      "rewards/forensics_tiou_reward": 0.24777784198522568,
      "step": 806
    },
    {
      "completion_length": 66.65625,
      "epoch": 1.6890052356020941,
      "grad_norm": 26.661287307739258,
      "kl": 7.1875,
      "learning_rate": 5.775681341719078e-07,
      "loss": 0.5248,
      "reward": 0.6583670079708099,
      "reward_std": 0.14367378503084183,
      "reward_std_all": 0.1427752748131752,
      "rewards-max/forensics_sinkhorn_reward": 0.988856703042984,
      "rewards-max/forensics_strict_edge_reward": 0.9890114367008209,
      "rewards-max/forensics_tiou_reward": 0.9994727075099945,
      "rewards/forensics_sinkhorn_reward": 0.5999474823474884,
      "rewards/forensics_strict_edge_reward": 0.3177037462592125,
      "rewards/forensics_tiou_reward": 0.4402129054069519,
      "step": 807
    },
    {
      "completion_length": 56.5625,
      "epoch": 1.6910994764397906,
      "grad_norm": 19.581573486328125,
      "kl": 6.015625,
      "learning_rate": 5.770440251572327e-07,
      "loss": 0.4276,
      "reward": 0.6458808481693268,
      "reward_std": 0.17329254746437073,
      "reward_std_all": 0.16540924459695816,
      "rewards-max/forensics_sinkhorn_reward": 0.9879795610904694,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9960180819034576,
      "rewards/forensics_sinkhorn_reward": 0.572122722864151,
      "rewards/forensics_strict_edge_reward": 0.37163616716861725,
      "rewards/forensics_tiou_reward": 0.41558191180229187,
      "step": 808
    },
    {
      "completion_length": 39.4375,
      "epoch": 1.693193717277487,
      "grad_norm": 23.343015670776367,
      "kl": 5.171875,
      "learning_rate": 5.765199161425576e-07,
      "loss": 0.3774,
      "reward": 0.6330270171165466,
      "reward_std": 0.17884717881679535,
      "reward_std_all": 0.1814000904560089,
      "rewards-max/forensics_sinkhorn_reward": 0.9866040050983429,
      "rewards-max/forensics_strict_edge_reward": 0.9169811904430389,
      "rewards-max/forensics_tiou_reward": 0.9906341135501862,
      "rewards/forensics_sinkhorn_reward": 0.5584091246128082,
      "rewards/forensics_strict_edge_reward": 0.3618175983428955,
      "rewards/forensics_tiou_reward": 0.4172467291355133,
      "step": 809
    },
    {
      "completion_length": 41.75,
      "epoch": 1.6952879581151832,
      "grad_norm": 38.47945785522461,
      "kl": 4.6328125,
      "learning_rate": 5.759958071278825e-07,
      "loss": 0.3784,
      "reward": 0.6753805875778198,
      "reward_std": 0.14417172223329544,
      "reward_std_all": 0.13699636235833168,
      "rewards-max/forensics_sinkhorn_reward": 0.9610951542854309,
      "rewards-max/forensics_strict_edge_reward": 0.8991738557815552,
      "rewards-max/forensics_tiou_reward": 0.9411786794662476,
      "rewards/forensics_sinkhorn_reward": 0.5230432450771332,
      "rewards/forensics_strict_edge_reward": 0.3468112051486969,
      "rewards/forensics_tiou_reward": 0.3797309845685959,
      "step": 810
    },
    {
      "completion_length": 49.4375,
      "epoch": 1.6973821989528797,
      "grad_norm": 24.610450744628906,
      "kl": 6.25,
      "learning_rate": 5.754716981132075e-07,
      "loss": 0.4983,
      "reward": 0.6224912405014038,
      "reward_std": 0.17504888027906418,
      "reward_std_all": 0.18210823088884354,
      "rewards-max/forensics_sinkhorn_reward": 0.9875161349773407,
      "rewards-max/forensics_strict_edge_reward": 0.9565287232398987,
      "rewards-max/forensics_tiou_reward": 0.9941957294940948,
      "rewards/forensics_sinkhorn_reward": 0.5151808112859726,
      "rewards/forensics_strict_edge_reward": 0.3225908428430557,
      "rewards/forensics_tiou_reward": 0.3935931921005249,
      "step": 811
    },
    {
      "completion_length": 53.140625,
      "epoch": 1.699476439790576,
      "grad_norm": 21.657072067260742,
      "kl": 5.234375,
      "learning_rate": 5.749475890985325e-07,
      "loss": 0.3566,
      "reward": 0.5965534448623657,
      "reward_std": 0.19371426105499268,
      "reward_std_all": 0.19957129657268524,
      "rewards-max/forensics_sinkhorn_reward": 0.9476569890975952,
      "rewards-max/forensics_strict_edge_reward": 0.9151731729507446,
      "rewards-max/forensics_tiou_reward": 0.9114572703838348,
      "rewards/forensics_sinkhorn_reward": 0.4418504238128662,
      "rewards/forensics_strict_edge_reward": 0.20528457313776016,
      "rewards/forensics_tiou_reward": 0.25493529438972473,
      "step": 812
    },
    {
      "completion_length": 30.46875,
      "epoch": 1.7015706806282722,
      "grad_norm": 16.110063552856445,
      "kl": 2.609375,
      "learning_rate": 5.744234800838574e-07,
      "loss": 0.2358,
      "reward": 0.6920002698898315,
      "reward_std": 0.13348305225372314,
      "reward_std_all": 0.12632297351956367,
      "rewards-max/forensics_sinkhorn_reward": 0.9867070317268372,
      "rewards-max/forensics_strict_edge_reward": 0.946017861366272,
      "rewards-max/forensics_tiou_reward": 0.9910354316234589,
      "rewards/forensics_sinkhorn_reward": 0.5927329659461975,
      "rewards/forensics_strict_edge_reward": 0.3445189744234085,
      "rewards/forensics_tiou_reward": 0.44965530931949615,
      "step": 813
    },
    {
      "completion_length": 48.796875,
      "epoch": 1.7036649214659687,
      "grad_norm": 26.33523941040039,
      "kl": 6.0,
      "learning_rate": 5.738993710691824e-07,
      "loss": 0.4138,
      "reward": 0.6424267590045929,
      "reward_std": 0.18885723501443863,
      "reward_std_all": 0.17773813754320145,
      "rewards-max/forensics_sinkhorn_reward": 0.9555442333221436,
      "rewards-max/forensics_strict_edge_reward": 0.8753803670406342,
      "rewards-max/forensics_tiou_reward": 0.922284871339798,
      "rewards/forensics_sinkhorn_reward": 0.5222964584827423,
      "rewards/forensics_strict_edge_reward": 0.32474927604198456,
      "rewards/forensics_tiou_reward": 0.3324471563100815,
      "step": 814
    },
    {
      "completion_length": 54.90625,
      "epoch": 1.7057591623036648,
      "grad_norm": 26.032363891601562,
      "kl": 5.0546875,
      "learning_rate": 5.733752620545072e-07,
      "loss": 0.3616,
      "reward": 0.6610211133956909,
      "reward_std": 0.162299282848835,
      "reward_std_all": 0.15897145867347717,
      "rewards-max/forensics_sinkhorn_reward": 0.9867839515209198,
      "rewards-max/forensics_strict_edge_reward": 0.9574736058712006,
      "rewards-max/forensics_tiou_reward": 0.991304337978363,
      "rewards/forensics_sinkhorn_reward": 0.5328651666641235,
      "rewards/forensics_strict_edge_reward": 0.31547272205352783,
      "rewards/forensics_tiou_reward": 0.372758150100708,
      "step": 815
    },
    {
      "completion_length": 30.234375,
      "epoch": 1.7078534031413612,
      "grad_norm": 19.242578506469727,
      "kl": 3.0859375,
      "learning_rate": 5.728511530398323e-07,
      "loss": 0.3104,
      "reward": 0.7121360898017883,
      "reward_std": 0.12374109402298927,
      "reward_std_all": 0.1147426888346672,
      "rewards-max/forensics_sinkhorn_reward": 0.9884204268455505,
      "rewards-max/forensics_strict_edge_reward": 0.9890114367008209,
      "rewards-max/forensics_tiou_reward": 0.9977528154850006,
      "rewards/forensics_sinkhorn_reward": 0.6561165452003479,
      "rewards/forensics_strict_edge_reward": 0.4020916074514389,
      "rewards/forensics_tiou_reward": 0.49152129888534546,
      "step": 816
    },
    {
      "completion_length": 44.59375,
      "epoch": 1.7099476439790577,
      "grad_norm": 18.659391403198242,
      "kl": 3.5703125,
      "learning_rate": 5.723270440251572e-07,
      "loss": 0.2865,
      "reward": 0.6726945638656616,
      "reward_std": 0.14936015754938126,
      "reward_std_all": 0.14245647564530373,
      "rewards-max/forensics_sinkhorn_reward": 0.9869434833526611,
      "rewards-max/forensics_strict_edge_reward": 0.9075756669044495,
      "rewards-max/forensics_tiou_reward": 0.991931825876236,
      "rewards/forensics_sinkhorn_reward": 0.5785684287548065,
      "rewards/forensics_strict_edge_reward": 0.3017326071858406,
      "rewards/forensics_tiou_reward": 0.4313415288925171,
      "step": 817
    },
    {
      "completion_length": 46.0625,
      "epoch": 1.7120418848167538,
      "grad_norm": 44.1705436706543,
      "kl": 4.359375,
      "learning_rate": 5.718029350104822e-07,
      "loss": 0.2727,
      "reward": 0.6147055923938751,
      "reward_std": 0.20671705901622772,
      "reward_std_all": 0.19984818249940872,
      "rewards-max/forensics_sinkhorn_reward": 0.9737487435340881,
      "rewards-max/forensics_strict_edge_reward": 0.810933917760849,
      "rewards-max/forensics_tiou_reward": 0.9568207561969757,
      "rewards/forensics_sinkhorn_reward": 0.48076875507831573,
      "rewards/forensics_strict_edge_reward": 0.1976134479045868,
      "rewards/forensics_tiou_reward": 0.23732871562242508,
      "step": 818
    },
    {
      "completion_length": 90.328125,
      "epoch": 1.7141361256544503,
      "grad_norm": 45.58772659301758,
      "kl": 8.015625,
      "learning_rate": 5.712788259958071e-07,
      "loss": 0.479,
      "reward": 0.6443088054656982,
      "reward_std": 0.168148435652256,
      "reward_std_all": 0.16805247217416763,
      "rewards-max/forensics_sinkhorn_reward": 0.9852947294712067,
      "rewards-max/forensics_strict_edge_reward": 0.9256840348243713,
      "rewards-max/forensics_tiou_reward": 0.9854353964328766,
      "rewards/forensics_sinkhorn_reward": 0.48702919483184814,
      "rewards/forensics_strict_edge_reward": 0.27325090765953064,
      "rewards/forensics_tiou_reward": 0.2842136174440384,
      "step": 819
    },
    {
      "completion_length": 57.921875,
      "epoch": 1.7162303664921466,
      "grad_norm": 21.912151336669922,
      "kl": 6.109375,
      "learning_rate": 5.707547169811321e-07,
      "loss": 0.4363,
      "reward": 0.6153487861156464,
      "reward_std": 0.12675385177135468,
      "reward_std_all": 0.1560574397444725,
      "rewards-max/forensics_sinkhorn_reward": 0.9086534082889557,
      "rewards-max/forensics_strict_edge_reward": 0.8394034504890442,
      "rewards-max/forensics_tiou_reward": 0.8371769189834595,
      "rewards/forensics_sinkhorn_reward": 0.4978974610567093,
      "rewards/forensics_strict_edge_reward": 0.2333904430270195,
      "rewards/forensics_tiou_reward": 0.2761754021048546,
      "step": 820
    },
    {
      "completion_length": 63.171875,
      "epoch": 1.7183246073298428,
      "grad_norm": 24.89455795288086,
      "kl": 6.1796875,
      "learning_rate": 5.70230607966457e-07,
      "loss": 0.3621,
      "reward": 0.6687883138656616,
      "reward_std": 0.1671227589249611,
      "reward_std_all": 0.16055617481470108,
      "rewards-max/forensics_sinkhorn_reward": 0.988235205411911,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9970235526561737,
      "rewards/forensics_sinkhorn_reward": 0.5538894236087799,
      "rewards/forensics_strict_edge_reward": 0.28101545572280884,
      "rewards/forensics_tiou_reward": 0.3759726881980896,
      "step": 821
    },
    {
      "completion_length": 60.1875,
      "epoch": 1.7204188481675393,
      "grad_norm": 24.121841430664062,
      "kl": 4.578125,
      "learning_rate": 5.697064989517819e-07,
      "loss": 0.3878,
      "reward": 0.6735987067222595,
      "reward_std": 0.13601043820381165,
      "reward_std_all": 0.13779070228338242,
      "rewards-max/forensics_sinkhorn_reward": 0.98834627866745,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9974629580974579,
      "rewards/forensics_sinkhorn_reward": 0.6732403337955475,
      "rewards/forensics_strict_edge_reward": 0.42776423692703247,
      "rewards/forensics_tiou_reward": 0.5239787399768829,
      "step": 822
    },
    {
      "completion_length": 28.4375,
      "epoch": 1.7225130890052356,
      "grad_norm": 19.67502784729004,
      "kl": 2.4296875,
      "learning_rate": 5.691823899371069e-07,
      "loss": 0.2401,
      "reward": 0.7044786810874939,
      "reward_std": 0.1069147065281868,
      "reward_std_all": 0.10528102144598961,
      "rewards-max/forensics_sinkhorn_reward": 0.9881636202335358,
      "rewards-max/forensics_strict_edge_reward": 0.9464850425720215,
      "rewards-max/forensics_tiou_reward": 0.9967445135116577,
      "rewards/forensics_sinkhorn_reward": 0.6769652366638184,
      "rewards/forensics_strict_edge_reward": 0.407526433467865,
      "rewards/forensics_tiou_reward": 0.49944984912872314,
      "step": 823
    },
    {
      "completion_length": 61.59375,
      "epoch": 1.7246073298429319,
      "grad_norm": 17.95820426940918,
      "kl": 5.5703125,
      "learning_rate": 5.686582809224318e-07,
      "loss": 0.4852,
      "reward": 0.6530170440673828,
      "reward_std": 0.15964911133050919,
      "reward_std_all": 0.15488436073064804,
      "rewards-max/forensics_sinkhorn_reward": 0.9637273252010345,
      "rewards-max/forensics_strict_edge_reward": 0.8920957744121552,
      "rewards-max/forensics_tiou_reward": 0.9445841908454895,
      "rewards/forensics_sinkhorn_reward": 0.5433329641819,
      "rewards/forensics_strict_edge_reward": 0.3437053710222244,
      "rewards/forensics_tiou_reward": 0.3843809962272644,
      "step": 824
    },
    {
      "completion_length": 74.03125,
      "epoch": 1.7267015706806284,
      "grad_norm": 27.031999588012695,
      "kl": 8.4375,
      "learning_rate": 5.681341719077568e-07,
      "loss": 0.4882,
      "reward": 0.6233128905296326,
      "reward_std": 0.2118956595659256,
      "reward_std_all": 0.2016039565205574,
      "rewards-max/forensics_sinkhorn_reward": 0.9883950054645538,
      "rewards-max/forensics_strict_edge_reward": 0.9890114367008209,
      "rewards-max/forensics_tiou_reward": 0.9976553320884705,
      "rewards/forensics_sinkhorn_reward": 0.5215715765953064,
      "rewards/forensics_strict_edge_reward": 0.3520469516515732,
      "rewards/forensics_tiou_reward": 0.39107346534729004,
      "step": 825
    },
    {
      "completion_length": 46.8125,
      "epoch": 1.7287958115183246,
      "grad_norm": 23.39073944091797,
      "kl": 5.140625,
      "learning_rate": 5.676100628930817e-07,
      "loss": 0.3967,
      "reward": 0.7270099222660065,
      "reward_std": 0.10841532424092293,
      "reward_std_all": 0.09894984588027,
      "rewards-max/forensics_sinkhorn_reward": 0.9884881675243378,
      "rewards-max/forensics_strict_edge_reward": 0.9782643616199493,
      "rewards-max/forensics_tiou_reward": 0.998019814491272,
      "rewards/forensics_sinkhorn_reward": 0.6230281591415405,
      "rewards/forensics_strict_edge_reward": 0.4227665066719055,
      "rewards/forensics_tiou_reward": 0.46833106875419617,
      "step": 826
    },
    {
      "completion_length": 55.796875,
      "epoch": 1.730890052356021,
      "grad_norm": 88.07475280761719,
      "kl": 5.9375,
      "learning_rate": 5.670859538784066e-07,
      "loss": 0.3122,
      "reward": 0.6279963850975037,
      "reward_std": 0.12063170596957207,
      "reward_std_all": 0.14724411070346832,
      "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": 0.5456505417823792,
      "rewards/forensics_strict_edge_reward": 0.3163062483072281,
      "rewards/forensics_tiou_reward": 0.3359935134649277,
      "step": 827
    },
    {
      "completion_length": 72.765625,
      "epoch": 1.7329842931937174,
      "grad_norm": 81.43993377685547,
      "kl": 7.265625,
      "learning_rate": 5.665618448637315e-07,
      "loss": 0.3677,
      "reward": 0.6620738506317139,
      "reward_std": 0.14438940957188606,
      "reward_std_all": 0.14830459654331207,
      "rewards-max/forensics_sinkhorn_reward": 0.9419349133968353,
      "rewards-max/forensics_strict_edge_reward": 0.8130750358104706,
      "rewards-max/forensics_tiou_reward": 0.8861533403396606,
      "rewards/forensics_sinkhorn_reward": 0.5266309678554535,
      "rewards/forensics_strict_edge_reward": 0.32736214995384216,
      "rewards/forensics_tiou_reward": 0.34488579630851746,
      "step": 828
    },
    {
      "completion_length": 74.609375,
      "epoch": 1.7350785340314137,
      "grad_norm": 37.453285217285156,
      "kl": 7.65625,
      "learning_rate": 5.660377358490566e-07,
      "loss": 0.5381,
      "reward": 0.6175114512443542,
      "reward_std": 0.18264759331941605,
      "reward_std_all": 0.19050436466932297,
      "rewards-max/forensics_sinkhorn_reward": 0.9887502789497375,
      "rewards-max/forensics_strict_edge_reward": 0.9890114367008209,
      "rewards-max/forensics_tiou_reward": 0.9990534782409668,
      "rewards/forensics_sinkhorn_reward": 0.48485109210014343,
      "rewards/forensics_strict_edge_reward": 0.2857949882745743,
      "rewards/forensics_tiou_reward": 0.30765550583601,
      "step": 829
    },
    {
      "completion_length": 75.171875,
      "epoch": 1.73717277486911,
      "grad_norm": 52.686363220214844,
      "kl": 7.921875,
      "learning_rate": 5.655136268343816e-07,
      "loss": 0.4011,
      "reward": 0.6348024010658264,
      "reward_std": 0.21314756572246552,
      "reward_std_all": 0.19858711957931519,
      "rewards-max/forensics_sinkhorn_reward": 0.9873627424240112,
      "rewards-max/forensics_strict_edge_reward": 0.9062341153621674,
      "rewards-max/forensics_tiou_reward": 0.9935837686061859,
      "rewards/forensics_sinkhorn_reward": 0.5586414933204651,
      "rewards/forensics_strict_edge_reward": 0.3411742150783539,
      "rewards/forensics_tiou_reward": 0.42482347786426544,
      "step": 830
    },
    {
      "completion_length": 83.484375,
      "epoch": 1.7392670157068064,
      "grad_norm": 18.04444694519043,
      "kl": 7.375,
      "learning_rate": 5.649895178197065e-07,
      "loss": 0.4391,
      "reward": 0.5603184998035431,
      "reward_std": 0.2285381704568863,
      "reward_std_all": 0.23312504589557648,
      "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": 0.5166285634040833,
      "rewards/forensics_strict_edge_reward": 0.32718514651060104,
      "rewards/forensics_tiou_reward": 0.3616478443145752,
      "step": 831
    },
    {
      "completion_length": 39.6875,
      "epoch": 1.7413612565445025,
      "grad_norm": 20.681543350219727,
      "kl": 3.234375,
      "learning_rate": 5.644654088050315e-07,
      "loss": 0.2011,
      "reward": 0.6556053459644318,
      "reward_std": 0.13956154137849808,
      "reward_std_all": 0.1494724154472351,
      "rewards-max/forensics_sinkhorn_reward": 0.9287256598472595,
      "rewards-max/forensics_strict_edge_reward": 0.8134243786334991,
      "rewards-max/forensics_tiou_reward": 0.868873655796051,
      "rewards/forensics_sinkhorn_reward": 0.5324345082044601,
      "rewards/forensics_strict_edge_reward": 0.29493098706007004,
      "rewards/forensics_tiou_reward": 0.2969532832503319,
      "step": 832
    },
    {
      "completion_length": 97.84375,
      "epoch": 1.743455497382199,
      "grad_norm": 15.917027473449707,
      "kl": 7.515625,
      "learning_rate": 5.639412997903563e-07,
      "loss": 0.4324,
      "reward": 0.61282879114151,
      "reward_std": 0.16886409372091293,
      "reward_std_all": 0.1796695813536644,
      "rewards-max/forensics_sinkhorn_reward": 0.9877251982688904,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9950198531150818,
      "rewards/forensics_sinkhorn_reward": 0.5182092189788818,
      "rewards/forensics_strict_edge_reward": 0.2848675698041916,
      "rewards/forensics_tiou_reward": 0.326650470495224,
      "step": 833
    },
    {
      "completion_length": 41.421875,
      "epoch": 1.7455497382198955,
      "grad_norm": 16.302080154418945,
      "kl": 2.87890625,
      "learning_rate": 5.634171907756813e-07,
      "loss": 0.3086,
      "reward": 0.653042733669281,
      "reward_std": 0.13111497089266777,
      "reward_std_all": 0.1420692354440689,
      "rewards-max/forensics_sinkhorn_reward": 0.97844198346138,
      "rewards-max/forensics_strict_edge_reward": 0.8915697336196899,
      "rewards-max/forensics_tiou_reward": 0.9766546487808228,
      "rewards/forensics_sinkhorn_reward": 0.6057130098342896,
      "rewards/forensics_strict_edge_reward": 0.36943473666906357,
      "rewards/forensics_tiou_reward": 0.44509728997945786,
      "step": 834
    },
    {
      "completion_length": 81.71875,
      "epoch": 1.7476439790575915,
      "grad_norm": 22.685558319091797,
      "kl": 5.796875,
      "learning_rate": 5.628930817610062e-07,
      "loss": 0.3691,
      "reward": 0.6640920639038086,
      "reward_std": 0.1332027092576027,
      "reward_std_all": 0.1354162096977234,
      "rewards-max/forensics_sinkhorn_reward": 0.9619537591934204,
      "rewards-max/forensics_strict_edge_reward": 0.9574736058712006,
      "rewards-max/forensics_tiou_reward": 0.9402965009212494,
      "rewards/forensics_sinkhorn_reward": 0.5697259604930878,
      "rewards/forensics_strict_edge_reward": 0.3069159835577011,
      "rewards/forensics_tiou_reward": 0.3466961979866028,
      "step": 835
    },
    {
      "completion_length": 62.578125,
      "epoch": 1.749738219895288,
      "grad_norm": 21.38347053527832,
      "kl": 4.921875,
      "learning_rate": 5.623689727463312e-07,
      "loss": 0.3294,
      "reward": 0.6525534987449646,
      "reward_std": 0.1200651042163372,
      "reward_std_all": 0.13678022474050522,
      "rewards-max/forensics_sinkhorn_reward": 0.9683636724948883,
      "rewards-max/forensics_strict_edge_reward": 0.8578422963619232,
      "rewards-max/forensics_tiou_reward": 0.9512321054935455,
      "rewards/forensics_sinkhorn_reward": 0.552907407283783,
      "rewards/forensics_strict_edge_reward": 0.28969667851924896,
      "rewards/forensics_tiou_reward": 0.35912057757377625,
      "step": 836
    },
    {
      "completion_length": 82.640625,
      "epoch": 1.7518324607329843,
      "grad_norm": 20.33755874633789,
      "kl": 5.8203125,
      "learning_rate": 5.618448637316562e-07,
      "loss": 0.3709,
      "reward": 0.6722258925437927,
      "reward_std": 0.1528000570833683,
      "reward_std_all": 0.146547332406044,
      "rewards-max/forensics_sinkhorn_reward": 0.9868887662887573,
      "rewards-max/forensics_strict_edge_reward": 0.9567649364471436,
      "rewards-max/forensics_tiou_reward": 0.9917485415935516,
      "rewards/forensics_sinkhorn_reward": 0.5929804146289825,
      "rewards/forensics_strict_edge_reward": 0.3279787600040436,
      "rewards/forensics_tiou_reward": 0.4140002578496933,
      "step": 837
    },
    {
      "completion_length": 93.3125,
      "epoch": 1.7539267015706805,
      "grad_norm": 16.919992446899414,
      "kl": 5.90625,
      "learning_rate": 5.613207547169811e-07,
      "loss": 0.4432,
      "reward": 0.580250471830368,
      "reward_std": 0.22913920879364014,
      "reward_std_all": 0.22698603570461273,
      "rewards-max/forensics_sinkhorn_reward": 0.9773475527763367,
      "rewards-max/forensics_strict_edge_reward": 0.9567649364471436,
      "rewards-max/forensics_tiou_reward": 0.974012553691864,
      "rewards/forensics_sinkhorn_reward": 0.48804783821105957,
      "rewards/forensics_strict_edge_reward": 0.3147244304418564,
      "rewards/forensics_tiou_reward": 0.3557267338037491,
      "step": 838
    },
    {
      "completion_length": 64.984375,
      "epoch": 1.756020942408377,
      "grad_norm": 18.51213836669922,
      "kl": 4.765625,
      "learning_rate": 5.60796645702306e-07,
      "loss": 0.338,
      "reward": 0.6833029091358185,
      "reward_std": 0.13828511908650398,
      "reward_std_all": 0.13435811921954155,
      "rewards-max/forensics_sinkhorn_reward": 0.8506304621696472,
      "rewards-max/forensics_strict_edge_reward": 0.7284609377384186,
      "rewards-max/forensics_tiou_reward": 0.7851124107837677,
      "rewards/forensics_sinkhorn_reward": 0.49897903203964233,
      "rewards/forensics_strict_edge_reward": 0.28083011507987976,
      "rewards/forensics_tiou_reward": 0.3206006810069084,
      "step": 839
    },
    {
      "completion_length": 57.78125,
      "epoch": 1.7581151832460733,
      "grad_norm": 15.018624305725098,
      "kl": 3.3984375,
      "learning_rate": 5.602725366876309e-07,
      "loss": 0.225,
      "reward": 0.6519008278846741,
      "reward_std": 0.10337556153535843,
      "reward_std_all": 0.1209525316953659,
      "rewards-max/forensics_sinkhorn_reward": 0.9874422550201416,
      "rewards-max/forensics_strict_edge_reward": 0.9357379674911499,
      "rewards-max/forensics_tiou_reward": 0.9939001202583313,
      "rewards/forensics_sinkhorn_reward": 0.525130569934845,
      "rewards/forensics_strict_edge_reward": 0.29837431013584137,
      "rewards/forensics_tiou_reward": 0.30732880532741547,
      "step": 840
    },
    {
      "completion_length": 46.21875,
      "epoch": 1.7602094240837696,
      "grad_norm": 28.55628204345703,
      "kl": 4.546875,
      "learning_rate": 5.59748427672956e-07,
      "loss": 0.2315,
      "reward": 0.7260590195655823,
      "reward_std": 0.10673835501074791,
      "reward_std_all": 0.09805629402399063,
      "rewards-max/forensics_sinkhorn_reward": 0.9889201521873474,
      "rewards-max/forensics_strict_edge_reward": 0.9890114367008209,
      "rewards-max/forensics_tiou_reward": 0.9997227191925049,
      "rewards/forensics_sinkhorn_reward": 0.6712216138839722,
      "rewards/forensics_strict_edge_reward": 0.40939459204673767,
      "rewards/forensics_tiou_reward": 0.513128325343132,
      "step": 841
    },
    {
      "completion_length": 103.34375,
      "epoch": 1.762303664921466,
      "grad_norm": 29.88141632080078,
      "kl": 8.84375,
      "learning_rate": 5.592243186582809e-07,
      "loss": 0.5316,
      "reward": 0.6133026480674744,
      "reward_std": 0.13698424398899078,
      "reward_std_all": 0.15898430347442627,
      "rewards-max/forensics_sinkhorn_reward": 0.977541983127594,
      "rewards-max/forensics_strict_edge_reward": 0.9256840348243713,
      "rewards-max/forensics_tiou_reward": 0.9730077385902405,
      "rewards/forensics_sinkhorn_reward": 0.4663040339946747,
      "rewards/forensics_strict_edge_reward": 0.3055869787931442,
      "rewards/forensics_tiou_reward": 0.33140093088150024,
      "step": 842
    },
    {
      "completion_length": 78.28125,
      "epoch": 1.7643979057591623,
      "grad_norm": 28.26556396484375,
      "kl": 6.390625,
      "learning_rate": 5.587002096436059e-07,
      "loss": 0.5044,
      "reward": 0.6073521077632904,
      "reward_std": 0.2007688656449318,
      "reward_std_all": 0.2028101533651352,
      "rewards-max/forensics_sinkhorn_reward": 0.9876551628112793,
      "rewards-max/forensics_strict_edge_reward": 0.9782643616199493,
      "rewards-max/forensics_tiou_reward": 0.9947368502616882,
      "rewards/forensics_sinkhorn_reward": 0.5228333324193954,
      "rewards/forensics_strict_edge_reward": 0.3577217757701874,
      "rewards/forensics_tiou_reward": 0.40394487977027893,
      "step": 843
    },
    {
      "completion_length": 57.21875,
      "epoch": 1.7664921465968586,
      "grad_norm": 19.752941131591797,
      "kl": 4.3359375,
      "learning_rate": 5.581761006289309e-07,
      "loss": 0.3343,
      "reward": 0.6784136295318604,
      "reward_std": 0.15247753262519836,
      "reward_std_all": 0.14223092794418335,
      "rewards-max/forensics_sinkhorn_reward": 0.9596501886844635,
      "rewards-max/forensics_strict_edge_reward": 0.9000824987888336,
      "rewards-max/forensics_tiou_reward": 0.9344562888145447,
      "rewards/forensics_sinkhorn_reward": 0.6038352251052856,
      "rewards/forensics_strict_edge_reward": 0.41051241755485535,
      "rewards/forensics_tiou_reward": 0.4500087946653366,
      "step": 844
    },
    {
      "completion_length": 60.171875,
      "epoch": 1.768586387434555,
      "grad_norm": 21.896236419677734,
      "kl": 4.5859375,
      "learning_rate": 5.576519916142557e-07,
      "loss": 0.3736,
      "reward": 0.5647617876529694,
      "reward_std": 0.22089388966560364,
      "reward_std_all": 0.22506506741046906,
      "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": 0.43940648436546326,
      "rewards/forensics_strict_edge_reward": 0.2692709416151047,
      "rewards/forensics_tiou_reward": 0.2886297330260277,
      "step": 845
    },
    {
      "completion_length": 76.953125,
      "epoch": 1.7706806282722511,
      "grad_norm": 30.391389846801758,
      "kl": 7.0625,
      "learning_rate": 5.571278825995807e-07,
      "loss": 0.4218,
      "reward": 0.6205381751060486,
      "reward_std": 0.18508898466825485,
      "reward_std_all": 0.18319042026996613,
      "rewards-max/forensics_sinkhorn_reward": 0.9878676235675812,
      "rewards-max/forensics_strict_edge_reward": 0.9364311099052429,
      "rewards-max/forensics_tiou_reward": 0.9955784976482391,
      "rewards/forensics_sinkhorn_reward": 0.5041462182998657,
      "rewards/forensics_strict_edge_reward": 0.2787366434931755,
      "rewards/forensics_tiou_reward": 0.344777911901474,
      "step": 846
    },
    {
      "completion_length": 68.109375,
      "epoch": 1.7727748691099476,
      "grad_norm": 18.59803009033203,
      "kl": 5.75,
      "learning_rate": 5.566037735849056e-07,
      "loss": 0.3934,
      "reward": 0.5733234882354736,
      "reward_std": 0.172095388174057,
      "reward_std_all": 0.19464586675167084,
      "rewards-max/forensics_sinkhorn_reward": 0.9640994966030121,
      "rewards-max/forensics_strict_edge_reward": 0.8134243786334991,
      "rewards-max/forensics_tiou_reward": 0.9413106143474579,
      "rewards/forensics_sinkhorn_reward": 0.4609702527523041,
      "rewards/forensics_strict_edge_reward": 0.2226264774799347,
      "rewards/forensics_tiou_reward": 0.24427929520606995,
      "step": 847
    },
    {
      "completion_length": 51.703125,
      "epoch": 1.7748691099476441,
      "grad_norm": 22.778627395629883,
      "kl": 4.65625,
      "learning_rate": 5.560796645702306e-07,
      "loss": 0.2419,
      "reward": 0.6435187458992004,
      "reward_std": 0.1705806404352188,
      "reward_std_all": 0.16706743091344833,
      "rewards-max/forensics_sinkhorn_reward": 0.9876309037208557,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9946415424346924,
      "rewards/forensics_sinkhorn_reward": 0.5476343035697937,
      "rewards/forensics_strict_edge_reward": 0.2642090171575546,
      "rewards/forensics_tiou_reward": 0.37410295009613037,
      "step": 848
    },
    {
      "completion_length": 57.828125,
      "epoch": 1.7769633507853402,
      "grad_norm": 18.80834197998047,
      "kl": 4.765625,
      "learning_rate": 5.555555555555555e-07,
      "loss": 0.2116,
      "reward": 0.6051084399223328,
      "reward_std": 0.1567114144563675,
      "reward_std_all": 0.17467167228460312,
      "rewards-max/forensics_sinkhorn_reward": 0.9459726214408875,
      "rewards-max/forensics_strict_edge_reward": 0.7978116869926453,
      "rewards-max/forensics_tiou_reward": 0.901844710111618,
      "rewards/forensics_sinkhorn_reward": 0.48559488356113434,
      "rewards/forensics_strict_edge_reward": 0.1989063248038292,
      "rewards/forensics_tiou_reward": 0.24250246584415436,
      "step": 849
    },
    {
      "completion_length": 76.109375,
      "epoch": 1.7790575916230367,
      "grad_norm": 21.57366943359375,
      "kl": 6.109375,
      "learning_rate": 5.550314465408805e-07,
      "loss": 0.3739,
      "reward": 0.5340695679187775,
      "reward_std": 0.23823441565036774,
      "reward_std_all": 0.24657493084669113,
      "rewards-max/forensics_sinkhorn_reward": 0.9880357086658478,
      "rewards-max/forensics_strict_edge_reward": 0.9567649364471436,
      "rewards-max/forensics_tiou_reward": 0.9962399005889893,
      "rewards/forensics_sinkhorn_reward": 0.4424520581960678,
      "rewards/forensics_strict_edge_reward": 0.19232027977705002,
      "rewards/forensics_tiou_reward": 0.2888500392436981,
      "step": 850
    },
    {
      "completion_length": 90.171875,
      "epoch": 1.781151832460733,
      "grad_norm": 23.405454635620117,
      "kl": 5.890625,
      "learning_rate": 5.545073375262053e-07,
      "loss": 0.3998,
      "reward": 0.5941275954246521,
      "reward_std": 0.1697566993534565,
      "reward_std_all": 0.18386729806661606,
      "rewards-max/forensics_sinkhorn_reward": 0.9838329255580902,
      "rewards-max/forensics_strict_edge_reward": 0.8854433596134186,
      "rewards-max/forensics_tiou_reward": 0.9798262715339661,
      "rewards/forensics_sinkhorn_reward": 0.4518076479434967,
      "rewards/forensics_strict_edge_reward": 0.24381646513938904,
      "rewards/forensics_tiou_reward": 0.32597101479768753,
      "step": 851
    },
    {
      "completion_length": 55.359375,
      "epoch": 1.7832460732984292,
      "grad_norm": 26.131935119628906,
      "kl": 4.703125,
      "learning_rate": 5.539832285115303e-07,
      "loss": 0.2809,
      "reward": 0.6005558967590332,
      "reward_std": 0.12382140010595322,
      "reward_std_all": 0.1643872633576393,
      "rewards-max/forensics_sinkhorn_reward": 0.9612783193588257,
      "rewards-max/forensics_strict_edge_reward": 0.947419673204422,
      "rewards-max/forensics_tiou_reward": 0.9422177374362946,
      "rewards/forensics_sinkhorn_reward": 0.5382785201072693,
      "rewards/forensics_strict_edge_reward": 0.26982206851243973,
      "rewards/forensics_tiou_reward": 0.3371746689081192,
      "step": 852
    },
    {
      "completion_length": 88.046875,
      "epoch": 1.7853403141361257,
      "grad_norm": 16.87331199645996,
      "kl": 7.59375,
      "learning_rate": 5.534591194968554e-07,
      "loss": 0.4269,
      "reward": 0.597354531288147,
      "reward_std": 0.1675269901752472,
      "reward_std_all": 0.186879001557827,
      "rewards-max/forensics_sinkhorn_reward": 0.9879344701766968,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9958392381668091,
      "rewards/forensics_sinkhorn_reward": 0.47223490476608276,
      "rewards/forensics_strict_edge_reward": 0.263204000890255,
      "rewards/forensics_tiou_reward": 0.33573637902736664,
      "step": 853
    },
    {
      "completion_length": 44.390625,
      "epoch": 1.787434554973822,
      "grad_norm": 20.57558822631836,
      "kl": 4.15625,
      "learning_rate": 5.529350104821803e-07,
      "loss": 0.2235,
      "reward": 0.6785168945789337,
      "reward_std": 0.1544942818582058,
      "reward_std_all": 0.14589520543813705,
      "rewards-max/forensics_sinkhorn_reward": 0.987892210483551,
      "rewards-max/forensics_strict_edge_reward": 0.9782643616199493,
      "rewards-max/forensics_tiou_reward": 0.9956803619861603,
      "rewards/forensics_sinkhorn_reward": 0.5981360971927643,
      "rewards/forensics_strict_edge_reward": 0.3330838084220886,
      "rewards/forensics_tiou_reward": 0.40819378197193146,
      "step": 854
    },
    {
      "completion_length": 43.09375,
      "epoch": 1.7895287958115182,
      "grad_norm": 19.156709671020508,
      "kl": 2.8359375,
      "learning_rate": 5.524109014675053e-07,
      "loss": 0.1491,
      "reward": 0.6820926070213318,
      "reward_std": 0.09499548748135567,
      "reward_std_all": 0.11308640986680984,
      "rewards-max/forensics_sinkhorn_reward": 0.9839230179786682,
      "rewards-max/forensics_strict_edge_reward": 0.9158510267734528,
      "rewards-max/forensics_tiou_reward": 0.9873239994049072,
      "rewards/forensics_sinkhorn_reward": 0.5564481616020203,
      "rewards/forensics_strict_edge_reward": 0.20910032093524933,
      "rewards/forensics_tiou_reward": 0.2695617377758026,
      "step": 855
    },
    {
      "completion_length": 84.046875,
      "epoch": 1.7916230366492147,
      "grad_norm": 19.617612838745117,
      "kl": 5.46875,
      "learning_rate": 5.518867924528302e-07,
      "loss": 0.2802,
      "reward": 0.6467067897319794,
      "reward_std": 0.15758485347032547,
      "reward_std_all": 0.15939293056726456,
      "rewards-max/forensics_sinkhorn_reward": 0.9788352847099304,
      "rewards-max/forensics_strict_edge_reward": 0.9158510267734528,
      "rewards-max/forensics_tiou_reward": 0.960704505443573,
      "rewards/forensics_sinkhorn_reward": 0.5393374860286713,
      "rewards/forensics_strict_edge_reward": 0.2669720947742462,
      "rewards/forensics_tiou_reward": 0.33316247165203094,
      "step": 856
    },
    {
      "completion_length": 72.875,
      "epoch": 1.793717277486911,
      "grad_norm": 18.780277252197266,
      "kl": 5.5,
      "learning_rate": 5.513626834381551e-07,
      "loss": 0.3855,
      "reward": 0.6460232734680176,
      "reward_std": 0.14077498018741608,
      "reward_std_all": 0.14559901505708694,
      "rewards-max/forensics_sinkhorn_reward": 0.9869773089885712,
      "rewards-max/forensics_strict_edge_reward": 0.9565287232398987,
      "rewards-max/forensics_tiou_reward": 0.9920647144317627,
      "rewards/forensics_sinkhorn_reward": 0.5233500897884369,
      "rewards/forensics_strict_edge_reward": 0.3242308348417282,
      "rewards/forensics_tiou_reward": 0.37452608346939087,
      "step": 857
    },
    {
      "completion_length": 64.484375,
      "epoch": 1.7958115183246073,
      "grad_norm": 29.765182495117188,
      "kl": 5.15625,
      "learning_rate": 5.5083857442348e-07,
      "loss": 0.3023,
      "reward": 0.6148253381252289,
      "reward_std": 0.20867447555065155,
      "reward_std_all": 0.1996917501091957,
      "rewards-max/forensics_sinkhorn_reward": 0.988743245601654,
      "rewards-max/forensics_strict_edge_reward": 0.9890114367008209,
      "rewards-max/forensics_tiou_reward": 0.999025821685791,
      "rewards/forensics_sinkhorn_reward": 0.5308735817670822,
      "rewards/forensics_strict_edge_reward": 0.3187635838985443,
      "rewards/forensics_tiou_reward": 0.3850001394748688,
      "step": 858
    },
    {
      "completion_length": 64.890625,
      "epoch": 1.7979057591623038,
      "grad_norm": 16.503368377685547,
      "kl": 5.1875,
      "learning_rate": 5.50314465408805e-07,
      "loss": 0.3336,
      "reward": 0.6201125681400299,
      "reward_std": 0.14200259372591972,
      "reward_std_all": 0.1621696762740612,
      "rewards-max/forensics_sinkhorn_reward": 0.888479471206665,
      "rewards-max/forensics_strict_edge_reward": 0.6709822118282318,
      "rewards-max/forensics_tiou_reward": 0.7753482758998871,
      "rewards/forensics_sinkhorn_reward": 0.42731958627700806,
      "rewards/forensics_strict_edge_reward": 0.16395673155784607,
      "rewards/forensics_tiou_reward": 0.22498873621225357,
      "step": 859
    },
    {
      "completion_length": 72.578125,
      "epoch": 1.8,
      "grad_norm": 19.014921188354492,
      "kl": 5.625,
      "learning_rate": 5.4979035639413e-07,
      "loss": 0.4843,
      "reward": 0.5932319462299347,
      "reward_std": 0.225785493850708,
      "reward_std_all": 0.21959883719682693,
      "rewards-max/forensics_sinkhorn_reward": 0.9733260273933411,
      "rewards-max/forensics_strict_edge_reward": 0.8288039863109589,
      "rewards-max/forensics_tiou_reward": 0.964801162481308,
      "rewards/forensics_sinkhorn_reward": 0.5034154653549194,
      "rewards/forensics_strict_edge_reward": 0.2560557425022125,
      "rewards/forensics_tiou_reward": 0.3462161123752594,
      "step": 860
    },
    {
      "completion_length": 48.1875,
      "epoch": 1.8020942408376963,
      "grad_norm": 16.563508987426758,
      "kl": 4.921875,
      "learning_rate": 5.492662473794549e-07,
      "loss": 0.3296,
      "reward": 0.6656391322612762,
      "reward_std": 0.12617044895887375,
      "reward_std_all": 0.12896598875522614,
      "rewards-max/forensics_sinkhorn_reward": 0.9690871238708496,
      "rewards-max/forensics_strict_edge_reward": 0.8553276658058167,
      "rewards-max/forensics_tiou_reward": 0.95280322432518,
      "rewards/forensics_sinkhorn_reward": 0.5569967329502106,
      "rewards/forensics_strict_edge_reward": 0.266587495803833,
      "rewards/forensics_tiou_reward": 0.3861006647348404,
      "step": 861
    },
    {
      "completion_length": 64.015625,
      "epoch": 1.8041884816753928,
      "grad_norm": 287.5954284667969,
      "kl": 11.03125,
      "learning_rate": 5.487421383647799e-07,
      "loss": 0.6213,
      "reward": 0.6100191473960876,
      "reward_std": 0.23183166235685349,
      "reward_std_all": 0.2172875925898552,
      "rewards-max/forensics_sinkhorn_reward": 0.9887523055076599,
      "rewards-max/forensics_strict_edge_reward": 0.9890114367008209,
      "rewards-max/forensics_tiou_reward": 0.9990610182285309,
      "rewards/forensics_sinkhorn_reward": 0.5486265122890472,
      "rewards/forensics_strict_edge_reward": 0.2787763923406601,
      "rewards/forensics_tiou_reward": 0.399514764547348,
      "step": 862
    },
    {
      "completion_length": 76.078125,
      "epoch": 1.8062827225130889,
      "grad_norm": 41.3411979675293,
      "kl": 7.4375,
      "learning_rate": 5.482180293501047e-07,
      "loss": 0.4541,
      "reward": 0.5957343876361847,
      "reward_std": 0.1491369754076004,
      "reward_std_all": 0.17857690900564194,
      "rewards-max/forensics_sinkhorn_reward": 0.9343783557415009,
      "rewards-max/forensics_strict_edge_reward": 0.9058879017829895,
      "rewards-max/forensics_tiou_reward": 0.8846005201339722,
      "rewards/forensics_sinkhorn_reward": 0.4676159918308258,
      "rewards/forensics_strict_edge_reward": 0.23597120493650436,
      "rewards/forensics_tiou_reward": 0.31394118070602417,
      "step": 863
    },
    {
      "completion_length": 65.671875,
      "epoch": 1.8083769633507853,
      "grad_norm": 16.452957153320312,
      "kl": 8.015625,
      "learning_rate": 5.476939203354297e-07,
      "loss": 0.5496,
      "reward": 0.6115556359291077,
      "reward_std": 0.20369575917720795,
      "reward_std_all": 0.19526997953653336,
      "rewards-max/forensics_sinkhorn_reward": 0.9761381447315216,
      "rewards-max/forensics_strict_edge_reward": 0.9151731729507446,
      "rewards-max/forensics_tiou_reward": 0.9673410058021545,
      "rewards/forensics_sinkhorn_reward": 0.4628875404596329,
      "rewards/forensics_strict_edge_reward": 0.27506010234355927,
      "rewards/forensics_tiou_reward": 0.32838669419288635,
      "step": 864
    },
    {
      "completion_length": 39.015625,
      "epoch": 1.8104712041884818,
      "grad_norm": 26.092924118041992,
      "kl": 4.28125,
      "learning_rate": 5.471698113207546e-07,
      "loss": 0.3401,
      "reward": 0.7219474017620087,
      "reward_std": 0.13620680943131447,
      "reward_std_all": 0.1204291582107544,
      "rewards-max/forensics_sinkhorn_reward": 0.9868029952049255,
      "rewards-max/forensics_strict_edge_reward": 0.9677534997463226,
      "rewards-max/forensics_tiou_reward": 0.9914164543151855,
      "rewards/forensics_sinkhorn_reward": 0.6477856338024139,
      "rewards/forensics_strict_edge_reward": 0.3821375370025635,
      "rewards/forensics_tiou_reward": 0.4889214485883713,
      "step": 865
    },
    {
      "completion_length": 42.640625,
      "epoch": 1.812565445026178,
      "grad_norm": 20.571977615356445,
      "kl": 4.828125,
      "learning_rate": 5.466457023060797e-07,
      "loss": 0.3323,
      "reward": 0.6256788372993469,
      "reward_std": 0.17299121618270874,
      "reward_std_all": 0.17211805284023285,
      "rewards-max/forensics_sinkhorn_reward": 0.9880302250385284,
      "rewards-max/forensics_strict_edge_reward": 0.9565287232398987,
      "rewards-max/forensics_tiou_reward": 0.9962145388126373,
      "rewards/forensics_sinkhorn_reward": 0.4682367593050003,
      "rewards/forensics_strict_edge_reward": 0.2681640759110451,
      "rewards/forensics_tiou_reward": 0.30327102541923523,
      "step": 866
    },
    {
      "completion_length": 41.484375,
      "epoch": 1.8146596858638744,
      "grad_norm": 17.024757385253906,
      "kl": 5.21875,
      "learning_rate": 5.461215932914046e-07,
      "loss": 0.4142,
      "reward": 0.6651987433433533,
      "reward_std": 0.19962728768587112,
      "reward_std_all": 0.18017210811376572,
      "rewards-max/forensics_sinkhorn_reward": 0.9841606318950653,
      "rewards-max/forensics_strict_edge_reward": 0.9565287232398987,
      "rewards-max/forensics_tiou_reward": 0.9879355132579803,
      "rewards/forensics_sinkhorn_reward": 0.5910686552524567,
      "rewards/forensics_strict_edge_reward": 0.33780041337013245,
      "rewards/forensics_tiou_reward": 0.4337436407804489,
      "step": 867
    },
    {
      "completion_length": 48.40625,
      "epoch": 1.8167539267015707,
      "grad_norm": 23.61299705505371,
      "kl": 7.546875,
      "learning_rate": 5.455974842767296e-07,
      "loss": 0.3777,
      "reward": 0.6145778894424438,
      "reward_std": 0.18328332901000977,
      "reward_std_all": 0.18105701357126236,
      "rewards-max/forensics_sinkhorn_reward": 0.9167237877845764,
      "rewards-max/forensics_strict_edge_reward": 0.8503920137882233,
      "rewards-max/forensics_tiou_reward": 0.8469734787940979,
      "rewards/forensics_sinkhorn_reward": 0.5089200288057327,
      "rewards/forensics_strict_edge_reward": 0.28438763320446014,
      "rewards/forensics_tiou_reward": 0.342031791806221,
      "step": 868
    },
    {
      "completion_length": 23.34375,
      "epoch": 1.818848167539267,
      "grad_norm": 28.287841796875,
      "kl": 3.9375,
      "learning_rate": 5.450733752620545e-07,
      "loss": 0.2886,
      "reward": 0.6071414053440094,
      "reward_std": 0.19541185349225998,
      "reward_std_all": 0.1981642171740532,
      "rewards-max/forensics_sinkhorn_reward": 0.9879425466060638,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9958733022212982,
      "rewards/forensics_sinkhorn_reward": 0.5439327955245972,
      "rewards/forensics_strict_edge_reward": 0.316249817609787,
      "rewards/forensics_tiou_reward": 0.39432255923748016,
      "step": 869
    },
    {
      "completion_length": 37.875,
      "epoch": 1.8209424083769634,
      "grad_norm": 26.958377838134766,
      "kl": 5.265625,
      "learning_rate": 5.445492662473794e-07,
      "loss": 0.2061,
      "reward": 0.6265735030174255,
      "reward_std": 0.17101890593767166,
      "reward_std_all": 0.17490672320127487,
      "rewards-max/forensics_sinkhorn_reward": 0.8967329859733582,
      "rewards-max/forensics_strict_edge_reward": 0.7160191237926483,
      "rewards-max/forensics_tiou_reward": 0.7964411973953247,
      "rewards/forensics_sinkhorn_reward": 0.47427308559417725,
      "rewards/forensics_strict_edge_reward": 0.13252109289169312,
      "rewards/forensics_tiou_reward": 0.2467975690960884,
      "step": 870
    },
    {
      "completion_length": 31.71875,
      "epoch": 1.8230366492146597,
      "grad_norm": 20.35581398010254,
      "kl": 4.140625,
      "learning_rate": 5.440251572327044e-07,
      "loss": 0.245,
      "reward": 0.6504344344139099,
      "reward_std": 0.16469499468803406,
      "reward_std_all": 0.160365030169487,
      "rewards-max/forensics_sinkhorn_reward": 0.9795598387718201,
      "rewards-max/forensics_strict_edge_reward": 0.830383837223053,
      "rewards-max/forensics_tiou_reward": 0.9764227569103241,
      "rewards/forensics_sinkhorn_reward": 0.5552995204925537,
      "rewards/forensics_strict_edge_reward": 0.2750912830233574,
      "rewards/forensics_tiou_reward": 0.36517588794231415,
      "step": 871
    },
    {
      "completion_length": 36.859375,
      "epoch": 1.825130890052356,
      "grad_norm": 24.923398971557617,
      "kl": 3.2734375,
      "learning_rate": 5.435010482180293e-07,
      "loss": 0.208,
      "reward": 0.6766555905342102,
      "reward_std": 0.14152280241250992,
      "reward_std_all": 0.1352665200829506,
      "rewards-max/forensics_sinkhorn_reward": 0.9871976375579834,
      "rewards-max/forensics_strict_edge_reward": 0.9464850425720215,
      "rewards-max/forensics_tiou_reward": 0.9929336905479431,
      "rewards/forensics_sinkhorn_reward": 0.5829043090343475,
      "rewards/forensics_strict_edge_reward": 0.2994643747806549,
      "rewards/forensics_tiou_reward": 0.3902880400419235,
      "step": 872
    },
    {
      "completion_length": 32.421875,
      "epoch": 1.8272251308900525,
      "grad_norm": 25.201135635375977,
      "kl": 3.8125,
      "learning_rate": 5.429769392033543e-07,
      "loss": 0.3035,
      "reward": 0.6696867048740387,
      "reward_std": 0.14017954468727112,
      "reward_std_all": 0.1403435915708542,
      "rewards-max/forensics_sinkhorn_reward": 0.9595607221126556,
      "rewards-max/forensics_strict_edge_reward": 0.8977590501308441,
      "rewards-max/forensics_tiou_reward": 0.9320161044597626,
      "rewards/forensics_sinkhorn_reward": 0.5380882173776627,
      "rewards/forensics_strict_edge_reward": 0.2886831909418106,
      "rewards/forensics_tiou_reward": 0.3105944097042084,
      "step": 873
    },
    {
      "completion_length": 34.640625,
      "epoch": 1.8293193717277487,
      "grad_norm": 21.06853485107422,
      "kl": 4.96875,
      "learning_rate": 5.424528301886792e-07,
      "loss": 0.4511,
      "reward": 0.6274375915527344,
      "reward_std": 0.1431245133280754,
      "reward_std_all": 0.15912870317697525,
      "rewards-max/forensics_sinkhorn_reward": 0.987797200679779,
      "rewards-max/forensics_strict_edge_reward": 0.9357379674911499,
      "rewards-max/forensics_tiou_reward": 0.9953024089336395,
      "rewards/forensics_sinkhorn_reward": 0.5249224305152893,
      "rewards/forensics_strict_edge_reward": 0.2716677561402321,
      "rewards/forensics_tiou_reward": 0.34632404148578644,
      "step": 874
    },
    {
      "completion_length": 46.078125,
      "epoch": 1.831413612565445,
      "grad_norm": 21.234569549560547,
      "kl": 5.53125,
      "learning_rate": 5.419287211740041e-07,
      "loss": 0.3404,
      "reward": 0.634967565536499,
      "reward_std": 0.16679541766643524,
      "reward_std_all": 0.1627141684293747,
      "rewards-max/forensics_sinkhorn_reward": 0.9475148320198059,
      "rewards-max/forensics_strict_edge_reward": 0.7317515909671783,
      "rewards-max/forensics_tiou_reward": 0.9070813953876495,
      "rewards/forensics_sinkhorn_reward": 0.5077743232250214,
      "rewards/forensics_strict_edge_reward": 0.26668232679367065,
      "rewards/forensics_tiou_reward": 0.35858912765979767,
      "step": 875
    },
    {
      "completion_length": 49.71875,
      "epoch": 1.8335078534031415,
      "grad_norm": 57.241600036621094,
      "kl": 9.15625,
      "learning_rate": 5.414046121593291e-07,
      "loss": 0.5273,
      "reward": 0.6920000910758972,
      "reward_std": 0.10327846929430962,
      "reward_std_all": 0.11252997070550919,
      "rewards-max/forensics_sinkhorn_reward": 0.9525035917758942,
      "rewards-max/forensics_strict_edge_reward": 0.8663162291049957,
      "rewards-max/forensics_tiou_reward": 0.9250144362449646,
      "rewards/forensics_sinkhorn_reward": 0.5781442523002625,
      "rewards/forensics_strict_edge_reward": 0.3142343610525131,
      "rewards/forensics_tiou_reward": 0.3967214971780777,
      "step": 876
    },
    {
      "completion_length": 34.5,
      "epoch": 1.8356020942408375,
      "grad_norm": 16.42867088317871,
      "kl": 3.0,
      "learning_rate": 5.40880503144654e-07,
      "loss": 0.2026,
      "reward": 0.7237122356891632,
      "reward_std": 0.10173710808157921,
      "reward_std_all": 0.09616556018590927,
      "rewards-max/forensics_sinkhorn_reward": 0.9877965748310089,
      "rewards-max/forensics_strict_edge_reward": 0.9425031542778015,
      "rewards-max/forensics_tiou_reward": 0.9952996671199799,
      "rewards/forensics_sinkhorn_reward": 0.5716698169708252,
      "rewards/forensics_strict_edge_reward": 0.2931778281927109,
      "rewards/forensics_tiou_reward": 0.39179740846157074,
      "step": 877
    },
    {
      "completion_length": 35.671875,
      "epoch": 1.837696335078534,
      "grad_norm": 15.991413116455078,
      "kl": 3.76171875,
      "learning_rate": 5.403563941299791e-07,
      "loss": 0.2088,
      "reward": 0.6759065091609955,
      "reward_std": 0.16638124361634254,
      "reward_std_all": 0.15578257851302624,
      "rewards-max/forensics_sinkhorn_reward": 0.9391858279705048,
      "rewards-max/forensics_strict_edge_reward": 0.7496984750032425,
      "rewards-max/forensics_tiou_reward": 0.8917707204818726,
      "rewards/forensics_sinkhorn_reward": 0.5491177141666412,
      "rewards/forensics_strict_edge_reward": 0.2367166429758072,
      "rewards/forensics_tiou_reward": 0.33999525010585785,
      "step": 878
    },
    {
      "completion_length": 32.15625,
      "epoch": 1.8397905759162305,
      "grad_norm": 20.722774505615234,
      "kl": 3.2265625,
      "learning_rate": 5.39832285115304e-07,
      "loss": 0.3362,
      "reward": 0.7307002544403076,
      "reward_std": 0.10478199645876884,
      "reward_std_all": 0.09715919755399227,
      "rewards-max/forensics_sinkhorn_reward": 0.9885175824165344,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9981365203857422,
      "rewards/forensics_sinkhorn_reward": 0.7026514410972595,
      "rewards/forensics_strict_edge_reward": 0.45269185304641724,
      "rewards/forensics_tiou_reward": 0.5675874650478363,
      "step": 879
    },
    {
      "completion_length": 54.375,
      "epoch": 1.8418848167539266,
      "grad_norm": 20.720142364501953,
      "kl": 5.46875,
      "learning_rate": 5.39308176100629e-07,
      "loss": 0.5024,
      "reward": 0.6424467265605927,
      "reward_std": 0.16601817682385445,
      "reward_std_all": 0.16530342027544975,
      "rewards-max/forensics_sinkhorn_reward": 0.9776895344257355,
      "rewards-max/forensics_strict_edge_reward": 0.9256840348243713,
      "rewards-max/forensics_tiou_reward": 0.9730700552463531,
      "rewards/forensics_sinkhorn_reward": 0.52682064473629,
      "rewards/forensics_strict_edge_reward": 0.3226427286863327,
      "rewards/forensics_tiou_reward": 0.39403946697711945,
      "step": 880
    },
    {
      "completion_length": 43.78125,
      "epoch": 1.843979057591623,
      "grad_norm": 30.119762420654297,
      "kl": 5.0625,
      "learning_rate": 5.387840670859538e-07,
      "loss": 0.402,
      "reward": 0.6442616283893585,
      "reward_std": 0.13327395170927048,
      "reward_std_all": 0.14589408785104752,
      "rewards-max/forensics_sinkhorn_reward": 0.9464631676673889,
      "rewards-max/forensics_strict_edge_reward": 0.783481240272522,
      "rewards-max/forensics_tiou_reward": 0.896709144115448,
      "rewards/forensics_sinkhorn_reward": 0.5100962519645691,
      "rewards/forensics_strict_edge_reward": 0.2511613517999649,
      "rewards/forensics_tiou_reward": 0.35598623752593994,
      "step": 881
    },
    {
      "completion_length": 61.828125,
      "epoch": 1.8460732984293193,
      "grad_norm": 31.017757415771484,
      "kl": 7.625,
      "learning_rate": 5.382599580712788e-07,
      "loss": 0.4431,
      "reward": 0.6401234865188599,
      "reward_std": 0.1380871869623661,
      "reward_std_all": 0.1516796424984932,
      "rewards-max/forensics_sinkhorn_reward": 0.9887248873710632,
      "rewards-max/forensics_strict_edge_reward": 0.9782643616199493,
      "rewards-max/forensics_tiou_reward": 0.9989528656005859,
      "rewards/forensics_sinkhorn_reward": 0.601901650428772,
      "rewards/forensics_strict_edge_reward": 0.32244114577770233,
      "rewards/forensics_tiou_reward": 0.4438712149858475,
      "step": 882
    },
    {
      "completion_length": 43.4375,
      "epoch": 1.8481675392670156,
      "grad_norm": 32.27002716064453,
      "kl": 3.8828125,
      "learning_rate": 5.377358490566037e-07,
      "loss": 0.2918,
      "reward": 0.6700767278671265,
      "reward_std": 0.11129387468099594,
      "reward_std_all": 0.12220768630504608,
      "rewards-max/forensics_sinkhorn_reward": 0.9094891250133514,
      "rewards-max/forensics_strict_edge_reward": 0.8472771048545837,
      "rewards-max/forensics_tiou_reward": 0.8353857398033142,
      "rewards/forensics_sinkhorn_reward": 0.534316211938858,
      "rewards/forensics_strict_edge_reward": 0.3479965329170227,
      "rewards/forensics_tiou_reward": 0.37013301253318787,
      "step": 883
    },
    {
      "completion_length": 51.953125,
      "epoch": 1.850261780104712,
      "grad_norm": 27.72258758544922,
      "kl": 6.796875,
      "learning_rate": 5.372117400419287e-07,
      "loss": 0.3618,
      "reward": 0.6014427542686462,
      "reward_std": 0.21315943449735641,
      "reward_std_all": 0.20712292939424515,
      "rewards-max/forensics_sinkhorn_reward": 0.9873640239238739,
      "rewards-max/forensics_strict_edge_reward": 0.927969753742218,
      "rewards-max/forensics_tiou_reward": 0.993610292673111,
      "rewards/forensics_sinkhorn_reward": 0.5203635692596436,
      "rewards/forensics_strict_edge_reward": 0.278923861682415,
      "rewards/forensics_tiou_reward": 0.3532957285642624,
      "step": 884
    },
    {
      "completion_length": 52.21875,
      "epoch": 1.8523560209424084,
      "grad_norm": 23.128623962402344,
      "kl": 5.734375,
      "learning_rate": 5.366876310272537e-07,
      "loss": 0.4288,
      "reward": 0.6097659468650818,
      "reward_std": 0.17605748772621155,
      "reward_std_all": 0.17839312553405762,
      "rewards-max/forensics_sinkhorn_reward": 0.986834317445755,
      "rewards-max/forensics_strict_edge_reward": 0.9252271056175232,
      "rewards-max/forensics_tiou_reward": 0.9915173947811127,
      "rewards/forensics_sinkhorn_reward": 0.46807774901390076,
      "rewards/forensics_strict_edge_reward": 0.2158804088830948,
      "rewards/forensics_tiou_reward": 0.2537855878472328,
      "step": 885
    },
    {
      "completion_length": 37.109375,
      "epoch": 1.8544502617801046,
      "grad_norm": 19.235721588134766,
      "kl": 4.78125,
      "learning_rate": 5.361635220125786e-07,
      "loss": 0.3027,
      "reward": 0.6535289585590363,
      "reward_std": 0.08493336290121078,
      "reward_std_all": 0.11694600433111191,
      "rewards-max/forensics_sinkhorn_reward": 0.9879721105098724,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9959854781627655,
      "rewards/forensics_sinkhorn_reward": 0.5979364812374115,
      "rewards/forensics_strict_edge_reward": 0.29779399931430817,
      "rewards/forensics_tiou_reward": 0.38870416581630707,
      "step": 886
    },
    {
      "completion_length": 75.3125,
      "epoch": 1.8565445026178011,
      "grad_norm": 20.182987213134766,
      "kl": 7.734375,
      "learning_rate": 5.356394129979035e-07,
      "loss": 0.5685,
      "reward": 0.6795392632484436,
      "reward_std": 0.104073915630579,
      "reward_std_all": 0.11307177692651749,
      "rewards-max/forensics_sinkhorn_reward": 0.9849372208118439,
      "rewards-max/forensics_strict_edge_reward": 0.820955753326416,
      "rewards-max/forensics_tiou_reward": 0.9840378165245056,
      "rewards/forensics_sinkhorn_reward": 0.5627636909484863,
      "rewards/forensics_strict_edge_reward": 0.25520357489585876,
      "rewards/forensics_tiou_reward": 0.38178861141204834,
      "step": 887
    },
    {
      "completion_length": 53.640625,
      "epoch": 1.8586387434554974,
      "grad_norm": 99.0846176147461,
      "kl": 7.5,
      "learning_rate": 5.351153039832284e-07,
      "loss": 0.5003,
      "reward": 0.5881058573722839,
      "reward_std": 0.2226724475622177,
      "reward_std_all": 0.21910127252340317,
      "rewards-max/forensics_sinkhorn_reward": 0.9853408634662628,
      "rewards-max/forensics_strict_edge_reward": 0.9567649364471436,
      "rewards-max/forensics_tiou_reward": 0.9910283982753754,
      "rewards/forensics_sinkhorn_reward": 0.4746786057949066,
      "rewards/forensics_strict_edge_reward": 0.19053608924150467,
      "rewards/forensics_tiou_reward": 0.25968270003795624,
      "step": 888
    },
    {
      "completion_length": 53.359375,
      "epoch": 1.8607329842931937,
      "grad_norm": 27.745708465576172,
      "kl": 6.015625,
      "learning_rate": 5.345911949685534e-07,
      "loss": 0.3208,
      "reward": 0.7139919102191925,
      "reward_std": 0.11396195739507675,
      "reward_std_all": 0.10693322867155075,
      "rewards-max/forensics_sinkhorn_reward": 0.9888876378536224,
      "rewards-max/forensics_strict_edge_reward": 0.9890114367008209,
      "rewards-max/forensics_tiou_reward": 0.9995945692062378,
      "rewards/forensics_sinkhorn_reward": 0.6713119149208069,
      "rewards/forensics_strict_edge_reward": 0.4257543832063675,
      "rewards/forensics_tiou_reward": 0.5629258751869202,
      "step": 889
    },
    {
      "completion_length": 64.015625,
      "epoch": 1.8628272251308902,
      "grad_norm": 20.926145553588867,
      "kl": 5.296875,
      "learning_rate": 5.340670859538783e-07,
      "loss": 0.3786,
      "reward": 0.5839616656303406,
      "reward_std": 0.19196245074272156,
      "reward_std_all": 0.207452692091465,
      "rewards-max/forensics_sinkhorn_reward": 0.9852812588214874,
      "rewards-max/forensics_strict_edge_reward": 0.9169811904430389,
      "rewards-max/forensics_tiou_reward": 0.9854869544506073,
      "rewards/forensics_sinkhorn_reward": 0.5220951437950134,
      "rewards/forensics_strict_edge_reward": 0.2421553134918213,
      "rewards/forensics_tiou_reward": 0.31553755700588226,
      "step": 890
    },
    {
      "completion_length": 93.875,
      "epoch": 1.8649214659685864,
      "grad_norm": 21.114336013793945,
      "kl": 6.734375,
      "learning_rate": 5.335429769392034e-07,
      "loss": 0.4144,
      "reward": 0.6086066961288452,
      "reward_std": 0.1756088137626648,
      "reward_std_all": 0.18386565148830414,
      "rewards-max/forensics_sinkhorn_reward": 0.9756518006324768,
      "rewards-max/forensics_strict_edge_reward": 0.9032877683639526,
      "rewards-max/forensics_tiou_reward": 0.9509862065315247,
      "rewards/forensics_sinkhorn_reward": 0.49663999676704407,
      "rewards/forensics_strict_edge_reward": 0.22443781048059464,
      "rewards/forensics_tiou_reward": 0.2835260182619095,
      "step": 891
    },
    {
      "completion_length": 60.390625,
      "epoch": 1.8670157068062827,
      "grad_norm": 19.790149688720703,
      "kl": 4.2578125,
      "learning_rate": 5.330188679245284e-07,
      "loss": 0.2132,
      "reward": 0.6403405964374542,
      "reward_std": 0.16792014241218567,
      "reward_std_all": 0.16994530707597733,
      "rewards-max/forensics_sinkhorn_reward": 0.9864315688610077,
      "rewards-max/forensics_strict_edge_reward": 0.9169811904430389,
      "rewards-max/forensics_tiou_reward": 0.9899486303329468,
      "rewards/forensics_sinkhorn_reward": 0.5422268211841583,
      "rewards/forensics_strict_edge_reward": 0.28223834931850433,
      "rewards/forensics_tiou_reward": 0.3610169440507889,
      "step": 892
    },
    {
      "completion_length": 64.890625,
      "epoch": 1.8691099476439792,
      "grad_norm": 16.655363082885742,
      "kl": 5.28125,
      "learning_rate": 5.324947589098532e-07,
      "loss": 0.3818,
      "reward": 0.6684193313121796,
      "reward_std": 0.1482723131775856,
      "reward_std_all": 0.14504383504390717,
      "rewards-max/forensics_sinkhorn_reward": 0.9876773357391357,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9948242902755737,
      "rewards/forensics_sinkhorn_reward": 0.5880582332611084,
      "rewards/forensics_strict_edge_reward": 0.33498090505599976,
      "rewards/forensics_tiou_reward": 0.4270658642053604,
      "step": 893
    },
    {
      "completion_length": 43.921875,
      "epoch": 1.8712041884816752,
      "grad_norm": 21.436674118041992,
      "kl": 4.171875,
      "learning_rate": 5.319706498951782e-07,
      "loss": 0.3135,
      "reward": 0.6346213817596436,
      "reward_std": 0.1995556429028511,
      "reward_std_all": 0.18755878508090973,
      "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": 0.5370962619781494,
      "rewards/forensics_strict_edge_reward": 0.28003208339214325,
      "rewards/forensics_tiou_reward": 0.37099607288837433,
      "step": 894
    },
    {
      "completion_length": 43.609375,
      "epoch": 1.8732984293193717,
      "grad_norm": 22.647567749023438,
      "kl": 4.140625,
      "learning_rate": 5.314465408805031e-07,
      "loss": 0.3291,
      "reward": 0.6573397517204285,
      "reward_std": 0.10177738964557648,
      "reward_std_all": 0.12255531921982765,
      "rewards-max/forensics_sinkhorn_reward": 0.9675260186195374,
      "rewards-max/forensics_strict_edge_reward": 0.8719756007194519,
      "rewards-max/forensics_tiou_reward": 0.9516711235046387,
      "rewards/forensics_sinkhorn_reward": 0.5269322842359543,
      "rewards/forensics_strict_edge_reward": 0.27445195615291595,
      "rewards/forensics_tiou_reward": 0.31925320625305176,
      "step": 895
    },
    {
      "completion_length": 57.09375,
      "epoch": 1.875392670157068,
      "grad_norm": 13.021939277648926,
      "kl": 4.515625,
      "learning_rate": 5.309224318658281e-07,
      "loss": 0.3504,
      "reward": 0.6811511814594269,
      "reward_std": 0.17773232609033585,
      "reward_std_all": 0.1605284884572029,
      "rewards-max/forensics_sinkhorn_reward": 0.9768860042095184,
      "rewards-max/forensics_strict_edge_reward": 0.9782643616199493,
      "rewards-max/forensics_tiou_reward": 0.9740161001682281,
      "rewards/forensics_sinkhorn_reward": 0.6222655773162842,
      "rewards/forensics_strict_edge_reward": 0.3983910381793976,
      "rewards/forensics_tiou_reward": 0.444828599691391,
      "step": 896
    },
    {
      "completion_length": 36.171875,
      "epoch": 1.8774869109947643,
      "grad_norm": 15.756429672241211,
      "kl": 2.59765625,
      "learning_rate": 5.30398322851153e-07,
      "loss": 0.2297,
      "reward": 0.6718061566352844,
      "reward_std": 0.06553313136100769,
      "reward_std_all": 0.09772377461194992,
      "rewards-max/forensics_sinkhorn_reward": 0.9880141317844391,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9961616396903992,
      "rewards/forensics_sinkhorn_reward": 0.576737105846405,
      "rewards/forensics_strict_edge_reward": 0.3075552433729172,
      "rewards/forensics_tiou_reward": 0.349240243434906,
      "step": 897
    },
    {
      "completion_length": 61.328125,
      "epoch": 1.8795811518324608,
      "grad_norm": 18.240482330322266,
      "kl": 3.8359375,
      "learning_rate": 5.29874213836478e-07,
      "loss": 0.1741,
      "reward": 0.5797562003135681,
      "reward_std": 0.14094045013189316,
      "reward_std_all": 0.17879020422697067,
      "rewards-max/forensics_sinkhorn_reward": 0.8926485478878021,
      "rewards-max/forensics_strict_edge_reward": 0.7455491125583649,
      "rewards-max/forensics_tiou_reward": 0.8031092584133148,
      "rewards/forensics_sinkhorn_reward": 0.48073677718639374,
      "rewards/forensics_strict_edge_reward": 0.20996107161045074,
      "rewards/forensics_tiou_reward": 0.25427091121673584,
      "step": 898
    },
    {
      "completion_length": 66.734375,
      "epoch": 1.881675392670157,
      "grad_norm": 25.56682586669922,
      "kl": 5.671875,
      "learning_rate": 5.293501048218028e-07,
      "loss": 0.3768,
      "reward": 0.693750649690628,
      "reward_std": 0.13880788534879684,
      "reward_std_all": 0.12944557517766953,
      "rewards-max/forensics_sinkhorn_reward": 0.9880664348602295,
      "rewards-max/forensics_strict_edge_reward": 0.9364311099052429,
      "rewards-max/forensics_tiou_reward": 0.9963579773902893,
      "rewards/forensics_sinkhorn_reward": 0.6213966906070709,
      "rewards/forensics_strict_edge_reward": 0.4008229076862335,
      "rewards/forensics_tiou_reward": 0.444866880774498,
      "step": 899
    },
    {
      "completion_length": 46.8125,
      "epoch": 1.8837696335078533,
      "grad_norm": 24.34605598449707,
      "kl": 4.203125,
      "learning_rate": 5.288259958071278e-07,
      "loss": 0.2699,
      "reward": 0.6500042080879211,
      "reward_std": 0.13754579052329063,
      "reward_std_all": 0.1445348933339119,
      "rewards-max/forensics_sinkhorn_reward": 0.9739023149013519,
      "rewards-max/forensics_strict_edge_reward": 0.9464850425720215,
      "rewards-max/forensics_tiou_reward": 0.9661313891410828,
      "rewards/forensics_sinkhorn_reward": 0.5695448815822601,
      "rewards/forensics_strict_edge_reward": 0.299983486533165,
      "rewards/forensics_tiou_reward": 0.35520949959754944,
      "step": 900
    },
    {
      "completion_length": 33.40625,
      "epoch": 1.8858638743455498,
      "grad_norm": 19.132522583007812,
      "kl": 2.59765625,
      "learning_rate": 5.283018867924528e-07,
      "loss": 0.1554,
      "reward": 0.6078625023365021,
      "reward_std": 0.1660318225622177,
      "reward_std_all": 0.17645590752363205,
      "rewards-max/forensics_sinkhorn_reward": 0.9741933047771454,
      "rewards-max/forensics_strict_edge_reward": 0.9252271056175232,
      "rewards-max/forensics_tiou_reward": 0.9664545357227325,
      "rewards/forensics_sinkhorn_reward": 0.47247619926929474,
      "rewards/forensics_strict_edge_reward": 0.2357330173254013,
      "rewards/forensics_tiou_reward": 0.2648300528526306,
      "step": 901
    },
    {
      "completion_length": 78.984375,
      "epoch": 1.887958115183246,
      "grad_norm": 42.640869140625,
      "kl": 6.203125,
      "learning_rate": 5.277777777777777e-07,
      "loss": 0.4599,
      "reward": 0.6191682815551758,
      "reward_std": 0.18450769782066345,
      "reward_std_all": 0.1880985125899315,
      "rewards-max/forensics_sinkhorn_reward": 0.9879379570484161,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9958563148975372,
      "rewards/forensics_sinkhorn_reward": 0.5535208880901337,
      "rewards/forensics_strict_edge_reward": 0.3205241709947586,
      "rewards/forensics_tiou_reward": 0.4200844317674637,
      "step": 902
    },
    {
      "completion_length": 89.671875,
      "epoch": 1.8900523560209423,
      "grad_norm": 31.37417221069336,
      "kl": 7.71875,
      "learning_rate": 5.272536687631028e-07,
      "loss": 0.474,
      "reward": 0.5673942863941193,
      "reward_std": 0.21584177762269974,
      "reward_std_all": 0.2306167483329773,
      "rewards-max/forensics_sinkhorn_reward": 0.9683980345726013,
      "rewards-max/forensics_strict_edge_reward": 0.9375866651535034,
      "rewards-max/forensics_tiou_reward": 0.9550569951534271,
      "rewards/forensics_sinkhorn_reward": 0.43344807624816895,
      "rewards/forensics_strict_edge_reward": 0.32083791494369507,
      "rewards/forensics_tiou_reward": 0.31326302886009216,
      "step": 903
    },
    {
      "completion_length": 39.640625,
      "epoch": 1.8921465968586388,
      "grad_norm": 24.357105255126953,
      "kl": 4.1953125,
      "learning_rate": 5.267295597484277e-07,
      "loss": 0.3493,
      "reward": 0.6736678183078766,
      "reward_std": 0.13249026238918304,
      "reward_std_all": 0.1391715407371521,
      "rewards-max/forensics_sinkhorn_reward": 0.987227588891983,
      "rewards-max/forensics_strict_edge_reward": 0.9169811904430389,
      "rewards-max/forensics_tiou_reward": 0.9930670261383057,
      "rewards/forensics_sinkhorn_reward": 0.5399187505245209,
      "rewards/forensics_strict_edge_reward": 0.2695562615990639,
      "rewards/forensics_tiou_reward": 0.3453843593597412,
      "step": 904
    },
    {
      "completion_length": 59.90625,
      "epoch": 1.8942408376963351,
      "grad_norm": 27.67887306213379,
      "kl": 6.703125,
      "learning_rate": 5.262054507337526e-07,
      "loss": 0.4046,
      "reward": 0.6275554299354553,
      "reward_std": 0.20929083228111267,
      "reward_std_all": 0.19959590584039688,
      "rewards-max/forensics_sinkhorn_reward": 0.9623541533946991,
      "rewards-max/forensics_strict_edge_reward": 0.9149472117424011,
      "rewards-max/forensics_tiou_reward": 0.9405130743980408,
      "rewards/forensics_sinkhorn_reward": 0.520091325044632,
      "rewards/forensics_strict_edge_reward": 0.29433903098106384,
      "rewards/forensics_tiou_reward": 0.34484240412712097,
      "step": 905
    },
    {
      "completion_length": 69.421875,
      "epoch": 1.8963350785340314,
      "grad_norm": 30.788774490356445,
      "kl": 5.3984375,
      "learning_rate": 5.256813417190775e-07,
      "loss": 0.4319,
      "reward": 0.6601437926292419,
      "reward_std": 0.2185760736465454,
      "reward_std_all": 0.1950492262840271,
      "rewards-max/forensics_sinkhorn_reward": 0.9713701009750366,
      "rewards-max/forensics_strict_edge_reward": 0.9158510267734528,
      "rewards-max/forensics_tiou_reward": 0.9579468667507172,
      "rewards/forensics_sinkhorn_reward": 0.5220654606819153,
      "rewards/forensics_strict_edge_reward": 0.36550532281398773,
      "rewards/forensics_tiou_reward": 0.3795766085386276,
      "step": 906
    },
    {
      "completion_length": 74.984375,
      "epoch": 1.8984293193717279,
      "grad_norm": 21.265548706054688,
      "kl": 6.9375,
      "learning_rate": 5.251572327044025e-07,
      "loss": 0.3009,
      "reward": 0.6927663087844849,
      "reward_std": 0.11719579249620438,
      "reward_std_all": 0.1163737028837204,
      "rewards-max/forensics_sinkhorn_reward": 0.9856821000576019,
      "rewards-max/forensics_strict_edge_reward": 0.9574736058712006,
      "rewards-max/forensics_tiou_reward": 0.9870495796203613,
      "rewards/forensics_sinkhorn_reward": 0.5551430284976959,
      "rewards/forensics_strict_edge_reward": 0.31087857484817505,
      "rewards/forensics_tiou_reward": 0.3782305121421814,
      "step": 907
    },
    {
      "completion_length": 148.6875,
      "epoch": 1.900523560209424,
      "grad_norm": 40.49901580810547,
      "kl": 11.8125,
      "learning_rate": 5.246331236897275e-07,
      "loss": 0.7222,
      "reward": 0.538163423538208,
      "reward_std": 0.22508905827999115,
      "reward_std_all": 0.24171270430088043,
      "rewards-max/forensics_sinkhorn_reward": 0.9536722004413605,
      "rewards-max/forensics_strict_edge_reward": 0.8317028284072876,
      "rewards-max/forensics_tiou_reward": 0.9241330921649933,
      "rewards/forensics_sinkhorn_reward": 0.43293602764606476,
      "rewards/forensics_strict_edge_reward": 0.217504620552063,
      "rewards/forensics_tiou_reward": 0.24945317953824997,
      "step": 908
    },
    {
      "completion_length": 83.9375,
      "epoch": 1.9026178010471204,
      "grad_norm": 29.54384994506836,
      "kl": 9.9375,
      "learning_rate": 5.241090146750524e-07,
      "loss": 0.5136,
      "reward": 0.5790545046329498,
      "reward_std": 0.2222602665424347,
      "reward_std_all": 0.21853332966566086,
      "rewards-max/forensics_sinkhorn_reward": 0.9243830144405365,
      "rewards-max/forensics_strict_edge_reward": 0.8893801271915436,
      "rewards-max/forensics_tiou_reward": 0.8672510981559753,
      "rewards/forensics_sinkhorn_reward": 0.4439432621002197,
      "rewards/forensics_strict_edge_reward": 0.208710178732872,
      "rewards/forensics_tiou_reward": 0.24578344076871872,
      "step": 909
    },
    {
      "completion_length": 70.046875,
      "epoch": 1.904712041884817,
      "grad_norm": 26.51218605041504,
      "kl": 6.265625,
      "learning_rate": 5.235849056603774e-07,
      "loss": 0.3837,
      "reward": 0.5931856036186218,
      "reward_std": 0.22192196547985077,
      "reward_std_all": 0.2188742458820343,
      "rewards-max/forensics_sinkhorn_reward": 0.9787915647029877,
      "rewards-max/forensics_strict_edge_reward": 0.9567649364471436,
      "rewards-max/forensics_tiou_reward": 0.9779334664344788,
      "rewards/forensics_sinkhorn_reward": 0.5187416076660156,
      "rewards/forensics_strict_edge_reward": 0.31625454872846603,
      "rewards/forensics_tiou_reward": 0.34835293889045715,
      "step": 910
    },
    {
      "completion_length": 42.609375,
      "epoch": 1.906806282722513,
      "grad_norm": 25.70760726928711,
      "kl": 2.80078125,
      "learning_rate": 5.230607966457022e-07,
      "loss": 0.4089,
      "reward": 0.6577344238758087,
      "reward_std": 0.1481654942035675,
      "reward_std_all": 0.14847500622272491,
      "rewards-max/forensics_sinkhorn_reward": 0.9624739289283752,
      "rewards-max/forensics_strict_edge_reward": 0.7883436977863312,
      "rewards-max/forensics_tiou_reward": 0.9422109425067902,
      "rewards/forensics_sinkhorn_reward": 0.5633711814880371,
      "rewards/forensics_strict_edge_reward": 0.2635662406682968,
      "rewards/forensics_tiou_reward": 0.32809048891067505,
      "step": 911
    },
    {
      "completion_length": 57.921875,
      "epoch": 1.9089005235602095,
      "grad_norm": 50.50601577758789,
      "kl": 5.3125,
      "learning_rate": 5.225366876310272e-07,
      "loss": 0.3581,
      "reward": 0.6431468427181244,
      "reward_std": 0.18459369987249374,
      "reward_std_all": 0.18295635282993317,
      "rewards-max/forensics_sinkhorn_reward": 0.9675886929035187,
      "rewards-max/forensics_strict_edge_reward": 0.8320213258266449,
      "rewards-max/forensics_tiou_reward": 0.9475508630275726,
      "rewards/forensics_sinkhorn_reward": 0.5597606003284454,
      "rewards/forensics_strict_edge_reward": 0.2855636477470398,
      "rewards/forensics_tiou_reward": 0.3779221624135971,
      "step": 912
    },
    {
      "completion_length": 91.203125,
      "epoch": 1.9109947643979057,
      "grad_norm": 25.077062606811523,
      "kl": 7.15625,
      "learning_rate": 5.220125786163521e-07,
      "loss": 0.442,
      "reward": 0.5863902866840363,
      "reward_std": 0.2137186899781227,
      "reward_std_all": 0.22235778719186783,
      "rewards-max/forensics_sinkhorn_reward": 0.9882677495479584,
      "rewards-max/forensics_strict_edge_reward": 0.9782643616199493,
      "rewards-max/forensics_tiou_reward": 0.9971550703048706,
      "rewards/forensics_sinkhorn_reward": 0.5236285179853439,
      "rewards/forensics_strict_edge_reward": 0.32776860892772675,
      "rewards/forensics_tiou_reward": 0.38001126050949097,
      "step": 913
    },
    {
      "completion_length": 93.484375,
      "epoch": 1.913089005235602,
      "grad_norm": 18.238584518432617,
      "kl": 5.578125,
      "learning_rate": 5.214884696016771e-07,
      "loss": 0.2939,
      "reward": 0.5676820278167725,
      "reward_std": 0.21064673364162445,
      "reward_std_all": 0.22100813686847687,
      "rewards-max/forensics_sinkhorn_reward": 0.9883580803871155,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9975085258483887,
      "rewards/forensics_sinkhorn_reward": 0.46310998499393463,
      "rewards/forensics_strict_edge_reward": 0.3077233210206032,
      "rewards/forensics_tiou_reward": 0.3428874760866165,
      "step": 914
    },
    {
      "completion_length": 91.71875,
      "epoch": 1.9151832460732985,
      "grad_norm": 20.406068801879883,
      "kl": 5.53125,
      "learning_rate": 5.20964360587002e-07,
      "loss": 0.2555,
      "reward": 0.6119487881660461,
      "reward_std": 0.16413472592830658,
      "reward_std_all": 0.1760188192129135,
      "rewards-max/forensics_sinkhorn_reward": 0.971521645784378,
      "rewards-max/forensics_strict_edge_reward": 0.905340164899826,
      "rewards-max/forensics_tiou_reward": 0.9616407454013824,
      "rewards/forensics_sinkhorn_reward": 0.49016351997852325,
      "rewards/forensics_strict_edge_reward": 0.19888121634721756,
      "rewards/forensics_tiou_reward": 0.2789447754621506,
      "step": 915
    },
    {
      "completion_length": 75.90625,
      "epoch": 1.9172774869109948,
      "grad_norm": 82.10935974121094,
      "kl": 10.25,
      "learning_rate": 5.204402515723271e-07,
      "loss": 0.4959,
      "reward": 0.552562803030014,
      "reward_std": 0.23433710634708405,
      "reward_std_all": 0.24833957850933075,
      "rewards-max/forensics_sinkhorn_reward": 0.9492090046405792,
      "rewards-max/forensics_strict_edge_reward": 0.8165846168994904,
      "rewards-max/forensics_tiou_reward": 0.9128231704235077,
      "rewards/forensics_sinkhorn_reward": 0.4643630236387253,
      "rewards/forensics_strict_edge_reward": 0.22324039787054062,
      "rewards/forensics_tiou_reward": 0.2942235916852951,
      "step": 916
    },
    {
      "completion_length": 80.46875,
      "epoch": 1.919371727748691,
      "grad_norm": 30.28687858581543,
      "kl": 4.828125,
      "learning_rate": 5.19916142557652e-07,
      "loss": 0.2902,
      "reward": 0.63922119140625,
      "reward_std": 0.1960030347108841,
      "reward_std_all": 0.18556586652994156,
      "rewards-max/forensics_sinkhorn_reward": 0.9738070666790009,
      "rewards-max/forensics_strict_edge_reward": 0.9075756669044495,
      "rewards-max/forensics_tiou_reward": 0.9652106463909149,
      "rewards/forensics_sinkhorn_reward": 0.544544130563736,
      "rewards/forensics_strict_edge_reward": 0.31255607306957245,
      "rewards/forensics_tiou_reward": 0.3700237274169922,
      "step": 917
    },
    {
      "completion_length": 58.046875,
      "epoch": 1.9214659685863875,
      "grad_norm": 16.449941635131836,
      "kl": 2.8671875,
      "learning_rate": 5.193920335429769e-07,
      "loss": 0.1778,
      "reward": 0.71101513504982,
      "reward_std": 0.11613625101745129,
      "reward_std_all": 0.10909752175211906,
      "rewards-max/forensics_sinkhorn_reward": 0.9878372550010681,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9954541027545929,
      "rewards/forensics_sinkhorn_reward": 0.5940185189247131,
      "rewards/forensics_strict_edge_reward": 0.35646890103816986,
      "rewards/forensics_tiou_reward": 0.40251851081848145,
      "step": 918
    },
    {
      "completion_length": 69.578125,
      "epoch": 1.9235602094240838,
      "grad_norm": 26.114484786987305,
      "kl": 2.9375,
      "learning_rate": 5.188679245283019e-07,
      "loss": 0.1663,
      "reward": 0.5781527161598206,
      "reward_std": 0.173491008579731,
      "reward_std_all": 0.1921803429722786,
      "rewards-max/forensics_sinkhorn_reward": 0.9829320907592773,
      "rewards-max/forensics_strict_edge_reward": 0.8137359023094177,
      "rewards-max/forensics_tiou_reward": 0.9763220846652985,
      "rewards/forensics_sinkhorn_reward": 0.4647044837474823,
      "rewards/forensics_strict_edge_reward": 0.21435753256082535,
      "rewards/forensics_tiou_reward": 0.2779756411910057,
      "step": 919
    },
    {
      "completion_length": 93.46875,
      "epoch": 1.92565445026178,
      "grad_norm": 31.927040100097656,
      "kl": 6.90625,
      "learning_rate": 5.183438155136268e-07,
      "loss": 0.5988,
      "reward": 0.5318958461284637,
      "reward_std": 0.16510984301567078,
      "reward_std_all": 0.20751766860485077,
      "rewards-max/forensics_sinkhorn_reward": 0.9547113180160522,
      "rewards-max/forensics_strict_edge_reward": 0.8237898349761963,
      "rewards-max/forensics_tiou_reward": 0.9184156656265259,
      "rewards/forensics_sinkhorn_reward": 0.4020075500011444,
      "rewards/forensics_strict_edge_reward": 0.1684878244996071,
      "rewards/forensics_tiou_reward": 0.19773997366428375,
      "step": 920
    },
    {
      "completion_length": 96.140625,
      "epoch": 1.9277486910994766,
      "grad_norm": 24.500938415527344,
      "kl": 5.46875,
      "learning_rate": 5.178197064989518e-07,
      "loss": 0.431,
      "reward": 0.5726402699947357,
      "reward_std": 0.244765043258667,
      "reward_std_all": 0.23498088866472244,
      "rewards-max/forensics_sinkhorn_reward": 0.9868092238903046,
      "rewards-max/forensics_strict_edge_reward": 0.9464850425720215,
      "rewards-max/forensics_tiou_reward": 0.9914053380489349,
      "rewards/forensics_sinkhorn_reward": 0.4993647187948227,
      "rewards/forensics_strict_edge_reward": 0.25759341567754745,
      "rewards/forensics_tiou_reward": 0.34677407145500183,
      "step": 921
    },
    {
      "completion_length": 50.828125,
      "epoch": 1.9298429319371728,
      "grad_norm": 29.442468643188477,
      "kl": 2.578125,
      "learning_rate": 5.172955974842767e-07,
      "loss": 0.1748,
      "reward": 0.6087378561496735,
      "reward_std": 0.158823661506176,
      "reward_std_all": 0.17038656771183014,
      "rewards-max/forensics_sinkhorn_reward": 0.984645813703537,
      "rewards-max/forensics_strict_edge_reward": 0.8404377698898315,
      "rewards-max/forensics_tiou_reward": 0.9829986393451691,
      "rewards/forensics_sinkhorn_reward": 0.510613352060318,
      "rewards/forensics_strict_edge_reward": 0.24120885133743286,
      "rewards/forensics_tiou_reward": 0.3140586018562317,
      "step": 922
    },
    {
      "completion_length": 88.890625,
      "epoch": 1.931937172774869,
      "grad_norm": 25.93185043334961,
      "kl": 5.3671875,
      "learning_rate": 5.167714884696016e-07,
      "loss": 0.2702,
      "reward": 0.6367720365524292,
      "reward_std": 0.16626883298158646,
      "reward_std_all": 0.17223957180976868,
      "rewards-max/forensics_sinkhorn_reward": 0.9877484440803528,
      "rewards-max/forensics_strict_edge_reward": 0.9565287232398987,
      "rewards-max/forensics_tiou_reward": 0.9951047301292419,
      "rewards/forensics_sinkhorn_reward": 0.5679985284805298,
      "rewards/forensics_strict_edge_reward": 0.35665176808834076,
      "rewards/forensics_tiou_reward": 0.42349591851234436,
      "step": 923
    },
    {
      "completion_length": 98.09375,
      "epoch": 1.9340314136125656,
      "grad_norm": 419.2882995605469,
      "kl": 23.171875,
      "learning_rate": 5.162473794549266e-07,
      "loss": 1.1594,
      "reward": 0.6456432938575745,
      "reward_std": 0.15937935188412666,
      "reward_std_all": 0.16060201823711395,
      "rewards-max/forensics_sinkhorn_reward": 0.9853411018848419,
      "rewards-max/forensics_strict_edge_reward": 0.8698340952396393,
      "rewards-max/forensics_tiou_reward": 0.9857126176357269,
      "rewards/forensics_sinkhorn_reward": 0.5136386156082153,
      "rewards/forensics_strict_edge_reward": 0.2632272392511368,
      "rewards/forensics_tiou_reward": 0.3368513733148575,
      "step": 924
    },
    {
      "completion_length": 64.859375,
      "epoch": 1.9361256544502616,
      "grad_norm": 25.510669708251953,
      "kl": 5.21875,
      "learning_rate": 5.157232704402515e-07,
      "loss": 0.3566,
      "reward": 0.6002289950847626,
      "reward_std": 0.23100031912326813,
      "reward_std_all": 0.2225274071097374,
      "rewards-max/forensics_sinkhorn_reward": 0.9886201620101929,
      "rewards-max/forensics_strict_edge_reward": 0.9782643616199493,
      "rewards-max/forensics_tiou_reward": 0.9985401332378387,
      "rewards/forensics_sinkhorn_reward": 0.4767673909664154,
      "rewards/forensics_strict_edge_reward": 0.26213137805461884,
      "rewards/forensics_tiou_reward": 0.30966461449861526,
      "step": 925
    },
    {
      "completion_length": 58.765625,
      "epoch": 1.9382198952879581,
      "grad_norm": 23.932668685913086,
      "kl": 4.640625,
      "learning_rate": 5.151991614255765e-07,
      "loss": 0.3151,
      "reward": 0.6142500638961792,
      "reward_std": 0.14163096249103546,
      "reward_std_all": 0.1646685004234314,
      "rewards-max/forensics_sinkhorn_reward": 0.9771960377693176,
      "rewards-max/forensics_strict_edge_reward": 0.8060585260391235,
      "rewards-max/forensics_tiou_reward": 0.9545061886310577,
      "rewards/forensics_sinkhorn_reward": 0.48421038687229156,
      "rewards/forensics_strict_edge_reward": 0.22150468826293945,
      "rewards/forensics_tiou_reward": 0.24584153294563293,
      "step": 926
    },
    {
      "completion_length": 46.859375,
      "epoch": 1.9403141361256544,
      "grad_norm": 31.033443450927734,
      "kl": 6.234375,
      "learning_rate": 5.146750524109015e-07,
      "loss": 0.4974,
      "reward": 0.6433509290218353,
      "reward_std": 0.09607931971549988,
      "reward_std_all": 0.1283179558813572,
      "rewards-max/forensics_sinkhorn_reward": 0.9880184531211853,
      "rewards-max/forensics_strict_edge_reward": 0.9782643616199493,
      "rewards-max/forensics_tiou_reward": 0.9961759150028229,
      "rewards/forensics_sinkhorn_reward": 0.5194802284240723,
      "rewards/forensics_strict_edge_reward": 0.2572869658470154,
      "rewards/forensics_tiou_reward": 0.3203981816768646,
      "step": 927
    },
    {
      "completion_length": 62.21875,
      "epoch": 1.9424083769633507,
      "grad_norm": 19.324596405029297,
      "kl": 4.9375,
      "learning_rate": 5.141509433962265e-07,
      "loss": 0.3092,
      "reward": 0.6079261302947998,
      "reward_std": 0.13418195396661758,
      "reward_std_all": 0.15916289016604424,
      "rewards-max/forensics_sinkhorn_reward": 0.9649558663368225,
      "rewards-max/forensics_strict_edge_reward": 0.9003686904907227,
      "rewards-max/forensics_tiou_reward": 0.9442928433418274,
      "rewards/forensics_sinkhorn_reward": 0.494440034031868,
      "rewards/forensics_strict_edge_reward": 0.238882914185524,
      "rewards/forensics_tiou_reward": 0.28159988671541214,
      "step": 928
    },
    {
      "completion_length": 49.796875,
      "epoch": 1.9445026178010472,
      "grad_norm": 36.35917282104492,
      "kl": 4.796875,
      "learning_rate": 5.136268343815513e-07,
      "loss": 0.4207,
      "reward": 0.6804943978786469,
      "reward_std": 0.1661982536315918,
      "reward_std_all": 0.15182919055223465,
      "rewards-max/forensics_sinkhorn_reward": 0.9877756237983704,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9952112138271332,
      "rewards/forensics_sinkhorn_reward": 0.5778321921825409,
      "rewards/forensics_strict_edge_reward": 0.3901509642601013,
      "rewards/forensics_tiou_reward": 0.44347673654556274,
      "step": 929
    },
    {
      "completion_length": 62.4375,
      "epoch": 1.9465968586387434,
      "grad_norm": 22.649267196655273,
      "kl": 6.734375,
      "learning_rate": 5.131027253668763e-07,
      "loss": 0.5007,
      "reward": 0.5777848958969116,
      "reward_std": 0.1415194533765316,
      "reward_std_all": 0.1753169372677803,
      "rewards-max/forensics_sinkhorn_reward": 0.9569208920001984,
      "rewards-max/forensics_strict_edge_reward": 0.8893801271915436,
      "rewards-max/forensics_tiou_reward": 0.928097665309906,
      "rewards/forensics_sinkhorn_reward": 0.4496042877435684,
      "rewards/forensics_strict_edge_reward": 0.20676221698522568,
      "rewards/forensics_tiou_reward": 0.275660865008831,
      "step": 930
    },
    {
      "completion_length": 66.9375,
      "epoch": 1.9486910994764397,
      "grad_norm": 29.937162399291992,
      "kl": 7.875,
      "learning_rate": 5.125786163522012e-07,
      "loss": 0.4909,
      "reward": 0.5499870777130127,
      "reward_std": 0.26143592596054077,
      "reward_std_all": 0.2603250741958618,
      "rewards-max/forensics_sinkhorn_reward": 0.9846166968345642,
      "rewards-max/forensics_strict_edge_reward": 0.8553276658058167,
      "rewards-max/forensics_tiou_reward": 0.9828915596008301,
      "rewards/forensics_sinkhorn_reward": 0.44656361639499664,
      "rewards/forensics_strict_edge_reward": 0.24515068531036377,
      "rewards/forensics_tiou_reward": 0.286267951130867,
      "step": 931
    },
    {
      "completion_length": 80.0625,
      "epoch": 1.9507853403141362,
      "grad_norm": 17.322513580322266,
      "kl": 7.640625,
      "learning_rate": 5.120545073375262e-07,
      "loss": 0.5521,
      "reward": 0.60548135638237,
      "reward_std": 0.17517954111099243,
      "reward_std_all": 0.187327578663826,
      "rewards-max/forensics_sinkhorn_reward": 0.9790380299091339,
      "rewards-max/forensics_strict_edge_reward": 0.9565287232398987,
      "rewards-max/forensics_tiou_reward": 0.9733319878578186,
      "rewards/forensics_sinkhorn_reward": 0.48090341687202454,
      "rewards/forensics_strict_edge_reward": 0.2583700120449066,
      "rewards/forensics_tiou_reward": 0.28821510076522827,
      "step": 932
    },
    {
      "completion_length": 84.15625,
      "epoch": 1.9528795811518325,
      "grad_norm": 26.080198287963867,
      "kl": 7.6875,
      "learning_rate": 5.115303983228512e-07,
      "loss": 0.5563,
      "reward": 0.6407525539398193,
      "reward_std": 0.16258706152439117,
      "reward_std_all": 0.16104251146316528,
      "rewards-max/forensics_sinkhorn_reward": 0.9498199820518494,
      "rewards-max/forensics_strict_edge_reward": 0.9425031542778015,
      "rewards-max/forensics_tiou_reward": 0.918828159570694,
      "rewards/forensics_sinkhorn_reward": 0.49956320226192474,
      "rewards/forensics_strict_edge_reward": 0.2696092873811722,
      "rewards/forensics_tiou_reward": 0.33011089265346527,
      "step": 933
    },
    {
      "completion_length": 29.96875,
      "epoch": 1.9549738219895287,
      "grad_norm": 19.428468704223633,
      "kl": 4.625,
      "learning_rate": 5.110062893081761e-07,
      "loss": 0.4769,
      "reward": 0.6023682355880737,
      "reward_std": 0.15385133773088455,
      "reward_std_all": 0.1747036650776863,
      "rewards-max/forensics_sinkhorn_reward": 0.9862821102142334,
      "rewards-max/forensics_strict_edge_reward": 0.946017861366272,
      "rewards-max/forensics_tiou_reward": 0.9893350303173065,
      "rewards/forensics_sinkhorn_reward": 0.49466651678085327,
      "rewards/forensics_strict_edge_reward": 0.2804746776819229,
      "rewards/forensics_tiou_reward": 0.3558690845966339,
      "step": 934
    },
    {
      "completion_length": 73.265625,
      "epoch": 1.9570680628272252,
      "grad_norm": 22.622154235839844,
      "kl": 6.5625,
      "learning_rate": 5.10482180293501e-07,
      "loss": 0.5079,
      "reward": 0.6134576201438904,
      "reward_std": 0.18402543663978577,
      "reward_std_all": 0.1820664256811142,
      "rewards-max/forensics_sinkhorn_reward": 0.9638190567493439,
      "rewards-max/forensics_strict_edge_reward": 0.9169811904430389,
      "rewards-max/forensics_tiou_reward": 0.9454345107078552,
      "rewards/forensics_sinkhorn_reward": 0.4996163100004196,
      "rewards/forensics_strict_edge_reward": 0.23767560720443726,
      "rewards/forensics_tiou_reward": 0.2791562080383301,
      "step": 935
    },
    {
      "completion_length": 49.609375,
      "epoch": 1.9591623036649215,
      "grad_norm": 22.27456283569336,
      "kl": 5.875,
      "learning_rate": 5.099580712788259e-07,
      "loss": 0.3717,
      "reward": 0.600279301404953,
      "reward_std": 0.17240753769874573,
      "reward_std_all": 0.184517003595829,
      "rewards-max/forensics_sinkhorn_reward": 0.9872286915779114,
      "rewards-max/forensics_strict_edge_reward": 0.9364311099052429,
      "rewards-max/forensics_tiou_reward": 0.9930562973022461,
      "rewards/forensics_sinkhorn_reward": 0.5180507600307465,
      "rewards/forensics_strict_edge_reward": 0.2732691764831543,
      "rewards/forensics_tiou_reward": 0.3268894702196121,
      "step": 936
    },
    {
      "completion_length": 55.59375,
      "epoch": 1.9612565445026178,
      "grad_norm": 45.45783996582031,
      "kl": 6.5,
      "learning_rate": 5.094339622641509e-07,
      "loss": 0.4013,
      "reward": 0.6328603327274323,
      "reward_std": 0.1819436475634575,
      "reward_std_all": 0.18733881413936615,
      "rewards-max/forensics_sinkhorn_reward": 0.9886631667613983,
      "rewards-max/forensics_strict_edge_reward": 0.9782643616199493,
      "rewards-max/forensics_tiou_reward": 0.9987096786499023,
      "rewards/forensics_sinkhorn_reward": 0.5616495609283447,
      "rewards/forensics_strict_edge_reward": 0.3141380697488785,
      "rewards/forensics_tiou_reward": 0.3877260088920593,
      "step": 937
    },
    {
      "completion_length": 51.0625,
      "epoch": 1.9633507853403143,
      "grad_norm": 24.199365615844727,
      "kl": 5.09375,
      "learning_rate": 5.089098532494758e-07,
      "loss": 0.3194,
      "reward": 0.6699071228504181,
      "reward_std": 0.13004855811595917,
      "reward_std_all": 0.1317719891667366,
      "rewards-max/forensics_sinkhorn_reward": 0.9878827631473541,
      "rewards-max/forensics_strict_edge_reward": 0.9158510267734528,
      "rewards-max/forensics_tiou_reward": 0.9956347644329071,
      "rewards/forensics_sinkhorn_reward": 0.5159643292427063,
      "rewards/forensics_strict_edge_reward": 0.181675985455513,
      "rewards/forensics_tiou_reward": 0.2882859706878662,
      "step": 938
    },
    {
      "completion_length": 56.609375,
      "epoch": 1.9654450261780103,
      "grad_norm": 26.94740867614746,
      "kl": 4.5078125,
      "learning_rate": 5.083857442348008e-07,
      "loss": 0.3173,
      "reward": 0.691331148147583,
      "reward_std": 0.11457014456391335,
      "reward_std_all": 0.11236120015382767,
      "rewards-max/forensics_sinkhorn_reward": 0.9884091913700104,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9977086782455444,
      "rewards/forensics_sinkhorn_reward": 0.595371425151825,
      "rewards/forensics_strict_edge_reward": 0.4248708635568619,
      "rewards/forensics_tiou_reward": 0.4503750503063202,
      "step": 939
    },
    {
      "completion_length": 70.125,
      "epoch": 1.9675392670157068,
      "grad_norm": 17.603050231933594,
      "kl": 4.8359375,
      "learning_rate": 5.078616352201259e-07,
      "loss": 0.2646,
      "reward": 0.556702196598053,
      "reward_std": 0.23624438047409058,
      "reward_std_all": 0.23849784582853317,
      "rewards-max/forensics_sinkhorn_reward": 0.944658637046814,
      "rewards-max/forensics_strict_edge_reward": 0.6976262629032135,
      "rewards-max/forensics_tiou_reward": 0.8986717164516449,
      "rewards/forensics_sinkhorn_reward": 0.3905180096626282,
      "rewards/forensics_strict_edge_reward": 0.15656042471528053,
      "rewards/forensics_tiou_reward": 0.16533388197422028,
      "step": 940
    },
    {
      "completion_length": 44.8125,
      "epoch": 1.9696335078534033,
      "grad_norm": 21.614316940307617,
      "kl": 3.71875,
      "learning_rate": 5.073375262054507e-07,
      "loss": 0.3745,
      "reward": 0.6615904569625854,
      "reward_std": 0.12341538444161415,
      "reward_std_all": 0.1375228315591812,
      "rewards-max/forensics_sinkhorn_reward": 0.9887057840824127,
      "rewards-max/forensics_strict_edge_reward": 0.9890114367008209,
      "rewards-max/forensics_tiou_reward": 0.9988783001899719,
      "rewards/forensics_sinkhorn_reward": 0.5638869106769562,
      "rewards/forensics_strict_edge_reward": 0.36684101819992065,
      "rewards/forensics_tiou_reward": 0.4317910969257355,
      "step": 941
    },
    {
      "completion_length": 55.46875,
      "epoch": 1.9717277486910993,
      "grad_norm": 19.107688903808594,
      "kl": 4.5,
      "learning_rate": 5.068134171907757e-07,
      "loss": 0.2485,
      "reward": 0.7044422030448914,
      "reward_std": 0.0953473299741745,
      "reward_std_all": 0.09683168306946754,
      "rewards-max/forensics_sinkhorn_reward": 0.9646989107131958,
      "rewards-max/forensics_strict_edge_reward": 0.8612285256385803,
      "rewards-max/forensics_tiou_reward": 0.9444151222705841,
      "rewards/forensics_sinkhorn_reward": 0.565524160861969,
      "rewards/forensics_strict_edge_reward": 0.3539542853832245,
      "rewards/forensics_tiou_reward": 0.3634341359138489,
      "step": 942
    },
    {
      "completion_length": 62.796875,
      "epoch": 1.9738219895287958,
      "grad_norm": 40.97380065917969,
      "kl": 3.6484375,
      "learning_rate": 5.062893081761006e-07,
      "loss": 0.2743,
      "reward": 0.6194148659706116,
      "reward_std": 0.14714496955275536,
      "reward_std_all": 0.16742782294750214,
      "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": 0.5642687082290649,
      "rewards/forensics_strict_edge_reward": 0.3057660162448883,
      "rewards/forensics_tiou_reward": 0.3677354156970978,
      "step": 943
    },
    {
      "completion_length": 61.890625,
      "epoch": 1.975916230366492,
      "grad_norm": 25.904571533203125,
      "kl": 4.640625,
      "learning_rate": 5.057651991614256e-07,
      "loss": 0.4551,
      "reward": 0.6294304132461548,
      "reward_std": 0.1913626566529274,
      "reward_std_all": 0.18183030933141708,
      "rewards-max/forensics_sinkhorn_reward": 0.8916900455951691,
      "rewards-max/forensics_strict_edge_reward": 0.8746401965618134,
      "rewards-max/forensics_tiou_reward": 0.8111991286277771,
      "rewards/forensics_sinkhorn_reward": 0.5392693877220154,
      "rewards/forensics_strict_edge_reward": 0.27842750400304794,
      "rewards/forensics_tiou_reward": 0.3399175703525543,
      "step": 944
    },
    {
      "completion_length": 57.453125,
      "epoch": 1.9780104712041884,
      "grad_norm": 28.880632400512695,
      "kl": 5.21875,
      "learning_rate": 5.052410901467505e-07,
      "loss": 0.4789,
      "reward": 0.6440774202346802,
      "reward_std": 0.17073211073875427,
      "reward_std_all": 0.1659955307841301,
      "rewards-max/forensics_sinkhorn_reward": 0.9735081791877747,
      "rewards-max/forensics_strict_edge_reward": 0.947419673204422,
      "rewards-max/forensics_tiou_reward": 0.9636746048927307,
      "rewards/forensics_sinkhorn_reward": 0.5800137519836426,
      "rewards/forensics_strict_edge_reward": 0.3524559587240219,
      "rewards/forensics_tiou_reward": 0.42201048135757446,
      "step": 945
    },
    {
      "completion_length": 62.671875,
      "epoch": 1.9801047120418849,
      "grad_norm": 14.572125434875488,
      "kl": 5.25,
      "learning_rate": 5.047169811320755e-07,
      "loss": 0.4218,
      "reward": 0.6249460577964783,
      "reward_std": 0.13143617659807205,
      "reward_std_all": 0.15022395178675652,
      "rewards-max/forensics_sinkhorn_reward": 0.9849022328853607,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.985834151506424,
      "rewards/forensics_sinkhorn_reward": 0.5119281262159348,
      "rewards/forensics_strict_edge_reward": 0.2175205573439598,
      "rewards/forensics_tiou_reward": 0.2935572490096092,
      "step": 946
    },
    {
      "completion_length": 48.796875,
      "epoch": 1.9821989528795811,
      "grad_norm": 17.873027801513672,
      "kl": 4.828125,
      "learning_rate": 5.041928721174003e-07,
      "loss": 0.3682,
      "reward": 0.6340548992156982,
      "reward_std": 0.13580834865570068,
      "reward_std_all": 0.15252754837274551,
      "rewards-max/forensics_sinkhorn_reward": 0.9809686839580536,
      "rewards-max/forensics_strict_edge_reward": 0.7783617675304413,
      "rewards-max/forensics_tiou_reward": 0.9688411355018616,
      "rewards/forensics_sinkhorn_reward": 0.5095580816268921,
      "rewards/forensics_strict_edge_reward": 0.16623496264219284,
      "rewards/forensics_tiou_reward": 0.3027777373790741,
      "step": 947
    },
    {
      "completion_length": 116.453125,
      "epoch": 1.9842931937172774,
      "grad_norm": 66.39860534667969,
      "kl": 8.375,
      "learning_rate": 5.036687631027253e-07,
      "loss": 0.5482,
      "reward": 0.5480929613113403,
      "reward_std": 0.2162591964006424,
      "reward_std_all": 0.22938548028469086,
      "rewards-max/forensics_sinkhorn_reward": 0.9786824584007263,
      "rewards-max/forensics_strict_edge_reward": 0.9464850425720215,
      "rewards-max/forensics_tiou_reward": 0.9752123057842255,
      "rewards/forensics_sinkhorn_reward": 0.4309849292039871,
      "rewards/forensics_strict_edge_reward": 0.2516764774918556,
      "rewards/forensics_tiou_reward": 0.28206130862236023,
      "step": 948
    },
    {
      "completion_length": 41.546875,
      "epoch": 1.986387434554974,
      "grad_norm": 17.82589340209961,
      "kl": 4.6875,
      "learning_rate": 5.031446540880503e-07,
      "loss": 0.328,
      "reward": 0.6855852007865906,
      "reward_std": 0.1527787633240223,
      "reward_std_all": 0.14092783629894257,
      "rewards-max/forensics_sinkhorn_reward": 0.9887070655822754,
      "rewards-max/forensics_strict_edge_reward": 0.9890114367008209,
      "rewards-max/forensics_tiou_reward": 0.9988826811313629,
      "rewards/forensics_sinkhorn_reward": 0.6633382141590118,
      "rewards/forensics_strict_edge_reward": 0.44502583146095276,
      "rewards/forensics_tiou_reward": 0.5351939499378204,
      "step": 949
    },
    {
      "completion_length": 59.109375,
      "epoch": 1.9884816753926702,
      "grad_norm": 27.697420120239258,
      "kl": 6.40625,
      "learning_rate": 5.026205450733752e-07,
      "loss": 0.5798,
      "reward": 0.6494515836238861,
      "reward_std": 0.09853703528642654,
      "reward_std_all": 0.1268155425786972,
      "rewards-max/forensics_sinkhorn_reward": 0.9880745708942413,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9963903427124023,
      "rewards/forensics_sinkhorn_reward": 0.540943905711174,
      "rewards/forensics_strict_edge_reward": 0.32557283341884613,
      "rewards/forensics_tiou_reward": 0.39799419045448303,
      "step": 950
    },
    {
      "completion_length": 93.203125,
      "epoch": 1.9905759162303664,
      "grad_norm": 68.09803009033203,
      "kl": 12.125,
      "learning_rate": 5.020964360587002e-07,
      "loss": 0.6488,
      "reward": 0.49865929782390594,
      "reward_std": 0.25587165355682373,
      "reward_std_all": 0.27899909019470215,
      "rewards-max/forensics_sinkhorn_reward": 0.9451834559440613,
      "rewards-max/forensics_strict_edge_reward": 0.7054409384727478,
      "rewards-max/forensics_tiou_reward": 0.89498171210289,
      "rewards/forensics_sinkhorn_reward": 0.38296787440776825,
      "rewards/forensics_strict_edge_reward": 0.15884723514318466,
      "rewards/forensics_tiou_reward": 0.205642968416214,
      "step": 951
    },
    {
      "completion_length": 77.4375,
      "epoch": 1.992670157068063,
      "grad_norm": 28.22779655456543,
      "kl": 7.375,
      "learning_rate": 5.015723270440252e-07,
      "loss": 0.5373,
      "reward": 0.5253495275974274,
      "reward_std": 0.1725853681564331,
      "reward_std_all": 0.22144793719053268,
      "rewards-max/forensics_sinkhorn_reward": 0.988709419965744,
      "rewards-max/forensics_strict_edge_reward": 0.9890114367008209,
      "rewards-max/forensics_tiou_reward": 0.9988919794559479,
      "rewards/forensics_sinkhorn_reward": 0.4425973892211914,
      "rewards/forensics_strict_edge_reward": 0.22929444909095764,
      "rewards/forensics_tiou_reward": 0.28683503717184067,
      "step": 952
    },
    {
      "completion_length": 49.390625,
      "epoch": 1.9947643979057592,
      "grad_norm": 23.148765563964844,
      "kl": 4.484375,
      "learning_rate": 5.010482180293501e-07,
      "loss": 0.2751,
      "reward": 0.6073294281959534,
      "reward_std": 0.19375015050172806,
      "reward_std_all": 0.20524662733078003,
      "rewards-max/forensics_sinkhorn_reward": 0.9470574259757996,
      "rewards-max/forensics_strict_edge_reward": 0.747104287147522,
      "rewards-max/forensics_tiou_reward": 0.9034852385520935,
      "rewards/forensics_sinkhorn_reward": 0.5056585669517517,
      "rewards/forensics_strict_edge_reward": 0.25548356026411057,
      "rewards/forensics_tiou_reward": 0.3106585890054703,
      "step": 953
    },
    {
      "completion_length": 77.875,
      "epoch": 1.9968586387434555,
      "grad_norm": 28.863056182861328,
      "kl": 6.9375,
      "learning_rate": 5.00524109014675e-07,
      "loss": 0.3766,
      "reward": 0.6111485958099365,
      "reward_std": 0.17919455468654633,
      "reward_std_all": 0.18700366467237473,
      "rewards-max/forensics_sinkhorn_reward": 0.9631517827510834,
      "rewards-max/forensics_strict_edge_reward": 0.9075756669044495,
      "rewards-max/forensics_tiou_reward": 0.9422396421432495,
      "rewards/forensics_sinkhorn_reward": 0.4827461242675781,
      "rewards/forensics_strict_edge_reward": 0.23584981262683868,
      "rewards/forensics_tiou_reward": 0.2500261142849922,
      "step": 954
    },
    {
      "completion_length": 73.109375,
      "epoch": 1.998952879581152,
      "grad_norm": 34.93071746826172,
      "kl": 7.9453125,
      "learning_rate": 5e-07,
      "loss": 0.3762,
      "reward": 0.6676797866821289,
      "reward_std": 0.13809340074658394,
      "reward_std_all": 0.1365494541823864,
      "rewards-max/forensics_sinkhorn_reward": 0.9870631992816925,
      "rewards-max/forensics_strict_edge_reward": 0.9567649364471436,
      "rewards-max/forensics_tiou_reward": 0.9924229681491852,
      "rewards/forensics_sinkhorn_reward": 0.6013411283493042,
      "rewards/forensics_strict_edge_reward": 0.4203185886144638,
      "rewards/forensics_tiou_reward": 0.4630992114543915,
      "step": 955
    },
    {
      "completion_length": 31.25,
      "epoch": 2.0,
      "grad_norm": 12.393827438354492,
      "kl": 3.171875,
      "learning_rate": 4.99475890985325e-07,
      "loss": 0.1513,
      "reward": 0.6445776224136353,
      "reward_std": 0.14696964621543884,
      "reward_std_all": 0.15532474219799042,
      "rewards-max/forensics_sinkhorn_reward": 0.9888147115707397,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9993070960044861,
      "rewards/forensics_sinkhorn_reward": 0.6190314292907715,
      "rewards/forensics_strict_edge_reward": 0.3812776505947113,
      "rewards/forensics_tiou_reward": 0.45776113867759705,
      "step": 956
    },
    {
      "completion_length": 72.578125,
      "epoch": 2.0020942408376965,
      "grad_norm": 30.999263763427734,
      "kl": 7.765625,
      "learning_rate": 4.989517819706499e-07,
      "loss": 0.5084,
      "reward": 0.5502365827560425,
      "reward_std": 0.2388116419315338,
      "reward_std_all": 0.2442302629351616,
      "rewards-max/forensics_sinkhorn_reward": 0.962476909160614,
      "rewards-max/forensics_strict_edge_reward": 0.8992868363857269,
      "rewards-max/forensics_tiou_reward": 0.940796971321106,
      "rewards/forensics_sinkhorn_reward": 0.46079790592193604,
      "rewards/forensics_strict_edge_reward": 0.2751259505748749,
      "rewards/forensics_tiou_reward": 0.29990698397159576,
      "step": 957
    },
    {
      "completion_length": 62.59375,
      "epoch": 2.0041884816753925,
      "grad_norm": 30.010358810424805,
      "kl": 4.7109375,
      "learning_rate": 4.984276729559748e-07,
      "loss": 0.3479,
      "reward": 0.6287204325199127,
      "reward_std": 0.11217410117387772,
      "reward_std_all": 0.1414669267833233,
      "rewards-max/forensics_sinkhorn_reward": 0.9625498950481415,
      "rewards-max/forensics_strict_edge_reward": 0.9075756669044495,
      "rewards-max/forensics_tiou_reward": 0.9417838156223297,
      "rewards/forensics_sinkhorn_reward": 0.5823079645633698,
      "rewards/forensics_strict_edge_reward": 0.30433887988328934,
      "rewards/forensics_tiou_reward": 0.3752775341272354,
      "step": 958
    },
    {
      "completion_length": 83.921875,
      "epoch": 2.006282722513089,
      "grad_norm": 19.66211700439453,
      "kl": 6.5625,
      "learning_rate": 4.979035639412998e-07,
      "loss": 0.5307,
      "reward": 0.5837657153606415,
      "reward_std": 0.19969656318426132,
      "reward_std_all": 0.20932172238826752,
      "rewards-max/forensics_sinkhorn_reward": 0.9878451228141785,
      "rewards-max/forensics_strict_edge_reward": 0.9565287232398987,
      "rewards-max/forensics_tiou_reward": 0.9954851865768433,
      "rewards/forensics_sinkhorn_reward": 0.5060717761516571,
      "rewards/forensics_strict_edge_reward": 0.25467800348997116,
      "rewards/forensics_tiou_reward": 0.35211774706840515,
      "step": 959
    },
    {
      "completion_length": 85.359375,
      "epoch": 2.0083769633507855,
      "grad_norm": 27.248680114746094,
      "kl": 7.484375,
      "learning_rate": 4.973794549266247e-07,
      "loss": 0.4709,
      "reward": 0.5737574398517609,
      "reward_std": 0.190405435860157,
      "reward_std_all": 0.21864806860685349,
      "rewards-max/forensics_sinkhorn_reward": 0.9709931313991547,
      "rewards-max/forensics_strict_edge_reward": 0.9364311099052429,
      "rewards-max/forensics_tiou_reward": 0.9598948061466217,
      "rewards/forensics_sinkhorn_reward": 0.44633692502975464,
      "rewards/forensics_strict_edge_reward": 0.22323215007781982,
      "rewards/forensics_tiou_reward": 0.24510382860898972,
      "step": 960
    },
    {
      "completion_length": 78.125,
      "epoch": 2.0104712041884816,
      "grad_norm": 20.8520450592041,
      "kl": 5.40625,
      "learning_rate": 4.968553459119496e-07,
      "loss": 0.3942,
      "reward": 0.662363737821579,
      "reward_std": 0.13358792662620544,
      "reward_std_all": 0.13743342086672783,
      "rewards-max/forensics_sinkhorn_reward": 0.9784373044967651,
      "rewards-max/forensics_strict_edge_reward": 0.946017861366272,
      "rewards-max/forensics_tiou_reward": 0.9749947488307953,
      "rewards/forensics_sinkhorn_reward": 0.5316314995288849,
      "rewards/forensics_strict_edge_reward": 0.29993336647748947,
      "rewards/forensics_tiou_reward": 0.3611717075109482,
      "step": 961
    },
    {
      "completion_length": 51.46875,
      "epoch": 2.012565445026178,
      "grad_norm": 26.57257843017578,
      "kl": 4.375,
      "learning_rate": 4.963312368972746e-07,
      "loss": 0.154,
      "reward": 0.5906961560249329,
      "reward_std": 0.20904717594385147,
      "reward_std_all": 0.21825896203517914,
      "rewards-max/forensics_sinkhorn_reward": 0.9009860754013062,
      "rewards-max/forensics_strict_edge_reward": 0.7763970792293549,
      "rewards-max/forensics_tiou_reward": 0.8166559934616089,
      "rewards/forensics_sinkhorn_reward": 0.47335490584373474,
      "rewards/forensics_strict_edge_reward": 0.1962616741657257,
      "rewards/forensics_tiou_reward": 0.23520222306251526,
      "step": 962
    },
    {
      "completion_length": 79.953125,
      "epoch": 2.0146596858638746,
      "grad_norm": 31.585599899291992,
      "kl": 5.59375,
      "learning_rate": 4.958071278825995e-07,
      "loss": 0.2921,
      "reward": 0.6214499473571777,
      "reward_std": 0.0822662003338337,
      "reward_std_all": 0.12860063463449478,
      "rewards-max/forensics_sinkhorn_reward": 0.9686821103096008,
      "rewards-max/forensics_strict_edge_reward": 0.8760378360748291,
      "rewards-max/forensics_tiou_reward": 0.9525090157985687,
      "rewards/forensics_sinkhorn_reward": 0.4537023454904556,
      "rewards/forensics_strict_edge_reward": 0.1690668910741806,
      "rewards/forensics_tiou_reward": 0.20082750171422958,
      "step": 963
    },
    {
      "completion_length": 76.59375,
      "epoch": 2.0167539267015706,
      "grad_norm": 20.7779483795166,
      "kl": 5.828125,
      "learning_rate": 4.952830188679246e-07,
      "loss": 0.3999,
      "reward": 0.5767645537853241,
      "reward_std": 0.20239870995283127,
      "reward_std_all": 0.21388526260852814,
      "rewards-max/forensics_sinkhorn_reward": 0.9873242974281311,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9954961836338043,
      "rewards/forensics_sinkhorn_reward": 0.48228493332862854,
      "rewards/forensics_strict_edge_reward": 0.23087003082036972,
      "rewards/forensics_tiou_reward": 0.2891530692577362,
      "step": 964
    },
    {
      "completion_length": 52.265625,
      "epoch": 2.018848167539267,
      "grad_norm": 19.676172256469727,
      "kl": 4.859375,
      "learning_rate": 4.947589098532495e-07,
      "loss": 0.2719,
      "reward": 0.590379148721695,
      "reward_std": 0.19739478081464767,
      "reward_std_all": 0.204165980219841,
      "rewards-max/forensics_sinkhorn_reward": 0.9813882410526276,
      "rewards-max/forensics_strict_edge_reward": 0.9364311099052429,
      "rewards-max/forensics_tiou_reward": 0.9825650155544281,
      "rewards/forensics_sinkhorn_reward": 0.4728502631187439,
      "rewards/forensics_strict_edge_reward": 0.2313208058476448,
      "rewards/forensics_tiou_reward": 0.29688918590545654,
      "step": 965
    },
    {
      "completion_length": 51.375,
      "epoch": 2.020942408376963,
      "grad_norm": 22.552534103393555,
      "kl": 3.609375,
      "learning_rate": 4.942348008385744e-07,
      "loss": 0.2973,
      "reward": 0.700736939907074,
      "reward_std": 0.13378266245126724,
      "reward_std_all": 0.1254706308245659,
      "rewards-max/forensics_sinkhorn_reward": 0.988313764333725,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9973332285881042,
      "rewards/forensics_sinkhorn_reward": 0.6322608888149261,
      "rewards/forensics_strict_edge_reward": 0.4060201495885849,
      "rewards/forensics_tiou_reward": 0.4970773309469223,
      "step": 966
    },
    {
      "completion_length": 54.421875,
      "epoch": 2.0230366492146596,
      "grad_norm": 24.48440170288086,
      "kl": 2.8515625,
      "learning_rate": 4.937106918238994e-07,
      "loss": 0.2562,
      "reward": 0.6616657674312592,
      "reward_std": 0.11458390206098557,
      "reward_std_all": 0.13012075424194336,
      "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": 0.5284381210803986,
      "rewards/forensics_strict_edge_reward": 0.3308308869600296,
      "rewards/forensics_tiou_reward": 0.3611812889575958,
      "step": 967
    },
    {
      "completion_length": 60.109375,
      "epoch": 2.025130890052356,
      "grad_norm": 24.24699592590332,
      "kl": 7.125,
      "learning_rate": 4.931865828092243e-07,
      "loss": 0.437,
      "reward": 0.5862197279930115,
      "reward_std": 0.21816309541463852,
      "reward_std_all": 0.2190789207816124,
      "rewards-max/forensics_sinkhorn_reward": 0.960658997297287,
      "rewards-max/forensics_strict_edge_reward": 0.9185642302036285,
      "rewards-max/forensics_tiou_reward": 0.9400303661823273,
      "rewards/forensics_sinkhorn_reward": 0.48836566507816315,
      "rewards/forensics_strict_edge_reward": 0.28697459399700165,
      "rewards/forensics_tiou_reward": 0.3485649824142456,
      "step": 968
    },
    {
      "completion_length": 71.484375,
      "epoch": 2.027225130890052,
      "grad_norm": 19.039640426635742,
      "kl": 4.734375,
      "learning_rate": 4.926624737945492e-07,
      "loss": 0.3663,
      "reward": 0.6556344628334045,
      "reward_std": 0.17047512531280518,
      "reward_std_all": 0.16395212709903717,
      "rewards-max/forensics_sinkhorn_reward": 0.9932576417922974,
      "rewards-max/forensics_strict_edge_reward": 0.9364311099052429,
      "rewards-max/forensics_tiou_reward": 0.9935219585895538,
      "rewards/forensics_sinkhorn_reward": 0.5368825793266296,
      "rewards/forensics_strict_edge_reward": 0.25507401674985886,
      "rewards/forensics_tiou_reward": 0.3427004814147949,
      "step": 969
    },
    {
      "completion_length": 73.796875,
      "epoch": 2.0293193717277487,
      "grad_norm": 25.259321212768555,
      "kl": 5.390625,
      "learning_rate": 4.921383647798742e-07,
      "loss": 0.3456,
      "reward": 0.6536512970924377,
      "reward_std": 0.159425500780344,
      "reward_std_all": 0.1543862149119377,
      "rewards-max/forensics_sinkhorn_reward": 0.9848774373531342,
      "rewards-max/forensics_strict_edge_reward": 0.927969753742218,
      "rewards-max/forensics_tiou_reward": 0.9839275777339935,
      "rewards/forensics_sinkhorn_reward": 0.5477719902992249,
      "rewards/forensics_strict_edge_reward": 0.3232829421758652,
      "rewards/forensics_tiou_reward": 0.39202769845724106,
      "step": 970
    },
    {
      "completion_length": 57.15625,
      "epoch": 2.031413612565445,
      "grad_norm": 23.54572868347168,
      "kl": 5.53125,
      "learning_rate": 4.916142557651992e-07,
      "loss": 0.43,
      "reward": 0.6233495771884918,
      "reward_std": 0.19381630420684814,
      "reward_std_all": 0.19168149679899216,
      "rewards-max/forensics_sinkhorn_reward": 0.9614469707012177,
      "rewards-max/forensics_strict_edge_reward": 0.8279296159744263,
      "rewards-max/forensics_tiou_reward": 0.9301971197128296,
      "rewards/forensics_sinkhorn_reward": 0.4906279593706131,
      "rewards/forensics_strict_edge_reward": 0.27560438960790634,
      "rewards/forensics_tiou_reward": 0.33514511585235596,
      "step": 971
    },
    {
      "completion_length": 65.546875,
      "epoch": 2.033507853403141,
      "grad_norm": 21.87407684326172,
      "kl": 7.28125,
      "learning_rate": 4.910901467505241e-07,
      "loss": 0.4285,
      "reward": 0.6601336300373077,
      "reward_std": 0.17429693788290024,
      "reward_std_all": 0.1621631607413292,
      "rewards-max/forensics_sinkhorn_reward": 0.9879729151725769,
      "rewards-max/forensics_strict_edge_reward": 0.9158510267734528,
      "rewards-max/forensics_tiou_reward": 0.9959928095340729,
      "rewards/forensics_sinkhorn_reward": 0.558613121509552,
      "rewards/forensics_strict_edge_reward": 0.28084592521190643,
      "rewards/forensics_tiou_reward": 0.34089431166648865,
      "step": 972
    },
    {
      "completion_length": 91.71875,
      "epoch": 2.0356020942408377,
      "grad_norm": 35.847442626953125,
      "kl": 8.5625,
      "learning_rate": 4.90566037735849e-07,
      "loss": 0.4845,
      "reward": 0.5968821048736572,
      "reward_std": 0.21507073938846588,
      "reward_std_all": 0.21517252177000046,
      "rewards-max/forensics_sinkhorn_reward": 0.9813450872898102,
      "rewards-max/forensics_strict_edge_reward": 0.8659839034080505,
      "rewards-max/forensics_tiou_reward": 0.9699247181415558,
      "rewards/forensics_sinkhorn_reward": 0.4375188499689102,
      "rewards/forensics_strict_edge_reward": 0.25786370038986206,
      "rewards/forensics_tiou_reward": 0.2704708054661751,
      "step": 973
    },
    {
      "completion_length": 46.8125,
      "epoch": 2.037696335078534,
      "grad_norm": 17.22433853149414,
      "kl": 3.765625,
      "learning_rate": 4.90041928721174e-07,
      "loss": 0.1953,
      "reward": 0.6807052791118622,
      "reward_std": 0.14441712945699692,
      "reward_std_all": 0.13710053265094757,
      "rewards-max/forensics_sinkhorn_reward": 0.9874483942985535,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9939287304878235,
      "rewards/forensics_sinkhorn_reward": 0.5703335404396057,
      "rewards/forensics_strict_edge_reward": 0.37723812460899353,
      "rewards/forensics_tiou_reward": 0.3947443217039108,
      "step": 974
    },
    {
      "completion_length": 54.328125,
      "epoch": 2.0397905759162303,
      "grad_norm": 27.00429916381836,
      "kl": 6.828125,
      "learning_rate": 4.895178197064989e-07,
      "loss": 0.3172,
      "reward": 0.6622607111930847,
      "reward_std": 0.178541611880064,
      "reward_std_all": 0.16774127632379532,
      "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": 0.5770617127418518,
      "rewards/forensics_strict_edge_reward": 0.3447968661785126,
      "rewards/forensics_tiou_reward": 0.4007684588432312,
      "step": 975
    },
    {
      "completion_length": 44.21875,
      "epoch": 2.0418848167539267,
      "grad_norm": 18.0899600982666,
      "kl": 4.0546875,
      "learning_rate": 4.889937106918238e-07,
      "loss": 0.4243,
      "reward": 0.6723274886608124,
      "reward_std": 0.1472972333431244,
      "reward_std_all": 0.1475609615445137,
      "rewards-max/forensics_sinkhorn_reward": 0.9886906445026398,
      "rewards-max/forensics_strict_edge_reward": 0.9890114367008209,
      "rewards-max/forensics_tiou_reward": 0.9988186657428741,
      "rewards/forensics_sinkhorn_reward": 0.5466989576816559,
      "rewards/forensics_strict_edge_reward": 0.34397175908088684,
      "rewards/forensics_tiou_reward": 0.3759468197822571,
      "step": 976
    },
    {
      "completion_length": 49.3125,
      "epoch": 2.0439790575916232,
      "grad_norm": 21.36685562133789,
      "kl": 6.109375,
      "learning_rate": 4.884696016771489e-07,
      "loss": 0.6031,
      "reward": 0.5730831623077393,
      "reward_std": 0.19701959192752838,
      "reward_std_all": 0.20747844874858856,
      "rewards-max/forensics_sinkhorn_reward": 0.9768860042095184,
      "rewards-max/forensics_strict_edge_reward": 0.9574736058712006,
      "rewards-max/forensics_tiou_reward": 0.9723249077796936,
      "rewards/forensics_sinkhorn_reward": 0.4335560202598572,
      "rewards/forensics_strict_edge_reward": 0.21552374958992004,
      "rewards/forensics_tiou_reward": 0.27683112025260925,
      "step": 977
    },
    {
      "completion_length": 63.671875,
      "epoch": 2.0460732984293193,
      "grad_norm": 23.79232406616211,
      "kl": 6.4375,
      "learning_rate": 4.879454926624738e-07,
      "loss": 0.4343,
      "reward": 0.6658956408500671,
      "reward_std": 0.15024268999695778,
      "reward_std_all": 0.14690939337015152,
      "rewards-max/forensics_sinkhorn_reward": 0.9885430037975311,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.998235821723938,
      "rewards/forensics_sinkhorn_reward": 0.5835840702056885,
      "rewards/forensics_strict_edge_reward": 0.38496848940849304,
      "rewards/forensics_tiou_reward": 0.4469209462404251,
      "step": 978
    },
    {
      "completion_length": 47.375,
      "epoch": 2.048167539267016,
      "grad_norm": 25.186885833740234,
      "kl": 5.109375,
      "learning_rate": 4.874213836477988e-07,
      "loss": 0.3809,
      "reward": 0.6673646569252014,
      "reward_std": 0.12046855688095093,
      "reward_std_all": 0.13015271723270416,
      "rewards-max/forensics_sinkhorn_reward": 0.946405440568924,
      "rewards-max/forensics_strict_edge_reward": 0.8373351395130157,
      "rewards-max/forensics_tiou_reward": 0.901488870382309,
      "rewards/forensics_sinkhorn_reward": 0.5169210433959961,
      "rewards/forensics_strict_edge_reward": 0.3505203425884247,
      "rewards/forensics_tiou_reward": 0.34963400661945343,
      "step": 979
    },
    {
      "completion_length": 122.265625,
      "epoch": 2.050261780104712,
      "grad_norm": 25.66746711730957,
      "kl": 10.28125,
      "learning_rate": 4.868972746331237e-07,
      "loss": 0.6566,
      "reward": 0.5732990205287933,
      "reward_std": 0.21088653802871704,
      "reward_std_all": 0.22115430980920792,
      "rewards-max/forensics_sinkhorn_reward": 0.9833173453807831,
      "rewards-max/forensics_strict_edge_reward": 0.878633052110672,
      "rewards-max/forensics_tiou_reward": 0.977828860282898,
      "rewards/forensics_sinkhorn_reward": 0.43678218126296997,
      "rewards/forensics_strict_edge_reward": 0.23930557817220688,
      "rewards/forensics_tiou_reward": 0.29837581515312195,
      "step": 980
    },
    {
      "completion_length": 66.796875,
      "epoch": 2.0523560209424083,
      "grad_norm": 24.67769432067871,
      "kl": 6.84375,
      "learning_rate": 4.863731656184486e-07,
      "loss": 0.5702,
      "reward": 0.5847662091255188,
      "reward_std": 0.2328474447131157,
      "reward_std_all": 0.2246486097574234,
      "rewards-max/forensics_sinkhorn_reward": 0.983694851398468,
      "rewards-max/forensics_strict_edge_reward": 0.8876297473907471,
      "rewards-max/forensics_tiou_reward": 0.979248970746994,
      "rewards/forensics_sinkhorn_reward": 0.48121222853660583,
      "rewards/forensics_strict_edge_reward": 0.23184452950954437,
      "rewards/forensics_tiou_reward": 0.28470101952552795,
      "step": 981
    },
    {
      "completion_length": 100.125,
      "epoch": 2.054450261780105,
      "grad_norm": 22.823705673217773,
      "kl": 9.375,
      "learning_rate": 4.858490566037736e-07,
      "loss": 0.4949,
      "reward": 0.566217452287674,
      "reward_std": 0.21605335921049118,
      "reward_std_all": 0.23639395833015442,
      "rewards-max/forensics_sinkhorn_reward": 0.9760569632053375,
      "rewards-max/forensics_strict_edge_reward": 0.8445805907249451,
      "rewards-max/forensics_tiou_reward": 0.9697646200656891,
      "rewards/forensics_sinkhorn_reward": 0.4378214329481125,
      "rewards/forensics_strict_edge_reward": 0.22516661882400513,
      "rewards/forensics_tiou_reward": 0.28330574929714203,
      "step": 982
    },
    {
      "completion_length": 82.453125,
      "epoch": 2.056544502617801,
      "grad_norm": 27.729564666748047,
      "kl": 6.703125,
      "learning_rate": 4.853249475890985e-07,
      "loss": 0.3635,
      "reward": 0.6058187782764435,
      "reward_std": 0.19593769311904907,
      "reward_std_all": 0.20382174104452133,
      "rewards-max/forensics_sinkhorn_reward": 0.9880331456661224,
      "rewards-max/forensics_strict_edge_reward": 0.9677534997463226,
      "rewards-max/forensics_tiou_reward": 0.99622642993927,
      "rewards/forensics_sinkhorn_reward": 0.509333997964859,
      "rewards/forensics_strict_edge_reward": 0.3004469573497772,
      "rewards/forensics_tiou_reward": 0.34953515231609344,
      "step": 983
    },
    {
      "completion_length": 120.09375,
      "epoch": 2.0586387434554974,
      "grad_norm": 62.163909912109375,
      "kl": 11.0625,
      "learning_rate": 4.848008385744234e-07,
      "loss": 0.6383,
      "reward": 0.634410560131073,
      "reward_std": 0.14120341837406158,
      "reward_std_all": 0.14703894406557083,
      "rewards-max/forensics_sinkhorn_reward": 0.9875328540802002,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9942625463008881,
      "rewards/forensics_sinkhorn_reward": 0.4829000234603882,
      "rewards/forensics_strict_edge_reward": 0.25856127589941025,
      "rewards/forensics_tiou_reward": 0.29440098255872726,
      "step": 984
    },
    {
      "completion_length": 145.34375,
      "epoch": 2.060732984293194,
      "grad_norm": 23.964574813842773,
      "kl": 8.625,
      "learning_rate": 4.842767295597484e-07,
      "loss": 0.5033,
      "reward": 0.4937015175819397,
      "reward_std": 0.2253812626004219,
      "reward_std_all": 0.2647882252931595,
      "rewards-max/forensics_sinkhorn_reward": 0.8963526487350464,
      "rewards-max/forensics_strict_edge_reward": 0.7473003268241882,
      "rewards-max/forensics_tiou_reward": 0.8034073114395142,
      "rewards/forensics_sinkhorn_reward": 0.3936740309000015,
      "rewards/forensics_strict_edge_reward": 0.20049337297677994,
      "rewards/forensics_tiou_reward": 0.25349320471286774,
      "step": 985
    },
    {
      "completion_length": 90.015625,
      "epoch": 2.06282722513089,
      "grad_norm": 16.40554428100586,
      "kl": 5.484375,
      "learning_rate": 4.837526205450733e-07,
      "loss": 0.4174,
      "reward": 0.5653370320796967,
      "reward_std": 0.25387606769800186,
      "reward_std_all": 0.2449454963207245,
      "rewards-max/forensics_sinkhorn_reward": 0.9883022606372833,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9972895681858063,
      "rewards/forensics_sinkhorn_reward": 0.49838492274284363,
      "rewards/forensics_strict_edge_reward": 0.2692905068397522,
      "rewards/forensics_tiou_reward": 0.3459305465221405,
      "step": 986
    },
    {
      "completion_length": 125.84375,
      "epoch": 2.0649214659685864,
      "grad_norm": 41.29396057128906,
      "kl": 7.8125,
      "learning_rate": 4.832285115303983e-07,
      "loss": 0.5077,
      "reward": 0.5605268478393555,
      "reward_std": 0.22389151901006699,
      "reward_std_all": 0.2354389876127243,
      "rewards-max/forensics_sinkhorn_reward": 0.9763658940792084,
      "rewards-max/forensics_strict_edge_reward": 0.7787764966487885,
      "rewards-max/forensics_tiou_reward": 0.9692755043506622,
      "rewards/forensics_sinkhorn_reward": 0.42361749708652496,
      "rewards/forensics_strict_edge_reward": 0.2614302560687065,
      "rewards/forensics_tiou_reward": 0.3008929491043091,
      "step": 987
    },
    {
      "completion_length": 137.765625,
      "epoch": 2.067015706806283,
      "grad_norm": 22.57717514038086,
      "kl": 6.1875,
      "learning_rate": 4.827044025157232e-07,
      "loss": 0.3946,
      "reward": 0.5431645512580872,
      "reward_std": 0.24136656522750854,
      "reward_std_all": 0.2432761937379837,
      "rewards-max/forensics_sinkhorn_reward": 0.9880331456661224,
      "rewards-max/forensics_strict_edge_reward": 0.9782643616199493,
      "rewards-max/forensics_tiou_reward": 0.99622642993927,
      "rewards/forensics_sinkhorn_reward": 0.4497302323579788,
      "rewards/forensics_strict_edge_reward": 0.21676570922136307,
      "rewards/forensics_tiou_reward": 0.2769046723842621,
      "step": 988
    },
    {
      "completion_length": 130.421875,
      "epoch": 2.069109947643979,
      "grad_norm": 19.74891471862793,
      "kl": 8.484375,
      "learning_rate": 4.821802935010483e-07,
      "loss": 0.4655,
      "reward": 0.5440541803836823,
      "reward_std": 0.2227773517370224,
      "reward_std_all": 0.2386736422777176,
      "rewards-max/forensics_sinkhorn_reward": 0.9596501886844635,
      "rewards-max/forensics_strict_edge_reward": 0.8745477199554443,
      "rewards-max/forensics_tiou_reward": 0.9328441917896271,
      "rewards/forensics_sinkhorn_reward": 0.4366987496614456,
      "rewards/forensics_strict_edge_reward": 0.18985111266374588,
      "rewards/forensics_tiou_reward": 0.24623564630746841,
      "step": 989
    },
    {
      "completion_length": 95.890625,
      "epoch": 2.0712041884816754,
      "grad_norm": 28.818761825561523,
      "kl": 6.265625,
      "learning_rate": 4.816561844863732e-07,
      "loss": 0.4496,
      "reward": 0.5672980546951294,
      "reward_std": 0.17659679055213928,
      "reward_std_all": 0.20430753380060196,
      "rewards-max/forensics_sinkhorn_reward": 0.9840780198574066,
      "rewards-max/forensics_strict_edge_reward": 0.8876297473907471,
      "rewards-max/forensics_tiou_reward": 0.9807202517986298,
      "rewards/forensics_sinkhorn_reward": 0.4652019292116165,
      "rewards/forensics_strict_edge_reward": 0.22520481795072556,
      "rewards/forensics_tiou_reward": 0.27756159007549286,
      "step": 990
    },
    {
      "completion_length": 106.765625,
      "epoch": 2.073298429319372,
      "grad_norm": 29.479694366455078,
      "kl": 8.03125,
      "learning_rate": 4.811320754716981e-07,
      "loss": 0.5931,
      "reward": 0.5426617562770844,
      "reward_std": 0.29225948452949524,
      "reward_std_all": 0.27577973902225494,
      "rewards-max/forensics_sinkhorn_reward": 0.9876865446567535,
      "rewards-max/forensics_strict_edge_reward": 0.9677534997463226,
      "rewards-max/forensics_tiou_reward": 0.9948737025260925,
      "rewards/forensics_sinkhorn_reward": 0.48092053830623627,
      "rewards/forensics_strict_edge_reward": 0.2907470166683197,
      "rewards/forensics_tiou_reward": 0.37018343806266785,
      "step": 991
    },
    {
      "completion_length": 93.25,
      "epoch": 2.075392670157068,
      "grad_norm": 16.773719787597656,
      "kl": 5.578125,
      "learning_rate": 4.806079664570231e-07,
      "loss": 0.4094,
      "reward": 0.6354628801345825,
      "reward_std": 0.13710755109786987,
      "reward_std_all": 0.15337128192186356,
      "rewards-max/forensics_sinkhorn_reward": 0.970997542142868,
      "rewards-max/forensics_strict_edge_reward": 0.9375866651535034,
      "rewards-max/forensics_tiou_reward": 0.9588354825973511,
      "rewards/forensics_sinkhorn_reward": 0.5339578092098236,
      "rewards/forensics_strict_edge_reward": 0.30960339307785034,
      "rewards/forensics_tiou_reward": 0.37460140883922577,
      "step": 992
    },
    {
      "completion_length": 118.265625,
      "epoch": 2.0774869109947645,
      "grad_norm": 36.9475212097168,
      "kl": 8.625,
      "learning_rate": 4.80083857442348e-07,
      "loss": 0.6183,
      "reward": 0.5625837296247482,
      "reward_std": 0.1875925064086914,
      "reward_std_all": 0.21295927464962006,
      "rewards-max/forensics_sinkhorn_reward": 0.9647227227687836,
      "rewards-max/forensics_strict_edge_reward": 0.9075756669044495,
      "rewards-max/forensics_tiou_reward": 0.9446149170398712,
      "rewards/forensics_sinkhorn_reward": 0.48255278170108795,
      "rewards/forensics_strict_edge_reward": 0.2722858190536499,
      "rewards/forensics_tiou_reward": 0.33204567432403564,
      "step": 993
    },
    {
      "completion_length": 132.078125,
      "epoch": 2.0795811518324605,
      "grad_norm": 23.07578468322754,
      "kl": 6.9375,
      "learning_rate": 4.795597484276729e-07,
      "loss": 0.302,
      "reward": 0.5583105832338333,
      "reward_std": 0.24308671057224274,
      "reward_std_all": 0.24455100297927856,
      "rewards-max/forensics_sinkhorn_reward": 0.9746058285236359,
      "rewards-max/forensics_strict_edge_reward": 0.9364311099052429,
      "rewards-max/forensics_tiou_reward": 0.9674992263317108,
      "rewards/forensics_sinkhorn_reward": 0.4323287457227707,
      "rewards/forensics_strict_edge_reward": 0.2399328052997589,
      "rewards/forensics_tiou_reward": 0.27836932241916656,
      "step": 994
    },
    {
      "completion_length": 81.78125,
      "epoch": 2.081675392670157,
      "grad_norm": 25.64715576171875,
      "kl": 5.4765625,
      "learning_rate": 4.790356394129979e-07,
      "loss": 0.4526,
      "reward": 0.5786546468734741,
      "reward_std": 0.2290555238723755,
      "reward_std_all": 0.22408176213502884,
      "rewards-max/forensics_sinkhorn_reward": 0.9628691673278809,
      "rewards-max/forensics_strict_edge_reward": 0.800643116235733,
      "rewards-max/forensics_tiou_reward": 0.9436235427856445,
      "rewards/forensics_sinkhorn_reward": 0.491530641913414,
      "rewards/forensics_strict_edge_reward": 0.26463910937309265,
      "rewards/forensics_tiou_reward": 0.3183329254388809,
      "step": 995
    },
    {
      "completion_length": 56.96875,
      "epoch": 2.0837696335078535,
      "grad_norm": 22.60244369506836,
      "kl": 5.234375,
      "learning_rate": 4.785115303983228e-07,
      "loss": 0.2185,
      "reward": 0.5768359899520874,
      "reward_std": 0.1655488759279251,
      "reward_std_all": 0.19952856749296188,
      "rewards-max/forensics_sinkhorn_reward": 0.8243288695812225,
      "rewards-max/forensics_strict_edge_reward": 0.7276096791028976,
      "rewards-max/forensics_tiou_reward": 0.8028478622436523,
      "rewards/forensics_sinkhorn_reward": 0.43745221197605133,
      "rewards/forensics_strict_edge_reward": 0.17506646737456322,
      "rewards/forensics_tiou_reward": 0.2122037038207054,
      "step": 996
    },
    {
      "completion_length": 88.578125,
      "epoch": 2.0858638743455495,
      "grad_norm": 22.740543365478516,
      "kl": 5.13671875,
      "learning_rate": 4.779874213836478e-07,
      "loss": 0.2749,
      "reward": 0.6245743930339813,
      "reward_std": 0.17727144062519073,
      "reward_std_all": 0.18394913524389267,
      "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": 0.5710867643356323,
      "rewards/forensics_strict_edge_reward": 0.26637229323387146,
      "rewards/forensics_tiou_reward": 0.38009290397167206,
      "step": 997
    },
    {
      "completion_length": 70.546875,
      "epoch": 2.087958115183246,
      "grad_norm": 23.287508010864258,
      "kl": 6.171875,
      "learning_rate": 4.774633123689727e-07,
      "loss": 0.308,
      "reward": 0.6175678372383118,
      "reward_std": 0.1467236801981926,
      "reward_std_all": 0.16093683242797852,
      "rewards-max/forensics_sinkhorn_reward": 0.946279376745224,
      "rewards-max/forensics_strict_edge_reward": 0.9169811904430389,
      "rewards-max/forensics_tiou_reward": 0.9093993902206421,
      "rewards/forensics_sinkhorn_reward": 0.49025100469589233,
      "rewards/forensics_strict_edge_reward": 0.22162996232509613,
      "rewards/forensics_tiou_reward": 0.2653249055147171,
      "step": 998
    },
    {
      "completion_length": 83.875,
      "epoch": 2.0900523560209425,
      "grad_norm": 18.615814208984375,
      "kl": 5.703125,
      "learning_rate": 4.769392033542976e-07,
      "loss": 0.4342,
      "reward": 0.6089533865451813,
      "reward_std": 0.13354850560426712,
      "reward_std_all": 0.1614186018705368,
      "rewards-max/forensics_sinkhorn_reward": 0.988754540681839,
      "rewards-max/forensics_strict_edge_reward": 0.9890114367008209,
      "rewards-max/forensics_tiou_reward": 0.9990697801113129,
      "rewards/forensics_sinkhorn_reward": 0.5520191192626953,
      "rewards/forensics_strict_edge_reward": 0.29866012930870056,
      "rewards/forensics_tiou_reward": 0.35944487154483795,
      "step": 999
    },
    {
      "completion_length": 78.0,
      "epoch": 2.0921465968586386,
      "grad_norm": 34.340576171875,
      "kl": 6.0,
      "learning_rate": 4.7641509433962264e-07,
      "loss": 0.3732,
      "reward": 0.5868495404720306,
      "reward_std": 0.17921315133571625,
      "reward_std_all": 0.19895002990961075,
      "rewards-max/forensics_sinkhorn_reward": 0.9883488118648529,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9974724650382996,
      "rewards/forensics_sinkhorn_reward": 0.5124197155237198,
      "rewards/forensics_strict_edge_reward": 0.24855097383260727,
      "rewards/forensics_tiou_reward": 0.3148678094148636,
      "step": 1000
    },
    {
      "completion_length": 86.4375,
      "epoch": 2.094240837696335,
      "grad_norm": 16.41309928894043,
      "kl": 6.140625,
      "learning_rate": 4.758909853249476e-07,
      "loss": 0.4598,
      "reward": 0.6292810142040253,
      "reward_std": 0.14345990121364594,
      "reward_std_all": 0.1585000902414322,
      "rewards-max/forensics_sinkhorn_reward": 0.9885641634464264,
      "rewards-max/forensics_strict_edge_reward": 0.9782643616199493,
      "rewards-max/forensics_tiou_reward": 0.9983193278312683,
      "rewards/forensics_sinkhorn_reward": 0.5652430653572083,
      "rewards/forensics_strict_edge_reward": 0.24171321839094162,
      "rewards/forensics_tiou_reward": 0.34157121181488037,
      "step": 1001
    },
    {
      "completion_length": 91.828125,
      "epoch": 2.0963350785340316,
      "grad_norm": 44.476802825927734,
      "kl": 7.25,
      "learning_rate": 4.753668763102725e-07,
      "loss": 0.5327,
      "reward": 0.6331798136234283,
      "reward_std": 0.1740846484899521,
      "reward_std_all": 0.17352013289928436,
      "rewards-max/forensics_sinkhorn_reward": 0.9648197293281555,
      "rewards-max/forensics_strict_edge_reward": 0.8060585260391235,
      "rewards-max/forensics_tiou_reward": 0.9441564977169037,
      "rewards/forensics_sinkhorn_reward": 0.5222151577472687,
      "rewards/forensics_strict_edge_reward": 0.2905459403991699,
      "rewards/forensics_tiou_reward": 0.3610275089740753,
      "step": 1002
    },
    {
      "completion_length": 72.203125,
      "epoch": 2.0984293193717276,
      "grad_norm": 18.291973114013672,
      "kl": 4.828125,
      "learning_rate": 4.7484276729559747e-07,
      "loss": 0.3874,
      "reward": 0.6376413404941559,
      "reward_std": 0.16493867337703705,
      "reward_std_all": 0.17214194685220718,
      "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": 0.5766511857509613,
      "rewards/forensics_strict_edge_reward": 0.3849985897541046,
      "rewards/forensics_tiou_reward": 0.44962188601493835,
      "step": 1003
    },
    {
      "completion_length": 88.859375,
      "epoch": 2.100523560209424,
      "grad_norm": 19.819072723388672,
      "kl": 6.828125,
      "learning_rate": 4.743186582809224e-07,
      "loss": 0.4354,
      "reward": 0.6067393720149994,
      "reward_std": 0.22192512452602386,
      "reward_std_all": 0.21432078629732132,
      "rewards-max/forensics_sinkhorn_reward": 0.9885757565498352,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.998365581035614,
      "rewards/forensics_sinkhorn_reward": 0.5422760844230652,
      "rewards/forensics_strict_edge_reward": 0.3205517902970314,
      "rewards/forensics_tiou_reward": 0.39711610972881317,
      "step": 1004
    },
    {
      "completion_length": 88.6875,
      "epoch": 2.1026178010471206,
      "grad_norm": 33.422264099121094,
      "kl": 6.171875,
      "learning_rate": 4.7379454926624733e-07,
      "loss": 0.3493,
      "reward": 0.5625446140766144,
      "reward_std": 0.24979600310325623,
      "reward_std_all": 0.25017454475164413,
      "rewards-max/forensics_sinkhorn_reward": 0.865157276391983,
      "rewards-max/forensics_strict_edge_reward": 0.7474770247936249,
      "rewards-max/forensics_tiou_reward": 0.7464866936206818,
      "rewards/forensics_sinkhorn_reward": 0.40418729186058044,
      "rewards/forensics_strict_edge_reward": 0.19982361793518066,
      "rewards/forensics_tiou_reward": 0.1884717494249344,
      "step": 1005
    },
    {
      "completion_length": 84.6875,
      "epoch": 2.1047120418848166,
      "grad_norm": 18.650964736938477,
      "kl": 4.703125,
      "learning_rate": 4.732704402515723e-07,
      "loss": 0.3064,
      "reward": 0.5929912030696869,
      "reward_std": 0.2765113115310669,
      "reward_std_all": 0.25377172231674194,
      "rewards-max/forensics_sinkhorn_reward": 0.9851574897766113,
      "rewards-max/forensics_strict_edge_reward": 0.8681221902370453,
      "rewards-max/forensics_tiou_reward": 0.984914630651474,
      "rewards/forensics_sinkhorn_reward": 0.4409525543451309,
      "rewards/forensics_strict_edge_reward": 0.2556133419275284,
      "rewards/forensics_tiou_reward": 0.3115421533584595,
      "step": 1006
    },
    {
      "completion_length": 48.390625,
      "epoch": 2.106806282722513,
      "grad_norm": 25.262609481811523,
      "kl": 5.703125,
      "learning_rate": 4.727463312368973e-07,
      "loss": 0.3698,
      "reward": 0.6077804565429688,
      "reward_std": 0.151320181787014,
      "reward_std_all": 0.17483659088611603,
      "rewards-max/forensics_sinkhorn_reward": 0.9695306420326233,
      "rewards-max/forensics_strict_edge_reward": 0.878633052110672,
      "rewards-max/forensics_tiou_reward": 0.9571479856967926,
      "rewards/forensics_sinkhorn_reward": 0.5115959048271179,
      "rewards/forensics_strict_edge_reward": 0.27957648038864136,
      "rewards/forensics_tiou_reward": 0.32200270891189575,
      "step": 1007
    },
    {
      "completion_length": 65.796875,
      "epoch": 2.1089005235602096,
      "grad_norm": 32.60049819946289,
      "kl": 5.234375,
      "learning_rate": 4.722222222222222e-07,
      "loss": 0.3204,
      "reward": 0.646287739276886,
      "reward_std": 0.1453419327735901,
      "reward_std_all": 0.14736326038837433,
      "rewards-max/forensics_sinkhorn_reward": 0.9880712926387787,
      "rewards-max/forensics_strict_edge_reward": 0.9782643616199493,
      "rewards-max/forensics_tiou_reward": 0.9963833689689636,
      "rewards/forensics_sinkhorn_reward": 0.5889192819595337,
      "rewards/forensics_strict_edge_reward": 0.340556338429451,
      "rewards/forensics_tiou_reward": 0.44575250148773193,
      "step": 1008
    },
    {
      "completion_length": 83.3125,
      "epoch": 2.1109947643979057,
      "grad_norm": 20.672786712646484,
      "kl": 7.359375,
      "learning_rate": 4.7169811320754717e-07,
      "loss": 0.5307,
      "reward": 0.6127558350563049,
      "reward_std": 0.18814583122730255,
      "reward_std_all": 0.19186685979366302,
      "rewards-max/forensics_sinkhorn_reward": 0.983046144247055,
      "rewards-max/forensics_strict_edge_reward": 0.9151731729507446,
      "rewards-max/forensics_tiou_reward": 0.9765892922878265,
      "rewards/forensics_sinkhorn_reward": 0.4606958478689194,
      "rewards/forensics_strict_edge_reward": 0.23927930742502213,
      "rewards/forensics_tiou_reward": 0.32610607147216797,
      "step": 1009
    },
    {
      "completion_length": 82.09375,
      "epoch": 2.113089005235602,
      "grad_norm": 22.343061447143555,
      "kl": 7.125,
      "learning_rate": 4.711740041928721e-07,
      "loss": 0.4365,
      "reward": 0.6272711753845215,
      "reward_std": 0.13577991351485252,
      "reward_std_all": 0.1508430503308773,
      "rewards-max/forensics_sinkhorn_reward": 0.963112473487854,
      "rewards-max/forensics_strict_edge_reward": 0.8575561046600342,
      "rewards-max/forensics_tiou_reward": 0.9416387379169464,
      "rewards/forensics_sinkhorn_reward": 0.5244887471199036,
      "rewards/forensics_strict_edge_reward": 0.2403934895992279,
      "rewards/forensics_tiou_reward": 0.3086051046848297,
      "step": 1010
    },
    {
      "completion_length": 85.9375,
      "epoch": 2.115183246073298,
      "grad_norm": 16.05336570739746,
      "kl": 7.953125,
      "learning_rate": 4.7064989517819703e-07,
      "loss": 0.5986,
      "reward": 0.6378948986530304,
      "reward_std": 0.15045557916164398,
      "reward_std_all": 0.16041625291109085,
      "rewards-max/forensics_sinkhorn_reward": 0.9885916411876678,
      "rewards-max/forensics_strict_edge_reward": 0.9890114367008209,
      "rewards-max/forensics_tiou_reward": 0.9984289109706879,
      "rewards/forensics_sinkhorn_reward": 0.6112089157104492,
      "rewards/forensics_strict_edge_reward": 0.3849460780620575,
      "rewards/forensics_tiou_reward": 0.47764043509960175,
      "step": 1011
    },
    {
      "completion_length": 53.46875,
      "epoch": 2.1172774869109947,
      "grad_norm": 33.069847106933594,
      "kl": 6.84375,
      "learning_rate": 4.70125786163522e-07,
      "loss": 0.5348,
      "reward": 0.6547310948371887,
      "reward_std": 0.13369107618927956,
      "reward_std_all": 0.1395617425441742,
      "rewards-max/forensics_sinkhorn_reward": 0.9740475714206696,
      "rewards-max/forensics_strict_edge_reward": 0.9364311099052429,
      "rewards-max/forensics_tiou_reward": 0.9653099477291107,
      "rewards/forensics_sinkhorn_reward": 0.48312368988990784,
      "rewards/forensics_strict_edge_reward": 0.294494166970253,
      "rewards/forensics_tiou_reward": 0.30171072483062744,
      "step": 1012
    },
    {
      "completion_length": 58.265625,
      "epoch": 2.119371727748691,
      "grad_norm": 18.270465850830078,
      "kl": 5.765625,
      "learning_rate": 4.6960167714884695e-07,
      "loss": 0.2382,
      "reward": 0.6310791075229645,
      "reward_std": 0.22168666124343872,
      "reward_std_all": 0.2038004919886589,
      "rewards-max/forensics_sinkhorn_reward": 0.9859380424022675,
      "rewards-max/forensics_strict_edge_reward": 0.927969753742218,
      "rewards-max/forensics_tiou_reward": 0.9880448281764984,
      "rewards/forensics_sinkhorn_reward": 0.5485662817955017,
      "rewards/forensics_strict_edge_reward": 0.2928631603717804,
      "rewards/forensics_tiou_reward": 0.3696511536836624,
      "step": 1013
    },
    {
      "completion_length": 69.796875,
      "epoch": 2.1214659685863873,
      "grad_norm": 18.60267448425293,
      "kl": 5.984375,
      "learning_rate": 4.6907756813417185e-07,
      "loss": 0.3695,
      "reward": 0.5982877314090729,
      "reward_std": 0.1413935124874115,
      "reward_std_all": 0.16461852937936783,
      "rewards-max/forensics_sinkhorn_reward": 0.9879764020442963,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9960065186023712,
      "rewards/forensics_sinkhorn_reward": 0.5141906887292862,
      "rewards/forensics_strict_edge_reward": 0.25017861276865005,
      "rewards/forensics_tiou_reward": 0.3141613006591797,
      "step": 1014
    },
    {
      "completion_length": 103.5625,
      "epoch": 2.1235602094240837,
      "grad_norm": 17.154298782348633,
      "kl": 7.3125,
      "learning_rate": 4.6855345911949686e-07,
      "loss": 0.4378,
      "reward": 0.7122683227062225,
      "reward_std": 0.08723020926117897,
      "reward_std_all": 0.09405714645981789,
      "rewards-max/forensics_sinkhorn_reward": 0.9861476123332977,
      "rewards-max/forensics_strict_edge_reward": 0.9062341153621674,
      "rewards-max/forensics_tiou_reward": 0.9888230562210083,
      "rewards/forensics_sinkhorn_reward": 0.5848614275455475,
      "rewards/forensics_strict_edge_reward": 0.32338647544384,
      "rewards/forensics_tiou_reward": 0.40388692915439606,
      "step": 1015
    },
    {
      "completion_length": 81.71875,
      "epoch": 2.1256544502617802,
      "grad_norm": 28.067224502563477,
      "kl": 6.421875,
      "learning_rate": 4.680293501048218e-07,
      "loss": 0.4436,
      "reward": 0.6708661913871765,
      "reward_std": 0.17661532759666443,
      "reward_std_all": 0.16193654388189316,
      "rewards-max/forensics_sinkhorn_reward": 0.9566204845905304,
      "rewards-max/forensics_strict_edge_reward": 0.8830007612705231,
      "rewards-max/forensics_tiou_reward": 0.9295732080936432,
      "rewards/forensics_sinkhorn_reward": 0.5922409296035767,
      "rewards/forensics_strict_edge_reward": 0.37362048029899597,
      "rewards/forensics_tiou_reward": 0.4434688687324524,
      "step": 1016
    },
    {
      "completion_length": 92.25,
      "epoch": 2.1277486910994763,
      "grad_norm": 23.098453521728516,
      "kl": 7.40625,
      "learning_rate": 4.6750524109014673e-07,
      "loss": 0.5031,
      "reward": 0.5520718693733215,
      "reward_std": 0.2782229855656624,
      "reward_std_all": 0.26817725598812103,
      "rewards-max/forensics_sinkhorn_reward": 0.9869882464408875,
      "rewards-max/forensics_strict_edge_reward": 0.9158510267734528,
      "rewards-max/forensics_tiou_reward": 0.9921274483203888,
      "rewards/forensics_sinkhorn_reward": 0.45268070697784424,
      "rewards/forensics_strict_edge_reward": 0.2549138143658638,
      "rewards/forensics_tiou_reward": 0.30155205726623535,
      "step": 1017
    },
    {
      "completion_length": 88.421875,
      "epoch": 2.129842931937173,
      "grad_norm": 37.13483428955078,
      "kl": 4.7734375,
      "learning_rate": 4.669811320754717e-07,
      "loss": 0.3542,
      "reward": 0.6449683904647827,
      "reward_std": 0.13913995027542114,
      "reward_std_all": 0.15315087139606476,
      "rewards-max/forensics_sinkhorn_reward": 0.9884204268455505,
      "rewards-max/forensics_strict_edge_reward": 0.9782643616199493,
      "rewards-max/forensics_tiou_reward": 0.9977528154850006,
      "rewards/forensics_sinkhorn_reward": 0.5636605620384216,
      "rewards/forensics_strict_edge_reward": 0.30727341026067734,
      "rewards/forensics_tiou_reward": 0.3792146146297455,
      "step": 1018
    },
    {
      "completion_length": 103.359375,
      "epoch": 2.1319371727748693,
      "grad_norm": 17.53838539123535,
      "kl": 6.3515625,
      "learning_rate": 4.6645702306079665e-07,
      "loss": 0.2991,
      "reward": 0.5575358271598816,
      "reward_std": 0.2252904772758484,
      "reward_std_all": 0.23165304958820343,
      "rewards-max/forensics_sinkhorn_reward": 0.9882677495479584,
      "rewards-max/forensics_strict_edge_reward": 0.9574736058712006,
      "rewards-max/forensics_tiou_reward": 0.9971550703048706,
      "rewards/forensics_sinkhorn_reward": 0.4345383942127228,
      "rewards/forensics_strict_edge_reward": 0.2809266895055771,
      "rewards/forensics_tiou_reward": 0.2934882640838623,
      "step": 1019
    },
    {
      "completion_length": 86.96875,
      "epoch": 2.1340314136125653,
      "grad_norm": 27.988767623901367,
      "kl": 6.0625,
      "learning_rate": 4.6593291404612155e-07,
      "loss": 0.4394,
      "reward": 0.6240952610969543,
      "reward_std": 0.1653667911887169,
      "reward_std_all": 0.17873825132846832,
      "rewards-max/forensics_sinkhorn_reward": 0.970859557390213,
      "rewards-max/forensics_strict_edge_reward": 0.8650398254394531,
      "rewards-max/forensics_tiou_reward": 0.9573504030704498,
      "rewards/forensics_sinkhorn_reward": 0.5612443685531616,
      "rewards/forensics_strict_edge_reward": 0.33412453532218933,
      "rewards/forensics_tiou_reward": 0.39810530841350555,
      "step": 1020
    },
    {
      "completion_length": 91.953125,
      "epoch": 2.136125654450262,
      "grad_norm": 25.4510498046875,
      "kl": 7.6875,
      "learning_rate": 4.654088050314465e-07,
      "loss": 0.4265,
      "reward": 0.5594937205314636,
      "reward_std": 0.2056591659784317,
      "reward_std_all": 0.2258945256471634,
      "rewards-max/forensics_sinkhorn_reward": 0.952529102563858,
      "rewards-max/forensics_strict_edge_reward": 0.842301219701767,
      "rewards-max/forensics_tiou_reward": 0.9167484641075134,
      "rewards/forensics_sinkhorn_reward": 0.4450176954269409,
      "rewards/forensics_strict_edge_reward": 0.1836487278342247,
      "rewards/forensics_tiou_reward": 0.24992969632148743,
      "step": 1021
    },
    {
      "completion_length": 59.296875,
      "epoch": 2.1382198952879583,
      "grad_norm": 29.80449104309082,
      "kl": 5.0,
      "learning_rate": 4.6488469601677147e-07,
      "loss": 0.3213,
      "reward": 0.6386842429637909,
      "reward_std": 0.15031235665082932,
      "reward_std_all": 0.16229917109012604,
      "rewards-max/forensics_sinkhorn_reward": 0.8989050686359406,
      "rewards-max/forensics_strict_edge_reward": 0.7127212733030319,
      "rewards-max/forensics_tiou_reward": 0.8052392899990082,
      "rewards/forensics_sinkhorn_reward": 0.5373720526695251,
      "rewards/forensics_strict_edge_reward": 0.2660103365778923,
      "rewards/forensics_tiou_reward": 0.3232949934899807,
      "step": 1022
    },
    {
      "completion_length": 78.625,
      "epoch": 2.1403141361256544,
      "grad_norm": 21.30466079711914,
      "kl": 5.78125,
      "learning_rate": 4.6436058700209643e-07,
      "loss": 0.4607,
      "reward": 0.6026343107223511,
      "reward_std": 0.17285223305225372,
      "reward_std_all": 0.18848221749067307,
      "rewards-max/forensics_sinkhorn_reward": 0.9873584508895874,
      "rewards-max/forensics_strict_edge_reward": 0.9565287232398987,
      "rewards-max/forensics_tiou_reward": 0.9935698509216309,
      "rewards/forensics_sinkhorn_reward": 0.49822378158569336,
      "rewards/forensics_strict_edge_reward": 0.2907692939043045,
      "rewards/forensics_tiou_reward": 0.3232213407754898,
      "step": 1023
    },
    {
      "completion_length": 31.875,
      "epoch": 2.142408376963351,
      "grad_norm": 21.2419490814209,
      "kl": 2.9765625,
      "learning_rate": 4.638364779874214e-07,
      "loss": 0.187,
      "reward": 0.6550524234771729,
      "reward_std": 0.1251489222049713,
      "reward_std_all": 0.1380968764424324,
      "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": 0.557280570268631,
      "rewards/forensics_strict_edge_reward": 0.3481796234846115,
      "rewards/forensics_tiou_reward": 0.3897251635789871,
      "step": 1024
    },
    {
      "completion_length": 76.765625,
      "epoch": 2.144502617801047,
      "grad_norm": 21.926300048828125,
      "kl": 5.203125,
      "learning_rate": 4.6331236897274635e-07,
      "loss": 0.3952,
      "reward": 0.6632147431373596,
      "reward_std": 0.14407142996788025,
      "reward_std_all": 0.1430174857378006,
      "rewards-max/forensics_sinkhorn_reward": 0.9842172265052795,
      "rewards-max/forensics_strict_edge_reward": 0.8968285918235779,
      "rewards-max/forensics_tiou_reward": 0.9813249409198761,
      "rewards/forensics_sinkhorn_reward": 0.542399525642395,
      "rewards/forensics_strict_edge_reward": 0.2614295408129692,
      "rewards/forensics_tiou_reward": 0.30948226153850555,
      "step": 1025
    },
    {
      "completion_length": 64.796875,
      "epoch": 2.1465968586387434,
      "grad_norm": 26.2082462310791,
      "kl": 5.6875,
      "learning_rate": 4.6278825995807125e-07,
      "loss": 0.3129,
      "reward": 0.571562796831131,
      "reward_std": 0.16628284007310867,
      "reward_std_all": 0.19776368141174316,
      "rewards-max/forensics_sinkhorn_reward": 0.959701657295227,
      "rewards-max/forensics_strict_edge_reward": 0.7652452886104584,
      "rewards-max/forensics_tiou_reward": 0.9304824769496918,
      "rewards/forensics_sinkhorn_reward": 0.4396500885486603,
      "rewards/forensics_strict_edge_reward": 0.17821035906672478,
      "rewards/forensics_tiou_reward": 0.22394324094057083,
      "step": 1026
    },
    {
      "completion_length": 62.4375,
      "epoch": 2.14869109947644,
      "grad_norm": 25.49532699584961,
      "kl": 5.0625,
      "learning_rate": 4.622641509433962e-07,
      "loss": 0.3318,
      "reward": 0.6049788594245911,
      "reward_std": 0.19497868418693542,
      "reward_std_all": 0.1945076584815979,
      "rewards-max/forensics_sinkhorn_reward": 0.9876801371574402,
      "rewards-max/forensics_strict_edge_reward": 0.9565287232398987,
      "rewards-max/forensics_tiou_reward": 0.9948413372039795,
      "rewards/forensics_sinkhorn_reward": 0.49004116654396057,
      "rewards/forensics_strict_edge_reward": 0.2632337659597397,
      "rewards/forensics_tiou_reward": 0.31039266288280487,
      "step": 1027
    },
    {
      "completion_length": 80.578125,
      "epoch": 2.150785340314136,
      "grad_norm": 19.32423210144043,
      "kl": 7.546875,
      "learning_rate": 4.6174004192872117e-07,
      "loss": 0.4542,
      "reward": 0.5758694112300873,
      "reward_std": 0.1702167522162199,
      "reward_std_all": 0.19155213981866837,
      "rewards-max/forensics_sinkhorn_reward": 0.9565849900245667,
      "rewards-max/forensics_strict_edge_reward": 0.9093653857707977,
      "rewards-max/forensics_tiou_reward": 0.9308341145515442,
      "rewards/forensics_sinkhorn_reward": 0.5016561448574066,
      "rewards/forensics_strict_edge_reward": 0.28281375765800476,
      "rewards/forensics_tiou_reward": 0.3327212855219841,
      "step": 1028
    },
    {
      "completion_length": 54.0,
      "epoch": 2.1528795811518324,
      "grad_norm": 19.987472534179688,
      "kl": 3.953125,
      "learning_rate": 4.612159329140461e-07,
      "loss": 0.1995,
      "reward": 0.6443432569503784,
      "reward_std": 0.17828378081321716,
      "reward_std_all": 0.17311272770166397,
      "rewards-max/forensics_sinkhorn_reward": 0.9887502789497375,
      "rewards-max/forensics_strict_edge_reward": 0.9890114367008209,
      "rewards-max/forensics_tiou_reward": 0.9990534782409668,
      "rewards/forensics_sinkhorn_reward": 0.5402871668338776,
      "rewards/forensics_strict_edge_reward": 0.3184938281774521,
      "rewards/forensics_tiou_reward": 0.37915948033332825,
      "step": 1029
    },
    {
      "completion_length": 48.203125,
      "epoch": 2.154973821989529,
      "grad_norm": 17.87361717224121,
      "kl": 4.296875,
      "learning_rate": 4.6069182389937103e-07,
      "loss": 0.3244,
      "reward": 0.6087417006492615,
      "reward_std": 0.19396497309207916,
      "reward_std_all": 0.19592084735631943,
      "rewards-max/forensics_sinkhorn_reward": 0.9883498847484589,
      "rewards-max/forensics_strict_edge_reward": 0.9574736058712006,
      "rewards-max/forensics_tiou_reward": 0.9974779486656189,
      "rewards/forensics_sinkhorn_reward": 0.5481481552124023,
      "rewards/forensics_strict_edge_reward": 0.26654234528541565,
      "rewards/forensics_tiou_reward": 0.3551781624555588,
      "step": 1030
    },
    {
      "completion_length": 60.09375,
      "epoch": 2.157068062827225,
      "grad_norm": 21.670995712280273,
      "kl": 8.484375,
      "learning_rate": 4.6016771488469605e-07,
      "loss": 0.5973,
      "reward": 0.6407280266284943,
      "reward_std": 0.20465749502182007,
      "reward_std_all": 0.1890726163983345,
      "rewards-max/forensics_sinkhorn_reward": 0.9886867105960846,
      "rewards-max/forensics_strict_edge_reward": 0.9890114367008209,
      "rewards-max/forensics_tiou_reward": 0.9988023936748505,
      "rewards/forensics_sinkhorn_reward": 0.5953458845615387,
      "rewards/forensics_strict_edge_reward": 0.3960459232330322,
      "rewards/forensics_tiou_reward": 0.4438652843236923,
      "step": 1031
    },
    {
      "completion_length": 71.0,
      "epoch": 2.1591623036649215,
      "grad_norm": 20.212129592895508,
      "kl": 5.6328125,
      "learning_rate": 4.5964360587002095e-07,
      "loss": 0.3965,
      "reward": 0.6141400933265686,
      "reward_std": 0.18329034000635147,
      "reward_std_all": 0.18503958731889725,
      "rewards-max/forensics_sinkhorn_reward": 0.9883707463741302,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9975584745407104,
      "rewards/forensics_sinkhorn_reward": 0.5017402172088623,
      "rewards/forensics_strict_edge_reward": 0.25786828249692917,
      "rewards/forensics_tiou_reward": 0.30384302139282227,
      "step": 1032
    },
    {
      "completion_length": 62.515625,
      "epoch": 2.161256544502618,
      "grad_norm": 16.435375213623047,
      "kl": 6.859375,
      "learning_rate": 4.591194968553459e-07,
      "loss": 0.4287,
      "reward": 0.675085723400116,
      "reward_std": 0.12956856563687325,
      "reward_std_all": 0.13315843045711517,
      "rewards-max/forensics_sinkhorn_reward": 0.9882985651493073,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9972721934318542,
      "rewards/forensics_sinkhorn_reward": 0.6188889443874359,
      "rewards/forensics_strict_edge_reward": 0.40269799530506134,
      "rewards/forensics_tiou_reward": 0.4785424917936325,
      "step": 1033
    },
    {
      "completion_length": 73.765625,
      "epoch": 2.163350785340314,
      "grad_norm": 36.764156341552734,
      "kl": 10.0625,
      "learning_rate": 4.5859538784067087e-07,
      "loss": 0.7069,
      "reward": 0.5841104984283447,
      "reward_std": 0.16622349619865417,
      "reward_std_all": 0.18725696951150894,
      "rewards-max/forensics_sinkhorn_reward": 0.9883950054645538,
      "rewards-max/forensics_strict_edge_reward": 0.9890114367008209,
      "rewards-max/forensics_tiou_reward": 0.9976553320884705,
      "rewards/forensics_sinkhorn_reward": 0.4914717674255371,
      "rewards/forensics_strict_edge_reward": 0.283113569021225,
      "rewards/forensics_tiou_reward": 0.30085863173007965,
      "step": 1034
    },
    {
      "completion_length": 56.765625,
      "epoch": 2.1654450261780105,
      "grad_norm": 22.12796401977539,
      "kl": 5.453125,
      "learning_rate": 4.580712788259958e-07,
      "loss": 0.3351,
      "reward": 0.6439395844936371,
      "reward_std": 0.12559646368026733,
      "reward_std_all": 0.1396087370812893,
      "rewards-max/forensics_sinkhorn_reward": 0.9877926111221313,
      "rewards-max/forensics_strict_edge_reward": 0.9567649364471436,
      "rewards-max/forensics_tiou_reward": 0.9952793121337891,
      "rewards/forensics_sinkhorn_reward": 0.5786498785018921,
      "rewards/forensics_strict_edge_reward": 0.3019614517688751,
      "rewards/forensics_tiou_reward": 0.3626369684934616,
      "step": 1035
    },
    {
      "completion_length": 46.015625,
      "epoch": 2.167539267015707,
      "grad_norm": 32.54085922241211,
      "kl": 6.265625,
      "learning_rate": 4.5754716981132073e-07,
      "loss": 0.4562,
      "reward": 0.6282128691673279,
      "reward_std": 0.13981427252292633,
      "reward_std_all": 0.15819044411182404,
      "rewards-max/forensics_sinkhorn_reward": 0.9752812683582306,
      "rewards-max/forensics_strict_edge_reward": 0.9567649364471436,
      "rewards-max/forensics_tiou_reward": 0.9711910784244537,
      "rewards/forensics_sinkhorn_reward": 0.4929487854242325,
      "rewards/forensics_strict_edge_reward": 0.25833966583013535,
      "rewards/forensics_tiou_reward": 0.2995792031288147,
      "step": 1036
    },
    {
      "completion_length": 57.359375,
      "epoch": 2.169633507853403,
      "grad_norm": 20.372467041015625,
      "kl": 7.03125,
      "learning_rate": 4.5702306079664564e-07,
      "loss": 0.5054,
      "reward": 0.5860788524150848,
      "reward_std": 0.1635824292898178,
      "reward_std_all": 0.1856476590037346,
      "rewards-max/forensics_sinkhorn_reward": 0.9883235692977905,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9973708987236023,
      "rewards/forensics_sinkhorn_reward": 0.5240086019039154,
      "rewards/forensics_strict_edge_reward": 0.2744199335575104,
      "rewards/forensics_tiou_reward": 0.32836824655532837,
      "step": 1037
    },
    {
      "completion_length": 80.890625,
      "epoch": 2.1717277486910995,
      "grad_norm": 25.678592681884766,
      "kl": 7.71875,
      "learning_rate": 4.564989517819706e-07,
      "loss": 0.4383,
      "reward": 0.6376702785491943,
      "reward_std": 0.1776065081357956,
      "reward_std_all": 0.17719487845897675,
      "rewards-max/forensics_sinkhorn_reward": 0.986532062292099,
      "rewards-max/forensics_strict_edge_reward": 0.9357379674911499,
      "rewards-max/forensics_tiou_reward": 0.9903355538845062,
      "rewards/forensics_sinkhorn_reward": 0.513346403837204,
      "rewards/forensics_strict_edge_reward": 0.2546152025461197,
      "rewards/forensics_tiou_reward": 0.32123179733753204,
      "step": 1038
    },
    {
      "completion_length": 67.96875,
      "epoch": 2.173821989528796,
      "grad_norm": 25.31199073791504,
      "kl": 6.71875,
      "learning_rate": 4.559748427672956e-07,
      "loss": 0.3112,
      "reward": 0.6152746379375458,
      "reward_std": 0.15615389496088028,
      "reward_std_all": 0.1729232743382454,
      "rewards-max/forensics_sinkhorn_reward": 0.9883261919021606,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9973829090595245,
      "rewards/forensics_sinkhorn_reward": 0.5372796058654785,
      "rewards/forensics_strict_edge_reward": 0.28014765679836273,
      "rewards/forensics_tiou_reward": 0.3434605002403259,
      "step": 1039
    },
    {
      "completion_length": 57.609375,
      "epoch": 2.175916230366492,
      "grad_norm": 23.714567184448242,
      "kl": 6.5703125,
      "learning_rate": 4.554507337526205e-07,
      "loss": 0.5167,
      "reward": 0.6571579277515411,
      "reward_std": 0.13240570947527885,
      "reward_std_all": 0.13962895423173904,
      "rewards-max/forensics_sinkhorn_reward": 0.9883860945701599,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9976174533367157,
      "rewards/forensics_sinkhorn_reward": 0.51341313123703,
      "rewards/forensics_strict_edge_reward": 0.31832583248615265,
      "rewards/forensics_tiou_reward": 0.3346560001373291,
      "step": 1040
    },
    {
      "completion_length": 71.734375,
      "epoch": 2.1780104712041886,
      "grad_norm": 17.013957977294922,
      "kl": 6.953125,
      "learning_rate": 4.549266247379455e-07,
      "loss": 0.4593,
      "reward": 0.6628890931606293,
      "reward_std": 0.13209164515137672,
      "reward_std_all": 0.1379808858036995,
      "rewards-max/forensics_sinkhorn_reward": 0.9885820150375366,
      "rewards-max/forensics_strict_edge_reward": 0.9782643616199493,
      "rewards-max/forensics_tiou_reward": 0.9983910024166107,
      "rewards/forensics_sinkhorn_reward": 0.5638786256313324,
      "rewards/forensics_strict_edge_reward": 0.36796337366104126,
      "rewards/forensics_tiou_reward": 0.397390753030777,
      "step": 1041
    },
    {
      "completion_length": 93.859375,
      "epoch": 2.180104712041885,
      "grad_norm": 18.93560791015625,
      "kl": 6.8125,
      "learning_rate": 4.5440251572327043e-07,
      "loss": 0.4444,
      "reward": 0.6268790364265442,
      "reward_std": 0.16844388097524643,
      "reward_std_all": 0.17050474137067795,
      "rewards-max/forensics_sinkhorn_reward": 0.9075802862644196,
      "rewards-max/forensics_strict_edge_reward": 0.8113168179988861,
      "rewards-max/forensics_tiou_reward": 0.8242135047912598,
      "rewards/forensics_sinkhorn_reward": 0.4564220607280731,
      "rewards/forensics_strict_edge_reward": 0.21289537847042084,
      "rewards/forensics_tiou_reward": 0.2572254091501236,
      "step": 1042
    },
    {
      "completion_length": 47.25,
      "epoch": 2.182198952879581,
      "grad_norm": 19.389629364013672,
      "kl": 3.421875,
      "learning_rate": 4.5387840670859534e-07,
      "loss": 0.3239,
      "reward": 0.6732956767082214,
      "reward_std": 0.12830384448170662,
      "reward_std_all": 0.12884247303009033,
      "rewards-max/forensics_sinkhorn_reward": 0.9877497255802155,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9951091110706329,
      "rewards/forensics_sinkhorn_reward": 0.6531208455562592,
      "rewards/forensics_strict_edge_reward": 0.4119913727045059,
      "rewards/forensics_tiou_reward": 0.4758168160915375,
      "step": 1043
    },
    {
      "completion_length": 90.84375,
      "epoch": 2.1842931937172776,
      "grad_norm": 23.47580909729004,
      "kl": 5.859375,
      "learning_rate": 4.533542976939203e-07,
      "loss": 0.4784,
      "reward": 0.6743860840797424,
      "reward_std": 0.14413990825414658,
      "reward_std_all": 0.14128538221120834,
      "rewards-max/forensics_sinkhorn_reward": 0.9885916411876678,
      "rewards-max/forensics_strict_edge_reward": 0.9890114367008209,
      "rewards-max/forensics_tiou_reward": 0.9984289109706879,
      "rewards/forensics_sinkhorn_reward": 0.5719740390777588,
      "rewards/forensics_strict_edge_reward": 0.3456294387578964,
      "rewards/forensics_tiou_reward": 0.42174510657787323,
      "step": 1044
    },
    {
      "completion_length": 92.265625,
      "epoch": 2.1863874345549736,
      "grad_norm": 29.01451873779297,
      "kl": 7.421875,
      "learning_rate": 4.5283018867924526e-07,
      "loss": 0.4984,
      "reward": 0.6150438189506531,
      "reward_std": 0.21403898298740387,
      "reward_std_all": 0.20199060440063477,
      "rewards-max/forensics_sinkhorn_reward": 0.9863774478435516,
      "rewards-max/forensics_strict_edge_reward": 0.946017861366272,
      "rewards-max/forensics_tiou_reward": 0.9897368550300598,
      "rewards/forensics_sinkhorn_reward": 0.4763755798339844,
      "rewards/forensics_strict_edge_reward": 0.3232601433992386,
      "rewards/forensics_tiou_reward": 0.35410644114017487,
      "step": 1045
    },
    {
      "completion_length": 77.53125,
      "epoch": 2.18848167539267,
      "grad_norm": 18.70359992980957,
      "kl": 6.109375,
      "learning_rate": 4.5230607966457016e-07,
      "loss": 0.4364,
      "reward": 0.6112054586410522,
      "reward_std": 0.15752942860126495,
      "reward_std_all": 0.17124389111995697,
      "rewards-max/forensics_sinkhorn_reward": 0.9827140271663666,
      "rewards-max/forensics_strict_edge_reward": 0.946017861366272,
      "rewards-max/forensics_tiou_reward": 0.984082818031311,
      "rewards/forensics_sinkhorn_reward": 0.5149649083614349,
      "rewards/forensics_strict_edge_reward": 0.2775490805506706,
      "rewards/forensics_tiou_reward": 0.36197519302368164,
      "step": 1046
    },
    {
      "completion_length": 105.4375,
      "epoch": 2.1905759162303666,
      "grad_norm": 30.604816436767578,
      "kl": 5.859375,
      "learning_rate": 4.5178197064989517e-07,
      "loss": 0.3341,
      "reward": 0.62168949842453,
      "reward_std": 0.17554568499326706,
      "reward_std_all": 0.1780264750123024,
      "rewards-max/forensics_sinkhorn_reward": 0.9881310164928436,
      "rewards-max/forensics_strict_edge_reward": 0.9464850425720215,
      "rewards-max/forensics_tiou_reward": 0.9966123998165131,
      "rewards/forensics_sinkhorn_reward": 0.530684232711792,
      "rewards/forensics_strict_edge_reward": 0.3454861640930176,
      "rewards/forensics_tiou_reward": 0.3959274888038635,
      "step": 1047
    },
    {
      "completion_length": 79.875,
      "epoch": 2.1926701570680627,
      "grad_norm": 23.877647399902344,
      "kl": 5.328125,
      "learning_rate": 4.5125786163522013e-07,
      "loss": 0.4729,
      "reward": 0.5924788415431976,
      "reward_std": 0.2228277325630188,
      "reward_std_all": 0.21430180966854095,
      "rewards-max/forensics_sinkhorn_reward": 0.9865047037601471,
      "rewards-max/forensics_strict_edge_reward": 0.946017861366272,
      "rewards-max/forensics_tiou_reward": 0.9902204871177673,
      "rewards/forensics_sinkhorn_reward": 0.4988718330860138,
      "rewards/forensics_strict_edge_reward": 0.3324854075908661,
      "rewards/forensics_tiou_reward": 0.37589307129383087,
      "step": 1048
    },
    {
      "completion_length": 55.71875,
      "epoch": 2.194764397905759,
      "grad_norm": 18.328184127807617,
      "kl": 4.296875,
      "learning_rate": 4.5073375262054504e-07,
      "loss": 0.4519,
      "reward": 0.6908497214317322,
      "reward_std": 0.11220605671405792,
      "reward_std_all": 0.12028755247592926,
      "rewards-max/forensics_sinkhorn_reward": 0.9880844354629517,
      "rewards-max/forensics_strict_edge_reward": 0.9782643616199493,
      "rewards-max/forensics_tiou_reward": 0.9964285790920258,
      "rewards/forensics_sinkhorn_reward": 0.6335214376449585,
      "rewards/forensics_strict_edge_reward": 0.44033674895763397,
      "rewards/forensics_tiou_reward": 0.5373464524745941,
      "step": 1049
    },
    {
      "completion_length": 80.078125,
      "epoch": 2.1968586387434557,
      "grad_norm": 15.727424621582031,
      "kl": 5.546875,
      "learning_rate": 4.5020964360587e-07,
      "loss": 0.3201,
      "reward": 0.6730127334594727,
      "reward_std": 0.14041505753993988,
      "reward_std_all": 0.1410263627767563,
      "rewards-max/forensics_sinkhorn_reward": 0.9876928925514221,
      "rewards-max/forensics_strict_edge_reward": 0.9565287232398987,
      "rewards-max/forensics_tiou_reward": 0.9948849081993103,
      "rewards/forensics_sinkhorn_reward": 0.6049100160598755,
      "rewards/forensics_strict_edge_reward": 0.3956729918718338,
      "rewards/forensics_tiou_reward": 0.45996488630771637,
      "step": 1050
    },
    {
      "completion_length": 117.609375,
      "epoch": 2.1989528795811517,
      "grad_norm": 31.260135650634766,
      "kl": 8.84375,
      "learning_rate": 4.4968553459119495e-07,
      "loss": 0.5674,
      "reward": 0.6196895837783813,
      "reward_std": 0.20850733667612076,
      "reward_std_all": 0.19813715666532516,
      "rewards-max/forensics_sinkhorn_reward": 0.9883580803871155,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9975085258483887,
      "rewards/forensics_sinkhorn_reward": 0.49931544065475464,
      "rewards/forensics_strict_edge_reward": 0.321720689535141,
      "rewards/forensics_tiou_reward": 0.38665641844272614,
      "step": 1051
    },
    {
      "completion_length": 106.578125,
      "epoch": 2.201047120418848,
      "grad_norm": 19.741451263427734,
      "kl": 6.640625,
      "learning_rate": 4.4916142557651986e-07,
      "loss": 0.4865,
      "reward": 0.5989394783973694,
      "reward_std": 0.2159448340535164,
      "reward_std_all": 0.20877178013324738,
      "rewards-max/forensics_sinkhorn_reward": 0.9879411458969116,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9958676695823669,
      "rewards/forensics_sinkhorn_reward": 0.5229483544826508,
      "rewards/forensics_strict_edge_reward": 0.3239094018936157,
      "rewards/forensics_tiou_reward": 0.38105612993240356,
      "step": 1052
    },
    {
      "completion_length": 76.890625,
      "epoch": 2.2031413612565447,
      "grad_norm": 29.737749099731445,
      "kl": 5.65625,
      "learning_rate": 4.486373165618448e-07,
      "loss": 0.4288,
      "reward": 0.5478824377059937,
      "reward_std": 0.18224620074033737,
      "reward_std_all": 0.21235424280166626,
      "rewards-max/forensics_sinkhorn_reward": 0.9880184531211853,
      "rewards-max/forensics_strict_edge_reward": 0.9782643616199493,
      "rewards-max/forensics_tiou_reward": 0.9961759150028229,
      "rewards/forensics_sinkhorn_reward": 0.44102543592453003,
      "rewards/forensics_strict_edge_reward": 0.18909801542758942,
      "rewards/forensics_tiou_reward": 0.23160579055547714,
      "step": 1053
    },
    {
      "completion_length": 80.796875,
      "epoch": 2.2052356020942407,
      "grad_norm": 24.427894592285156,
      "kl": 5.609375,
      "learning_rate": 4.481132075471698e-07,
      "loss": 0.4396,
      "reward": 0.6141109764575958,
      "reward_std": 0.2537960112094879,
      "reward_std_all": 0.23118475824594498,
      "rewards-max/forensics_sinkhorn_reward": 0.9880042374134064,
      "rewards-max/forensics_strict_edge_reward": 0.9565287232398987,
      "rewards-max/forensics_tiou_reward": 0.9961137771606445,
      "rewards/forensics_sinkhorn_reward": 0.538529634475708,
      "rewards/forensics_strict_edge_reward": 0.3531864434480667,
      "rewards/forensics_tiou_reward": 0.42026206851005554,
      "step": 1054
    },
    {
      "completion_length": 61.609375,
      "epoch": 2.2073298429319372,
      "grad_norm": 24.040536880493164,
      "kl": 6.484375,
      "learning_rate": 4.4758909853249474e-07,
      "loss": 0.4845,
      "reward": 0.6554490923881531,
      "reward_std": 0.15511569380760193,
      "reward_std_all": 0.15454450249671936,
      "rewards-max/forensics_sinkhorn_reward": 0.9868140816688538,
      "rewards-max/forensics_strict_edge_reward": 0.9357379674911499,
      "rewards-max/forensics_tiou_reward": 0.9914548397064209,
      "rewards/forensics_sinkhorn_reward": 0.5286179333925247,
      "rewards/forensics_strict_edge_reward": 0.2789747044444084,
      "rewards/forensics_tiou_reward": 0.31934142112731934,
      "step": 1055
    },
    {
      "completion_length": 69.46875,
      "epoch": 2.2094240837696333,
      "grad_norm": 26.530115127563477,
      "kl": 5.7421875,
      "learning_rate": 4.470649895178197e-07,
      "loss": 0.4222,
      "reward": 0.622946709394455,
      "reward_std": 0.14635667949914932,
      "reward_std_all": 0.1642000377178192,
      "rewards-max/forensics_sinkhorn_reward": 0.9794841110706329,
      "rewards-max/forensics_strict_edge_reward": 0.842301219701767,
      "rewards-max/forensics_tiou_reward": 0.9629928767681122,
      "rewards/forensics_sinkhorn_reward": 0.4997304379940033,
      "rewards/forensics_strict_edge_reward": 0.24043643474578857,
      "rewards/forensics_tiou_reward": 0.29378579556941986,
      "step": 1056
    },
    {
      "completion_length": 117.28125,
      "epoch": 2.2115183246073298,
      "grad_norm": 20.637483596801758,
      "kl": 8.890625,
      "learning_rate": 4.4654088050314465e-07,
      "loss": 0.5274,
      "reward": 0.5876744091510773,
      "reward_std": 0.21095839142799377,
      "reward_std_all": 0.21404480934143066,
      "rewards-max/forensics_sinkhorn_reward": 0.9872908294200897,
      "rewards-max/forensics_strict_edge_reward": 0.9355069994926453,
      "rewards-max/forensics_tiou_reward": 0.993314802646637,
      "rewards/forensics_sinkhorn_reward": 0.4308520555496216,
      "rewards/forensics_strict_edge_reward": 0.23388073593378067,
      "rewards/forensics_tiou_reward": 0.27185747027397156,
      "step": 1057
    },
    {
      "completion_length": 70.3125,
      "epoch": 2.2136125654450263,
      "grad_norm": 21.703001022338867,
      "kl": 5.640625,
      "learning_rate": 4.4601677148846956e-07,
      "loss": 0.3647,
      "reward": 0.6873413920402527,
      "reward_std": 0.14801695197820663,
      "reward_std_all": 0.14415422081947327,
      "rewards-max/forensics_sinkhorn_reward": 0.9872665703296661,
      "rewards-max/forensics_strict_edge_reward": 0.9357379674911499,
      "rewards-max/forensics_tiou_reward": 0.9932162463665009,
      "rewards/forensics_sinkhorn_reward": 0.5829799473285675,
      "rewards/forensics_strict_edge_reward": 0.3545856922864914,
      "rewards/forensics_tiou_reward": 0.42616353929042816,
      "step": 1058
    },
    {
      "completion_length": 88.09375,
      "epoch": 2.2157068062827223,
      "grad_norm": 22.971986770629883,
      "kl": 6.3125,
      "learning_rate": 4.454926624737945e-07,
      "loss": 0.4574,
      "reward": 0.638249009847641,
      "reward_std": 0.17992721498012543,
      "reward_std_all": 0.17249976098537445,
      "rewards-max/forensics_sinkhorn_reward": 0.9868064224720001,
      "rewards-max/forensics_strict_edge_reward": 0.9464850425720215,
      "rewards-max/forensics_tiou_reward": 0.9914200901985168,
      "rewards/forensics_sinkhorn_reward": 0.5574645400047302,
      "rewards/forensics_strict_edge_reward": 0.2649519070982933,
      "rewards/forensics_tiou_reward": 0.3698982298374176,
      "step": 1059
    },
    {
      "completion_length": 51.609375,
      "epoch": 2.217801047120419,
      "grad_norm": 29.216251373291016,
      "kl": 5.484375,
      "learning_rate": 4.449685534591195e-07,
      "loss": 0.237,
      "reward": 0.6932553946971893,
      "reward_std": 0.10364505276083946,
      "reward_std_all": 0.11021661758422852,
      "rewards-max/forensics_sinkhorn_reward": 0.98738032579422,
      "rewards-max/forensics_strict_edge_reward": 0.9677534997463226,
      "rewards-max/forensics_tiou_reward": 0.993674099445343,
      "rewards/forensics_sinkhorn_reward": 0.5721528828144073,
      "rewards/forensics_strict_edge_reward": 0.25777480006217957,
      "rewards/forensics_tiou_reward": 0.34011687338352203,
      "step": 1060
    },
    {
      "completion_length": 51.984375,
      "epoch": 2.2198952879581153,
      "grad_norm": 19.422998428344727,
      "kl": 5.0,
      "learning_rate": 4.444444444444444e-07,
      "loss": 0.2528,
      "reward": 0.620909720659256,
      "reward_std": 0.1501096561551094,
      "reward_std_all": 0.16564220190048218,
      "rewards-max/forensics_sinkhorn_reward": 0.968480259180069,
      "rewards-max/forensics_strict_edge_reward": 0.8983768224716187,
      "rewards-max/forensics_tiou_reward": 0.9543398022651672,
      "rewards/forensics_sinkhorn_reward": 0.5237612128257751,
      "rewards/forensics_strict_edge_reward": 0.28522706031799316,
      "rewards/forensics_tiou_reward": 0.3221900314092636,
      "step": 1061
    },
    {
      "completion_length": 52.71875,
      "epoch": 2.2219895287958114,
      "grad_norm": 29.917232513427734,
      "kl": 6.96875,
      "learning_rate": 4.4392033542976934e-07,
      "loss": 0.4084,
      "reward": 0.6539541184902191,
      "reward_std": 0.15856575220823288,
      "reward_std_all": 0.16090166568756104,
      "rewards-max/forensics_sinkhorn_reward": 0.9875067472457886,
      "rewards-max/forensics_strict_edge_reward": 0.9574736058712006,
      "rewards-max/forensics_tiou_reward": 0.994169145822525,
      "rewards/forensics_sinkhorn_reward": 0.5817475616931915,
      "rewards/forensics_strict_edge_reward": 0.3373613506555557,
      "rewards/forensics_tiou_reward": 0.3999772369861603,
      "step": 1062
    },
    {
      "completion_length": 35.0,
      "epoch": 2.224083769633508,
      "grad_norm": 18.254348754882812,
      "kl": 2.4921875,
      "learning_rate": 4.4339622641509435e-07,
      "loss": 0.1516,
      "reward": 0.6860161125659943,
      "reward_std": 0.09837546665221453,
      "reward_std_all": 0.10999575816094875,
      "rewards-max/forensics_sinkhorn_reward": 0.9866842925548553,
      "rewards-max/forensics_strict_edge_reward": 0.87538942694664,
      "rewards-max/forensics_tiou_reward": 0.9909128546714783,
      "rewards/forensics_sinkhorn_reward": 0.5731053799390793,
      "rewards/forensics_strict_edge_reward": 0.24681617319583893,
      "rewards/forensics_tiou_reward": 0.3178688585758209,
      "step": 1063
    },
    {
      "completion_length": 44.390625,
      "epoch": 2.2261780104712043,
      "grad_norm": 25.371967315673828,
      "kl": 4.65625,
      "learning_rate": 4.4287211740041926e-07,
      "loss": 0.2503,
      "reward": 0.7099323272705078,
      "reward_std": 0.09834147989749908,
      "reward_std_all": 0.09726360067725182,
      "rewards-max/forensics_sinkhorn_reward": 0.9372245967388153,
      "rewards-max/forensics_strict_edge_reward": 0.8317028284072876,
      "rewards-max/forensics_tiou_reward": 0.8874431252479553,
      "rewards/forensics_sinkhorn_reward": 0.5513713657855988,
      "rewards/forensics_strict_edge_reward": 0.2940147817134857,
      "rewards/forensics_tiou_reward": 0.35301342606544495,
      "step": 1064
    },
    {
      "completion_length": 78.375,
      "epoch": 2.2282722513089004,
      "grad_norm": 21.900400161743164,
      "kl": 6.53125,
      "learning_rate": 4.423480083857442e-07,
      "loss": 0.3351,
      "reward": 0.696547120809555,
      "reward_std": 0.08681401237845421,
      "reward_std_all": 0.0976962111890316,
      "rewards-max/forensics_sinkhorn_reward": 0.9875160753726959,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9941997528076172,
      "rewards/forensics_sinkhorn_reward": 0.6078618764877319,
      "rewards/forensics_strict_edge_reward": 0.35471709072589874,
      "rewards/forensics_tiou_reward": 0.40739384293556213,
      "step": 1065
    },
    {
      "completion_length": 48.203125,
      "epoch": 2.230366492146597,
      "grad_norm": 17.283851623535156,
      "kl": 5.84375,
      "learning_rate": 4.418238993710692e-07,
      "loss": 0.3747,
      "reward": 0.6977900266647339,
      "reward_std": 0.1125875785946846,
      "reward_std_all": 0.11096558347344398,
      "rewards-max/forensics_sinkhorn_reward": 0.9865060746669769,
      "rewards-max/forensics_strict_edge_reward": 0.9385594427585602,
      "rewards-max/forensics_tiou_reward": 0.9902097880840302,
      "rewards/forensics_sinkhorn_reward": 0.6329500377178192,
      "rewards/forensics_strict_edge_reward": 0.3289604038000107,
      "rewards/forensics_tiou_reward": 0.4251208007335663,
      "step": 1066
    },
    {
      "completion_length": 46.703125,
      "epoch": 2.2324607329842934,
      "grad_norm": 22.30479621887207,
      "kl": 3.921875,
      "learning_rate": 4.412997903563941e-07,
      "loss": 0.2531,
      "reward": 0.680753618478775,
      "reward_std": 0.12620678916573524,
      "reward_std_all": 0.12507733702659607,
      "rewards-max/forensics_sinkhorn_reward": 0.9560587704181671,
      "rewards-max/forensics_strict_edge_reward": 0.882964164018631,
      "rewards-max/forensics_tiou_reward": 0.9279997944831848,
      "rewards/forensics_sinkhorn_reward": 0.5591786503791809,
      "rewards/forensics_strict_edge_reward": 0.2814596891403198,
      "rewards/forensics_tiou_reward": 0.3461020141839981,
      "step": 1067
    },
    {
      "completion_length": 44.984375,
      "epoch": 2.2345549738219894,
      "grad_norm": 24.15452766418457,
      "kl": 4.03125,
      "learning_rate": 4.4077568134171904e-07,
      "loss": 0.2167,
      "reward": 0.6442411243915558,
      "reward_std": 0.09969259798526764,
      "reward_std_all": 0.12431800365447998,
      "rewards-max/forensics_sinkhorn_reward": 0.9883543848991394,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9974940717220306,
      "rewards/forensics_sinkhorn_reward": 0.543773740530014,
      "rewards/forensics_strict_edge_reward": 0.2315339595079422,
      "rewards/forensics_tiou_reward": 0.27374184131622314,
      "step": 1068
    },
    {
      "completion_length": 44.578125,
      "epoch": 2.236649214659686,
      "grad_norm": 26.678688049316406,
      "kl": 4.125,
      "learning_rate": 4.40251572327044e-07,
      "loss": 0.2119,
      "reward": 0.5816152393817902,
      "reward_std": 0.18548761308193207,
      "reward_std_all": 0.19963456690311432,
      "rewards-max/forensics_sinkhorn_reward": 0.9069967269897461,
      "rewards-max/forensics_strict_edge_reward": 0.7442108243703842,
      "rewards-max/forensics_tiou_reward": 0.8206529319286346,
      "rewards/forensics_sinkhorn_reward": 0.48035310208797455,
      "rewards/forensics_strict_edge_reward": 0.21852251887321472,
      "rewards/forensics_tiou_reward": 0.25603001564741135,
      "step": 1069
    },
    {
      "completion_length": 72.390625,
      "epoch": 2.2387434554973824,
      "grad_norm": 32.81379318237305,
      "kl": 7.265625,
      "learning_rate": 4.397274633123689e-07,
      "loss": 0.423,
      "reward": 0.5820603668689728,
      "reward_std": 0.15526989847421646,
      "reward_std_all": 0.18035247921943665,
      "rewards-max/forensics_sinkhorn_reward": 0.9878110587596893,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9953519999980927,
      "rewards/forensics_sinkhorn_reward": 0.5355324745178223,
      "rewards/forensics_strict_edge_reward": 0.3483922630548477,
      "rewards/forensics_tiou_reward": 0.41800521314144135,
      "step": 1070
    },
    {
      "completion_length": 64.046875,
      "epoch": 2.2408376963350785,
      "grad_norm": 27.61440086364746,
      "kl": 6.171875,
      "learning_rate": 4.392033542976939e-07,
      "loss": 0.3088,
      "reward": 0.5791561901569366,
      "reward_std": 0.1671472080051899,
      "reward_std_all": 0.18942580372095108,
      "rewards-max/forensics_sinkhorn_reward": 0.8669003546237946,
      "rewards-max/forensics_strict_edge_reward": 0.7443208694458008,
      "rewards-max/forensics_tiou_reward": 0.7939290106296539,
      "rewards/forensics_sinkhorn_reward": 0.44041043519973755,
      "rewards/forensics_strict_edge_reward": 0.10423406586050987,
      "rewards/forensics_tiou_reward": 0.21790383756160736,
      "step": 1071
    },
    {
      "completion_length": 49.0,
      "epoch": 2.242931937172775,
      "grad_norm": 24.77703094482422,
      "kl": 5.625,
      "learning_rate": 4.386792452830189e-07,
      "loss": 0.3209,
      "reward": 0.6246744990348816,
      "reward_std": 0.18620823323726654,
      "reward_std_all": 0.1840531900525093,
      "rewards-max/forensics_sinkhorn_reward": 0.955891877412796,
      "rewards-max/forensics_strict_edge_reward": 0.7641629874706268,
      "rewards-max/forensics_tiou_reward": 0.9258496165275574,
      "rewards/forensics_sinkhorn_reward": 0.49336063861846924,
      "rewards/forensics_strict_edge_reward": 0.17041879519820213,
      "rewards/forensics_tiou_reward": 0.2517430931329727,
      "step": 1072
    },
    {
      "completion_length": 42.265625,
      "epoch": 2.245026178010471,
      "grad_norm": 20.828794479370117,
      "kl": 3.3125,
      "learning_rate": 4.381551362683438e-07,
      "loss": 0.2209,
      "reward": 0.6929055154323578,
      "reward_std": 0.12161632627248764,
      "reward_std_all": 0.11685995757579803,
      "rewards-max/forensics_sinkhorn_reward": 0.9826365411281586,
      "rewards-max/forensics_strict_edge_reward": 0.9364311099052429,
      "rewards-max/forensics_tiou_reward": 0.9846474528312683,
      "rewards/forensics_sinkhorn_reward": 0.5587261915206909,
      "rewards/forensics_strict_edge_reward": 0.24837034940719604,
      "rewards/forensics_tiou_reward": 0.35922659933567047,
      "step": 1073
    },
    {
      "completion_length": 72.671875,
      "epoch": 2.2471204188481675,
      "grad_norm": 23.58356285095215,
      "kl": 7.453125,
      "learning_rate": 4.3763102725366874e-07,
      "loss": 0.5579,
      "reward": 0.6124308705329895,
      "reward_std": 0.1374768428504467,
      "reward_std_all": 0.15929604694247246,
      "rewards-max/forensics_sinkhorn_reward": 0.9538429379463196,
      "rewards-max/forensics_strict_edge_reward": 0.8453949391841888,
      "rewards-max/forensics_tiou_reward": 0.9176575541496277,
      "rewards/forensics_sinkhorn_reward": 0.48286962509155273,
      "rewards/forensics_strict_edge_reward": 0.24956076592206955,
      "rewards/forensics_tiou_reward": 0.2824058532714844,
      "step": 1074
    },
    {
      "completion_length": 56.015625,
      "epoch": 2.249214659685864,
      "grad_norm": 28.199382781982422,
      "kl": 5.53125,
      "learning_rate": 4.371069182389937e-07,
      "loss": 0.3739,
      "reward": 0.6420235335826874,
      "reward_std": 0.15494807064533234,
      "reward_std_all": 0.1543118879199028,
      "rewards-max/forensics_sinkhorn_reward": 0.9878257513046265,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9954099953174591,
      "rewards/forensics_sinkhorn_reward": 0.5545873939990997,
      "rewards/forensics_strict_edge_reward": 0.33869296312332153,
      "rewards/forensics_tiou_reward": 0.415077805519104,
      "step": 1075
    },
    {
      "completion_length": 54.78125,
      "epoch": 2.25130890052356,
      "grad_norm": 25.871273040771484,
      "kl": 6.03125,
      "learning_rate": 4.365828092243186e-07,
      "loss": 0.4558,
      "reward": 0.6330041885375977,
      "reward_std": 0.13346990197896957,
      "reward_std_all": 0.14758796989917755,
      "rewards-max/forensics_sinkhorn_reward": 0.9864206612110138,
      "rewards-max/forensics_strict_edge_reward": 0.9357379674911499,
      "rewards-max/forensics_tiou_reward": 0.9898712038993835,
      "rewards/forensics_sinkhorn_reward": 0.5146424919366837,
      "rewards/forensics_strict_edge_reward": 0.2797946035861969,
      "rewards/forensics_tiou_reward": 0.36867041885852814,
      "step": 1076
    },
    {
      "completion_length": 38.46875,
      "epoch": 2.2534031413612565,
      "grad_norm": 16.577823638916016,
      "kl": 4.3203125,
      "learning_rate": 4.3605870020964356e-07,
      "loss": 0.4974,
      "reward": 0.647504448890686,
      "reward_std": 0.12489710748195648,
      "reward_std_all": 0.1449100449681282,
      "rewards-max/forensics_sinkhorn_reward": 0.9880331456661224,
      "rewards-max/forensics_strict_edge_reward": 0.9782643616199493,
      "rewards-max/forensics_tiou_reward": 0.9972189962863922,
      "rewards/forensics_sinkhorn_reward": 0.5732837468385696,
      "rewards/forensics_strict_edge_reward": 0.33500276505947113,
      "rewards/forensics_tiou_reward": 0.42310458421707153,
      "step": 1077
    },
    {
      "completion_length": 43.703125,
      "epoch": 2.255497382198953,
      "grad_norm": 15.736173629760742,
      "kl": 4.390625,
      "learning_rate": 4.355345911949685e-07,
      "loss": 0.35,
      "reward": 0.6775218546390533,
      "reward_std": 0.12149771489202976,
      "reward_std_all": 0.12729648500680923,
      "rewards-max/forensics_sinkhorn_reward": 0.9816546738147736,
      "rewards-max/forensics_strict_edge_reward": 0.8698340952396393,
      "rewards-max/forensics_tiou_reward": 0.9715320467948914,
      "rewards/forensics_sinkhorn_reward": 0.5438140481710434,
      "rewards/forensics_strict_edge_reward": 0.2985801622271538,
      "rewards/forensics_tiou_reward": 0.36231499910354614,
      "step": 1078
    },
    {
      "completion_length": 45.3125,
      "epoch": 2.257591623036649,
      "grad_norm": 17.477771759033203,
      "kl": 6.265625,
      "learning_rate": 4.350104821802935e-07,
      "loss": 0.4548,
      "reward": 0.6898863315582275,
      "reward_std": 0.12163442373275757,
      "reward_std_all": 0.12085495889186859,
      "rewards-max/forensics_sinkhorn_reward": 0.9874621629714966,
      "rewards-max/forensics_strict_edge_reward": 0.9464850425720215,
      "rewards-max/forensics_tiou_reward": 0.9939860999584198,
      "rewards/forensics_sinkhorn_reward": 0.5682075917720795,
      "rewards/forensics_strict_edge_reward": 0.3700565993785858,
      "rewards/forensics_tiou_reward": 0.3733699321746826,
      "step": 1079
    },
    {
      "completion_length": 60.15625,
      "epoch": 2.2596858638743456,
      "grad_norm": 29.259719848632812,
      "kl": 6.2265625,
      "learning_rate": 4.3448637316561844e-07,
      "loss": 0.3261,
      "reward": 0.6817278563976288,
      "reward_std": 0.1223946139216423,
      "reward_std_all": 0.12540754675865173,
      "rewards-max/forensics_sinkhorn_reward": 0.9872604310512543,
      "rewards-max/forensics_strict_edge_reward": 0.9256840348243713,
      "rewards-max/forensics_tiou_reward": 0.9931838512420654,
      "rewards/forensics_sinkhorn_reward": 0.560062825679779,
      "rewards/forensics_strict_edge_reward": 0.3110770881175995,
      "rewards/forensics_tiou_reward": 0.3718399256467819,
      "step": 1080
    },
    {
      "completion_length": 67.640625,
      "epoch": 2.261780104712042,
      "grad_norm": 31.32664680480957,
      "kl": 4.96875,
      "learning_rate": 4.339622641509434e-07,
      "loss": 0.2989,
      "reward": 0.6848176121711731,
      "reward_std": 0.15303314477205276,
      "reward_std_all": 0.1431138962507248,
      "rewards-max/forensics_sinkhorn_reward": 0.9547443091869354,
      "rewards-max/forensics_strict_edge_reward": 0.9054669737815857,
      "rewards-max/forensics_tiou_reward": 0.9261676669120789,
      "rewards/forensics_sinkhorn_reward": 0.5609169602394104,
      "rewards/forensics_strict_edge_reward": 0.3571542948484421,
      "rewards/forensics_tiou_reward": 0.390750527381897,
      "step": 1081
    },
    {
      "completion_length": 65.984375,
      "epoch": 2.263874345549738,
      "grad_norm": 24.457242965698242,
      "kl": 5.578125,
      "learning_rate": 4.334381551362683e-07,
      "loss": 0.4787,
      "reward": 0.6401488482952118,
      "reward_std": 0.15880709886550903,
      "reward_std_all": 0.1642427295446396,
      "rewards-max/forensics_sinkhorn_reward": 0.9876551330089569,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9947370290756226,
      "rewards/forensics_sinkhorn_reward": 0.5191541910171509,
      "rewards/forensics_strict_edge_reward": 0.3598470389842987,
      "rewards/forensics_tiou_reward": 0.37031176686286926,
      "step": 1082
    },
    {
      "completion_length": 60.28125,
      "epoch": 2.2659685863874346,
      "grad_norm": 31.18367576599121,
      "kl": 4.859375,
      "learning_rate": 4.3291404612159326e-07,
      "loss": 0.3164,
      "reward": 0.5977980494499207,
      "reward_std": 0.17866158485412598,
      "reward_std_all": 0.18495646119117737,
      "rewards-max/forensics_sinkhorn_reward": 0.9886631667613983,
      "rewards-max/forensics_strict_edge_reward": 0.9890114367008209,
      "rewards-max/forensics_tiou_reward": 0.9987096786499023,
      "rewards/forensics_sinkhorn_reward": 0.5032509714365005,
      "rewards/forensics_strict_edge_reward": 0.2325826808810234,
      "rewards/forensics_tiou_reward": 0.2955086678266525,
      "step": 1083
    },
    {
      "completion_length": 35.6875,
      "epoch": 2.268062827225131,
      "grad_norm": 20.935466766357422,
      "kl": 5.15625,
      "learning_rate": 4.323899371069182e-07,
      "loss": 0.2747,
      "reward": 0.6392318606376648,
      "reward_std": 0.149637870490551,
      "reward_std_all": 0.15677396953105927,
      "rewards-max/forensics_sinkhorn_reward": 0.9397857189178467,
      "rewards-max/forensics_strict_edge_reward": 0.8168056011199951,
      "rewards-max/forensics_tiou_reward": 0.897149384021759,
      "rewards/forensics_sinkhorn_reward": 0.527890682220459,
      "rewards/forensics_strict_edge_reward": 0.2926603853702545,
      "rewards/forensics_tiou_reward": 0.32616251707077026,
      "step": 1084
    },
    {
      "completion_length": 56.3125,
      "epoch": 2.270157068062827,
      "grad_norm": 20.408767700195312,
      "kl": 5.515625,
      "learning_rate": 4.3186582809224313e-07,
      "loss": 0.2466,
      "reward": 0.731092095375061,
      "reward_std": 0.11933739483356476,
      "reward_std_all": 0.10586622729897499,
      "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": 0.6163258850574493,
      "rewards/forensics_strict_edge_reward": 0.33746594190597534,
      "rewards/forensics_tiou_reward": 0.4405110478401184,
      "step": 1085
    },
    {
      "completion_length": 62.75,
      "epoch": 2.2722513089005236,
      "grad_norm": 23.40474510192871,
      "kl": 5.921875,
      "learning_rate": 4.313417190775681e-07,
      "loss": 0.4662,
      "reward": 0.5947987735271454,
      "reward_std": 0.14602789282798767,
      "reward_std_all": 0.18264348804950714,
      "rewards-max/forensics_sinkhorn_reward": 0.9880532026290894,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9963085651397705,
      "rewards/forensics_sinkhorn_reward": 0.5055934637784958,
      "rewards/forensics_strict_edge_reward": 0.3219156116247177,
      "rewards/forensics_tiou_reward": 0.35790589451789856,
      "step": 1086
    },
    {
      "completion_length": 69.703125,
      "epoch": 2.2743455497382197,
      "grad_norm": 35.908504486083984,
      "kl": 5.03125,
      "learning_rate": 4.308176100628931e-07,
      "loss": 0.3948,
      "reward": 0.6400189101696014,
      "reward_std": 0.1493905633687973,
      "reward_std_all": 0.16104546934366226,
      "rewards-max/forensics_sinkhorn_reward": 0.9874355792999268,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9938717484474182,
      "rewards/forensics_sinkhorn_reward": 0.5355018675327301,
      "rewards/forensics_strict_edge_reward": 0.3121238350868225,
      "rewards/forensics_tiou_reward": 0.3615082949399948,
      "step": 1087
    },
    {
      "completion_length": 62.171875,
      "epoch": 2.276439790575916,
      "grad_norm": 18.272462844848633,
      "kl": 5.3984375,
      "learning_rate": 4.30293501048218e-07,
      "loss": 0.4199,
      "reward": 0.6201441586017609,
      "reward_std": 0.17169547826051712,
      "reward_std_all": 0.17719395458698273,
      "rewards-max/forensics_sinkhorn_reward": 0.9866400361061096,
      "rewards-max/forensics_strict_edge_reward": 0.9364311099052429,
      "rewards-max/forensics_tiou_reward": 0.9907769858837128,
      "rewards/forensics_sinkhorn_reward": 0.5863593220710754,
      "rewards/forensics_strict_edge_reward": 0.329868346452713,
      "rewards/forensics_tiou_reward": 0.40991203486919403,
      "step": 1088
    },
    {
      "completion_length": 65.109375,
      "epoch": 2.2785340314136127,
      "grad_norm": 28.747058868408203,
      "kl": 5.796875,
      "learning_rate": 4.2976939203354296e-07,
      "loss": 0.5352,
      "reward": 0.6051739454269409,
      "reward_std": 0.21458441019058228,
      "reward_std_all": 0.2094293162226677,
      "rewards-max/forensics_sinkhorn_reward": 0.9883325397968292,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9974063634872437,
      "rewards/forensics_sinkhorn_reward": 0.4914080500602722,
      "rewards/forensics_strict_edge_reward": 0.25329499691724777,
      "rewards/forensics_tiou_reward": 0.33431369066238403,
      "step": 1089
    },
    {
      "completion_length": 37.71875,
      "epoch": 2.2806282722513087,
      "grad_norm": 18.850444793701172,
      "kl": 3.53125,
      "learning_rate": 4.292452830188679e-07,
      "loss": 0.1836,
      "reward": 0.6229122877120972,
      "reward_std": 0.1568341627717018,
      "reward_std_all": 0.16723797470331192,
      "rewards-max/forensics_sinkhorn_reward": 0.9634325206279755,
      "rewards-max/forensics_strict_edge_reward": 0.878633052110672,
      "rewards-max/forensics_tiou_reward": 0.9414501786231995,
      "rewards/forensics_sinkhorn_reward": 0.518833339214325,
      "rewards/forensics_strict_edge_reward": 0.20179720222949982,
      "rewards/forensics_tiou_reward": 0.3036584109067917,
      "step": 1090
    },
    {
      "completion_length": 53.796875,
      "epoch": 2.282722513089005,
      "grad_norm": 22.847105026245117,
      "kl": 6.921875,
      "learning_rate": 4.2872117400419283e-07,
      "loss": 0.4912,
      "reward": 0.6287870109081268,
      "reward_std": 0.18327432870864868,
      "reward_std_all": 0.18298973143100739,
      "rewards-max/forensics_sinkhorn_reward": 0.9872422516345978,
      "rewards-max/forensics_strict_edge_reward": 0.9565287232398987,
      "rewards-max/forensics_tiou_reward": 0.9931207299232483,
      "rewards/forensics_sinkhorn_reward": 0.5512930452823639,
      "rewards/forensics_strict_edge_reward": 0.2943962514400482,
      "rewards/forensics_tiou_reward": 0.36204421520233154,
      "step": 1091
    },
    {
      "completion_length": 50.453125,
      "epoch": 2.2848167539267017,
      "grad_norm": 28.05998420715332,
      "kl": 7.59375,
      "learning_rate": 4.281970649895178e-07,
      "loss": 0.5269,
      "reward": 0.6999174356460571,
      "reward_std": 0.11926589906215668,
      "reward_std_all": 0.11598043516278267,
      "rewards-max/forensics_sinkhorn_reward": 0.9864610731601715,
      "rewards-max/forensics_strict_edge_reward": 0.9565287232398987,
      "rewards-max/forensics_tiou_reward": 0.9900303781032562,
      "rewards/forensics_sinkhorn_reward": 0.5762961208820343,
      "rewards/forensics_strict_edge_reward": 0.3904359042644501,
      "rewards/forensics_tiou_reward": 0.38969385623931885,
      "step": 1092
    },
    {
      "completion_length": 39.375,
      "epoch": 2.2869109947643977,
      "grad_norm": 22.615005493164062,
      "kl": 6.09375,
      "learning_rate": 4.2767295597484274e-07,
      "loss": 0.3904,
      "reward": 0.6162544190883636,
      "reward_std": 0.20154199749231339,
      "reward_std_all": 0.19477900862693787,
      "rewards-max/forensics_sinkhorn_reward": 0.9877369999885559,
      "rewards-max/forensics_strict_edge_reward": 0.9256840348243713,
      "rewards-max/forensics_tiou_reward": 0.9950588047504425,
      "rewards/forensics_sinkhorn_reward": 0.5247534811496735,
      "rewards/forensics_strict_edge_reward": 0.31652598083019257,
      "rewards/forensics_tiou_reward": 0.32774700224399567,
      "step": 1093
    },
    {
      "completion_length": 43.265625,
      "epoch": 2.2890052356020942,
      "grad_norm": 18.480449676513672,
      "kl": 4.8359375,
      "learning_rate": 4.2714884696016765e-07,
      "loss": 0.2738,
      "reward": 0.7082317769527435,
      "reward_std": 0.0840340293943882,
      "reward_std_all": 0.08920419216156006,
      "rewards-max/forensics_sinkhorn_reward": 0.988288551568985,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9972359836101532,
      "rewards/forensics_sinkhorn_reward": 0.679681658744812,
      "rewards/forensics_strict_edge_reward": 0.4777224063873291,
      "rewards/forensics_tiou_reward": 0.5580035299062729,
      "step": 1094
    },
    {
      "completion_length": 46.109375,
      "epoch": 2.2910994764397907,
      "grad_norm": 19.762428283691406,
      "kl": 4.1484375,
      "learning_rate": 4.2662473794549266e-07,
      "loss": 0.3725,
      "reward": 0.702723890542984,
      "reward_std": 0.10291695408523083,
      "reward_std_all": 0.10586245357990265,
      "rewards-max/forensics_sinkhorn_reward": 0.9869979619979858,
      "rewards-max/forensics_strict_edge_reward": 0.9364311099052429,
      "rewards-max/forensics_tiou_reward": 0.9921655356884003,
      "rewards/forensics_sinkhorn_reward": 0.6488893032073975,
      "rewards/forensics_strict_edge_reward": 0.42336326837539673,
      "rewards/forensics_tiou_reward": 0.5163032710552216,
      "step": 1095
    },
    {
      "completion_length": 84.109375,
      "epoch": 2.2931937172774868,
      "grad_norm": 38.15386199951172,
      "kl": 12.875,
      "learning_rate": 4.261006289308176e-07,
      "loss": 0.7504,
      "reward": 0.5821148753166199,
      "reward_std": 0.1355305165052414,
      "reward_std_all": 0.1783621683716774,
      "rewards-max/forensics_sinkhorn_reward": 0.8105257153511047,
      "rewards-max/forensics_strict_edge_reward": 0.7416317760944366,
      "rewards-max/forensics_tiou_reward": 0.6501265168190002,
      "rewards/forensics_sinkhorn_reward": 0.4248322695493698,
      "rewards/forensics_strict_edge_reward": 0.21364206075668335,
      "rewards/forensics_tiou_reward": 0.19687510281801224,
      "step": 1096
    },
    {
      "completion_length": 49.765625,
      "epoch": 2.2952879581151833,
      "grad_norm": 31.50185775756836,
      "kl": 5.21875,
      "learning_rate": 4.255765199161425e-07,
      "loss": 0.3061,
      "reward": 0.6607999503612518,
      "reward_std": 0.10870014876127243,
      "reward_std_all": 0.1260659620165825,
      "rewards-max/forensics_sinkhorn_reward": 0.9888241589069366,
      "rewards-max/forensics_strict_edge_reward": 0.9890114367008209,
      "rewards-max/forensics_tiou_reward": 0.9993442595005035,
      "rewards/forensics_sinkhorn_reward": 0.5654025375843048,
      "rewards/forensics_strict_edge_reward": 0.2994462698698044,
      "rewards/forensics_tiou_reward": 0.3824390172958374,
      "step": 1097
    },
    {
      "completion_length": 35.515625,
      "epoch": 2.2973821989528798,
      "grad_norm": 21.968822479248047,
      "kl": 5.3125,
      "learning_rate": 4.250524109014675e-07,
      "loss": 0.3927,
      "reward": 0.6363988518714905,
      "reward_std": 0.11815806850790977,
      "reward_std_all": 0.14245445281267166,
      "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": 0.513397827744484,
      "rewards/forensics_strict_edge_reward": 0.31519271433353424,
      "rewards/forensics_tiou_reward": 0.34784577786922455,
      "step": 1098
    },
    {
      "completion_length": 54.359375,
      "epoch": 2.299476439790576,
      "grad_norm": 19.716739654541016,
      "kl": 7.34375,
      "learning_rate": 4.2452830188679244e-07,
      "loss": 0.4389,
      "reward": 0.6870060563087463,
      "reward_std": 0.0903395302593708,
      "reward_std_all": 0.10519666224718094,
      "rewards-max/forensics_sinkhorn_reward": 0.9636228978633881,
      "rewards-max/forensics_strict_edge_reward": 0.8876297473907471,
      "rewards-max/forensics_tiou_reward": 0.9410256147384644,
      "rewards/forensics_sinkhorn_reward": 0.5593705475330353,
      "rewards/forensics_strict_edge_reward": 0.32700713723897934,
      "rewards/forensics_tiou_reward": 0.3905893862247467,
      "step": 1099
    },
    {
      "completion_length": 51.96875,
      "epoch": 2.3015706806282723,
      "grad_norm": 20.948354721069336,
      "kl": 7.40625,
      "learning_rate": 4.2400419287211735e-07,
      "loss": 0.5754,
      "reward": 0.6269020438194275,
      "reward_std": 0.1653965227305889,
      "reward_std_all": 0.17507920414209366,
      "rewards-max/forensics_sinkhorn_reward": 0.984642893075943,
      "rewards-max/forensics_strict_edge_reward": 0.9169811904430389,
      "rewards-max/forensics_tiou_reward": 0.9828710556030273,
      "rewards/forensics_sinkhorn_reward": 0.5377974212169647,
      "rewards/forensics_strict_edge_reward": 0.292703315615654,
      "rewards/forensics_tiou_reward": 0.3925726115703583,
      "step": 1100
    },
    {
      "completion_length": 81.484375,
      "epoch": 2.303664921465969,
      "grad_norm": 40.06446075439453,
      "kl": 8.953125,
      "learning_rate": 4.234800838574423e-07,
      "loss": 0.6156,
      "reward": 0.560155063867569,
      "reward_std": 0.19160175323486328,
      "reward_std_all": 0.21128610521554947,
      "rewards-max/forensics_sinkhorn_reward": 0.9883732497692108,
      "rewards-max/forensics_strict_edge_reward": 0.9782643616199493,
      "rewards-max/forensics_tiou_reward": 0.9975698590278625,
      "rewards/forensics_sinkhorn_reward": 0.4665004312992096,
      "rewards/forensics_strict_edge_reward": 0.288300596177578,
      "rewards/forensics_tiou_reward": 0.2941928803920746,
      "step": 1101
    },
    {
      "completion_length": 56.328125,
      "epoch": 2.305759162303665,
      "grad_norm": 22.824644088745117,
      "kl": 6.875,
      "learning_rate": 4.229559748427673e-07,
      "loss": 0.4708,
      "reward": 0.6643255352973938,
      "reward_std": 0.10565832257270813,
      "reward_std_all": 0.12241328880190849,
      "rewards-max/forensics_sinkhorn_reward": 0.9513340890407562,
      "rewards-max/forensics_strict_edge_reward": 0.8719756007194519,
      "rewards-max/forensics_tiou_reward": 0.9161786139011383,
      "rewards/forensics_sinkhorn_reward": 0.5420990586280823,
      "rewards/forensics_strict_edge_reward": 0.25198739767074585,
      "rewards/forensics_tiou_reward": 0.3140212297439575,
      "step": 1102
    },
    {
      "completion_length": 65.4375,
      "epoch": 2.3078534031413613,
      "grad_norm": 26.730016708374023,
      "kl": 5.7734375,
      "learning_rate": 4.224318658280922e-07,
      "loss": 0.314,
      "reward": 0.5649189949035645,
      "reward_std": 0.14218974858522415,
      "reward_std_all": 0.18509140610694885,
      "rewards-max/forensics_sinkhorn_reward": 0.9473811388015747,
      "rewards-max/forensics_strict_edge_reward": 0.8805383145809174,
      "rewards-max/forensics_tiou_reward": 0.9142818748950958,
      "rewards/forensics_sinkhorn_reward": 0.49240386486053467,
      "rewards/forensics_strict_edge_reward": 0.29102832078933716,
      "rewards/forensics_tiou_reward": 0.3242511972784996,
      "step": 1103
    },
    {
      "completion_length": 78.78125,
      "epoch": 2.309947643979058,
      "grad_norm": 31.02947235107422,
      "kl": 7.375,
      "learning_rate": 4.219077568134172e-07,
      "loss": 0.4471,
      "reward": 0.6548890769481659,
      "reward_std": 0.1544998660683632,
      "reward_std_all": 0.14973726123571396,
      "rewards-max/forensics_sinkhorn_reward": 0.9630879461765289,
      "rewards-max/forensics_strict_edge_reward": 0.9421285092830658,
      "rewards-max/forensics_tiou_reward": 0.9423101246356964,
      "rewards/forensics_sinkhorn_reward": 0.5391974449157715,
      "rewards/forensics_strict_edge_reward": 0.3198231905698776,
      "rewards/forensics_tiou_reward": 0.4125156104564667,
      "step": 1104
    },
    {
      "completion_length": 64.890625,
      "epoch": 2.312041884816754,
      "grad_norm": 20.88872718811035,
      "kl": 6.5,
      "learning_rate": 4.2138364779874214e-07,
      "loss": 0.409,
      "reward": 0.6633198261260986,
      "reward_std": 0.10442202910780907,
      "reward_std_all": 0.12419084087014198,
      "rewards-max/forensics_sinkhorn_reward": 0.9877058863639832,
      "rewards-max/forensics_strict_edge_reward": 0.9574736058712006,
      "rewards-max/forensics_tiou_reward": 0.9949367046356201,
      "rewards/forensics_sinkhorn_reward": 0.49505719542503357,
      "rewards/forensics_strict_edge_reward": 0.29930828511714935,
      "rewards/forensics_tiou_reward": 0.3537103980779648,
      "step": 1105
    },
    {
      "completion_length": 35.515625,
      "epoch": 2.3141361256544504,
      "grad_norm": 29.633995056152344,
      "kl": 4.5234375,
      "learning_rate": 4.2085953878406705e-07,
      "loss": 0.2725,
      "reward": 0.6975095272064209,
      "reward_std": 0.0993833877146244,
      "reward_std_all": 0.10404720902442932,
      "rewards-max/forensics_sinkhorn_reward": 0.9871379435062408,
      "rewards-max/forensics_strict_edge_reward": 0.946017861366272,
      "rewards-max/forensics_tiou_reward": 0.9926978945732117,
      "rewards/forensics_sinkhorn_reward": 0.656327098608017,
      "rewards/forensics_strict_edge_reward": 0.3875593990087509,
      "rewards/forensics_tiou_reward": 0.46822357177734375,
      "step": 1106
    },
    {
      "completion_length": 76.0,
      "epoch": 2.3162303664921464,
      "grad_norm": 23.995891571044922,
      "kl": 6.84375,
      "learning_rate": 4.20335429769392e-07,
      "loss": 0.467,
      "reward": 0.6433842480182648,
      "reward_std": 0.1418396383523941,
      "reward_std_all": 0.14644183218479156,
      "rewards-max/forensics_sinkhorn_reward": 0.9879208505153656,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.995790421962738,
      "rewards/forensics_sinkhorn_reward": 0.5791513025760651,
      "rewards/forensics_strict_edge_reward": 0.28384020924568176,
      "rewards/forensics_tiou_reward": 0.34161530435085297,
      "step": 1107
    },
    {
      "completion_length": 74.859375,
      "epoch": 2.318324607329843,
      "grad_norm": 23.608951568603516,
      "kl": 6.875,
      "learning_rate": 4.1981132075471697e-07,
      "loss": 0.3538,
      "reward": 0.6459634900093079,
      "reward_std": 0.13982806354761124,
      "reward_std_all": 0.1461925059556961,
      "rewards-max/forensics_sinkhorn_reward": 0.9869734048843384,
      "rewards-max/forensics_strict_edge_reward": 0.9357379674911499,
      "rewards-max/forensics_tiou_reward": 0.9920763075351715,
      "rewards/forensics_sinkhorn_reward": 0.5095958113670349,
      "rewards/forensics_strict_edge_reward": 0.2876841425895691,
      "rewards/forensics_tiou_reward": 0.34413252770900726,
      "step": 1108
    },
    {
      "completion_length": 37.546875,
      "epoch": 2.3204188481675394,
      "grad_norm": 51.921775817871094,
      "kl": 5.15625,
      "learning_rate": 4.1928721174004187e-07,
      "loss": 0.3368,
      "reward": 0.6291978359222412,
      "reward_std": 0.1124216727912426,
      "reward_std_all": 0.1408684104681015,
      "rewards-max/forensics_sinkhorn_reward": 0.9874860644340515,
      "rewards-max/forensics_strict_edge_reward": 0.9252271056175232,
      "rewards-max/forensics_tiou_reward": 0.9940716326236725,
      "rewards/forensics_sinkhorn_reward": 0.5128206014633179,
      "rewards/forensics_strict_edge_reward": 0.2763223946094513,
      "rewards/forensics_tiou_reward": 0.30843448638916016,
      "step": 1109
    },
    {
      "completion_length": 58.296875,
      "epoch": 2.3225130890052355,
      "grad_norm": 21.236631393432617,
      "kl": 6.4765625,
      "learning_rate": 4.1876310272536683e-07,
      "loss": 0.4253,
      "reward": 0.6080335080623627,
      "reward_std": 0.18332047760486603,
      "reward_std_all": 0.19114384055137634,
      "rewards-max/forensics_sinkhorn_reward": 0.8872709274291992,
      "rewards-max/forensics_strict_edge_reward": 0.6636994034051895,
      "rewards-max/forensics_tiou_reward": 0.7796536684036255,
      "rewards/forensics_sinkhorn_reward": 0.46906450390815735,
      "rewards/forensics_strict_edge_reward": 0.2748064026236534,
      "rewards/forensics_tiou_reward": 0.30195795744657516,
      "step": 1110
    },
    {
      "completion_length": 46.890625,
      "epoch": 2.324607329842932,
      "grad_norm": 26.111963272094727,
      "kl": 4.3203125,
      "learning_rate": 4.1823899371069184e-07,
      "loss": 0.3675,
      "reward": 0.7113629281520844,
      "reward_std": 0.11802345141768456,
      "reward_std_all": 0.11242728680372238,
      "rewards-max/forensics_sinkhorn_reward": 0.9885846078395844,
      "rewards-max/forensics_strict_edge_reward": 0.9890114367008209,
      "rewards-max/forensics_tiou_reward": 0.9984000027179718,
      "rewards/forensics_sinkhorn_reward": 0.588567703962326,
      "rewards/forensics_strict_edge_reward": 0.35341353714466095,
      "rewards/forensics_tiou_reward": 0.4009888172149658,
      "step": 1111
    },
    {
      "completion_length": 56.953125,
      "epoch": 2.3267015706806284,
      "grad_norm": 21.83535385131836,
      "kl": 3.8046875,
      "learning_rate": 4.1771488469601675e-07,
      "loss": 0.318,
      "reward": 0.6851881146430969,
      "reward_std": 0.13620057702064514,
      "reward_std_all": 0.13182317838072777,
      "rewards-max/forensics_sinkhorn_reward": 0.9878386855125427,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9954647421836853,
      "rewards/forensics_sinkhorn_reward": 0.6416785717010498,
      "rewards/forensics_strict_edge_reward": 0.36573272943496704,
      "rewards/forensics_tiou_reward": 0.48147815465927124,
      "step": 1112
    },
    {
      "completion_length": 48.640625,
      "epoch": 2.3287958115183245,
      "grad_norm": 22.814611434936523,
      "kl": 4.1875,
      "learning_rate": 4.171907756813417e-07,
      "loss": 0.2594,
      "reward": 0.6298871636390686,
      "reward_std": 0.13482753932476044,
      "reward_std_all": 0.15204331278800964,
      "rewards-max/forensics_sinkhorn_reward": 0.9714987874031067,
      "rewards-max/forensics_strict_edge_reward": 0.8850063383579254,
      "rewards-max/forensics_tiou_reward": 0.9566957950592041,
      "rewards/forensics_sinkhorn_reward": 0.5077159106731415,
      "rewards/forensics_strict_edge_reward": 0.22408227622509003,
      "rewards/forensics_tiou_reward": 0.2827332317829132,
      "step": 1113
    },
    {
      "completion_length": 74.65625,
      "epoch": 2.330890052356021,
      "grad_norm": 19.06960105895996,
      "kl": 7.140625,
      "learning_rate": 4.1666666666666667e-07,
      "loss": 0.4819,
      "reward": 0.662622481584549,
      "reward_std": 0.2009594589471817,
      "reward_std_all": 0.18131759762763977,
      "rewards-max/forensics_sinkhorn_reward": 0.9885820150375366,
      "rewards-max/forensics_strict_edge_reward": 0.9782643616199493,
      "rewards-max/forensics_tiou_reward": 0.9983910024166107,
      "rewards/forensics_sinkhorn_reward": 0.5938245356082916,
      "rewards/forensics_strict_edge_reward": 0.36498188972473145,
      "rewards/forensics_tiou_reward": 0.4271140843629837,
      "step": 1114
    },
    {
      "completion_length": 55.90625,
      "epoch": 2.332984293193717,
      "grad_norm": 20.78588104248047,
      "kl": 6.09375,
      "learning_rate": 4.1614255765199157e-07,
      "loss": 0.4677,
      "reward": 0.6554870009422302,
      "reward_std": 0.18693051114678383,
      "reward_std_all": 0.1758815422654152,
      "rewards-max/forensics_sinkhorn_reward": 0.972029834985733,
      "rewards-max/forensics_strict_edge_reward": 0.877170205116272,
      "rewards-max/forensics_tiou_reward": 0.9614129960536957,
      "rewards/forensics_sinkhorn_reward": 0.5245424807071686,
      "rewards/forensics_strict_edge_reward": 0.2501723915338516,
      "rewards/forensics_tiou_reward": 0.33579157292842865,
      "step": 1115
    },
    {
      "completion_length": 57.1875,
      "epoch": 2.3350785340314135,
      "grad_norm": 18.421478271484375,
      "kl": 4.625,
      "learning_rate": 4.1561844863731653e-07,
      "loss": 0.3874,
      "reward": 0.6043416857719421,
      "reward_std": 0.1647159308195114,
      "reward_std_all": 0.1837635189294815,
      "rewards-max/forensics_sinkhorn_reward": 0.9874432384967804,
      "rewards-max/forensics_strict_edge_reward": 0.9565287232398987,
      "rewards-max/forensics_tiou_reward": 0.9939104616641998,
      "rewards/forensics_sinkhorn_reward": 0.5060158222913742,
      "rewards/forensics_strict_edge_reward": 0.2683994323015213,
      "rewards/forensics_tiou_reward": 0.3018934428691864,
      "step": 1116
    },
    {
      "completion_length": 74.328125,
      "epoch": 2.33717277486911,
      "grad_norm": 22.30417823791504,
      "kl": 6.671875,
      "learning_rate": 4.150943396226415e-07,
      "loss": 0.4341,
      "reward": 0.6958515048027039,
      "reward_std": 0.1515977755188942,
      "reward_std_all": 0.13799704611301422,
      "rewards-max/forensics_sinkhorn_reward": 0.9889025986194611,
      "rewards-max/forensics_strict_edge_reward": 0.9890114367008209,
      "rewards-max/forensics_tiou_reward": 0.999653548002243,
      "rewards/forensics_sinkhorn_reward": 0.5170373469591141,
      "rewards/forensics_strict_edge_reward": 0.33334946632385254,
      "rewards/forensics_tiou_reward": 0.3857884407043457,
      "step": 1117
    },
    {
      "completion_length": 79.71875,
      "epoch": 2.339267015706806,
      "grad_norm": 28.816177368164062,
      "kl": 9.09375,
      "learning_rate": 4.145702306079664e-07,
      "loss": 0.5563,
      "reward": 0.6883254945278168,
      "reward_std": 0.13201484829187393,
      "reward_std_all": 0.1263762228190899,
      "rewards-max/forensics_sinkhorn_reward": 0.9850743412971497,
      "rewards-max/forensics_strict_edge_reward": 0.9464850425720215,
      "rewards-max/forensics_tiou_reward": 0.9846859872341156,
      "rewards/forensics_sinkhorn_reward": 0.5691097676753998,
      "rewards/forensics_strict_edge_reward": 0.3220219016075134,
      "rewards/forensics_tiou_reward": 0.3974042534828186,
      "step": 1118
    },
    {
      "completion_length": 58.71875,
      "epoch": 2.3413612565445026,
      "grad_norm": 35.01732635498047,
      "kl": 8.046875,
      "learning_rate": 4.140461215932914e-07,
      "loss": 0.5236,
      "reward": 0.6040820181369781,
      "reward_std": 0.20730968564748764,
      "reward_std_all": 0.20330295711755753,
      "rewards-max/forensics_sinkhorn_reward": 0.9885612726211548,
      "rewards-max/forensics_strict_edge_reward": 0.9782643616199493,
      "rewards-max/forensics_tiou_reward": 0.9983093738555908,
      "rewards/forensics_sinkhorn_reward": 0.4634346663951874,
      "rewards/forensics_strict_edge_reward": 0.25266971439123154,
      "rewards/forensics_tiou_reward": 0.2735547721385956,
      "step": 1119
    },
    {
      "completion_length": 25.25,
      "epoch": 2.343455497382199,
      "grad_norm": 18.268524169921875,
      "kl": 3.3828125,
      "learning_rate": 4.1352201257861636e-07,
      "loss": 0.2001,
      "reward": 0.7275251150131226,
      "reward_std": 0.08031529188156128,
      "reward_std_all": 0.08358766511082649,
      "rewards-max/forensics_sinkhorn_reward": 0.988766998052597,
      "rewards-max/forensics_strict_edge_reward": 0.9890114367008209,
      "rewards-max/forensics_tiou_reward": 0.9991189539432526,
      "rewards/forensics_sinkhorn_reward": 0.6761002540588379,
      "rewards/forensics_strict_edge_reward": 0.43715178966522217,
      "rewards/forensics_tiou_reward": 0.5054436922073364,
      "step": 1120
    },
    {
      "completion_length": 54.6875,
      "epoch": 2.345549738219895,
      "grad_norm": 39.679283142089844,
      "kl": 3.921875,
      "learning_rate": 4.1299790356394127e-07,
      "loss": 0.4357,
      "reward": 0.6251360774040222,
      "reward_std": 0.17637403309345245,
      "reward_std_all": 0.17523808777332306,
      "rewards-max/forensics_sinkhorn_reward": 0.9884248673915863,
      "rewards-max/forensics_strict_edge_reward": 0.9782643616199493,
      "rewards-max/forensics_tiou_reward": 0.9977728426456451,
      "rewards/forensics_sinkhorn_reward": 0.561337023973465,
      "rewards/forensics_strict_edge_reward": 0.27059246599674225,
      "rewards/forensics_tiou_reward": 0.3529722988605499,
      "step": 1121
    },
    {
      "completion_length": 51.703125,
      "epoch": 2.3476439790575916,
      "grad_norm": 28.1778564453125,
      "kl": 4.984375,
      "learning_rate": 4.1247379454926623e-07,
      "loss": 0.35,
      "reward": 0.6357801556587219,
      "reward_std": 0.19291012734174728,
      "reward_std_all": 0.18422894179821014,
      "rewards-max/forensics_sinkhorn_reward": 0.9887832403182983,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9991829693317413,
      "rewards/forensics_sinkhorn_reward": 0.5734236240386963,
      "rewards/forensics_strict_edge_reward": 0.31954115629196167,
      "rewards/forensics_tiou_reward": 0.3903222233057022,
      "step": 1122
    },
    {
      "completion_length": 50.046875,
      "epoch": 2.349738219895288,
      "grad_norm": 22.454490661621094,
      "kl": 5.421875,
      "learning_rate": 4.119496855345912e-07,
      "loss": 0.5014,
      "reward": 0.6307432651519775,
      "reward_std": 0.18087328970432281,
      "reward_std_all": 0.18203472346067429,
      "rewards-max/forensics_sinkhorn_reward": 0.9887523055076599,
      "rewards-max/forensics_strict_edge_reward": 0.9677534997463226,
      "rewards-max/forensics_tiou_reward": 0.9990610182285309,
      "rewards/forensics_sinkhorn_reward": 0.541606217622757,
      "rewards/forensics_strict_edge_reward": 0.3331691175699234,
      "rewards/forensics_tiou_reward": 0.394428089261055,
      "step": 1123
    },
    {
      "completion_length": 60.515625,
      "epoch": 2.351832460732984,
      "grad_norm": 36.80162048339844,
      "kl": 5.71875,
      "learning_rate": 4.114255765199161e-07,
      "loss": 0.4611,
      "reward": 0.6053643524646759,
      "reward_std": 0.19836686551570892,
      "reward_std_all": 0.19959227740764618,
      "rewards-max/forensics_sinkhorn_reward": 0.9873267114162445,
      "rewards-max/forensics_strict_edge_reward": 0.9185642302036285,
      "rewards-max/forensics_tiou_reward": 0.9934426248073578,
      "rewards/forensics_sinkhorn_reward": 0.5161315053701401,
      "rewards/forensics_strict_edge_reward": 0.31686753034591675,
      "rewards/forensics_tiou_reward": 0.35386331379413605,
      "step": 1124
    },
    {
      "completion_length": 70.8125,
      "epoch": 2.3539267015706806,
      "grad_norm": 24.927452087402344,
      "kl": 7.078125,
      "learning_rate": 4.1090146750524105e-07,
      "loss": 0.4869,
      "reward": 0.6516015529632568,
      "reward_std": 0.10406148061156273,
      "reward_std_all": 0.12386801093816757,
      "rewards-max/forensics_sinkhorn_reward": 0.9440446197986603,
      "rewards-max/forensics_strict_edge_reward": 0.9015313386917114,
      "rewards-max/forensics_tiou_reward": 0.9049264192581177,
      "rewards/forensics_sinkhorn_reward": 0.49399349093437195,
      "rewards/forensics_strict_edge_reward": 0.24119603633880615,
      "rewards/forensics_tiou_reward": 0.28541938960552216,
      "step": 1125
    },
    {
      "completion_length": 37.78125,
      "epoch": 2.356020942408377,
      "grad_norm": 17.414974212646484,
      "kl": 4.1953125,
      "learning_rate": 4.1037735849056606e-07,
      "loss": 0.3714,
      "reward": 0.67502161860466,
      "reward_std": 0.12547161430120468,
      "reward_std_all": 0.12502488121390343,
      "rewards-max/forensics_sinkhorn_reward": 0.9880712926387787,
      "rewards-max/forensics_strict_edge_reward": 0.9782643616199493,
      "rewards-max/forensics_tiou_reward": 0.9963833689689636,
      "rewards/forensics_sinkhorn_reward": 0.5682013034820557,
      "rewards/forensics_strict_edge_reward": 0.31629765033721924,
      "rewards/forensics_tiou_reward": 0.3845781832933426,
      "step": 1126
    },
    {
      "completion_length": 85.609375,
      "epoch": 2.358115183246073,
      "grad_norm": 21.36552619934082,
      "kl": 8.71875,
      "learning_rate": 4.0985324947589097e-07,
      "loss": 0.6166,
      "reward": 0.625391811132431,
      "reward_std": 0.16396629065275192,
      "reward_std_all": 0.1767466813325882,
      "rewards-max/forensics_sinkhorn_reward": 0.9852806925773621,
      "rewards-max/forensics_strict_edge_reward": 0.9062341153621674,
      "rewards-max/forensics_tiou_reward": 0.9854688048362732,
      "rewards/forensics_sinkhorn_reward": 0.5557727515697479,
      "rewards/forensics_strict_edge_reward": 0.28727221488952637,
      "rewards/forensics_tiou_reward": 0.34256476908922195,
      "step": 1127
    },
    {
      "completion_length": 87.765625,
      "epoch": 2.3602094240837697,
      "grad_norm": 18.80129051208496,
      "kl": 9.4375,
      "learning_rate": 4.0932914046121593e-07,
      "loss": 0.4616,
      "reward": 0.6020644009113312,
      "reward_std": 0.21927470713853836,
      "reward_std_all": 0.21987715363502502,
      "rewards-max/forensics_sinkhorn_reward": 0.9683145880699158,
      "rewards-max/forensics_strict_edge_reward": 0.8751733303070068,
      "rewards-max/forensics_tiou_reward": 0.9516995549201965,
      "rewards/forensics_sinkhorn_reward": 0.4874742329120636,
      "rewards/forensics_strict_edge_reward": 0.23485352843999863,
      "rewards/forensics_tiou_reward": 0.3054545670747757,
      "step": 1128
    },
    {
      "completion_length": 98.8125,
      "epoch": 2.362303664921466,
      "grad_norm": 35.28549575805664,
      "kl": 10.859375,
      "learning_rate": 4.088050314465409e-07,
      "loss": 0.5152,
      "reward": 0.5995491147041321,
      "reward_std": 0.1630365587770939,
      "reward_std_all": 0.1778075248003006,
      "rewards-max/forensics_sinkhorn_reward": 0.9733351469039917,
      "rewards-max/forensics_strict_edge_reward": 0.7610765993595123,
      "rewards-max/forensics_tiou_reward": 0.9618556499481201,
      "rewards/forensics_sinkhorn_reward": 0.4786229580640793,
      "rewards/forensics_strict_edge_reward": 0.18792053684592247,
      "rewards/forensics_tiou_reward": 0.2408778890967369,
      "step": 1129
    },
    {
      "completion_length": 65.71875,
      "epoch": 2.364397905759162,
      "grad_norm": 25.65913200378418,
      "kl": 7.0625,
      "learning_rate": 4.082809224318658e-07,
      "loss": 0.459,
      "reward": 0.64092618227005,
      "reward_std": 0.14107724279165268,
      "reward_std_all": 0.14931099116802216,
      "rewards-max/forensics_sinkhorn_reward": 0.9498848021030426,
      "rewards-max/forensics_strict_edge_reward": 0.8681221902370453,
      "rewards-max/forensics_tiou_reward": 0.915711909532547,
      "rewards/forensics_sinkhorn_reward": 0.5212867259979248,
      "rewards/forensics_strict_edge_reward": 0.2631956934928894,
      "rewards/forensics_tiou_reward": 0.27259624004364014,
      "step": 1130
    },
    {
      "completion_length": 62.546875,
      "epoch": 2.3664921465968587,
      "grad_norm": 39.209712982177734,
      "kl": 4.5703125,
      "learning_rate": 4.0775681341719075e-07,
      "loss": 0.4114,
      "reward": 0.6515818536281586,
      "reward_std": 0.12127113342285156,
      "reward_std_all": 0.13546396791934967,
      "rewards-max/forensics_sinkhorn_reward": 0.9876144230365753,
      "rewards-max/forensics_strict_edge_reward": 0.9565287232398987,
      "rewards-max/forensics_tiou_reward": 0.9945784211158752,
      "rewards/forensics_sinkhorn_reward": 0.5795254111289978,
      "rewards/forensics_strict_edge_reward": 0.27864695340394974,
      "rewards/forensics_tiou_reward": 0.35338982939720154,
      "step": 1131
    },
    {
      "completion_length": 76.03125,
      "epoch": 2.368586387434555,
      "grad_norm": 23.23716926574707,
      "kl": 7.0,
      "learning_rate": 4.072327044025157e-07,
      "loss": 0.5452,
      "reward": 0.5722172856330872,
      "reward_std": 0.19330624490976334,
      "reward_std_all": 0.21259280294179916,
      "rewards-max/forensics_sinkhorn_reward": 0.9835458695888519,
      "rewards-max/forensics_strict_edge_reward": 0.8876297473907471,
      "rewards-max/forensics_tiou_reward": 0.9785555303096771,
      "rewards/forensics_sinkhorn_reward": 0.490206316113472,
      "rewards/forensics_strict_edge_reward": 0.244877427816391,
      "rewards/forensics_tiou_reward": 0.2888805791735649,
      "step": 1132
    },
    {
      "completion_length": 88.40625,
      "epoch": 2.3706806282722512,
      "grad_norm": 26.184764862060547,
      "kl": 7.71875,
      "learning_rate": 4.067085953878406e-07,
      "loss": 0.494,
      "reward": 0.6106933951377869,
      "reward_std": 0.13830403983592987,
      "reward_std_all": 0.16030173003673553,
      "rewards-max/forensics_sinkhorn_reward": 0.9876458644866943,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9946997165679932,
      "rewards/forensics_sinkhorn_reward": 0.5285041034221649,
      "rewards/forensics_strict_edge_reward": 0.3070373088121414,
      "rewards/forensics_tiou_reward": 0.3318954110145569,
      "step": 1133
    },
    {
      "completion_length": 73.96875,
      "epoch": 2.3727748691099477,
      "grad_norm": 19.730518341064453,
      "kl": 7.703125,
      "learning_rate": 4.0618448637316563e-07,
      "loss": 0.4307,
      "reward": 0.6290268003940582,
      "reward_std": 0.16885946691036224,
      "reward_std_all": 0.1671750769019127,
      "rewards-max/forensics_sinkhorn_reward": 0.9872674345970154,
      "rewards-max/forensics_strict_edge_reward": 0.9567649364471436,
      "rewards-max/forensics_tiou_reward": 0.993230551481247,
      "rewards/forensics_sinkhorn_reward": 0.5824204087257385,
      "rewards/forensics_strict_edge_reward": 0.34860342741012573,
      "rewards/forensics_tiou_reward": 0.43229468166828156,
      "step": 1134
    },
    {
      "completion_length": 87.015625,
      "epoch": 2.374869109947644,
      "grad_norm": 20.683761596679688,
      "kl": 5.875,
      "learning_rate": 4.056603773584906e-07,
      "loss": 0.4043,
      "reward": 0.6717678606510162,
      "reward_std": 0.11052298545837402,
      "reward_std_all": 0.12444893270730972,
      "rewards-max/forensics_sinkhorn_reward": 0.9692928493022919,
      "rewards-max/forensics_strict_edge_reward": 0.8745477199554443,
      "rewards-max/forensics_tiou_reward": 0.9574134349822998,
      "rewards/forensics_sinkhorn_reward": 0.5806102901697159,
      "rewards/forensics_strict_edge_reward": 0.3547995537519455,
      "rewards/forensics_tiou_reward": 0.4221123158931732,
      "step": 1135
    },
    {
      "completion_length": 70.984375,
      "epoch": 2.3769633507853403,
      "grad_norm": 22.946306228637695,
      "kl": 5.421875,
      "learning_rate": 4.051362683438155e-07,
      "loss": 0.4023,
      "reward": 0.5971646904945374,
      "reward_std": 0.19356942176818848,
      "reward_std_all": 0.1992686316370964,
      "rewards-max/forensics_sinkhorn_reward": 0.9880594611167908,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9963334500789642,
      "rewards/forensics_sinkhorn_reward": 0.47832298278808594,
      "rewards/forensics_strict_edge_reward": 0.24948854744434357,
      "rewards/forensics_tiou_reward": 0.2944512963294983,
      "step": 1136
    },
    {
      "completion_length": 118.390625,
      "epoch": 2.3790575916230368,
      "grad_norm": 22.786972045898438,
      "kl": 7.90625,
      "learning_rate": 4.0461215932914045e-07,
      "loss": 0.4444,
      "reward": 0.6324759721755981,
      "reward_std": 0.22073815017938614,
      "reward_std_all": 0.20334212481975555,
      "rewards-max/forensics_sinkhorn_reward": 0.9635829925537109,
      "rewards-max/forensics_strict_edge_reward": 0.8805811703205109,
      "rewards-max/forensics_tiou_reward": 0.9433399140834808,
      "rewards/forensics_sinkhorn_reward": 0.5249076336622238,
      "rewards/forensics_strict_edge_reward": 0.27376779168844223,
      "rewards/forensics_tiou_reward": 0.3495335429906845,
      "step": 1137
    },
    {
      "completion_length": 83.9375,
      "epoch": 2.381151832460733,
      "grad_norm": 17.641990661621094,
      "kl": 7.34375,
      "learning_rate": 4.040880503144654e-07,
      "loss": 0.4977,
      "reward": 0.5650578737258911,
      "reward_std": 0.21140826493501663,
      "reward_std_all": 0.22018776834011078,
      "rewards-max/forensics_sinkhorn_reward": 0.9887801706790924,
      "rewards-max/forensics_strict_edge_reward": 0.9890114367008209,
      "rewards-max/forensics_tiou_reward": 0.9991711676120758,
      "rewards/forensics_sinkhorn_reward": 0.493291437625885,
      "rewards/forensics_strict_edge_reward": 0.29254236817359924,
      "rewards/forensics_tiou_reward": 0.33306068181991577,
      "step": 1138
    },
    {
      "completion_length": 117.015625,
      "epoch": 2.3832460732984293,
      "grad_norm": 25.48151397705078,
      "kl": 6.578125,
      "learning_rate": 4.035639412997903e-07,
      "loss": 0.3207,
      "reward": 0.5747748613357544,
      "reward_std": 0.24825980514287949,
      "reward_std_all": 0.23684999346733093,
      "rewards-max/forensics_sinkhorn_reward": 0.8526791036128998,
      "rewards-max/forensics_strict_edge_reward": 0.47016218304634094,
      "rewards-max/forensics_tiou_reward": 0.7101674675941467,
      "rewards/forensics_sinkhorn_reward": 0.39008285105228424,
      "rewards/forensics_strict_edge_reward": 0.15544898808002472,
      "rewards/forensics_tiou_reward": 0.2074359580874443,
      "step": 1139
    },
    {
      "completion_length": 118.09375,
      "epoch": 2.385340314136126,
      "grad_norm": 22.007511138916016,
      "kl": 6.640625,
      "learning_rate": 4.030398322851153e-07,
      "loss": 0.3713,
      "reward": 0.58156518638134,
      "reward_std": 0.14870623499155045,
      "reward_std_all": 0.1767464578151703,
      "rewards-max/forensics_sinkhorn_reward": 0.988498866558075,
      "rewards-max/forensics_strict_edge_reward": 0.9890114367008209,
      "rewards-max/forensics_tiou_reward": 0.9980638921260834,
      "rewards/forensics_sinkhorn_reward": 0.4673585593700409,
      "rewards/forensics_strict_edge_reward": 0.2360985428094864,
      "rewards/forensics_tiou_reward": 0.2900897115468979,
      "step": 1140
    },
    {
      "completion_length": 56.90625,
      "epoch": 2.387434554973822,
      "grad_norm": 21.959802627563477,
      "kl": 4.0390625,
      "learning_rate": 4.0251572327044023e-07,
      "loss": 0.3137,
      "reward": 0.5921936929225922,
      "reward_std": 0.1563895046710968,
      "reward_std_all": 0.1782848984003067,
      "rewards-max/forensics_sinkhorn_reward": 0.9086254239082336,
      "rewards-max/forensics_strict_edge_reward": 0.8539886474609375,
      "rewards-max/forensics_tiou_reward": 0.8392998278141022,
      "rewards/forensics_sinkhorn_reward": 0.43880388140678406,
      "rewards/forensics_strict_edge_reward": 0.18402621150016785,
      "rewards/forensics_tiou_reward": 0.2463451474905014,
      "step": 1141
    },
    {
      "completion_length": 122.828125,
      "epoch": 2.3895287958115183,
      "grad_norm": 23.089933395385742,
      "kl": 7.484375,
      "learning_rate": 4.019916142557652e-07,
      "loss": 0.5975,
      "reward": 0.5142835080623627,
      "reward_std": 0.2128700390458107,
      "reward_std_all": 0.24615135043859482,
      "rewards-max/forensics_sinkhorn_reward": 0.9885768890380859,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9983696937561035,
      "rewards/forensics_sinkhorn_reward": 0.4103909581899643,
      "rewards/forensics_strict_edge_reward": 0.24903596937656403,
      "rewards/forensics_tiou_reward": 0.2780604511499405,
      "step": 1142
    },
    {
      "completion_length": 87.15625,
      "epoch": 2.391623036649215,
      "grad_norm": 30.567092895507812,
      "kl": 5.375,
      "learning_rate": 4.0146750524109015e-07,
      "loss": 0.3523,
      "reward": 0.6300424039363861,
      "reward_std": 0.19806210696697235,
      "reward_std_all": 0.1883416846394539,
      "rewards-max/forensics_sinkhorn_reward": 0.987475723028183,
      "rewards-max/forensics_strict_edge_reward": 0.9782643616199493,
      "rewards-max/forensics_tiou_reward": 0.9940564930438995,
      "rewards/forensics_sinkhorn_reward": 0.5423743724822998,
      "rewards/forensics_strict_edge_reward": 0.36648179590702057,
      "rewards/forensics_tiou_reward": 0.40186072885990143,
      "step": 1143
    },
    {
      "completion_length": 114.71875,
      "epoch": 2.393717277486911,
      "grad_norm": 22.827592849731445,
      "kl": 7.9375,
      "learning_rate": 4.009433962264151e-07,
      "loss": 0.5145,
      "reward": 0.5722521841526031,
      "reward_std": 0.29273900389671326,
      "reward_std_all": 0.2692308723926544,
      "rewards-max/forensics_sinkhorn_reward": 0.9885163307189941,
      "rewards-max/forensics_strict_edge_reward": 0.9890114367008209,
      "rewards-max/forensics_tiou_reward": 0.9981308281421661,
      "rewards/forensics_sinkhorn_reward": 0.4875660240650177,
      "rewards/forensics_strict_edge_reward": 0.32811393588781357,
      "rewards/forensics_tiou_reward": 0.3678508698940277,
      "step": 1144
    },
    {
      "completion_length": 75.21875,
      "epoch": 2.3958115183246074,
      "grad_norm": 42.47976303100586,
      "kl": 6.546875,
      "learning_rate": 4.0041928721174e-07,
      "loss": 0.3682,
      "reward": 0.5777037441730499,
      "reward_std": 0.22661308199167252,
      "reward_std_all": 0.22662822157144547,
      "rewards-max/forensics_sinkhorn_reward": 0.9866782426834106,
      "rewards-max/forensics_strict_edge_reward": 0.9357379674911499,
      "rewards-max/forensics_tiou_reward": 0.9909185171127319,
      "rewards/forensics_sinkhorn_reward": 0.45888084173202515,
      "rewards/forensics_strict_edge_reward": 0.2962091788649559,
      "rewards/forensics_tiou_reward": 0.3099770098924637,
      "step": 1145
    },
    {
      "completion_length": 78.21875,
      "epoch": 2.3979057591623034,
      "grad_norm": 40.032588958740234,
      "kl": 6.921875,
      "learning_rate": 3.99895178197065e-07,
      "loss": 0.3996,
      "reward": 0.598268061876297,
      "reward_std": 0.23967302590608597,
      "reward_std_all": 0.22551779448986053,
      "rewards-max/forensics_sinkhorn_reward": 0.9562885463237762,
      "rewards-max/forensics_strict_edge_reward": 0.8064817786216736,
      "rewards-max/forensics_tiou_reward": 0.9292518496513367,
      "rewards/forensics_sinkhorn_reward": 0.4820069223642349,
      "rewards/forensics_strict_edge_reward": 0.22307804226875305,
      "rewards/forensics_tiou_reward": 0.29003213346004486,
      "step": 1146
    },
    {
      "completion_length": 96.546875,
      "epoch": 2.4,
      "grad_norm": 19.994848251342773,
      "kl": 6.203125,
      "learning_rate": 3.9937106918238993e-07,
      "loss": 0.4448,
      "reward": 0.5550407767295837,
      "reward_std": 0.23320290446281433,
      "reward_std_all": 0.24085357785224915,
      "rewards-max/forensics_sinkhorn_reward": 0.9440407454967499,
      "rewards-max/forensics_strict_edge_reward": 0.8317028284072876,
      "rewards-max/forensics_tiou_reward": 0.8950686156749725,
      "rewards/forensics_sinkhorn_reward": 0.4795314818620682,
      "rewards/forensics_strict_edge_reward": 0.2876584976911545,
      "rewards/forensics_tiou_reward": 0.3420626074075699,
      "step": 1147
    },
    {
      "completion_length": 51.296875,
      "epoch": 2.4020942408376964,
      "grad_norm": 22.900236129760742,
      "kl": 3.5546875,
      "learning_rate": 3.9884696016771484e-07,
      "loss": 0.3285,
      "reward": 0.7130833566188812,
      "reward_std": 0.09768734779208899,
      "reward_std_all": 0.09715833514928818,
      "rewards-max/forensics_sinkhorn_reward": 0.9880184531211853,
      "rewards-max/forensics_strict_edge_reward": 0.9782643616199493,
      "rewards-max/forensics_tiou_reward": 0.9961759150028229,
      "rewards/forensics_sinkhorn_reward": 0.6530126333236694,
      "rewards/forensics_strict_edge_reward": 0.3666772395372391,
      "rewards/forensics_tiou_reward": 0.4814610481262207,
      "step": 1148
    },
    {
      "completion_length": 74.65625,
      "epoch": 2.4041884816753925,
      "grad_norm": 37.296749114990234,
      "kl": 7.875,
      "learning_rate": 3.983228511530398e-07,
      "loss": 0.5939,
      "reward": 0.5718129277229309,
      "reward_std": 0.17969992011785507,
      "reward_std_all": 0.19698255509138107,
      "rewards-max/forensics_sinkhorn_reward": 0.984412431716919,
      "rewards-max/forensics_strict_edge_reward": 0.8854433596134186,
      "rewards-max/forensics_tiou_reward": 0.9819993078708649,
      "rewards/forensics_sinkhorn_reward": 0.4897965043783188,
      "rewards/forensics_strict_edge_reward": 0.2340996339917183,
      "rewards/forensics_tiou_reward": 0.2744082361459732,
      "step": 1149
    },
    {
      "completion_length": 55.453125,
      "epoch": 2.406282722513089,
      "grad_norm": 28.234880447387695,
      "kl": 8.140625,
      "learning_rate": 3.977987421383648e-07,
      "loss": 0.4951,
      "reward": 0.6275592148303986,
      "reward_std": 0.20862717926502228,
      "reward_std_all": 0.19819528609514236,
      "rewards-max/forensics_sinkhorn_reward": 0.8910587131977081,
      "rewards-max/forensics_strict_edge_reward": 0.8827816545963287,
      "rewards-max/forensics_tiou_reward": 0.8154967129230499,
      "rewards/forensics_sinkhorn_reward": 0.4552396237850189,
      "rewards/forensics_strict_edge_reward": 0.24393336474895477,
      "rewards/forensics_tiou_reward": 0.29444532096385956,
      "step": 1150
    },
    {
      "completion_length": 71.578125,
      "epoch": 2.4083769633507854,
      "grad_norm": 23.69866943359375,
      "kl": 7.078125,
      "learning_rate": 3.972746331236897e-07,
      "loss": 0.4178,
      "reward": 0.5912122130393982,
      "reward_std": 0.18554094433784485,
      "reward_std_all": 0.19759375602006912,
      "rewards-max/forensics_sinkhorn_reward": 0.8974780738353729,
      "rewards-max/forensics_strict_edge_reward": 0.8214614391326904,
      "rewards-max/forensics_tiou_reward": 0.8109697997570038,
      "rewards/forensics_sinkhorn_reward": 0.4276982247829437,
      "rewards/forensics_strict_edge_reward": 0.19623883068561554,
      "rewards/forensics_tiou_reward": 0.21019555628299713,
      "step": 1151
    },
    {
      "completion_length": 40.984375,
      "epoch": 2.4104712041884815,
      "grad_norm": 36.02736282348633,
      "kl": 5.5390625,
      "learning_rate": 3.9675052410901467e-07,
      "loss": 0.4507,
      "reward": 0.6774882972240448,
      "reward_std": 0.12674643844366074,
      "reward_std_all": 0.13386466354131699,
      "rewards-max/forensics_sinkhorn_reward": 0.9886201620101929,
      "rewards-max/forensics_strict_edge_reward": 0.9890114367008209,
      "rewards-max/forensics_tiou_reward": 0.9985401332378387,
      "rewards/forensics_sinkhorn_reward": 0.6389488279819489,
      "rewards/forensics_strict_edge_reward": 0.4166591167449951,
      "rewards/forensics_tiou_reward": 0.47776423394680023,
      "step": 1152
    },
    {
      "completion_length": 102.859375,
      "epoch": 2.412565445026178,
      "grad_norm": 22.703536987304688,
      "kl": 7.65625,
      "learning_rate": 3.9622641509433963e-07,
      "loss": 0.5162,
      "reward": 0.5532156825065613,
      "reward_std": 0.204451322555542,
      "reward_std_all": 0.22538124024868011,
      "rewards-max/forensics_sinkhorn_reward": 0.9875025451183319,
      "rewards-max/forensics_strict_edge_reward": 0.9565287232398987,
      "rewards-max/forensics_tiou_reward": 0.9941399693489075,
      "rewards/forensics_sinkhorn_reward": 0.455904096364975,
      "rewards/forensics_strict_edge_reward": 0.2818175405263901,
      "rewards/forensics_tiou_reward": 0.32399411499500275,
      "step": 1153
    },
    {
      "completion_length": 92.703125,
      "epoch": 2.4146596858638745,
      "grad_norm": 38.78409194946289,
      "kl": 9.359375,
      "learning_rate": 3.9570230607966454e-07,
      "loss": 0.4816,
      "reward": 0.5024362206459045,
      "reward_std": 0.251980721950531,
      "reward_std_all": 0.2651967853307724,
      "rewards-max/forensics_sinkhorn_reward": 0.9391838014125824,
      "rewards-max/forensics_strict_edge_reward": 0.717239111661911,
      "rewards-max/forensics_tiou_reward": 0.8941473066806793,
      "rewards/forensics_sinkhorn_reward": 0.403731107711792,
      "rewards/forensics_strict_edge_reward": 0.15146543830633163,
      "rewards/forensics_tiou_reward": 0.19515763223171234,
      "step": 1154
    },
    {
      "completion_length": 56.75,
      "epoch": 2.4167539267015705,
      "grad_norm": 22.15398406982422,
      "kl": 4.6875,
      "learning_rate": 3.951781970649895e-07,
      "loss": 0.203,
      "reward": 0.7343192398548126,
      "reward_std": 0.11362109053879976,
      "reward_std_all": 0.10096327494829893,
      "rewards-max/forensics_sinkhorn_reward": 0.9778542518615723,
      "rewards-max/forensics_strict_edge_reward": 0.9062341153621674,
      "rewards-max/forensics_tiou_reward": 0.9737332463264465,
      "rewards/forensics_sinkhorn_reward": 0.6389129161834717,
      "rewards/forensics_strict_edge_reward": 0.3823116570711136,
      "rewards/forensics_tiou_reward": 0.49114128947257996,
      "step": 1155
    },
    {
      "completion_length": 63.21875,
      "epoch": 2.418848167539267,
      "grad_norm": 22.45012092590332,
      "kl": 5.703125,
      "learning_rate": 3.9465408805031445e-07,
      "loss": 0.4667,
      "reward": 0.5931847095489502,
      "reward_std": 0.19098999351263046,
      "reward_std_all": 0.19269481301307678,
      "rewards-max/forensics_sinkhorn_reward": 0.9875406622886658,
      "rewards-max/forensics_strict_edge_reward": 0.947419673204422,
      "rewards-max/forensics_tiou_reward": 0.9942857027053833,
      "rewards/forensics_sinkhorn_reward": 0.49650391936302185,
      "rewards/forensics_strict_edge_reward": 0.30403925478458405,
      "rewards/forensics_tiou_reward": 0.3759464919567108,
      "step": 1156
    },
    {
      "completion_length": 91.46875,
      "epoch": 2.4209424083769635,
      "grad_norm": 24.458942413330078,
      "kl": 6.515625,
      "learning_rate": 3.9412997903563936e-07,
      "loss": 0.3147,
      "reward": 0.5293758511543274,
      "reward_std": 0.18948203325271606,
      "reward_std_all": 0.23319388926029205,
      "rewards-max/forensics_sinkhorn_reward": 0.950533926486969,
      "rewards-max/forensics_strict_edge_reward": 0.8327312767505646,
      "rewards-max/forensics_tiou_reward": 0.9098523557186127,
      "rewards/forensics_sinkhorn_reward": 0.4244399964809418,
      "rewards/forensics_strict_edge_reward": 0.1735663041472435,
      "rewards/forensics_tiou_reward": 0.23721818625926971,
      "step": 1157
    },
    {
      "completion_length": 62.921875,
      "epoch": 2.4230366492146596,
      "grad_norm": 25.092391967773438,
      "kl": 4.84375,
      "learning_rate": 3.9360587002096437e-07,
      "loss": 0.4365,
      "reward": 0.6447305977344513,
      "reward_std": 0.18759462237358093,
      "reward_std_all": 0.17898091673851013,
      "rewards-max/forensics_sinkhorn_reward": 0.987281858921051,
      "rewards-max/forensics_strict_edge_reward": 0.9782643616199493,
      "rewards-max/forensics_tiou_reward": 0.993288666009903,
      "rewards/forensics_sinkhorn_reward": 0.6117730736732483,
      "rewards/forensics_strict_edge_reward": 0.4319290369749069,
      "rewards/forensics_tiou_reward": 0.5086233019828796,
      "step": 1158
    },
    {
      "completion_length": 89.0,
      "epoch": 2.425130890052356,
      "grad_norm": 36.53801345825195,
      "kl": 5.875,
      "learning_rate": 3.9308176100628933e-07,
      "loss": 0.3153,
      "reward": 0.6059535145759583,
      "reward_std": 0.20128133893013,
      "reward_std_all": 0.2101963609457016,
      "rewards-max/forensics_sinkhorn_reward": 0.985368013381958,
      "rewards-max/forensics_strict_edge_reward": 0.905340164899826,
      "rewards-max/forensics_tiou_reward": 0.9857572019100189,
      "rewards/forensics_sinkhorn_reward": 0.5374476909637451,
      "rewards/forensics_strict_edge_reward": 0.3332795649766922,
      "rewards/forensics_tiou_reward": 0.39117753505706787,
      "step": 1159
    },
    {
      "completion_length": 89.0625,
      "epoch": 2.4272251308900525,
      "grad_norm": 24.862974166870117,
      "kl": 9.109375,
      "learning_rate": 3.9255765199161424e-07,
      "loss": 0.4857,
      "reward": 0.5591188669204712,
      "reward_std": 0.24320191144943237,
      "reward_std_all": 0.2435600459575653,
      "rewards-max/forensics_sinkhorn_reward": 0.9887801706790924,
      "rewards-max/forensics_strict_edge_reward": 0.9890114367008209,
      "rewards-max/forensics_tiou_reward": 0.9991711676120758,
      "rewards/forensics_sinkhorn_reward": 0.499640554189682,
      "rewards/forensics_strict_edge_reward": 0.2604617476463318,
      "rewards/forensics_tiou_reward": 0.3462396413087845,
      "step": 1160
    },
    {
      "completion_length": 69.25,
      "epoch": 2.4293193717277486,
      "grad_norm": 25.41468620300293,
      "kl": 5.5625,
      "learning_rate": 3.920335429769392e-07,
      "loss": 0.3185,
      "reward": 0.5991115570068359,
      "reward_std": 0.21543943881988525,
      "reward_std_all": 0.21371467411518097,
      "rewards-max/forensics_sinkhorn_reward": 0.9882445633411407,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9970608651638031,
      "rewards/forensics_sinkhorn_reward": 0.5382251739501953,
      "rewards/forensics_strict_edge_reward": 0.29357127845287323,
      "rewards/forensics_tiou_reward": 0.3518955111503601,
      "step": 1161
    },
    {
      "completion_length": 33.734375,
      "epoch": 2.431413612565445,
      "grad_norm": 20.317075729370117,
      "kl": 4.1796875,
      "learning_rate": 3.9150943396226415e-07,
      "loss": 0.3745,
      "reward": 0.6670593321323395,
      "reward_std": 0.11497106775641441,
      "reward_std_all": 0.12511616200208664,
      "rewards-max/forensics_sinkhorn_reward": 0.9884390234947205,
      "rewards-max/forensics_strict_edge_reward": 0.9782643616199493,
      "rewards-max/forensics_tiou_reward": 0.997826099395752,
      "rewards/forensics_sinkhorn_reward": 0.5617338716983795,
      "rewards/forensics_strict_edge_reward": 0.3576466590166092,
      "rewards/forensics_tiou_reward": 0.4186442196369171,
      "step": 1162
    },
    {
      "completion_length": 47.265625,
      "epoch": 2.4335078534031416,
      "grad_norm": 21.34722328186035,
      "kl": 5.640625,
      "learning_rate": 3.9098532494758906e-07,
      "loss": 0.3928,
      "reward": 0.6603001654148102,
      "reward_std": 0.12742344290018082,
      "reward_std_all": 0.13783591985702515,
      "rewards-max/forensics_sinkhorn_reward": 0.988379955291748,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9975936114788055,
      "rewards/forensics_sinkhorn_reward": 0.5786080658435822,
      "rewards/forensics_strict_edge_reward": 0.29620539397001266,
      "rewards/forensics_tiou_reward": 0.3839588910341263,
      "step": 1163
    },
    {
      "completion_length": 87.875,
      "epoch": 2.4356020942408376,
      "grad_norm": 32.11140441894531,
      "kl": 7.359375,
      "learning_rate": 3.90461215932914e-07,
      "loss": 0.4854,
      "reward": 0.5776374340057373,
      "reward_std": 0.1881919875741005,
      "reward_std_all": 0.2086673453450203,
      "rewards-max/forensics_sinkhorn_reward": 0.9560785293579102,
      "rewards-max/forensics_strict_edge_reward": 0.7675493955612183,
      "rewards-max/forensics_tiou_reward": 0.9299719333648682,
      "rewards/forensics_sinkhorn_reward": 0.44045010209083557,
      "rewards/forensics_strict_edge_reward": 0.19731508940458298,
      "rewards/forensics_tiou_reward": 0.26618143171072006,
      "step": 1164
    },
    {
      "completion_length": 52.34375,
      "epoch": 2.437696335078534,
      "grad_norm": 31.378110885620117,
      "kl": 5.484375,
      "learning_rate": 3.89937106918239e-07,
      "loss": 0.2328,
      "reward": 0.682261049747467,
      "reward_std": 0.1166495531797409,
      "reward_std_all": 0.1202801838517189,
      "rewards-max/forensics_sinkhorn_reward": 0.9886201620101929,
      "rewards-max/forensics_strict_edge_reward": 0.9890114367008209,
      "rewards-max/forensics_tiou_reward": 0.9985401332378387,
      "rewards/forensics_sinkhorn_reward": 0.6440793871879578,
      "rewards/forensics_strict_edge_reward": 0.4159035235643387,
      "rewards/forensics_tiou_reward": 0.5219546556472778,
      "step": 1165
    },
    {
      "completion_length": 96.609375,
      "epoch": 2.4397905759162306,
      "grad_norm": 25.440275192260742,
      "kl": 6.390625,
      "learning_rate": 3.8941299790356394e-07,
      "loss": 0.4191,
      "reward": 0.4985083192586899,
      "reward_std": 0.26158633828163147,
      "reward_std_all": 0.2799850255250931,
      "rewards-max/forensics_sinkhorn_reward": 0.9678208827972412,
      "rewards-max/forensics_strict_edge_reward": 0.7852677702903748,
      "rewards-max/forensics_tiou_reward": 0.9505491554737091,
      "rewards/forensics_sinkhorn_reward": 0.4046408087015152,
      "rewards/forensics_strict_edge_reward": 0.15380245447158813,
      "rewards/forensics_tiou_reward": 0.25140465795993805,
      "step": 1166
    },
    {
      "completion_length": 42.28125,
      "epoch": 2.4418848167539267,
      "grad_norm": 16.93971061706543,
      "kl": 3.5625,
      "learning_rate": 3.888888888888889e-07,
      "loss": 0.1622,
      "reward": 0.6141751408576965,
      "reward_std": 0.17411736398935318,
      "reward_std_all": 0.18323107063770294,
      "rewards-max/forensics_sinkhorn_reward": 0.981528103351593,
      "rewards-max/forensics_strict_edge_reward": 0.9464850425720215,
      "rewards-max/forensics_tiou_reward": 0.9836231768131256,
      "rewards/forensics_sinkhorn_reward": 0.4816581457853317,
      "rewards/forensics_strict_edge_reward": 0.22634827345609665,
      "rewards/forensics_tiou_reward": 0.2680651545524597,
      "step": 1167
    },
    {
      "completion_length": 50.671875,
      "epoch": 2.443979057591623,
      "grad_norm": 29.308752059936523,
      "kl": 5.0,
      "learning_rate": 3.8836477987421385e-07,
      "loss": 0.1838,
      "reward": 0.6293146908283234,
      "reward_std": 0.13974222540855408,
      "reward_std_all": 0.15967725217342377,
      "rewards-max/forensics_sinkhorn_reward": 0.9748754501342773,
      "rewards-max/forensics_strict_edge_reward": 0.8805811703205109,
      "rewards-max/forensics_tiou_reward": 0.966166615486145,
      "rewards/forensics_sinkhorn_reward": 0.5350403189659119,
      "rewards/forensics_strict_edge_reward": 0.21921831369400024,
      "rewards/forensics_tiou_reward": 0.26535117626190186,
      "step": 1168
    },
    {
      "completion_length": 75.8125,
      "epoch": 2.446073298429319,
      "grad_norm": 27.94449234008789,
      "kl": 5.609375,
      "learning_rate": 3.8784067085953876e-07,
      "loss": 0.4542,
      "reward": 0.6150202751159668,
      "reward_std": 0.16376382112503052,
      "reward_std_all": 0.18128979206085205,
      "rewards-max/forensics_sinkhorn_reward": 0.9749318361282349,
      "rewards-max/forensics_strict_edge_reward": 0.9062341153621674,
      "rewards-max/forensics_tiou_reward": 0.9691889584064484,
      "rewards/forensics_sinkhorn_reward": 0.5725013017654419,
      "rewards/forensics_strict_edge_reward": 0.28536927700042725,
      "rewards/forensics_tiou_reward": 0.4373050183057785,
      "step": 1169
    },
    {
      "completion_length": 53.6875,
      "epoch": 2.4481675392670157,
      "grad_norm": 34.6297721862793,
      "kl": 5.15625,
      "learning_rate": 3.873165618448637e-07,
      "loss": 0.3044,
      "reward": 0.6433026194572449,
      "reward_std": 0.1932523027062416,
      "reward_std_all": 0.18603961169719696,
      "rewards-max/forensics_sinkhorn_reward": 0.9534305334091187,
      "rewards-max/forensics_strict_edge_reward": 0.802157998085022,
      "rewards-max/forensics_tiou_reward": 0.9217284917831421,
      "rewards/forensics_sinkhorn_reward": 0.5665171146392822,
      "rewards/forensics_strict_edge_reward": 0.22553765028715134,
      "rewards/forensics_tiou_reward": 0.36745208501815796,
      "step": 1170
    },
    {
      "completion_length": 39.53125,
      "epoch": 2.450261780104712,
      "grad_norm": 15.706419944763184,
      "kl": 2.9140625,
      "learning_rate": 3.867924528301887e-07,
      "loss": 0.1691,
      "reward": 0.6702202558517456,
      "reward_std": 0.1105368435382843,
      "reward_std_all": 0.12069212272763252,
      "rewards-max/forensics_sinkhorn_reward": 0.986999124288559,
      "rewards-max/forensics_strict_edge_reward": 0.8983768224716187,
      "rewards-max/forensics_tiou_reward": 0.9921697676181793,
      "rewards/forensics_sinkhorn_reward": 0.5919755697250366,
      "rewards/forensics_strict_edge_reward": 0.3229702338576317,
      "rewards/forensics_tiou_reward": 0.36809661984443665,
      "step": 1171
    },
    {
      "completion_length": 80.09375,
      "epoch": 2.4523560209424082,
      "grad_norm": 26.214269638061523,
      "kl": 8.390625,
      "learning_rate": 3.862683438155136e-07,
      "loss": 0.5786,
      "reward": 0.5641180872917175,
      "reward_std": 0.18419714272022247,
      "reward_std_all": 0.20446207374334335,
      "rewards-max/forensics_sinkhorn_reward": 0.9840809404850006,
      "rewards-max/forensics_strict_edge_reward": 0.8957232534885406,
      "rewards-max/forensics_tiou_reward": 0.9807792007923126,
      "rewards/forensics_sinkhorn_reward": 0.4170901030302048,
      "rewards/forensics_strict_edge_reward": 0.1886829137802124,
      "rewards/forensics_tiou_reward": 0.20220033824443817,
      "step": 1172
    },
    {
      "completion_length": 61.78125,
      "epoch": 2.4544502617801047,
      "grad_norm": 27.1899356842041,
      "kl": 4.5703125,
      "learning_rate": 3.8574423480083854e-07,
      "loss": 0.3899,
      "reward": 0.6684710085391998,
      "reward_std": 0.18571317940950394,
      "reward_std_all": 0.16908207535743713,
      "rewards-max/forensics_sinkhorn_reward": 0.9889025986194611,
      "rewards-max/forensics_strict_edge_reward": 0.9890114367008209,
      "rewards-max/forensics_tiou_reward": 0.999653548002243,
      "rewards/forensics_sinkhorn_reward": 0.6015601456165314,
      "rewards/forensics_strict_edge_reward": 0.38932740688323975,
      "rewards/forensics_tiou_reward": 0.43381141126155853,
      "step": 1173
    },
    {
      "completion_length": 44.21875,
      "epoch": 2.456544502617801,
      "grad_norm": 14.871994018554688,
      "kl": 3.859375,
      "learning_rate": 3.8522012578616355e-07,
      "loss": 0.2494,
      "reward": 0.6335309147834778,
      "reward_std": 0.14106608554720879,
      "reward_std_all": 0.15428058803081512,
      "rewards-max/forensics_sinkhorn_reward": 0.9866423010826111,
      "rewards-max/forensics_strict_edge_reward": 0.9256840348243713,
      "rewards-max/forensics_tiou_reward": 0.9907465875148773,
      "rewards/forensics_sinkhorn_reward": 0.565073698759079,
      "rewards/forensics_strict_edge_reward": 0.276249922811985,
      "rewards/forensics_tiou_reward": 0.36931100487709045,
      "step": 1174
    },
    {
      "completion_length": 51.34375,
      "epoch": 2.4586387434554973,
      "grad_norm": 20.43601417541504,
      "kl": 5.734375,
      "learning_rate": 3.8469601677148846e-07,
      "loss": 0.5004,
      "reward": 0.6384837627410889,
      "reward_std": 0.14342057704925537,
      "reward_std_all": 0.15508411079645157,
      "rewards-max/forensics_sinkhorn_reward": 0.9871483147144318,
      "rewards-max/forensics_strict_edge_reward": 0.9565287232398987,
      "rewards-max/forensics_tiou_reward": 0.9927583038806915,
      "rewards/forensics_sinkhorn_reward": 0.5718087255954742,
      "rewards/forensics_strict_edge_reward": 0.36180080473423004,
      "rewards/forensics_tiou_reward": 0.40913882851600647,
      "step": 1175
    },
    {
      "completion_length": 33.078125,
      "epoch": 2.4607329842931938,
      "grad_norm": 15.054343223571777,
      "kl": 3.1875,
      "learning_rate": 3.841719077568134e-07,
      "loss": 0.2179,
      "reward": 0.630696564912796,
      "reward_std": 0.16890899091959,
      "reward_std_all": 0.17517820000648499,
      "rewards-max/forensics_sinkhorn_reward": 0.9881939589977264,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9968628883361816,
      "rewards/forensics_sinkhorn_reward": 0.5830380916595459,
      "rewards/forensics_strict_edge_reward": 0.40720927715301514,
      "rewards/forensics_tiou_reward": 0.4429738223552704,
      "step": 1176
    },
    {
      "completion_length": 44.015625,
      "epoch": 2.46282722513089,
      "grad_norm": 40.65653991699219,
      "kl": 5.046875,
      "learning_rate": 3.836477987421384e-07,
      "loss": 0.3853,
      "reward": 0.6239669919013977,
      "reward_std": 0.12905904836952686,
      "reward_std_all": 0.16160347312688828,
      "rewards-max/forensics_sinkhorn_reward": 0.9874242842197418,
      "rewards-max/forensics_strict_edge_reward": 0.947419673204422,
      "rewards-max/forensics_tiou_reward": 0.993846207857132,
      "rewards/forensics_sinkhorn_reward": 0.5628260374069214,
      "rewards/forensics_strict_edge_reward": 0.3314446657896042,
      "rewards/forensics_tiou_reward": 0.40541040897369385,
      "step": 1177
    },
    {
      "completion_length": 57.84375,
      "epoch": 2.4649214659685863,
      "grad_norm": 26.33439826965332,
      "kl": 7.203125,
      "learning_rate": 3.831236897274633e-07,
      "loss": 0.4782,
      "reward": 0.5772064924240112,
      "reward_std": 0.2051214724779129,
      "reward_std_all": 0.21131650358438492,
      "rewards-max/forensics_sinkhorn_reward": 0.8773207664489746,
      "rewards-max/forensics_strict_edge_reward": 0.8491117358207703,
      "rewards-max/forensics_tiou_reward": 0.7762356102466583,
      "rewards/forensics_sinkhorn_reward": 0.46690574288368225,
      "rewards/forensics_strict_edge_reward": 0.16731618344783783,
      "rewards/forensics_tiou_reward": 0.2447357401251793,
      "step": 1178
    },
    {
      "completion_length": 30.6875,
      "epoch": 2.467015706806283,
      "grad_norm": 30.3643856048584,
      "kl": 2.458984375,
      "learning_rate": 3.8259958071278824e-07,
      "loss": 0.1425,
      "reward": 0.73784339427948,
      "reward_std": 0.10222591646015644,
      "reward_std_all": 0.09189976379275322,
      "rewards-max/forensics_sinkhorn_reward": 0.9650532305240631,
      "rewards-max/forensics_strict_edge_reward": 0.9158510267734528,
      "rewards-max/forensics_tiou_reward": 0.947680652141571,
      "rewards/forensics_sinkhorn_reward": 0.6398223042488098,
      "rewards/forensics_strict_edge_reward": 0.4247434735298157,
      "rewards/forensics_tiou_reward": 0.480451375246048,
      "step": 1179
    },
    {
      "completion_length": 55.0,
      "epoch": 2.469109947643979,
      "grad_norm": 30.54594612121582,
      "kl": 8.203125,
      "learning_rate": 3.820754716981132e-07,
      "loss": 0.57,
      "reward": 0.596677839756012,
      "reward_std": 0.16455156728625298,
      "reward_std_all": 0.1872246339917183,
      "rewards-max/forensics_sinkhorn_reward": 0.8584553301334381,
      "rewards-max/forensics_strict_edge_reward": 0.835724264383316,
      "rewards-max/forensics_tiou_reward": 0.7430367022752762,
      "rewards/forensics_sinkhorn_reward": 0.5723888874053955,
      "rewards/forensics_strict_edge_reward": 0.353520467877388,
      "rewards/forensics_tiou_reward": 0.4277169406414032,
      "step": 1180
    },
    {
      "completion_length": 42.34375,
      "epoch": 2.4712041884816753,
      "grad_norm": 20.251632690429688,
      "kl": 4.734375,
      "learning_rate": 3.815513626834381e-07,
      "loss": 0.4258,
      "reward": 0.5741243660449982,
      "reward_std": 0.19136770069599152,
      "reward_std_all": 0.20345183461904526,
      "rewards-max/forensics_sinkhorn_reward": 0.986691951751709,
      "rewards-max/forensics_strict_edge_reward": 0.9256840348243713,
      "rewards-max/forensics_tiou_reward": 0.9909732937812805,
      "rewards/forensics_sinkhorn_reward": 0.5008310973644257,
      "rewards/forensics_strict_edge_reward": 0.20605652779340744,
      "rewards/forensics_tiou_reward": 0.3064137101173401,
      "step": 1181
    },
    {
      "completion_length": 38.96875,
      "epoch": 2.473298429319372,
      "grad_norm": 22.234710693359375,
      "kl": 4.1875,
      "learning_rate": 3.810272536687631e-07,
      "loss": 0.2394,
      "reward": 0.6709283292293549,
      "reward_std": 0.1192951537668705,
      "reward_std_all": 0.12362856417894363,
      "rewards-max/forensics_sinkhorn_reward": 0.9816085994243622,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9830001592636108,
      "rewards/forensics_sinkhorn_reward": 0.6108793616294861,
      "rewards/forensics_strict_edge_reward": 0.43389010429382324,
      "rewards/forensics_tiou_reward": 0.4353261888027191,
      "step": 1182
    },
    {
      "completion_length": 41.515625,
      "epoch": 2.475392670157068,
      "grad_norm": 27.013938903808594,
      "kl": 4.3359375,
      "learning_rate": 3.805031446540881e-07,
      "loss": 0.2999,
      "reward": 0.6547936201095581,
      "reward_std": 0.14033004641532898,
      "reward_std_all": 0.14970462769269943,
      "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": 0.6015934348106384,
      "rewards/forensics_strict_edge_reward": 0.3471987247467041,
      "rewards/forensics_tiou_reward": 0.4255152642726898,
      "step": 1183
    },
    {
      "completion_length": 42.171875,
      "epoch": 2.4774869109947644,
      "grad_norm": 19.000200271606445,
      "kl": 4.25390625,
      "learning_rate": 3.79979035639413e-07,
      "loss": 0.3074,
      "reward": 0.6093397736549377,
      "reward_std": 0.15490638464689255,
      "reward_std_all": 0.17084331810474396,
      "rewards-max/forensics_sinkhorn_reward": 0.9868682324886322,
      "rewards-max/forensics_strict_edge_reward": 0.9464850425720215,
      "rewards-max/forensics_tiou_reward": 0.9916614890098572,
      "rewards/forensics_sinkhorn_reward": 0.5072462856769562,
      "rewards/forensics_strict_edge_reward": 0.2287762388586998,
      "rewards/forensics_tiou_reward": 0.31563861668109894,
      "step": 1184
    },
    {
      "completion_length": 80.40625,
      "epoch": 2.479581151832461,
      "grad_norm": 31.204221725463867,
      "kl": 8.078125,
      "learning_rate": 3.7945492662473794e-07,
      "loss": 0.4845,
      "reward": 0.6155402064323425,
      "reward_std": 0.18207884952425957,
      "reward_std_all": 0.17936433479189873,
      "rewards-max/forensics_sinkhorn_reward": 0.9622414112091064,
      "rewards-max/forensics_strict_edge_reward": 0.7999143600463867,
      "rewards-max/forensics_tiou_reward": 0.9388310015201569,
      "rewards/forensics_sinkhorn_reward": 0.5157965123653412,
      "rewards/forensics_strict_edge_reward": 0.19925352185964584,
      "rewards/forensics_tiou_reward": 0.323619969189167,
      "step": 1185
    },
    {
      "completion_length": 83.453125,
      "epoch": 2.481675392670157,
      "grad_norm": 24.67180633544922,
      "kl": 10.78125,
      "learning_rate": 3.789308176100629e-07,
      "loss": 0.6138,
      "reward": 0.5921484529972076,
      "reward_std": 0.23504018783569336,
      "reward_std_all": 0.22605819255113602,
      "rewards-max/forensics_sinkhorn_reward": 0.988250732421875,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9970836937427521,
      "rewards/forensics_sinkhorn_reward": 0.522342175245285,
      "rewards/forensics_strict_edge_reward": 0.3041018396615982,
      "rewards/forensics_tiou_reward": 0.36149153113365173,
      "step": 1186
    },
    {
      "completion_length": 47.21875,
      "epoch": 2.4837696335078534,
      "grad_norm": 21.110815048217773,
      "kl": 7.71875,
      "learning_rate": 3.784067085953878e-07,
      "loss": 0.516,
      "reward": 0.5808124840259552,
      "reward_std": 0.19406241923570633,
      "reward_std_all": 0.20383420586585999,
      "rewards-max/forensics_sinkhorn_reward": 0.9877660572528839,
      "rewards-max/forensics_strict_edge_reward": 0.9565287232398987,
      "rewards-max/forensics_tiou_reward": 0.9951809048652649,
      "rewards/forensics_sinkhorn_reward": 0.5073081254959106,
      "rewards/forensics_strict_edge_reward": 0.267045721411705,
      "rewards/forensics_tiou_reward": 0.3411223441362381,
      "step": 1187
    },
    {
      "completion_length": 72.984375,
      "epoch": 2.48586387434555,
      "grad_norm": 46.16224670410156,
      "kl": 9.84375,
      "learning_rate": 3.7788259958071276e-07,
      "loss": 0.592,
      "reward": 0.6025945544242859,
      "reward_std": 0.20288123935461044,
      "reward_std_all": 0.21178381890058517,
      "rewards-max/forensics_sinkhorn_reward": 0.9819359183311462,
      "rewards-max/forensics_strict_edge_reward": 0.9890114367008209,
      "rewards-max/forensics_tiou_reward": 0.9848646819591522,
      "rewards/forensics_sinkhorn_reward": 0.5221903324127197,
      "rewards/forensics_strict_edge_reward": 0.26242808997631073,
      "rewards/forensics_tiou_reward": 0.31727591156959534,
      "step": 1188
    },
    {
      "completion_length": 42.46875,
      "epoch": 2.487958115183246,
      "grad_norm": 31.291757583618164,
      "kl": 8.046875,
      "learning_rate": 3.773584905660377e-07,
      "loss": 0.4509,
      "reward": 0.6879203617572784,
      "reward_std": 0.16316598653793335,
      "reward_std_all": 0.1479327231645584,
      "rewards-max/forensics_sinkhorn_reward": 0.9883950054645538,
      "rewards-max/forensics_strict_edge_reward": 0.9890114367008209,
      "rewards-max/forensics_tiou_reward": 0.9976553320884705,
      "rewards/forensics_sinkhorn_reward": 0.5966038703918457,
      "rewards/forensics_strict_edge_reward": 0.38079874217510223,
      "rewards/forensics_tiou_reward": 0.42038694024086,
      "step": 1189
    },
    {
      "completion_length": 52.671875,
      "epoch": 2.4900523560209424,
      "grad_norm": 19.232473373413086,
      "kl": 4.703125,
      "learning_rate": 3.768343815513627e-07,
      "loss": 0.2888,
      "reward": 0.6587303280830383,
      "reward_std": 0.10786037519574165,
      "reward_std_all": 0.11957376450300217,
      "rewards-max/forensics_sinkhorn_reward": 0.9877736568450928,
      "rewards-max/forensics_strict_edge_reward": 0.9574736058712006,
      "rewards-max/forensics_tiou_reward": 0.9952153265476227,
      "rewards/forensics_sinkhorn_reward": 0.5575846433639526,
      "rewards/forensics_strict_edge_reward": 0.39936043322086334,
      "rewards/forensics_tiou_reward": 0.4509214758872986,
      "step": 1190
    },
    {
      "completion_length": 42.09375,
      "epoch": 2.492146596858639,
      "grad_norm": 15.932609558105469,
      "kl": 4.890625,
      "learning_rate": 3.7631027253668764e-07,
      "loss": 0.3476,
      "reward": 0.7031199932098389,
      "reward_std": 0.135312058031559,
      "reward_std_all": 0.1241857260465622,
      "rewards-max/forensics_sinkhorn_reward": 0.9823912680149078,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9842860698699951,
      "rewards/forensics_sinkhorn_reward": 0.6475357711315155,
      "rewards/forensics_strict_edge_reward": 0.41203613579273224,
      "rewards/forensics_tiou_reward": 0.5157907456159592,
      "step": 1191
    },
    {
      "completion_length": 73.609375,
      "epoch": 2.494240837696335,
      "grad_norm": 42.93019104003906,
      "kl": 6.71875,
      "learning_rate": 3.757861635220126e-07,
      "loss": 0.3929,
      "reward": 0.5884063541889191,
      "reward_std": 0.22166325896978378,
      "reward_std_all": 0.21495644003152847,
      "rewards-max/forensics_sinkhorn_reward": 0.9869706034660339,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9920385181903839,
      "rewards/forensics_sinkhorn_reward": 0.4905330389738083,
      "rewards/forensics_strict_edge_reward": 0.26500674337148666,
      "rewards/forensics_tiou_reward": 0.3007969409227371,
      "step": 1192
    },
    {
      "completion_length": 85.8125,
      "epoch": 2.4963350785340315,
      "grad_norm": 21.448131561279297,
      "kl": 6.28125,
      "learning_rate": 3.752620545073375e-07,
      "loss": 0.3676,
      "reward": 0.6109731793403625,
      "reward_std": 0.20593054592609406,
      "reward_std_all": 0.20261816680431366,
      "rewards-max/forensics_sinkhorn_reward": 0.9866218864917755,
      "rewards-max/forensics_strict_edge_reward": 0.9464850425720215,
      "rewards-max/forensics_tiou_reward": 0.9906964004039764,
      "rewards/forensics_sinkhorn_reward": 0.5178303122520447,
      "rewards/forensics_strict_edge_reward": 0.25804470479488373,
      "rewards/forensics_tiou_reward": 0.3278023302555084,
      "step": 1193
    },
    {
      "completion_length": 79.234375,
      "epoch": 2.498429319371728,
      "grad_norm": 26.05807876586914,
      "kl": 5.9375,
      "learning_rate": 3.7473794549266246e-07,
      "loss": 0.4727,
      "reward": 0.6039115786552429,
      "reward_std": 0.20343151688575745,
      "reward_std_all": 0.2042590081691742,
      "rewards-max/forensics_sinkhorn_reward": 0.9810994267463684,
      "rewards-max/forensics_strict_edge_reward": 0.9151731729507446,
      "rewards-max/forensics_tiou_reward": 0.9789406061172485,
      "rewards/forensics_sinkhorn_reward": 0.5308538377285004,
      "rewards/forensics_strict_edge_reward": 0.2959897518157959,
      "rewards/forensics_tiou_reward": 0.38978777825832367,
      "step": 1194
    },
    {
      "completion_length": 94.953125,
      "epoch": 2.500523560209424,
      "grad_norm": 45.77395248413086,
      "kl": 7.796875,
      "learning_rate": 3.7421383647798737e-07,
      "loss": 0.5763,
      "reward": 0.5836105346679688,
      "reward_std": 0.2054467797279358,
      "reward_std_all": 0.21807924658060074,
      "rewards-max/forensics_sinkhorn_reward": 0.9583784341812134,
      "rewards-max/forensics_strict_edge_reward": 0.9093653857707977,
      "rewards-max/forensics_tiou_reward": 0.9329230189323425,
      "rewards/forensics_sinkhorn_reward": 0.4706001877784729,
      "rewards/forensics_strict_edge_reward": 0.28034521639347076,
      "rewards/forensics_tiou_reward": 0.2921150028705597,
      "step": 1195
    },
    {
      "completion_length": 40.1875,
      "epoch": 2.5026178010471205,
      "grad_norm": 24.28745460510254,
      "kl": 5.390625,
      "learning_rate": 3.7368972746331233e-07,
      "loss": 0.3395,
      "reward": 0.5971441566944122,
      "reward_std": 0.21138029545545578,
      "reward_std_all": 0.2086411863565445,
      "rewards-max/forensics_sinkhorn_reward": 0.987930417060852,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9958254992961884,
      "rewards/forensics_sinkhorn_reward": 0.5352807343006134,
      "rewards/forensics_strict_edge_reward": 0.26594628393650055,
      "rewards/forensics_tiou_reward": 0.3230052888393402,
      "step": 1196
    },
    {
      "completion_length": 53.328125,
      "epoch": 2.504712041884817,
      "grad_norm": 24.267728805541992,
      "kl": 5.59375,
      "learning_rate": 3.731656184486373e-07,
      "loss": 0.4098,
      "reward": 0.6248288750648499,
      "reward_std": 0.19338425993919373,
      "reward_std_all": 0.19029059261083603,
      "rewards-max/forensics_sinkhorn_reward": 0.9857869148254395,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9924074113368988,
      "rewards/forensics_sinkhorn_reward": 0.5362978875637054,
      "rewards/forensics_strict_edge_reward": 0.3877152353525162,
      "rewards/forensics_tiou_reward": 0.397528737783432,
      "step": 1197
    },
    {
      "completion_length": 83.375,
      "epoch": 2.506806282722513,
      "grad_norm": 116.47693634033203,
      "kl": 9.375,
      "learning_rate": 3.7264150943396224e-07,
      "loss": 0.7003,
      "reward": 0.5343402624130249,
      "reward_std": 0.2728988379240036,
      "reward_std_all": 0.26513465493917465,
      "rewards-max/forensics_sinkhorn_reward": 0.9701376557350159,
      "rewards-max/forensics_strict_edge_reward": 0.882964164018631,
      "rewards-max/forensics_tiou_reward": 0.9570401012897491,
      "rewards/forensics_sinkhorn_reward": 0.4608503133058548,
      "rewards/forensics_strict_edge_reward": 0.2608598992228508,
      "rewards/forensics_tiou_reward": 0.33011655509471893,
      "step": 1198
    },
    {
      "completion_length": 40.046875,
      "epoch": 2.5089005235602095,
      "grad_norm": 33.87000274658203,
      "kl": 4.875,
      "learning_rate": 3.721174004192872e-07,
      "loss": 0.3663,
      "reward": 0.6096901893615723,
      "reward_std": 0.16598854959011078,
      "reward_std_all": 0.18124817311763763,
      "rewards-max/forensics_sinkhorn_reward": 0.9865781962871552,
      "rewards-max/forensics_strict_edge_reward": 0.8553276658058167,
      "rewards-max/forensics_tiou_reward": 0.990521639585495,
      "rewards/forensics_sinkhorn_reward": 0.5221296846866608,
      "rewards/forensics_strict_edge_reward": 0.22021403908729553,
      "rewards/forensics_tiou_reward": 0.3110640123486519,
      "step": 1199
    },
    {
      "completion_length": 80.9375,
      "epoch": 2.510994764397906,
      "grad_norm": 21.87468910217285,
      "kl": 5.5390625,
      "learning_rate": 3.7159329140461216e-07,
      "loss": 0.4198,
      "reward": 0.6009677648544312,
      "reward_std": 0.2040087655186653,
      "reward_std_all": 0.2039216160774231,
      "rewards-max/forensics_sinkhorn_reward": 0.9879462718963623,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9958836436271667,
      "rewards/forensics_sinkhorn_reward": 0.49386273324489594,
      "rewards/forensics_strict_edge_reward": 0.26040390133857727,
      "rewards/forensics_tiou_reward": 0.29147475957870483,
      "step": 1200
    },
    {
      "completion_length": 33.375,
      "epoch": 2.513089005235602,
      "grad_norm": 21.348134994506836,
      "kl": 3.53125,
      "learning_rate": 3.7106918238993707e-07,
      "loss": 0.2484,
      "reward": 0.6773529648780823,
      "reward_std": 0.18704240769147873,
      "reward_std_all": 0.16986095160245895,
      "rewards-max/forensics_sinkhorn_reward": 0.98704993724823,
      "rewards-max/forensics_strict_edge_reward": 0.9464850425720215,
      "rewards-max/forensics_tiou_reward": 0.9923514127731323,
      "rewards/forensics_sinkhorn_reward": 0.6028636991977692,
      "rewards/forensics_strict_edge_reward": 0.4113699495792389,
      "rewards/forensics_tiou_reward": 0.4839200973510742,
      "step": 1201
    },
    {
      "completion_length": 43.796875,
      "epoch": 2.5151832460732986,
      "grad_norm": 24.758769989013672,
      "kl": 4.328125,
      "learning_rate": 3.70545073375262e-07,
      "loss": 0.4062,
      "reward": 0.6393206119537354,
      "reward_std": 0.19660823792219162,
      "reward_std_all": 0.184883251786232,
      "rewards-max/forensics_sinkhorn_reward": 0.9776987135410309,
      "rewards-max/forensics_strict_edge_reward": 0.905340164899826,
      "rewards-max/forensics_tiou_reward": 0.9721274971961975,
      "rewards/forensics_sinkhorn_reward": 0.525581419467926,
      "rewards/forensics_strict_edge_reward": 0.2794181555509567,
      "rewards/forensics_tiou_reward": 0.35591158270835876,
      "step": 1202
    },
    {
      "completion_length": 98.28125,
      "epoch": 2.5172774869109946,
      "grad_norm": 14.844841003417969,
      "kl": 7.21875,
      "learning_rate": 3.70020964360587e-07,
      "loss": 0.501,
      "reward": 0.5205984562635422,
      "reward_std": 0.2623475044965744,
      "reward_std_all": 0.2615876793861389,
      "rewards-max/forensics_sinkhorn_reward": 0.9858523607254028,
      "rewards-max/forensics_strict_edge_reward": 0.9364311099052429,
      "rewards-max/forensics_tiou_reward": 0.9876352250576019,
      "rewards/forensics_sinkhorn_reward": 0.43868987262248993,
      "rewards/forensics_strict_edge_reward": 0.2434873878955841,
      "rewards/forensics_tiou_reward": 0.30426596105098724,
      "step": 1203
    },
    {
      "completion_length": 63.703125,
      "epoch": 2.519371727748691,
      "grad_norm": 20.06438446044922,
      "kl": 6.296875,
      "learning_rate": 3.694968553459119e-07,
      "loss": 0.5489,
      "reward": 0.6229508221149445,
      "reward_std": 0.13869373500347137,
      "reward_std_all": 0.1624358743429184,
      "rewards-max/forensics_sinkhorn_reward": 0.9879546165466309,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.995920717716217,
      "rewards/forensics_sinkhorn_reward": 0.5671457648277283,
      "rewards/forensics_strict_edge_reward": 0.3508254587650299,
      "rewards/forensics_tiou_reward": 0.4387401193380356,
      "step": 1204
    },
    {
      "completion_length": 39.0,
      "epoch": 2.521465968586387,
      "grad_norm": 18.02562713623047,
      "kl": 5.015625,
      "learning_rate": 3.6897274633123685e-07,
      "loss": 0.3085,
      "reward": 0.6093287467956543,
      "reward_std": 0.19603891670703888,
      "reward_std_all": 0.19457345455884933,
      "rewards-max/forensics_sinkhorn_reward": 0.8867174088954926,
      "rewards-max/forensics_strict_edge_reward": 0.7866832911968231,
      "rewards-max/forensics_tiou_reward": 0.7645731866359711,
      "rewards/forensics_sinkhorn_reward": 0.48582063615322113,
      "rewards/forensics_strict_edge_reward": 0.22266234457492828,
      "rewards/forensics_tiou_reward": 0.24043558537960052,
      "step": 1205
    },
    {
      "completion_length": 62.03125,
      "epoch": 2.5235602094240837,
      "grad_norm": 25.596479415893555,
      "kl": 6.0,
      "learning_rate": 3.6844863731656186e-07,
      "loss": 0.4896,
      "reward": 0.6092592179775238,
      "reward_std": 0.22481010109186172,
      "reward_std_all": 0.21447620540857315,
      "rewards-max/forensics_sinkhorn_reward": 0.9708690643310547,
      "rewards-max/forensics_strict_edge_reward": 0.8968285918235779,
      "rewards-max/forensics_tiou_reward": 0.9602746367454529,
      "rewards/forensics_sinkhorn_reward": 0.5477548241615295,
      "rewards/forensics_strict_edge_reward": 0.2560628652572632,
      "rewards/forensics_tiou_reward": 0.38007429242134094,
      "step": 1206
    },
    {
      "completion_length": 50.921875,
      "epoch": 2.52565445026178,
      "grad_norm": 19.75487518310547,
      "kl": 6.265625,
      "learning_rate": 3.6792452830188677e-07,
      "loss": 0.4587,
      "reward": 0.5858227014541626,
      "reward_std": 0.24200007319450378,
      "reward_std_all": 0.23370875418186188,
      "rewards-max/forensics_sinkhorn_reward": 0.9634186625480652,
      "rewards-max/forensics_strict_edge_reward": 0.8983768224716187,
      "rewards-max/forensics_tiou_reward": 0.9451687633991241,
      "rewards/forensics_sinkhorn_reward": 0.48453284800052643,
      "rewards/forensics_strict_edge_reward": 0.24728615581989288,
      "rewards/forensics_tiou_reward": 0.30295833945274353,
      "step": 1207
    },
    {
      "completion_length": 51.96875,
      "epoch": 2.527748691099476,
      "grad_norm": 30.418699264526367,
      "kl": 7.125,
      "learning_rate": 3.674004192872117e-07,
      "loss": 0.3922,
      "reward": 0.5488406419754028,
      "reward_std": 0.236514113843441,
      "reward_std_all": 0.2394876927137375,
      "rewards-max/forensics_sinkhorn_reward": 0.9792092144489288,
      "rewards-max/forensics_strict_edge_reward": 0.9158510267734528,
      "rewards-max/forensics_tiou_reward": 0.9759107828140259,
      "rewards/forensics_sinkhorn_reward": 0.47898606956005096,
      "rewards/forensics_strict_edge_reward": 0.16361106932163239,
      "rewards/forensics_tiou_reward": 0.27515701949596405,
      "step": 1208
    },
    {
      "completion_length": 52.6875,
      "epoch": 2.5298429319371727,
      "grad_norm": 24.102739334106445,
      "kl": 4.9453125,
      "learning_rate": 3.668763102725367e-07,
      "loss": 0.3641,
      "reward": 0.5972012281417847,
      "reward_std": 0.21229156851768494,
      "reward_std_all": 0.20623933523893356,
      "rewards-max/forensics_sinkhorn_reward": 0.9485583007335663,
      "rewards-max/forensics_strict_edge_reward": 0.7807358801364899,
      "rewards-max/forensics_tiou_reward": 0.9099702537059784,
      "rewards/forensics_sinkhorn_reward": 0.49058589339256287,
      "rewards/forensics_strict_edge_reward": 0.24588056653738022,
      "rewards/forensics_tiou_reward": 0.33247077465057373,
      "step": 1209
    },
    {
      "completion_length": 74.265625,
      "epoch": 2.531937172774869,
      "grad_norm": 24.358135223388672,
      "kl": 10.4375,
      "learning_rate": 3.663522012578616e-07,
      "loss": 0.6076,
      "reward": 0.5595197081565857,
      "reward_std": 0.2345634549856186,
      "reward_std_all": 0.23947474360466003,
      "rewards-max/forensics_sinkhorn_reward": 0.9816701114177704,
      "rewards-max/forensics_strict_edge_reward": 0.9062341153621674,
      "rewards-max/forensics_tiou_reward": 0.9714884161949158,
      "rewards/forensics_sinkhorn_reward": 0.4527329355478287,
      "rewards/forensics_strict_edge_reward": 0.19591286033391953,
      "rewards/forensics_tiou_reward": 0.25760067999362946,
      "step": 1210
    },
    {
      "completion_length": 65.6875,
      "epoch": 2.5340314136125652,
      "grad_norm": 18.081754684448242,
      "kl": 6.4375,
      "learning_rate": 3.6582809224318655e-07,
      "loss": 0.4017,
      "reward": 0.6291607320308685,
      "reward_std": 0.18120046332478523,
      "reward_std_all": 0.17713894695043564,
      "rewards-max/forensics_sinkhorn_reward": 0.9873696267604828,
      "rewards-max/forensics_strict_edge_reward": 0.8915697336196899,
      "rewards-max/forensics_tiou_reward": 0.9936117827892303,
      "rewards/forensics_sinkhorn_reward": 0.5790619105100632,
      "rewards/forensics_strict_edge_reward": 0.3285900726914406,
      "rewards/forensics_tiou_reward": 0.4178582727909088,
      "step": 1211
    },
    {
      "completion_length": 89.734375,
      "epoch": 2.5361256544502617,
      "grad_norm": 18.606372833251953,
      "kl": 8.28125,
      "learning_rate": 3.653039832285115e-07,
      "loss": 0.5506,
      "reward": 0.553670734167099,
      "reward_std": 0.28226953744888306,
      "reward_std_all": 0.26754774153232574,
      "rewards-max/forensics_sinkhorn_reward": 0.9879408478736877,
      "rewards-max/forensics_strict_edge_reward": 0.9364311099052429,
      "rewards-max/forensics_tiou_reward": 0.995862603187561,
      "rewards/forensics_sinkhorn_reward": 0.46126262843608856,
      "rewards/forensics_strict_edge_reward": 0.28503671288490295,
      "rewards/forensics_tiou_reward": 0.3306756764650345,
      "step": 1212
    },
    {
      "completion_length": 87.765625,
      "epoch": 2.538219895287958,
      "grad_norm": 21.88634490966797,
      "kl": 9.4375,
      "learning_rate": 3.647798742138364e-07,
      "loss": 0.6163,
      "reward": 0.5743437111377716,
      "reward_std": 0.22573985904455185,
      "reward_std_all": 0.22461175173521042,
      "rewards-max/forensics_sinkhorn_reward": 0.9869220554828644,
      "rewards-max/forensics_strict_edge_reward": 0.946017861366272,
      "rewards-max/forensics_tiou_reward": 0.9918624758720398,
      "rewards/forensics_sinkhorn_reward": 0.50388503074646,
      "rewards/forensics_strict_edge_reward": 0.23491012305021286,
      "rewards/forensics_tiou_reward": 0.2987275570631027,
      "step": 1213
    },
    {
      "completion_length": 80.8125,
      "epoch": 2.5403141361256543,
      "grad_norm": 16.871156692504883,
      "kl": 6.53125,
      "learning_rate": 3.642557651991614e-07,
      "loss": 0.416,
      "reward": 0.5984601676464081,
      "reward_std": 0.17740297317504883,
      "reward_std_all": 0.2067197486758232,
      "rewards-max/forensics_sinkhorn_reward": 0.9886791110038757,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9987725019454956,
      "rewards/forensics_sinkhorn_reward": 0.5204092711210251,
      "rewards/forensics_strict_edge_reward": 0.30877967178821564,
      "rewards/forensics_tiou_reward": 0.3917538821697235,
      "step": 1214
    },
    {
      "completion_length": 98.203125,
      "epoch": 2.5424083769633508,
      "grad_norm": 20.37884521484375,
      "kl": 8.453125,
      "learning_rate": 3.637316561844864e-07,
      "loss": 0.5531,
      "reward": 0.5306174159049988,
      "reward_std": 0.24233487993478775,
      "reward_std_all": 0.2574499100446701,
      "rewards-max/forensics_sinkhorn_reward": 0.9573160409927368,
      "rewards-max/forensics_strict_edge_reward": 0.8355957567691803,
      "rewards-max/forensics_tiou_reward": 0.9325156509876251,
      "rewards/forensics_sinkhorn_reward": 0.46973922848701477,
      "rewards/forensics_strict_edge_reward": 0.2512223422527313,
      "rewards/forensics_tiou_reward": 0.3732926547527313,
      "step": 1215
    },
    {
      "completion_length": 55.9375,
      "epoch": 2.5445026178010473,
      "grad_norm": 17.204729080200195,
      "kl": 6.171875,
      "learning_rate": 3.632075471698113e-07,
      "loss": 0.3969,
      "reward": 0.5948110222816467,
      "reward_std": 0.11243988946080208,
      "reward_std_all": 0.15425966680049896,
      "rewards-max/forensics_sinkhorn_reward": 0.988743245601654,
      "rewards-max/forensics_strict_edge_reward": 0.9890114367008209,
      "rewards-max/forensics_tiou_reward": 0.999025821685791,
      "rewards/forensics_sinkhorn_reward": 0.5281867384910583,
      "rewards/forensics_strict_edge_reward": 0.3502536118030548,
      "rewards/forensics_tiou_reward": 0.38050632178783417,
      "step": 1216
    },
    {
      "completion_length": 91.546875,
      "epoch": 2.5465968586387433,
      "grad_norm": 21.70318603515625,
      "kl": 8.296875,
      "learning_rate": 3.6268343815513625e-07,
      "loss": 0.3633,
      "reward": 0.5419908761978149,
      "reward_std": 0.2092052772641182,
      "reward_std_all": 0.2346128523349762,
      "rewards-max/forensics_sinkhorn_reward": 0.9707611501216888,
      "rewards-max/forensics_strict_edge_reward": 0.8957232534885406,
      "rewards-max/forensics_tiou_reward": 0.9574176967144012,
      "rewards/forensics_sinkhorn_reward": 0.405648335814476,
      "rewards/forensics_strict_edge_reward": 0.17997893691062927,
      "rewards/forensics_tiou_reward": 0.2063157707452774,
      "step": 1217
    },
    {
      "completion_length": 139.4375,
      "epoch": 2.54869109947644,
      "grad_norm": 22.378271102905273,
      "kl": 10.625,
      "learning_rate": 3.621593291404612e-07,
      "loss": 0.6665,
      "reward": 0.5217956155538559,
      "reward_std": 0.18456945195794106,
      "reward_std_all": 0.23036862909793854,
      "rewards-max/forensics_sinkhorn_reward": 0.9883905947208405,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.997635155916214,
      "rewards/forensics_sinkhorn_reward": 0.48645392060279846,
      "rewards/forensics_strict_edge_reward": 0.3122949153184891,
      "rewards/forensics_tiou_reward": 0.3798424154520035,
      "step": 1218
    },
    {
      "completion_length": 104.71875,
      "epoch": 2.5507853403141363,
      "grad_norm": 38.299842834472656,
      "kl": 7.515625,
      "learning_rate": 3.616352201257861e-07,
      "loss": 0.4786,
      "reward": 0.5207718312740326,
      "reward_std": 0.21173956990242004,
      "reward_std_all": 0.24090750515460968,
      "rewards-max/forensics_sinkhorn_reward": 0.8920832276344299,
      "rewards-max/forensics_strict_edge_reward": 0.6934256851673126,
      "rewards-max/forensics_tiou_reward": 0.7939923405647278,
      "rewards/forensics_sinkhorn_reward": 0.3943530023097992,
      "rewards/forensics_strict_edge_reward": 0.1714814230799675,
      "rewards/forensics_tiou_reward": 0.18391767144203186,
      "step": 1219
    },
    {
      "completion_length": 108.625,
      "epoch": 2.5528795811518323,
      "grad_norm": 20.55086898803711,
      "kl": 6.0,
      "learning_rate": 3.6111111111111107e-07,
      "loss": 0.4033,
      "reward": 0.5738863050937653,
      "reward_std": 0.2579878866672516,
      "reward_std_all": 0.24493750929832458,
      "rewards-max/forensics_sinkhorn_reward": 0.9885718822479248,
      "rewards-max/forensics_strict_edge_reward": 0.9890114367008209,
      "rewards-max/forensics_tiou_reward": 0.9983511865139008,
      "rewards/forensics_sinkhorn_reward": 0.48080991208553314,
      "rewards/forensics_strict_edge_reward": 0.2758997976779938,
      "rewards/forensics_tiou_reward": 0.356775164604187,
      "step": 1220
    },
    {
      "completion_length": 114.203125,
      "epoch": 2.554973821989529,
      "grad_norm": 24.816394805908203,
      "kl": 8.125,
      "learning_rate": 3.6058700209643603e-07,
      "loss": 0.5125,
      "reward": 0.5230138897895813,
      "reward_std": 0.2824143022298813,
      "reward_std_all": 0.2757539823651314,
      "rewards-max/forensics_sinkhorn_reward": 0.9519762098789215,
      "rewards-max/forensics_strict_edge_reward": 0.8578422963619232,
      "rewards-max/forensics_tiou_reward": 0.9213186204433441,
      "rewards/forensics_sinkhorn_reward": 0.41009965538978577,
      "rewards/forensics_strict_edge_reward": 0.2336791232228279,
      "rewards/forensics_tiou_reward": 0.2713053822517395,
      "step": 1221
    },
    {
      "completion_length": 127.265625,
      "epoch": 2.5570680628272253,
      "grad_norm": 24.178518295288086,
      "kl": 7.15625,
      "learning_rate": 3.60062893081761e-07,
      "loss": 0.5347,
      "reward": 0.5124701261520386,
      "reward_std": 0.19611679762601852,
      "reward_std_all": 0.24102488905191422,
      "rewards-max/forensics_sinkhorn_reward": 0.982099860906601,
      "rewards-max/forensics_strict_edge_reward": 0.9256840348243713,
      "rewards-max/forensics_tiou_reward": 0.9731373190879822,
      "rewards/forensics_sinkhorn_reward": 0.4217745363712311,
      "rewards/forensics_strict_edge_reward": 0.23339763283729553,
      "rewards/forensics_tiou_reward": 0.3007354214787483,
      "step": 1222
    },
    {
      "completion_length": 82.1875,
      "epoch": 2.5591623036649214,
      "grad_norm": 15.24878215789795,
      "kl": 5.203125,
      "learning_rate": 3.5953878406708595e-07,
      "loss": 0.2985,
      "reward": 0.5914103090763092,
      "reward_std": 0.1917724460363388,
      "reward_std_all": 0.19863354414701462,
      "rewards-max/forensics_sinkhorn_reward": 0.9883290529251099,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9973942637443542,
      "rewards/forensics_sinkhorn_reward": 0.5167109370231628,
      "rewards/forensics_strict_edge_reward": 0.23936571180820465,
      "rewards/forensics_tiou_reward": 0.3369234949350357,
      "step": 1223
    },
    {
      "completion_length": 111.546875,
      "epoch": 2.561256544502618,
      "grad_norm": 19.45612335205078,
      "kl": 7.0625,
      "learning_rate": 3.590146750524109e-07,
      "loss": 0.5589,
      "reward": 0.508268341422081,
      "reward_std": 0.26590219140052795,
      "reward_std_all": 0.2893649563193321,
      "rewards-max/forensics_sinkhorn_reward": 0.9879748523235321,
      "rewards-max/forensics_strict_edge_reward": 0.946017861366272,
      "rewards-max/forensics_tiou_reward": 0.9960010051727295,
      "rewards/forensics_sinkhorn_reward": 0.4054865837097168,
      "rewards/forensics_strict_edge_reward": 0.26844992488622665,
      "rewards/forensics_tiou_reward": 0.26389723271131516,
      "step": 1224
    },
    {
      "completion_length": 109.96875,
      "epoch": 2.5633507853403144,
      "grad_norm": 21.729101181030273,
      "kl": 7.0,
      "learning_rate": 3.584905660377358e-07,
      "loss": 0.4291,
      "reward": 0.5161105990409851,
      "reward_std": 0.2549189403653145,
      "reward_std_all": 0.26915687322616577,
      "rewards-max/forensics_sinkhorn_reward": 0.9883535802364349,
      "rewards-max/forensics_strict_edge_reward": 0.9464850425720215,
      "rewards-max/forensics_tiou_reward": 0.9974893033504486,
      "rewards/forensics_sinkhorn_reward": 0.3678300082683563,
      "rewards/forensics_strict_edge_reward": 0.2176513820886612,
      "rewards/forensics_tiou_reward": 0.23823705315589905,
      "step": 1225
    },
    {
      "completion_length": 77.984375,
      "epoch": 2.5654450261780104,
      "grad_norm": 33.179290771484375,
      "kl": 7.484375,
      "learning_rate": 3.5796645702306077e-07,
      "loss": 0.45,
      "reward": 0.5690670013427734,
      "reward_std": 0.21799816191196442,
      "reward_std_all": 0.22340981662273407,
      "rewards-max/forensics_sinkhorn_reward": 0.9742459058761597,
      "rewards-max/forensics_strict_edge_reward": 0.8983768224716187,
      "rewards-max/forensics_tiou_reward": 0.9651688039302826,
      "rewards/forensics_sinkhorn_reward": 0.4872948080301285,
      "rewards/forensics_strict_edge_reward": 0.24257422238588333,
      "rewards/forensics_tiou_reward": 0.32094424962997437,
      "step": 1226
    },
    {
      "completion_length": 83.875,
      "epoch": 2.567539267015707,
      "grad_norm": 20.589250564575195,
      "kl": 7.234375,
      "learning_rate": 3.5744234800838573e-07,
      "loss": 0.4645,
      "reward": 0.5176877379417419,
      "reward_std": 0.301319420337677,
      "reward_std_all": 0.29240545630455017,
      "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": 0.46623294055461884,
      "rewards/forensics_strict_edge_reward": 0.2405804991722107,
      "rewards/forensics_tiou_reward": 0.31993862986564636,
      "step": 1227
    },
    {
      "completion_length": 112.671875,
      "epoch": 2.5696335078534034,
      "grad_norm": 28.122333526611328,
      "kl": 7.78125,
      "learning_rate": 3.5691823899371064e-07,
      "loss": 0.4651,
      "reward": 0.5007597655057907,
      "reward_std": 0.21798840910196304,
      "reward_std_all": 0.25538745522499084,
      "rewards-max/forensics_sinkhorn_reward": 0.9883950054645538,
      "rewards-max/forensics_strict_edge_reward": 0.9890114367008209,
      "rewards-max/forensics_tiou_reward": 0.9976553320884705,
      "rewards/forensics_sinkhorn_reward": 0.4181196689605713,
      "rewards/forensics_strict_edge_reward": 0.28186365962028503,
      "rewards/forensics_tiou_reward": 0.3160294443368912,
      "step": 1228
    },
    {
      "completion_length": 59.21875,
      "epoch": 2.5717277486910994,
      "grad_norm": 15.320289611816406,
      "kl": 5.40625,
      "learning_rate": 3.563941299790356e-07,
      "loss": 0.3441,
      "reward": 0.5664918422698975,
      "reward_std": 0.20368914306163788,
      "reward_std_all": 0.2220754250884056,
      "rewards-max/forensics_sinkhorn_reward": 0.9861707389354706,
      "rewards-max/forensics_strict_edge_reward": 0.947419673204422,
      "rewards-max/forensics_tiou_reward": 0.9889506101608276,
      "rewards/forensics_sinkhorn_reward": 0.4936292767524719,
      "rewards/forensics_strict_edge_reward": 0.32806992530822754,
      "rewards/forensics_tiou_reward": 0.3745678961277008,
      "step": 1229
    },
    {
      "completion_length": 80.546875,
      "epoch": 2.573821989528796,
      "grad_norm": 47.97762680053711,
      "kl": 6.296875,
      "learning_rate": 3.558700209643606e-07,
      "loss": 0.3537,
      "reward": 0.593235969543457,
      "reward_std": 0.18740924075245857,
      "reward_std_all": 0.1983785480260849,
      "rewards-max/forensics_sinkhorn_reward": 0.9666005373001099,
      "rewards-max/forensics_strict_edge_reward": 0.9185642302036285,
      "rewards-max/forensics_tiou_reward": 0.948058545589447,
      "rewards/forensics_sinkhorn_reward": 0.47507548332214355,
      "rewards/forensics_strict_edge_reward": 0.2925109192728996,
      "rewards/forensics_tiou_reward": 0.3333883583545685,
      "step": 1230
    },
    {
      "completion_length": 67.21875,
      "epoch": 2.5759162303664924,
      "grad_norm": 22.494136810302734,
      "kl": 7.296875,
      "learning_rate": 3.553459119496855e-07,
      "loss": 0.4923,
      "reward": 0.49065305292606354,
      "reward_std": 0.18205126374959946,
      "reward_std_all": 0.24306753277778625,
      "rewards-max/forensics_sinkhorn_reward": 0.9853531718254089,
      "rewards-max/forensics_strict_edge_reward": 0.946017861366272,
      "rewards-max/forensics_tiou_reward": 0.9857021570205688,
      "rewards/forensics_sinkhorn_reward": 0.4262390434741974,
      "rewards/forensics_strict_edge_reward": 0.19739412516355515,
      "rewards/forensics_tiou_reward": 0.2598745971918106,
      "step": 1231
    },
    {
      "completion_length": 121.625,
      "epoch": 2.5780104712041885,
      "grad_norm": 24.920392990112305,
      "kl": 10.3125,
      "learning_rate": 3.5482180293501047e-07,
      "loss": 0.6987,
      "reward": 0.5179228782653809,
      "reward_std": 0.20619191229343414,
      "reward_std_all": 0.24782877415418625,
      "rewards-max/forensics_sinkhorn_reward": 0.9871909618377686,
      "rewards-max/forensics_strict_edge_reward": 0.9574736058712006,
      "rewards-max/forensics_tiou_reward": 0.9929329454898834,
      "rewards/forensics_sinkhorn_reward": 0.45185594260692596,
      "rewards/forensics_strict_edge_reward": 0.2716040313243866,
      "rewards/forensics_tiou_reward": 0.32963719964027405,
      "step": 1232
    },
    {
      "completion_length": 79.796875,
      "epoch": 2.580104712041885,
      "grad_norm": 22.200674057006836,
      "kl": 9.09375,
      "learning_rate": 3.5429769392033543e-07,
      "loss": 0.6028,
      "reward": 0.55750972032547,
      "reward_std": 0.2978934347629547,
      "reward_std_all": 0.27689461410045624,
      "rewards-max/forensics_sinkhorn_reward": 0.9887057840824127,
      "rewards-max/forensics_strict_edge_reward": 0.9782643616199493,
      "rewards-max/forensics_tiou_reward": 0.9988783001899719,
      "rewards/forensics_sinkhorn_reward": 0.5205007195472717,
      "rewards/forensics_strict_edge_reward": 0.4107557088136673,
      "rewards/forensics_tiou_reward": 0.43718643486499786,
      "step": 1233
    },
    {
      "completion_length": 94.203125,
      "epoch": 2.582198952879581,
      "grad_norm": 31.40544891357422,
      "kl": 8.203125,
      "learning_rate": 3.5377358490566033e-07,
      "loss": 0.5777,
      "reward": 0.5554779171943665,
      "reward_std": 0.18907011300325394,
      "reward_std_all": 0.2180519625544548,
      "rewards-max/forensics_sinkhorn_reward": 0.9635681211948395,
      "rewards-max/forensics_strict_edge_reward": 0.8698340952396393,
      "rewards-max/forensics_tiou_reward": 0.9429433941841125,
      "rewards/forensics_sinkhorn_reward": 0.44929562509059906,
      "rewards/forensics_strict_edge_reward": 0.21182816475629807,
      "rewards/forensics_tiou_reward": 0.29993098974227905,
      "step": 1234
    },
    {
      "completion_length": 67.828125,
      "epoch": 2.5842931937172775,
      "grad_norm": 23.922990798950195,
      "kl": 8.234375,
      "learning_rate": 3.532494758909853e-07,
      "loss": 0.5175,
      "reward": 0.5844087600708008,
      "reward_std": 0.21263347566127777,
      "reward_std_all": 0.2197437360882759,
      "rewards-max/forensics_sinkhorn_reward": 0.9884568452835083,
      "rewards-max/forensics_strict_edge_reward": 0.9565287232398987,
      "rewards-max/forensics_tiou_reward": 0.9978968799114227,
      "rewards/forensics_sinkhorn_reward": 0.5374895632266998,
      "rewards/forensics_strict_edge_reward": 0.29381901025772095,
      "rewards/forensics_tiou_reward": 0.38980674743652344,
      "step": 1235
    },
    {
      "completion_length": 89.953125,
      "epoch": 2.5863874345549736,
      "grad_norm": 29.536542892456055,
      "kl": 7.265625,
      "learning_rate": 3.5272536687631025e-07,
      "loss": 0.5366,
      "reward": 0.5574192702770233,
      "reward_std": 0.202636256814003,
      "reward_std_all": 0.22051507979631424,
      "rewards-max/forensics_sinkhorn_reward": 0.987560510635376,
      "rewards-max/forensics_strict_edge_reward": 0.927969753742218,
      "rewards-max/forensics_tiou_reward": 0.9943798184394836,
      "rewards/forensics_sinkhorn_reward": 0.4577566534280777,
      "rewards/forensics_strict_edge_reward": 0.279746413230896,
      "rewards/forensics_tiou_reward": 0.3335610628128052,
      "step": 1236
    },
    {
      "completion_length": 73.875,
      "epoch": 2.58848167539267,
      "grad_norm": 26.736949920654297,
      "kl": 7.859375,
      "learning_rate": 3.5220125786163516e-07,
      "loss": 0.4951,
      "reward": 0.6413305401802063,
      "reward_std": 0.18104633688926697,
      "reward_std_all": 0.17847250401973724,
      "rewards-max/forensics_sinkhorn_reward": 0.9882451593875885,
      "rewards-max/forensics_strict_edge_reward": 0.9567649364471436,
      "rewards-max/forensics_tiou_reward": 0.9970638751983643,
      "rewards/forensics_sinkhorn_reward": 0.5970941185951233,
      "rewards/forensics_strict_edge_reward": 0.39607132971286774,
      "rewards/forensics_tiou_reward": 0.4324878454208374,
      "step": 1237
    },
    {
      "completion_length": 71.4375,
      "epoch": 2.5905759162303665,
      "grad_norm": 17.08641815185547,
      "kl": 7.09375,
      "learning_rate": 3.5167714884696017e-07,
      "loss": 0.4959,
      "reward": 0.6390206515789032,
      "reward_std": 0.18777363747358322,
      "reward_std_all": 0.1778058409690857,
      "rewards-max/forensics_sinkhorn_reward": 0.9815438687801361,
      "rewards-max/forensics_strict_edge_reward": 0.8365300297737122,
      "rewards-max/forensics_tiou_reward": 0.9709928631782532,
      "rewards/forensics_sinkhorn_reward": 0.560016468167305,
      "rewards/forensics_strict_edge_reward": 0.28001271188259125,
      "rewards/forensics_tiou_reward": 0.4056102931499481,
      "step": 1238
    },
    {
      "completion_length": 88.671875,
      "epoch": 2.5926701570680626,
      "grad_norm": 18.244998931884766,
      "kl": 9.609375,
      "learning_rate": 3.5115303983228513e-07,
      "loss": 0.5308,
      "reward": 0.5356300920248032,
      "reward_std": 0.22450587898492813,
      "reward_std_all": 0.24476298689842224,
      "rewards-max/forensics_sinkhorn_reward": 0.9568423628807068,
      "rewards-max/forensics_strict_edge_reward": 0.8286563754081726,
      "rewards-max/forensics_tiou_reward": 0.9232126474380493,
      "rewards/forensics_sinkhorn_reward": 0.43988001346588135,
      "rewards/forensics_strict_edge_reward": 0.19005443900823593,
      "rewards/forensics_tiou_reward": 0.2347036525607109,
      "step": 1239
    },
    {
      "completion_length": 55.3125,
      "epoch": 2.594764397905759,
      "grad_norm": 24.00920295715332,
      "kl": 5.375,
      "learning_rate": 3.5062893081761003e-07,
      "loss": 0.4852,
      "reward": 0.6492556035518646,
      "reward_std": 0.1881389245390892,
      "reward_std_all": 0.1767161265015602,
      "rewards-max/forensics_sinkhorn_reward": 0.9559738039970398,
      "rewards-max/forensics_strict_edge_reward": 0.8348778784275055,
      "rewards-max/forensics_tiou_reward": 0.9264995753765106,
      "rewards/forensics_sinkhorn_reward": 0.5559793263673782,
      "rewards/forensics_strict_edge_reward": 0.3258248269557953,
      "rewards/forensics_tiou_reward": 0.37850579619407654,
      "step": 1240
    },
    {
      "completion_length": 58.40625,
      "epoch": 2.5968586387434556,
      "grad_norm": 25.908597946166992,
      "kl": 5.34375,
      "learning_rate": 3.50104821802935e-07,
      "loss": 0.3236,
      "reward": 0.5820081532001495,
      "reward_std": 0.23254981637001038,
      "reward_std_all": 0.2330646887421608,
      "rewards-max/forensics_sinkhorn_reward": 0.965835839509964,
      "rewards-max/forensics_strict_edge_reward": 0.8659839034080505,
      "rewards-max/forensics_tiou_reward": 0.937386155128479,
      "rewards/forensics_sinkhorn_reward": 0.4908915013074875,
      "rewards/forensics_strict_edge_reward": 0.20174506306648254,
      "rewards/forensics_tiou_reward": 0.2794041186571121,
      "step": 1241
    },
    {
      "completion_length": 50.171875,
      "epoch": 2.5989528795811516,
      "grad_norm": 17.430145263671875,
      "kl": 4.8828125,
      "learning_rate": 3.4958071278825995e-07,
      "loss": 0.394,
      "reward": 0.6087310314178467,
      "reward_std": 0.1415572538971901,
      "reward_std_all": 0.16020194441080093,
      "rewards-max/forensics_sinkhorn_reward": 0.9875659644603729,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9943926930427551,
      "rewards/forensics_sinkhorn_reward": 0.5409113764762878,
      "rewards/forensics_strict_edge_reward": 0.31511247158050537,
      "rewards/forensics_tiou_reward": 0.35664939880371094,
      "step": 1242
    },
    {
      "completion_length": 56.8125,
      "epoch": 2.601047120418848,
      "grad_norm": 17.86470603942871,
      "kl": 7.125,
      "learning_rate": 3.4905660377358486e-07,
      "loss": 0.4966,
      "reward": 0.6184073984622955,
      "reward_std": 0.22269275784492493,
      "reward_std_all": 0.21052907407283783,
      "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": 0.548532634973526,
      "rewards/forensics_strict_edge_reward": 0.339164063334465,
      "rewards/forensics_tiou_reward": 0.39377541840076447,
      "step": 1243
    },
    {
      "completion_length": 56.890625,
      "epoch": 2.6031413612565446,
      "grad_norm": 19.390363693237305,
      "kl": 5.8828125,
      "learning_rate": 3.485324947589098e-07,
      "loss": 0.4778,
      "reward": 0.6104201972484589,
      "reward_std": 0.2261485457420349,
      "reward_std_all": 0.21315031498670578,
      "rewards-max/forensics_sinkhorn_reward": 0.9884685277938843,
      "rewards-max/forensics_strict_edge_reward": 0.9890114367008209,
      "rewards-max/forensics_tiou_reward": 0.9979445040225983,
      "rewards/forensics_sinkhorn_reward": 0.5324276983737946,
      "rewards/forensics_strict_edge_reward": 0.3314647972583771,
      "rewards/forensics_tiou_reward": 0.36838240921497345,
      "step": 1244
    },
    {
      "completion_length": 42.265625,
      "epoch": 2.6052356020942407,
      "grad_norm": 16.5644474029541,
      "kl": 4.265625,
      "learning_rate": 3.480083857442348e-07,
      "loss": 0.3556,
      "reward": 0.6452324986457825,
      "reward_std": 0.1619025021791458,
      "reward_std_all": 0.16319236159324646,
      "rewards-max/forensics_sinkhorn_reward": 0.9695079624652863,
      "rewards-max/forensics_strict_edge_reward": 0.9093653857707977,
      "rewards-max/forensics_tiou_reward": 0.9557092189788818,
      "rewards/forensics_sinkhorn_reward": 0.5671239197254181,
      "rewards/forensics_strict_edge_reward": 0.2792022004723549,
      "rewards/forensics_tiou_reward": 0.4023246318101883,
      "step": 1245
    },
    {
      "completion_length": 38.09375,
      "epoch": 2.607329842931937,
      "grad_norm": 26.473175048828125,
      "kl": 4.578125,
      "learning_rate": 3.4748427672955973e-07,
      "loss": 0.3714,
      "reward": 0.6408407092094421,
      "reward_std": 0.14111047983169556,
      "reward_std_all": 0.16007331758737564,
      "rewards-max/forensics_sinkhorn_reward": 0.9874671995639801,
      "rewards-max/forensics_strict_edge_reward": 0.9565287232398987,
      "rewards-max/forensics_tiou_reward": 0.9939984381198883,
      "rewards/forensics_sinkhorn_reward": 0.5557152628898621,
      "rewards/forensics_strict_edge_reward": 0.33708956837654114,
      "rewards/forensics_tiou_reward": 0.3918236643075943,
      "step": 1246
    },
    {
      "completion_length": 48.5,
      "epoch": 2.6094240837696336,
      "grad_norm": 19.621427536010742,
      "kl": 7.1875,
      "learning_rate": 3.469601677148847e-07,
      "loss": 0.3759,
      "reward": 0.5693824887275696,
      "reward_std": 0.19617243111133575,
      "reward_std_all": 0.2072189599275589,
      "rewards-max/forensics_sinkhorn_reward": 0.8529216349124908,
      "rewards-max/forensics_strict_edge_reward": 0.6798972487449646,
      "rewards-max/forensics_tiou_reward": 0.7211716622114182,
      "rewards/forensics_sinkhorn_reward": 0.47928497195243835,
      "rewards/forensics_strict_edge_reward": 0.22177522629499435,
      "rewards/forensics_tiou_reward": 0.2909240201115608,
      "step": 1247
    },
    {
      "completion_length": 44.90625,
      "epoch": 2.6115183246073297,
      "grad_norm": 19.449337005615234,
      "kl": 6.21875,
      "learning_rate": 3.4643605870020965e-07,
      "loss": 0.3822,
      "reward": 0.6288270652294159,
      "reward_std": 0.20602940022945404,
      "reward_std_all": 0.19364851713180542,
      "rewards-max/forensics_sinkhorn_reward": 0.9586383700370789,
      "rewards-max/forensics_strict_edge_reward": 0.7933230996131897,
      "rewards-max/forensics_tiou_reward": 0.933355838060379,
      "rewards/forensics_sinkhorn_reward": 0.5616990327835083,
      "rewards/forensics_strict_edge_reward": 0.30078161880373955,
      "rewards/forensics_tiou_reward": 0.37680791318416595,
      "step": 1248
    },
    {
      "completion_length": 50.34375,
      "epoch": 2.613612565445026,
      "grad_norm": 18.710617065429688,
      "kl": 6.5,
      "learning_rate": 3.4591194968553456e-07,
      "loss": 0.4337,
      "reward": 0.5984080731868744,
      "reward_std": 0.205111064016819,
      "reward_std_all": 0.2083391770720482,
      "rewards-max/forensics_sinkhorn_reward": 0.9701376557350159,
      "rewards-max/forensics_strict_edge_reward": 0.9574736058712006,
      "rewards-max/forensics_tiou_reward": 0.9597141146659851,
      "rewards/forensics_sinkhorn_reward": 0.5426880121231079,
      "rewards/forensics_strict_edge_reward": 0.26240406930446625,
      "rewards/forensics_tiou_reward": 0.33842483162879944,
      "step": 1249
    },
    {
      "completion_length": 34.6875,
      "epoch": 2.6157068062827227,
      "grad_norm": 15.034059524536133,
      "kl": 4.734375,
      "learning_rate": 3.453878406708595e-07,
      "loss": 0.3105,
      "reward": 0.6174344420433044,
      "reward_std": 0.09196992218494415,
      "reward_std_all": 0.14204108715057373,
      "rewards-max/forensics_sinkhorn_reward": 0.9662360846996307,
      "rewards-max/forensics_strict_edge_reward": 0.8394034504890442,
      "rewards-max/forensics_tiou_reward": 0.9462952613830566,
      "rewards/forensics_sinkhorn_reward": 0.5334819853305817,
      "rewards/forensics_strict_edge_reward": 0.29059287160634995,
      "rewards/forensics_tiou_reward": 0.3462473154067993,
      "step": 1250
    },
    {
      "completion_length": 64.203125,
      "epoch": 2.6178010471204187,
      "grad_norm": 19.098674774169922,
      "kl": 8.09375,
      "learning_rate": 3.448637316561845e-07,
      "loss": 0.5834,
      "reward": 0.6446086168289185,
      "reward_std": 0.21729791164398193,
      "reward_std_all": 0.19745223224163055,
      "rewards-max/forensics_sinkhorn_reward": 0.987061083316803,
      "rewards-max/forensics_strict_edge_reward": 0.8876297473907471,
      "rewards-max/forensics_tiou_reward": 0.9924112260341644,
      "rewards/forensics_sinkhorn_reward": 0.5415572226047516,
      "rewards/forensics_strict_edge_reward": 0.27480171620845795,
      "rewards/forensics_tiou_reward": 0.37748056650161743,
      "step": 1251
    },
    {
      "completion_length": 47.515625,
      "epoch": 2.619895287958115,
      "grad_norm": 26.03034210205078,
      "kl": 6.484375,
      "learning_rate": 3.443396226415094e-07,
      "loss": 0.4768,
      "reward": 0.5997948050498962,
      "reward_std": 0.24327701330184937,
      "reward_std_all": 0.23070194572210312,
      "rewards-max/forensics_sinkhorn_reward": 0.9871731698513031,
      "rewards-max/forensics_strict_edge_reward": 0.9565287232398987,
      "rewards-max/forensics_tiou_reward": 0.992836594581604,
      "rewards/forensics_sinkhorn_reward": 0.540298730134964,
      "rewards/forensics_strict_edge_reward": 0.28242821991443634,
      "rewards/forensics_tiou_reward": 0.3546317517757416,
      "step": 1252
    },
    {
      "completion_length": 53.109375,
      "epoch": 2.6219895287958117,
      "grad_norm": 23.36463165283203,
      "kl": 6.46875,
      "learning_rate": 3.4381551362683434e-07,
      "loss": 0.6025,
      "reward": 0.5921585559844971,
      "reward_std": 0.1944163590669632,
      "reward_std_all": 0.20425160229206085,
      "rewards-max/forensics_sinkhorn_reward": 0.9752831757068634,
      "rewards-max/forensics_strict_edge_reward": 0.7557639181613922,
      "rewards-max/forensics_tiou_reward": 0.9670019447803497,
      "rewards/forensics_sinkhorn_reward": 0.49141958355903625,
      "rewards/forensics_strict_edge_reward": 0.21936355531215668,
      "rewards/forensics_tiou_reward": 0.29564401507377625,
      "step": 1253
    },
    {
      "completion_length": 67.90625,
      "epoch": 2.6240837696335078,
      "grad_norm": 21.681156158447266,
      "kl": 7.828125,
      "learning_rate": 3.4329140461215935e-07,
      "loss": 0.5878,
      "reward": 0.6059839129447937,
      "reward_std": 0.20826204866170883,
      "reward_std_all": 0.21821600198745728,
      "rewards-max/forensics_sinkhorn_reward": 0.9883412420749664,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.997442215681076,
      "rewards/forensics_sinkhorn_reward": 0.5661620497703552,
      "rewards/forensics_strict_edge_reward": 0.42484065890312195,
      "rewards/forensics_tiou_reward": 0.4665292948484421,
      "step": 1254
    },
    {
      "completion_length": 56.921875,
      "epoch": 2.6261780104712042,
      "grad_norm": 16.699718475341797,
      "kl": 5.8984375,
      "learning_rate": 3.4276729559748426e-07,
      "loss": 0.3564,
      "reward": 0.6742988526821136,
      "reward_std": 0.14103231206536293,
      "reward_std_all": 0.1404084712266922,
      "rewards-max/forensics_sinkhorn_reward": 0.9877608120441437,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9951553344726562,
      "rewards/forensics_sinkhorn_reward": 0.6201274991035461,
      "rewards/forensics_strict_edge_reward": 0.3910636603832245,
      "rewards/forensics_tiou_reward": 0.4740491807460785,
      "step": 1255
    },
    {
      "completion_length": 48.75,
      "epoch": 2.6282722513089007,
      "grad_norm": 25.345470428466797,
      "kl": 9.328125,
      "learning_rate": 3.422431865828092e-07,
      "loss": 0.5553,
      "reward": 0.5053954720497131,
      "reward_std": 0.21925204992294312,
      "reward_std_all": 0.2544980049133301,
      "rewards-max/forensics_sinkhorn_reward": 0.9856393933296204,
      "rewards-max/forensics_strict_edge_reward": 0.8668389916419983,
      "rewards-max/forensics_tiou_reward": 0.9868683815002441,
      "rewards/forensics_sinkhorn_reward": 0.4328960031270981,
      "rewards/forensics_strict_edge_reward": 0.22789104282855988,
      "rewards/forensics_tiou_reward": 0.30421485751867294,
      "step": 1256
    },
    {
      "completion_length": 76.671875,
      "epoch": 2.630366492146597,
      "grad_norm": 23.376941680908203,
      "kl": 8.984375,
      "learning_rate": 3.4171907756813417e-07,
      "loss": 0.6193,
      "reward": 0.531199499964714,
      "reward_std": 0.2170112207531929,
      "reward_std_all": 0.23871779441833496,
      "rewards-max/forensics_sinkhorn_reward": 0.9788262844085693,
      "rewards-max/forensics_strict_edge_reward": 0.9677534997463226,
      "rewards-max/forensics_tiou_reward": 0.9774467945098877,
      "rewards/forensics_sinkhorn_reward": 0.4206540435552597,
      "rewards/forensics_strict_edge_reward": 0.20056022703647614,
      "rewards/forensics_tiou_reward": 0.23528511077165604,
      "step": 1257
    },
    {
      "completion_length": 58.8125,
      "epoch": 2.6324607329842933,
      "grad_norm": 26.561403274536133,
      "kl": 9.0,
      "learning_rate": 3.411949685534591e-07,
      "loss": 0.6027,
      "reward": 0.5606471002101898,
      "reward_std": 0.17523977905511856,
      "reward_std_all": 0.20310338586568832,
      "rewards-max/forensics_sinkhorn_reward": 0.9688045382499695,
      "rewards-max/forensics_strict_edge_reward": 0.8181455135345459,
      "rewards-max/forensics_tiou_reward": 0.9564363062381744,
      "rewards/forensics_sinkhorn_reward": 0.4272339642047882,
      "rewards/forensics_strict_edge_reward": 0.19326484203338623,
      "rewards/forensics_tiou_reward": 0.2527475878596306,
      "step": 1258
    },
    {
      "completion_length": 35.859375,
      "epoch": 2.6345549738219898,
      "grad_norm": 20.666038513183594,
      "kl": 5.5625,
      "learning_rate": 3.4067085953878404e-07,
      "loss": 0.3811,
      "reward": 0.5868399739265442,
      "reward_std": 0.19360297918319702,
      "reward_std_all": 0.2001684159040451,
      "rewards-max/forensics_sinkhorn_reward": 0.9274708926677704,
      "rewards-max/forensics_strict_edge_reward": 0.7926336228847504,
      "rewards-max/forensics_tiou_reward": 0.8619591295719147,
      "rewards/forensics_sinkhorn_reward": 0.4827524870634079,
      "rewards/forensics_strict_edge_reward": 0.19001302123069763,
      "rewards/forensics_tiou_reward": 0.2657516598701477,
      "step": 1259
    },
    {
      "completion_length": 42.71875,
      "epoch": 2.636649214659686,
      "grad_norm": 23.013017654418945,
      "kl": 5.0,
      "learning_rate": 3.40146750524109e-07,
      "loss": 0.4942,
      "reward": 0.6083165407180786,
      "reward_std": 0.22296372056007385,
      "reward_std_all": 0.21044737100601196,
      "rewards-max/forensics_sinkhorn_reward": 0.9880735576152802,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9963909089565277,
      "rewards/forensics_sinkhorn_reward": 0.5655878782272339,
      "rewards/forensics_strict_edge_reward": 0.367695689201355,
      "rewards/forensics_tiou_reward": 0.3884739279747009,
      "step": 1260
    },
    {
      "completion_length": 95.9375,
      "epoch": 2.6387434554973823,
      "grad_norm": 25.95974349975586,
      "kl": 10.75,
      "learning_rate": 3.396226415094339e-07,
      "loss": 0.697,
      "reward": 0.5507622063159943,
      "reward_std": 0.21839986741542816,
      "reward_std_all": 0.23820039629936218,
      "rewards-max/forensics_sinkhorn_reward": 0.9777558445930481,
      "rewards-max/forensics_strict_edge_reward": 0.855939507484436,
      "rewards-max/forensics_tiou_reward": 0.9692651927471161,
      "rewards/forensics_sinkhorn_reward": 0.4894917458295822,
      "rewards/forensics_strict_edge_reward": 0.2816292941570282,
      "rewards/forensics_tiou_reward": 0.3754780739545822,
      "step": 1261
    },
    {
      "completion_length": 65.046875,
      "epoch": 2.640837696335079,
      "grad_norm": 24.334571838378906,
      "kl": 9.28125,
      "learning_rate": 3.390985324947589e-07,
      "loss": 0.6647,
      "reward": 0.568273514509201,
      "reward_std": 0.15955663472414017,
      "reward_std_all": 0.2027025818824768,
      "rewards-max/forensics_sinkhorn_reward": 0.9887009263038635,
      "rewards-max/forensics_strict_edge_reward": 0.9890114367008209,
      "rewards-max/forensics_tiou_reward": 0.9988591074943542,
      "rewards/forensics_sinkhorn_reward": 0.4983314126729965,
      "rewards/forensics_strict_edge_reward": 0.32875387370586395,
      "rewards/forensics_tiou_reward": 0.41480329632759094,
      "step": 1262
    },
    {
      "completion_length": 85.125,
      "epoch": 2.642931937172775,
      "grad_norm": 17.830656051635742,
      "kl": 9.0,
      "learning_rate": 3.3857442348008387e-07,
      "loss": 0.6602,
      "reward": 0.5475435256958008,
      "reward_std": 0.18035534396767616,
      "reward_std_all": 0.22625886648893356,
      "rewards-max/forensics_sinkhorn_reward": 0.9414176940917969,
      "rewards-max/forensics_strict_edge_reward": 0.8405323326587677,
      "rewards-max/forensics_tiou_reward": 0.8985713720321655,
      "rewards/forensics_sinkhorn_reward": 0.4477292150259018,
      "rewards/forensics_strict_edge_reward": 0.29764944314956665,
      "rewards/forensics_tiou_reward": 0.3350019007921219,
      "step": 1263
    },
    {
      "completion_length": 52.296875,
      "epoch": 2.6450261780104714,
      "grad_norm": 14.86623477935791,
      "kl": 5.46875,
      "learning_rate": 3.380503144654088e-07,
      "loss": 0.3859,
      "reward": 0.6703952848911285,
      "reward_std": 0.14787057787179947,
      "reward_std_all": 0.14289386942982674,
      "rewards-max/forensics_sinkhorn_reward": 0.9727113842964172,
      "rewards-max/forensics_strict_edge_reward": 0.8876297473907471,
      "rewards-max/forensics_tiou_reward": 0.9614598453044891,
      "rewards/forensics_sinkhorn_reward": 0.6290513277053833,
      "rewards/forensics_strict_edge_reward": 0.3505728840827942,
      "rewards/forensics_tiou_reward": 0.46175041794776917,
      "step": 1264
    },
    {
      "completion_length": 70.46875,
      "epoch": 2.6471204188481674,
      "grad_norm": 14.902538299560547,
      "kl": 8.5625,
      "learning_rate": 3.3752620545073374e-07,
      "loss": 0.5058,
      "reward": 0.5179713368415833,
      "reward_std": 0.23762457817792892,
      "reward_std_all": 0.27292972803115845,
      "rewards-max/forensics_sinkhorn_reward": 0.9672872126102448,
      "rewards-max/forensics_strict_edge_reward": 0.8582656681537628,
      "rewards-max/forensics_tiou_reward": 0.9534454941749573,
      "rewards/forensics_sinkhorn_reward": 0.4480225443840027,
      "rewards/forensics_strict_edge_reward": 0.23681925237178802,
      "rewards/forensics_tiou_reward": 0.32471027970314026,
      "step": 1265
    },
    {
      "completion_length": 53.015625,
      "epoch": 2.649214659685864,
      "grad_norm": 22.546035766601562,
      "kl": 6.453125,
      "learning_rate": 3.370020964360587e-07,
      "loss": 0.415,
      "reward": 0.5976660251617432,
      "reward_std": 0.2099028080701828,
      "reward_std_all": 0.2157355546951294,
      "rewards-max/forensics_sinkhorn_reward": 0.9350093007087708,
      "rewards-max/forensics_strict_edge_reward": 0.878633052110672,
      "rewards-max/forensics_tiou_reward": 0.8826995193958282,
      "rewards/forensics_sinkhorn_reward": 0.4992874562740326,
      "rewards/forensics_strict_edge_reward": 0.3127364367246628,
      "rewards/forensics_tiou_reward": 0.3434244990348816,
      "step": 1266
    },
    {
      "completion_length": 88.09375,
      "epoch": 2.65130890052356,
      "grad_norm": 18.251920700073242,
      "kl": 9.3125,
      "learning_rate": 3.364779874213836e-07,
      "loss": 0.6236,
      "reward": 0.5572891384363174,
      "reward_std": 0.20778947323560715,
      "reward_std_all": 0.2247520536184311,
      "rewards-max/forensics_sinkhorn_reward": 0.9561797678470612,
      "rewards-max/forensics_strict_edge_reward": 0.8349016904830933,
      "rewards-max/forensics_tiou_reward": 0.9221575260162354,
      "rewards/forensics_sinkhorn_reward": 0.5067551583051682,
      "rewards/forensics_strict_edge_reward": 0.29045863449573517,
      "rewards/forensics_tiou_reward": 0.3494233191013336,
      "step": 1267
    },
    {
      "completion_length": 94.765625,
      "epoch": 2.6534031413612564,
      "grad_norm": 17.70178985595703,
      "kl": 8.1875,
      "learning_rate": 3.3595387840670856e-07,
      "loss": 0.5,
      "reward": 0.6063670516014099,
      "reward_std": 0.2800971046090126,
      "reward_std_all": 0.25246234238147736,
      "rewards-max/forensics_sinkhorn_reward": 0.9340264797210693,
      "rewards-max/forensics_strict_edge_reward": 0.7802780568599701,
      "rewards-max/forensics_tiou_reward": 0.8790368735790253,
      "rewards/forensics_sinkhorn_reward": 0.5397224426269531,
      "rewards/forensics_strict_edge_reward": 0.3066922277212143,
      "rewards/forensics_tiou_reward": 0.4215269982814789,
      "step": 1268
    },
    {
      "completion_length": 103.890625,
      "epoch": 2.655497382198953,
      "grad_norm": 26.292409896850586,
      "kl": 8.515625,
      "learning_rate": 3.3542976939203357e-07,
      "loss": 0.6328,
      "reward": 0.48959141969680786,
      "reward_std": 0.28692902624607086,
      "reward_std_all": 0.2977210655808449,
      "rewards-max/forensics_sinkhorn_reward": 0.9552919566631317,
      "rewards-max/forensics_strict_edge_reward": 0.8472771048545837,
      "rewards-max/forensics_tiou_reward": 0.9244275391101837,
      "rewards/forensics_sinkhorn_reward": 0.3786873370409012,
      "rewards/forensics_strict_edge_reward": 0.2134372517466545,
      "rewards/forensics_tiou_reward": 0.22257843613624573,
      "step": 1269
    },
    {
      "completion_length": 94.984375,
      "epoch": 2.657591623036649,
      "grad_norm": 20.893634796142578,
      "kl": 9.28125,
      "learning_rate": 3.349056603773585e-07,
      "loss": 0.6399,
      "reward": 0.5207200944423676,
      "reward_std": 0.2690311372280121,
      "reward_std_all": 0.28187546133995056,
      "rewards-max/forensics_sinkhorn_reward": 0.9881614446640015,
      "rewards-max/forensics_strict_edge_reward": 0.9782643616199493,
      "rewards-max/forensics_tiou_reward": 0.9967373609542847,
      "rewards/forensics_sinkhorn_reward": 0.4484841972589493,
      "rewards/forensics_strict_edge_reward": 0.2788108065724373,
      "rewards/forensics_tiou_reward": 0.308610200881958,
      "step": 1270
    },
    {
      "completion_length": 73.5,
      "epoch": 2.6596858638743455,
      "grad_norm": 21.371885299682617,
      "kl": 5.984375,
      "learning_rate": 3.3438155136268344e-07,
      "loss": 0.4516,
      "reward": 0.5222319811582565,
      "reward_std": 0.19107724726200104,
      "reward_std_all": 0.22664916515350342,
      "rewards-max/forensics_sinkhorn_reward": 0.8883761167526245,
      "rewards-max/forensics_strict_edge_reward": 0.6504650861024857,
      "rewards-max/forensics_tiou_reward": 0.7859825491905212,
      "rewards/forensics_sinkhorn_reward": 0.41462452709674835,
      "rewards/forensics_strict_edge_reward": 0.10664820298552513,
      "rewards/forensics_tiou_reward": 0.17482899874448776,
      "step": 1271
    },
    {
      "completion_length": 78.265625,
      "epoch": 2.661780104712042,
      "grad_norm": 48.98177719116211,
      "kl": 7.828125,
      "learning_rate": 3.338574423480084e-07,
      "loss": 0.5513,
      "reward": 0.5407509207725525,
      "reward_std": 0.23687251657247543,
      "reward_std_all": 0.2518821805715561,
      "rewards-max/forensics_sinkhorn_reward": 0.9256420731544495,
      "rewards-max/forensics_strict_edge_reward": 0.7492803931236267,
      "rewards-max/forensics_tiou_reward": 0.8618404269218445,
      "rewards/forensics_sinkhorn_reward": 0.4274447113275528,
      "rewards/forensics_strict_edge_reward": 0.2509615868330002,
      "rewards/forensics_tiou_reward": 0.2859659790992737,
      "step": 1272
    },
    {
      "completion_length": 65.484375,
      "epoch": 2.663874345549738,
      "grad_norm": 34.35560989379883,
      "kl": 7.171875,
      "learning_rate": 3.333333333333333e-07,
      "loss": 0.4034,
      "reward": 0.5449842810630798,
      "reward_std": 0.2120780199766159,
      "reward_std_all": 0.23510338366031647,
      "rewards-max/forensics_sinkhorn_reward": 0.9856996536254883,
      "rewards-max/forensics_strict_edge_reward": 0.9357379674911499,
      "rewards-max/forensics_tiou_reward": 0.9870953261852264,
      "rewards/forensics_sinkhorn_reward": 0.46834956109523773,
      "rewards/forensics_strict_edge_reward": 0.2590901665389538,
      "rewards/forensics_tiou_reward": 0.3043902963399887,
      "step": 1273
    },
    {
      "completion_length": 44.78125,
      "epoch": 2.6659685863874345,
      "grad_norm": 22.834123611450195,
      "kl": 4.84375,
      "learning_rate": 3.3280922431865826e-07,
      "loss": 0.2973,
      "reward": 0.6420163810253143,
      "reward_std": 0.17931945621967316,
      "reward_std_all": 0.17436707019805908,
      "rewards-max/forensics_sinkhorn_reward": 0.9850520193576813,
      "rewards-max/forensics_strict_edge_reward": 0.9364311099052429,
      "rewards-max/forensics_tiou_reward": 0.9845832586288452,
      "rewards/forensics_sinkhorn_reward": 0.5048909187316895,
      "rewards/forensics_strict_edge_reward": 0.30254998803138733,
      "rewards/forensics_tiou_reward": 0.3793680816888809,
      "step": 1274
    },
    {
      "completion_length": 73.34375,
      "epoch": 2.668062827225131,
      "grad_norm": 179.9895782470703,
      "kl": 8.21875,
      "learning_rate": 3.322851153039832e-07,
      "loss": 0.5132,
      "reward": 0.5195730626583099,
      "reward_std": 0.24364255368709564,
      "reward_std_all": 0.2563757598400116,
      "rewards-max/forensics_sinkhorn_reward": 0.8903968632221222,
      "rewards-max/forensics_strict_edge_reward": 0.722899854183197,
      "rewards-max/forensics_tiou_reward": 0.7957625687122345,
      "rewards/forensics_sinkhorn_reward": 0.4050210565328598,
      "rewards/forensics_strict_edge_reward": 0.12390196695923805,
      "rewards/forensics_tiou_reward": 0.21230146288871765,
      "step": 1275
    },
    {
      "completion_length": 83.046875,
      "epoch": 2.670157068062827,
      "grad_norm": 17.765695571899414,
      "kl": 6.21875,
      "learning_rate": 3.317610062893081e-07,
      "loss": 0.4103,
      "reward": 0.6159892678260803,
      "reward_std": 0.2353893369436264,
      "reward_std_all": 0.22208546847105026,
      "rewards-max/forensics_sinkhorn_reward": 0.9886867105960846,
      "rewards-max/forensics_strict_edge_reward": 0.9782643616199493,
      "rewards-max/forensics_tiou_reward": 0.9988023936748505,
      "rewards/forensics_sinkhorn_reward": 0.5823454260826111,
      "rewards/forensics_strict_edge_reward": 0.37904058396816254,
      "rewards/forensics_tiou_reward": 0.42813004553318024,
      "step": 1276
    },
    {
      "completion_length": 88.703125,
      "epoch": 2.6722513089005235,
      "grad_norm": 23.077585220336914,
      "kl": 8.46875,
      "learning_rate": 3.3123689727463314e-07,
      "loss": 0.4697,
      "reward": 0.5707588791847229,
      "reward_std": 0.2041504979133606,
      "reward_std_all": 0.21115487813949585,
      "rewards-max/forensics_sinkhorn_reward": 0.9883845746517181,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9976125955581665,
      "rewards/forensics_sinkhorn_reward": 0.5260169208049774,
      "rewards/forensics_strict_edge_reward": 0.29020680487155914,
      "rewards/forensics_tiou_reward": 0.3751804381608963,
      "step": 1277
    },
    {
      "completion_length": 58.53125,
      "epoch": 2.67434554973822,
      "grad_norm": 20.41922378540039,
      "kl": 6.265625,
      "learning_rate": 3.307127882599581e-07,
      "loss": 0.319,
      "reward": 0.6991526782512665,
      "reward_std": 0.130293857306242,
      "reward_std_all": 0.1216140165925026,
      "rewards-max/forensics_sinkhorn_reward": 0.9835916757583618,
      "rewards-max/forensics_strict_edge_reward": 0.9364311099052429,
      "rewards-max/forensics_tiou_reward": 0.98695307970047,
      "rewards/forensics_sinkhorn_reward": 0.5875853151082993,
      "rewards/forensics_strict_edge_reward": 0.35324129462242126,
      "rewards/forensics_tiou_reward": 0.44034361839294434,
      "step": 1278
    },
    {
      "completion_length": 70.9375,
      "epoch": 2.676439790575916,
      "grad_norm": 21.50800895690918,
      "kl": 8.515625,
      "learning_rate": 3.30188679245283e-07,
      "loss": 0.3937,
      "reward": 0.6179014146327972,
      "reward_std": 0.17276369407773018,
      "reward_std_all": 0.1755715161561966,
      "rewards-max/forensics_sinkhorn_reward": 0.9874527752399445,
      "rewards-max/forensics_strict_edge_reward": 0.9677534997463226,
      "rewards-max/forensics_tiou_reward": 0.9939393997192383,
      "rewards/forensics_sinkhorn_reward": 0.5731253325939178,
      "rewards/forensics_strict_edge_reward": 0.3627157509326935,
      "rewards/forensics_tiou_reward": 0.454249769449234,
      "step": 1279
    },
    {
      "completion_length": 93.796875,
      "epoch": 2.6785340314136126,
      "grad_norm": 489.0637512207031,
      "kl": 12.828125,
      "learning_rate": 3.2966457023060796e-07,
      "loss": 0.7454,
      "reward": 0.5585447549819946,
      "reward_std": 0.1086216140538454,
      "reward_std_all": 0.16821888461709023,
      "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": 0.4902152270078659,
      "rewards/forensics_strict_edge_reward": 0.24772576242685318,
      "rewards/forensics_tiou_reward": 0.34350165724754333,
      "step": 1280
    },
    {
      "completion_length": 78.0,
      "epoch": 2.680628272251309,
      "grad_norm": 22.089496612548828,
      "kl": 9.25,
      "learning_rate": 3.291404612159329e-07,
      "loss": 0.5759,
      "reward": 0.52837273478508,
      "reward_std": 0.20908358693122864,
      "reward_std_all": 0.24730288237333298,
      "rewards-max/forensics_sinkhorn_reward": 0.9880844354629517,
      "rewards-max/forensics_strict_edge_reward": 0.9782643616199493,
      "rewards-max/forensics_tiou_reward": 0.9964285790920258,
      "rewards/forensics_sinkhorn_reward": 0.45698995888233185,
      "rewards/forensics_strict_edge_reward": 0.2247445285320282,
      "rewards/forensics_tiou_reward": 0.3075172156095505,
      "step": 1281
    },
    {
      "completion_length": 99.5625,
      "epoch": 2.682722513089005,
      "grad_norm": 18.84103012084961,
      "kl": 9.5625,
      "learning_rate": 3.286163522012578e-07,
      "loss": 0.617,
      "reward": 0.5503786206245422,
      "reward_std": 0.27507156133651733,
      "reward_std_all": 0.2651084065437317,
      "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": 0.46954596042633057,
      "rewards/forensics_strict_edge_reward": 0.2650803253054619,
      "rewards/forensics_tiou_reward": 0.3287535309791565,
      "step": 1282
    },
    {
      "completion_length": 87.625,
      "epoch": 2.6848167539267016,
      "grad_norm": 19.248592376708984,
      "kl": 7.53125,
      "learning_rate": 3.280922431865828e-07,
      "loss": 0.5309,
      "reward": 0.6334178447723389,
      "reward_std": 0.17884094268083572,
      "reward_std_all": 0.1843252032995224,
      "rewards-max/forensics_sinkhorn_reward": 0.9816426634788513,
      "rewards-max/forensics_strict_edge_reward": 0.8472771048545837,
      "rewards-max/forensics_tiou_reward": 0.9829316735267639,
      "rewards/forensics_sinkhorn_reward": 0.5874635279178619,
      "rewards/forensics_strict_edge_reward": 0.36430270969867706,
      "rewards/forensics_tiou_reward": 0.46209123730659485,
      "step": 1283
    },
    {
      "completion_length": 77.734375,
      "epoch": 2.686910994764398,
      "grad_norm": 28.862396240234375,
      "kl": 7.984375,
      "learning_rate": 3.2756813417190774e-07,
      "loss": 0.6037,
      "reward": 0.5762501955032349,
      "reward_std": 0.21174320578575134,
      "reward_std_all": 0.22823592275381088,
      "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": 0.5365437269210815,
      "rewards/forensics_strict_edge_reward": 0.3253185749053955,
      "rewards/forensics_tiou_reward": 0.42810551822185516,
      "step": 1284
    },
    {
      "completion_length": 54.578125,
      "epoch": 2.689005235602094,
      "grad_norm": 23.60405731201172,
      "kl": 4.078125,
      "learning_rate": 3.2704402515723265e-07,
      "loss": 0.2642,
      "reward": 0.6099646687507629,
      "reward_std": 0.20585856586694717,
      "reward_std_all": 0.20353619754314423,
      "rewards-max/forensics_sinkhorn_reward": 0.9771441221237183,
      "rewards-max/forensics_strict_edge_reward": 0.8893801271915436,
      "rewards-max/forensics_tiou_reward": 0.9744827449321747,
      "rewards/forensics_sinkhorn_reward": 0.4853847026824951,
      "rewards/forensics_strict_edge_reward": 0.24283242225646973,
      "rewards/forensics_tiou_reward": 0.27583371102809906,
      "step": 1285
    },
    {
      "completion_length": 98.203125,
      "epoch": 2.6910994764397906,
      "grad_norm": 18.54867935180664,
      "kl": 8.265625,
      "learning_rate": 3.2651991614255766e-07,
      "loss": 0.3549,
      "reward": 0.4872683584690094,
      "reward_std": 0.20790906623005867,
      "reward_std_all": 0.2555091977119446,
      "rewards-max/forensics_sinkhorn_reward": 0.9863245189189911,
      "rewards-max/forensics_strict_edge_reward": 0.9357379674911499,
      "rewards-max/forensics_tiou_reward": 0.989515095949173,
      "rewards/forensics_sinkhorn_reward": 0.4097610414028168,
      "rewards/forensics_strict_edge_reward": 0.14823614805936813,
      "rewards/forensics_tiou_reward": 0.21111533045768738,
      "step": 1286
    },
    {
      "completion_length": 62.3125,
      "epoch": 2.693193717277487,
      "grad_norm": 23.6451358795166,
      "kl": 3.890625,
      "learning_rate": 3.259958071278826e-07,
      "loss": 0.3876,
      "reward": 0.6595732867717743,
      "reward_std": 0.1985800862312317,
      "reward_std_all": 0.18361486494541168,
      "rewards-max/forensics_sinkhorn_reward": 0.9883348047733307,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.997417539358139,
      "rewards/forensics_sinkhorn_reward": 0.6051118969917297,
      "rewards/forensics_strict_edge_reward": 0.42294828593730927,
      "rewards/forensics_tiou_reward": 0.48232702910900116,
      "step": 1287
    },
    {
      "completion_length": 59.65625,
      "epoch": 2.695287958115183,
      "grad_norm": 15.695637702941895,
      "kl": 4.7421875,
      "learning_rate": 3.254716981132075e-07,
      "loss": 0.3524,
      "reward": 0.6301153302192688,
      "reward_std": 0.18206438422203064,
      "reward_std_all": 0.1855783686041832,
      "rewards-max/forensics_sinkhorn_reward": 0.9883900582790375,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9976343810558319,
      "rewards/forensics_sinkhorn_reward": 0.6241146177053452,
      "rewards/forensics_strict_edge_reward": 0.39199863374233246,
      "rewards/forensics_tiou_reward": 0.490627184510231,
      "step": 1288
    },
    {
      "completion_length": 57.703125,
      "epoch": 2.6973821989528797,
      "grad_norm": 17.45207977294922,
      "kl": 5.125,
      "learning_rate": 3.249475890985325e-07,
      "loss": 0.3716,
      "reward": 0.6083173155784607,
      "reward_std": 0.1559111848473549,
      "reward_std_all": 0.1764409840106964,
      "rewards-max/forensics_sinkhorn_reward": 0.9691835641860962,
      "rewards-max/forensics_strict_edge_reward": 0.8805811703205109,
      "rewards-max/forensics_tiou_reward": 0.9536576569080353,
      "rewards/forensics_sinkhorn_reward": 0.5525402873754501,
      "rewards/forensics_strict_edge_reward": 0.28475798666477203,
      "rewards/forensics_tiou_reward": 0.36676670610904694,
      "step": 1289
    },
    {
      "completion_length": 86.21875,
      "epoch": 2.699476439790576,
      "grad_norm": 20.097070693969727,
      "kl": 6.609375,
      "learning_rate": 3.2442348008385744e-07,
      "loss": 0.5218,
      "reward": 0.5531505346298218,
      "reward_std": 0.2937885820865631,
      "reward_std_all": 0.27564260363578796,
      "rewards-max/forensics_sinkhorn_reward": 0.9787565469741821,
      "rewards-max/forensics_strict_edge_reward": 0.9357379674911499,
      "rewards-max/forensics_tiou_reward": 0.9774263203144073,
      "rewards/forensics_sinkhorn_reward": 0.5126412957906723,
      "rewards/forensics_strict_edge_reward": 0.3461071848869324,
      "rewards/forensics_tiou_reward": 0.4286508709192276,
      "step": 1290
    },
    {
      "completion_length": 55.703125,
      "epoch": 2.701570680628272,
      "grad_norm": 24.24530601501465,
      "kl": 4.875,
      "learning_rate": 3.2389937106918235e-07,
      "loss": 0.3606,
      "reward": 0.6419605314731598,
      "reward_std": 0.2172526866197586,
      "reward_std_all": 0.20129191130399704,
      "rewards-max/forensics_sinkhorn_reward": 0.9885387420654297,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.998219907283783,
      "rewards/forensics_sinkhorn_reward": 0.5729666948318481,
      "rewards/forensics_strict_edge_reward": 0.36435239017009735,
      "rewards/forensics_tiou_reward": 0.43661606311798096,
      "step": 1291
    },
    {
      "completion_length": 39.53125,
      "epoch": 2.7036649214659687,
      "grad_norm": 18.694156646728516,
      "kl": 2.9453125,
      "learning_rate": 3.233752620545073e-07,
      "loss": 0.171,
      "reward": 0.7113091051578522,
      "reward_std": 0.0850019846111536,
      "reward_std_all": 0.0934053584933281,
      "rewards-max/forensics_sinkhorn_reward": 0.9887022376060486,
      "rewards-max/forensics_strict_edge_reward": 0.9574736058712006,
      "rewards-max/forensics_tiou_reward": 0.9988636374473572,
      "rewards/forensics_sinkhorn_reward": 0.6507864892482758,
      "rewards/forensics_strict_edge_reward": 0.32529790699481964,
      "rewards/forensics_tiou_reward": 0.4758525937795639,
      "step": 1292
    },
    {
      "completion_length": 80.96875,
      "epoch": 2.7057591623036648,
      "grad_norm": 16.45174217224121,
      "kl": 6.296875,
      "learning_rate": 3.228511530398323e-07,
      "loss": 0.4523,
      "reward": 0.5689907968044281,
      "reward_std": 0.18622969835996628,
      "reward_std_all": 0.20925787091255188,
      "rewards-max/forensics_sinkhorn_reward": 0.9878159165382385,
      "rewards-max/forensics_strict_edge_reward": 0.9782643616199493,
      "rewards-max/forensics_tiou_reward": 0.995381087064743,
      "rewards/forensics_sinkhorn_reward": 0.47563138604164124,
      "rewards/forensics_strict_edge_reward": 0.2611225098371506,
      "rewards/forensics_tiou_reward": 0.30058491230010986,
      "step": 1293
    },
    {
      "completion_length": 51.046875,
      "epoch": 2.7078534031413612,
      "grad_norm": 20.047988891601562,
      "kl": 6.796875,
      "learning_rate": 3.223270440251572e-07,
      "loss": 0.3983,
      "reward": 0.6264511048793793,
      "reward_std": 0.18758856505155563,
      "reward_std_all": 0.1799398511648178,
      "rewards-max/forensics_sinkhorn_reward": 0.9880357086658478,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9962399005889893,
      "rewards/forensics_sinkhorn_reward": 0.5896551012992859,
      "rewards/forensics_strict_edge_reward": 0.35188838839530945,
      "rewards/forensics_tiou_reward": 0.4367208629846573,
      "step": 1294
    },
    {
      "completion_length": 69.921875,
      "epoch": 2.7099476439790577,
      "grad_norm": 20.748756408691406,
      "kl": 9.4375,
      "learning_rate": 3.218029350104822e-07,
      "loss": 0.7025,
      "reward": 0.4631652981042862,
      "reward_std": 0.25493165105581284,
      "reward_std_all": 0.293760746717453,
      "rewards-max/forensics_sinkhorn_reward": 0.9738684594631195,
      "rewards-max/forensics_strict_edge_reward": 0.9151731729507446,
      "rewards-max/forensics_tiou_reward": 0.9636233448982239,
      "rewards/forensics_sinkhorn_reward": 0.36349061131477356,
      "rewards/forensics_strict_edge_reward": 0.2063640058040619,
      "rewards/forensics_tiou_reward": 0.23946750164031982,
      "step": 1295
    },
    {
      "completion_length": 60.75,
      "epoch": 2.712041884816754,
      "grad_norm": 20.746150970458984,
      "kl": 8.421875,
      "learning_rate": 3.2127882599580714e-07,
      "loss": 0.5269,
      "reward": 0.5081442594528198,
      "reward_std": 0.20730211585760117,
      "reward_std_all": 0.24743227660655975,
      "rewards-max/forensics_sinkhorn_reward": 0.9572811126708984,
      "rewards-max/forensics_strict_edge_reward": 0.7743246257305145,
      "rewards-max/forensics_tiou_reward": 0.9285056591033936,
      "rewards/forensics_sinkhorn_reward": 0.3935575485229492,
      "rewards/forensics_strict_edge_reward": 0.1418853960931301,
      "rewards/forensics_tiou_reward": 0.2238277792930603,
      "step": 1296
    },
    {
      "completion_length": 71.0625,
      "epoch": 2.7141361256544503,
      "grad_norm": 17.4101619720459,
      "kl": 6.265625,
      "learning_rate": 3.2075471698113204e-07,
      "loss": 0.3932,
      "reward": 0.6199911534786224,
      "reward_std": 0.18016085773706436,
      "reward_std_all": 0.1798458993434906,
      "rewards-max/forensics_sinkhorn_reward": 0.9843052923679352,
      "rewards-max/forensics_strict_edge_reward": 0.9169811904430389,
      "rewards-max/forensics_tiou_reward": 0.9815429747104645,
      "rewards/forensics_sinkhorn_reward": 0.5427803695201874,
      "rewards/forensics_strict_edge_reward": 0.27911555022001266,
      "rewards/forensics_tiou_reward": 0.3643312454223633,
      "step": 1297
    },
    {
      "completion_length": 38.046875,
      "epoch": 2.7162303664921463,
      "grad_norm": 22.08083724975586,
      "kl": 4.53125,
      "learning_rate": 3.20230607966457e-07,
      "loss": 0.2623,
      "reward": 0.6335904598236084,
      "reward_std": 0.17123518884181976,
      "reward_std_all": 0.17479314655065536,
      "rewards-max/forensics_sinkhorn_reward": 0.819476455450058,
      "rewards-max/forensics_strict_edge_reward": 0.7676876485347748,
      "rewards-max/forensics_tiou_reward": 0.7217820286750793,
      "rewards/forensics_sinkhorn_reward": 0.4941983073949814,
      "rewards/forensics_strict_edge_reward": 0.20613499730825424,
      "rewards/forensics_tiou_reward": 0.2296205312013626,
      "step": 1298
    },
    {
      "completion_length": 70.3125,
      "epoch": 2.718324607329843,
      "grad_norm": 22.564266204833984,
      "kl": 10.546875,
      "learning_rate": 3.1970649895178196e-07,
      "loss": 0.5542,
      "reward": 0.5740117430686951,
      "reward_std": 0.17747575044631958,
      "reward_std_all": 0.20395110547542572,
      "rewards-max/forensics_sinkhorn_reward": 0.9871697127819061,
      "rewards-max/forensics_strict_edge_reward": 0.9357379674911499,
      "rewards-max/forensics_tiou_reward": 0.9928330183029175,
      "rewards/forensics_sinkhorn_reward": 0.46522097289562225,
      "rewards/forensics_strict_edge_reward": 0.25727929174900055,
      "rewards/forensics_tiou_reward": 0.2869717925786972,
      "step": 1299
    },
    {
      "completion_length": 86.6875,
      "epoch": 2.7204188481675393,
      "grad_norm": 26.43166160583496,
      "kl": 12.375,
      "learning_rate": 3.1918238993710687e-07,
      "loss": 0.7115,
      "reward": 0.4998215138912201,
      "reward_std": 0.2162136733531952,
      "reward_std_all": 0.2679375857114792,
      "rewards-max/forensics_sinkhorn_reward": 0.9415769577026367,
      "rewards-max/forensics_strict_edge_reward": 0.659758597612381,
      "rewards-max/forensics_tiou_reward": 0.8966650366783142,
      "rewards/forensics_sinkhorn_reward": 0.4136267751455307,
      "rewards/forensics_strict_edge_reward": 0.1708132103085518,
      "rewards/forensics_tiou_reward": 0.25851500034332275,
      "step": 1300
    },
    {
      "completion_length": 70.21875,
      "epoch": 2.7225130890052354,
      "grad_norm": 19.541624069213867,
      "kl": 8.890625,
      "learning_rate": 3.186582809224319e-07,
      "loss": 0.5596,
      "reward": 0.5417301952838898,
      "reward_std": 0.21638262271881104,
      "reward_std_all": 0.23717549443244934,
      "rewards-max/forensics_sinkhorn_reward": 0.987552136182785,
      "rewards-max/forensics_strict_edge_reward": 0.9464850425720215,
      "rewards-max/forensics_tiou_reward": 0.9943302571773529,
      "rewards/forensics_sinkhorn_reward": 0.48766113817691803,
      "rewards/forensics_strict_edge_reward": 0.21838834881782532,
      "rewards/forensics_tiou_reward": 0.31435707211494446,
      "step": 1301
    },
    {
      "completion_length": 65.234375,
      "epoch": 2.724607329842932,
      "grad_norm": 31.013935089111328,
      "kl": 11.96875,
      "learning_rate": 3.1813417190775684e-07,
      "loss": 0.8461,
      "reward": 0.4811951816082001,
      "reward_std": 0.2414916455745697,
      "reward_std_all": 0.27542997896671295,
      "rewards-max/forensics_sinkhorn_reward": 0.9724754691123962,
      "rewards-max/forensics_strict_edge_reward": 0.8659839034080505,
      "rewards-max/forensics_tiou_reward": 0.9565494656562805,
      "rewards/forensics_sinkhorn_reward": 0.37718988955020905,
      "rewards/forensics_strict_edge_reward": 0.20511404424905777,
      "rewards/forensics_tiou_reward": 0.27219465374946594,
      "step": 1302
    },
    {
      "completion_length": 83.28125,
      "epoch": 2.7267015706806284,
      "grad_norm": 92.29999542236328,
      "kl": 9.109375,
      "learning_rate": 3.1761006289308174e-07,
      "loss": 0.6332,
      "reward": 0.5858410596847534,
      "reward_std": 0.23124819993972778,
      "reward_std_all": 0.2263241559267044,
      "rewards-max/forensics_sinkhorn_reward": 0.9839982688426971,
      "rewards-max/forensics_strict_edge_reward": 0.8876297473907471,
      "rewards-max/forensics_tiou_reward": 0.9803350865840912,
      "rewards/forensics_sinkhorn_reward": 0.5268000662326813,
      "rewards/forensics_strict_edge_reward": 0.3444209694862366,
      "rewards/forensics_tiou_reward": 0.40302789211273193,
      "step": 1303
    },
    {
      "completion_length": 51.34375,
      "epoch": 2.7287958115183244,
      "grad_norm": 15.225410461425781,
      "kl": 5.0859375,
      "learning_rate": 3.170859538784067e-07,
      "loss": 0.4016,
      "reward": 0.6138789653778076,
      "reward_std": 0.15053410083055496,
      "reward_std_all": 0.1700529232621193,
      "rewards-max/forensics_sinkhorn_reward": 0.9858845770359039,
      "rewards-max/forensics_strict_edge_reward": 0.8854433596134186,
      "rewards-max/forensics_tiou_reward": 0.9878208935260773,
      "rewards/forensics_sinkhorn_reward": 0.5065647661685944,
      "rewards/forensics_strict_edge_reward": 0.25271232426166534,
      "rewards/forensics_tiou_reward": 0.31329189240932465,
      "step": 1304
    },
    {
      "completion_length": 61.453125,
      "epoch": 2.730890052356021,
      "grad_norm": 21.26799201965332,
      "kl": 5.28125,
      "learning_rate": 3.1656184486373166e-07,
      "loss": 0.4038,
      "reward": 0.6100776195526123,
      "reward_std": 0.18686795979738235,
      "reward_std_all": 0.19101300835609436,
      "rewards-max/forensics_sinkhorn_reward": 0.960053414106369,
      "rewards-max/forensics_strict_edge_reward": 0.8328719735145569,
      "rewards-max/forensics_tiou_reward": 0.933395266532898,
      "rewards/forensics_sinkhorn_reward": 0.4960438758134842,
      "rewards/forensics_strict_edge_reward": 0.22065535932779312,
      "rewards/forensics_tiou_reward": 0.3144885450601578,
      "step": 1305
    },
    {
      "completion_length": 70.125,
      "epoch": 2.7329842931937174,
      "grad_norm": 18.0524845123291,
      "kl": 6.25,
      "learning_rate": 3.1603773584905657e-07,
      "loss": 0.6006,
      "reward": 0.6041953563690186,
      "reward_std": 0.20965129882097244,
      "reward_std_all": 0.215439572930336,
      "rewards-max/forensics_sinkhorn_reward": 0.9879591763019562,
      "rewards-max/forensics_strict_edge_reward": 0.9782643616199493,
      "rewards-max/forensics_tiou_reward": 0.9959432184696198,
      "rewards/forensics_sinkhorn_reward": 0.512576162815094,
      "rewards/forensics_strict_edge_reward": 0.34166625142097473,
      "rewards/forensics_tiou_reward": 0.3924964517354965,
      "step": 1306
    },
    {
      "completion_length": 53.59375,
      "epoch": 2.7350785340314134,
      "grad_norm": 22.23948860168457,
      "kl": 4.546875,
      "learning_rate": 3.155136268343815e-07,
      "loss": 0.4344,
      "reward": 0.5839881300926208,
      "reward_std": 0.18153922259807587,
      "reward_std_all": 0.19222605973482132,
      "rewards-max/forensics_sinkhorn_reward": 0.9716424345970154,
      "rewards-max/forensics_strict_edge_reward": 0.946017861366272,
      "rewards-max/forensics_tiou_reward": 0.9598299860954285,
      "rewards/forensics_sinkhorn_reward": 0.4740251749753952,
      "rewards/forensics_strict_edge_reward": 0.2904548794031143,
      "rewards/forensics_tiou_reward": 0.2957029491662979,
      "step": 1307
    },
    {
      "completion_length": 68.75,
      "epoch": 2.73717277486911,
      "grad_norm": 14.482725143432617,
      "kl": 5.875,
      "learning_rate": 3.149895178197065e-07,
      "loss": 0.5751,
      "reward": 0.5287081897258759,
      "reward_std": 0.20443099737167358,
      "reward_std_all": 0.2347780019044876,
      "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": 0.4965197890996933,
      "rewards/forensics_strict_edge_reward": 0.3305091857910156,
      "rewards/forensics_tiou_reward": 0.38890695571899414,
      "step": 1308
    },
    {
      "completion_length": 54.96875,
      "epoch": 2.7392670157068064,
      "grad_norm": 63.69305419921875,
      "kl": 7.890625,
      "learning_rate": 3.1446540880503144e-07,
      "loss": 0.4312,
      "reward": 0.6211203932762146,
      "reward_std": 0.2158592939376831,
      "reward_std_all": 0.20257791876792908,
      "rewards-max/forensics_sinkhorn_reward": 0.9884881675243378,
      "rewards-max/forensics_strict_edge_reward": 0.9890114367008209,
      "rewards-max/forensics_tiou_reward": 0.998019814491272,
      "rewards/forensics_sinkhorn_reward": 0.5944517254829407,
      "rewards/forensics_strict_edge_reward": 0.39035220444202423,
      "rewards/forensics_tiou_reward": 0.4608701169490814,
      "step": 1309
    },
    {
      "completion_length": 83.609375,
      "epoch": 2.7413612565445025,
      "grad_norm": 29.901582717895508,
      "kl": 7.171875,
      "learning_rate": 3.139412997903564e-07,
      "loss": 0.5806,
      "reward": 0.5104382336139679,
      "reward_std": 0.17067930102348328,
      "reward_std_all": 0.23320304602384567,
      "rewards-max/forensics_sinkhorn_reward": 0.9886201620101929,
      "rewards-max/forensics_strict_edge_reward": 0.9890114367008209,
      "rewards-max/forensics_tiou_reward": 0.9985401332378387,
      "rewards/forensics_sinkhorn_reward": 0.46152976155281067,
      "rewards/forensics_strict_edge_reward": 0.2792980596423149,
      "rewards/forensics_tiou_reward": 0.310890756547451,
      "step": 1310
    },
    {
      "completion_length": 64.421875,
      "epoch": 2.743455497382199,
      "grad_norm": 16.581703186035156,
      "kl": 6.375,
      "learning_rate": 3.1341719077568136e-07,
      "loss": 0.4714,
      "reward": 0.6137395799160004,
      "reward_std": 0.21214976161718369,
      "reward_std_all": 0.20664981752634048,
      "rewards-max/forensics_sinkhorn_reward": 0.9811348021030426,
      "rewards-max/forensics_strict_edge_reward": 0.9677534997463226,
      "rewards-max/forensics_tiou_reward": 0.9822951555252075,
      "rewards/forensics_sinkhorn_reward": 0.5231429636478424,
      "rewards/forensics_strict_edge_reward": 0.3318502604961395,
      "rewards/forensics_tiou_reward": 0.3873409181833267,
      "step": 1311
    },
    {
      "completion_length": 72.203125,
      "epoch": 2.7455497382198955,
      "grad_norm": 106.35419464111328,
      "kl": 8.90625,
      "learning_rate": 3.1289308176100627e-07,
      "loss": 0.5426,
      "reward": 0.6147771179676056,
      "reward_std": 0.1804947704076767,
      "reward_std_all": 0.18076694756746292,
      "rewards-max/forensics_sinkhorn_reward": 0.9881276190280914,
      "rewards-max/forensics_strict_edge_reward": 0.9265981018543243,
      "rewards-max/forensics_tiou_reward": 0.9966033399105072,
      "rewards/forensics_sinkhorn_reward": 0.5641342550516129,
      "rewards/forensics_strict_edge_reward": 0.2257976457476616,
      "rewards/forensics_tiou_reward": 0.39217764139175415,
      "step": 1312
    },
    {
      "completion_length": 78.3125,
      "epoch": 2.7476439790575915,
      "grad_norm": 21.971027374267578,
      "kl": 10.0,
      "learning_rate": 3.123689727463312e-07,
      "loss": 0.6045,
      "reward": 0.4897749722003937,
      "reward_std": 0.20202389359474182,
      "reward_std_all": 0.2641981244087219,
      "rewards-max/forensics_sinkhorn_reward": 0.987771213054657,
      "rewards-max/forensics_strict_edge_reward": 0.9565287232398987,
      "rewards-max/forensics_tiou_reward": 0.9952017366886139,
      "rewards/forensics_sinkhorn_reward": 0.4291197210550308,
      "rewards/forensics_strict_edge_reward": 0.29123513400554657,
      "rewards/forensics_tiou_reward": 0.31952688097953796,
      "step": 1313
    },
    {
      "completion_length": 129.21875,
      "epoch": 2.749738219895288,
      "grad_norm": 30.44618034362793,
      "kl": 12.8125,
      "learning_rate": 3.118448637316562e-07,
      "loss": 0.8485,
      "reward": 0.42212237417697906,
      "reward_std": 0.2561826556921005,
      "reward_std_all": 0.30304335057735443,
      "rewards-max/forensics_sinkhorn_reward": 0.9659880995750427,
      "rewards-max/forensics_strict_edge_reward": 0.8507176637649536,
      "rewards-max/forensics_tiou_reward": 0.9445928335189819,
      "rewards/forensics_sinkhorn_reward": 0.3120441883802414,
      "rewards/forensics_strict_edge_reward": 0.164373941719532,
      "rewards/forensics_tiou_reward": 0.18698831647634506,
      "step": 1314
    },
    {
      "completion_length": 64.84375,
      "epoch": 2.7518324607329845,
      "grad_norm": 31.753416061401367,
      "kl": 7.7734375,
      "learning_rate": 3.113207547169811e-07,
      "loss": 0.5594,
      "reward": 0.5521230101585388,
      "reward_std": 0.2654801979660988,
      "reward_std_all": 0.2666424959897995,
      "rewards-max/forensics_sinkhorn_reward": 0.9861707389354706,
      "rewards-max/forensics_strict_edge_reward": 0.947419673204422,
      "rewards-max/forensics_tiou_reward": 0.9889506101608276,
      "rewards/forensics_sinkhorn_reward": 0.4864547550678253,
      "rewards/forensics_strict_edge_reward": 0.3962443172931671,
      "rewards/forensics_tiou_reward": 0.4163612872362137,
      "step": 1315
    },
    {
      "completion_length": 45.296875,
      "epoch": 2.7539267015706805,
      "grad_norm": 23.819412231445312,
      "kl": 7.296875,
      "learning_rate": 3.1079664570230605e-07,
      "loss": 0.4349,
      "reward": 0.5731469690799713,
      "reward_std": 0.19655897468328476,
      "reward_std_all": 0.21864806860685349,
      "rewards-max/forensics_sinkhorn_reward": 0.9461376965045929,
      "rewards-max/forensics_strict_edge_reward": 0.8805811703205109,
      "rewards-max/forensics_tiou_reward": 0.9057105183601379,
      "rewards/forensics_sinkhorn_reward": 0.48272745311260223,
      "rewards/forensics_strict_edge_reward": 0.24616503715515137,
      "rewards/forensics_tiou_reward": 0.28373972326517105,
      "step": 1316
    },
    {
      "completion_length": 45.765625,
      "epoch": 2.756020942408377,
      "grad_norm": 28.729665756225586,
      "kl": 6.078125,
      "learning_rate": 3.1027253668763106e-07,
      "loss": 0.417,
      "reward": 0.6336241364479065,
      "reward_std": 0.1568453088402748,
      "reward_std_all": 0.160274475812912,
      "rewards-max/forensics_sinkhorn_reward": 0.9863399565219879,
      "rewards-max/forensics_strict_edge_reward": 0.9464850425720215,
      "rewards-max/forensics_tiou_reward": 0.9895958006381989,
      "rewards/forensics_sinkhorn_reward": 0.5808095335960388,
      "rewards/forensics_strict_edge_reward": 0.34982460737228394,
      "rewards/forensics_tiou_reward": 0.42444486916065216,
      "step": 1317
    },
    {
      "completion_length": 101.15625,
      "epoch": 2.7581151832460735,
      "grad_norm": 35.02638626098633,
      "kl": 10.078125,
      "learning_rate": 3.0974842767295597e-07,
      "loss": 0.6218,
      "reward": 0.5393231958150864,
      "reward_std": 0.20442575216293335,
      "reward_std_all": 0.2467067986726761,
      "rewards-max/forensics_sinkhorn_reward": 0.9880594909191132,
      "rewards-max/forensics_strict_edge_reward": 0.9574736058712006,
      "rewards-max/forensics_tiou_reward": 0.9963302612304688,
      "rewards/forensics_sinkhorn_reward": 0.4987562298774719,
      "rewards/forensics_strict_edge_reward": 0.3836126923561096,
      "rewards/forensics_tiou_reward": 0.4307510256767273,
      "step": 1318
    },
    {
      "completion_length": 67.546875,
      "epoch": 2.7602094240837696,
      "grad_norm": 18.43256950378418,
      "kl": 5.921875,
      "learning_rate": 3.092243186582809e-07,
      "loss": 0.4443,
      "reward": 0.6694449782371521,
      "reward_std": 0.1623961329460144,
      "reward_std_all": 0.15801849961280823,
      "rewards-max/forensics_sinkhorn_reward": 0.9854982495307922,
      "rewards-max/forensics_strict_edge_reward": 0.8698340952396393,
      "rewards-max/forensics_tiou_reward": 0.9862909018993378,
      "rewards/forensics_sinkhorn_reward": 0.6094968914985657,
      "rewards/forensics_strict_edge_reward": 0.3382881432771683,
      "rewards/forensics_tiou_reward": 0.45855510234832764,
      "step": 1319
    },
    {
      "completion_length": 72.859375,
      "epoch": 2.762303664921466,
      "grad_norm": 17.943683624267578,
      "kl": 7.34375,
      "learning_rate": 3.087002096436059e-07,
      "loss": 0.5544,
      "reward": 0.48439061641693115,
      "reward_std": 0.2096829116344452,
      "reward_std_all": 0.2660374045372009,
      "rewards-max/forensics_sinkhorn_reward": 0.9757114350795746,
      "rewards-max/forensics_strict_edge_reward": 0.946017861366272,
      "rewards-max/forensics_tiou_reward": 0.9691295623779297,
      "rewards/forensics_sinkhorn_reward": 0.4124313294887543,
      "rewards/forensics_strict_edge_reward": 0.2319992482662201,
      "rewards/forensics_tiou_reward": 0.2638852596282959,
      "step": 1320
    },
    {
      "completion_length": 50.96875,
      "epoch": 2.7643979057591626,
      "grad_norm": 59.216800689697266,
      "kl": 8.90625,
      "learning_rate": 3.081761006289308e-07,
      "loss": 0.4593,
      "reward": 0.5416120290756226,
      "reward_std": 0.22101380676031113,
      "reward_std_all": 0.24052973091602325,
      "rewards-max/forensics_sinkhorn_reward": 0.980856329202652,
      "rewards-max/forensics_strict_edge_reward": 0.9158510267734528,
      "rewards-max/forensics_tiou_reward": 0.9790972769260406,
      "rewards/forensics_sinkhorn_reward": 0.4410802274942398,
      "rewards/forensics_strict_edge_reward": 0.2614900469779968,
      "rewards/forensics_tiou_reward": 0.3027088940143585,
      "step": 1321
    },
    {
      "completion_length": 56.578125,
      "epoch": 2.7664921465968586,
      "grad_norm": 24.51561737060547,
      "kl": 7.53125,
      "learning_rate": 3.0765199161425575e-07,
      "loss": 0.5658,
      "reward": 0.4763963669538498,
      "reward_std": 0.26031072437763214,
      "reward_std_all": 0.28151218593120575,
      "rewards-max/forensics_sinkhorn_reward": 0.9767563343048096,
      "rewards-max/forensics_strict_edge_reward": 0.9464850425720215,
      "rewards-max/forensics_tiou_reward": 0.9715022444725037,
      "rewards/forensics_sinkhorn_reward": 0.4206753820180893,
      "rewards/forensics_strict_edge_reward": 0.26422375440597534,
      "rewards/forensics_tiou_reward": 0.30426716804504395,
      "step": 1322
    },
    {
      "completion_length": 69.734375,
      "epoch": 2.768586387434555,
      "grad_norm": 29.130599975585938,
      "kl": 9.625,
      "learning_rate": 3.071278825995807e-07,
      "loss": 0.6544,
      "reward": 0.5179043114185333,
      "reward_std": 0.25203315913677216,
      "reward_std_all": 0.2769412398338318,
      "rewards-max/forensics_sinkhorn_reward": 0.9847494065761566,
      "rewards-max/forensics_strict_edge_reward": 0.9169811904430389,
      "rewards-max/forensics_tiou_reward": 0.9834125339984894,
      "rewards/forensics_sinkhorn_reward": 0.45425157248973846,
      "rewards/forensics_strict_edge_reward": 0.32459698617458344,
      "rewards/forensics_tiou_reward": 0.36815112829208374,
      "step": 1323
    },
    {
      "completion_length": 72.6875,
      "epoch": 2.770680628272251,
      "grad_norm": 27.67304229736328,
      "kl": 7.796875,
      "learning_rate": 3.066037735849056e-07,
      "loss": 0.4866,
      "reward": 0.533921480178833,
      "reward_std": 0.23309510946273804,
      "reward_std_all": 0.2515189200639725,
      "rewards-max/forensics_sinkhorn_reward": 0.9567941129207611,
      "rewards-max/forensics_strict_edge_reward": 0.856150895357132,
      "rewards-max/forensics_tiou_reward": 0.9265459179878235,
      "rewards/forensics_sinkhorn_reward": 0.4294874668121338,
      "rewards/forensics_strict_edge_reward": 0.24033693969249725,
      "rewards/forensics_tiou_reward": 0.2970236837863922,
      "step": 1324
    },
    {
      "completion_length": 76.140625,
      "epoch": 2.7727748691099476,
      "grad_norm": 33.01266860961914,
      "kl": 6.109375,
      "learning_rate": 3.060796645702306e-07,
      "loss": 0.4949,
      "reward": 0.5049696266651154,
      "reward_std": 0.2538096159696579,
      "reward_std_all": 0.2670057415962219,
      "rewards-max/forensics_sinkhorn_reward": 0.9372918605804443,
      "rewards-max/forensics_strict_edge_reward": 0.7348747849464417,
      "rewards-max/forensics_tiou_reward": 0.8822607100009918,
      "rewards/forensics_sinkhorn_reward": 0.43219272792339325,
      "rewards/forensics_strict_edge_reward": 0.18298490718007088,
      "rewards/forensics_tiou_reward": 0.268669031560421,
      "step": 1325
    },
    {
      "completion_length": 44.875,
      "epoch": 2.774869109947644,
      "grad_norm": 31.324203491210938,
      "kl": 6.5703125,
      "learning_rate": 3.055555555555556e-07,
      "loss": 0.3438,
      "reward": 0.6417626142501831,
      "reward_std": 0.2208225280046463,
      "reward_std_all": 0.2017439678311348,
      "rewards-max/forensics_sinkhorn_reward": 0.9883950054645538,
      "rewards-max/forensics_strict_edge_reward": 0.9890114367008209,
      "rewards-max/forensics_tiou_reward": 0.9976553320884705,
      "rewards/forensics_sinkhorn_reward": 0.5536598265171051,
      "rewards/forensics_strict_edge_reward": 0.370571032166481,
      "rewards/forensics_tiou_reward": 0.3971753269433975,
      "step": 1326
    },
    {
      "completion_length": 76.96875,
      "epoch": 2.77696335078534,
      "grad_norm": 23.013296127319336,
      "kl": 6.578125,
      "learning_rate": 3.050314465408805e-07,
      "loss": 0.4086,
      "reward": 0.50565105676651,
      "reward_std": 0.2803558260202408,
      "reward_std_all": 0.2910561263561249,
      "rewards-max/forensics_sinkhorn_reward": 0.9847266674041748,
      "rewards-max/forensics_strict_edge_reward": 0.9169811904430389,
      "rewards-max/forensics_tiou_reward": 0.9832018315792084,
      "rewards/forensics_sinkhorn_reward": 0.42988765239715576,
      "rewards/forensics_strict_edge_reward": 0.16855355352163315,
      "rewards/forensics_tiou_reward": 0.3040922060608864,
      "step": 1327
    },
    {
      "completion_length": 27.0,
      "epoch": 2.7790575916230367,
      "grad_norm": 28.86484718322754,
      "kl": 5.328125,
      "learning_rate": 3.0450733752620545e-07,
      "loss": 0.3423,
      "reward": 0.6047804057598114,
      "reward_std": 0.1492060273885727,
      "reward_std_all": 0.17767224460840225,
      "rewards-max/forensics_sinkhorn_reward": 0.9870898425579071,
      "rewards-max/forensics_strict_edge_reward": 0.9574736058712006,
      "rewards-max/forensics_tiou_reward": 0.9925374090671539,
      "rewards/forensics_sinkhorn_reward": 0.5042160451412201,
      "rewards/forensics_strict_edge_reward": 0.24348001182079315,
      "rewards/forensics_tiou_reward": 0.28937700390815735,
      "step": 1328
    },
    {
      "completion_length": 51.6875,
      "epoch": 2.7811518324607327,
      "grad_norm": 24.185434341430664,
      "kl": 5.75,
      "learning_rate": 3.039832285115304e-07,
      "loss": 0.3402,
      "reward": 0.5329595506191254,
      "reward_std": 0.24862322211265564,
      "reward_std_all": 0.2575627416372299,
      "rewards-max/forensics_sinkhorn_reward": 0.9849976599216461,
      "rewards-max/forensics_strict_edge_reward": 0.9375866651535034,
      "rewards-max/forensics_tiou_reward": 0.9842696487903595,
      "rewards/forensics_sinkhorn_reward": 0.42382316291332245,
      "rewards/forensics_strict_edge_reward": 0.19527367502450943,
      "rewards/forensics_tiou_reward": 0.2437010109424591,
      "step": 1329
    },
    {
      "completion_length": 62.265625,
      "epoch": 2.783246073298429,
      "grad_norm": 16.05907440185547,
      "kl": 7.921875,
      "learning_rate": 3.034591194968553e-07,
      "loss": 0.5275,
      "reward": 0.5720700025558472,
      "reward_std": 0.23279651999473572,
      "reward_std_all": 0.22836077958345413,
      "rewards-max/forensics_sinkhorn_reward": 0.9877665340900421,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9951765835285187,
      "rewards/forensics_sinkhorn_reward": 0.5026627779006958,
      "rewards/forensics_strict_edge_reward": 0.30596327781677246,
      "rewards/forensics_tiou_reward": 0.36410951614379883,
      "step": 1330
    },
    {
      "completion_length": 91.90625,
      "epoch": 2.7853403141361257,
      "grad_norm": 17.778759002685547,
      "kl": 9.9375,
      "learning_rate": 3.0293501048218027e-07,
      "loss": 0.5858,
      "reward": 0.5553786158561707,
      "reward_std": 0.22323264181613922,
      "reward_std_all": 0.2369002103805542,
      "rewards-max/forensics_sinkhorn_reward": 0.9885500967502594,
      "rewards-max/forensics_strict_edge_reward": 0.9890114367008209,
      "rewards-max/forensics_tiou_reward": 0.9982653856277466,
      "rewards/forensics_sinkhorn_reward": 0.4685271829366684,
      "rewards/forensics_strict_edge_reward": 0.26143306493759155,
      "rewards/forensics_tiou_reward": 0.34612831473350525,
      "step": 1331
    },
    {
      "completion_length": 39.59375,
      "epoch": 2.7874345549738218,
      "grad_norm": 18.890331268310547,
      "kl": 5.34375,
      "learning_rate": 3.0241090146750523e-07,
      "loss": 0.2847,
      "reward": 0.6199979186058044,
      "reward_std": 0.19198968261480331,
      "reward_std_all": 0.19228330999612808,
      "rewards-max/forensics_sinkhorn_reward": 0.9884204268455505,
      "rewards-max/forensics_strict_edge_reward": 0.9890114367008209,
      "rewards-max/forensics_tiou_reward": 0.9977528154850006,
      "rewards/forensics_sinkhorn_reward": 0.5496630668640137,
      "rewards/forensics_strict_edge_reward": 0.3246176391839981,
      "rewards/forensics_tiou_reward": 0.3626185804605484,
      "step": 1332
    },
    {
      "completion_length": 42.09375,
      "epoch": 2.7895287958115182,
      "grad_norm": 17.790634155273438,
      "kl": 5.984375,
      "learning_rate": 3.018867924528302e-07,
      "loss": 0.3656,
      "reward": 0.6202479302883148,
      "reward_std": 0.16310671716928482,
      "reward_std_all": 0.178282730281353,
      "rewards-max/forensics_sinkhorn_reward": 0.9777273833751678,
      "rewards-max/forensics_strict_edge_reward": 0.9357379674911499,
      "rewards-max/forensics_tiou_reward": 0.9738377332687378,
      "rewards/forensics_sinkhorn_reward": 0.4775436967611313,
      "rewards/forensics_strict_edge_reward": 0.24524562805891037,
      "rewards/forensics_tiou_reward": 0.2733565419912338,
      "step": 1333
    },
    {
      "completion_length": 68.109375,
      "epoch": 2.7916230366492147,
      "grad_norm": 25.49362564086914,
      "kl": 8.03125,
      "learning_rate": 3.0136268343815515e-07,
      "loss": 0.6038,
      "reward": 0.5431192219257355,
      "reward_std": 0.25698527693748474,
      "reward_std_all": 0.2654934823513031,
      "rewards-max/forensics_sinkhorn_reward": 0.987747311592102,
      "rewards-max/forensics_strict_edge_reward": 0.9567649364471436,
      "rewards-max/forensics_tiou_reward": 0.9950995743274689,
      "rewards/forensics_sinkhorn_reward": 0.4895147979259491,
      "rewards/forensics_strict_edge_reward": 0.3249963968992233,
      "rewards/forensics_tiou_reward": 0.37826021015644073,
      "step": 1334
    },
    {
      "completion_length": 45.625,
      "epoch": 2.793717277486911,
      "grad_norm": 20.623998641967773,
      "kl": 5.734375,
      "learning_rate": 3.008385744234801e-07,
      "loss": 0.4227,
      "reward": 0.6110557615756989,
      "reward_std": 0.2471616417169571,
      "reward_std_all": 0.22764630615711212,
      "rewards-max/forensics_sinkhorn_reward": 0.980085015296936,
      "rewards-max/forensics_strict_edge_reward": 0.8573527932167053,
      "rewards-max/forensics_tiou_reward": 0.9780684411525726,
      "rewards/forensics_sinkhorn_reward": 0.5152172893285751,
      "rewards/forensics_strict_edge_reward": 0.2946293354034424,
      "rewards/forensics_tiou_reward": 0.36412563920021057,
      "step": 1335
    },
    {
      "completion_length": 41.140625,
      "epoch": 2.7958115183246073,
      "grad_norm": 18.777767181396484,
      "kl": 4.09375,
      "learning_rate": 3.00314465408805e-07,
      "loss": 0.3926,
      "reward": 0.6168049275875092,
      "reward_std": 0.1870252937078476,
      "reward_std_all": 0.19142255187034607,
      "rewards-max/forensics_sinkhorn_reward": 0.9885142743587494,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.998123437166214,
      "rewards/forensics_sinkhorn_reward": 0.4964062124490738,
      "rewards/forensics_strict_edge_reward": 0.3085639625787735,
      "rewards/forensics_tiou_reward": 0.3318643420934677,
      "step": 1336
    },
    {
      "completion_length": 39.796875,
      "epoch": 2.7979057591623038,
      "grad_norm": 21.18354606628418,
      "kl": 5.734375,
      "learning_rate": 2.9979035639412997e-07,
      "loss": 0.3844,
      "reward": 0.6119330823421478,
      "reward_std": 0.223244346678257,
      "reward_std_all": 0.2160612717270851,
      "rewards-max/forensics_sinkhorn_reward": 0.9882974326610565,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9972686767578125,
      "rewards/forensics_sinkhorn_reward": 0.5127495974302292,
      "rewards/forensics_strict_edge_reward": 0.31849417090415955,
      "rewards/forensics_tiou_reward": 0.3673195242881775,
      "step": 1337
    },
    {
      "completion_length": 44.359375,
      "epoch": 2.8,
      "grad_norm": 20.628225326538086,
      "kl": 7.390625,
      "learning_rate": 2.9926624737945493e-07,
      "loss": 0.4848,
      "reward": 0.5798866152763367,
      "reward_std": 0.24443408846855164,
      "reward_std_all": 0.24533645063638687,
      "rewards-max/forensics_sinkhorn_reward": 0.9623684287071228,
      "rewards-max/forensics_strict_edge_reward": 0.8983768224716187,
      "rewards-max/forensics_tiou_reward": 0.9416195154190063,
      "rewards/forensics_sinkhorn_reward": 0.508405789732933,
      "rewards/forensics_strict_edge_reward": 0.3020388334989548,
      "rewards/forensics_tiou_reward": 0.37466655671596527,
      "step": 1338
    },
    {
      "completion_length": 43.515625,
      "epoch": 2.8020942408376963,
      "grad_norm": 26.701618194580078,
      "kl": 7.109375,
      "learning_rate": 2.9874213836477983e-07,
      "loss": 0.4435,
      "reward": 0.582472562789917,
      "reward_std": 0.10852385312318802,
      "reward_std_all": 0.16429872065782547,
      "rewards-max/forensics_sinkhorn_reward": 0.987425684928894,
      "rewards-max/forensics_strict_edge_reward": 0.9677534997463226,
      "rewards-max/forensics_tiou_reward": 0.9938325881958008,
      "rewards/forensics_sinkhorn_reward": 0.4941737800836563,
      "rewards/forensics_strict_edge_reward": 0.29428625106811523,
      "rewards/forensics_tiou_reward": 0.3305477052927017,
      "step": 1339
    },
    {
      "completion_length": 64.03125,
      "epoch": 2.804188481675393,
      "grad_norm": 14.29396915435791,
      "kl": 7.015625,
      "learning_rate": 2.982180293501048e-07,
      "loss": 0.5238,
      "reward": 0.5401595085859299,
      "reward_std": 0.2318214252591133,
      "reward_std_all": 0.24881816655397415,
      "rewards-max/forensics_sinkhorn_reward": 0.9402118027210236,
      "rewards-max/forensics_strict_edge_reward": 0.8320213258266449,
      "rewards-max/forensics_tiou_reward": 0.8952057957649231,
      "rewards/forensics_sinkhorn_reward": 0.41388365626335144,
      "rewards/forensics_strict_edge_reward": 0.2461504265666008,
      "rewards/forensics_tiou_reward": 0.2685633525252342,
      "step": 1340
    },
    {
      "completion_length": 38.703125,
      "epoch": 2.806282722513089,
      "grad_norm": 19.255950927734375,
      "kl": 4.09375,
      "learning_rate": 2.976939203354298e-07,
      "loss": 0.3257,
      "reward": 0.6172360181808472,
      "reward_std": 0.20361243933439255,
      "reward_std_all": 0.20362938195466995,
      "rewards-max/forensics_sinkhorn_reward": 0.9884685277938843,
      "rewards-max/forensics_strict_edge_reward": 0.9890114367008209,
      "rewards-max/forensics_tiou_reward": 0.9979445040225983,
      "rewards/forensics_sinkhorn_reward": 0.548259973526001,
      "rewards/forensics_strict_edge_reward": 0.30433791130781174,
      "rewards/forensics_tiou_reward": 0.35395048558712006,
      "step": 1341
    },
    {
      "completion_length": 31.75,
      "epoch": 2.8083769633507853,
      "grad_norm": 25.1273250579834,
      "kl": 5.5625,
      "learning_rate": 2.971698113207547e-07,
      "loss": 0.365,
      "reward": 0.6101548969745636,
      "reward_std": 0.1990819051861763,
      "reward_std_all": 0.19518087804317474,
      "rewards-max/forensics_sinkhorn_reward": 0.9566424190998077,
      "rewards-max/forensics_strict_edge_reward": 0.8277941644191742,
      "rewards-max/forensics_tiou_reward": 0.9285508990287781,
      "rewards/forensics_sinkhorn_reward": 0.46894361078739166,
      "rewards/forensics_strict_edge_reward": 0.21107559651136398,
      "rewards/forensics_tiou_reward": 0.2832081764936447,
      "step": 1342
    },
    {
      "completion_length": 46.921875,
      "epoch": 2.810471204188482,
      "grad_norm": 24.002498626708984,
      "kl": 8.375,
      "learning_rate": 2.9664570230607967e-07,
      "loss": 0.5219,
      "reward": 0.5973129868507385,
      "reward_std": 0.23216912895441055,
      "reward_std_all": 0.22127756476402283,
      "rewards-max/forensics_sinkhorn_reward": 0.9857907295227051,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.991890162229538,
      "rewards/forensics_sinkhorn_reward": 0.5306534469127655,
      "rewards/forensics_strict_edge_reward": 0.32092732191085815,
      "rewards/forensics_tiou_reward": 0.376359760761261,
      "step": 1343
    },
    {
      "completion_length": 52.953125,
      "epoch": 2.812565445026178,
      "grad_norm": 21.884401321411133,
      "kl": 6.296875,
      "learning_rate": 2.9612159329140463e-07,
      "loss": 0.4039,
      "reward": 0.6333035826683044,
      "reward_std": 0.1545981541275978,
      "reward_std_all": 0.16780821233987808,
      "rewards-max/forensics_sinkhorn_reward": 0.9877356886863708,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9950621724128723,
      "rewards/forensics_sinkhorn_reward": 0.5440214276313782,
      "rewards/forensics_strict_edge_reward": 0.3316922187805176,
      "rewards/forensics_tiou_reward": 0.41595759987831116,
      "step": 1344
    },
    {
      "completion_length": 36.140625,
      "epoch": 2.8146596858638744,
      "grad_norm": 21.621057510375977,
      "kl": 4.2421875,
      "learning_rate": 2.9559748427672953e-07,
      "loss": 0.3529,
      "reward": 0.5630923807621002,
      "reward_std": 0.21474214643239975,
      "reward_std_all": 0.23037239909172058,
      "rewards-max/forensics_sinkhorn_reward": 0.988766998052597,
      "rewards-max/forensics_strict_edge_reward": 0.9890114367008209,
      "rewards-max/forensics_tiou_reward": 0.9991189539432526,
      "rewards/forensics_sinkhorn_reward": 0.5183054804801941,
      "rewards/forensics_strict_edge_reward": 0.3168201521039009,
      "rewards/forensics_tiou_reward": 0.36421963572502136,
      "step": 1345
    },
    {
      "completion_length": 66.125,
      "epoch": 2.816753926701571,
      "grad_norm": 16.611045837402344,
      "kl": 7.75,
      "learning_rate": 2.950733752620545e-07,
      "loss": 0.4535,
      "reward": 0.5479705333709717,
      "reward_std": 0.22062601149082184,
      "reward_std_all": 0.23058895766735077,
      "rewards-max/forensics_sinkhorn_reward": 0.9338798820972443,
      "rewards-max/forensics_strict_edge_reward": 0.8133637309074402,
      "rewards-max/forensics_tiou_reward": 0.8829990923404694,
      "rewards/forensics_sinkhorn_reward": 0.4461563676595688,
      "rewards/forensics_strict_edge_reward": 0.1922537237405777,
      "rewards/forensics_tiou_reward": 0.25164514780044556,
      "step": 1346
    },
    {
      "completion_length": 76.6875,
      "epoch": 2.818848167539267,
      "grad_norm": 20.125850677490234,
      "kl": 6.234375,
      "learning_rate": 2.9454926624737945e-07,
      "loss": 0.4098,
      "reward": 0.5693325102329254,
      "reward_std": 0.22908758372068405,
      "reward_std_all": 0.22542710602283478,
      "rewards-max/forensics_sinkhorn_reward": 0.9873475134372711,
      "rewards-max/forensics_strict_edge_reward": 0.9565287232398987,
      "rewards-max/forensics_tiou_reward": 0.9935237169265747,
      "rewards/forensics_sinkhorn_reward": 0.5106285214424133,
      "rewards/forensics_strict_edge_reward": 0.2877522110939026,
      "rewards/forensics_tiou_reward": 0.35914796590805054,
      "step": 1347
    },
    {
      "completion_length": 62.3125,
      "epoch": 2.8209424083769634,
      "grad_norm": 16.26201629638672,
      "kl": 7.03125,
      "learning_rate": 2.9402515723270436e-07,
      "loss": 0.4072,
      "reward": 0.600204199552536,
      "reward_std": 0.20549985021352768,
      "reward_std_all": 0.20049618184566498,
      "rewards-max/forensics_sinkhorn_reward": 0.9881440997123718,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9966652989387512,
      "rewards/forensics_sinkhorn_reward": 0.4937979280948639,
      "rewards/forensics_strict_edge_reward": 0.2740931734442711,
      "rewards/forensics_tiou_reward": 0.33197709918022156,
      "step": 1348
    },
    {
      "completion_length": 45.421875,
      "epoch": 2.82303664921466,
      "grad_norm": 16.626569747924805,
      "kl": 5.09375,
      "learning_rate": 2.9350104821802937e-07,
      "loss": 0.2482,
      "reward": 0.5976144075393677,
      "reward_std": 0.20401443541049957,
      "reward_std_all": 0.20923113077878952,
      "rewards-max/forensics_sinkhorn_reward": 0.9435722231864929,
      "rewards-max/forensics_strict_edge_reward": 0.8188429176807404,
      "rewards-max/forensics_tiou_reward": 0.9000403583049774,
      "rewards/forensics_sinkhorn_reward": 0.5213905870914459,
      "rewards/forensics_strict_edge_reward": 0.24246925115585327,
      "rewards/forensics_tiou_reward": 0.33251817524433136,
      "step": 1349
    },
    {
      "completion_length": 46.4375,
      "epoch": 2.825130890052356,
      "grad_norm": 20.447982788085938,
      "kl": 6.703125,
      "learning_rate": 2.9297693920335433e-07,
      "loss": 0.4254,
      "reward": 0.6066146492958069,
      "reward_std": 0.20390845090150833,
      "reward_std_all": 0.20340051501989365,
      "rewards-max/forensics_sinkhorn_reward": 0.9285193681716919,
      "rewards-max/forensics_strict_edge_reward": 0.7926336228847504,
      "rewards-max/forensics_tiou_reward": 0.8654478192329407,
      "rewards/forensics_sinkhorn_reward": 0.47654275596141815,
      "rewards/forensics_strict_edge_reward": 0.21796011924743652,
      "rewards/forensics_tiou_reward": 0.2701951265335083,
      "step": 1350
    },
    {
      "completion_length": 55.953125,
      "epoch": 2.8272251308900525,
      "grad_norm": 17.166086196899414,
      "kl": 8.28125,
      "learning_rate": 2.9245283018867923e-07,
      "loss": 0.523,
      "reward": 0.5782141983509064,
      "reward_std": 0.26838889718055725,
      "reward_std_all": 0.25382089614868164,
      "rewards-max/forensics_sinkhorn_reward": 0.9872046411037445,
      "rewards-max/forensics_strict_edge_reward": 0.9464850425720215,
      "rewards-max/forensics_tiou_reward": 0.9929777085781097,
      "rewards/forensics_sinkhorn_reward": 0.48392510414123535,
      "rewards/forensics_strict_edge_reward": 0.2932697683572769,
      "rewards/forensics_tiou_reward": 0.34540796279907227,
      "step": 1351
    },
    {
      "completion_length": 44.671875,
      "epoch": 2.829319371727749,
      "grad_norm": 22.063570022583008,
      "kl": 6.15625,
      "learning_rate": 2.919287211740042e-07,
      "loss": 0.3613,
      "reward": 0.5715661644935608,
      "reward_std": 0.17772610485553741,
      "reward_std_all": 0.19037501141428947,
      "rewards-max/forensics_sinkhorn_reward": 0.9877093732357025,
      "rewards-max/forensics_strict_edge_reward": 0.9565287232398987,
      "rewards-max/forensics_tiou_reward": 0.9949526190757751,
      "rewards/forensics_sinkhorn_reward": 0.4638409912586212,
      "rewards/forensics_strict_edge_reward": 0.258044958114624,
      "rewards/forensics_tiou_reward": 0.26754825562238693,
      "step": 1352
    },
    {
      "completion_length": 32.53125,
      "epoch": 2.831413612565445,
      "grad_norm": 24.598352432250977,
      "kl": 6.703125,
      "learning_rate": 2.914046121593291e-07,
      "loss": 0.4677,
      "reward": 0.5752487778663635,
      "reward_std": 0.19689160585403442,
      "reward_std_all": 0.21248649060726166,
      "rewards-max/forensics_sinkhorn_reward": 0.9885275661945343,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9981758296489716,
      "rewards/forensics_sinkhorn_reward": 0.5339240729808807,
      "rewards/forensics_strict_edge_reward": 0.2911630868911743,
      "rewards/forensics_tiou_reward": 0.3528725504875183,
      "step": 1353
    },
    {
      "completion_length": 51.921875,
      "epoch": 2.8335078534031415,
      "grad_norm": 21.461273193359375,
      "kl": 6.671875,
      "learning_rate": 2.9088050314465406e-07,
      "loss": 0.4982,
      "reward": 0.6657275557518005,
      "reward_std": 0.18948683887720108,
      "reward_std_all": 0.17230583727359772,
      "rewards-max/forensics_sinkhorn_reward": 0.9885486662387848,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9982582032680511,
      "rewards/forensics_sinkhorn_reward": 0.5787135064601898,
      "rewards/forensics_strict_edge_reward": 0.3978540599346161,
      "rewards/forensics_tiou_reward": 0.43697279691696167,
      "step": 1354
    },
    {
      "completion_length": 49.921875,
      "epoch": 2.8356020942408375,
      "grad_norm": 21.667951583862305,
      "kl": 6.921875,
      "learning_rate": 2.90356394129979e-07,
      "loss": 0.457,
      "reward": 0.6033447980880737,
      "reward_std": 0.21810464560985565,
      "reward_std_all": 0.21997518092393875,
      "rewards-max/forensics_sinkhorn_reward": 0.9853854179382324,
      "rewards-max/forensics_strict_edge_reward": 0.8719756007194519,
      "rewards-max/forensics_tiou_reward": 0.9858815371990204,
      "rewards/forensics_sinkhorn_reward": 0.5124164521694183,
      "rewards/forensics_strict_edge_reward": 0.31246958673000336,
      "rewards/forensics_tiou_reward": 0.3695552200078964,
      "step": 1355
    },
    {
      "completion_length": 58.515625,
      "epoch": 2.837696335078534,
      "grad_norm": 18.98493766784668,
      "kl": 4.6796875,
      "learning_rate": 2.898322851153039e-07,
      "loss": 0.3412,
      "reward": 0.5819854140281677,
      "reward_std": 0.1784091666340828,
      "reward_std_all": 0.19618623703718185,
      "rewards-max/forensics_sinkhorn_reward": 0.9862165749073029,
      "rewards-max/forensics_strict_edge_reward": 0.9565287232398987,
      "rewards-max/forensics_tiou_reward": 0.9891084730625153,
      "rewards/forensics_sinkhorn_reward": 0.5561883747577667,
      "rewards/forensics_strict_edge_reward": 0.32873156666755676,
      "rewards/forensics_tiou_reward": 0.40207983553409576,
      "step": 1356
    },
    {
      "completion_length": 71.890625,
      "epoch": 2.8397905759162305,
      "grad_norm": 23.35491180419922,
      "kl": 7.5,
      "learning_rate": 2.8930817610062893e-07,
      "loss": 0.5165,
      "reward": 0.5093870759010315,
      "reward_std": 0.2686862349510193,
      "reward_std_all": 0.2729345113039017,
      "rewards-max/forensics_sinkhorn_reward": 0.9885500967502594,
      "rewards-max/forensics_strict_edge_reward": 0.9890114367008209,
      "rewards-max/forensics_tiou_reward": 0.9982653856277466,
      "rewards/forensics_sinkhorn_reward": 0.4503367245197296,
      "rewards/forensics_strict_edge_reward": 0.2938931882381439,
      "rewards/forensics_tiou_reward": 0.31322692334651947,
      "step": 1357
    },
    {
      "completion_length": 81.09375,
      "epoch": 2.8418848167539266,
      "grad_norm": 25.826297760009766,
      "kl": 6.609375,
      "learning_rate": 2.887840670859539e-07,
      "loss": 0.4201,
      "reward": 0.5915236175060272,
      "reward_std": 0.2389901503920555,
      "reward_std_all": 0.22596947848796844,
      "rewards-max/forensics_sinkhorn_reward": 0.9879440069198608,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9958779513835907,
      "rewards/forensics_sinkhorn_reward": 0.5413879603147507,
      "rewards/forensics_strict_edge_reward": 0.360088974237442,
      "rewards/forensics_tiou_reward": 0.42259082198143005,
      "step": 1358
    },
    {
      "completion_length": 72.40625,
      "epoch": 2.843979057591623,
      "grad_norm": 18.398935317993164,
      "kl": 6.75,
      "learning_rate": 2.882599580712788e-07,
      "loss": 0.196,
      "reward": 0.4956316351890564,
      "reward_std": 0.23274309933185577,
      "reward_std_all": 0.27067314833402634,
      "rewards-max/forensics_sinkhorn_reward": 0.9885163307189941,
      "rewards-max/forensics_strict_edge_reward": 0.9890114367008209,
      "rewards-max/forensics_tiou_reward": 0.9981308281421661,
      "rewards/forensics_sinkhorn_reward": 0.38550643622875214,
      "rewards/forensics_strict_edge_reward": 0.19747474789619446,
      "rewards/forensics_tiou_reward": 0.21318057924509048,
      "step": 1359
    },
    {
      "completion_length": 49.28125,
      "epoch": 2.846073298429319,
      "grad_norm": 11.454885482788086,
      "kl": 4.609375,
      "learning_rate": 2.8773584905660376e-07,
      "loss": 0.4271,
      "reward": 0.6157549023628235,
      "reward_std": 0.13192427903413773,
      "reward_std_all": 0.15918008238077164,
      "rewards-max/forensics_sinkhorn_reward": 0.9335589706897736,
      "rewards-max/forensics_strict_edge_reward": 0.8078543543815613,
      "rewards-max/forensics_tiou_reward": 0.8778645098209381,
      "rewards/forensics_sinkhorn_reward": 0.5228718370199203,
      "rewards/forensics_strict_edge_reward": 0.2097935602068901,
      "rewards/forensics_tiou_reward": 0.3081718385219574,
      "step": 1360
    },
    {
      "completion_length": 45.65625,
      "epoch": 2.8481675392670156,
      "grad_norm": 23.9539794921875,
      "kl": 4.0234375,
      "learning_rate": 2.872117400419287e-07,
      "loss": 0.3278,
      "reward": 0.6488936245441437,
      "reward_std": 0.1820726841688156,
      "reward_std_all": 0.17198903858661652,
      "rewards-max/forensics_sinkhorn_reward": 0.9884198606014252,
      "rewards-max/forensics_strict_edge_reward": 0.9464850425720215,
      "rewards-max/forensics_tiou_reward": 0.9977510869503021,
      "rewards/forensics_sinkhorn_reward": 0.5747836530208588,
      "rewards/forensics_strict_edge_reward": 0.3563009351491928,
      "rewards/forensics_tiou_reward": 0.4181113690137863,
      "step": 1361
    },
    {
      "completion_length": 92.28125,
      "epoch": 2.850261780104712,
      "grad_norm": 26.352733612060547,
      "kl": 8.875,
      "learning_rate": 2.866876310272536e-07,
      "loss": 0.637,
      "reward": 0.4642772376537323,
      "reward_std": 0.24758580327033997,
      "reward_std_all": 0.2901323437690735,
      "rewards-max/forensics_sinkhorn_reward": 0.9602549970149994,
      "rewards-max/forensics_strict_edge_reward": 0.8668389916419983,
      "rewards-max/forensics_tiou_reward": 0.9373038113117218,
      "rewards/forensics_sinkhorn_reward": 0.3912799805402756,
      "rewards/forensics_strict_edge_reward": 0.18264901638031006,
      "rewards/forensics_tiou_reward": 0.23821648955345154,
      "step": 1362
    },
    {
      "completion_length": 68.46875,
      "epoch": 2.852356020942408,
      "grad_norm": 18.243505477905273,
      "kl": 5.75,
      "learning_rate": 2.861635220125786e-07,
      "loss": 0.5424,
      "reward": 0.58837890625,
      "reward_std": 0.27654044330120087,
      "reward_std_all": 0.2538336366415024,
      "rewards-max/forensics_sinkhorn_reward": 0.9773142635822296,
      "rewards-max/forensics_strict_edge_reward": 0.8805811703205109,
      "rewards-max/forensics_tiou_reward": 0.9726038873195648,
      "rewards/forensics_sinkhorn_reward": 0.509277269244194,
      "rewards/forensics_strict_edge_reward": 0.31362731009721756,
      "rewards/forensics_tiou_reward": 0.3756791204214096,
      "step": 1363
    },
    {
      "completion_length": 54.375,
      "epoch": 2.8544502617801046,
      "grad_norm": 23.44729232788086,
      "kl": 4.5234375,
      "learning_rate": 2.8563941299790354e-07,
      "loss": 0.4367,
      "reward": 0.5678999871015549,
      "reward_std": 0.17791210114955902,
      "reward_std_all": 0.20446323603391647,
      "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": 0.4928680658340454,
      "rewards/forensics_strict_edge_reward": 0.2945331186056137,
      "rewards/forensics_tiou_reward": 0.3372521251440048,
      "step": 1364
    },
    {
      "completion_length": 55.890625,
      "epoch": 2.856544502617801,
      "grad_norm": 15.514001846313477,
      "kl": 7.453125,
      "learning_rate": 2.851153039832285e-07,
      "loss": 0.5849,
      "reward": 0.5509395003318787,
      "reward_std": 0.2260921448469162,
      "reward_std_all": 0.2439383640885353,
      "rewards-max/forensics_sinkhorn_reward": 0.9841863214969635,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9895786345005035,
      "rewards/forensics_sinkhorn_reward": 0.4921395033597946,
      "rewards/forensics_strict_edge_reward": 0.33897267282009125,
      "rewards/forensics_tiou_reward": 0.38215284049510956,
      "step": 1365
    },
    {
      "completion_length": 63.796875,
      "epoch": 2.858638743455497,
      "grad_norm": 20.85346794128418,
      "kl": 7.28125,
      "learning_rate": 2.8459119496855345e-07,
      "loss": 0.4833,
      "reward": 0.5781612694263458,
      "reward_std": 0.20889290422201157,
      "reward_std_all": 0.2157329097390175,
      "rewards-max/forensics_sinkhorn_reward": 0.8256183862686157,
      "rewards-max/forensics_strict_edge_reward": 0.7680542469024658,
      "rewards-max/forensics_tiou_reward": 0.7941464483737946,
      "rewards/forensics_sinkhorn_reward": 0.46388451755046844,
      "rewards/forensics_strict_edge_reward": 0.2721472233533859,
      "rewards/forensics_tiou_reward": 0.317277267575264,
      "step": 1366
    },
    {
      "completion_length": 47.234375,
      "epoch": 2.8607329842931937,
      "grad_norm": 18.987642288208008,
      "kl": 4.734375,
      "learning_rate": 2.840670859538784e-07,
      "loss": 0.3477,
      "reward": 0.5343583524227142,
      "reward_std": 0.22824648767709732,
      "reward_std_all": 0.2441493198275566,
      "rewards-max/forensics_sinkhorn_reward": 0.987107902765274,
      "rewards-max/forensics_strict_edge_reward": 0.9567649364471436,
      "rewards-max/forensics_tiou_reward": 0.9926294982433319,
      "rewards/forensics_sinkhorn_reward": 0.4645397216081619,
      "rewards/forensics_strict_edge_reward": 0.2585115283727646,
      "rewards/forensics_tiou_reward": 0.30392301082611084,
      "step": 1367
    },
    {
      "completion_length": 65.515625,
      "epoch": 2.86282722513089,
      "grad_norm": 16.501989364624023,
      "kl": 6.703125,
      "learning_rate": 2.835429769392033e-07,
      "loss": 0.3329,
      "reward": 0.5502490401268005,
      "reward_std": 0.2614169865846634,
      "reward_std_all": 0.26190054416656494,
      "rewards-max/forensics_sinkhorn_reward": 0.9870212376117706,
      "rewards-max/forensics_strict_edge_reward": 0.9062341153621674,
      "rewards-max/forensics_tiou_reward": 0.9922494888305664,
      "rewards/forensics_sinkhorn_reward": 0.4939242899417877,
      "rewards/forensics_strict_edge_reward": 0.26435884833335876,
      "rewards/forensics_tiou_reward": 0.34824252128601074,
      "step": 1368
    },
    {
      "completion_length": 82.96875,
      "epoch": 2.864921465968586,
      "grad_norm": 35.28187942504883,
      "kl": 10.25,
      "learning_rate": 2.830188679245283e-07,
      "loss": 0.6228,
      "reward": 0.4191201329231262,
      "reward_std": 0.22347354888916016,
      "reward_std_all": 0.29027532041072845,
      "rewards-max/forensics_sinkhorn_reward": 0.909047931432724,
      "rewards-max/forensics_strict_edge_reward": 0.5769832506775856,
      "rewards-max/forensics_tiou_reward": 0.8360873460769653,
      "rewards/forensics_sinkhorn_reward": 0.2995830476284027,
      "rewards/forensics_strict_edge_reward": 0.059905603528022766,
      "rewards/forensics_tiou_reward": 0.13861799240112305,
      "step": 1369
    },
    {
      "completion_length": 79.828125,
      "epoch": 2.8670157068062827,
      "grad_norm": 76.15958404541016,
      "kl": 9.4375,
      "learning_rate": 2.8249475890985324e-07,
      "loss": 0.6355,
      "reward": 0.4905475080013275,
      "reward_std": 0.19985433667898178,
      "reward_std_all": 0.2536064237356186,
      "rewards-max/forensics_sinkhorn_reward": 0.981622576713562,
      "rewards-max/forensics_strict_edge_reward": 0.8404377698898315,
      "rewards-max/forensics_tiou_reward": 0.9712882936000824,
      "rewards/forensics_sinkhorn_reward": 0.452422633767128,
      "rewards/forensics_strict_edge_reward": 0.18871580809354782,
      "rewards/forensics_tiou_reward": 0.29994285106658936,
      "step": 1370
    },
    {
      "completion_length": 56.03125,
      "epoch": 2.869109947643979,
      "grad_norm": 19.77923583984375,
      "kl": 8.59375,
      "learning_rate": 2.8197064989517814e-07,
      "loss": 0.5221,
      "reward": 0.5153457522392273,
      "reward_std": 0.31865517795085907,
      "reward_std_all": 0.30555596947669983,
      "rewards-max/forensics_sinkhorn_reward": 0.9604323208332062,
      "rewards-max/forensics_strict_edge_reward": 0.8877308070659637,
      "rewards-max/forensics_tiou_reward": 0.9326680302619934,
      "rewards/forensics_sinkhorn_reward": 0.4139277786016464,
      "rewards/forensics_strict_edge_reward": 0.29554393887519836,
      "rewards/forensics_tiou_reward": 0.3055959641933441,
      "step": 1371
    },
    {
      "completion_length": 58.8125,
      "epoch": 2.8712041884816752,
      "grad_norm": 35.49671173095703,
      "kl": 9.109375,
      "learning_rate": 2.814465408805031e-07,
      "loss": 0.5457,
      "reward": 0.5631524324417114,
      "reward_std": 0.2707916647195816,
      "reward_std_all": 0.26222164928913116,
      "rewards-max/forensics_sinkhorn_reward": 0.9702070951461792,
      "rewards-max/forensics_strict_edge_reward": 0.87538942694664,
      "rewards-max/forensics_tiou_reward": 0.9552817940711975,
      "rewards/forensics_sinkhorn_reward": 0.4993681013584137,
      "rewards/forensics_strict_edge_reward": 0.2694967985153198,
      "rewards/forensics_tiou_reward": 0.32349297404289246,
      "step": 1372
    },
    {
      "completion_length": 75.421875,
      "epoch": 2.8732984293193717,
      "grad_norm": 25.412918090820312,
      "kl": 10.5625,
      "learning_rate": 2.809224318658281e-07,
      "loss": 0.6577,
      "reward": 0.5052342116832733,
      "reward_std": 0.22384757548570633,
      "reward_std_all": 0.25162288546562195,
      "rewards-max/forensics_sinkhorn_reward": 0.9874694645404816,
      "rewards-max/forensics_strict_edge_reward": 0.9464850425720215,
      "rewards-max/forensics_tiou_reward": 0.994004637002945,
      "rewards/forensics_sinkhorn_reward": 0.4349445551633835,
      "rewards/forensics_strict_edge_reward": 0.20872415602207184,
      "rewards/forensics_tiou_reward": 0.25821302086114883,
      "step": 1373
    },
    {
      "completion_length": 65.0,
      "epoch": 2.8753926701570682,
      "grad_norm": 25.423898696899414,
      "kl": 7.328125,
      "learning_rate": 2.80398322851153e-07,
      "loss": 0.5652,
      "reward": 0.603486031293869,
      "reward_std": 0.1988554149866104,
      "reward_std_all": 0.2082328274846077,
      "rewards-max/forensics_sinkhorn_reward": 0.9542986452579498,
      "rewards-max/forensics_strict_edge_reward": 0.8983768224716187,
      "rewards-max/forensics_tiou_reward": 0.9230732917785645,
      "rewards/forensics_sinkhorn_reward": 0.5166263282299042,
      "rewards/forensics_strict_edge_reward": 0.35469596087932587,
      "rewards/forensics_tiou_reward": 0.3944771885871887,
      "step": 1374
    },
    {
      "completion_length": 66.84375,
      "epoch": 2.8774869109947643,
      "grad_norm": 24.999574661254883,
      "kl": 9.4375,
      "learning_rate": 2.79874213836478e-07,
      "loss": 0.6586,
      "reward": 0.48438985645771027,
      "reward_std": 0.23237420618534088,
      "reward_std_all": 0.2737784832715988,
      "rewards-max/forensics_sinkhorn_reward": 0.9869841039180756,
      "rewards-max/forensics_strict_edge_reward": 0.9464850425720215,
      "rewards-max/forensics_tiou_reward": 0.9921203851699829,
      "rewards/forensics_sinkhorn_reward": 0.4191068559885025,
      "rewards/forensics_strict_edge_reward": 0.24012621492147446,
      "rewards/forensics_tiou_reward": 0.3148421496152878,
      "step": 1375
    },
    {
      "completion_length": 20.40625,
      "epoch": 2.8795811518324608,
      "grad_norm": 19.893217086791992,
      "kl": 3.75,
      "learning_rate": 2.7935010482180294e-07,
      "loss": 0.2175,
      "reward": 0.6491144895553589,
      "reward_std": 0.15588634461164474,
      "reward_std_all": 0.1529354266822338,
      "rewards-max/forensics_sinkhorn_reward": 0.9880951642990112,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9964739978313446,
      "rewards/forensics_sinkhorn_reward": 0.5615958869457245,
      "rewards/forensics_strict_edge_reward": 0.33594179153442383,
      "rewards/forensics_tiou_reward": 0.38960880041122437,
      "step": 1376
    },
    {
      "completion_length": 70.59375,
      "epoch": 2.8816753926701573,
      "grad_norm": 29.56081199645996,
      "kl": 9.84375,
      "learning_rate": 2.7882599580712784e-07,
      "loss": 0.5873,
      "reward": 0.479453906416893,
      "reward_std": 0.2461947426199913,
      "reward_std_all": 0.2842007726430893,
      "rewards-max/forensics_sinkhorn_reward": 0.9875406622886658,
      "rewards-max/forensics_strict_edge_reward": 0.9782643616199493,
      "rewards-max/forensics_tiou_reward": 0.9942857027053833,
      "rewards/forensics_sinkhorn_reward": 0.41540829837322235,
      "rewards/forensics_strict_edge_reward": 0.23684945702552795,
      "rewards/forensics_tiou_reward": 0.30113062262535095,
      "step": 1377
    },
    {
      "completion_length": 47.28125,
      "epoch": 2.8837696335078533,
      "grad_norm": 37.08782958984375,
      "kl": 5.265625,
      "learning_rate": 2.783018867924528e-07,
      "loss": 0.3589,
      "reward": 0.6147418022155762,
      "reward_std": 0.22069744765758514,
      "reward_std_all": 0.21277585625648499,
      "rewards-max/forensics_sinkhorn_reward": 0.9884204268455505,
      "rewards-max/forensics_strict_edge_reward": 0.9890114367008209,
      "rewards-max/forensics_tiou_reward": 0.9977528154850006,
      "rewards/forensics_sinkhorn_reward": 0.5461552739143372,
      "rewards/forensics_strict_edge_reward": 0.350205659866333,
      "rewards/forensics_tiou_reward": 0.3847314715385437,
      "step": 1378
    },
    {
      "completion_length": 82.21875,
      "epoch": 2.88586387434555,
      "grad_norm": 16.889301300048828,
      "kl": 8.09375,
      "learning_rate": 2.7777777777777776e-07,
      "loss": 0.564,
      "reward": 0.5375858545303345,
      "reward_std": 0.34255650639533997,
      "reward_std_all": 0.3152378797531128,
      "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": 0.4432009160518646,
      "rewards/forensics_strict_edge_reward": 0.30989956855773926,
      "rewards/forensics_tiou_reward": 0.32149460911750793,
      "step": 1379
    },
    {
      "completion_length": 59.1875,
      "epoch": 2.8879581151832463,
      "grad_norm": 147.3479461669922,
      "kl": 11.375,
      "learning_rate": 2.7725366876310267e-07,
      "loss": 0.7185,
      "reward": 0.5767538249492645,
      "reward_std": 0.21088982373476028,
      "reward_std_all": 0.22907379269599915,
      "rewards-max/forensics_sinkhorn_reward": 0.9598978161811829,
      "rewards-max/forensics_strict_edge_reward": 0.8635591566562653,
      "rewards-max/forensics_tiou_reward": 0.9315694272518158,
      "rewards/forensics_sinkhorn_reward": 0.49867507815361023,
      "rewards/forensics_strict_edge_reward": 0.3264891803264618,
      "rewards/forensics_tiou_reward": 0.38201597332954407,
      "step": 1380
    },
    {
      "completion_length": 49.171875,
      "epoch": 2.8900523560209423,
      "grad_norm": 17.794885635375977,
      "kl": 5.328125,
      "learning_rate": 2.767295597484277e-07,
      "loss": 0.3745,
      "reward": 0.6285741925239563,
      "reward_std": 0.1435038074851036,
      "reward_std_all": 0.1550099104642868,
      "rewards-max/forensics_sinkhorn_reward": 0.9877210557460785,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9950027465820312,
      "rewards/forensics_sinkhorn_reward": 0.5559320151805878,
      "rewards/forensics_strict_edge_reward": 0.32852721214294434,
      "rewards/forensics_tiou_reward": 0.4023655205965042,
      "step": 1381
    },
    {
      "completion_length": 59.921875,
      "epoch": 2.892146596858639,
      "grad_norm": 22.238855361938477,
      "kl": 9.09375,
      "learning_rate": 2.7620545073375264e-07,
      "loss": 0.4672,
      "reward": 0.5104639828205109,
      "reward_std": 0.222059927880764,
      "reward_std_all": 0.2516903430223465,
      "rewards-max/forensics_sinkhorn_reward": 0.9877868890762329,
      "rewards-max/forensics_strict_edge_reward": 0.9252271056175232,
      "rewards-max/forensics_tiou_reward": 0.9952556192874908,
      "rewards/forensics_sinkhorn_reward": 0.4198513925075531,
      "rewards/forensics_strict_edge_reward": 0.17538642138242722,
      "rewards/forensics_tiou_reward": 0.23888269066810608,
      "step": 1382
    },
    {
      "completion_length": 82.328125,
      "epoch": 2.8942408376963353,
      "grad_norm": 33.110408782958984,
      "kl": 8.859375,
      "learning_rate": 2.7568134171907754e-07,
      "loss": 0.5621,
      "reward": 0.4814856946468353,
      "reward_std": 0.2768539488315582,
      "reward_std_all": 0.29429425299167633,
      "rewards-max/forensics_sinkhorn_reward": 0.9869535565376282,
      "rewards-max/forensics_strict_edge_reward": 0.9565287232398987,
      "rewards-max/forensics_tiou_reward": 0.9919711053371429,
      "rewards/forensics_sinkhorn_reward": 0.38368038833141327,
      "rewards/forensics_strict_edge_reward": 0.22188662737607956,
      "rewards/forensics_tiou_reward": 0.24530483782291412,
      "step": 1383
    },
    {
      "completion_length": 99.59375,
      "epoch": 2.8963350785340314,
      "grad_norm": 20.827682495117188,
      "kl": 10.75,
      "learning_rate": 2.751572327044025e-07,
      "loss": 0.7743,
      "reward": 0.49806734919548035,
      "reward_std": 0.3297322690486908,
      "reward_std_all": 0.3260404020547867,
      "rewards-max/forensics_sinkhorn_reward": 0.9883696734905243,
      "rewards-max/forensics_strict_edge_reward": 0.9567649364471436,
      "rewards-max/forensics_tiou_reward": 0.9975549280643463,
      "rewards/forensics_sinkhorn_reward": 0.4349494129419327,
      "rewards/forensics_strict_edge_reward": 0.29411303997039795,
      "rewards/forensics_tiou_reward": 0.333419531583786,
      "step": 1384
    },
    {
      "completion_length": 44.015625,
      "epoch": 2.898429319371728,
      "grad_norm": 7623.33544921875,
      "kl": 83.65625,
      "learning_rate": 2.7463312368972746e-07,
      "loss": 3.5772,
      "reward": 0.5296094119548798,
      "reward_std": 0.28361064940690994,
      "reward_std_all": 0.28507499396800995,
      "rewards-max/forensics_sinkhorn_reward": 0.9859474897384644,
      "rewards-max/forensics_strict_edge_reward": 0.9565287232398987,
      "rewards-max/forensics_tiou_reward": 0.988073080778122,
      "rewards/forensics_sinkhorn_reward": 0.4374076724052429,
      "rewards/forensics_strict_edge_reward": 0.21660243719816208,
      "rewards/forensics_tiou_reward": 0.32145628333091736,
      "step": 1385
    },
    {
      "completion_length": 48.375,
      "epoch": 2.900523560209424,
      "grad_norm": 20.76363182067871,
      "kl": 3.84375,
      "learning_rate": 2.7410901467505236e-07,
      "loss": 0.3392,
      "reward": 0.5679864287376404,
      "reward_std": 0.25571437180042267,
      "reward_std_all": 0.24552933126688004,
      "rewards-max/forensics_sinkhorn_reward": 0.9882552921772003,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9971019923686981,
      "rewards/forensics_sinkhorn_reward": 0.5338945537805557,
      "rewards/forensics_strict_edge_reward": 0.35103701055049896,
      "rewards/forensics_tiou_reward": 0.39253199100494385,
      "step": 1386
    },
    {
      "completion_length": 79.640625,
      "epoch": 2.9026178010471204,
      "grad_norm": 15.744552612304688,
      "kl": 8.40625,
      "learning_rate": 2.735849056603773e-07,
      "loss": 0.5416,
      "reward": 0.5390357375144958,
      "reward_std": 0.2631016671657562,
      "reward_std_all": 0.2706121802330017,
      "rewards-max/forensics_sinkhorn_reward": 0.9843749105930328,
      "rewards-max/forensics_strict_edge_reward": 0.9782643616199493,
      "rewards-max/forensics_tiou_reward": 0.9895894527435303,
      "rewards/forensics_sinkhorn_reward": 0.46125298738479614,
      "rewards/forensics_strict_edge_reward": 0.23750029504299164,
      "rewards/forensics_tiou_reward": 0.3222070187330246,
      "step": 1387
    },
    {
      "completion_length": 41.328125,
      "epoch": 2.904712041884817,
      "grad_norm": 19.37196159362793,
      "kl": 5.9375,
      "learning_rate": 2.730607966457023e-07,
      "loss": 0.4848,
      "reward": 0.5123544335365295,
      "reward_std": 0.2064184918999672,
      "reward_std_all": 0.2461337372660637,
      "rewards-max/forensics_sinkhorn_reward": 0.9783722460269928,
      "rewards-max/forensics_strict_edge_reward": 0.9151731729507446,
      "rewards-max/forensics_tiou_reward": 0.9753868877887726,
      "rewards/forensics_sinkhorn_reward": 0.3956169933080673,
      "rewards/forensics_strict_edge_reward": 0.17205391079187393,
      "rewards/forensics_tiou_reward": 0.24928749352693558,
      "step": 1388
    },
    {
      "completion_length": 54.21875,
      "epoch": 2.906806282722513,
      "grad_norm": 19.950504302978516,
      "kl": 6.1875,
      "learning_rate": 2.7253668763102724e-07,
      "loss": 0.4395,
      "reward": 0.5677011907100677,
      "reward_std": 0.2545323371887207,
      "reward_std_all": 0.24415725469589233,
      "rewards-max/forensics_sinkhorn_reward": 0.9883468747138977,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9974633157253265,
      "rewards/forensics_sinkhorn_reward": 0.5076137185096741,
      "rewards/forensics_strict_edge_reward": 0.3041704148054123,
      "rewards/forensics_tiou_reward": 0.374360129237175,
      "step": 1389
    },
    {
      "completion_length": 62.109375,
      "epoch": 2.9089005235602095,
      "grad_norm": 14.46674919128418,
      "kl": 7.1875,
      "learning_rate": 2.720125786163522e-07,
      "loss": 0.458,
      "reward": 0.47872698307037354,
      "reward_std": 0.26839372515678406,
      "reward_std_all": 0.29532216489315033,
      "rewards-max/forensics_sinkhorn_reward": 0.9601269960403442,
      "rewards-max/forensics_strict_edge_reward": 0.8104448914527893,
      "rewards-max/forensics_tiou_reward": 0.9328935444355011,
      "rewards/forensics_sinkhorn_reward": 0.4178762882947922,
      "rewards/forensics_strict_edge_reward": 0.20353297889232635,
      "rewards/forensics_tiou_reward": 0.2722625136375427,
      "step": 1390
    },
    {
      "completion_length": 56.859375,
      "epoch": 2.9109947643979055,
      "grad_norm": 13773.5810546875,
      "kl": 36.59375,
      "learning_rate": 2.7148846960167716e-07,
      "loss": 1.7323,
      "reward": 0.6040307879447937,
      "reward_std": 0.2115292102098465,
      "reward_std_all": 0.20282268524169922,
      "rewards-max/forensics_sinkhorn_reward": 0.9888094365596771,
      "rewards-max/forensics_strict_edge_reward": 0.9890114367008209,
      "rewards-max/forensics_tiou_reward": 0.9992864727973938,
      "rewards/forensics_sinkhorn_reward": 0.5733765661716461,
      "rewards/forensics_strict_edge_reward": 0.3899662643671036,
      "rewards/forensics_tiou_reward": 0.4761187732219696,
      "step": 1391
    },
    {
      "completion_length": 27.578125,
      "epoch": 2.913089005235602,
      "grad_norm": 26.726993560791016,
      "kl": 5.921875,
      "learning_rate": 2.7096436058700206e-07,
      "loss": 0.3775,
      "reward": 0.5935579538345337,
      "reward_std": 0.22420553117990494,
      "reward_std_all": 0.2221822589635849,
      "rewards-max/forensics_sinkhorn_reward": 0.9823239743709564,
      "rewards-max/forensics_strict_edge_reward": 0.9375866651535034,
      "rewards-max/forensics_tiou_reward": 0.9740930497646332,
      "rewards/forensics_sinkhorn_reward": 0.5243088006973267,
      "rewards/forensics_strict_edge_reward": 0.3124602138996124,
      "rewards/forensics_tiou_reward": 0.3853701651096344,
      "step": 1392
    },
    {
      "completion_length": 74.34375,
      "epoch": 2.9151832460732985,
      "grad_norm": 43.56959915161133,
      "kl": 8.546875,
      "learning_rate": 2.70440251572327e-07,
      "loss": 0.5415,
      "reward": 0.5739351809024811,
      "reward_std": 0.27505163103342056,
      "reward_std_all": 0.25648797303438187,
      "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": 0.5197288393974304,
      "rewards/forensics_strict_edge_reward": 0.33398081362247467,
      "rewards/forensics_tiou_reward": 0.40351884067058563,
      "step": 1393
    },
    {
      "completion_length": 51.296875,
      "epoch": 2.9172774869109945,
      "grad_norm": 23.550121307373047,
      "kl": 8.0,
      "learning_rate": 2.69916142557652e-07,
      "loss": 0.2743,
      "reward": 0.49730654060840607,
      "reward_std": 0.2629213482141495,
      "reward_std_all": 0.28638549894094467,
      "rewards-max/forensics_sinkhorn_reward": 0.988667756319046,
      "rewards-max/forensics_strict_edge_reward": 0.9890114367008209,
      "rewards-max/forensics_tiou_reward": 0.9987285435199738,
      "rewards/forensics_sinkhorn_reward": 0.4174441546201706,
      "rewards/forensics_strict_edge_reward": 0.27727460861206055,
      "rewards/forensics_tiou_reward": 0.28388921171426773,
      "step": 1394
    },
    {
      "completion_length": 32.234375,
      "epoch": 2.919371727748691,
      "grad_norm": 31.130216598510742,
      "kl": 5.28125,
      "learning_rate": 2.693920335429769e-07,
      "loss": 0.3268,
      "reward": 0.6209984719753265,
      "reward_std": 0.16798897087574005,
      "reward_std_all": 0.17614258080720901,
      "rewards-max/forensics_sinkhorn_reward": 0.9878984689712524,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9956973195075989,
      "rewards/forensics_sinkhorn_reward": 0.5269971340894699,
      "rewards/forensics_strict_edge_reward": 0.34000660479068756,
      "rewards/forensics_tiou_reward": 0.38589201867580414,
      "step": 1395
    },
    {
      "completion_length": 30.859375,
      "epoch": 2.9214659685863875,
      "grad_norm": 19.889326095581055,
      "kl": 5.734375,
      "learning_rate": 2.6886792452830185e-07,
      "loss": 0.4621,
      "reward": 0.5767315626144409,
      "reward_std": 0.25673075020313263,
      "reward_std_all": 0.24566535651683807,
      "rewards-max/forensics_sinkhorn_reward": 0.9663920402526855,
      "rewards-max/forensics_strict_edge_reward": 0.8477883636951447,
      "rewards-max/forensics_tiou_reward": 0.945445328950882,
      "rewards/forensics_sinkhorn_reward": 0.5044793635606766,
      "rewards/forensics_strict_edge_reward": 0.2658744752407074,
      "rewards/forensics_tiou_reward": 0.3816804587841034,
      "step": 1396
    },
    {
      "completion_length": 48.8125,
      "epoch": 2.9235602094240836,
      "grad_norm": 19.727542877197266,
      "kl": 7.84375,
      "learning_rate": 2.6834381551362686e-07,
      "loss": 0.6624,
      "reward": 0.4859852194786072,
      "reward_std": 0.2526341378688812,
      "reward_std_all": 0.27988193929195404,
      "rewards-max/forensics_sinkhorn_reward": 0.8741036951541901,
      "rewards-max/forensics_strict_edge_reward": 0.7291640937328339,
      "rewards-max/forensics_tiou_reward": 0.7795020043849945,
      "rewards/forensics_sinkhorn_reward": 0.4439644441008568,
      "rewards/forensics_strict_edge_reward": 0.2853486202657223,
      "rewards/forensics_tiou_reward": 0.34226229786872864,
      "step": 1397
    },
    {
      "completion_length": 50.5,
      "epoch": 2.92565445026178,
      "grad_norm": 25.82490348815918,
      "kl": 6.53125,
      "learning_rate": 2.6781970649895176e-07,
      "loss": 0.5942,
      "reward": 0.5480324625968933,
      "reward_std": 0.22552987933158875,
      "reward_std_all": 0.2500620409846306,
      "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": 0.47587577998638153,
      "rewards/forensics_strict_edge_reward": 0.3234555125236511,
      "rewards/forensics_tiou_reward": 0.34867069125175476,
      "step": 1398
    },
    {
      "completion_length": 75.125,
      "epoch": 2.9277486910994766,
      "grad_norm": 31.271499633789062,
      "kl": 11.78125,
      "learning_rate": 2.672955974842767e-07,
      "loss": 0.741,
      "reward": 0.47171005606651306,
      "reward_std": 0.2633417993783951,
      "reward_std_all": 0.2967122942209244,
      "rewards-max/forensics_sinkhorn_reward": 0.986993134021759,
      "rewards-max/forensics_strict_edge_reward": 0.8926426768302917,
      "rewards-max/forensics_tiou_reward": 0.9921486377716064,
      "rewards/forensics_sinkhorn_reward": 0.3805810660123825,
      "rewards/forensics_strict_edge_reward": 0.214448019862175,
      "rewards/forensics_tiou_reward": 0.28980444371700287,
      "step": 1399
    },
    {
      "completion_length": 49.65625,
      "epoch": 2.9298429319371726,
      "grad_norm": 32.98735427856445,
      "kl": 8.671875,
      "learning_rate": 2.667714884696017e-07,
      "loss": 0.4326,
      "reward": 0.5690441131591797,
      "reward_std": 0.1919301301240921,
      "reward_std_all": 0.21095316857099533,
      "rewards-max/forensics_sinkhorn_reward": 0.9886260330677032,
      "rewards-max/forensics_strict_edge_reward": 0.9782643616199493,
      "rewards-max/forensics_tiou_reward": 0.9985642433166504,
      "rewards/forensics_sinkhorn_reward": 0.5413450449705124,
      "rewards/forensics_strict_edge_reward": 0.3374311178922653,
      "rewards/forensics_tiou_reward": 0.427728608250618,
      "step": 1400
    },
    {
      "completion_length": 37.65625,
      "epoch": 2.931937172774869,
      "grad_norm": 21.87281608581543,
      "kl": 5.75,
      "learning_rate": 2.662473794549266e-07,
      "loss": 0.4108,
      "reward": 0.5315108597278595,
      "reward_std": 0.19353480637073517,
      "reward_std_all": 0.2347404807806015,
      "rewards-max/forensics_sinkhorn_reward": 0.9884631037712097,
      "rewards-max/forensics_strict_edge_reward": 0.9375866651535034,
      "rewards-max/forensics_tiou_reward": 0.9979231655597687,
      "rewards/forensics_sinkhorn_reward": 0.42718465626239777,
      "rewards/forensics_strict_edge_reward": 0.21767187118530273,
      "rewards/forensics_tiou_reward": 0.2964447885751724,
      "step": 1401
    },
    {
      "completion_length": 44.9375,
      "epoch": 2.9340314136125656,
      "grad_norm": 33.42776870727539,
      "kl": 6.765625,
      "learning_rate": 2.6572327044025154e-07,
      "loss": 0.5731,
      "reward": 0.5258381217718124,
      "reward_std": 0.16255054622888565,
      "reward_std_all": 0.22490882128477097,
      "rewards-max/forensics_sinkhorn_reward": 0.9880844354629517,
      "rewards-max/forensics_strict_edge_reward": 0.9782643616199493,
      "rewards-max/forensics_tiou_reward": 0.9964285790920258,
      "rewards/forensics_sinkhorn_reward": 0.4570486694574356,
      "rewards/forensics_strict_edge_reward": 0.28570425510406494,
      "rewards/forensics_tiou_reward": 0.3449641466140747,
      "step": 1402
    },
    {
      "completion_length": 63.40625,
      "epoch": 2.9361256544502616,
      "grad_norm": 18.345821380615234,
      "kl": 7.390625,
      "learning_rate": 2.651991614255765e-07,
      "loss": 0.5461,
      "reward": 0.6109501123428345,
      "reward_std": 0.20589915663003922,
      "reward_std_all": 0.1983962431550026,
      "rewards-max/forensics_sinkhorn_reward": 0.9882899522781372,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9972403645515442,
      "rewards/forensics_sinkhorn_reward": 0.5502497553825378,
      "rewards/forensics_strict_edge_reward": 0.38896042108535767,
      "rewards/forensics_tiou_reward": 0.43302637338638306,
      "step": 1403
    },
    {
      "completion_length": 36.046875,
      "epoch": 2.938219895287958,
      "grad_norm": 23.439205169677734,
      "kl": 4.203125,
      "learning_rate": 2.646750524109014e-07,
      "loss": 0.2903,
      "reward": 0.6249519884586334,
      "reward_std": 0.1619822420179844,
      "reward_std_all": 0.17202851176261902,
      "rewards-max/forensics_sinkhorn_reward": 0.9800664186477661,
      "rewards-max/forensics_strict_edge_reward": 0.9464850425720215,
      "rewards-max/forensics_tiou_reward": 0.9785102307796478,
      "rewards/forensics_sinkhorn_reward": 0.5387339293956757,
      "rewards/forensics_strict_edge_reward": 0.2949296832084656,
      "rewards/forensics_tiou_reward": 0.3632734715938568,
      "step": 1404
    },
    {
      "completion_length": 52.40625,
      "epoch": 2.9403141361256546,
      "grad_norm": 20.867456436157227,
      "kl": 5.640625,
      "learning_rate": 2.641509433962264e-07,
      "loss": 0.4244,
      "reward": 0.5815181136131287,
      "reward_std": 0.24046321213245392,
      "reward_std_all": 0.24200760573148727,
      "rewards-max/forensics_sinkhorn_reward": 0.983494907617569,
      "rewards-max/forensics_strict_edge_reward": 0.9158510267734528,
      "rewards-max/forensics_tiou_reward": 0.9866335690021515,
      "rewards/forensics_sinkhorn_reward": 0.4793207049369812,
      "rewards/forensics_strict_edge_reward": 0.271496057510376,
      "rewards/forensics_tiou_reward": 0.3188251852989197,
      "step": 1405
    },
    {
      "completion_length": 42.984375,
      "epoch": 2.9424083769633507,
      "grad_norm": 24.790136337280273,
      "kl": 4.734375,
      "learning_rate": 2.636268343815514e-07,
      "loss": 0.305,
      "reward": 0.574085533618927,
      "reward_std": 0.19538726657629013,
      "reward_std_all": 0.2158021777868271,
      "rewards-max/forensics_sinkhorn_reward": 0.9880184531211853,
      "rewards-max/forensics_strict_edge_reward": 0.9782643616199493,
      "rewards-max/forensics_tiou_reward": 0.9961759150028229,
      "rewards/forensics_sinkhorn_reward": 0.48139065504074097,
      "rewards/forensics_strict_edge_reward": 0.20355506241321564,
      "rewards/forensics_tiou_reward": 0.28108712285757065,
      "step": 1406
    },
    {
      "completion_length": 33.390625,
      "epoch": 2.944502617801047,
      "grad_norm": 23.468107223510742,
      "kl": 3.453125,
      "learning_rate": 2.631027253668763e-07,
      "loss": 0.3032,
      "reward": 0.6255509555339813,
      "reward_std": 0.24469111859798431,
      "reward_std_all": 0.22235184907913208,
      "rewards-max/forensics_sinkhorn_reward": 0.9886753857135773,
      "rewards-max/forensics_strict_edge_reward": 0.9890114367008209,
      "rewards-max/forensics_tiou_reward": 0.9987577497959137,
      "rewards/forensics_sinkhorn_reward": 0.535180926322937,
      "rewards/forensics_strict_edge_reward": 0.3126184493303299,
      "rewards/forensics_tiou_reward": 0.3827246278524399,
      "step": 1407
    },
    {
      "completion_length": 55.25,
      "epoch": 2.9465968586387437,
      "grad_norm": 24.777339935302734,
      "kl": 5.5625,
      "learning_rate": 2.6257861635220124e-07,
      "loss": 0.3607,
      "reward": 0.5496143698692322,
      "reward_std": 0.19987038522958755,
      "reward_std_all": 0.2222725823521614,
      "rewards-max/forensics_sinkhorn_reward": 0.9888150691986084,
      "rewards-max/forensics_strict_edge_reward": 0.9782643616199493,
      "rewards-max/forensics_tiou_reward": 0.9993086755275726,
      "rewards/forensics_sinkhorn_reward": 0.442316934466362,
      "rewards/forensics_strict_edge_reward": 0.2328399419784546,
      "rewards/forensics_tiou_reward": 0.26173585653305054,
      "step": 1408
    },
    {
      "completion_length": 72.84375,
      "epoch": 2.9486910994764397,
      "grad_norm": 16.005048751831055,
      "kl": 7.46875,
      "learning_rate": 2.620545073375262e-07,
      "loss": 0.5954,
      "reward": 0.5122175216674805,
      "reward_std": 0.19532250612974167,
      "reward_std_all": 0.24590536952018738,
      "rewards-max/forensics_sinkhorn_reward": 0.987695038318634,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9948945045471191,
      "rewards/forensics_sinkhorn_reward": 0.41210031509399414,
      "rewards/forensics_strict_edge_reward": 0.22064394503831863,
      "rewards/forensics_tiou_reward": 0.3035116270184517,
      "step": 1409
    },
    {
      "completion_length": 47.46875,
      "epoch": 2.950785340314136,
      "grad_norm": 22.33098793029785,
      "kl": 6.03125,
      "learning_rate": 2.615303983228511e-07,
      "loss": 0.513,
      "reward": 0.6320856809616089,
      "reward_std": 0.18980170786380768,
      "reward_std_all": 0.1807568073272705,
      "rewards-max/forensics_sinkhorn_reward": 0.9875768423080444,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9944289028644562,
      "rewards/forensics_sinkhorn_reward": 0.579847514629364,
      "rewards/forensics_strict_edge_reward": 0.3454584330320358,
      "rewards/forensics_tiou_reward": 0.4523811340332031,
      "step": 1410
    },
    {
      "completion_length": 62.578125,
      "epoch": 2.9528795811518327,
      "grad_norm": 21.536678314208984,
      "kl": 6.203125,
      "learning_rate": 2.6100628930817607e-07,
      "loss": 0.3629,
      "reward": 0.5854036509990692,
      "reward_std": 0.17149287089705467,
      "reward_std_all": 0.19128188118338585,
      "rewards-max/forensics_sinkhorn_reward": 0.9887720048427582,
      "rewards-max/forensics_strict_edge_reward": 0.9782643616199493,
      "rewards-max/forensics_tiou_reward": 0.9991390407085419,
      "rewards/forensics_sinkhorn_reward": 0.5528264939785004,
      "rewards/forensics_strict_edge_reward": 0.3504885882139206,
      "rewards/forensics_tiou_reward": 0.4429137110710144,
      "step": 1411
    },
    {
      "completion_length": 45.0,
      "epoch": 2.9549738219895287,
      "grad_norm": 18.19939422607422,
      "kl": 6.578125,
      "learning_rate": 2.60482180293501e-07,
      "loss": 0.5473,
      "reward": 0.5495697557926178,
      "reward_std": 0.23305443674325943,
      "reward_std_all": 0.24623648822307587,
      "rewards-max/forensics_sinkhorn_reward": 0.9880940318107605,
      "rewards-max/forensics_strict_edge_reward": 0.947419673204422,
      "rewards-max/forensics_tiou_reward": 0.9964664280414581,
      "rewards/forensics_sinkhorn_reward": 0.5026305764913559,
      "rewards/forensics_strict_edge_reward": 0.2681691199541092,
      "rewards/forensics_tiou_reward": 0.3875614106655121,
      "step": 1412
    },
    {
      "completion_length": 45.0,
      "epoch": 2.9570680628272252,
      "grad_norm": 29.83319091796875,
      "kl": 7.890625,
      "learning_rate": 2.59958071278826e-07,
      "loss": 0.3876,
      "reward": 0.5143021643161774,
      "reward_std": 0.22461151331663132,
      "reward_std_all": 0.25473351776599884,
      "rewards-max/forensics_sinkhorn_reward": 0.9756503403186798,
      "rewards-max/forensics_strict_edge_reward": 0.8876297473907471,
      "rewards-max/forensics_tiou_reward": 0.9681123793125153,
      "rewards/forensics_sinkhorn_reward": 0.3951011896133423,
      "rewards/forensics_strict_edge_reward": 0.1764039397239685,
      "rewards/forensics_tiou_reward": 0.22964631021022797,
      "step": 1413
    },
    {
      "completion_length": 68.453125,
      "epoch": 2.9591623036649217,
      "grad_norm": 23.319425582885742,
      "kl": 9.375,
      "learning_rate": 2.5943396226415094e-07,
      "loss": 0.5487,
      "reward": 0.48494987189769745,
      "reward_std": 0.22399093210697174,
      "reward_std_all": 0.26464203000068665,
      "rewards-max/forensics_sinkhorn_reward": 0.9878598153591156,
      "rewards-max/forensics_strict_edge_reward": 0.9464850425720215,
      "rewards-max/forensics_tiou_reward": 0.9955428838729858,
      "rewards/forensics_sinkhorn_reward": 0.3930231183767319,
      "rewards/forensics_strict_edge_reward": 0.19558284431695938,
      "rewards/forensics_tiou_reward": 0.23833703994750977,
      "step": 1414
    },
    {
      "completion_length": 53.953125,
      "epoch": 2.9612565445026178,
      "grad_norm": 28.964981079101562,
      "kl": 8.9375,
      "learning_rate": 2.589098532494759e-07,
      "loss": 0.5174,
      "reward": 0.551887184381485,
      "reward_std": 0.25342872738838196,
      "reward_std_all": 0.25186687707901,
      "rewards-max/forensics_sinkhorn_reward": 0.9714072346687317,
      "rewards-max/forensics_strict_edge_reward": 0.9048932790756226,
      "rewards-max/forensics_tiou_reward": 0.9595470428466797,
      "rewards/forensics_sinkhorn_reward": 0.4756375849246979,
      "rewards/forensics_strict_edge_reward": 0.25548920780420303,
      "rewards/forensics_tiou_reward": 0.33359144628047943,
      "step": 1415
    },
    {
      "completion_length": 81.265625,
      "epoch": 2.9633507853403143,
      "grad_norm": 25.000537872314453,
      "kl": 11.09375,
      "learning_rate": 2.583857442348008e-07,
      "loss": 0.7123,
      "reward": 0.4529017508029938,
      "reward_std": 0.36173124611377716,
      "reward_std_all": 0.3477793037891388,
      "rewards-max/forensics_sinkhorn_reward": 0.9887140095233917,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9989117085933685,
      "rewards/forensics_sinkhorn_reward": 0.38222700357437134,
      "rewards/forensics_strict_edge_reward": 0.2870132848620415,
      "rewards/forensics_tiou_reward": 0.3095630705356598,
      "step": 1416
    },
    {
      "completion_length": 44.765625,
      "epoch": 2.9654450261780103,
      "grad_norm": 32.30437469482422,
      "kl": 9.921875,
      "learning_rate": 2.5786163522012577e-07,
      "loss": 0.6269,
      "reward": 0.533989891409874,
      "reward_std": 0.11823762580752373,
      "reward_std_all": 0.19990326091647148,
      "rewards-max/forensics_sinkhorn_reward": 0.9798576533794403,
      "rewards-max/forensics_strict_edge_reward": 0.9185642302036285,
      "rewards-max/forensics_tiou_reward": 0.9785802662372589,
      "rewards/forensics_sinkhorn_reward": 0.46182912588119507,
      "rewards/forensics_strict_edge_reward": 0.23275388777256012,
      "rewards/forensics_tiou_reward": 0.36697329580783844,
      "step": 1417
    },
    {
      "completion_length": 51.203125,
      "epoch": 2.967539267015707,
      "grad_norm": 17.135238647460938,
      "kl": 5.671875,
      "learning_rate": 2.573375262054507e-07,
      "loss": 0.3147,
      "reward": 0.5630097985267639,
      "reward_std": 0.15787423402071,
      "reward_std_all": 0.20217283815145493,
      "rewards-max/forensics_sinkhorn_reward": 0.9865982234477997,
      "rewards-max/forensics_strict_edge_reward": 0.9464850425720215,
      "rewards-max/forensics_tiou_reward": 0.9906013011932373,
      "rewards/forensics_sinkhorn_reward": 0.5098793804645538,
      "rewards/forensics_strict_edge_reward": 0.2999449819326401,
      "rewards/forensics_tiou_reward": 0.35958343744277954,
      "step": 1418
    },
    {
      "completion_length": 43.875,
      "epoch": 2.9696335078534033,
      "grad_norm": 22.825735092163086,
      "kl": 6.546875,
      "learning_rate": 2.5681341719077563e-07,
      "loss": 0.4397,
      "reward": 0.5893252193927765,
      "reward_std": 0.2806745246052742,
      "reward_std_all": 0.25646813213825226,
      "rewards-max/forensics_sinkhorn_reward": 0.9878159165382385,
      "rewards-max/forensics_strict_edge_reward": 0.9782643616199493,
      "rewards-max/forensics_tiou_reward": 0.995381087064743,
      "rewards/forensics_sinkhorn_reward": 0.520210012793541,
      "rewards/forensics_strict_edge_reward": 0.3549728989601135,
      "rewards/forensics_tiou_reward": 0.38259483873844147,
      "step": 1419
    },
    {
      "completion_length": 65.671875,
      "epoch": 2.9717277486910993,
      "grad_norm": 19.559391021728516,
      "kl": 9.046875,
      "learning_rate": 2.562893081761006e-07,
      "loss": 0.7262,
      "reward": 0.46962326765060425,
      "reward_std": 0.26719212532043457,
      "reward_std_all": 0.29539790004491806,
      "rewards-max/forensics_sinkhorn_reward": 0.9524320065975189,
      "rewards-max/forensics_strict_edge_reward": 0.7736190855503082,
      "rewards-max/forensics_tiou_reward": 0.8666625916957855,
      "rewards/forensics_sinkhorn_reward": 0.3840155750513077,
      "rewards/forensics_strict_edge_reward": 0.19163282215595245,
      "rewards/forensics_tiou_reward": 0.24932505190372467,
      "step": 1420
    },
    {
      "completion_length": 71.875,
      "epoch": 2.973821989528796,
      "grad_norm": 24.593236923217773,
      "kl": 7.109375,
      "learning_rate": 2.557651991614256e-07,
      "loss": 0.5116,
      "reward": 0.5368538498878479,
      "reward_std": 0.22226973623037338,
      "reward_std_all": 0.2353670373558998,
      "rewards-max/forensics_sinkhorn_reward": 0.9889057874679565,
      "rewards-max/forensics_strict_edge_reward": 0.9890114367008209,
      "rewards-max/forensics_tiou_reward": 0.9996661245822906,
      "rewards/forensics_sinkhorn_reward": 0.43189944326877594,
      "rewards/forensics_strict_edge_reward": 0.29024261236190796,
      "rewards/forensics_tiou_reward": 0.33871451020240784,
      "step": 1421
    },
    {
      "completion_length": 35.453125,
      "epoch": 2.975916230366492,
      "grad_norm": 18.902616500854492,
      "kl": 5.4921875,
      "learning_rate": 2.552410901467505e-07,
      "loss": 0.4381,
      "reward": 0.6094809472560883,
      "reward_std": 0.16205069422721863,
      "reward_std_all": 0.17328208684921265,
      "rewards-max/forensics_sinkhorn_reward": 0.9878931939601898,
      "rewards-max/forensics_strict_edge_reward": 0.9158510267734528,
      "rewards-max/forensics_tiou_reward": 0.9956744015216827,
      "rewards/forensics_sinkhorn_reward": 0.6166715919971466,
      "rewards/forensics_strict_edge_reward": 0.3589000701904297,
      "rewards/forensics_tiou_reward": 0.5040715038776398,
      "step": 1422
    },
    {
      "completion_length": 38.625,
      "epoch": 2.9780104712041884,
      "grad_norm": 23.147315979003906,
      "kl": 4.234375,
      "learning_rate": 2.5471698113207547e-07,
      "loss": 0.4555,
      "reward": 0.5949344038963318,
      "reward_std": 0.1909618079662323,
      "reward_std_all": 0.19508716464042664,
      "rewards-max/forensics_sinkhorn_reward": 0.9879591763019562,
      "rewards-max/forensics_strict_edge_reward": 0.9782643616199493,
      "rewards-max/forensics_tiou_reward": 0.9959432184696198,
      "rewards/forensics_sinkhorn_reward": 0.4926588535308838,
      "rewards/forensics_strict_edge_reward": 0.2855006158351898,
      "rewards/forensics_tiou_reward": 0.3325433135032654,
      "step": 1423
    },
    {
      "completion_length": 49.921875,
      "epoch": 2.980104712041885,
      "grad_norm": 16.989238739013672,
      "kl": 4.984375,
      "learning_rate": 2.541928721174004e-07,
      "loss": 0.3192,
      "reward": 0.542847752571106,
      "reward_std": 0.2359236478805542,
      "reward_std_all": 0.24132371693849564,
      "rewards-max/forensics_sinkhorn_reward": 0.9885381460189819,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9982177913188934,
      "rewards/forensics_sinkhorn_reward": 0.4936957359313965,
      "rewards/forensics_strict_edge_reward": 0.30159181356430054,
      "rewards/forensics_tiou_reward": 0.34734009206295013,
      "step": 1424
    },
    {
      "completion_length": 63.890625,
      "epoch": 2.982198952879581,
      "grad_norm": 14.94110107421875,
      "kl": 8.578125,
      "learning_rate": 2.5366876310272533e-07,
      "loss": 0.6455,
      "reward": 0.49494920670986176,
      "reward_std": 0.2745027393102646,
      "reward_std_all": 0.2846909761428833,
      "rewards-max/forensics_sinkhorn_reward": 0.9670982360839844,
      "rewards-max/forensics_strict_edge_reward": 0.9364311099052429,
      "rewards-max/forensics_tiou_reward": 0.9484754502773285,
      "rewards/forensics_sinkhorn_reward": 0.4432511627674103,
      "rewards/forensics_strict_edge_reward": 0.31690508127212524,
      "rewards/forensics_tiou_reward": 0.3300279378890991,
      "step": 1425
    },
    {
      "completion_length": 29.671875,
      "epoch": 2.9842931937172774,
      "grad_norm": 22.150205612182617,
      "kl": 3.421875,
      "learning_rate": 2.531446540880503e-07,
      "loss": 0.3473,
      "reward": 0.6399809122085571,
      "reward_std": 0.1715266853570938,
      "reward_std_all": 0.16803184896707535,
      "rewards-max/forensics_sinkhorn_reward": 0.988539457321167,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9982229769229889,
      "rewards/forensics_sinkhorn_reward": 0.5570693016052246,
      "rewards/forensics_strict_edge_reward": 0.3406398445367813,
      "rewards/forensics_tiou_reward": 0.4262565076351166,
      "step": 1426
    },
    {
      "completion_length": 98.375,
      "epoch": 2.986387434554974,
      "grad_norm": 27.89459800720215,
      "kl": 7.90625,
      "learning_rate": 2.5262054507337525e-07,
      "loss": 0.3725,
      "reward": 0.4851295053958893,
      "reward_std": 0.2899644076824188,
      "reward_std_all": 0.2940749228000641,
      "rewards-max/forensics_sinkhorn_reward": 0.9806618988513947,
      "rewards-max/forensics_strict_edge_reward": 0.9782643616199493,
      "rewards-max/forensics_tiou_reward": 0.980923593044281,
      "rewards/forensics_sinkhorn_reward": 0.3799609839916229,
      "rewards/forensics_strict_edge_reward": 0.196435809135437,
      "rewards/forensics_tiou_reward": 0.20916473120450974,
      "step": 1427
    },
    {
      "completion_length": 55.234375,
      "epoch": 2.98848167539267,
      "grad_norm": 25.76447296142578,
      "kl": 7.875,
      "learning_rate": 2.5209643605870015e-07,
      "loss": 0.4865,
      "reward": 0.5247577726840973,
      "reward_std": 0.26249146461486816,
      "reward_std_all": 0.27148228883743286,
      "rewards-max/forensics_sinkhorn_reward": 0.96497443318367,
      "rewards-max/forensics_strict_edge_reward": 0.946017861366272,
      "rewards-max/forensics_tiou_reward": 0.9487355351448059,
      "rewards/forensics_sinkhorn_reward": 0.4324037432670593,
      "rewards/forensics_strict_edge_reward": 0.24240954220294952,
      "rewards/forensics_tiou_reward": 0.30308856070041656,
      "step": 1428
    },
    {
      "completion_length": 60.734375,
      "epoch": 2.9905759162303664,
      "grad_norm": 23.774808883666992,
      "kl": 7.78125,
      "learning_rate": 2.5157232704402517e-07,
      "loss": 0.4723,
      "reward": 0.6120780110359192,
      "reward_std": 0.2513093799352646,
      "reward_std_all": 0.23380839824676514,
      "rewards-max/forensics_sinkhorn_reward": 0.9881721436977386,
      "rewards-max/forensics_strict_edge_reward": 0.9677534997463226,
      "rewards-max/forensics_tiou_reward": 0.9967741966247559,
      "rewards/forensics_sinkhorn_reward": 0.56056347489357,
      "rewards/forensics_strict_edge_reward": 0.38883934915065765,
      "rewards/forensics_tiou_reward": 0.4448397308588028,
      "step": 1429
    },
    {
      "completion_length": 61.234375,
      "epoch": 2.992670157068063,
      "grad_norm": 28.29539680480957,
      "kl": 9.40625,
      "learning_rate": 2.510482180293501e-07,
      "loss": 0.4158,
      "reward": 0.4768924415111542,
      "reward_std": 0.26495788991451263,
      "reward_std_all": 0.2949923127889633,
      "rewards-max/forensics_sinkhorn_reward": 0.9582638740539551,
      "rewards-max/forensics_strict_edge_reward": 0.8440330624580383,
      "rewards-max/forensics_tiou_reward": 0.926243007183075,
      "rewards/forensics_sinkhorn_reward": 0.3786640912294388,
      "rewards/forensics_strict_edge_reward": 0.1455133967101574,
      "rewards/forensics_tiou_reward": 0.19773171842098236,
      "step": 1430
    },
    {
      "completion_length": 48.21875,
      "epoch": 2.994764397905759,
      "grad_norm": 20.56655502319336,
      "kl": 8.3125,
      "learning_rate": 2.5052410901467503e-07,
      "loss": 0.61,
      "reward": 0.5160104483366013,
      "reward_std": 0.15465538948774338,
      "reward_std_all": 0.2172693982720375,
      "rewards-max/forensics_sinkhorn_reward": 0.9885565638542175,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9982900321483612,
      "rewards/forensics_sinkhorn_reward": 0.5106247961521149,
      "rewards/forensics_strict_edge_reward": 0.3170626014471054,
      "rewards/forensics_tiou_reward": 0.4302344024181366,
      "step": 1431
    },
    {
      "completion_length": 116.421875,
      "epoch": 2.9968586387434555,
      "grad_norm": 29.734914779663086,
      "kl": 13.15625,
      "learning_rate": 2.5e-07,
      "loss": 0.8,
      "reward": 0.4003414213657379,
      "reward_std": 0.3004348948597908,
      "reward_std_all": 0.34900496900081635,
      "rewards-max/forensics_sinkhorn_reward": 0.9871109426021576,
      "rewards-max/forensics_strict_edge_reward": 0.9375866651535034,
      "rewards-max/forensics_tiou_reward": 0.9926200211048126,
      "rewards/forensics_sinkhorn_reward": 0.3460751175880432,
      "rewards/forensics_strict_edge_reward": 0.235902801156044,
      "rewards/forensics_tiou_reward": 0.25999076664447784,
      "step": 1432
    },
    {
      "completion_length": 66.453125,
      "epoch": 2.998952879581152,
      "grad_norm": 14.338260650634766,
      "kl": 6.34375,
      "learning_rate": 2.4947589098532495e-07,
      "loss": 0.5081,
      "reward": 0.5734186172485352,
      "reward_std": 0.17857256531715393,
      "reward_std_all": 0.21028784662485123,
      "rewards-max/forensics_sinkhorn_reward": 0.9876804053783417,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9948384165763855,
      "rewards/forensics_sinkhorn_reward": 0.4889654666185379,
      "rewards/forensics_strict_edge_reward": 0.37812499701976776,
      "rewards/forensics_tiou_reward": 0.4046074450016022,
      "step": 1433
    },
    {
      "completion_length": 47.53125,
      "epoch": 3.0,
      "grad_norm": 14.287420272827148,
      "kl": 6.4375,
      "learning_rate": 2.489517819706499e-07,
      "loss": 0.2191,
      "reward": 0.6010856628417969,
      "reward_std": 0.18684926629066467,
      "reward_std_all": 0.19602355360984802,
      "rewards-max/forensics_sinkhorn_reward": 0.9877995252609253,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9953106641769409,
      "rewards/forensics_sinkhorn_reward": 0.566704273223877,
      "rewards/forensics_strict_edge_reward": 0.33180004358291626,
      "rewards/forensics_tiou_reward": 0.43125590682029724,
      "step": 1434
    },
    {
      "completion_length": 62.515625,
      "epoch": 3.0020942408376965,
      "grad_norm": 26.296669006347656,
      "kl": 8.1875,
      "learning_rate": 2.484276729559748e-07,
      "loss": 0.3555,
      "reward": 0.5085199475288391,
      "reward_std": 0.20699654519557953,
      "reward_std_all": 0.24473337084054947,
      "rewards-max/forensics_sinkhorn_reward": 0.9485934674739838,
      "rewards-max/forensics_strict_edge_reward": 0.819539487361908,
      "rewards-max/forensics_tiou_reward": 0.9067255854606628,
      "rewards/forensics_sinkhorn_reward": 0.4199853837490082,
      "rewards/forensics_strict_edge_reward": 0.1162959448993206,
      "rewards/forensics_tiou_reward": 0.18723483383655548,
      "step": 1435
    },
    {
      "completion_length": 67.65625,
      "epoch": 3.0041884816753925,
      "grad_norm": 22.172452926635742,
      "kl": 7.390625,
      "learning_rate": 2.4790356394129977e-07,
      "loss": 0.5604,
      "reward": 0.550852358341217,
      "reward_std": 0.18115856498479843,
      "reward_std_all": 0.21593530476093292,
      "rewards-max/forensics_sinkhorn_reward": 0.9794143438339233,
      "rewards-max/forensics_strict_edge_reward": 0.9256840348243713,
      "rewards-max/forensics_tiou_reward": 0.9771640002727509,
      "rewards/forensics_sinkhorn_reward": 0.4995841979980469,
      "rewards/forensics_strict_edge_reward": 0.2730341702699661,
      "rewards/forensics_tiou_reward": 0.37019798159599304,
      "step": 1436
    },
    {
      "completion_length": 39.96875,
      "epoch": 3.006282722513089,
      "grad_norm": 18.274240493774414,
      "kl": 5.59375,
      "learning_rate": 2.4737945492662473e-07,
      "loss": 0.4956,
      "reward": 0.586889237165451,
      "reward_std": 0.16075419634580612,
      "reward_std_all": 0.19521407783031464,
      "rewards-max/forensics_sinkhorn_reward": 0.976089209318161,
      "rewards-max/forensics_strict_edge_reward": 0.8426913917064667,
      "rewards-max/forensics_tiou_reward": 0.9698267877101898,
      "rewards/forensics_sinkhorn_reward": 0.5184656381607056,
      "rewards/forensics_strict_edge_reward": 0.289664588868618,
      "rewards/forensics_tiou_reward": 0.385147325694561,
      "step": 1437
    },
    {
      "completion_length": 37.46875,
      "epoch": 3.0083769633507855,
      "grad_norm": 18.66582679748535,
      "kl": 3.7578125,
      "learning_rate": 2.468553459119497e-07,
      "loss": 0.1772,
      "reward": 0.5769146382808685,
      "reward_std": 0.15418047830462456,
      "reward_std_all": 0.18399638682603836,
      "rewards-max/forensics_sinkhorn_reward": 0.9887248873710632,
      "rewards-max/forensics_strict_edge_reward": 0.9890114367008209,
      "rewards-max/forensics_tiou_reward": 0.9989528656005859,
      "rewards/forensics_sinkhorn_reward": 0.5176651775836945,
      "rewards/forensics_strict_edge_reward": 0.23878157138824463,
      "rewards/forensics_tiou_reward": 0.32012465596199036,
      "step": 1438
    },
    {
      "completion_length": 54.640625,
      "epoch": 3.0104712041884816,
      "grad_norm": 22.047300338745117,
      "kl": 4.28125,
      "learning_rate": 2.463312368972746e-07,
      "loss": 0.1723,
      "reward": 0.6671261787414551,
      "reward_std": 0.09991355799138546,
      "reward_std_all": 0.12079774588346481,
      "rewards-max/forensics_sinkhorn_reward": 0.9671065211296082,
      "rewards-max/forensics_strict_edge_reward": 0.8805811703205109,
      "rewards-max/forensics_tiou_reward": 0.9485682249069214,
      "rewards/forensics_sinkhorn_reward": 0.5402904599905014,
      "rewards/forensics_strict_edge_reward": 0.274116650223732,
      "rewards/forensics_tiou_reward": 0.3543384373188019,
      "step": 1439
    },
    {
      "completion_length": 40.578125,
      "epoch": 3.012565445026178,
      "grad_norm": 42.00322341918945,
      "kl": 5.5625,
      "learning_rate": 2.458071278825996e-07,
      "loss": 0.4543,
      "reward": 0.6568899154663086,
      "reward_std": 0.16509800404310226,
      "reward_std_all": 0.1638626977801323,
      "rewards-max/forensics_sinkhorn_reward": 0.9765586853027344,
      "rewards-max/forensics_strict_edge_reward": 0.9464850425720215,
      "rewards-max/forensics_tiou_reward": 0.9708301424980164,
      "rewards/forensics_sinkhorn_reward": 0.5854098796844482,
      "rewards/forensics_strict_edge_reward": 0.3089488595724106,
      "rewards/forensics_tiou_reward": 0.3946010172367096,
      "step": 1440
    },
    {
      "completion_length": 40.3125,
      "epoch": 3.0146596858638746,
      "grad_norm": 23.953107833862305,
      "kl": 5.765625,
      "learning_rate": 2.452830188679245e-07,
      "loss": 0.4554,
      "reward": 0.5757677257061005,
      "reward_std": 0.24573256075382233,
      "reward_std_all": 0.24441415071487427,
      "rewards-max/forensics_sinkhorn_reward": 0.9679512083530426,
      "rewards-max/forensics_strict_edge_reward": 0.856150895357132,
      "rewards-max/forensics_tiou_reward": 0.9489278793334961,
      "rewards/forensics_sinkhorn_reward": 0.47448909282684326,
      "rewards/forensics_strict_edge_reward": 0.2928549349308014,
      "rewards/forensics_tiou_reward": 0.32903875410556793,
      "step": 1441
    },
    {
      "completion_length": 52.96875,
      "epoch": 3.0167539267015706,
      "grad_norm": 22.389108657836914,
      "kl": 4.9609375,
      "learning_rate": 2.4475890985324947e-07,
      "loss": 0.4947,
      "reward": 0.5949447602033615,
      "reward_std": 0.20282428711652756,
      "reward_std_all": 0.21661170572042465,
      "rewards-max/forensics_sinkhorn_reward": 0.9878112971782684,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9953532516956329,
      "rewards/forensics_sinkhorn_reward": 0.5713833868503571,
      "rewards/forensics_strict_edge_reward": 0.36154481768608093,
      "rewards/forensics_tiou_reward": 0.46875855326652527,
      "step": 1442
    },
    {
      "completion_length": 55.671875,
      "epoch": 3.018848167539267,
      "grad_norm": 18.154592514038086,
      "kl": 5.90625,
      "learning_rate": 2.4423480083857443e-07,
      "loss": 0.3296,
      "reward": 0.5896003544330597,
      "reward_std": 0.11591238528490067,
      "reward_std_all": 0.17601722478866577,
      "rewards-max/forensics_sinkhorn_reward": 0.9843240976333618,
      "rewards-max/forensics_strict_edge_reward": 0.8078656196594238,
      "rewards-max/forensics_tiou_reward": 0.9816226959228516,
      "rewards/forensics_sinkhorn_reward": 0.5137150585651398,
      "rewards/forensics_strict_edge_reward": 0.30357226729393005,
      "rewards/forensics_tiou_reward": 0.3793380856513977,
      "step": 1443
    },
    {
      "completion_length": 56.21875,
      "epoch": 3.020942408376963,
      "grad_norm": 27.29682159423828,
      "kl": 8.078125,
      "learning_rate": 2.437106918238994e-07,
      "loss": 0.5799,
      "reward": 0.46956197917461395,
      "reward_std": 0.21957701444625854,
      "reward_std_all": 0.267034113407135,
      "rewards-max/forensics_sinkhorn_reward": 0.9665517210960388,
      "rewards-max/forensics_strict_edge_reward": 0.8744657039642334,
      "rewards-max/forensics_tiou_reward": 0.9519102573394775,
      "rewards/forensics_sinkhorn_reward": 0.364630863070488,
      "rewards/forensics_strict_edge_reward": 0.17828437685966492,
      "rewards/forensics_tiou_reward": 0.25719453394412994,
      "step": 1444
    },
    {
      "completion_length": 40.703125,
      "epoch": 3.0230366492146596,
      "grad_norm": 17.2643985748291,
      "kl": 6.125,
      "learning_rate": 2.431865828092243e-07,
      "loss": 0.5176,
      "reward": 0.5834676921367645,
      "reward_std": 0.22150607407093048,
      "reward_std_all": 0.2224879115819931,
      "rewards-max/forensics_sinkhorn_reward": 0.9887801706790924,
      "rewards-max/forensics_strict_edge_reward": 0.9890114367008209,
      "rewards-max/forensics_tiou_reward": 0.9991711676120758,
      "rewards/forensics_sinkhorn_reward": 0.5187820792198181,
      "rewards/forensics_strict_edge_reward": 0.3079551160335541,
      "rewards/forensics_tiou_reward": 0.37119437754154205,
      "step": 1445
    },
    {
      "completion_length": 27.5625,
      "epoch": 3.025130890052356,
      "grad_norm": 24.761995315551758,
      "kl": 4.53125,
      "learning_rate": 2.4266247379454925e-07,
      "loss": 0.3939,
      "reward": 0.5462559163570404,
      "reward_std": 0.30053725838661194,
      "reward_std_all": 0.2842142730951309,
      "rewards-max/forensics_sinkhorn_reward": 0.928186446428299,
      "rewards-max/forensics_strict_edge_reward": 0.8635591566562653,
      "rewards-max/forensics_tiou_reward": 0.870676577091217,
      "rewards/forensics_sinkhorn_reward": 0.4339026212692261,
      "rewards/forensics_strict_edge_reward": 0.27392714470624924,
      "rewards/forensics_tiou_reward": 0.28122203052043915,
      "step": 1446
    },
    {
      "completion_length": 60.125,
      "epoch": 3.027225130890052,
      "grad_norm": 16.045480728149414,
      "kl": 6.265625,
      "learning_rate": 2.421383647798742e-07,
      "loss": 0.3947,
      "reward": 0.6054920554161072,
      "reward_std": 0.1954350471496582,
      "reward_std_all": 0.20551464706659317,
      "rewards-max/forensics_sinkhorn_reward": 0.9866978228092194,
      "rewards-max/forensics_strict_edge_reward": 0.9565287232398987,
      "rewards-max/forensics_tiou_reward": 0.9909632802009583,
      "rewards/forensics_sinkhorn_reward": 0.5247984975576401,
      "rewards/forensics_strict_edge_reward": 0.3473622426390648,
      "rewards/forensics_tiou_reward": 0.38283970952033997,
      "step": 1447
    },
    {
      "completion_length": 66.21875,
      "epoch": 3.0293193717277487,
      "grad_norm": 314.7582092285156,
      "kl": 14.921875,
      "learning_rate": 2.4161425576519917e-07,
      "loss": 0.7858,
      "reward": 0.5009365081787109,
      "reward_std": 0.21129991114139557,
      "reward_std_all": 0.2534175515174866,
      "rewards-max/forensics_sinkhorn_reward": 0.9633947014808655,
      "rewards-max/forensics_strict_edge_reward": 0.909769743680954,
      "rewards-max/forensics_tiou_reward": 0.9409647583961487,
      "rewards/forensics_sinkhorn_reward": 0.4156201034784317,
      "rewards/forensics_strict_edge_reward": 0.23573417589068413,
      "rewards/forensics_tiou_reward": 0.31402866542339325,
      "step": 1448
    },
    {
      "completion_length": 98.0625,
      "epoch": 3.031413612565445,
      "grad_norm": 28.396970748901367,
      "kl": 10.75,
      "learning_rate": 2.4109014675052413e-07,
      "loss": 0.8204,
      "reward": 0.37258343398571014,
      "reward_std": 0.28651776909828186,
      "reward_std_all": 0.36005187034606934,
      "rewards-max/forensics_sinkhorn_reward": 0.9873267114162445,
      "rewards-max/forensics_strict_edge_reward": 0.9574736058712006,
      "rewards-max/forensics_tiou_reward": 0.9934426248073578,
      "rewards/forensics_sinkhorn_reward": 0.33967868983745575,
      "rewards/forensics_strict_edge_reward": 0.2676899880170822,
      "rewards/forensics_tiou_reward": 0.2986014112830162,
      "step": 1449
    },
    {
      "completion_length": 40.75,
      "epoch": 3.033507853403141,
      "grad_norm": 25.75379180908203,
      "kl": 6.2265625,
      "learning_rate": 2.4056603773584903e-07,
      "loss": 0.4047,
      "reward": 0.5391556024551392,
      "reward_std": 0.19212353229522705,
      "reward_std_all": 0.2301005870103836,
      "rewards-max/forensics_sinkhorn_reward": 0.9780712723731995,
      "rewards-max/forensics_strict_edge_reward": 0.878633052110672,
      "rewards-max/forensics_tiou_reward": 0.9746480882167816,
      "rewards/forensics_sinkhorn_reward": 0.46477971971035004,
      "rewards/forensics_strict_edge_reward": 0.2607678174972534,
      "rewards/forensics_tiou_reward": 0.338122233748436,
      "step": 1450
    },
    {
      "completion_length": 69.96875,
      "epoch": 3.0356020942408377,
      "grad_norm": 24.308279037475586,
      "kl": 9.6875,
      "learning_rate": 2.40041928721174e-07,
      "loss": 0.5912,
      "reward": 0.43960531055927277,
      "reward_std": 0.23812683671712875,
      "reward_std_all": 0.29991593956947327,
      "rewards-max/forensics_sinkhorn_reward": 0.9879591763019562,
      "rewards-max/forensics_strict_edge_reward": 0.9782643616199493,
      "rewards-max/forensics_tiou_reward": 0.9959432184696198,
      "rewards/forensics_sinkhorn_reward": 0.3347112834453583,
      "rewards/forensics_strict_edge_reward": 0.18667342513799667,
      "rewards/forensics_tiou_reward": 0.21171771734952927,
      "step": 1451
    },
    {
      "completion_length": 79.109375,
      "epoch": 3.037696335078534,
      "grad_norm": 30.33078956604004,
      "kl": 10.4375,
      "learning_rate": 2.3951781970649895e-07,
      "loss": 0.5839,
      "reward": 0.43569278717041016,
      "reward_std": 0.20967627316713333,
      "reward_std_all": 0.29081542789936066,
      "rewards-max/forensics_sinkhorn_reward": 0.9885683953762054,
      "rewards-max/forensics_strict_edge_reward": 0.9375866651535034,
      "rewards-max/forensics_tiou_reward": 0.9983374774456024,
      "rewards/forensics_sinkhorn_reward": 0.36525169014930725,
      "rewards/forensics_strict_edge_reward": 0.22397346794605255,
      "rewards/forensics_tiou_reward": 0.2784219980239868,
      "step": 1452
    },
    {
      "completion_length": 42.296875,
      "epoch": 3.0397905759162303,
      "grad_norm": 28.277585983276367,
      "kl": 6.359375,
      "learning_rate": 2.389937106918239e-07,
      "loss": 0.4566,
      "reward": 0.481212854385376,
      "reward_std": 0.12788162380456924,
      "reward_std_all": 0.21898306161165237,
      "rewards-max/forensics_sinkhorn_reward": 0.9881507754325867,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9967074990272522,
      "rewards/forensics_sinkhorn_reward": 0.4512353539466858,
      "rewards/forensics_strict_edge_reward": 0.25890669226646423,
      "rewards/forensics_tiou_reward": 0.37816569209098816,
      "step": 1453
    },
    {
      "completion_length": 81.90625,
      "epoch": 3.0418848167539267,
      "grad_norm": 131.5067901611328,
      "kl": 11.8125,
      "learning_rate": 2.384696016771488e-07,
      "loss": 0.775,
      "reward": 0.45005249977111816,
      "reward_std": 0.2546931207180023,
      "reward_std_all": 0.3016376793384552,
      "rewards-max/forensics_sinkhorn_reward": 0.8906381726264954,
      "rewards-max/forensics_strict_edge_reward": 0.684678316116333,
      "rewards-max/forensics_tiou_reward": 0.7862131595611572,
      "rewards/forensics_sinkhorn_reward": 0.3270690143108368,
      "rewards/forensics_strict_edge_reward": 0.1790214478969574,
      "rewards/forensics_tiou_reward": 0.2038925439119339,
      "step": 1454
    },
    {
      "completion_length": 50.046875,
      "epoch": 3.0439790575916232,
      "grad_norm": 19.207225799560547,
      "kl": 6.0,
      "learning_rate": 2.379454926624738e-07,
      "loss": 0.3521,
      "reward": 0.5874959230422974,
      "reward_std": 0.20309878140687943,
      "reward_std_all": 0.21274512261152267,
      "rewards-max/forensics_sinkhorn_reward": 0.9867562055587769,
      "rewards-max/forensics_strict_edge_reward": 0.9565287232398987,
      "rewards-max/forensics_tiou_reward": 0.9912131726741791,
      "rewards/forensics_sinkhorn_reward": 0.5412299782037735,
      "rewards/forensics_strict_edge_reward": 0.3017123341560364,
      "rewards/forensics_tiou_reward": 0.3920490890741348,
      "step": 1455
    },
    {
      "completion_length": 50.296875,
      "epoch": 3.0460732984293193,
      "grad_norm": 13.903536796569824,
      "kl": 5.65625,
      "learning_rate": 2.3742138364779873e-07,
      "loss": 0.4918,
      "reward": 0.5814933180809021,
      "reward_std": 0.25348784029483795,
      "reward_std_all": 0.23927458375692368,
      "rewards-max/forensics_sinkhorn_reward": 0.9882864058017731,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9972266256809235,
      "rewards/forensics_sinkhorn_reward": 0.5613145530223846,
      "rewards/forensics_strict_edge_reward": 0.43180684745311737,
      "rewards/forensics_tiou_reward": 0.47711871564388275,
      "step": 1456
    },
    {
      "completion_length": 46.59375,
      "epoch": 3.048167539267016,
      "grad_norm": 32.20172119140625,
      "kl": 6.359375,
      "learning_rate": 2.3689727463312367e-07,
      "loss": 0.4761,
      "reward": 0.574041485786438,
      "reward_std": 0.28183579444885254,
      "reward_std_all": 0.2610717937350273,
      "rewards-max/forensics_sinkhorn_reward": 0.9741075336933136,
      "rewards-max/forensics_strict_edge_reward": 0.8824498057365417,
      "rewards-max/forensics_tiou_reward": 0.9660161733627319,
      "rewards/forensics_sinkhorn_reward": 0.5329010188579559,
      "rewards/forensics_strict_edge_reward": 0.30726952850818634,
      "rewards/forensics_tiou_reward": 0.41034509241580963,
      "step": 1457
    },
    {
      "completion_length": 45.15625,
      "epoch": 3.050261780104712,
      "grad_norm": 21.216211318969727,
      "kl": 4.53125,
      "learning_rate": 2.3637316561844865e-07,
      "loss": 0.1987,
      "reward": 0.613831639289856,
      "reward_std": 0.19914133846759796,
      "reward_std_all": 0.19291356950998306,
      "rewards-max/forensics_sinkhorn_reward": 0.9839077293872833,
      "rewards-max/forensics_strict_edge_reward": 0.9265981018543243,
      "rewards-max/forensics_tiou_reward": 0.9870653450489044,
      "rewards/forensics_sinkhorn_reward": 0.5419706106185913,
      "rewards/forensics_strict_edge_reward": 0.31074458360671997,
      "rewards/forensics_tiou_reward": 0.3905446231365204,
      "step": 1458
    },
    {
      "completion_length": 64.421875,
      "epoch": 3.0523560209424083,
      "grad_norm": 20.16241455078125,
      "kl": 6.421875,
      "learning_rate": 2.3584905660377358e-07,
      "loss": 0.5443,
      "reward": 0.5446953177452087,
      "reward_std": 0.20893627405166626,
      "reward_std_all": 0.2481546625494957,
      "rewards-max/forensics_sinkhorn_reward": 0.9850628674030304,
      "rewards-max/forensics_strict_edge_reward": 0.894839346408844,
      "rewards-max/forensics_tiou_reward": 0.9845837950706482,
      "rewards/forensics_sinkhorn_reward": 0.46751992404460907,
      "rewards/forensics_strict_edge_reward": 0.3187219500541687,
      "rewards/forensics_tiou_reward": 0.36798252165317535,
      "step": 1459
    },
    {
      "completion_length": 44.6875,
      "epoch": 3.054450261780105,
      "grad_norm": 21.103578567504883,
      "kl": 5.6328125,
      "learning_rate": 2.3532494758909851e-07,
      "loss": 0.4923,
      "reward": 0.5574761033058167,
      "reward_std": 0.28836749494075775,
      "reward_std_all": 0.27606774866580963,
      "rewards-max/forensics_sinkhorn_reward": 0.9658199548721313,
      "rewards-max/forensics_strict_edge_reward": 0.8968285918235779,
      "rewards-max/forensics_tiou_reward": 0.9476862847805023,
      "rewards/forensics_sinkhorn_reward": 0.4524529278278351,
      "rewards/forensics_strict_edge_reward": 0.27015022933483124,
      "rewards/forensics_tiou_reward": 0.3510039895772934,
      "step": 1460
    },
    {
      "completion_length": 30.65625,
      "epoch": 3.056544502617801,
      "grad_norm": 20.49752426147461,
      "kl": 4.109375,
      "learning_rate": 2.3480083857442347e-07,
      "loss": 0.4037,
      "reward": 0.5875689685344696,
      "reward_std": 0.22685133665800095,
      "reward_std_all": 0.21809223294258118,
      "rewards-max/forensics_sinkhorn_reward": 0.9824229776859283,
      "rewards-max/forensics_strict_edge_reward": 0.9375866651535034,
      "rewards-max/forensics_tiou_reward": 0.9848058819770813,
      "rewards/forensics_sinkhorn_reward": 0.49938802421092987,
      "rewards/forensics_strict_edge_reward": 0.31011562049388885,
      "rewards/forensics_tiou_reward": 0.3980510085821152,
      "step": 1461
    },
    {
      "completion_length": 46.078125,
      "epoch": 3.0586387434554974,
      "grad_norm": 22.75297737121582,
      "kl": 6.3125,
      "learning_rate": 2.3427672955974843e-07,
      "loss": 0.4663,
      "reward": 0.5386762320995331,
      "reward_std": 0.26674073934555054,
      "reward_std_all": 0.26525215059518814,
      "rewards-max/forensics_sinkhorn_reward": 0.9842601418495178,
      "rewards-max/forensics_strict_edge_reward": 0.9375866651535034,
      "rewards-max/forensics_tiou_reward": 0.9896698892116547,
      "rewards/forensics_sinkhorn_reward": 0.4358914643526077,
      "rewards/forensics_strict_edge_reward": 0.2505653649568558,
      "rewards/forensics_tiou_reward": 0.28660985827445984,
      "step": 1462
    },
    {
      "completion_length": 35.328125,
      "epoch": 3.060732984293194,
      "grad_norm": 15.4576997756958,
      "kl": 5.328125,
      "learning_rate": 2.3375262054507336e-07,
      "loss": 0.3423,
      "reward": 0.5544577836990356,
      "reward_std": 0.2319405972957611,
      "reward_std_all": 0.2437484711408615,
      "rewards-max/forensics_sinkhorn_reward": 0.9883562922477722,
      "rewards-max/forensics_strict_edge_reward": 0.9185642302036285,
      "rewards-max/forensics_tiou_reward": 0.9975000023841858,
      "rewards/forensics_sinkhorn_reward": 0.49645498394966125,
      "rewards/forensics_strict_edge_reward": 0.29127706587314606,
      "rewards/forensics_tiou_reward": 0.36131468415260315,
      "step": 1463
    },
    {
      "completion_length": 79.609375,
      "epoch": 3.06282722513089,
      "grad_norm": 28.256746292114258,
      "kl": 8.671875,
      "learning_rate": 2.3322851153039832e-07,
      "loss": 0.7068,
      "reward": 0.4937284290790558,
      "reward_std": 0.17835816740989685,
      "reward_std_all": 0.24624623358249664,
      "rewards-max/forensics_sinkhorn_reward": 0.9797022342681885,
      "rewards-max/forensics_strict_edge_reward": 0.8655564188957214,
      "rewards-max/forensics_tiou_reward": 0.9731245636940002,
      "rewards/forensics_sinkhorn_reward": 0.4372204840183258,
      "rewards/forensics_strict_edge_reward": 0.27100858092308044,
      "rewards/forensics_tiou_reward": 0.3056223690509796,
      "step": 1464
    },
    {
      "completion_length": 77.828125,
      "epoch": 3.0649214659685864,
      "grad_norm": 19.150209426879883,
      "kl": 10.40625,
      "learning_rate": 2.3270440251572326e-07,
      "loss": 0.7146,
      "reward": 0.448453351855278,
      "reward_std": 0.24157869815826416,
      "reward_std_all": 0.29391469806432724,
      "rewards-max/forensics_sinkhorn_reward": 0.9428059756755829,
      "rewards-max/forensics_strict_edge_reward": 0.8755655884742737,
      "rewards-max/forensics_tiou_reward": 0.9031253755092621,
      "rewards/forensics_sinkhorn_reward": 0.3810001462697983,
      "rewards/forensics_strict_edge_reward": 0.23730813711881638,
      "rewards/forensics_tiou_reward": 0.26066168397665024,
      "step": 1465
    },
    {
      "completion_length": 77.453125,
      "epoch": 3.067015706806283,
      "grad_norm": 23.4419002532959,
      "kl": 8.1875,
      "learning_rate": 2.3218029350104821e-07,
      "loss": 0.6018,
      "reward": 0.5120820701122284,
      "reward_std": 0.23893781751394272,
      "reward_std_all": 0.2580416277050972,
      "rewards-max/forensics_sinkhorn_reward": 0.9583784341812134,
      "rewards-max/forensics_strict_edge_reward": 0.9093653857707977,
      "rewards-max/forensics_tiou_reward": 0.9312390983104706,
      "rewards/forensics_sinkhorn_reward": 0.42210080474615097,
      "rewards/forensics_strict_edge_reward": 0.3258567899465561,
      "rewards/forensics_tiou_reward": 0.33406367897987366,
      "step": 1466
    },
    {
      "completion_length": 62.15625,
      "epoch": 3.069109947643979,
      "grad_norm": 16.10323715209961,
      "kl": 7.28125,
      "learning_rate": 2.3165618448637317e-07,
      "loss": 0.5791,
      "reward": 0.49446868896484375,
      "reward_std": 0.2912779599428177,
      "reward_std_all": 0.29865823686122894,
      "rewards-max/forensics_sinkhorn_reward": 0.9879591763019562,
      "rewards-max/forensics_strict_edge_reward": 0.9782643616199493,
      "rewards-max/forensics_tiou_reward": 0.9959432184696198,
      "rewards/forensics_sinkhorn_reward": 0.3943168818950653,
      "rewards/forensics_strict_edge_reward": 0.24570368975400925,
      "rewards/forensics_tiou_reward": 0.2688399404287338,
      "step": 1467
    },
    {
      "completion_length": 65.75,
      "epoch": 3.0712041884816754,
      "grad_norm": 29.828502655029297,
      "kl": 8.28125,
      "learning_rate": 2.311320754716981e-07,
      "loss": 0.5677,
      "reward": 0.5556953996419907,
      "reward_std": 0.2001761794090271,
      "reward_std_all": 0.21898587048053741,
      "rewards-max/forensics_sinkhorn_reward": 0.9873475134372711,
      "rewards-max/forensics_strict_edge_reward": 0.9565287232398987,
      "rewards-max/forensics_tiou_reward": 0.9935237169265747,
      "rewards/forensics_sinkhorn_reward": 0.5063436031341553,
      "rewards/forensics_strict_edge_reward": 0.31705112755298615,
      "rewards/forensics_tiou_reward": 0.39473727345466614,
      "step": 1468
    },
    {
      "completion_length": 57.4375,
      "epoch": 3.073298429319372,
      "grad_norm": 21.994857788085938,
      "kl": 7.34375,
      "learning_rate": 2.3060796645702304e-07,
      "loss": 0.558,
      "reward": 0.5705572366714478,
      "reward_std": 0.22228213399648666,
      "reward_std_all": 0.23221522569656372,
      "rewards-max/forensics_sinkhorn_reward": 0.9770578742027283,
      "rewards-max/forensics_strict_edge_reward": 0.8983768224716187,
      "rewards-max/forensics_tiou_reward": 0.9723438918590546,
      "rewards/forensics_sinkhorn_reward": 0.5150921791791916,
      "rewards/forensics_strict_edge_reward": 0.33001038432121277,
      "rewards/forensics_tiou_reward": 0.39023658633232117,
      "step": 1469
    },
    {
      "completion_length": 72.265625,
      "epoch": 3.075392670157068,
      "grad_norm": 25.396106719970703,
      "kl": 11.71875,
      "learning_rate": 2.3008385744234802e-07,
      "loss": 0.7318,
      "reward": 0.4425610452890396,
      "reward_std": 0.2805434763431549,
      "reward_std_all": 0.3202529847621918,
      "rewards-max/forensics_sinkhorn_reward": 0.9776321053504944,
      "rewards-max/forensics_strict_edge_reward": 0.8404377698898315,
      "rewards-max/forensics_tiou_reward": 0.9721408188343048,
      "rewards/forensics_sinkhorn_reward": 0.3530634641647339,
      "rewards/forensics_strict_edge_reward": 0.1809522807598114,
      "rewards/forensics_tiou_reward": 0.24123543500900269,
      "step": 1470
    },
    {
      "completion_length": 44.703125,
      "epoch": 3.0774869109947645,
      "grad_norm": 20.05362892150879,
      "kl": 6.625,
      "learning_rate": 2.2955974842767295e-07,
      "loss": 0.5092,
      "reward": 0.5856417715549469,
      "reward_std": 0.24315230548381805,
      "reward_std_all": 0.24239402264356613,
      "rewards-max/forensics_sinkhorn_reward": 0.9855566620826721,
      "rewards-max/forensics_strict_edge_reward": 0.8286563754081726,
      "rewards-max/forensics_tiou_reward": 0.9865420758724213,
      "rewards/forensics_sinkhorn_reward": 0.5360403656959534,
      "rewards/forensics_strict_edge_reward": 0.2914707809686661,
      "rewards/forensics_tiou_reward": 0.39150361716747284,
      "step": 1471
    },
    {
      "completion_length": 86.3125,
      "epoch": 3.0795811518324605,
      "grad_norm": 20.12171173095703,
      "kl": 8.5625,
      "learning_rate": 2.290356394129979e-07,
      "loss": 0.5896,
      "reward": 0.5314029455184937,
      "reward_std": 0.2241794541478157,
      "reward_std_all": 0.2480441927909851,
      "rewards-max/forensics_sinkhorn_reward": 0.968452513217926,
      "rewards-max/forensics_strict_edge_reward": 0.9256840348243713,
      "rewards-max/forensics_tiou_reward": 0.9530744552612305,
      "rewards/forensics_sinkhorn_reward": 0.41306111216545105,
      "rewards/forensics_strict_edge_reward": 0.22702983021736145,
      "rewards/forensics_tiou_reward": 0.2808680683374405,
      "step": 1472
    },
    {
      "completion_length": 49.59375,
      "epoch": 3.081675392670157,
      "grad_norm": 18.451414108276367,
      "kl": 7.609375,
      "learning_rate": 2.2851153039832282e-07,
      "loss": 0.6031,
      "reward": 0.5199495851993561,
      "reward_std": 0.2735186368227005,
      "reward_std_all": 0.2699131369590759,
      "rewards-max/forensics_sinkhorn_reward": 0.9861113727092743,
      "rewards-max/forensics_strict_edge_reward": 0.9357379674911499,
      "rewards-max/forensics_tiou_reward": 0.9886776208877563,
      "rewards/forensics_sinkhorn_reward": 0.45842456817626953,
      "rewards/forensics_strict_edge_reward": 0.2633323222398758,
      "rewards/forensics_tiou_reward": 0.3372969925403595,
      "step": 1473
    },
    {
      "completion_length": 38.140625,
      "epoch": 3.0837696335078535,
      "grad_norm": 18.364730834960938,
      "kl": 5.828125,
      "learning_rate": 2.279874213836478e-07,
      "loss": 0.4594,
      "reward": 0.6107094287872314,
      "reward_std": 0.2194553092122078,
      "reward_std_all": 0.20707383751869202,
      "rewards-max/forensics_sinkhorn_reward": 0.9720126688480377,
      "rewards-max/forensics_strict_edge_reward": 0.9256840348243713,
      "rewards-max/forensics_tiou_reward": 0.9603479206562042,
      "rewards/forensics_sinkhorn_reward": 0.5568542182445526,
      "rewards/forensics_strict_edge_reward": 0.34476783871650696,
      "rewards/forensics_tiou_reward": 0.4279375225305557,
      "step": 1474
    },
    {
      "completion_length": 63.0,
      "epoch": 3.0858638743455495,
      "grad_norm": 18.93796730041504,
      "kl": 7.90625,
      "learning_rate": 2.2746331236897274e-07,
      "loss": 0.5273,
      "reward": 0.5660877525806427,
      "reward_std": 0.17197050154209137,
      "reward_std_all": 0.2114424705505371,
      "rewards-max/forensics_sinkhorn_reward": 0.987395316362381,
      "rewards-max/forensics_strict_edge_reward": 0.9355069994926453,
      "rewards-max/forensics_tiou_reward": 0.9937319159507751,
      "rewards/forensics_sinkhorn_reward": 0.5031486451625824,
      "rewards/forensics_strict_edge_reward": 0.28915630280971527,
      "rewards/forensics_tiou_reward": 0.32601362466812134,
      "step": 1475
    },
    {
      "completion_length": 51.3125,
      "epoch": 3.087958115183246,
      "grad_norm": 29.627939224243164,
      "kl": 6.859375,
      "learning_rate": 2.2693920335429767e-07,
      "loss": 0.5918,
      "reward": 0.4805138409137726,
      "reward_std": 0.20797187834978104,
      "reward_std_all": 0.2624686285853386,
      "rewards-max/forensics_sinkhorn_reward": 0.9838442802429199,
      "rewards-max/forensics_strict_edge_reward": 0.9093653857707977,
      "rewards-max/forensics_tiou_reward": 0.9799351692199707,
      "rewards/forensics_sinkhorn_reward": 0.3955300599336624,
      "rewards/forensics_strict_edge_reward": 0.15868611633777618,
      "rewards/forensics_tiou_reward": 0.1875811591744423,
      "step": 1476
    },
    {
      "completion_length": 78.796875,
      "epoch": 3.0900523560209425,
      "grad_norm": 46.40288543701172,
      "kl": 12.78125,
      "learning_rate": 2.2641509433962263e-07,
      "loss": 0.7068,
      "reward": 0.4291457086801529,
      "reward_std": 0.23196063935756683,
      "reward_std_all": 0.3105389028787613,
      "rewards-max/forensics_sinkhorn_reward": 0.915777713060379,
      "rewards-max/forensics_strict_edge_reward": 0.8260076344013214,
      "rewards-max/forensics_tiou_reward": 0.8405795693397522,
      "rewards/forensics_sinkhorn_reward": 0.3327896296977997,
      "rewards/forensics_strict_edge_reward": 0.1909751296043396,
      "rewards/forensics_tiou_reward": 0.23095161467790604,
      "step": 1477
    },
    {
      "completion_length": 77.328125,
      "epoch": 3.0921465968586386,
      "grad_norm": 23.35735511779785,
      "kl": 9.4375,
      "learning_rate": 2.2589098532494759e-07,
      "loss": 0.6517,
      "reward": 0.5039342939853668,
      "reward_std": 0.15909036993980408,
      "reward_std_all": 0.2169637829065323,
      "rewards-max/forensics_sinkhorn_reward": 0.9683229625225067,
      "rewards-max/forensics_strict_edge_reward": 0.8528120815753937,
      "rewards-max/forensics_tiou_reward": 0.9520633220672607,
      "rewards/forensics_sinkhorn_reward": 0.39020177721977234,
      "rewards/forensics_strict_edge_reward": 0.20851725339889526,
      "rewards/forensics_tiou_reward": 0.2542576342821121,
      "step": 1478
    },
    {
      "completion_length": 50.671875,
      "epoch": 3.094240837696335,
      "grad_norm": 54.531890869140625,
      "kl": 8.71875,
      "learning_rate": 2.2536687631027252e-07,
      "loss": 0.5396,
      "reward": 0.5320512354373932,
      "reward_std": 0.25855443626642227,
      "reward_std_all": 0.2671818360686302,
      "rewards-max/forensics_sinkhorn_reward": 0.9856287240982056,
      "rewards-max/forensics_strict_edge_reward": 0.8876297473907471,
      "rewards-max/forensics_tiou_reward": 0.9868117272853851,
      "rewards/forensics_sinkhorn_reward": 0.4781161844730377,
      "rewards/forensics_strict_edge_reward": 0.27104412019252777,
      "rewards/forensics_tiou_reward": 0.348557785153389,
      "step": 1479
    },
    {
      "completion_length": 60.203125,
      "epoch": 3.0963350785340316,
      "grad_norm": 18.230998992919922,
      "kl": 7.40625,
      "learning_rate": 2.2484276729559748e-07,
      "loss": 0.3676,
      "reward": 0.6420551538467407,
      "reward_std": 0.19053082913160324,
      "reward_std_all": 0.18433088809251785,
      "rewards-max/forensics_sinkhorn_reward": 0.9863251149654388,
      "rewards-max/forensics_strict_edge_reward": 0.878633052110672,
      "rewards-max/forensics_tiou_reward": 0.9894946813583374,
      "rewards/forensics_sinkhorn_reward": 0.5510440170764923,
      "rewards/forensics_strict_edge_reward": 0.2585725486278534,
      "rewards/forensics_tiou_reward": 0.3485654145479202,
      "step": 1480
    },
    {
      "completion_length": 82.640625,
      "epoch": 3.0984293193717276,
      "grad_norm": 22.757036209106445,
      "kl": 11.53125,
      "learning_rate": 2.243186582809224e-07,
      "loss": 0.8011,
      "reward": 0.43817779421806335,
      "reward_std": 0.21268105506896973,
      "reward_std_all": 0.2935318648815155,
      "rewards-max/forensics_sinkhorn_reward": 0.9260866940021515,
      "rewards-max/forensics_strict_edge_reward": 0.8285773992538452,
      "rewards-max/forensics_tiou_reward": 0.8705973923206329,
      "rewards/forensics_sinkhorn_reward": 0.3489355593919754,
      "rewards/forensics_strict_edge_reward": 0.18230605125427246,
      "rewards/forensics_tiou_reward": 0.24392467737197876,
      "step": 1481
    },
    {
      "completion_length": 69.390625,
      "epoch": 3.100523560209424,
      "grad_norm": 18.07185935974121,
      "kl": 7.890625,
      "learning_rate": 2.2379454926624737e-07,
      "loss": 0.5477,
      "reward": 0.47508007287979126,
      "reward_std": 0.23742344975471497,
      "reward_std_all": 0.2779115289449692,
      "rewards-max/forensics_sinkhorn_reward": 0.987443208694458,
      "rewards-max/forensics_strict_edge_reward": 0.9565287232398987,
      "rewards-max/forensics_tiou_reward": 0.9939008355140686,
      "rewards/forensics_sinkhorn_reward": 0.39019547402858734,
      "rewards/forensics_strict_edge_reward": 0.2512326240539551,
      "rewards/forensics_tiou_reward": 0.27609285712242126,
      "step": 1482
    },
    {
      "completion_length": 41.375,
      "epoch": 3.1026178010471206,
      "grad_norm": 21.320417404174805,
      "kl": 6.203125,
      "learning_rate": 2.2327044025157233e-07,
      "loss": 0.3273,
      "reward": 0.5638999044895172,
      "reward_std": 0.21867398917675018,
      "reward_std_all": 0.23280596733093262,
      "rewards-max/forensics_sinkhorn_reward": 0.9664556384086609,
      "rewards-max/forensics_strict_edge_reward": 0.8698340952396393,
      "rewards-max/forensics_tiou_reward": 0.9487111568450928,
      "rewards/forensics_sinkhorn_reward": 0.4436013847589493,
      "rewards/forensics_strict_edge_reward": 0.25134437531232834,
      "rewards/forensics_tiou_reward": 0.26949694752693176,
      "step": 1483
    },
    {
      "completion_length": 30.09375,
      "epoch": 3.1047120418848166,
      "grad_norm": 25.09512710571289,
      "kl": 4.828125,
      "learning_rate": 2.2274633123689726e-07,
      "loss": 0.2946,
      "reward": 0.5979141294956207,
      "reward_std": 0.1801031082868576,
      "reward_std_all": 0.19414857774972916,
      "rewards-max/forensics_sinkhorn_reward": 0.9671518504619598,
      "rewards-max/forensics_strict_edge_reward": 0.9149472117424011,
      "rewards-max/forensics_tiou_reward": 0.9529761970043182,
      "rewards/forensics_sinkhorn_reward": 0.49283748865127563,
      "rewards/forensics_strict_edge_reward": 0.24168019741773605,
      "rewards/forensics_tiou_reward": 0.3073095381259918,
      "step": 1484
    },
    {
      "completion_length": 49.921875,
      "epoch": 3.106806282722513,
      "grad_norm": 17.047401428222656,
      "kl": 6.71875,
      "learning_rate": 2.222222222222222e-07,
      "loss": 0.4931,
      "reward": 0.548913836479187,
      "reward_std": 0.20862974226474762,
      "reward_std_all": 0.22780166566371918,
      "rewards-max/forensics_sinkhorn_reward": 0.8813445270061493,
      "rewards-max/forensics_strict_edge_reward": 0.820955753326416,
      "rewards-max/forensics_tiou_reward": 0.8568381369113922,
      "rewards/forensics_sinkhorn_reward": 0.4088567793369293,
      "rewards/forensics_strict_edge_reward": 0.2949126064777374,
      "rewards/forensics_tiou_reward": 0.3086778521537781,
      "step": 1485
    },
    {
      "completion_length": 71.8125,
      "epoch": 3.1089005235602096,
      "grad_norm": 20.66794776916504,
      "kl": 7.796875,
      "learning_rate": 2.2169811320754718e-07,
      "loss": 0.5485,
      "reward": 0.5448790192604065,
      "reward_std": 0.2393144816160202,
      "reward_std_all": 0.25178661197423935,
      "rewards-max/forensics_sinkhorn_reward": 0.9879891276359558,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9960553348064423,
      "rewards/forensics_sinkhorn_reward": 0.49182629585266113,
      "rewards/forensics_strict_edge_reward": 0.3569314777851105,
      "rewards/forensics_tiou_reward": 0.36743292212486267,
      "step": 1486
    },
    {
      "completion_length": 59.234375,
      "epoch": 3.1109947643979057,
      "grad_norm": 26.80450439453125,
      "kl": 7.9375,
      "learning_rate": 2.211740041928721e-07,
      "loss": 0.5771,
      "reward": 0.5121368169784546,
      "reward_std": 0.2822640389204025,
      "reward_std_all": 0.28992196917533875,
      "rewards-max/forensics_sinkhorn_reward": 0.9475095272064209,
      "rewards-max/forensics_strict_edge_reward": 0.7803006768226624,
      "rewards-max/forensics_tiou_reward": 0.9055235385894775,
      "rewards/forensics_sinkhorn_reward": 0.42159128189086914,
      "rewards/forensics_strict_edge_reward": 0.20499558001756668,
      "rewards/forensics_tiou_reward": 0.27517732977867126,
      "step": 1487
    },
    {
      "completion_length": 31.90625,
      "epoch": 3.113089005235602,
      "grad_norm": 22.35331153869629,
      "kl": 5.609375,
      "learning_rate": 2.2064989517819704e-07,
      "loss": 0.2402,
      "reward": 0.5183451771736145,
      "reward_std": 0.20428411662578583,
      "reward_std_all": 0.22783903777599335,
      "rewards-max/forensics_sinkhorn_reward": 0.9754113256931305,
      "rewards-max/forensics_strict_edge_reward": 0.9364311099052429,
      "rewards-max/forensics_tiou_reward": 0.9707275629043579,
      "rewards/forensics_sinkhorn_reward": 0.4191490411758423,
      "rewards/forensics_strict_edge_reward": 0.17697813361883163,
      "rewards/forensics_tiou_reward": 0.21078843623399734,
      "step": 1488
    },
    {
      "completion_length": 45.765625,
      "epoch": 3.115183246073298,
      "grad_norm": 21.69141387939453,
      "kl": 6.125,
      "learning_rate": 2.20125786163522e-07,
      "loss": 0.4589,
      "reward": 0.563008189201355,
      "reward_std": 0.15302052348852158,
      "reward_std_all": 0.19609618186950684,
      "rewards-max/forensics_sinkhorn_reward": 0.9878120720386505,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9953621625900269,
      "rewards/forensics_sinkhorn_reward": 0.4816882610321045,
      "rewards/forensics_strict_edge_reward": 0.2213243916630745,
      "rewards/forensics_tiou_reward": 0.31002240628004074,
      "step": 1489
    },
    {
      "completion_length": 36.359375,
      "epoch": 3.1172774869109947,
      "grad_norm": 16.940013885498047,
      "kl": 4.390625,
      "learning_rate": 2.1960167714884696e-07,
      "loss": 0.3228,
      "reward": 0.6241314709186554,
      "reward_std": 0.1796145923435688,
      "reward_std_all": 0.1805744394659996,
      "rewards-max/forensics_sinkhorn_reward": 0.9784019589424133,
      "rewards-max/forensics_strict_edge_reward": 0.9003686904907227,
      "rewards-max/forensics_tiou_reward": 0.9759786128997803,
      "rewards/forensics_sinkhorn_reward": 0.5061423182487488,
      "rewards/forensics_strict_edge_reward": 0.2927483022212982,
      "rewards/forensics_tiou_reward": 0.33436594903469086,
      "step": 1490
    },
    {
      "completion_length": 37.203125,
      "epoch": 3.119371727748691,
      "grad_norm": 17.618040084838867,
      "kl": 6.640625,
      "learning_rate": 2.190775681341719e-07,
      "loss": 0.4174,
      "reward": 0.545308381319046,
      "reward_std": 0.26075099408626556,
      "reward_std_all": 0.2592564821243286,
      "rewards-max/forensics_sinkhorn_reward": 0.9884204268455505,
      "rewards-max/forensics_strict_edge_reward": 0.927969753742218,
      "rewards-max/forensics_tiou_reward": 0.9977528154850006,
      "rewards/forensics_sinkhorn_reward": 0.49442707002162933,
      "rewards/forensics_strict_edge_reward": 0.2705548405647278,
      "rewards/forensics_tiou_reward": 0.37899209558963776,
      "step": 1491
    },
    {
      "completion_length": 21.578125,
      "epoch": 3.1214659685863873,
      "grad_norm": 32.18949508666992,
      "kl": 5.0703125,
      "learning_rate": 2.1855345911949685e-07,
      "loss": 0.3237,
      "reward": 0.6241130232810974,
      "reward_std": 0.17026406526565552,
      "reward_std_all": 0.18071583658456802,
      "rewards-max/forensics_sinkhorn_reward": 0.9730073809623718,
      "rewards-max/forensics_strict_edge_reward": 0.7666849195957184,
      "rewards-max/forensics_tiou_reward": 0.9602329432964325,
      "rewards/forensics_sinkhorn_reward": 0.5049979090690613,
      "rewards/forensics_strict_edge_reward": 0.26026230305433273,
      "rewards/forensics_tiou_reward": 0.3505018502473831,
      "step": 1492
    },
    {
      "completion_length": 34.578125,
      "epoch": 3.1235602094240837,
      "grad_norm": 23.078542709350586,
      "kl": 5.921875,
      "learning_rate": 2.1802935010482178e-07,
      "loss": 0.4429,
      "reward": 0.5710014998912811,
      "reward_std": 0.20780684053897858,
      "reward_std_all": 0.220248781144619,
      "rewards-max/forensics_sinkhorn_reward": 0.9679911136627197,
      "rewards-max/forensics_strict_edge_reward": 0.9185642302036285,
      "rewards-max/forensics_tiou_reward": 0.9530540108680725,
      "rewards/forensics_sinkhorn_reward": 0.5160960108041763,
      "rewards/forensics_strict_edge_reward": 0.36970191448926926,
      "rewards/forensics_tiou_reward": 0.408655509352684,
      "step": 1493
    },
    {
      "completion_length": 37.484375,
      "epoch": 3.1256544502617802,
      "grad_norm": 36.13386917114258,
      "kl": 6.2265625,
      "learning_rate": 2.1750524109014674e-07,
      "loss": 0.2061,
      "reward": 0.6309430301189423,
      "reward_std": 0.1732543781399727,
      "reward_std_all": 0.17754462361335754,
      "rewards-max/forensics_sinkhorn_reward": 0.9774373471736908,
      "rewards-max/forensics_strict_edge_reward": 0.8853305876255035,
      "rewards-max/forensics_tiou_reward": 0.9741575717926025,
      "rewards/forensics_sinkhorn_reward": 0.5253775417804718,
      "rewards/forensics_strict_edge_reward": 0.2504759877920151,
      "rewards/forensics_tiou_reward": 0.3727010190486908,
      "step": 1494
    },
    {
      "completion_length": 29.203125,
      "epoch": 3.1277486910994763,
      "grad_norm": 18.878511428833008,
      "kl": 5.15625,
      "learning_rate": 2.169811320754717e-07,
      "loss": 0.4139,
      "reward": 0.6039375960826874,
      "reward_std": 0.1850561425089836,
      "reward_std_all": 0.19762159138917923,
      "rewards-max/forensics_sinkhorn_reward": 0.9880712926387787,
      "rewards-max/forensics_strict_edge_reward": 0.9782643616199493,
      "rewards-max/forensics_tiou_reward": 0.9963833689689636,
      "rewards/forensics_sinkhorn_reward": 0.5663114190101624,
      "rewards/forensics_strict_edge_reward": 0.3382353335618973,
      "rewards/forensics_tiou_reward": 0.4549497961997986,
      "step": 1495
    },
    {
      "completion_length": 61.21875,
      "epoch": 3.129842931937173,
      "grad_norm": 15.47856616973877,
      "kl": 7.46875,
      "learning_rate": 2.1645702306079663e-07,
      "loss": 0.4538,
      "reward": 0.5903449058532715,
      "reward_std": 0.26566487550735474,
      "reward_std_all": 0.24587763845920563,
      "rewards-max/forensics_sinkhorn_reward": 0.988130509853363,
      "rewards-max/forensics_strict_edge_reward": 0.9782643616199493,
      "rewards-max/forensics_tiou_reward": 0.996610164642334,
      "rewards/forensics_sinkhorn_reward": 0.5020225793123245,
      "rewards/forensics_strict_edge_reward": 0.2892490327358246,
      "rewards/forensics_tiou_reward": 0.35874637961387634,
      "step": 1496
    },
    {
      "completion_length": 66.703125,
      "epoch": 3.1319371727748693,
      "grad_norm": 27.343406677246094,
      "kl": 7.75,
      "learning_rate": 2.1593291404612156e-07,
      "loss": 0.3771,
      "reward": 0.5682045221328735,
      "reward_std": 0.1977568119764328,
      "reward_std_all": 0.22209803014993668,
      "rewards-max/forensics_sinkhorn_reward": 0.9463672637939453,
      "rewards-max/forensics_strict_edge_reward": 0.9185642302036285,
      "rewards-max/forensics_tiou_reward": 0.9130415320396423,
      "rewards/forensics_sinkhorn_reward": 0.5307205021381378,
      "rewards/forensics_strict_edge_reward": 0.3123921975493431,
      "rewards/forensics_tiou_reward": 0.38797180354595184,
      "step": 1497
    },
    {
      "completion_length": 45.59375,
      "epoch": 3.1340314136125653,
      "grad_norm": 39.190895080566406,
      "kl": 6.421875,
      "learning_rate": 2.1540880503144655e-07,
      "loss": 0.4063,
      "reward": 0.6274948418140411,
      "reward_std": 0.1510540395975113,
      "reward_std_all": 0.1640535220503807,
      "rewards-max/forensics_sinkhorn_reward": 0.9876911342144012,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9948855936527252,
      "rewards/forensics_sinkhorn_reward": 0.5846775472164154,
      "rewards/forensics_strict_edge_reward": 0.3842950016260147,
      "rewards/forensics_tiou_reward": 0.44547218084335327,
      "step": 1498
    },
    {
      "completion_length": 24.765625,
      "epoch": 3.136125654450262,
      "grad_norm": 26.476566314697266,
      "kl": 5.28125,
      "learning_rate": 2.1488469601677148e-07,
      "loss": 0.372,
      "reward": 0.5701529234647751,
      "reward_std": 0.1972031369805336,
      "reward_std_all": 0.22152934223413467,
      "rewards-max/forensics_sinkhorn_reward": 0.9885096848011017,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9981054365634918,
      "rewards/forensics_sinkhorn_reward": 0.48471832275390625,
      "rewards/forensics_strict_edge_reward": 0.2902360185980797,
      "rewards/forensics_tiou_reward": 0.3263065069913864,
      "step": 1499
    },
    {
      "completion_length": 61.015625,
      "epoch": 3.1382198952879583,
      "grad_norm": 16.895265579223633,
      "kl": 8.75,
      "learning_rate": 2.1436058700209641e-07,
      "loss": 0.6794,
      "reward": 0.5562845170497894,
      "reward_std": 0.23103483021259308,
      "reward_std_all": 0.24073395878076553,
      "rewards-max/forensics_sinkhorn_reward": 0.9861470460891724,
      "rewards-max/forensics_strict_edge_reward": 0.946017861366272,
      "rewards-max/forensics_tiou_reward": 0.9901401400566101,
      "rewards/forensics_sinkhorn_reward": 0.5122996270656586,
      "rewards/forensics_strict_edge_reward": 0.3340926170349121,
      "rewards/forensics_tiou_reward": 0.42245206236839294,
      "step": 1500
    },
    {
      "completion_length": 56.359375,
      "epoch": 3.1403141361256544,
      "grad_norm": 26.952857971191406,
      "kl": 8.3125,
      "learning_rate": 2.1383647798742137e-07,
      "loss": 0.4986,
      "reward": 0.5579676032066345,
      "reward_std": 0.2711389809846878,
      "reward_std_all": 0.2581193745136261,
      "rewards-max/forensics_sinkhorn_reward": 0.9884916841983795,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9980336129665375,
      "rewards/forensics_sinkhorn_reward": 0.46158739924430847,
      "rewards/forensics_strict_edge_reward": 0.34082718193531036,
      "rewards/forensics_tiou_reward": 0.3713715076446533,
      "step": 1501
    },
    {
      "completion_length": 71.59375,
      "epoch": 3.142408376963351,
      "grad_norm": 17.251218795776367,
      "kl": 10.6875,
      "learning_rate": 2.1331236897274633e-07,
      "loss": 0.6807,
      "reward": 0.44194942712783813,
      "reward_std": 0.2859746068716049,
      "reward_std_all": 0.3291446417570114,
      "rewards-max/forensics_sinkhorn_reward": 0.9294571280479431,
      "rewards-max/forensics_strict_edge_reward": 0.8719756007194519,
      "rewards-max/forensics_tiou_reward": 0.8771942555904388,
      "rewards/forensics_sinkhorn_reward": 0.34221772849559784,
      "rewards/forensics_strict_edge_reward": 0.2317357286810875,
      "rewards/forensics_tiou_reward": 0.23934665322303772,
      "step": 1502
    },
    {
      "completion_length": 100.46875,
      "epoch": 3.144502617801047,
      "grad_norm": 639.650390625,
      "kl": 16.03125,
      "learning_rate": 2.1278825995807126e-07,
      "loss": 0.9639,
      "reward": 0.38867107033729553,
      "reward_std": 0.2781359553337097,
      "reward_std_all": 0.33238130807876587,
      "rewards-max/forensics_sinkhorn_reward": 0.9365682601928711,
      "rewards-max/forensics_strict_edge_reward": 0.7744058072566986,
      "rewards-max/forensics_tiou_reward": 0.8847456574440002,
      "rewards/forensics_sinkhorn_reward": 0.28848162293434143,
      "rewards/forensics_strict_edge_reward": 0.10830526426434517,
      "rewards/forensics_tiou_reward": 0.16476260870695114,
      "step": 1503
    },
    {
      "completion_length": 53.46875,
      "epoch": 3.1465968586387434,
      "grad_norm": 34.042964935302734,
      "kl": 8.28125,
      "learning_rate": 2.1226415094339622e-07,
      "loss": 0.617,
      "reward": 0.5131707787513733,
      "reward_std": 0.2547140419483185,
      "reward_std_all": 0.27996407449245453,
      "rewards-max/forensics_sinkhorn_reward": 0.9888888597488403,
      "rewards-max/forensics_strict_edge_reward": 0.9890114367008209,
      "rewards-max/forensics_tiou_reward": 0.999599426984787,
      "rewards/forensics_sinkhorn_reward": 0.4431392252445221,
      "rewards/forensics_strict_edge_reward": 0.27040117233991623,
      "rewards/forensics_tiou_reward": 0.3211747258901596,
      "step": 1504
    },
    {
      "completion_length": 50.5,
      "epoch": 3.14869109947644,
      "grad_norm": 19.973346710205078,
      "kl": 6.53125,
      "learning_rate": 2.1174004192872115e-07,
      "loss": 0.3251,
      "reward": 0.5967036187648773,
      "reward_std": 0.20451593399047852,
      "reward_std_all": 0.2063525840640068,
      "rewards-max/forensics_sinkhorn_reward": 0.9570946395397186,
      "rewards-max/forensics_strict_edge_reward": 0.935806930065155,
      "rewards-max/forensics_tiou_reward": 0.9271453320980072,
      "rewards/forensics_sinkhorn_reward": 0.47073037922382355,
      "rewards/forensics_strict_edge_reward": 0.2816891074180603,
      "rewards/forensics_tiou_reward": 0.3087007850408554,
      "step": 1505
    },
    {
      "completion_length": 45.515625,
      "epoch": 3.150785340314136,
      "grad_norm": 19.427682876586914,
      "kl": 6.890625,
      "learning_rate": 2.112159329140461e-07,
      "loss": 0.529,
      "reward": 0.5495114326477051,
      "reward_std": 0.27214573323726654,
      "reward_std_all": 0.2721564471721649,
      "rewards-max/forensics_sinkhorn_reward": 0.9881254434585571,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9965901374816895,
      "rewards/forensics_sinkhorn_reward": 0.4956264793872833,
      "rewards/forensics_strict_edge_reward": 0.26119523495435715,
      "rewards/forensics_tiou_reward": 0.33963824808597565,
      "step": 1506
    },
    {
      "completion_length": 77.59375,
      "epoch": 3.1528795811518324,
      "grad_norm": 14.022793769836426,
      "kl": 9.03125,
      "learning_rate": 2.1069182389937107e-07,
      "loss": 0.6909,
      "reward": 0.4348718822002411,
      "reward_std": 0.2473166063427925,
      "reward_std_all": 0.3087593615055084,
      "rewards-max/forensics_sinkhorn_reward": 0.9884745478630066,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9979666471481323,
      "rewards/forensics_sinkhorn_reward": 0.39699891209602356,
      "rewards/forensics_strict_edge_reward": 0.2557419091463089,
      "rewards/forensics_tiou_reward": 0.3072744458913803,
      "step": 1507
    },
    {
      "completion_length": 55.375,
      "epoch": 3.154973821989529,
      "grad_norm": 25.670238494873047,
      "kl": 5.828125,
      "learning_rate": 2.10167714884696e-07,
      "loss": 0.5179,
      "reward": 0.5660648792982101,
      "reward_std": 0.2756164148449898,
      "reward_std_all": 0.2620372921228409,
      "rewards-max/forensics_sinkhorn_reward": 0.9877458214759827,
      "rewards-max/forensics_strict_edge_reward": 0.9565287232398987,
      "rewards-max/forensics_tiou_reward": 0.9950935244560242,
      "rewards/forensics_sinkhorn_reward": 0.5183579623699188,
      "rewards/forensics_strict_edge_reward": 0.3288863003253937,
      "rewards/forensics_tiou_reward": 0.3810860812664032,
      "step": 1508
    },
    {
      "completion_length": 71.140625,
      "epoch": 3.157068062827225,
      "grad_norm": 43.59366989135742,
      "kl": 10.09375,
      "learning_rate": 2.0964360587002094e-07,
      "loss": 0.714,
      "reward": 0.42918457090854645,
      "reward_std": 0.1814054474234581,
      "reward_std_all": 0.27148866653442383,
      "rewards-max/forensics_sinkhorn_reward": 0.986377090215683,
      "rewards-max/forensics_strict_edge_reward": 0.9357379674911499,
      "rewards-max/forensics_tiou_reward": 0.9897412359714508,
      "rewards/forensics_sinkhorn_reward": 0.32932157814502716,
      "rewards/forensics_strict_edge_reward": 0.2166137769818306,
      "rewards/forensics_tiou_reward": 0.2588864862918854,
      "step": 1509
    },
    {
      "completion_length": 51.34375,
      "epoch": 3.1591623036649215,
      "grad_norm": 19.268342971801758,
      "kl": 5.75,
      "learning_rate": 2.0911949685534592e-07,
      "loss": 0.4476,
      "reward": 0.5641405582427979,
      "reward_std": 0.22810839861631393,
      "reward_std_all": 0.2324843928217888,
      "rewards-max/forensics_sinkhorn_reward": 0.9476083219051361,
      "rewards-max/forensics_strict_edge_reward": 0.8457958102226257,
      "rewards-max/forensics_tiou_reward": 0.9033626019954681,
      "rewards/forensics_sinkhorn_reward": 0.49806439876556396,
      "rewards/forensics_strict_edge_reward": 0.28244373947381973,
      "rewards/forensics_tiou_reward": 0.3749482035636902,
      "step": 1510
    },
    {
      "completion_length": 39.109375,
      "epoch": 3.161256544502618,
      "grad_norm": 17.243183135986328,
      "kl": 3.8125,
      "learning_rate": 2.0859538784067085e-07,
      "loss": 0.1495,
      "reward": 0.6354475915431976,
      "reward_std": 0.17958280444145203,
      "reward_std_all": 0.17798474431037903,
      "rewards-max/forensics_sinkhorn_reward": 0.987801730632782,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9953139126300812,
      "rewards/forensics_sinkhorn_reward": 0.5485895872116089,
      "rewards/forensics_strict_edge_reward": 0.31793397665023804,
      "rewards/forensics_tiou_reward": 0.3596327155828476,
      "step": 1511
    },
    {
      "completion_length": 62.453125,
      "epoch": 3.163350785340314,
      "grad_norm": 22.11139678955078,
      "kl": 6.9765625,
      "learning_rate": 2.0807127882599579e-07,
      "loss": 0.5857,
      "reward": 0.5088794082403183,
      "reward_std": 0.2078588455915451,
      "reward_std_all": 0.25028523802757263,
      "rewards-max/forensics_sinkhorn_reward": 0.9879445135593414,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9958774745464325,
      "rewards/forensics_sinkhorn_reward": 0.4363745152950287,
      "rewards/forensics_strict_edge_reward": 0.2741795852780342,
      "rewards/forensics_tiou_reward": 0.30082110315561295,
      "step": 1512
    },
    {
      "completion_length": 47.96875,
      "epoch": 3.1654450261780105,
      "grad_norm": 21.70502471923828,
      "kl": 5.578125,
      "learning_rate": 2.0754716981132074e-07,
      "loss": 0.4841,
      "reward": 0.5534527003765106,
      "reward_std": 0.24511798471212387,
      "reward_std_all": 0.24714581668376923,
      "rewards-max/forensics_sinkhorn_reward": 0.9829250872135162,
      "rewards-max/forensics_strict_edge_reward": 0.8328413963317871,
      "rewards-max/forensics_tiou_reward": 0.9761038720607758,
      "rewards/forensics_sinkhorn_reward": 0.47438910603523254,
      "rewards/forensics_strict_edge_reward": 0.2944640517234802,
      "rewards/forensics_tiou_reward": 0.3638670742511749,
      "step": 1513
    },
    {
      "completion_length": 47.03125,
      "epoch": 3.167539267015707,
      "grad_norm": 19.771602630615234,
      "kl": 5.7421875,
      "learning_rate": 2.070230607966457e-07,
      "loss": 0.4604,
      "reward": 0.545329749584198,
      "reward_std": 0.19941619038581848,
      "reward_std_all": 0.22247151285409927,
      "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": 0.4646099805831909,
      "rewards/forensics_strict_edge_reward": 0.26380209624767303,
      "rewards/forensics_tiou_reward": 0.3378572463989258,
      "step": 1514
    },
    {
      "completion_length": 48.078125,
      "epoch": 3.169633507853403,
      "grad_norm": 2410.5537109375,
      "kl": 29.5,
      "learning_rate": 2.0649895178197064e-07,
      "loss": 1.5317,
      "reward": 0.5442421734333038,
      "reward_std": 0.199390210211277,
      "reward_std_all": 0.23385798931121826,
      "rewards-max/forensics_sinkhorn_reward": 0.9885500967502594,
      "rewards-max/forensics_strict_edge_reward": 0.9782643616199493,
      "rewards-max/forensics_tiou_reward": 0.9982653856277466,
      "rewards/forensics_sinkhorn_reward": 0.47247514128685,
      "rewards/forensics_strict_edge_reward": 0.2642035037279129,
      "rewards/forensics_tiou_reward": 0.373113676905632,
      "step": 1515
    },
    {
      "completion_length": 39.25,
      "epoch": 3.1717277486910995,
      "grad_norm": 22.336631774902344,
      "kl": 5.765625,
      "learning_rate": 2.059748427672956e-07,
      "loss": 0.2859,
      "reward": 0.5523003339767456,
      "reward_std": 0.24297881126403809,
      "reward_std_all": 0.23998530209064484,
      "rewards-max/forensics_sinkhorn_reward": 0.9739590883255005,
      "rewards-max/forensics_strict_edge_reward": 0.9357379674911499,
      "rewards-max/forensics_tiou_reward": 0.9641979336738586,
      "rewards/forensics_sinkhorn_reward": 0.46431417763233185,
      "rewards/forensics_strict_edge_reward": 0.22851350903511047,
      "rewards/forensics_tiou_reward": 0.28598426282405853,
      "step": 1516
    },
    {
      "completion_length": 69.28125,
      "epoch": 3.173821989528796,
      "grad_norm": 23.69939422607422,
      "kl": 7.59375,
      "learning_rate": 2.0545073375262053e-07,
      "loss": 0.4917,
      "reward": 0.5059142261743546,
      "reward_std": 0.19823160022497177,
      "reward_std_all": 0.24436869472265244,
      "rewards-max/forensics_sinkhorn_reward": 0.9724448025226593,
      "rewards-max/forensics_strict_edge_reward": 0.9149472117424011,
      "rewards-max/forensics_tiou_reward": 0.9604169726371765,
      "rewards/forensics_sinkhorn_reward": 0.40545275807380676,
      "rewards/forensics_strict_edge_reward": 0.27759896218776703,
      "rewards/forensics_tiou_reward": 0.30385899543762207,
      "step": 1517
    },
    {
      "completion_length": 28.828125,
      "epoch": 3.175916230366492,
      "grad_norm": 30.49044418334961,
      "kl": 5.90625,
      "learning_rate": 2.0492662473794548e-07,
      "loss": 0.468,
      "reward": 0.5799639523029327,
      "reward_std": 0.1745186150074005,
      "reward_std_all": 0.20004839450120926,
      "rewards-max/forensics_sinkhorn_reward": 0.9487152397632599,
      "rewards-max/forensics_strict_edge_reward": 0.7148464620113373,
      "rewards-max/forensics_tiou_reward": 0.9026798903942108,
      "rewards/forensics_sinkhorn_reward": 0.4512600898742676,
      "rewards/forensics_strict_edge_reward": 0.21445663273334503,
      "rewards/forensics_tiou_reward": 0.3398042768239975,
      "step": 1518
    },
    {
      "completion_length": 27.359375,
      "epoch": 3.1780104712041886,
      "grad_norm": 17.574615478515625,
      "kl": 6.75,
      "learning_rate": 2.0440251572327044e-07,
      "loss": 0.5753,
      "reward": 0.5374644100666046,
      "reward_std": 0.09356077201664448,
      "reward_std_all": 0.1744418516755104,
      "rewards-max/forensics_sinkhorn_reward": 0.9884831309318542,
      "rewards-max/forensics_strict_edge_reward": 0.9574736058712006,
      "rewards-max/forensics_tiou_reward": 0.9979999959468842,
      "rewards/forensics_sinkhorn_reward": 0.4711303263902664,
      "rewards/forensics_strict_edge_reward": 0.2873464822769165,
      "rewards/forensics_tiou_reward": 0.3508681505918503,
      "step": 1519
    },
    {
      "completion_length": 35.796875,
      "epoch": 3.180104712041885,
      "grad_norm": 18.658235549926758,
      "kl": 6.734375,
      "learning_rate": 2.0387840670859538e-07,
      "loss": 0.4442,
      "reward": 0.5654226839542389,
      "reward_std": 0.23582099378108978,
      "reward_std_all": 0.23057931661605835,
      "rewards-max/forensics_sinkhorn_reward": 0.9845298528671265,
      "rewards-max/forensics_strict_edge_reward": 0.9677534997463226,
      "rewards-max/forensics_tiou_reward": 0.9825825095176697,
      "rewards/forensics_sinkhorn_reward": 0.48979945480823517,
      "rewards/forensics_strict_edge_reward": 0.25504642724990845,
      "rewards/forensics_tiou_reward": 0.3579907864332199,
      "step": 1520
    },
    {
      "completion_length": 37.34375,
      "epoch": 3.182198952879581,
      "grad_norm": 25.22737693786621,
      "kl": 7.0078125,
      "learning_rate": 2.033542976939203e-07,
      "loss": 0.3859,
      "reward": 0.6018501222133636,
      "reward_std": 0.19099151343107224,
      "reward_std_all": 0.19478866457939148,
      "rewards-max/forensics_sinkhorn_reward": 0.9604355096817017,
      "rewards-max/forensics_strict_edge_reward": 0.9158510267734528,
      "rewards-max/forensics_tiou_reward": 0.9388128221035004,
      "rewards/forensics_sinkhorn_reward": 0.49241092801094055,
      "rewards/forensics_strict_edge_reward": 0.2641877084970474,
      "rewards/forensics_tiou_reward": 0.31745126843452454,
      "step": 1521
    },
    {
      "completion_length": 39.015625,
      "epoch": 3.1842931937172776,
      "grad_norm": 20.25463104248047,
      "kl": 7.6875,
      "learning_rate": 2.028301886792453e-07,
      "loss": 0.5988,
      "reward": 0.5220964550971985,
      "reward_std": 0.30860787630081177,
      "reward_std_all": 0.29373955726623535,
      "rewards-max/forensics_sinkhorn_reward": 0.9809812307357788,
      "rewards-max/forensics_strict_edge_reward": 0.947419673204422,
      "rewards-max/forensics_tiou_reward": 0.9801123738288879,
      "rewards/forensics_sinkhorn_reward": 0.4275854229927063,
      "rewards/forensics_strict_edge_reward": 0.3115391284227371,
      "rewards/forensics_tiou_reward": 0.33366599678993225,
      "step": 1522
    },
    {
      "completion_length": 29.265625,
      "epoch": 3.1863874345549736,
      "grad_norm": 26.369617462158203,
      "kl": 5.203125,
      "learning_rate": 2.0230607966457023e-07,
      "loss": 0.2468,
      "reward": 0.5553805530071259,
      "reward_std": 0.17299625277519226,
      "reward_std_all": 0.20783231407403946,
      "rewards-max/forensics_sinkhorn_reward": 0.9860706031322479,
      "rewards-max/forensics_strict_edge_reward": 0.9357379674911499,
      "rewards-max/forensics_tiou_reward": 0.9885235726833344,
      "rewards/forensics_sinkhorn_reward": 0.5045179724693298,
      "rewards/forensics_strict_edge_reward": 0.2906777262687683,
      "rewards/forensics_tiou_reward": 0.36091963946819305,
      "step": 1523
    },
    {
      "completion_length": 26.0,
      "epoch": 3.18848167539267,
      "grad_norm": 25.05600929260254,
      "kl": 7.53125,
      "learning_rate": 2.0178197064989516e-07,
      "loss": 0.5185,
      "reward": 0.49263398349285126,
      "reward_std": 0.25831490755081177,
      "reward_std_all": 0.27943067252635956,
      "rewards-max/forensics_sinkhorn_reward": 0.9884204268455505,
      "rewards-max/forensics_strict_edge_reward": 0.9782643616199493,
      "rewards-max/forensics_tiou_reward": 0.9977528154850006,
      "rewards/forensics_sinkhorn_reward": 0.406289279460907,
      "rewards/forensics_strict_edge_reward": 0.2900649905204773,
      "rewards/forensics_tiou_reward": 0.31758102774620056,
      "step": 1524
    },
    {
      "completion_length": 43.796875,
      "epoch": 3.1905759162303666,
      "grad_norm": 21.61482810974121,
      "kl": 9.46875,
      "learning_rate": 2.0125786163522012e-07,
      "loss": 0.5538,
      "reward": 0.513243705034256,
      "reward_std": 0.19660799950361252,
      "reward_std_all": 0.23128674924373627,
      "rewards-max/forensics_sinkhorn_reward": 0.8779958784580231,
      "rewards-max/forensics_strict_edge_reward": 0.8366184830665588,
      "rewards-max/forensics_tiou_reward": 0.7776240706443787,
      "rewards/forensics_sinkhorn_reward": 0.3690912127494812,
      "rewards/forensics_strict_edge_reward": 0.2267836183309555,
      "rewards/forensics_tiou_reward": 0.2244899570941925,
      "step": 1525
    },
    {
      "completion_length": 30.046875,
      "epoch": 3.1926701570680627,
      "grad_norm": 33.36606216430664,
      "kl": 8.3125,
      "learning_rate": 2.0073375262054508e-07,
      "loss": 0.5496,
      "reward": 0.4734383523464203,
      "reward_std": 0.20307499170303345,
      "reward_std_all": 0.2656368538737297,
      "rewards-max/forensics_sinkhorn_reward": 0.9880029261112213,
      "rewards-max/forensics_strict_edge_reward": 0.9565287232398987,
      "rewards-max/forensics_tiou_reward": 0.9961099922657013,
      "rewards/forensics_sinkhorn_reward": 0.42067283391952515,
      "rewards/forensics_strict_edge_reward": 0.22249694168567657,
      "rewards/forensics_tiou_reward": 0.3018312156200409,
      "step": 1526
    },
    {
      "completion_length": 32.8125,
      "epoch": 3.194764397905759,
      "grad_norm": 25.24671745300293,
      "kl": 6.6796875,
      "learning_rate": 2.0020964360587e-07,
      "loss": 0.4347,
      "reward": 0.610481321811676,
      "reward_std": 0.11735716462135315,
      "reward_std_all": 0.15795094147324562,
      "rewards-max/forensics_sinkhorn_reward": 0.9889057874679565,
      "rewards-max/forensics_strict_edge_reward": 0.9890114367008209,
      "rewards-max/forensics_tiou_reward": 0.9996661245822906,
      "rewards/forensics_sinkhorn_reward": 0.5820502936840057,
      "rewards/forensics_strict_edge_reward": 0.37419217824935913,
      "rewards/forensics_tiou_reward": 0.46798190474510193,
      "step": 1527
    },
    {
      "completion_length": 34.921875,
      "epoch": 3.1968586387434557,
      "grad_norm": 19.992952346801758,
      "kl": 6.34375,
      "learning_rate": 1.9968553459119497e-07,
      "loss": 0.5376,
      "reward": 0.5390060991048813,
      "reward_std": 0.24826619774103165,
      "reward_std_all": 0.26028626412153244,
      "rewards-max/forensics_sinkhorn_reward": 0.9885718822479248,
      "rewards-max/forensics_strict_edge_reward": 0.9890114367008209,
      "rewards-max/forensics_tiou_reward": 0.9983511865139008,
      "rewards/forensics_sinkhorn_reward": 0.4884493201971054,
      "rewards/forensics_strict_edge_reward": 0.2977568805217743,
      "rewards/forensics_tiou_reward": 0.3542429357767105,
      "step": 1528
    },
    {
      "completion_length": 33.25,
      "epoch": 3.1989528795811517,
      "grad_norm": 21.38730239868164,
      "kl": 5.890625,
      "learning_rate": 1.991614255765199e-07,
      "loss": 0.3757,
      "reward": 0.6061401665210724,
      "reward_std": 0.14134018123149872,
      "reward_std_all": 0.17549337074160576,
      "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": 0.5487026125192642,
      "rewards/forensics_strict_edge_reward": 0.37105444073677063,
      "rewards/forensics_tiou_reward": 0.46526408195495605,
      "step": 1529
    },
    {
      "completion_length": 28.765625,
      "epoch": 3.201047120418848,
      "grad_norm": 25.072650909423828,
      "kl": 6.296875,
      "learning_rate": 1.9863731656184486e-07,
      "loss": 0.4724,
      "reward": 0.5689922571182251,
      "reward_std": 0.17243780940771103,
      "reward_std_all": 0.2076769843697548,
      "rewards-max/forensics_sinkhorn_reward": 0.9881722629070282,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9967752695083618,
      "rewards/forensics_sinkhorn_reward": 0.4948495328426361,
      "rewards/forensics_strict_edge_reward": 0.3491814136505127,
      "rewards/forensics_tiou_reward": 0.4030865877866745,
      "step": 1530
    },
    {
      "completion_length": 28.1875,
      "epoch": 3.2031413612565447,
      "grad_norm": 34.23932647705078,
      "kl": 5.640625,
      "learning_rate": 1.9811320754716982e-07,
      "loss": 0.3244,
      "reward": 0.5487059652805328,
      "reward_std": 0.2447764277458191,
      "reward_std_all": 0.2500475198030472,
      "rewards-max/forensics_sinkhorn_reward": 0.9167782366275787,
      "rewards-max/forensics_strict_edge_reward": 0.78504678606987,
      "rewards-max/forensics_tiou_reward": 0.8423675000667572,
      "rewards/forensics_sinkhorn_reward": 0.41810595989227295,
      "rewards/forensics_strict_edge_reward": 0.21867765113711357,
      "rewards/forensics_tiou_reward": 0.24674896895885468,
      "step": 1531
    },
    {
      "completion_length": 35.203125,
      "epoch": 3.2052356020942407,
      "grad_norm": 18.064416885375977,
      "kl": 4.84375,
      "learning_rate": 1.9758909853249475e-07,
      "loss": 0.3067,
      "reward": 0.5630946159362793,
      "reward_std": 0.23392202705144882,
      "reward_std_all": 0.24161836504936218,
      "rewards-max/forensics_sinkhorn_reward": 0.8085603713989258,
      "rewards-max/forensics_strict_edge_reward": 0.7872911989688873,
      "rewards-max/forensics_tiou_reward": 0.7735947370529175,
      "rewards/forensics_sinkhorn_reward": 0.4353116750717163,
      "rewards/forensics_strict_edge_reward": 0.2392309159040451,
      "rewards/forensics_tiou_reward": 0.2524132877588272,
      "step": 1532
    },
    {
      "completion_length": 35.75,
      "epoch": 3.2073298429319372,
      "grad_norm": 25.041704177856445,
      "kl": 4.7578125,
      "learning_rate": 1.9706498951781968e-07,
      "loss": 0.2717,
      "reward": 0.5646469295024872,
      "reward_std": 0.22779588401317596,
      "reward_std_all": 0.23088563978672028,
      "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": 0.5072959661483765,
      "rewards/forensics_strict_edge_reward": 0.341725155711174,
      "rewards/forensics_tiou_reward": 0.3602353036403656,
      "step": 1533
    },
    {
      "completion_length": 54.796875,
      "epoch": 3.2094240837696333,
      "grad_norm": 28.751659393310547,
      "kl": 7.609375,
      "learning_rate": 1.9654088050314467e-07,
      "loss": 0.5672,
      "reward": 0.5250496864318848,
      "reward_std": 0.20197435468435287,
      "reward_std_all": 0.24548477679491043,
      "rewards-max/forensics_sinkhorn_reward": 0.9887536466121674,
      "rewards-max/forensics_strict_edge_reward": 0.9782643616199493,
      "rewards-max/forensics_tiou_reward": 0.9990667402744293,
      "rewards/forensics_sinkhorn_reward": 0.43765583634376526,
      "rewards/forensics_strict_edge_reward": 0.28155723214149475,
      "rewards/forensics_tiou_reward": 0.29411499202251434,
      "step": 1534
    },
    {
      "completion_length": 44.09375,
      "epoch": 3.2115183246073298,
      "grad_norm": 20.949371337890625,
      "kl": 4.875,
      "learning_rate": 1.960167714884696e-07,
      "loss": 0.2855,
      "reward": 0.5617708265781403,
      "reward_std": 0.17960556596517563,
      "reward_std_all": 0.2063591629266739,
      "rewards-max/forensics_sinkhorn_reward": 0.9862017333507538,
      "rewards-max/forensics_strict_edge_reward": 0.927969753742218,
      "rewards-max/forensics_tiou_reward": 0.9890713691711426,
      "rewards/forensics_sinkhorn_reward": 0.5062582194805145,
      "rewards/forensics_strict_edge_reward": 0.24940352141857147,
      "rewards/forensics_tiou_reward": 0.32961368560791016,
      "step": 1535
    },
    {
      "completion_length": 24.03125,
      "epoch": 3.2136125654450263,
      "grad_norm": 17.369464874267578,
      "kl": 5.671875,
      "learning_rate": 1.9549266247379453e-07,
      "loss": 0.5316,
      "reward": 0.5402369499206543,
      "reward_std": 0.26455481350421906,
      "reward_std_all": 0.2675248384475708,
      "rewards-max/forensics_sinkhorn_reward": 0.9743717312812805,
      "rewards-max/forensics_strict_edge_reward": 0.9357379674911499,
      "rewards-max/forensics_tiou_reward": 0.9657612144947052,
      "rewards/forensics_sinkhorn_reward": 0.4513622671365738,
      "rewards/forensics_strict_edge_reward": 0.2859968841075897,
      "rewards/forensics_tiou_reward": 0.3285854160785675,
      "step": 1536
    },
    {
      "completion_length": 43.484375,
      "epoch": 3.2157068062827223,
      "grad_norm": 26.264373779296875,
      "kl": 6.375,
      "learning_rate": 1.949685534591195e-07,
      "loss": 0.3853,
      "reward": 0.5614231526851654,
      "reward_std": 0.20126798748970032,
      "reward_std_all": 0.22075558453798294,
      "rewards-max/forensics_sinkhorn_reward": 0.9633858501911163,
      "rewards-max/forensics_strict_edge_reward": 0.8237898349761963,
      "rewards-max/forensics_tiou_reward": 0.9431642591953278,
      "rewards/forensics_sinkhorn_reward": 0.4261692017316818,
      "rewards/forensics_strict_edge_reward": 0.2152150198817253,
      "rewards/forensics_tiou_reward": 0.2644096314907074,
      "step": 1537
    },
    {
      "completion_length": 31.109375,
      "epoch": 3.217801047120419,
      "grad_norm": 18.84313201904297,
      "kl": 6.4375,
      "learning_rate": 1.9444444444444445e-07,
      "loss": 0.4132,
      "reward": 0.5230812281370163,
      "reward_std": 0.1382135972380638,
      "reward_std_all": 0.1995764970779419,
      "rewards-max/forensics_sinkhorn_reward": 0.9869057238101959,
      "rewards-max/forensics_strict_edge_reward": 0.927969753742218,
      "rewards-max/forensics_tiou_reward": 0.9918178021907806,
      "rewards/forensics_sinkhorn_reward": 0.4623826742172241,
      "rewards/forensics_strict_edge_reward": 0.2961907312273979,
      "rewards/forensics_tiou_reward": 0.3674709349870682,
      "step": 1538
    },
    {
      "completion_length": 53.75,
      "epoch": 3.2198952879581153,
      "grad_norm": 21.832624435424805,
      "kl": 6.9375,
      "learning_rate": 1.9392033542976938e-07,
      "loss": 0.5878,
      "reward": 0.5246109664440155,
      "reward_std": 0.26190702617168427,
      "reward_std_all": 0.272076852619648,
      "rewards-max/forensics_sinkhorn_reward": 0.9871945083141327,
      "rewards-max/forensics_strict_edge_reward": 0.9048932790756226,
      "rewards-max/forensics_tiou_reward": 0.9929206371307373,
      "rewards/forensics_sinkhorn_reward": 0.46399979293346405,
      "rewards/forensics_strict_edge_reward": 0.31758204102516174,
      "rewards/forensics_tiou_reward": 0.3546822518110275,
      "step": 1539
    },
    {
      "completion_length": 32.0,
      "epoch": 3.2219895287958114,
      "grad_norm": 31.385513305664062,
      "kl": 7.21875,
      "learning_rate": 1.9339622641509434e-07,
      "loss": 0.5511,
      "reward": 0.57545405626297,
      "reward_std": 0.2306927964091301,
      "reward_std_all": 0.23723971843719482,
      "rewards-max/forensics_sinkhorn_reward": 0.9838809370994568,
      "rewards-max/forensics_strict_edge_reward": 0.8719756007194519,
      "rewards-max/forensics_tiou_reward": 0.980047345161438,
      "rewards/forensics_sinkhorn_reward": 0.5657529234886169,
      "rewards/forensics_strict_edge_reward": 0.29532116651535034,
      "rewards/forensics_tiou_reward": 0.4483341872692108,
      "step": 1540
    },
    {
      "completion_length": 32.515625,
      "epoch": 3.224083769633508,
      "grad_norm": 20.104904174804688,
      "kl": 7.1875,
      "learning_rate": 1.9287211740041927e-07,
      "loss": 0.4737,
      "reward": 0.5210121870040894,
      "reward_std": 0.19961363077163696,
      "reward_std_all": 0.24423901736736298,
      "rewards-max/forensics_sinkhorn_reward": 0.9796169102191925,
      "rewards-max/forensics_strict_edge_reward": 0.9256840348243713,
      "rewards-max/forensics_tiou_reward": 0.9757668077945709,
      "rewards/forensics_sinkhorn_reward": 0.43924081325531006,
      "rewards/forensics_strict_edge_reward": 0.22345112264156342,
      "rewards/forensics_tiou_reward": 0.3121207058429718,
      "step": 1541
    },
    {
      "completion_length": 28.6875,
      "epoch": 3.2261780104712043,
      "grad_norm": 23.979896545410156,
      "kl": 7.484375,
      "learning_rate": 1.9234800838574423e-07,
      "loss": 0.4799,
      "reward": 0.5640218257904053,
      "reward_std": 0.1706867441534996,
      "reward_std_all": 0.1977902352809906,
      "rewards-max/forensics_sinkhorn_reward": 0.9116638600826263,
      "rewards-max/forensics_strict_edge_reward": 0.8351600170135498,
      "rewards-max/forensics_tiou_reward": 0.8351527750492096,
      "rewards/forensics_sinkhorn_reward": 0.48754236102104187,
      "rewards/forensics_strict_edge_reward": 0.28837574273347855,
      "rewards/forensics_tiou_reward": 0.3288329690694809,
      "step": 1542
    },
    {
      "completion_length": 20.875,
      "epoch": 3.2282722513089004,
      "grad_norm": 21.103771209716797,
      "kl": 3.96875,
      "learning_rate": 1.918238993710692e-07,
      "loss": 0.2058,
      "reward": 0.575791984796524,
      "reward_std": 0.14923598617315292,
      "reward_std_all": 0.18142130970954895,
      "rewards-max/forensics_sinkhorn_reward": 0.9871518313884735,
      "rewards-max/forensics_strict_edge_reward": 0.947419673204422,
      "rewards-max/forensics_tiou_reward": 0.9927798807621002,
      "rewards/forensics_sinkhorn_reward": 0.5125586092472076,
      "rewards/forensics_strict_edge_reward": 0.3069506883621216,
      "rewards/forensics_tiou_reward": 0.3416650742292404,
      "step": 1543
    },
    {
      "completion_length": 38.5,
      "epoch": 3.230366492146597,
      "grad_norm": 21.007320404052734,
      "kl": 7.078125,
      "learning_rate": 1.9129979035639412e-07,
      "loss": 0.4509,
      "reward": 0.5062156021595001,
      "reward_std": 0.1940794587135315,
      "reward_std_all": 0.2506379932165146,
      "rewards-max/forensics_sinkhorn_reward": 0.9749866127967834,
      "rewards-max/forensics_strict_edge_reward": 0.9048932790756226,
      "rewards-max/forensics_tiou_reward": 0.9650008380413055,
      "rewards/forensics_sinkhorn_reward": 0.43020372092723846,
      "rewards/forensics_strict_edge_reward": 0.2825204133987427,
      "rewards/forensics_tiou_reward": 0.30370117723941803,
      "step": 1544
    },
    {
      "completion_length": 27.515625,
      "epoch": 3.2324607329842934,
      "grad_norm": 17.84189224243164,
      "kl": 4.2890625,
      "learning_rate": 1.9077568134171905e-07,
      "loss": 0.3761,
      "reward": 0.6219145059585571,
      "reward_std": 0.22242328524589539,
      "reward_std_all": 0.2073432058095932,
      "rewards-max/forensics_sinkhorn_reward": 0.9878075122833252,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9953464865684509,
      "rewards/forensics_sinkhorn_reward": 0.6055843234062195,
      "rewards/forensics_strict_edge_reward": 0.39318931102752686,
      "rewards/forensics_tiou_reward": 0.4941730499267578,
      "step": 1545
    },
    {
      "completion_length": 43.21875,
      "epoch": 3.2345549738219894,
      "grad_norm": 46.492923736572266,
      "kl": 9.0625,
      "learning_rate": 1.9025157232704404e-07,
      "loss": 0.6488,
      "reward": 0.4873914420604706,
      "reward_std": 0.17383316159248352,
      "reward_std_all": 0.23597650229930878,
      "rewards-max/forensics_sinkhorn_reward": 0.9483098387718201,
      "rewards-max/forensics_strict_edge_reward": 0.8668389916419983,
      "rewards-max/forensics_tiou_reward": 0.9077761769294739,
      "rewards/forensics_sinkhorn_reward": 0.3779823035001755,
      "rewards/forensics_strict_edge_reward": 0.1728176772594452,
      "rewards/forensics_tiou_reward": 0.2104821875691414,
      "step": 1546
    },
    {
      "completion_length": 28.078125,
      "epoch": 3.236649214659686,
      "grad_norm": 33.997528076171875,
      "kl": 9.0625,
      "learning_rate": 1.8972746331236897e-07,
      "loss": 0.565,
      "reward": 0.5335899591445923,
      "reward_std": 0.1870841085910797,
      "reward_std_all": 0.23749027401208878,
      "rewards-max/forensics_sinkhorn_reward": 0.9581237435340881,
      "rewards-max/forensics_strict_edge_reward": 0.8983768224716187,
      "rewards-max/forensics_tiou_reward": 0.9311740100383759,
      "rewards/forensics_sinkhorn_reward": 0.49829959869384766,
      "rewards/forensics_strict_edge_reward": 0.307290680706501,
      "rewards/forensics_tiou_reward": 0.3744118735194206,
      "step": 1547
    },
    {
      "completion_length": 36.703125,
      "epoch": 3.2387434554973824,
      "grad_norm": 22.45997428894043,
      "kl": 6.609375,
      "learning_rate": 1.892033542976939e-07,
      "loss": 0.4753,
      "reward": 0.5477606356143951,
      "reward_std": 0.21125590801239014,
      "reward_std_all": 0.2400559037923813,
      "rewards-max/forensics_sinkhorn_reward": 0.9889025986194611,
      "rewards-max/forensics_strict_edge_reward": 0.9890114367008209,
      "rewards-max/forensics_tiou_reward": 0.999653548002243,
      "rewards/forensics_sinkhorn_reward": 0.5006005018949509,
      "rewards/forensics_strict_edge_reward": 0.29819657653570175,
      "rewards/forensics_tiou_reward": 0.38885049521923065,
      "step": 1548
    },
    {
      "completion_length": 40.0,
      "epoch": 3.2408376963350785,
      "grad_norm": 26.72867774963379,
      "kl": 7.640625,
      "learning_rate": 1.8867924528301886e-07,
      "loss": 0.2913,
      "reward": 0.5406402945518494,
      "reward_std": 0.23021672666072845,
      "reward_std_all": 0.24763643741607666,
      "rewards-max/forensics_sinkhorn_reward": 0.9884685277938843,
      "rewards-max/forensics_strict_edge_reward": 0.9890114367008209,
      "rewards-max/forensics_tiou_reward": 0.9979445040225983,
      "rewards/forensics_sinkhorn_reward": 0.4394083470106125,
      "rewards/forensics_strict_edge_reward": 0.2577030211687088,
      "rewards/forensics_tiou_reward": 0.3078417479991913,
      "step": 1549
    },
    {
      "completion_length": 43.28125,
      "epoch": 3.242931937172775,
      "grad_norm": 41.520198822021484,
      "kl": 8.53125,
      "learning_rate": 1.8815513626834382e-07,
      "loss": 0.5596,
      "reward": 0.5993658006191254,
      "reward_std": 0.21878859400749207,
      "reward_std_all": 0.20942073315382004,
      "rewards-max/forensics_sinkhorn_reward": 0.9873595535755157,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9935785233974457,
      "rewards/forensics_sinkhorn_reward": 0.5546801686286926,
      "rewards/forensics_strict_edge_reward": 0.3600655645132065,
      "rewards/forensics_tiou_reward": 0.42959973216056824,
      "step": 1550
    },
    {
      "completion_length": 27.90625,
      "epoch": 3.245026178010471,
      "grad_norm": 15.374850273132324,
      "kl": 5.140625,
      "learning_rate": 1.8763102725366875e-07,
      "loss": 0.3699,
      "reward": 0.625706136226654,
      "reward_std": 0.1690131351351738,
      "reward_std_all": 0.17883439362049103,
      "rewards-max/forensics_sinkhorn_reward": 0.9879001080989838,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9957041442394257,
      "rewards/forensics_sinkhorn_reward": 0.612621545791626,
      "rewards/forensics_strict_edge_reward": 0.4065370708703995,
      "rewards/forensics_tiou_reward": 0.48508691787719727,
      "step": 1551
    },
    {
      "completion_length": 26.640625,
      "epoch": 3.2471204188481675,
      "grad_norm": 22.572385787963867,
      "kl": 6.53125,
      "learning_rate": 1.8710691823899368e-07,
      "loss": 0.4051,
      "reward": 0.5601266920566559,
      "reward_std": 0.19793804734945297,
      "reward_std_all": 0.214798241853714,
      "rewards-max/forensics_sinkhorn_reward": 0.9859185814857483,
      "rewards-max/forensics_strict_edge_reward": 0.9256840348243713,
      "rewards-max/forensics_tiou_reward": 0.9879278242588043,
      "rewards/forensics_sinkhorn_reward": 0.4314761757850647,
      "rewards/forensics_strict_edge_reward": 0.3029915988445282,
      "rewards/forensics_tiou_reward": 0.283611536026001,
      "step": 1552
    },
    {
      "completion_length": 38.53125,
      "epoch": 3.249214659685864,
      "grad_norm": 27.61224365234375,
      "kl": 6.546875,
      "learning_rate": 1.8658280922431864e-07,
      "loss": 0.3556,
      "reward": 0.4887446016073227,
      "reward_std": 0.21660171449184418,
      "reward_std_all": 0.2646871656179428,
      "rewards-max/forensics_sinkhorn_reward": 0.9239965081214905,
      "rewards-max/forensics_strict_edge_reward": 0.8128527700901031,
      "rewards-max/forensics_tiou_reward": 0.8548692166805267,
      "rewards/forensics_sinkhorn_reward": 0.3833720535039902,
      "rewards/forensics_strict_edge_reward": 0.1766260340809822,
      "rewards/forensics_tiou_reward": 0.22111494839191437,
      "step": 1553
    },
    {
      "completion_length": 28.625,
      "epoch": 3.25130890052356,
      "grad_norm": 18.652015686035156,
      "kl": 6.2734375,
      "learning_rate": 1.860587002096436e-07,
      "loss": 0.3972,
      "reward": 0.578880786895752,
      "reward_std": 0.24347981810569763,
      "reward_std_all": 0.2359498292207718,
      "rewards-max/forensics_sinkhorn_reward": 0.9610174894332886,
      "rewards-max/forensics_strict_edge_reward": 0.8876297473907471,
      "rewards-max/forensics_tiou_reward": 0.9350630342960358,
      "rewards/forensics_sinkhorn_reward": 0.5225570499897003,
      "rewards/forensics_strict_edge_reward": 0.3629327714443207,
      "rewards/forensics_tiou_reward": 0.39767923206090927,
      "step": 1554
    },
    {
      "completion_length": 29.421875,
      "epoch": 3.2534031413612565,
      "grad_norm": 32.3222770690918,
      "kl": 5.59375,
      "learning_rate": 1.8553459119496853e-07,
      "loss": 0.479,
      "reward": 0.5856021642684937,
      "reward_std": 0.19446133822202682,
      "reward_std_all": 0.2103719264268875,
      "rewards-max/forensics_sinkhorn_reward": 0.9646013677120209,
      "rewards-max/forensics_strict_edge_reward": 0.7842414975166321,
      "rewards-max/forensics_tiou_reward": 0.9423400163650513,
      "rewards/forensics_sinkhorn_reward": 0.4890389144420624,
      "rewards/forensics_strict_edge_reward": 0.20690539479255676,
      "rewards/forensics_tiou_reward": 0.28958458453416824,
      "step": 1555
    },
    {
      "completion_length": 28.453125,
      "epoch": 3.255497382198953,
      "grad_norm": 59.95005416870117,
      "kl": 6.09375,
      "learning_rate": 1.850104821802935e-07,
      "loss": 0.4834,
      "reward": 0.5748514533042908,
      "reward_std": 0.2926225960254669,
      "reward_std_all": 0.27149729430675507,
      "rewards-max/forensics_sinkhorn_reward": 0.9885820150375366,
      "rewards-max/forensics_strict_edge_reward": 0.9782643616199493,
      "rewards-max/forensics_tiou_reward": 0.9983910024166107,
      "rewards/forensics_sinkhorn_reward": 0.5101998448371887,
      "rewards/forensics_strict_edge_reward": 0.3096526712179184,
      "rewards/forensics_tiou_reward": 0.3875579386949539,
      "step": 1556
    },
    {
      "completion_length": 22.640625,
      "epoch": 3.257591623036649,
      "grad_norm": 26.605754852294922,
      "kl": 3.6171875,
      "learning_rate": 1.8448637316561842e-07,
      "loss": 0.3506,
      "reward": 0.5544118285179138,
      "reward_std": 0.23660030961036682,
      "reward_std_all": 0.24207425862550735,
      "rewards-max/forensics_sinkhorn_reward": 0.988304853439331,
      "rewards-max/forensics_strict_edge_reward": 0.9782643616199493,
      "rewards-max/forensics_tiou_reward": 0.9972972869873047,
      "rewards/forensics_sinkhorn_reward": 0.44679516553878784,
      "rewards/forensics_strict_edge_reward": 0.2563192546367645,
      "rewards/forensics_tiou_reward": 0.30103790014982224,
      "step": 1557
    },
    {
      "completion_length": 32.640625,
      "epoch": 3.2596858638743456,
      "grad_norm": 18.532012939453125,
      "kl": 5.828125,
      "learning_rate": 1.8396226415094338e-07,
      "loss": 0.3569,
      "reward": 0.5665716528892517,
      "reward_std": 0.1534014716744423,
      "reward_std_all": 0.1932455375790596,
      "rewards-max/forensics_sinkhorn_reward": 0.988575279712677,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9983632266521454,
      "rewards/forensics_sinkhorn_reward": 0.4679015874862671,
      "rewards/forensics_strict_edge_reward": 0.2873358577489853,
      "rewards/forensics_tiou_reward": 0.2972307503223419,
      "step": 1558
    },
    {
      "completion_length": 22.5625,
      "epoch": 3.261780104712042,
      "grad_norm": 21.93145179748535,
      "kl": 4.0703125,
      "learning_rate": 1.8343815513626834e-07,
      "loss": 0.1586,
      "reward": 0.6331246495246887,
      "reward_std": 0.12327741831541061,
      "reward_std_all": 0.14954353123903275,
      "rewards-max/forensics_sinkhorn_reward": 0.9869315028190613,
      "rewards-max/forensics_strict_edge_reward": 0.9357379674911499,
      "rewards-max/forensics_tiou_reward": 0.9919041395187378,
      "rewards/forensics_sinkhorn_reward": 0.5318267643451691,
      "rewards/forensics_strict_edge_reward": 0.2650764286518097,
      "rewards/forensics_tiou_reward": 0.3362131416797638,
      "step": 1559
    },
    {
      "completion_length": 25.765625,
      "epoch": 3.263874345549738,
      "grad_norm": 23.434768676757812,
      "kl": 5.125,
      "learning_rate": 1.8291404612159327e-07,
      "loss": 0.3544,
      "reward": 0.5657823383808136,
      "reward_std": 0.18261145800352097,
      "reward_std_all": 0.20606816560029984,
      "rewards-max/forensics_sinkhorn_reward": 0.9871780872344971,
      "rewards-max/forensics_strict_edge_reward": 0.9375866651535034,
      "rewards-max/forensics_tiou_reward": 0.9928571283817291,
      "rewards/forensics_sinkhorn_reward": 0.48989368975162506,
      "rewards/forensics_strict_edge_reward": 0.2932280898094177,
      "rewards/forensics_tiou_reward": 0.34326253831386566,
      "step": 1560
    },
    {
      "completion_length": 23.640625,
      "epoch": 3.2659685863874346,
      "grad_norm": 27.797943115234375,
      "kl": 7.0,
      "learning_rate": 1.823899371069182e-07,
      "loss": 0.3904,
      "reward": 0.5917228162288666,
      "reward_std": 0.23308885097503662,
      "reward_std_all": 0.2350602075457573,
      "rewards-max/forensics_sinkhorn_reward": 0.9752461314201355,
      "rewards-max/forensics_strict_edge_reward": 0.9579255282878876,
      "rewards-max/forensics_tiou_reward": 0.9474068582057953,
      "rewards/forensics_sinkhorn_reward": 0.5527692884206772,
      "rewards/forensics_strict_edge_reward": 0.3627658784389496,
      "rewards/forensics_tiou_reward": 0.41986918449401855,
      "step": 1561
    },
    {
      "completion_length": 22.125,
      "epoch": 3.268062827225131,
      "grad_norm": 20.794645309448242,
      "kl": 5.296875,
      "learning_rate": 1.818658280922432e-07,
      "loss": 0.4198,
      "reward": 0.6225533783435822,
      "reward_std": 0.16534190624952316,
      "reward_std_all": 0.17764300107955933,
      "rewards-max/forensics_sinkhorn_reward": 0.988395482301712,
      "rewards-max/forensics_strict_edge_reward": 0.9464850425720215,
      "rewards-max/forensics_tiou_reward": 0.997654527425766,
      "rewards/forensics_sinkhorn_reward": 0.5351910293102264,
      "rewards/forensics_strict_edge_reward": 0.38872160017490387,
      "rewards/forensics_tiou_reward": 0.4196845293045044,
      "step": 1562
    },
    {
      "completion_length": 23.34375,
      "epoch": 3.270157068062827,
      "grad_norm": 25.74304962158203,
      "kl": 3.7109375,
      "learning_rate": 1.8134171907756812e-07,
      "loss": 0.242,
      "reward": 0.6470121741294861,
      "reward_std": 0.22048766911029816,
      "reward_std_all": 0.20229148119688034,
      "rewards-max/forensics_sinkhorn_reward": 0.9885916411876678,
      "rewards-max/forensics_strict_edge_reward": 0.9890114367008209,
      "rewards-max/forensics_tiou_reward": 0.9984289109706879,
      "rewards/forensics_sinkhorn_reward": 0.5663200318813324,
      "rewards/forensics_strict_edge_reward": 0.3958672881126404,
      "rewards/forensics_tiou_reward": 0.4177950918674469,
      "step": 1563
    },
    {
      "completion_length": 40.46875,
      "epoch": 3.2722513089005236,
      "grad_norm": 23.53306007385254,
      "kl": 5.640625,
      "learning_rate": 1.8081761006289306e-07,
      "loss": 0.5477,
      "reward": 0.5252721011638641,
      "reward_std": 0.24047747999429703,
      "reward_std_all": 0.258327916264534,
      "rewards-max/forensics_sinkhorn_reward": 0.9880844354629517,
      "rewards-max/forensics_strict_edge_reward": 0.9782643616199493,
      "rewards-max/forensics_tiou_reward": 0.9964285790920258,
      "rewards/forensics_sinkhorn_reward": 0.45586369931697845,
      "rewards/forensics_strict_edge_reward": 0.30485785007476807,
      "rewards/forensics_tiou_reward": 0.3210619240999222,
      "step": 1564
    },
    {
      "completion_length": 36.40625,
      "epoch": 3.2743455497382197,
      "grad_norm": 25.180429458618164,
      "kl": 4.9375,
      "learning_rate": 1.8029350104821801e-07,
      "loss": 0.3409,
      "reward": 0.5732241272926331,
      "reward_std": 0.17910508811473846,
      "reward_std_all": 0.20494069159030914,
      "rewards-max/forensics_sinkhorn_reward": 0.9325568974018097,
      "rewards-max/forensics_strict_edge_reward": 0.8715300559997559,
      "rewards-max/forensics_tiou_reward": 0.8725190460681915,
      "rewards/forensics_sinkhorn_reward": 0.44838380813598633,
      "rewards/forensics_strict_edge_reward": 0.23260696977376938,
      "rewards/forensics_tiou_reward": 0.2861662656068802,
      "step": 1565
    },
    {
      "completion_length": 35.125,
      "epoch": 3.276439790575916,
      "grad_norm": 27.753278732299805,
      "kl": 4.953125,
      "learning_rate": 1.7976939203354297e-07,
      "loss": 0.3804,
      "reward": 0.5953508913516998,
      "reward_std": 0.2470349222421646,
      "reward_std_all": 0.2359037697315216,
      "rewards-max/forensics_sinkhorn_reward": 0.9454612135887146,
      "rewards-max/forensics_strict_edge_reward": 0.7466395199298859,
      "rewards-max/forensics_tiou_reward": 0.9067018330097198,
      "rewards/forensics_sinkhorn_reward": 0.5513452589511871,
      "rewards/forensics_strict_edge_reward": 0.3038012459874153,
      "rewards/forensics_tiou_reward": 0.3965727388858795,
      "step": 1566
    },
    {
      "completion_length": 51.34375,
      "epoch": 3.2785340314136127,
      "grad_norm": 36.129791259765625,
      "kl": 7.15625,
      "learning_rate": 1.792452830188679e-07,
      "loss": 0.4053,
      "reward": 0.49590954184532166,
      "reward_std": 0.2634517475962639,
      "reward_std_all": 0.29251620173454285,
      "rewards-max/forensics_sinkhorn_reward": 0.9451876580715179,
      "rewards-max/forensics_strict_edge_reward": 0.716708242893219,
      "rewards-max/forensics_tiou_reward": 0.8981533646583557,
      "rewards/forensics_sinkhorn_reward": 0.3840400278568268,
      "rewards/forensics_strict_edge_reward": 0.16630207002162933,
      "rewards/forensics_tiou_reward": 0.24254315346479416,
      "step": 1567
    },
    {
      "completion_length": 51.734375,
      "epoch": 3.2806282722513087,
      "grad_norm": 23.32465171813965,
      "kl": 11.03125,
      "learning_rate": 1.7872117400419286e-07,
      "loss": 0.7532,
      "reward": 0.4035985767841339,
      "reward_std": 0.15802600234746933,
      "reward_std_all": 0.2726680636405945,
      "rewards-max/forensics_sinkhorn_reward": 0.95685675740242,
      "rewards-max/forensics_strict_edge_reward": 0.8553276658058167,
      "rewards-max/forensics_tiou_reward": 0.9258165061473846,
      "rewards/forensics_sinkhorn_reward": 0.3549364507198334,
      "rewards/forensics_strict_edge_reward": 0.2062424197793007,
      "rewards/forensics_tiou_reward": 0.2765219807624817,
      "step": 1568
    },
    {
      "completion_length": 41.359375,
      "epoch": 3.282722513089005,
      "grad_norm": 33.41118621826172,
      "kl": 10.15625,
      "learning_rate": 1.781970649895178e-07,
      "loss": 0.6191,
      "reward": 0.39979781210422516,
      "reward_std": 0.1861361414194107,
      "reward_std_all": 0.28249233961105347,
      "rewards-max/forensics_sinkhorn_reward": 0.959164023399353,
      "rewards-max/forensics_strict_edge_reward": 0.9075756669044495,
      "rewards-max/forensics_tiou_reward": 0.9347774088382721,
      "rewards/forensics_sinkhorn_reward": 0.3039228469133377,
      "rewards/forensics_strict_edge_reward": 0.15239306539297104,
      "rewards/forensics_tiou_reward": 0.17146873474121094,
      "step": 1569
    },
    {
      "completion_length": 51.0625,
      "epoch": 3.2848167539267017,
      "grad_norm": 39.559757232666016,
      "kl": 9.53125,
      "learning_rate": 1.7767295597484276e-07,
      "loss": 0.5732,
      "reward": 0.41999387741088867,
      "reward_std": 0.21450930833816528,
      "reward_std_all": 0.29621192812919617,
      "rewards-max/forensics_sinkhorn_reward": 0.9869451820850372,
      "rewards-max/forensics_strict_edge_reward": 0.9565287232398987,
      "rewards-max/forensics_tiou_reward": 0.9919410347938538,
      "rewards/forensics_sinkhorn_reward": 0.3592711240053177,
      "rewards/forensics_strict_edge_reward": 0.23203401267528534,
      "rewards/forensics_tiou_reward": 0.2717612162232399,
      "step": 1570
    },
    {
      "completion_length": 55.03125,
      "epoch": 3.2869109947643977,
      "grad_norm": 30.607742309570312,
      "kl": 13.5,
      "learning_rate": 1.7714884696016771e-07,
      "loss": 1.0189,
      "reward": 0.3711450397968292,
      "reward_std": 0.2740514576435089,
      "reward_std_all": 0.340093731880188,
      "rewards-max/forensics_sinkhorn_reward": 0.981555163860321,
      "rewards-max/forensics_strict_edge_reward": 0.9256840348243713,
      "rewards-max/forensics_tiou_reward": 0.9781899452209473,
      "rewards/forensics_sinkhorn_reward": 0.29588914662599564,
      "rewards/forensics_strict_edge_reward": 0.1832607388496399,
      "rewards/forensics_tiou_reward": 0.21580462157726288,
      "step": 1571
    },
    {
      "completion_length": 36.953125,
      "epoch": 3.2890052356020942,
      "grad_norm": 36.35264587402344,
      "kl": 7.6875,
      "learning_rate": 1.7662473794549265e-07,
      "loss": 0.5218,
      "reward": 0.5445657670497894,
      "reward_std": 0.2844715118408203,
      "reward_std_all": 0.2742665559053421,
      "rewards-max/forensics_sinkhorn_reward": 0.9878616034984589,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9955560266971588,
      "rewards/forensics_sinkhorn_reward": 0.5134782493114471,
      "rewards/forensics_strict_edge_reward": 0.3446999341249466,
      "rewards/forensics_tiou_reward": 0.39497481286525726,
      "step": 1572
    },
    {
      "completion_length": 34.0,
      "epoch": 3.2910994764397907,
      "grad_norm": 24.66290855407715,
      "kl": 6.40625,
      "learning_rate": 1.7610062893081758e-07,
      "loss": 0.4866,
      "reward": 0.4706111252307892,
      "reward_std": 0.292984277009964,
      "reward_std_all": 0.31654173135757446,
      "rewards-max/forensics_sinkhorn_reward": 0.936944842338562,
      "rewards-max/forensics_strict_edge_reward": 0.8850063383579254,
      "rewards-max/forensics_tiou_reward": 0.891284167766571,
      "rewards/forensics_sinkhorn_reward": 0.4092153310775757,
      "rewards/forensics_strict_edge_reward": 0.24690186977386475,
      "rewards/forensics_tiou_reward": 0.26494861394166946,
      "step": 1573
    },
    {
      "completion_length": 34.484375,
      "epoch": 3.2931937172774868,
      "grad_norm": 25.841440200805664,
      "kl": 4.78125,
      "learning_rate": 1.7557651991614256e-07,
      "loss": 0.3335,
      "reward": 0.5630410611629486,
      "reward_std": 0.14754191413521767,
      "reward_std_all": 0.19617784023284912,
      "rewards-max/forensics_sinkhorn_reward": 0.9767186641693115,
      "rewards-max/forensics_strict_edge_reward": 0.927969753742218,
      "rewards-max/forensics_tiou_reward": 0.973129153251648,
      "rewards/forensics_sinkhorn_reward": 0.4860318899154663,
      "rewards/forensics_strict_edge_reward": 0.2060006558895111,
      "rewards/forensics_tiou_reward": 0.3057192265987396,
      "step": 1574
    },
    {
      "completion_length": 38.640625,
      "epoch": 3.2952879581151833,
      "grad_norm": 12.55605411529541,
      "kl": 5.390625,
      "learning_rate": 1.750524109014675e-07,
      "loss": 0.4977,
      "reward": 0.5875160098075867,
      "reward_std": 0.1711699441075325,
      "reward_std_all": 0.19324615597724915,
      "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": 0.5953919291496277,
      "rewards/forensics_strict_edge_reward": 0.3930102288722992,
      "rewards/forensics_tiou_reward": 0.5245430767536163,
      "step": 1575
    },
    {
      "completion_length": 34.46875,
      "epoch": 3.2973821989528798,
      "grad_norm": 22.566455841064453,
      "kl": 6.578125,
      "learning_rate": 1.7452830188679243e-07,
      "loss": 0.3984,
      "reward": 0.5337995290756226,
      "reward_std": 0.23509332537651062,
      "reward_std_all": 0.25291866809129715,
      "rewards-max/forensics_sinkhorn_reward": 0.9645700454711914,
      "rewards-max/forensics_strict_edge_reward": 0.9357379674911499,
      "rewards-max/forensics_tiou_reward": 0.9455301463603973,
      "rewards/forensics_sinkhorn_reward": 0.4017358720302582,
      "rewards/forensics_strict_edge_reward": 0.21703271567821503,
      "rewards/forensics_tiou_reward": 0.28144344687461853,
      "step": 1576
    },
    {
      "completion_length": 34.9375,
      "epoch": 3.299476439790576,
      "grad_norm": 32.537662506103516,
      "kl": 8.71875,
      "learning_rate": 1.740041928721174e-07,
      "loss": 0.4629,
      "reward": 0.5818411409854889,
      "reward_std": 0.22072894871234894,
      "reward_std_all": 0.21711304783821106,
      "rewards-max/forensics_sinkhorn_reward": 0.987893134355545,
      "rewards-max/forensics_strict_edge_reward": 0.9567649364471436,
      "rewards-max/forensics_tiou_reward": 0.9956807792186737,
      "rewards/forensics_sinkhorn_reward": 0.5192793756723404,
      "rewards/forensics_strict_edge_reward": 0.31496914476156235,
      "rewards/forensics_tiou_reward": 0.3667229562997818,
      "step": 1577
    },
    {
      "completion_length": 44.328125,
      "epoch": 3.3015706806282723,
      "grad_norm": 26.286441802978516,
      "kl": 4.65625,
      "learning_rate": 1.7348008385744235e-07,
      "loss": 0.3707,
      "reward": 0.5378932803869247,
      "reward_std": 0.23083511739969254,
      "reward_std_all": 0.25103025138378143,
      "rewards-max/forensics_sinkhorn_reward": 0.9875659644603729,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9943926930427551,
      "rewards/forensics_sinkhorn_reward": 0.4914616346359253,
      "rewards/forensics_strict_edge_reward": 0.30170175433158875,
      "rewards/forensics_tiou_reward": 0.36755791306495667,
      "step": 1578
    },
    {
      "completion_length": 40.34375,
      "epoch": 3.303664921465969,
      "grad_norm": 27.561767578125,
      "kl": 6.796875,
      "learning_rate": 1.7295597484276728e-07,
      "loss": 0.5424,
      "reward": 0.585158109664917,
      "reward_std": 0.1809106022119522,
      "reward_std_all": 0.20382564514875412,
      "rewards-max/forensics_sinkhorn_reward": 0.987930417060852,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9958254992961884,
      "rewards/forensics_sinkhorn_reward": 0.496036633849144,
      "rewards/forensics_strict_edge_reward": 0.315690815448761,
      "rewards/forensics_tiou_reward": 0.3541591316461563,
      "step": 1579
    },
    {
      "completion_length": 39.453125,
      "epoch": 3.305759162303665,
      "grad_norm": 18.71931266784668,
      "kl": 7.0625,
      "learning_rate": 1.7243186582809224e-07,
      "loss": 0.486,
      "reward": 0.5666094422340393,
      "reward_std": 0.23147320747375488,
      "reward_std_all": 0.24386468529701233,
      "rewards-max/forensics_sinkhorn_reward": 0.9882758855819702,
      "rewards-max/forensics_strict_edge_reward": 0.9782643616199493,
      "rewards-max/forensics_tiou_reward": 0.997183084487915,
      "rewards/forensics_sinkhorn_reward": 0.47326862812042236,
      "rewards/forensics_strict_edge_reward": 0.3095383942127228,
      "rewards/forensics_tiou_reward": 0.34444208443164825,
      "step": 1580
    },
    {
      "completion_length": 23.078125,
      "epoch": 3.3078534031413613,
      "grad_norm": 46.49346160888672,
      "kl": 5.671875,
      "learning_rate": 1.7190775681341717e-07,
      "loss": 0.4209,
      "reward": 0.6059135794639587,
      "reward_std": 0.1245818305760622,
      "reward_std_all": 0.15762240439653397,
      "rewards-max/forensics_sinkhorn_reward": 0.9880331456661224,
      "rewards-max/forensics_strict_edge_reward": 0.9782643616199493,
      "rewards-max/forensics_tiou_reward": 0.99622642993927,
      "rewards/forensics_sinkhorn_reward": 0.5200110971927643,
      "rewards/forensics_strict_edge_reward": 0.31962059438228607,
      "rewards/forensics_tiou_reward": 0.3614516854286194,
      "step": 1581
    },
    {
      "completion_length": 29.453125,
      "epoch": 3.309947643979058,
      "grad_norm": 24.673608779907227,
      "kl": 5.609375,
      "learning_rate": 1.7138364779874213e-07,
      "loss": 0.4357,
      "reward": 0.5366454422473907,
      "reward_std": 0.24889567494392395,
      "reward_std_all": 0.2572287768125534,
      "rewards-max/forensics_sinkhorn_reward": 0.9575698971748352,
      "rewards-max/forensics_strict_edge_reward": 0.8553276658058167,
      "rewards-max/forensics_tiou_reward": 0.9287335276603699,
      "rewards/forensics_sinkhorn_reward": 0.4628842920064926,
      "rewards/forensics_strict_edge_reward": 0.15224188566207886,
      "rewards/forensics_tiou_reward": 0.24145226925611496,
      "step": 1582
    },
    {
      "completion_length": 30.96875,
      "epoch": 3.312041884816754,
      "grad_norm": 17.89468002319336,
      "kl": 4.28125,
      "learning_rate": 1.7085953878406709e-07,
      "loss": 0.3356,
      "reward": 0.6037164628505707,
      "reward_std": 0.19035394489765167,
      "reward_std_all": 0.19873324781656265,
      "rewards-max/forensics_sinkhorn_reward": 0.9882923364639282,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9972497820854187,
      "rewards/forensics_sinkhorn_reward": 0.5386136472225189,
      "rewards/forensics_strict_edge_reward": 0.2905276268720627,
      "rewards/forensics_tiou_reward": 0.38289743661880493,
      "step": 1583
    },
    {
      "completion_length": 25.34375,
      "epoch": 3.3141361256544504,
      "grad_norm": 24.960622787475586,
      "kl": 5.578125,
      "learning_rate": 1.7033542976939202e-07,
      "loss": 0.3213,
      "reward": 0.5743042826652527,
      "reward_std": 0.17697104811668396,
      "reward_std_all": 0.20617395639419556,
      "rewards-max/forensics_sinkhorn_reward": 0.9876264333724976,
      "rewards-max/forensics_strict_edge_reward": 0.9890114367008209,
      "rewards-max/forensics_tiou_reward": 0.996780514717102,
      "rewards/forensics_sinkhorn_reward": 0.5021663904190063,
      "rewards/forensics_strict_edge_reward": 0.2862872779369354,
      "rewards/forensics_tiou_reward": 0.3603716790676117,
      "step": 1584
    },
    {
      "completion_length": 37.65625,
      "epoch": 3.3162303664921464,
      "grad_norm": 23.71660804748535,
      "kl": 7.90625,
      "learning_rate": 1.6981132075471695e-07,
      "loss": 0.4847,
      "reward": 0.49664488434791565,
      "reward_std": 0.1984189823269844,
      "reward_std_all": 0.25035329163074493,
      "rewards-max/forensics_sinkhorn_reward": 0.9608278870582581,
      "rewards-max/forensics_strict_edge_reward": 0.820955753326416,
      "rewards-max/forensics_tiou_reward": 0.9337621927261353,
      "rewards/forensics_sinkhorn_reward": 0.3791579604148865,
      "rewards/forensics_strict_edge_reward": 0.21379666030406952,
      "rewards/forensics_tiou_reward": 0.23289483785629272,
      "step": 1585
    },
    {
      "completion_length": 40.25,
      "epoch": 3.318324607329843,
      "grad_norm": 25.52252197265625,
      "kl": 6.78125,
      "learning_rate": 1.6928721174004194e-07,
      "loss": 0.4361,
      "reward": 0.46602652966976166,
      "reward_std": 0.26362690329551697,
      "reward_std_all": 0.2889488935470581,
      "rewards-max/forensics_sinkhorn_reward": 0.9621294140815735,
      "rewards-max/forensics_strict_edge_reward": 0.7421593070030212,
      "rewards-max/forensics_tiou_reward": 0.9400675594806671,
      "rewards/forensics_sinkhorn_reward": 0.3595859557390213,
      "rewards/forensics_strict_edge_reward": 0.13699832186102867,
      "rewards/forensics_tiou_reward": 0.18845482170581818,
      "step": 1586
    },
    {
      "completion_length": 45.6875,
      "epoch": 3.3204188481675394,
      "grad_norm": 25.761720657348633,
      "kl": 6.734375,
      "learning_rate": 1.6876310272536687e-07,
      "loss": 0.5459,
      "reward": 0.565182089805603,
      "reward_std": 0.19048543274402618,
      "reward_std_all": 0.21964029222726822,
      "rewards-max/forensics_sinkhorn_reward": 0.9885258972644806,
      "rewards-max/forensics_strict_edge_reward": 0.9782643616199493,
      "rewards-max/forensics_tiou_reward": 0.9981701672077179,
      "rewards/forensics_sinkhorn_reward": 0.49922531843185425,
      "rewards/forensics_strict_edge_reward": 0.3408816158771515,
      "rewards/forensics_tiou_reward": 0.37458673119544983,
      "step": 1587
    },
    {
      "completion_length": 23.734375,
      "epoch": 3.3225130890052355,
      "grad_norm": 33.31441879272461,
      "kl": 4.0078125,
      "learning_rate": 1.682389937106918e-07,
      "loss": 0.4269,
      "reward": 0.6417243778705597,
      "reward_std": 0.16600356996059418,
      "reward_std_all": 0.16327672451734543,
      "rewards-max/forensics_sinkhorn_reward": 0.9883950054645538,
      "rewards-max/forensics_strict_edge_reward": 0.9890114367008209,
      "rewards-max/forensics_tiou_reward": 0.9976553320884705,
      "rewards/forensics_sinkhorn_reward": 0.579342246055603,
      "rewards/forensics_strict_edge_reward": 0.31450607627630234,
      "rewards/forensics_tiou_reward": 0.36068810522556305,
      "step": 1588
    },
    {
      "completion_length": 47.765625,
      "epoch": 3.324607329842932,
      "grad_norm": 27.025144577026367,
      "kl": 5.515625,
      "learning_rate": 1.6771488469601679e-07,
      "loss": 0.3897,
      "reward": 0.5590603649616241,
      "reward_std": 0.29365894198417664,
      "reward_std_all": 0.2760763317346573,
      "rewards-max/forensics_sinkhorn_reward": 0.9089462161064148,
      "rewards-max/forensics_strict_edge_reward": 0.8049824237823486,
      "rewards-max/forensics_tiou_reward": 0.8255545496940613,
      "rewards/forensics_sinkhorn_reward": 0.4755818098783493,
      "rewards/forensics_strict_edge_reward": 0.22259709239006042,
      "rewards/forensics_tiou_reward": 0.32614878565073013,
      "step": 1589
    },
    {
      "completion_length": 46.5625,
      "epoch": 3.3267015706806284,
      "grad_norm": 16.96185874938965,
      "kl": 8.390625,
      "learning_rate": 1.6719077568134172e-07,
      "loss": 0.5453,
      "reward": 0.476061686873436,
      "reward_std": 0.19354454427957535,
      "reward_std_all": 0.26297152042388916,
      "rewards-max/forensics_sinkhorn_reward": 0.9358669221401215,
      "rewards-max/forensics_strict_edge_reward": 0.74742591381073,
      "rewards-max/forensics_tiou_reward": 0.8900120258331299,
      "rewards/forensics_sinkhorn_reward": 0.34800319373607635,
      "rewards/forensics_strict_edge_reward": 0.18562230467796326,
      "rewards/forensics_tiou_reward": 0.2080737017095089,
      "step": 1590
    },
    {
      "completion_length": 26.78125,
      "epoch": 3.3287958115183245,
      "grad_norm": 23.679697036743164,
      "kl": 4.6640625,
      "learning_rate": 1.6666666666666665e-07,
      "loss": 0.36,
      "reward": 0.5847734808921814,
      "reward_std": 0.1679912805557251,
      "reward_std_all": 0.1981581747531891,
      "rewards-max/forensics_sinkhorn_reward": 0.9838162362575531,
      "rewards-max/forensics_strict_edge_reward": 0.8288396298885345,
      "rewards-max/forensics_tiou_reward": 0.979610413312912,
      "rewards/forensics_sinkhorn_reward": 0.5054275095462799,
      "rewards/forensics_strict_edge_reward": 0.23784124851226807,
      "rewards/forensics_tiou_reward": 0.3529873192310333,
      "step": 1591
    },
    {
      "completion_length": 31.5,
      "epoch": 3.330890052356021,
      "grad_norm": 15.23149299621582,
      "kl": 5.3125,
      "learning_rate": 1.661425576519916e-07,
      "loss": 0.3301,
      "reward": 0.6049919426441193,
      "reward_std": 0.22750356048345566,
      "reward_std_all": 0.21886827796697617,
      "rewards-max/forensics_sinkhorn_reward": 0.9886867105960846,
      "rewards-max/forensics_strict_edge_reward": 0.9890114367008209,
      "rewards-max/forensics_tiou_reward": 0.9988023936748505,
      "rewards/forensics_sinkhorn_reward": 0.5700017809867859,
      "rewards/forensics_strict_edge_reward": 0.31938838958740234,
      "rewards/forensics_tiou_reward": 0.41778723895549774,
      "step": 1592
    },
    {
      "completion_length": 41.78125,
      "epoch": 3.332984293193717,
      "grad_norm": 32.32637405395508,
      "kl": 8.203125,
      "learning_rate": 1.6561844863731657e-07,
      "loss": 0.4485,
      "reward": 0.5254632085561752,
      "reward_std": 0.2598143517971039,
      "reward_std_all": 0.28117503225803375,
      "rewards-max/forensics_sinkhorn_reward": 0.988640159368515,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9986195862293243,
      "rewards/forensics_sinkhorn_reward": 0.4278118312358856,
      "rewards/forensics_strict_edge_reward": 0.29574069380760193,
      "rewards/forensics_tiou_reward": 0.32139482349157333,
      "step": 1593
    },
    {
      "completion_length": 26.28125,
      "epoch": 3.3350785340314135,
      "grad_norm": 21.854997634887695,
      "kl": 7.609375,
      "learning_rate": 1.650943396226415e-07,
      "loss": 0.4932,
      "reward": 0.5727958977222443,
      "reward_std": 0.23522904515266418,
      "reward_std_all": 0.22906047850847244,
      "rewards-max/forensics_sinkhorn_reward": 0.9748952388763428,
      "rewards-max/forensics_strict_edge_reward": 0.9151731729507446,
      "rewards-max/forensics_tiou_reward": 0.9690010249614716,
      "rewards/forensics_sinkhorn_reward": 0.49126438796520233,
      "rewards/forensics_strict_edge_reward": 0.27351053804159164,
      "rewards/forensics_tiou_reward": 0.35467788577079773,
      "step": 1594
    },
    {
      "completion_length": 28.8125,
      "epoch": 3.33717277486911,
      "grad_norm": 35.28709411621094,
      "kl": 5.8125,
      "learning_rate": 1.6457023060796646e-07,
      "loss": 0.3928,
      "reward": 0.5968367159366608,
      "reward_std": 0.22868389636278152,
      "reward_std_all": 0.22325943410396576,
      "rewards-max/forensics_sinkhorn_reward": 0.9879463613033295,
      "rewards-max/forensics_strict_edge_reward": 0.9169811904430389,
      "rewards-max/forensics_tiou_reward": 0.995884120464325,
      "rewards/forensics_sinkhorn_reward": 0.5092414021492004,
      "rewards/forensics_strict_edge_reward": 0.25615646690130234,
      "rewards/forensics_tiou_reward": 0.36720961332321167,
      "step": 1595
    },
    {
      "completion_length": 27.59375,
      "epoch": 3.339267015706806,
      "grad_norm": 25.14521026611328,
      "kl": 5.578125,
      "learning_rate": 1.640461215932914e-07,
      "loss": 0.3318,
      "reward": 0.5633371472358704,
      "reward_std": 0.1938587725162506,
      "reward_std_all": 0.2206537425518036,
      "rewards-max/forensics_sinkhorn_reward": 0.9877453148365021,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9950932562351227,
      "rewards/forensics_sinkhorn_reward": 0.43850789964199066,
      "rewards/forensics_strict_edge_reward": 0.2933477759361267,
      "rewards/forensics_tiou_reward": 0.29515276849269867,
      "step": 1596
    },
    {
      "completion_length": 32.53125,
      "epoch": 3.3413612565445026,
      "grad_norm": 21.76490020751953,
      "kl": 6.546875,
      "learning_rate": 1.6352201257861632e-07,
      "loss": 0.3823,
      "reward": 0.5552469342947006,
      "reward_std": 0.14210981875658035,
      "reward_std_all": 0.19628556072711945,
      "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": 0.492854043841362,
      "rewards/forensics_strict_edge_reward": 0.28946737200021744,
      "rewards/forensics_tiou_reward": 0.3403383791446686,
      "step": 1597
    },
    {
      "completion_length": 29.8125,
      "epoch": 3.343455497382199,
      "grad_norm": 21.675369262695312,
      "kl": 6.109375,
      "learning_rate": 1.629979035639413e-07,
      "loss": 0.465,
      "reward": 0.525920569896698,
      "reward_std": 0.24097636342048645,
      "reward_std_all": 0.26160232722759247,
      "rewards-max/forensics_sinkhorn_reward": 0.9884175956249237,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9977425038814545,
      "rewards/forensics_sinkhorn_reward": 0.45620352029800415,
      "rewards/forensics_strict_edge_reward": 0.34120510518550873,
      "rewards/forensics_tiou_reward": 0.38594500720500946,
      "step": 1598
    },
    {
      "completion_length": 20.734375,
      "epoch": 3.345549738219895,
      "grad_norm": 24.285871505737305,
      "kl": 3.9296875,
      "learning_rate": 1.6247379454926624e-07,
      "loss": 0.1696,
      "reward": 0.6875171661376953,
      "reward_std": 0.11081161350011826,
      "reward_std_all": 0.11495120450854301,
      "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": 0.6253528296947479,
      "rewards/forensics_strict_edge_reward": 0.2891254425048828,
      "rewards/forensics_tiou_reward": 0.4174778461456299,
      "step": 1599
    },
    {
      "completion_length": 36.390625,
      "epoch": 3.3476439790575916,
      "grad_norm": 19.8402042388916,
      "kl": 6.078125,
      "learning_rate": 1.6194968553459117e-07,
      "loss": 0.4261,
      "reward": 0.5608541965484619,
      "reward_std": 0.18261928111314774,
      "reward_std_all": 0.21240241825580597,
      "rewards-max/forensics_sinkhorn_reward": 0.9801670014858246,
      "rewards-max/forensics_strict_edge_reward": 0.9252271056175232,
      "rewards-max/forensics_tiou_reward": 0.9658111035823822,
      "rewards/forensics_sinkhorn_reward": 0.49073927104473114,
      "rewards/forensics_strict_edge_reward": 0.22760920226573944,
      "rewards/forensics_tiou_reward": 0.29302578419446945,
      "step": 1600
    },
    {
      "completion_length": 29.03125,
      "epoch": 3.349738219895288,
      "grad_norm": 20.096981048583984,
      "kl": 7.140625,
      "learning_rate": 1.6142557651991616e-07,
      "loss": 0.4207,
      "reward": 0.5138025283813477,
      "reward_std": 0.27659235894680023,
      "reward_std_all": 0.2886171191930771,
      "rewards-max/forensics_sinkhorn_reward": 0.9866098165512085,
      "rewards-max/forensics_strict_edge_reward": 0.946017861366272,
      "rewards-max/forensics_tiou_reward": 0.9906164705753326,
      "rewards/forensics_sinkhorn_reward": 0.42210978269577026,
      "rewards/forensics_strict_edge_reward": 0.24476386606693268,
      "rewards/forensics_tiou_reward": 0.28751347959041595,
      "step": 1601
    },
    {
      "completion_length": 35.796875,
      "epoch": 3.351832460732984,
      "grad_norm": 29.89870834350586,
      "kl": 5.21875,
      "learning_rate": 1.609014675052411e-07,
      "loss": 0.2861,
      "reward": 0.5217001736164093,
      "reward_std": 0.20415885746479034,
      "reward_std_all": 0.23981793224811554,
      "rewards-max/forensics_sinkhorn_reward": 0.9240247309207916,
      "rewards-max/forensics_strict_edge_reward": 0.8340164124965668,
      "rewards-max/forensics_tiou_reward": 0.8591342270374298,
      "rewards/forensics_sinkhorn_reward": 0.3746369183063507,
      "rewards/forensics_strict_edge_reward": 0.21152199059724808,
      "rewards/forensics_tiou_reward": 0.21224838495254517,
      "step": 1602
    },
    {
      "completion_length": 28.84375,
      "epoch": 3.3539267015706806,
      "grad_norm": 34.8635368347168,
      "kl": 6.09375,
      "learning_rate": 1.6037735849056602e-07,
      "loss": 0.4952,
      "reward": 0.594520777463913,
      "reward_std": 0.1632368564605713,
      "reward_std_all": 0.18509358912706375,
      "rewards-max/forensics_sinkhorn_reward": 0.9710339605808258,
      "rewards-max/forensics_strict_edge_reward": 0.8876297473907471,
      "rewards-max/forensics_tiou_reward": 0.9593373835086823,
      "rewards/forensics_sinkhorn_reward": 0.5210044533014297,
      "rewards/forensics_strict_edge_reward": 0.28439853340387344,
      "rewards/forensics_tiou_reward": 0.350552573800087,
      "step": 1603
    },
    {
      "completion_length": 29.0625,
      "epoch": 3.356020942408377,
      "grad_norm": 21.968122482299805,
      "kl": 4.0859375,
      "learning_rate": 1.5985324947589098e-07,
      "loss": 0.1708,
      "reward": 0.6748268306255341,
      "reward_std": 0.15240159630775452,
      "reward_std_all": 0.1464863084256649,
      "rewards-max/forensics_sinkhorn_reward": 0.9881426692008972,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9966590106487274,
      "rewards/forensics_sinkhorn_reward": 0.6097803115844727,
      "rewards/forensics_strict_edge_reward": 0.41174982488155365,
      "rewards/forensics_tiou_reward": 0.4637559652328491,
      "step": 1604
    },
    {
      "completion_length": 21.96875,
      "epoch": 3.358115183246073,
      "grad_norm": 46.87697982788086,
      "kl": 7.359375,
      "learning_rate": 1.5932914046121594e-07,
      "loss": 0.4947,
      "reward": 0.5660101175308228,
      "reward_std": 0.24721424281597137,
      "reward_std_all": 0.253144234418869,
      "rewards-max/forensics_sinkhorn_reward": 0.988594114780426,
      "rewards-max/forensics_strict_edge_reward": 0.9782643616199493,
      "rewards-max/forensics_tiou_reward": 0.9984374940395355,
      "rewards/forensics_sinkhorn_reward": 0.5441394746303558,
      "rewards/forensics_strict_edge_reward": 0.3697098046541214,
      "rewards/forensics_tiou_reward": 0.45084576308727264,
      "step": 1605
    },
    {
      "completion_length": 23.265625,
      "epoch": 3.3602094240837697,
      "grad_norm": 35.744384765625,
      "kl": 5.0859375,
      "learning_rate": 1.5880503144654087e-07,
      "loss": 0.3099,
      "reward": 0.5436549484729767,
      "reward_std": 0.3175102174282074,
      "reward_std_all": 0.30246950685977936,
      "rewards-max/forensics_sinkhorn_reward": 0.9887395799160004,
      "rewards-max/forensics_strict_edge_reward": 0.9890114367008209,
      "rewards-max/forensics_tiou_reward": 0.999011367559433,
      "rewards/forensics_sinkhorn_reward": 0.4751627445220947,
      "rewards/forensics_strict_edge_reward": 0.31587915122509,
      "rewards/forensics_tiou_reward": 0.35539481043815613,
      "step": 1606
    },
    {
      "completion_length": 23.421875,
      "epoch": 3.362303664921466,
      "grad_norm": 17.210472106933594,
      "kl": 5.09375,
      "learning_rate": 1.5828092243186583e-07,
      "loss": 0.2947,
      "reward": 0.600691556930542,
      "reward_std": 0.20924770832061768,
      "reward_std_all": 0.20480797439813614,
      "rewards-max/forensics_sinkhorn_reward": 0.9884831309318542,
      "rewards-max/forensics_strict_edge_reward": 0.9574736058712006,
      "rewards-max/forensics_tiou_reward": 0.9979999959468842,
      "rewards/forensics_sinkhorn_reward": 0.5136231780052185,
      "rewards/forensics_strict_edge_reward": 0.28583507239818573,
      "rewards/forensics_tiou_reward": 0.3904792219400406,
      "step": 1607
    },
    {
      "completion_length": 37.515625,
      "epoch": 3.364397905759162,
      "grad_norm": 26.7334041595459,
      "kl": 7.21875,
      "learning_rate": 1.5775681341719076e-07,
      "loss": 0.4906,
      "reward": 0.6110974550247192,
      "reward_std": 0.2625797614455223,
      "reward_std_all": 0.23903732001781464,
      "rewards-max/forensics_sinkhorn_reward": 0.9576717019081116,
      "rewards-max/forensics_strict_edge_reward": 0.8968285918235779,
      "rewards-max/forensics_tiou_reward": 0.9308655560016632,
      "rewards/forensics_sinkhorn_reward": 0.4925662726163864,
      "rewards/forensics_strict_edge_reward": 0.2972078323364258,
      "rewards/forensics_tiou_reward": 0.33977213501930237,
      "step": 1608
    },
    {
      "completion_length": 44.75,
      "epoch": 3.3664921465968587,
      "grad_norm": 3703.6591796875,
      "kl": 13.6875,
      "learning_rate": 1.5723270440251572e-07,
      "loss": 0.8759,
      "reward": 0.47025686502456665,
      "reward_std": 0.2930480092763901,
      "reward_std_all": 0.3198286294937134,
      "rewards-max/forensics_sinkhorn_reward": 0.9764118492603302,
      "rewards-max/forensics_strict_edge_reward": 0.8719756007194519,
      "rewards-max/forensics_tiou_reward": 0.9727193117141724,
      "rewards/forensics_sinkhorn_reward": 0.3892928510904312,
      "rewards/forensics_strict_edge_reward": 0.22590677440166473,
      "rewards/forensics_tiou_reward": 0.2644606679677963,
      "step": 1609
    },
    {
      "completion_length": 28.8125,
      "epoch": 3.368586387434555,
      "grad_norm": 21.748355865478516,
      "kl": 4.6015625,
      "learning_rate": 1.5670859538784068e-07,
      "loss": 0.3488,
      "reward": 0.5955804288387299,
      "reward_std": 0.19701384007930756,
      "reward_std_all": 0.20649732649326324,
      "rewards-max/forensics_sinkhorn_reward": 0.9880638420581818,
      "rewards-max/forensics_strict_edge_reward": 0.9567649364471436,
      "rewards-max/forensics_tiou_reward": 0.9963495433330536,
      "rewards/forensics_sinkhorn_reward": 0.5593771934509277,
      "rewards/forensics_strict_edge_reward": 0.32648760080337524,
      "rewards/forensics_tiou_reward": 0.42982588708400726,
      "step": 1610
    },
    {
      "completion_length": 25.4375,
      "epoch": 3.3706806282722512,
      "grad_norm": 23.98055076599121,
      "kl": 7.40625,
      "learning_rate": 1.561844863731656e-07,
      "loss": 0.5111,
      "reward": 0.4902091324329376,
      "reward_std": 0.22874826192855835,
      "reward_std_all": 0.2627240940928459,
      "rewards-max/forensics_sinkhorn_reward": 0.9888150691986084,
      "rewards-max/forensics_strict_edge_reward": 0.9890114367008209,
      "rewards-max/forensics_tiou_reward": 0.9993086755275726,
      "rewards/forensics_sinkhorn_reward": 0.40815992653369904,
      "rewards/forensics_strict_edge_reward": 0.2595501244068146,
      "rewards/forensics_tiou_reward": 0.29643240571022034,
      "step": 1611
    },
    {
      "completion_length": 55.25,
      "epoch": 3.3727748691099477,
      "grad_norm": 34.82034683227539,
      "kl": 9.046875,
      "learning_rate": 1.5566037735849055e-07,
      "loss": 0.6023,
      "reward": 0.5004026889801025,
      "reward_std": 0.20413347333669662,
      "reward_std_all": 0.2527617961168289,
      "rewards-max/forensics_sinkhorn_reward": 0.9833824038505554,
      "rewards-max/forensics_strict_edge_reward": 0.946017861366272,
      "rewards-max/forensics_tiou_reward": 0.9852999448776245,
      "rewards/forensics_sinkhorn_reward": 0.4249987155199051,
      "rewards/forensics_strict_edge_reward": 0.23288293182849884,
      "rewards/forensics_tiou_reward": 0.327462837100029,
      "step": 1612
    },
    {
      "completion_length": 28.1875,
      "epoch": 3.374869109947644,
      "grad_norm": 27.29722785949707,
      "kl": 6.90625,
      "learning_rate": 1.5513626834381553e-07,
      "loss": 0.2728,
      "reward": 0.544864684343338,
      "reward_std": 0.2362833023071289,
      "reward_std_all": 0.2560271769762039,
      "rewards-max/forensics_sinkhorn_reward": 0.9868785738945007,
      "rewards-max/forensics_strict_edge_reward": 0.9464850425720215,
      "rewards-max/forensics_tiou_reward": 0.991708368062973,
      "rewards/forensics_sinkhorn_reward": 0.4705178886651993,
      "rewards/forensics_strict_edge_reward": 0.289600133895874,
      "rewards/forensics_tiou_reward": 0.31796327233314514,
      "step": 1613
    },
    {
      "completion_length": 28.765625,
      "epoch": 3.3769633507853403,
      "grad_norm": 25.696651458740234,
      "kl": 6.734375,
      "learning_rate": 1.5461215932914046e-07,
      "loss": 0.3089,
      "reward": 0.6090722978115082,
      "reward_std": 0.2121383622288704,
      "reward_std_all": 0.20449219644069672,
      "rewards-max/forensics_sinkhorn_reward": 0.9884347915649414,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9978104829788208,
      "rewards/forensics_sinkhorn_reward": 0.536281943321228,
      "rewards/forensics_strict_edge_reward": 0.3230898678302765,
      "rewards/forensics_tiou_reward": 0.3786178380250931,
      "step": 1614
    },
    {
      "completion_length": 16.703125,
      "epoch": 3.3790575916230368,
      "grad_norm": 18.99317741394043,
      "kl": 4.34375,
      "learning_rate": 1.540880503144654e-07,
      "loss": 0.3145,
      "reward": 0.6362840533256531,
      "reward_std": 0.12240314483642578,
      "reward_std_all": 0.14593539386987686,
      "rewards-max/forensics_sinkhorn_reward": 0.9885229766368866,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9981578290462494,
      "rewards/forensics_sinkhorn_reward": 0.6322118043899536,
      "rewards/forensics_strict_edge_reward": 0.43549929559230804,
      "rewards/forensics_tiou_reward": 0.5339649319648743,
      "step": 1615
    },
    {
      "completion_length": 40.15625,
      "epoch": 3.381151832460733,
      "grad_norm": 28.180999755859375,
      "kl": 9.4375,
      "learning_rate": 1.5356394129979035e-07,
      "loss": 0.5584,
      "reward": 0.47021153569221497,
      "reward_std": 0.18706297874450684,
      "reward_std_all": 0.25606629252433777,
      "rewards-max/forensics_sinkhorn_reward": 0.9698804020881653,
      "rewards-max/forensics_strict_edge_reward": 0.9364311099052429,
      "rewards-max/forensics_tiou_reward": 0.9572458565235138,
      "rewards/forensics_sinkhorn_reward": 0.42964187264442444,
      "rewards/forensics_strict_edge_reward": 0.296421580016613,
      "rewards/forensics_tiou_reward": 0.3302580714225769,
      "step": 1616
    },
    {
      "completion_length": 24.3125,
      "epoch": 3.3832460732984293,
      "grad_norm": 24.592111587524414,
      "kl": 2.82421875,
      "learning_rate": 1.530398322851153e-07,
      "loss": 0.2335,
      "reward": 0.6576777398586273,
      "reward_std": 0.13061054795980453,
      "reward_std_all": 0.14207768812775612,
      "rewards-max/forensics_sinkhorn_reward": 0.9861707389354706,
      "rewards-max/forensics_strict_edge_reward": 0.947419673204422,
      "rewards-max/forensics_tiou_reward": 0.9889506101608276,
      "rewards/forensics_sinkhorn_reward": 0.5936569571495056,
      "rewards/forensics_strict_edge_reward": 0.28865495324134827,
      "rewards/forensics_tiou_reward": 0.3717462569475174,
      "step": 1617
    },
    {
      "completion_length": 18.0,
      "epoch": 3.385340314136126,
      "grad_norm": 29.81988525390625,
      "kl": 4.3984375,
      "learning_rate": 1.5251572327044024e-07,
      "loss": 0.2564,
      "reward": 0.6625377833843231,
      "reward_std": 0.14163381606340408,
      "reward_std_all": 0.14540079608559608,
      "rewards-max/forensics_sinkhorn_reward": 0.9884491562843323,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9978664517402649,
      "rewards/forensics_sinkhorn_reward": 0.586029440164566,
      "rewards/forensics_strict_edge_reward": 0.26893360912799835,
      "rewards/forensics_tiou_reward": 0.40259991586208344,
      "step": 1618
    },
    {
      "completion_length": 38.21875,
      "epoch": 3.387434554973822,
      "grad_norm": 22.20128059387207,
      "kl": 5.90625,
      "learning_rate": 1.519916142557652e-07,
      "loss": 0.3237,
      "reward": 0.6052270233631134,
      "reward_std": 0.18875937908887863,
      "reward_std_all": 0.20007344335317612,
      "rewards-max/forensics_sinkhorn_reward": 0.9768489897251129,
      "rewards-max/forensics_strict_edge_reward": 0.8320213258266449,
      "rewards-max/forensics_tiou_reward": 0.9723981022834778,
      "rewards/forensics_sinkhorn_reward": 0.49461551010608673,
      "rewards/forensics_strict_edge_reward": 0.2862103208899498,
      "rewards/forensics_tiou_reward": 0.3657887279987335,
      "step": 1619
    },
    {
      "completion_length": 30.90625,
      "epoch": 3.3895287958115183,
      "grad_norm": 31.808441162109375,
      "kl": 7.59375,
      "learning_rate": 1.5146750524109014e-07,
      "loss": 0.4423,
      "reward": 0.5368466675281525,
      "reward_std": 0.16288475692272186,
      "reward_std_all": 0.21265341341495514,
      "rewards-max/forensics_sinkhorn_reward": 0.984390914440155,
      "rewards-max/forensics_strict_edge_reward": 0.905340164899826,
      "rewards-max/forensics_tiou_reward": 0.9887841939926147,
      "rewards/forensics_sinkhorn_reward": 0.4670238345861435,
      "rewards/forensics_strict_edge_reward": 0.16659356653690338,
      "rewards/forensics_tiou_reward": 0.2691906690597534,
      "step": 1620
    },
    {
      "completion_length": 30.15625,
      "epoch": 3.391623036649215,
      "grad_norm": 29.384258270263672,
      "kl": 7.8125,
      "learning_rate": 1.509433962264151e-07,
      "loss": 0.5795,
      "reward": 0.4577786326408386,
      "reward_std": 0.12101666629314423,
      "reward_std_all": 0.23132362216711044,
      "rewards-max/forensics_sinkhorn_reward": 0.9866346418857574,
      "rewards-max/forensics_strict_edge_reward": 0.9149472117424011,
      "rewards-max/forensics_tiou_reward": 0.9907423555850983,
      "rewards/forensics_sinkhorn_reward": 0.415229007601738,
      "rewards/forensics_strict_edge_reward": 0.24876676499843597,
      "rewards/forensics_tiou_reward": 0.31234823912382126,
      "step": 1621
    },
    {
      "completion_length": 29.421875,
      "epoch": 3.393717277486911,
      "grad_norm": 18.54192543029785,
      "kl": 5.9375,
      "learning_rate": 1.5041928721174005e-07,
      "loss": 0.4141,
      "reward": 0.5099932849407196,
      "reward_std": 0.21807341277599335,
      "reward_std_all": 0.24877287447452545,
      "rewards-max/forensics_sinkhorn_reward": 0.9883071184158325,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9973088800907135,
      "rewards/forensics_sinkhorn_reward": 0.43035759031772614,
      "rewards/forensics_strict_edge_reward": 0.2665923088788986,
      "rewards/forensics_tiou_reward": 0.30452917516231537,
      "step": 1622
    },
    {
      "completion_length": 43.984375,
      "epoch": 3.3958115183246074,
      "grad_norm": 19.282482147216797,
      "kl": 6.8125,
      "learning_rate": 1.4989517819706498e-07,
      "loss": 0.4474,
      "reward": 0.5023360103368759,
      "reward_std": 0.19462237507104874,
      "reward_std_all": 0.2524920850992203,
      "rewards-max/forensics_sinkhorn_reward": 0.9856500029563904,
      "rewards-max/forensics_strict_edge_reward": 0.9169811904430389,
      "rewards-max/forensics_tiou_reward": 0.9868887662887573,
      "rewards/forensics_sinkhorn_reward": 0.41895171999931335,
      "rewards/forensics_strict_edge_reward": 0.2658129185438156,
      "rewards/forensics_tiou_reward": 0.27720458805561066,
      "step": 1623
    },
    {
      "completion_length": 20.78125,
      "epoch": 3.3979057591623034,
      "grad_norm": 30.44255256652832,
      "kl": 4.796875,
      "learning_rate": 1.4937106918238992e-07,
      "loss": 0.2243,
      "reward": 0.6166242361068726,
      "reward_std": 0.15012472867965698,
      "reward_std_all": 0.16979075968265533,
      "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": 0.6046143174171448,
      "rewards/forensics_strict_edge_reward": 0.4066558927297592,
      "rewards/forensics_tiou_reward": 0.458665207028389,
      "step": 1624
    },
    {
      "completion_length": 20.984375,
      "epoch": 3.4,
      "grad_norm": 20.486083984375,
      "kl": 6.34375,
      "learning_rate": 1.488469601677149e-07,
      "loss": 0.3335,
      "reward": 0.5786019265651703,
      "reward_std": 0.2180301770567894,
      "reward_std_all": 0.21365345269441605,
      "rewards-max/forensics_sinkhorn_reward": 0.9878159165382385,
      "rewards-max/forensics_strict_edge_reward": 0.9890114367008209,
      "rewards-max/forensics_tiou_reward": 0.995381087064743,
      "rewards/forensics_sinkhorn_reward": 0.5176241397857666,
      "rewards/forensics_strict_edge_reward": 0.2726759761571884,
      "rewards/forensics_tiou_reward": 0.32848040759563446,
      "step": 1625
    },
    {
      "completion_length": 27.421875,
      "epoch": 3.4020942408376964,
      "grad_norm": 25.410049438476562,
      "kl": 8.0,
      "learning_rate": 1.4832285115303983e-07,
      "loss": 0.6029,
      "reward": 0.5239991992712021,
      "reward_std": 0.29797790944576263,
      "reward_std_all": 0.2871355563402176,
      "rewards-max/forensics_sinkhorn_reward": 0.954609602689743,
      "rewards-max/forensics_strict_edge_reward": 0.8983768224716187,
      "rewards-max/forensics_tiou_reward": 0.9240858852863312,
      "rewards/forensics_sinkhorn_reward": 0.4101519137620926,
      "rewards/forensics_strict_edge_reward": 0.23764817416667938,
      "rewards/forensics_tiou_reward": 0.25004614889621735,
      "step": 1626
    },
    {
      "completion_length": 27.015625,
      "epoch": 3.4041884816753925,
      "grad_norm": 27.0473690032959,
      "kl": 5.234375,
      "learning_rate": 1.4779874213836477e-07,
      "loss": 0.4227,
      "reward": 0.5756624341011047,
      "reward_std": 0.18330401182174683,
      "reward_std_all": 0.20948530733585358,
      "rewards-max/forensics_sinkhorn_reward": 0.9875693023204803,
      "rewards-max/forensics_strict_edge_reward": 0.9357379674911499,
      "rewards-max/forensics_tiou_reward": 0.9944010376930237,
      "rewards/forensics_sinkhorn_reward": 0.5079308748245239,
      "rewards/forensics_strict_edge_reward": 0.34449613094329834,
      "rewards/forensics_tiou_reward": 0.4096237123012543,
      "step": 1627
    },
    {
      "completion_length": 45.953125,
      "epoch": 3.406282722513089,
      "grad_norm": 21.068967819213867,
      "kl": 5.15625,
      "learning_rate": 1.4727463312368973e-07,
      "loss": 0.4377,
      "reward": 0.5582936406135559,
      "reward_std": 0.18970459699630737,
      "reward_std_all": 0.2172406241297722,
      "rewards-max/forensics_sinkhorn_reward": 0.8871442675590515,
      "rewards-max/forensics_strict_edge_reward": 0.8066739737987518,
      "rewards-max/forensics_tiou_reward": 0.7979816496372223,
      "rewards/forensics_sinkhorn_reward": 0.4537719190120697,
      "rewards/forensics_strict_edge_reward": 0.3042427897453308,
      "rewards/forensics_tiou_reward": 0.3180885910987854,
      "step": 1628
    },
    {
      "completion_length": 35.5,
      "epoch": 3.4083769633507854,
      "grad_norm": 15.257034301757812,
      "kl": 5.390625,
      "learning_rate": 1.4675052410901468e-07,
      "loss": 0.4272,
      "reward": 0.627335399389267,
      "reward_std": 0.25825558602809906,
      "reward_std_all": 0.23174190521240234,
      "rewards-max/forensics_sinkhorn_reward": 0.9862060844898224,
      "rewards-max/forensics_strict_edge_reward": 0.8849806487560272,
      "rewards-max/forensics_tiou_reward": 0.9890844225883484,
      "rewards/forensics_sinkhorn_reward": 0.5289450883865356,
      "rewards/forensics_strict_edge_reward": 0.3502102345228195,
      "rewards/forensics_tiou_reward": 0.3889544904232025,
      "step": 1629
    },
    {
      "completion_length": 41.90625,
      "epoch": 3.4104712041884815,
      "grad_norm": 20.288904190063477,
      "kl": 7.640625,
      "learning_rate": 1.4622641509433962e-07,
      "loss": 0.5337,
      "reward": 0.5334320962429047,
      "reward_std": 0.19287168234586716,
      "reward_std_all": 0.23109924793243408,
      "rewards-max/forensics_sinkhorn_reward": 0.9886753857135773,
      "rewards-max/forensics_strict_edge_reward": 0.9890114367008209,
      "rewards-max/forensics_tiou_reward": 0.9987577497959137,
      "rewards/forensics_sinkhorn_reward": 0.48153993487358093,
      "rewards/forensics_strict_edge_reward": 0.3079678416252136,
      "rewards/forensics_tiou_reward": 0.36576448380947113,
      "step": 1630
    },
    {
      "completion_length": 22.65625,
      "epoch": 3.412565445026178,
      "grad_norm": 29.63103485107422,
      "kl": 5.78125,
      "learning_rate": 1.4570230607966455e-07,
      "loss": 0.3357,
      "reward": 0.5575507283210754,
      "reward_std": 0.2233855277299881,
      "reward_std_all": 0.23231229931116104,
      "rewards-max/forensics_sinkhorn_reward": 0.9881294071674347,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9966081976890564,
      "rewards/forensics_sinkhorn_reward": 0.48382580280303955,
      "rewards/forensics_strict_edge_reward": 0.2652910351753235,
      "rewards/forensics_tiou_reward": 0.335853636264801,
      "step": 1631
    },
    {
      "completion_length": 59.359375,
      "epoch": 3.4146596858638745,
      "grad_norm": 24.52819061279297,
      "kl": 6.203125,
      "learning_rate": 1.451781970649895e-07,
      "loss": 0.2143,
      "reward": 0.5106228291988373,
      "reward_std": 0.2240620106458664,
      "reward_std_all": 0.2594011723995209,
      "rewards-max/forensics_sinkhorn_reward": 0.9753220081329346,
      "rewards-max/forensics_strict_edge_reward": 0.8445805907249451,
      "rewards-max/forensics_tiou_reward": 0.9688588082790375,
      "rewards/forensics_sinkhorn_reward": 0.3958028554916382,
      "rewards/forensics_strict_edge_reward": 0.19331125915050507,
      "rewards/forensics_tiou_reward": 0.21625982224941254,
      "step": 1632
    },
    {
      "completion_length": 40.703125,
      "epoch": 3.4167539267015705,
      "grad_norm": 27.482423782348633,
      "kl": 7.0625,
      "learning_rate": 1.4465408805031447e-07,
      "loss": 0.447,
      "reward": 0.5967268347740173,
      "reward_std": 0.22401777654886246,
      "reward_std_all": 0.225441575050354,
      "rewards-max/forensics_sinkhorn_reward": 0.9811348021030426,
      "rewards-max/forensics_strict_edge_reward": 0.9782643616199493,
      "rewards-max/forensics_tiou_reward": 0.9824495613574982,
      "rewards/forensics_sinkhorn_reward": 0.5015006363391876,
      "rewards/forensics_strict_edge_reward": 0.34950289130210876,
      "rewards/forensics_tiou_reward": 0.37439069151878357,
      "step": 1633
    },
    {
      "completion_length": 25.796875,
      "epoch": 3.418848167539267,
      "grad_norm": 28.093841552734375,
      "kl": 5.640625,
      "learning_rate": 1.441299790356394e-07,
      "loss": 0.34,
      "reward": 0.5515455007553101,
      "reward_std": 0.20861905813217163,
      "reward_std_all": 0.23891886323690414,
      "rewards-max/forensics_sinkhorn_reward": 0.9879239797592163,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9958035945892334,
      "rewards/forensics_sinkhorn_reward": 0.5393775403499603,
      "rewards/forensics_strict_edge_reward": 0.33138036727905273,
      "rewards/forensics_tiou_reward": 0.4248789995908737,
      "step": 1634
    },
    {
      "completion_length": 22.78125,
      "epoch": 3.4209424083769635,
      "grad_norm": 17.819150924682617,
      "kl": 5.125,
      "learning_rate": 1.4360587002096436e-07,
      "loss": 0.4482,
      "reward": 0.5762585401535034,
      "reward_std": 0.2002243772149086,
      "reward_std_all": 0.20516525208950043,
      "rewards-max/forensics_sinkhorn_reward": 0.9887720048427582,
      "rewards-max/forensics_strict_edge_reward": 0.9782643616199493,
      "rewards-max/forensics_tiou_reward": 0.9991390407085419,
      "rewards/forensics_sinkhorn_reward": 0.5563153326511383,
      "rewards/forensics_strict_edge_reward": 0.35241274535655975,
      "rewards/forensics_tiou_reward": 0.4618990868330002,
      "step": 1635
    },
    {
      "completion_length": 48.296875,
      "epoch": 3.4230366492146596,
      "grad_norm": 19.773630142211914,
      "kl": 10.75,
      "learning_rate": 1.430817610062893e-07,
      "loss": 0.7735,
      "reward": 0.4840185195207596,
      "reward_std": 0.33907386660575867,
      "reward_std_all": 0.33724047243595123,
      "rewards-max/forensics_sinkhorn_reward": 0.9577760994434357,
      "rewards-max/forensics_strict_edge_reward": 0.814805805683136,
      "rewards-max/forensics_tiou_reward": 0.9357163310050964,
      "rewards/forensics_sinkhorn_reward": 0.40185675024986267,
      "rewards/forensics_strict_edge_reward": 0.21755782514810562,
      "rewards/forensics_tiou_reward": 0.2832441180944443,
      "step": 1636
    },
    {
      "completion_length": 47.875,
      "epoch": 3.425130890052356,
      "grad_norm": 17.656713485717773,
      "kl": 7.53125,
      "learning_rate": 1.4255765199161425e-07,
      "loss": 0.6089,
      "reward": 0.5509243011474609,
      "reward_std": 0.2555241584777832,
      "reward_std_all": 0.25217901915311813,
      "rewards-max/forensics_sinkhorn_reward": 0.9886260330677032,
      "rewards-max/forensics_strict_edge_reward": 0.9890114367008209,
      "rewards-max/forensics_tiou_reward": 0.9985642433166504,
      "rewards/forensics_sinkhorn_reward": 0.48073233664035797,
      "rewards/forensics_strict_edge_reward": 0.2862154245376587,
      "rewards/forensics_tiou_reward": 0.3557841032743454,
      "step": 1637
    },
    {
      "completion_length": 41.296875,
      "epoch": 3.4272251308900525,
      "grad_norm": 20.31422233581543,
      "kl": 5.8828125,
      "learning_rate": 1.420335429769392e-07,
      "loss": 0.2111,
      "reward": 0.6495244204998016,
      "reward_std": 0.18484267219901085,
      "reward_std_all": 0.1730760596692562,
      "rewards-max/forensics_sinkhorn_reward": 0.983809232711792,
      "rewards-max/forensics_strict_edge_reward": 0.9677534997463226,
      "rewards-max/forensics_tiou_reward": 0.9889814257621765,
      "rewards/forensics_sinkhorn_reward": 0.6340531557798386,
      "rewards/forensics_strict_edge_reward": 0.40982668846845627,
      "rewards/forensics_tiou_reward": 0.49582722783088684,
      "step": 1638
    },
    {
      "completion_length": 33.078125,
      "epoch": 3.4293193717277486,
      "grad_norm": 26.23870086669922,
      "kl": 7.328125,
      "learning_rate": 1.4150943396226414e-07,
      "loss": 0.5091,
      "reward": 0.5449463725090027,
      "reward_std": 0.20092609524726868,
      "reward_std_all": 0.23881690204143524,
      "rewards-max/forensics_sinkhorn_reward": 0.986704409122467,
      "rewards-max/forensics_strict_edge_reward": 0.9574736058712006,
      "rewards-max/forensics_tiou_reward": 0.9910315573215485,
      "rewards/forensics_sinkhorn_reward": 0.4977220743894577,
      "rewards/forensics_strict_edge_reward": 0.2700253427028656,
      "rewards/forensics_tiou_reward": 0.35009826719760895,
      "step": 1639
    },
    {
      "completion_length": 43.546875,
      "epoch": 3.431413612565445,
      "grad_norm": 21.523313522338867,
      "kl": 5.671875,
      "learning_rate": 1.4098532494758907e-07,
      "loss": 0.4707,
      "reward": 0.5668940246105194,
      "reward_std": 0.21077817678451538,
      "reward_std_all": 0.22506625205278397,
      "rewards-max/forensics_sinkhorn_reward": 0.9884794652462006,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.997985452413559,
      "rewards/forensics_sinkhorn_reward": 0.5370350480079651,
      "rewards/forensics_strict_edge_reward": 0.28624725341796875,
      "rewards/forensics_tiou_reward": 0.45647601783275604,
      "step": 1640
    },
    {
      "completion_length": 21.4375,
      "epoch": 3.4335078534031416,
      "grad_norm": 19.703798294067383,
      "kl": 3.8046875,
      "learning_rate": 1.4046121593291406e-07,
      "loss": 0.3159,
      "reward": 0.6326483190059662,
      "reward_std": 0.11124916560947895,
      "reward_std_all": 0.14194494858384132,
      "rewards-max/forensics_sinkhorn_reward": 0.9874975681304932,
      "rewards-max/forensics_strict_edge_reward": 0.946017861366272,
      "rewards-max/forensics_tiou_reward": 0.9941251575946808,
      "rewards/forensics_sinkhorn_reward": 0.5775556862354279,
      "rewards/forensics_strict_edge_reward": 0.3371922969818115,
      "rewards/forensics_tiou_reward": 0.37623997032642365,
      "step": 1641
    },
    {
      "completion_length": 44.65625,
      "epoch": 3.4356020942408376,
      "grad_norm": 18.32278060913086,
      "kl": 6.359375,
      "learning_rate": 1.39937106918239e-07,
      "loss": 0.4881,
      "reward": 0.544829249382019,
      "reward_std": 0.15704768896102905,
      "reward_std_all": 0.20991280674934387,
      "rewards-max/forensics_sinkhorn_reward": 0.9136435389518738,
      "rewards-max/forensics_strict_edge_reward": 0.7545184791088104,
      "rewards-max/forensics_tiou_reward": 0.836900144815445,
      "rewards/forensics_sinkhorn_reward": 0.448333740234375,
      "rewards/forensics_strict_edge_reward": 0.248533483594656,
      "rewards/forensics_tiou_reward": 0.3036152124404907,
      "step": 1642
    },
    {
      "completion_length": 42.8125,
      "epoch": 3.437696335078534,
      "grad_norm": 15.522309303283691,
      "kl": 4.1328125,
      "learning_rate": 1.3941299790356392e-07,
      "loss": 0.3793,
      "reward": 0.5679643452167511,
      "reward_std": 0.1974235326051712,
      "reward_std_all": 0.22184006869792938,
      "rewards-max/forensics_sinkhorn_reward": 0.9880676865577698,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9963650107383728,
      "rewards/forensics_sinkhorn_reward": 0.542280375957489,
      "rewards/forensics_strict_edge_reward": 0.31099507212638855,
      "rewards/forensics_tiou_reward": 0.4328882247209549,
      "step": 1643
    },
    {
      "completion_length": 31.203125,
      "epoch": 3.4397905759162306,
      "grad_norm": 18.495731353759766,
      "kl": 6.0,
      "learning_rate": 1.3888888888888888e-07,
      "loss": 0.3759,
      "reward": 0.536875993013382,
      "reward_std": 0.2443757727742195,
      "reward_std_all": 0.25167927891016006,
      "rewards-max/forensics_sinkhorn_reward": 0.9580118656158447,
      "rewards-max/forensics_strict_edge_reward": 0.7466395199298859,
      "rewards-max/forensics_tiou_reward": 0.9336148202419281,
      "rewards/forensics_sinkhorn_reward": 0.4198392927646637,
      "rewards/forensics_strict_edge_reward": 0.24595744162797928,
      "rewards/forensics_tiou_reward": 0.31215815246105194,
      "step": 1644
    },
    {
      "completion_length": 27.46875,
      "epoch": 3.4418848167539267,
      "grad_norm": 20.44890594482422,
      "kl": 4.3828125,
      "learning_rate": 1.3836477987421384e-07,
      "loss": 0.1298,
      "reward": 0.6787979006767273,
      "reward_std": 0.14449146389961243,
      "reward_std_all": 0.13959479331970215,
      "rewards-max/forensics_sinkhorn_reward": 0.9758872985839844,
      "rewards-max/forensics_strict_edge_reward": 0.9256840348243713,
      "rewards-max/forensics_tiou_reward": 0.9694130718708038,
      "rewards/forensics_sinkhorn_reward": 0.5937447547912598,
      "rewards/forensics_strict_edge_reward": 0.27656223624944687,
      "rewards/forensics_tiou_reward": 0.4027988016605377,
      "step": 1645
    },
    {
      "completion_length": 44.75,
      "epoch": 3.443979057591623,
      "grad_norm": 20.545982360839844,
      "kl": 6.5,
      "learning_rate": 1.3784067085953877e-07,
      "loss": 0.2792,
      "reward": 0.5398530066013336,
      "reward_std": 0.25917984545230865,
      "reward_std_all": 0.25508658587932587,
      "rewards-max/forensics_sinkhorn_reward": 0.8301719427108765,
      "rewards-max/forensics_strict_edge_reward": 0.7530666887760162,
      "rewards-max/forensics_tiou_reward": 0.7839871346950531,
      "rewards/forensics_sinkhorn_reward": 0.37135279178619385,
      "rewards/forensics_strict_edge_reward": 0.21584924310445786,
      "rewards/forensics_tiou_reward": 0.21985777467489243,
      "step": 1646
    },
    {
      "completion_length": 19.96875,
      "epoch": 3.446073298429319,
      "grad_norm": 21.673564910888672,
      "kl": 2.4375,
      "learning_rate": 1.3731656184486373e-07,
      "loss": 0.2281,
      "reward": 0.6822000443935394,
      "reward_std": 0.11251937597990036,
      "reward_std_all": 0.11849676072597504,
      "rewards-max/forensics_sinkhorn_reward": 0.9878634810447693,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9955651164054871,
      "rewards/forensics_sinkhorn_reward": 0.6350829303264618,
      "rewards/forensics_strict_edge_reward": 0.3298463523387909,
      "rewards/forensics_tiou_reward": 0.4434065818786621,
      "step": 1647
    },
    {
      "completion_length": 29.3125,
      "epoch": 3.4481675392670157,
      "grad_norm": 20.047653198242188,
      "kl": 5.40625,
      "learning_rate": 1.3679245283018866e-07,
      "loss": 0.42,
      "reward": 0.5889819860458374,
      "reward_std": 0.29377785325050354,
      "reward_std_all": 0.26624874025583267,
      "rewards-max/forensics_sinkhorn_reward": 0.9821679890155792,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.983741283416748,
      "rewards/forensics_sinkhorn_reward": 0.5015271008014679,
      "rewards/forensics_strict_edge_reward": 0.30936679244041443,
      "rewards/forensics_tiou_reward": 0.3636331409215927,
      "step": 1648
    },
    {
      "completion_length": 31.078125,
      "epoch": 3.450261780104712,
      "grad_norm": 16.94072914123535,
      "kl": 6.59375,
      "learning_rate": 1.3626834381551362e-07,
      "loss": 0.4605,
      "reward": 0.49987006187438965,
      "reward_std": 0.3026110976934433,
      "reward_std_all": 0.31096673011779785,
      "rewards-max/forensics_sinkhorn_reward": 0.9871451258659363,
      "rewards-max/forensics_strict_edge_reward": 0.9574736058712006,
      "rewards-max/forensics_tiou_reward": 0.9927272796630859,
      "rewards/forensics_sinkhorn_reward": 0.43019305169582367,
      "rewards/forensics_strict_edge_reward": 0.25088198482990265,
      "rewards/forensics_tiou_reward": 0.29235057532787323,
      "step": 1649
    },
    {
      "completion_length": 33.546875,
      "epoch": 3.4523560209424082,
      "grad_norm": 23.07125473022461,
      "kl": 6.359375,
      "learning_rate": 1.3574423480083858e-07,
      "loss": 0.4996,
      "reward": 0.5937164723873138,
      "reward_std": 0.2172391712665558,
      "reward_std_all": 0.22131940722465515,
      "rewards-max/forensics_sinkhorn_reward": 0.9774203896522522,
      "rewards-max/forensics_strict_edge_reward": 0.8805811703205109,
      "rewards-max/forensics_tiou_reward": 0.9732509255409241,
      "rewards/forensics_sinkhorn_reward": 0.4687231034040451,
      "rewards/forensics_strict_edge_reward": 0.26903625577688217,
      "rewards/forensics_tiou_reward": 0.34909188747406006,
      "step": 1650
    },
    {
      "completion_length": 55.34375,
      "epoch": 3.4544502617801047,
      "grad_norm": 21.365896224975586,
      "kl": 7.71875,
      "learning_rate": 1.352201257861635e-07,
      "loss": 0.5433,
      "reward": 0.5576633214950562,
      "reward_std": 0.2763667553663254,
      "reward_std_all": 0.26316432654857635,
      "rewards-max/forensics_sinkhorn_reward": 0.9878886044025421,
      "rewards-max/forensics_strict_edge_reward": 0.9464850425720215,
      "rewards-max/forensics_tiou_reward": 0.9956563711166382,
      "rewards/forensics_sinkhorn_reward": 0.5278530120849609,
      "rewards/forensics_strict_edge_reward": 0.34201914072036743,
      "rewards/forensics_tiou_reward": 0.4277011454105377,
      "step": 1651
    },
    {
      "completion_length": 56.640625,
      "epoch": 3.456544502617801,
      "grad_norm": 16.726322174072266,
      "kl": 6.71875,
      "learning_rate": 1.3469601677148844e-07,
      "loss": 0.5315,
      "reward": 0.5095238387584686,
      "reward_std": 0.28029564023017883,
      "reward_std_all": 0.2832697331905365,
      "rewards-max/forensics_sinkhorn_reward": 0.9763996005058289,
      "rewards-max/forensics_strict_edge_reward": 0.8957232534885406,
      "rewards-max/forensics_tiou_reward": 0.9685012400150299,
      "rewards/forensics_sinkhorn_reward": 0.41446442902088165,
      "rewards/forensics_strict_edge_reward": 0.203546654433012,
      "rewards/forensics_tiou_reward": 0.23718994110822678,
      "step": 1652
    },
    {
      "completion_length": 53.90625,
      "epoch": 3.4586387434554973,
      "grad_norm": 17.456588745117188,
      "kl": 6.765625,
      "learning_rate": 1.3417190775681343e-07,
      "loss": 0.5564,
      "reward": 0.512054443359375,
      "reward_std": 0.2578486353158951,
      "reward_std_all": 0.2823583334684372,
      "rewards-max/forensics_sinkhorn_reward": 0.9883950054645538,
      "rewards-max/forensics_strict_edge_reward": 0.9890114367008209,
      "rewards-max/forensics_tiou_reward": 0.9976553320884705,
      "rewards/forensics_sinkhorn_reward": 0.45957815647125244,
      "rewards/forensics_strict_edge_reward": 0.3352584093809128,
      "rewards/forensics_tiou_reward": 0.3860906958580017,
      "step": 1653
    },
    {
      "completion_length": 23.0625,
      "epoch": 3.4607329842931938,
      "grad_norm": 19.5649356842041,
      "kl": 6.0078125,
      "learning_rate": 1.3364779874213836e-07,
      "loss": 0.3629,
      "reward": 0.590384230017662,
      "reward_std": 0.1653449535369873,
      "reward_std_all": 0.19394156336784363,
      "rewards-max/forensics_sinkhorn_reward": 0.9879995882511139,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9960942566394806,
      "rewards/forensics_sinkhorn_reward": 0.5007366240024567,
      "rewards/forensics_strict_edge_reward": 0.3157999813556671,
      "rewards/forensics_tiou_reward": 0.35943514108657837,
      "step": 1654
    },
    {
      "completion_length": 35.890625,
      "epoch": 3.46282722513089,
      "grad_norm": 25.35915184020996,
      "kl": 7.96875,
      "learning_rate": 1.331236897274633e-07,
      "loss": 0.5563,
      "reward": 0.5502378046512604,
      "reward_std": 0.23103734850883484,
      "reward_std_all": 0.24287676066160202,
      "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": 0.5528410375118256,
      "rewards/forensics_strict_edge_reward": 0.4231775104999542,
      "rewards/forensics_tiou_reward": 0.49734702706336975,
      "step": 1655
    },
    {
      "completion_length": 34.453125,
      "epoch": 3.4649214659685863,
      "grad_norm": 22.231035232543945,
      "kl": 5.078125,
      "learning_rate": 1.3259958071278825e-07,
      "loss": 0.3819,
      "reward": 0.5890368521213531,
      "reward_std": 0.1793934926390648,
      "reward_std_all": 0.20200227200984955,
      "rewards-max/forensics_sinkhorn_reward": 0.9883950054645538,
      "rewards-max/forensics_strict_edge_reward": 0.9890114367008209,
      "rewards-max/forensics_tiou_reward": 0.9976553320884705,
      "rewards/forensics_sinkhorn_reward": 0.5216448903083801,
      "rewards/forensics_strict_edge_reward": 0.3400525450706482,
      "rewards/forensics_tiou_reward": 0.43506433069705963,
      "step": 1656
    },
    {
      "completion_length": 29.84375,
      "epoch": 3.467015706806283,
      "grad_norm": 34.886131286621094,
      "kl": 7.359375,
      "learning_rate": 1.320754716981132e-07,
      "loss": 0.4671,
      "reward": 0.5574808716773987,
      "reward_std": 0.22608879208564758,
      "reward_std_all": 0.23718544095754623,
      "rewards-max/forensics_sinkhorn_reward": 0.9880317747592926,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9962243437767029,
      "rewards/forensics_sinkhorn_reward": 0.5104649066925049,
      "rewards/forensics_strict_edge_reward": 0.29054613411426544,
      "rewards/forensics_tiou_reward": 0.317804291844368,
      "step": 1657
    },
    {
      "completion_length": 29.28125,
      "epoch": 3.469109947643979,
      "grad_norm": 25.144502639770508,
      "kl": 7.53125,
      "learning_rate": 1.3155136268343814e-07,
      "loss": 0.5295,
      "reward": 0.44771595299243927,
      "reward_std": 0.2621098458766937,
      "reward_std_all": 0.30621933937072754,
      "rewards-max/forensics_sinkhorn_reward": 0.9338429272174835,
      "rewards-max/forensics_strict_edge_reward": 0.7839131057262421,
      "rewards-max/forensics_tiou_reward": 0.8766552805900574,
      "rewards/forensics_sinkhorn_reward": 0.2979820817708969,
      "rewards/forensics_strict_edge_reward": 0.1769295409321785,
      "rewards/forensics_tiou_reward": 0.16005827486515045,
      "step": 1658
    },
    {
      "completion_length": 21.078125,
      "epoch": 3.4712041884816753,
      "grad_norm": 32.57039260864258,
      "kl": 8.09375,
      "learning_rate": 1.310272536687631e-07,
      "loss": 0.6355,
      "reward": 0.46176427602767944,
      "reward_std": 0.346561461687088,
      "reward_std_all": 0.3472086638212204,
      "rewards-max/forensics_sinkhorn_reward": 0.9865386784076691,
      "rewards-max/forensics_strict_edge_reward": 0.9574736058712006,
      "rewards-max/forensics_tiou_reward": 0.9903848469257355,
      "rewards/forensics_sinkhorn_reward": 0.39140576124191284,
      "rewards/forensics_strict_edge_reward": 0.2437509223818779,
      "rewards/forensics_tiou_reward": 0.30720896273851395,
      "step": 1659
    },
    {
      "completion_length": 27.3125,
      "epoch": 3.473298429319372,
      "grad_norm": 25.08402442932129,
      "kl": 5.75,
      "learning_rate": 1.3050314465408803e-07,
      "loss": 0.3824,
      "reward": 0.5459299683570862,
      "reward_std": 0.20251938700675964,
      "reward_std_all": 0.23420988023281097,
      "rewards-max/forensics_sinkhorn_reward": 0.9695034623146057,
      "rewards-max/forensics_strict_edge_reward": 0.8698340952396393,
      "rewards-max/forensics_tiou_reward": 0.9261694252490997,
      "rewards/forensics_sinkhorn_reward": 0.4375583976507187,
      "rewards/forensics_strict_edge_reward": 0.26928867399692535,
      "rewards/forensics_tiou_reward": 0.30926109850406647,
      "step": 1660
    },
    {
      "completion_length": 21.890625,
      "epoch": 3.475392670157068,
      "grad_norm": 21.628263473510742,
      "kl": 7.09375,
      "learning_rate": 1.29979035639413e-07,
      "loss": 0.5559,
      "reward": 0.49183380603790283,
      "reward_std": 0.2548399791121483,
      "reward_std_all": 0.27883418649435043,
      "rewards-max/forensics_sinkhorn_reward": 0.9885127544403076,
      "rewards-max/forensics_strict_edge_reward": 0.9677534997463226,
      "rewards-max/forensics_tiou_reward": 0.99811851978302,
      "rewards/forensics_sinkhorn_reward": 0.4182181656360626,
      "rewards/forensics_strict_edge_reward": 0.27547572553157806,
      "rewards/forensics_tiou_reward": 0.32650845497846603,
      "step": 1661
    },
    {
      "completion_length": 30.75,
      "epoch": 3.4774869109947644,
      "grad_norm": 24.753358840942383,
      "kl": 5.421875,
      "learning_rate": 1.2945492662473795e-07,
      "loss": 0.3233,
      "reward": 0.5727781355381012,
      "reward_std": 0.21471627801656723,
      "reward_std_all": 0.22818128764629364,
      "rewards-max/forensics_sinkhorn_reward": 0.9880876541137695,
      "rewards-max/forensics_strict_edge_reward": 0.9567649364471436,
      "rewards-max/forensics_tiou_reward": 0.9964441359043121,
      "rewards/forensics_sinkhorn_reward": 0.48285046219825745,
      "rewards/forensics_strict_edge_reward": 0.27426691353321075,
      "rewards/forensics_tiou_reward": 0.3290821760892868,
      "step": 1662
    },
    {
      "completion_length": 27.546875,
      "epoch": 3.479581151832461,
      "grad_norm": 31.819345474243164,
      "kl": 7.484375,
      "learning_rate": 1.2893081761006288e-07,
      "loss": 0.5234,
      "reward": 0.5118332058191299,
      "reward_std": 0.28853797912597656,
      "reward_std_all": 0.2970478981733322,
      "rewards-max/forensics_sinkhorn_reward": 0.9737953841686249,
      "rewards-max/forensics_strict_edge_reward": 0.8472771048545837,
      "rewards-max/forensics_tiou_reward": 0.9650979042053223,
      "rewards/forensics_sinkhorn_reward": 0.41279880702495575,
      "rewards/forensics_strict_edge_reward": 0.21417294442653656,
      "rewards/forensics_tiou_reward": 0.2729237973690033,
      "step": 1663
    },
    {
      "completion_length": 31.15625,
      "epoch": 3.481675392670157,
      "grad_norm": 23.578763961791992,
      "kl": 8.578125,
      "learning_rate": 1.2840670859538782e-07,
      "loss": 0.4815,
      "reward": 0.5525682866573334,
      "reward_std": 0.20854692161083221,
      "reward_std_all": 0.22975897043943405,
      "rewards-max/forensics_sinkhorn_reward": 0.9874047040939331,
      "rewards-max/forensics_strict_edge_reward": 0.9574736058712006,
      "rewards-max/forensics_tiou_reward": 0.9937500059604645,
      "rewards/forensics_sinkhorn_reward": 0.5056800097227097,
      "rewards/forensics_strict_edge_reward": 0.3240649104118347,
      "rewards/forensics_tiou_reward": 0.3810541182756424,
      "step": 1664
    },
    {
      "completion_length": 26.0625,
      "epoch": 3.4837696335078534,
      "grad_norm": 19.09633445739746,
      "kl": 5.640625,
      "learning_rate": 1.278825995807128e-07,
      "loss": 0.3319,
      "reward": 0.5814192295074463,
      "reward_std": 0.21439459174871445,
      "reward_std_all": 0.21545615792274475,
      "rewards-max/forensics_sinkhorn_reward": 0.9883498847484589,
      "rewards-max/forensics_strict_edge_reward": 0.9677534997463226,
      "rewards-max/forensics_tiou_reward": 0.9974779486656189,
      "rewards/forensics_sinkhorn_reward": 0.4953932613134384,
      "rewards/forensics_strict_edge_reward": 0.28589656203985214,
      "rewards/forensics_tiou_reward": 0.3446476310491562,
      "step": 1665
    },
    {
      "completion_length": 43.078125,
      "epoch": 3.48586387434555,
      "grad_norm": 20.979490280151367,
      "kl": 9.34375,
      "learning_rate": 1.2735849056603773e-07,
      "loss": 0.6711,
      "reward": 0.47011883556842804,
      "reward_std": 0.19180336222052574,
      "reward_std_all": 0.2542174905538559,
      "rewards-max/forensics_sinkhorn_reward": 0.9884341657161713,
      "rewards-max/forensics_strict_edge_reward": 0.9890114367008209,
      "rewards-max/forensics_tiou_reward": 0.9978094100952148,
      "rewards/forensics_sinkhorn_reward": 0.39777252078056335,
      "rewards/forensics_strict_edge_reward": 0.30739739537239075,
      "rewards/forensics_tiou_reward": 0.325292631983757,
      "step": 1666
    },
    {
      "completion_length": 22.28125,
      "epoch": 3.487958115183246,
      "grad_norm": 18.470849990844727,
      "kl": 5.6875,
      "learning_rate": 1.2683438155136267e-07,
      "loss": 0.3728,
      "reward": 0.5778807103633881,
      "reward_std": 0.20060597360134125,
      "reward_std_all": 0.2163560912013054,
      "rewards-max/forensics_sinkhorn_reward": 0.9817769825458527,
      "rewards-max/forensics_strict_edge_reward": 0.9464850425720215,
      "rewards-max/forensics_tiou_reward": 0.983155369758606,
      "rewards/forensics_sinkhorn_reward": 0.5553667545318604,
      "rewards/forensics_strict_edge_reward": 0.39427053183317184,
      "rewards/forensics_tiou_reward": 0.4505922794342041,
      "step": 1667
    },
    {
      "completion_length": 25.265625,
      "epoch": 3.4900523560209424,
      "grad_norm": 20.725168228149414,
      "kl": 5.21875,
      "learning_rate": 1.2631027253668762e-07,
      "loss": 0.3819,
      "reward": 0.5793972015380859,
      "reward_std": 0.21530504524707794,
      "reward_std_all": 0.2301645278930664,
      "rewards-max/forensics_sinkhorn_reward": 0.9700117707252502,
      "rewards-max/forensics_strict_edge_reward": 0.8968285918235779,
      "rewards-max/forensics_tiou_reward": 0.9555910229682922,
      "rewards/forensics_sinkhorn_reward": 0.5282177329063416,
      "rewards/forensics_strict_edge_reward": 0.2586687505245209,
      "rewards/forensics_tiou_reward": 0.36564628779888153,
      "step": 1668
    },
    {
      "completion_length": 26.484375,
      "epoch": 3.492146596858639,
      "grad_norm": 28.40323257446289,
      "kl": 5.671875,
      "learning_rate": 1.2578616352201258e-07,
      "loss": 0.4868,
      "reward": 0.5510201007127762,
      "reward_std": 0.2579655051231384,
      "reward_std_all": 0.26167648285627365,
      "rewards-max/forensics_sinkhorn_reward": 0.9877949655056,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9952948689460754,
      "rewards/forensics_sinkhorn_reward": 0.5324132144451141,
      "rewards/forensics_strict_edge_reward": 0.3718266487121582,
      "rewards/forensics_tiou_reward": 0.4243031442165375,
      "step": 1669
    },
    {
      "completion_length": 32.875,
      "epoch": 3.494240837696335,
      "grad_norm": 21.926225662231445,
      "kl": 7.921875,
      "learning_rate": 1.2526205450733751e-07,
      "loss": 0.5598,
      "reward": 0.5149552822113037,
      "reward_std": 0.21735472977161407,
      "reward_std_all": 0.25216689705848694,
      "rewards-max/forensics_sinkhorn_reward": 0.962911069393158,
      "rewards-max/forensics_strict_edge_reward": 0.8893801271915436,
      "rewards-max/forensics_tiou_reward": 0.9429795444011688,
      "rewards/forensics_sinkhorn_reward": 0.4285642206668854,
      "rewards/forensics_strict_edge_reward": 0.24374975264072418,
      "rewards/forensics_tiou_reward": 0.30968476831912994,
      "step": 1670
    },
    {
      "completion_length": 28.6875,
      "epoch": 3.4963350785340315,
      "grad_norm": 37.20658874511719,
      "kl": 7.703125,
      "learning_rate": 1.2473794549266247e-07,
      "loss": 0.4431,
      "reward": 0.5654116570949554,
      "reward_std": 0.2531561553478241,
      "reward_std_all": 0.24582330882549286,
      "rewards-max/forensics_sinkhorn_reward": 0.988498866558075,
      "rewards-max/forensics_strict_edge_reward": 0.9782643616199493,
      "rewards-max/forensics_tiou_reward": 0.9980638921260834,
      "rewards/forensics_sinkhorn_reward": 0.5230060517787933,
      "rewards/forensics_strict_edge_reward": 0.38529127836227417,
      "rewards/forensics_tiou_reward": 0.411098450422287,
      "step": 1671
    },
    {
      "completion_length": 30.40625,
      "epoch": 3.498429319371728,
      "grad_norm": 24.706796646118164,
      "kl": 7.796875,
      "learning_rate": 1.242138364779874e-07,
      "loss": 0.5,
      "reward": 0.5545433759689331,
      "reward_std": 0.25139474868774414,
      "reward_std_all": 0.2566746547818184,
      "rewards-max/forensics_sinkhorn_reward": 0.9611122608184814,
      "rewards-max/forensics_strict_edge_reward": 0.9085061252117157,
      "rewards-max/forensics_tiou_reward": 0.940268486738205,
      "rewards/forensics_sinkhorn_reward": 0.5140906721353531,
      "rewards/forensics_strict_edge_reward": 0.3321193903684616,
      "rewards/forensics_tiou_reward": 0.42768794298171997,
      "step": 1672
    },
    {
      "completion_length": 19.34375,
      "epoch": 3.500523560209424,
      "grad_norm": 24.97080421447754,
      "kl": 6.265625,
      "learning_rate": 1.2368972746331236e-07,
      "loss": 0.4171,
      "reward": 0.5646125078201294,
      "reward_std": 0.2334062159061432,
      "reward_std_all": 0.2424679920077324,
      "rewards-max/forensics_sinkhorn_reward": 0.9877483546733856,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9951096177101135,
      "rewards/forensics_sinkhorn_reward": 0.5191138088703156,
      "rewards/forensics_strict_edge_reward": 0.35450728237628937,
      "rewards/forensics_tiou_reward": 0.42188267409801483,
      "step": 1673
    },
    {
      "completion_length": 32.046875,
      "epoch": 3.5026178010471205,
      "grad_norm": 23.675626754760742,
      "kl": 6.1796875,
      "learning_rate": 1.231656184486373e-07,
      "loss": 0.4944,
      "reward": 0.5624526739120483,
      "reward_std": 0.1925356760621071,
      "reward_std_all": 0.22300390899181366,
      "rewards-max/forensics_sinkhorn_reward": 0.9874028563499451,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9937526285648346,
      "rewards/forensics_sinkhorn_reward": 0.5158796310424805,
      "rewards/forensics_strict_edge_reward": 0.3755563795566559,
      "rewards/forensics_tiou_reward": 0.42456860840320587,
      "step": 1674
    },
    {
      "completion_length": 33.90625,
      "epoch": 3.504712041884817,
      "grad_norm": 17.772029876708984,
      "kl": 8.59375,
      "learning_rate": 1.2264150943396226e-07,
      "loss": 0.6013,
      "reward": 0.5344276130199432,
      "reward_std": 0.15703772008419037,
      "reward_std_all": 0.21800972521305084,
      "rewards-max/forensics_sinkhorn_reward": 0.9616991877555847,
      "rewards-max/forensics_strict_edge_reward": 0.8578422963619232,
      "rewards-max/forensics_tiou_reward": 0.9353415071964264,
      "rewards/forensics_sinkhorn_reward": 0.4588790684938431,
      "rewards/forensics_strict_edge_reward": 0.287103608250618,
      "rewards/forensics_tiou_reward": 0.35516759753227234,
      "step": 1675
    },
    {
      "completion_length": 31.171875,
      "epoch": 3.506806282722513,
      "grad_norm": 30.205982208251953,
      "kl": 10.828125,
      "learning_rate": 1.2211740041928721e-07,
      "loss": 0.5014,
      "reward": 0.4585917294025421,
      "reward_std": 0.20322441309690475,
      "reward_std_all": 0.27142590284347534,
      "rewards-max/forensics_sinkhorn_reward": 0.9877626001834869,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9951662719249725,
      "rewards/forensics_sinkhorn_reward": 0.38841041922569275,
      "rewards/forensics_strict_edge_reward": 0.2208937257528305,
      "rewards/forensics_tiou_reward": 0.2353416383266449,
      "step": 1676
    },
    {
      "completion_length": 28.453125,
      "epoch": 3.5089005235602095,
      "grad_norm": 28.00127601623535,
      "kl": 10.34375,
      "learning_rate": 1.2159329140461215e-07,
      "loss": 0.746,
      "reward": 0.3606208860874176,
      "reward_std": 0.17513708025217056,
      "reward_std_all": 0.305758535861969,
      "rewards-max/forensics_sinkhorn_reward": 0.9882758855819702,
      "rewards-max/forensics_strict_edge_reward": 0.9782643616199493,
      "rewards-max/forensics_tiou_reward": 0.997183084487915,
      "rewards/forensics_sinkhorn_reward": 0.3062794804573059,
      "rewards/forensics_strict_edge_reward": 0.17642058432102203,
      "rewards/forensics_tiou_reward": 0.20170475542545319,
      "step": 1677
    },
    {
      "completion_length": 25.9375,
      "epoch": 3.510994764397906,
      "grad_norm": 18.286394119262695,
      "kl": 6.9375,
      "learning_rate": 1.210691823899371e-07,
      "loss": 0.4944,
      "reward": 0.5762183219194412,
      "reward_std": 0.2711898535490036,
      "reward_std_all": 0.2564208507537842,
      "rewards-max/forensics_sinkhorn_reward": 0.9857390522956848,
      "rewards-max/forensics_strict_edge_reward": 0.8856562972068787,
      "rewards-max/forensics_tiou_reward": 0.9872169494628906,
      "rewards/forensics_sinkhorn_reward": 0.47569170594215393,
      "rewards/forensics_strict_edge_reward": 0.2641187384724617,
      "rewards/forensics_tiou_reward": 0.3215121477842331,
      "step": 1678
    },
    {
      "completion_length": 30.5,
      "epoch": 3.513089005235602,
      "grad_norm": 21.24293327331543,
      "kl": 6.4375,
      "learning_rate": 1.2054507337526206e-07,
      "loss": 0.5851,
      "reward": 0.508778914809227,
      "reward_std": 0.2791191339492798,
      "reward_std_all": 0.29296331107616425,
      "rewards-max/forensics_sinkhorn_reward": 0.9865207672119141,
      "rewards-max/forensics_strict_edge_reward": 0.8876297473907471,
      "rewards-max/forensics_tiou_reward": 0.990274965763092,
      "rewards/forensics_sinkhorn_reward": 0.4375976622104645,
      "rewards/forensics_strict_edge_reward": 0.28253407776355743,
      "rewards/forensics_tiou_reward": 0.31793196499347687,
      "step": 1679
    },
    {
      "completion_length": 26.328125,
      "epoch": 3.5151832460732986,
      "grad_norm": 16.019725799560547,
      "kl": 4.375,
      "learning_rate": 1.20020964360587e-07,
      "loss": 0.4481,
      "reward": 0.6207168102264404,
      "reward_std": 0.23236578702926636,
      "reward_std_all": 0.21393360942602158,
      "rewards-max/forensics_sinkhorn_reward": 0.9884204268455505,
      "rewards-max/forensics_strict_edge_reward": 0.9782643616199493,
      "rewards-max/forensics_tiou_reward": 0.9977528154850006,
      "rewards/forensics_sinkhorn_reward": 0.5614878684282303,
      "rewards/forensics_strict_edge_reward": 0.39834195375442505,
      "rewards/forensics_tiou_reward": 0.46655744314193726,
      "step": 1680
    },
    {
      "completion_length": 37.375,
      "epoch": 3.5172774869109946,
      "grad_norm": 32.343963623046875,
      "kl": 7.140625,
      "learning_rate": 1.1949685534591195e-07,
      "loss": 0.4468,
      "reward": 0.5423403382301331,
      "reward_std": 0.3121293783187866,
      "reward_std_all": 0.29044950008392334,
      "rewards-max/forensics_sinkhorn_reward": 0.9680292010307312,
      "rewards-max/forensics_strict_edge_reward": 0.8394034504890442,
      "rewards-max/forensics_tiou_reward": 0.9532474279403687,
      "rewards/forensics_sinkhorn_reward": 0.4273485392332077,
      "rewards/forensics_strict_edge_reward": 0.27424807846546173,
      "rewards/forensics_tiou_reward": 0.32896894216537476,
      "step": 1681
    },
    {
      "completion_length": 27.453125,
      "epoch": 3.519371727748691,
      "grad_norm": 27.494714736938477,
      "kl": 6.984375,
      "learning_rate": 1.189727463312369e-07,
      "loss": 0.4194,
      "reward": 0.559579074382782,
      "reward_std": 0.1690787971019745,
      "reward_std_all": 0.2106521874666214,
      "rewards-max/forensics_sinkhorn_reward": 0.9860574901103973,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9929398596286774,
      "rewards/forensics_sinkhorn_reward": 0.45785626769065857,
      "rewards/forensics_strict_edge_reward": 0.21613728255033493,
      "rewards/forensics_tiou_reward": 0.30515143275260925,
      "step": 1682
    },
    {
      "completion_length": 28.484375,
      "epoch": 3.521465968586387,
      "grad_norm": 17.695865631103516,
      "kl": 6.359375,
      "learning_rate": 1.1844863731656183e-07,
      "loss": 0.4128,
      "reward": 0.5797021985054016,
      "reward_std": 0.24661675095558167,
      "reward_std_all": 0.238797627389431,
      "rewards-max/forensics_sinkhorn_reward": 0.8740085363388062,
      "rewards-max/forensics_strict_edge_reward": 0.7143478095531464,
      "rewards-max/forensics_tiou_reward": 0.7584776878356934,
      "rewards/forensics_sinkhorn_reward": 0.4004160612821579,
      "rewards/forensics_strict_edge_reward": 0.20619703829288483,
      "rewards/forensics_tiou_reward": 0.22602718323469162,
      "step": 1683
    },
    {
      "completion_length": 35.21875,
      "epoch": 3.5235602094240837,
      "grad_norm": 16.81981086730957,
      "kl": 5.875,
      "learning_rate": 1.1792452830188679e-07,
      "loss": 0.4405,
      "reward": 0.5933617353439331,
      "reward_std": 0.18795405328273773,
      "reward_std_all": 0.20341100543737411,
      "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": 0.5644425451755524,
      "rewards/forensics_strict_edge_reward": 0.35317549109458923,
      "rewards/forensics_tiou_reward": 0.45873764157295227,
      "step": 1684
    },
    {
      "completion_length": 37.015625,
      "epoch": 3.52565445026178,
      "grad_norm": 35.54890441894531,
      "kl": 8.546875,
      "learning_rate": 1.1740041928721174e-07,
      "loss": 0.5482,
      "reward": 0.5147078633308411,
      "reward_std": 0.24624907970428467,
      "reward_std_all": 0.2631721794605255,
      "rewards-max/forensics_sinkhorn_reward": 0.988801896572113,
      "rewards-max/forensics_strict_edge_reward": 0.9890114367008209,
      "rewards-max/forensics_tiou_reward": 0.9992581605911255,
      "rewards/forensics_sinkhorn_reward": 0.4331589788198471,
      "rewards/forensics_strict_edge_reward": 0.3107891157269478,
      "rewards/forensics_tiou_reward": 0.3368830010294914,
      "step": 1685
    },
    {
      "completion_length": 19.140625,
      "epoch": 3.527748691099476,
      "grad_norm": 33.437538146972656,
      "kl": 6.953125,
      "learning_rate": 1.1687631027253668e-07,
      "loss": 0.3127,
      "reward": 0.5323087275028229,
      "reward_std": 0.13808459043502808,
      "reward_std_all": 0.19814608991146088,
      "rewards-max/forensics_sinkhorn_reward": 0.986058384180069,
      "rewards-max/forensics_strict_edge_reward": 0.8983768224716187,
      "rewards-max/forensics_tiou_reward": 0.9884916841983795,
      "rewards/forensics_sinkhorn_reward": 0.42574696242809296,
      "rewards/forensics_strict_edge_reward": 0.1427607387304306,
      "rewards/forensics_tiou_reward": 0.20901135355234146,
      "step": 1686
    },
    {
      "completion_length": 22.796875,
      "epoch": 3.5298429319371727,
      "grad_norm": 25.319862365722656,
      "kl": 6.421875,
      "learning_rate": 1.1635220125786163e-07,
      "loss": 0.3325,
      "reward": 0.541168749332428,
      "reward_std": 0.17413976788520813,
      "reward_std_all": 0.22219200432300568,
      "rewards-max/forensics_sinkhorn_reward": 0.952131599187851,
      "rewards-max/forensics_strict_edge_reward": 0.8677490055561066,
      "rewards-max/forensics_tiou_reward": 0.9128174483776093,
      "rewards/forensics_sinkhorn_reward": 0.4532502591609955,
      "rewards/forensics_strict_edge_reward": 0.28500503301620483,
      "rewards/forensics_tiou_reward": 0.2940252721309662,
      "step": 1687
    },
    {
      "completion_length": 37.03125,
      "epoch": 3.531937172774869,
      "grad_norm": 17.1223087310791,
      "kl": 4.5703125,
      "learning_rate": 1.1582809224318659e-07,
      "loss": 0.5064,
      "reward": 0.5863508582115173,
      "reward_std": 0.2024591863155365,
      "reward_std_all": 0.21235280483961105,
      "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": 0.5983397960662842,
      "rewards/forensics_strict_edge_reward": 0.4076634645462036,
      "rewards/forensics_tiou_reward": 0.48494458198547363,
      "step": 1688
    },
    {
      "completion_length": 51.78125,
      "epoch": 3.5340314136125652,
      "grad_norm": 28.950658798217773,
      "kl": 9.125,
      "learning_rate": 1.1530398322851152e-07,
      "loss": 0.6922,
      "reward": 0.5649940967559814,
      "reward_std": 0.2299366444349289,
      "reward_std_all": 0.2394634485244751,
      "rewards-max/forensics_sinkhorn_reward": 0.9838802516460419,
      "rewards-max/forensics_strict_edge_reward": 0.9357379674911499,
      "rewards-max/forensics_tiou_reward": 0.9871259927749634,
      "rewards/forensics_sinkhorn_reward": 0.4902675151824951,
      "rewards/forensics_strict_edge_reward": 0.27467264235019684,
      "rewards/forensics_tiou_reward": 0.35155831277370453,
      "step": 1689
    },
    {
      "completion_length": 19.125,
      "epoch": 3.5361256544502617,
      "grad_norm": 17.520612716674805,
      "kl": 3.8671875,
      "learning_rate": 1.1477987421383648e-07,
      "loss": 0.2188,
      "reward": 0.6986099183559418,
      "reward_std": 0.11328406631946564,
      "reward_std_all": 0.11621993407607079,
      "rewards-max/forensics_sinkhorn_reward": 0.9729317724704742,
      "rewards-max/forensics_strict_edge_reward": 0.9256840348243713,
      "rewards-max/forensics_tiou_reward": 0.9621921479701996,
      "rewards/forensics_sinkhorn_reward": 0.6098202168941498,
      "rewards/forensics_strict_edge_reward": 0.2977060377597809,
      "rewards/forensics_tiou_reward": 0.39921845495700836,
      "step": 1690
    },
    {
      "completion_length": 29.484375,
      "epoch": 3.538219895287958,
      "grad_norm": 23.232431411743164,
      "kl": 6.265625,
      "learning_rate": 1.1425576519916141e-07,
      "loss": 0.5435,
      "reward": 0.5477115213871002,
      "reward_std": 0.20342518389225006,
      "reward_std_all": 0.23972336947917938,
      "rewards-max/forensics_sinkhorn_reward": 0.9835109710693359,
      "rewards-max/forensics_strict_edge_reward": 0.8578422963619232,
      "rewards-max/forensics_tiou_reward": 0.9784487783908844,
      "rewards/forensics_sinkhorn_reward": 0.4754121005535126,
      "rewards/forensics_strict_edge_reward": 0.2654125243425369,
      "rewards/forensics_tiou_reward": 0.3372710794210434,
      "step": 1691
    },
    {
      "completion_length": 33.4375,
      "epoch": 3.5403141361256543,
      "grad_norm": 20.000675201416016,
      "kl": 4.9296875,
      "learning_rate": 1.1373165618448637e-07,
      "loss": 0.2829,
      "reward": 0.6384735405445099,
      "reward_std": 0.2151089608669281,
      "reward_std_all": 0.19903625547885895,
      "rewards-max/forensics_sinkhorn_reward": 0.9888071715831757,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9992772936820984,
      "rewards/forensics_sinkhorn_reward": 0.602200984954834,
      "rewards/forensics_strict_edge_reward": 0.37570399045944214,
      "rewards/forensics_tiou_reward": 0.46223220229148865,
      "step": 1692
    },
    {
      "completion_length": 23.3125,
      "epoch": 3.5424083769633508,
      "grad_norm": 23.93439292907715,
      "kl": 5.5,
      "learning_rate": 1.1320754716981131e-07,
      "loss": 0.3818,
      "reward": 0.6084205806255341,
      "reward_std": 0.18456773459911346,
      "reward_std_all": 0.18927302211523056,
      "rewards-max/forensics_sinkhorn_reward": 0.9718754291534424,
      "rewards-max/forensics_strict_edge_reward": 0.8987811803817749,
      "rewards-max/forensics_tiou_reward": 0.961304634809494,
      "rewards/forensics_sinkhorn_reward": 0.4875521808862686,
      "rewards/forensics_strict_edge_reward": 0.27716386318206787,
      "rewards/forensics_tiou_reward": 0.3240126967430115,
      "step": 1693
    },
    {
      "completion_length": 34.5,
      "epoch": 3.5445026178010473,
      "grad_norm": 19.667631149291992,
      "kl": 7.375,
      "learning_rate": 1.1268343815513626e-07,
      "loss": 0.5277,
      "reward": 0.5368121266365051,
      "reward_std": 0.17343533039093018,
      "reward_std_all": 0.2208331674337387,
      "rewards-max/forensics_sinkhorn_reward": 0.9883732497692108,
      "rewards-max/forensics_strict_edge_reward": 0.9782643616199493,
      "rewards-max/forensics_tiou_reward": 0.9975698590278625,
      "rewards/forensics_sinkhorn_reward": 0.4388182610273361,
      "rewards/forensics_strict_edge_reward": 0.2774985730648041,
      "rewards/forensics_tiou_reward": 0.3002014011144638,
      "step": 1694
    },
    {
      "completion_length": 28.15625,
      "epoch": 3.5465968586387433,
      "grad_norm": 23.76840591430664,
      "kl": 5.265625,
      "learning_rate": 1.121593291404612e-07,
      "loss": 0.4484,
      "reward": 0.6099373698234558,
      "reward_std": 0.20125500112771988,
      "reward_std_all": 0.20030120015144348,
      "rewards-max/forensics_sinkhorn_reward": 0.9889025986194611,
      "rewards-max/forensics_strict_edge_reward": 0.9890114367008209,
      "rewards-max/forensics_tiou_reward": 0.999653548002243,
      "rewards/forensics_sinkhorn_reward": 0.5221367329359055,
      "rewards/forensics_strict_edge_reward": 0.31037137657403946,
      "rewards/forensics_tiou_reward": 0.38395586609840393,
      "step": 1695
    },
    {
      "completion_length": 21.6875,
      "epoch": 3.54869109947644,
      "grad_norm": 18.7154483795166,
      "kl": 6.46875,
      "learning_rate": 1.1163522012578616e-07,
      "loss": 0.3378,
      "reward": 0.5001566410064697,
      "reward_std": 0.21783814579248428,
      "reward_std_all": 0.2556108981370926,
      "rewards-max/forensics_sinkhorn_reward": 0.9881778657436371,
      "rewards-max/forensics_strict_edge_reward": 0.946017861366272,
      "rewards-max/forensics_tiou_reward": 0.9967990219593048,
      "rewards/forensics_sinkhorn_reward": 0.46265409886837006,
      "rewards/forensics_strict_edge_reward": 0.3190998435020447,
      "rewards/forensics_tiou_reward": 0.37084856629371643,
      "step": 1696
    },
    {
      "completion_length": 22.15625,
      "epoch": 3.5507853403141363,
      "grad_norm": 21.906312942504883,
      "kl": 6.46875,
      "learning_rate": 1.111111111111111e-07,
      "loss": 0.5986,
      "reward": 0.5345268249511719,
      "reward_std": 0.24375098943710327,
      "reward_std_all": 0.2671353444457054,
      "rewards-max/forensics_sinkhorn_reward": 0.98738893866539,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9936891794204712,
      "rewards/forensics_sinkhorn_reward": 0.4438133090734482,
      "rewards/forensics_strict_edge_reward": 0.2874017655849457,
      "rewards/forensics_tiou_reward": 0.3186919391155243,
      "step": 1697
    },
    {
      "completion_length": 34.078125,
      "epoch": 3.5528795811518323,
      "grad_norm": 64.31542205810547,
      "kl": 10.59375,
      "learning_rate": 1.1058700209643605e-07,
      "loss": 0.6124,
      "reward": 0.5272939205169678,
      "reward_std": 0.2150404378771782,
      "reward_std_all": 0.251049667596817,
      "rewards-max/forensics_sinkhorn_reward": 0.9713550209999084,
      "rewards-max/forensics_strict_edge_reward": 0.8464850187301636,
      "rewards-max/forensics_tiou_reward": 0.9588558077812195,
      "rewards/forensics_sinkhorn_reward": 0.467030793428421,
      "rewards/forensics_strict_edge_reward": 0.21527115255594254,
      "rewards/forensics_tiou_reward": 0.3593771457672119,
      "step": 1698
    },
    {
      "completion_length": 23.578125,
      "epoch": 3.554973821989529,
      "grad_norm": 20.55716323852539,
      "kl": 7.703125,
      "learning_rate": 1.10062893081761e-07,
      "loss": 0.4884,
      "reward": 0.5227232277393341,
      "reward_std": 0.1895875483751297,
      "reward_std_all": 0.22399678081274033,
      "rewards-max/forensics_sinkhorn_reward": 0.9867525100708008,
      "rewards-max/forensics_strict_edge_reward": 0.9357379674911499,
      "rewards-max/forensics_tiou_reward": 0.9911848604679108,
      "rewards/forensics_sinkhorn_reward": 0.43638888001441956,
      "rewards/forensics_strict_edge_reward": 0.2875514030456543,
      "rewards/forensics_tiou_reward": 0.34902241826057434,
      "step": 1699
    },
    {
      "completion_length": 41.46875,
      "epoch": 3.5570680628272253,
      "grad_norm": 18.063676834106445,
      "kl": 9.4375,
      "learning_rate": 1.0953878406708595e-07,
      "loss": 0.621,
      "reward": 0.5216300189495087,
      "reward_std": 0.20365433022379875,
      "reward_std_all": 0.2366236373782158,
      "rewards-max/forensics_sinkhorn_reward": 0.9315408170223236,
      "rewards-max/forensics_strict_edge_reward": 0.8469520509243011,
      "rewards-max/forensics_tiou_reward": 0.8709174990653992,
      "rewards/forensics_sinkhorn_reward": 0.39967484772205353,
      "rewards/forensics_strict_edge_reward": 0.2125648409128189,
      "rewards/forensics_tiou_reward": 0.2647849917411804,
      "step": 1700
    },
    {
      "completion_length": 42.375,
      "epoch": 3.5591623036649214,
      "grad_norm": 17.513792037963867,
      "kl": 7.59375,
      "learning_rate": 1.0901467505241089e-07,
      "loss": 0.4912,
      "reward": 0.5341638624668121,
      "reward_std": 0.2181660532951355,
      "reward_std_all": 0.24171382188796997,
      "rewards-max/forensics_sinkhorn_reward": 0.98573237657547,
      "rewards-max/forensics_strict_edge_reward": 0.9357379674911499,
      "rewards-max/forensics_tiou_reward": 0.9872074723243713,
      "rewards/forensics_sinkhorn_reward": 0.47722117602825165,
      "rewards/forensics_strict_edge_reward": 0.28187161684036255,
      "rewards/forensics_tiou_reward": 0.33539025485515594,
      "step": 1701
    },
    {
      "completion_length": 22.734375,
      "epoch": 3.561256544502618,
      "grad_norm": 18.267711639404297,
      "kl": 3.953125,
      "learning_rate": 1.0849056603773585e-07,
      "loss": 0.325,
      "reward": 0.5687775015830994,
      "reward_std": 0.19390273839235306,
      "reward_std_all": 0.21692921221256256,
      "rewards-max/forensics_sinkhorn_reward": 0.9878958463668823,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9956902265548706,
      "rewards/forensics_sinkhorn_reward": 0.4708753526210785,
      "rewards/forensics_strict_edge_reward": 0.2533906251192093,
      "rewards/forensics_tiou_reward": 0.2888592332601547,
      "step": 1702
    },
    {
      "completion_length": 34.703125,
      "epoch": 3.5633507853403144,
      "grad_norm": 23.63783836364746,
      "kl": 7.75,
      "learning_rate": 1.0796645702306078e-07,
      "loss": 0.5662,
      "reward": 0.4397996813058853,
      "reward_std": 0.180124431848526,
      "reward_std_all": 0.2607031613588333,
      "rewards-max/forensics_sinkhorn_reward": 0.9878676235675812,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9955784976482391,
      "rewards/forensics_sinkhorn_reward": 0.3696416914463043,
      "rewards/forensics_strict_edge_reward": 0.2470456287264824,
      "rewards/forensics_tiou_reward": 0.2840932011604309,
      "step": 1703
    },
    {
      "completion_length": 20.828125,
      "epoch": 3.5654450261780104,
      "grad_norm": 35.77230453491211,
      "kl": 8.078125,
      "learning_rate": 1.0744234800838574e-07,
      "loss": 0.5062,
      "reward": 0.541754275560379,
      "reward_std": 0.1718888357281685,
      "reward_std_all": 0.21448712050914764,
      "rewards-max/forensics_sinkhorn_reward": 0.9493160545825958,
      "rewards-max/forensics_strict_edge_reward": 0.771587461233139,
      "rewards-max/forensics_tiou_reward": 0.9174241721630096,
      "rewards/forensics_sinkhorn_reward": 0.43714170157909393,
      "rewards/forensics_strict_edge_reward": 0.2239060401916504,
      "rewards/forensics_tiou_reward": 0.2719572111964226,
      "step": 1704
    },
    {
      "completion_length": 21.5625,
      "epoch": 3.567539267015707,
      "grad_norm": 23.844345092773438,
      "kl": 4.625,
      "learning_rate": 1.0691823899371069e-07,
      "loss": 0.3791,
      "reward": 0.6546887457370758,
      "reward_std": 0.191397562623024,
      "reward_std_all": 0.17915398627519608,
      "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": 0.5717667043209076,
      "rewards/forensics_strict_edge_reward": 0.3163120895624161,
      "rewards/forensics_tiou_reward": 0.42471177875995636,
      "step": 1705
    },
    {
      "completion_length": 32.578125,
      "epoch": 3.5696335078534034,
      "grad_norm": 37709.74609375,
      "kl": 199.0,
      "learning_rate": 1.0639412997903563e-07,
      "loss": 8.2558,
      "reward": 0.5513167083263397,
      "reward_std": 0.18335233628749847,
      "reward_std_all": 0.21908578276634216,
      "rewards-max/forensics_sinkhorn_reward": 0.987560510635376,
      "rewards-max/forensics_strict_edge_reward": 0.9574736058712006,
      "rewards-max/forensics_tiou_reward": 0.9943798184394836,
      "rewards/forensics_sinkhorn_reward": 0.46098601818084717,
      "rewards/forensics_strict_edge_reward": 0.33482758700847626,
      "rewards/forensics_tiou_reward": 0.36059650778770447,
      "step": 1706
    },
    {
      "completion_length": 25.03125,
      "epoch": 3.5717277486910994,
      "grad_norm": 17.337108612060547,
      "kl": 5.734375,
      "learning_rate": 1.0587002096436058e-07,
      "loss": 0.4352,
      "reward": 0.5856369733810425,
      "reward_std": 0.2500511258840561,
      "reward_std_all": 0.24748355150222778,
      "rewards-max/forensics_sinkhorn_reward": 0.9887502789497375,
      "rewards-max/forensics_strict_edge_reward": 0.9890114367008209,
      "rewards-max/forensics_tiou_reward": 0.9990534782409668,
      "rewards/forensics_sinkhorn_reward": 0.5228730738162994,
      "rewards/forensics_strict_edge_reward": 0.3405735194683075,
      "rewards/forensics_tiou_reward": 0.3720491975545883,
      "step": 1707
    },
    {
      "completion_length": 17.375,
      "epoch": 3.573821989528796,
      "grad_norm": 20.990249633789062,
      "kl": 2.14453125,
      "learning_rate": 1.0534591194968554e-07,
      "loss": 0.1354,
      "reward": 0.6572076678276062,
      "reward_std": 0.15406998246908188,
      "reward_std_all": 0.15293224155902863,
      "rewards-max/forensics_sinkhorn_reward": 0.9883616268634796,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9975225031375885,
      "rewards/forensics_sinkhorn_reward": 0.5961431562900543,
      "rewards/forensics_strict_edge_reward": 0.3669711500406265,
      "rewards/forensics_tiou_reward": 0.4048211872577667,
      "step": 1708
    },
    {
      "completion_length": 31.09375,
      "epoch": 3.5759162303664924,
      "grad_norm": 16.358861923217773,
      "kl": 7.390625,
      "learning_rate": 1.0482180293501047e-07,
      "loss": 0.5436,
      "reward": 0.5492926836013794,
      "reward_std": 0.25649892538785934,
      "reward_std_all": 0.2619926556944847,
      "rewards-max/forensics_sinkhorn_reward": 0.9866172969341278,
      "rewards-max/forensics_strict_edge_reward": 0.9256840348243713,
      "rewards-max/forensics_tiou_reward": 0.990655243396759,
      "rewards/forensics_sinkhorn_reward": 0.48775653541088104,
      "rewards/forensics_strict_edge_reward": 0.2935449630022049,
      "rewards/forensics_tiou_reward": 0.34763601422309875,
      "step": 1709
    },
    {
      "completion_length": 29.078125,
      "epoch": 3.5780104712041885,
      "grad_norm": 23.36945915222168,
      "kl": 5.859375,
      "learning_rate": 1.0429769392033543e-07,
      "loss": 0.5247,
      "reward": 0.5381369590759277,
      "reward_std": 0.2774025425314903,
      "reward_std_all": 0.2695305868983269,
      "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": 0.4760727733373642,
      "rewards/forensics_strict_edge_reward": 0.2953171283006668,
      "rewards/forensics_tiou_reward": 0.3500438779592514,
      "step": 1710
    },
    {
      "completion_length": 17.390625,
      "epoch": 3.580104712041885,
      "grad_norm": 27.29941177368164,
      "kl": 3.9765625,
      "learning_rate": 1.0377358490566037e-07,
      "loss": 0.1923,
      "reward": 0.6857210099697113,
      "reward_std": 0.15310407802462578,
      "reward_std_all": 0.14219174906611443,
      "rewards-max/forensics_sinkhorn_reward": 0.988380640745163,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9975980520248413,
      "rewards/forensics_sinkhorn_reward": 0.66180020570755,
      "rewards/forensics_strict_edge_reward": 0.45647425949573517,
      "rewards/forensics_tiou_reward": 0.532352551817894,
      "step": 1711
    },
    {
      "completion_length": 28.453125,
      "epoch": 3.582198952879581,
      "grad_norm": 25.009416580200195,
      "kl": 4.6640625,
      "learning_rate": 1.0324947589098532e-07,
      "loss": 0.3977,
      "reward": 0.5224288702011108,
      "reward_std": 0.25301267206668854,
      "reward_std_all": 0.2689223513007164,
      "rewards-max/forensics_sinkhorn_reward": 0.9861707389354706,
      "rewards-max/forensics_strict_edge_reward": 0.947419673204422,
      "rewards-max/forensics_tiou_reward": 0.9889506101608276,
      "rewards/forensics_sinkhorn_reward": 0.4710509926080704,
      "rewards/forensics_strict_edge_reward": 0.2627083510160446,
      "rewards/forensics_tiou_reward": 0.33111873269081116,
      "step": 1712
    },
    {
      "completion_length": 20.78125,
      "epoch": 3.5842931937172775,
      "grad_norm": 373.6114196777344,
      "kl": 8.921875,
      "learning_rate": 1.0272536687631026e-07,
      "loss": 0.5414,
      "reward": 0.5975425541400909,
      "reward_std": 0.20395013689994812,
      "reward_std_all": 0.21246837079524994,
      "rewards-max/forensics_sinkhorn_reward": 0.9872018992900848,
      "rewards-max/forensics_strict_edge_reward": 0.9565287232398987,
      "rewards-max/forensics_tiou_reward": 0.9929635524749756,
      "rewards/forensics_sinkhorn_reward": 0.4982435703277588,
      "rewards/forensics_strict_edge_reward": 0.27782224863767624,
      "rewards/forensics_tiou_reward": 0.32482584565877914,
      "step": 1713
    },
    {
      "completion_length": 36.640625,
      "epoch": 3.5863874345549736,
      "grad_norm": 27.0257511138916,
      "kl": 5.484375,
      "learning_rate": 1.0220125786163522e-07,
      "loss": 0.4563,
      "reward": 0.595398485660553,
      "reward_std": 0.15759418159723282,
      "reward_std_all": 0.1778511255979538,
      "rewards-max/forensics_sinkhorn_reward": 0.9882758855819702,
      "rewards-max/forensics_strict_edge_reward": 0.9782643616199493,
      "rewards-max/forensics_tiou_reward": 0.997183084487915,
      "rewards/forensics_sinkhorn_reward": 0.6117993295192719,
      "rewards/forensics_strict_edge_reward": 0.39874178171157837,
      "rewards/forensics_tiou_reward": 0.5140480697154999,
      "step": 1714
    },
    {
      "completion_length": 37.796875,
      "epoch": 3.58848167539267,
      "grad_norm": 25.196462631225586,
      "kl": 6.359375,
      "learning_rate": 1.0167714884696015e-07,
      "loss": 0.4656,
      "reward": 0.5556438267230988,
      "reward_std": 0.25992774963378906,
      "reward_std_all": 0.250394307076931,
      "rewards-max/forensics_sinkhorn_reward": 0.9841480553150177,
      "rewards-max/forensics_strict_edge_reward": 0.87538942694664,
      "rewards-max/forensics_tiou_reward": 0.9867304563522339,
      "rewards/forensics_sinkhorn_reward": 0.4948889762163162,
      "rewards/forensics_strict_edge_reward": 0.27398718893527985,
      "rewards/forensics_tiou_reward": 0.3560629338026047,
      "step": 1715
    },
    {
      "completion_length": 19.140625,
      "epoch": 3.5905759162303665,
      "grad_norm": 28.093734741210938,
      "kl": 5.5,
      "learning_rate": 1.0115303983228511e-07,
      "loss": 0.2994,
      "reward": 0.5546252429485321,
      "reward_std": 0.20006077736616135,
      "reward_std_all": 0.2251049131155014,
      "rewards-max/forensics_sinkhorn_reward": 0.9864151477813721,
      "rewards-max/forensics_strict_edge_reward": 0.9252271056175232,
      "rewards-max/forensics_tiou_reward": 0.9898813664913177,
      "rewards/forensics_sinkhorn_reward": 0.47847485542297363,
      "rewards/forensics_strict_edge_reward": 0.18651564419269562,
      "rewards/forensics_tiou_reward": 0.2317415103316307,
      "step": 1716
    },
    {
      "completion_length": 29.3125,
      "epoch": 3.5926701570680626,
      "grad_norm": 20.1336727142334,
      "kl": 5.5625,
      "learning_rate": 1.0062893081761006e-07,
      "loss": 0.3906,
      "reward": 0.5605760514736176,
      "reward_std": 0.2502291053533554,
      "reward_std_all": 0.25367170572280884,
      "rewards-max/forensics_sinkhorn_reward": 0.987165242433548,
      "rewards-max/forensics_strict_edge_reward": 0.9357379674911499,
      "rewards-max/forensics_tiou_reward": 0.9928119480609894,
      "rewards/forensics_sinkhorn_reward": 0.4494790881872177,
      "rewards/forensics_strict_edge_reward": 0.24104857444763184,
      "rewards/forensics_tiou_reward": 0.2922085076570511,
      "step": 1717
    },
    {
      "completion_length": 30.6875,
      "epoch": 3.594764397905759,
      "grad_norm": 21.273508071899414,
      "kl": 8.78125,
      "learning_rate": 1.00104821802935e-07,
      "loss": 0.6851,
      "reward": 0.4530925899744034,
      "reward_std": 0.15746742486953735,
      "reward_std_all": 0.24601663649082184,
      "rewards-max/forensics_sinkhorn_reward": 0.9867600798606873,
      "rewards-max/forensics_strict_edge_reward": 0.9565287232398987,
      "rewards-max/forensics_tiou_reward": 0.9912087917327881,
      "rewards/forensics_sinkhorn_reward": 0.37655192613601685,
      "rewards/forensics_strict_edge_reward": 0.24666163325309753,
      "rewards/forensics_tiou_reward": 0.26338373124599457,
      "step": 1718
    },
    {
      "completion_length": 37.140625,
      "epoch": 3.5968586387434556,
      "grad_norm": 27.610504150390625,
      "kl": 7.71875,
      "learning_rate": 9.958071278825995e-08,
      "loss": 0.4343,
      "reward": 0.5278341770172119,
      "reward_std": 0.25508594512939453,
      "reward_std_all": 0.2598431929945946,
      "rewards-max/forensics_sinkhorn_reward": 0.9775598049163818,
      "rewards-max/forensics_strict_edge_reward": 0.9464850425720215,
      "rewards-max/forensics_tiou_reward": 0.9744256436824799,
      "rewards/forensics_sinkhorn_reward": 0.4461568295955658,
      "rewards/forensics_strict_edge_reward": 0.29703274369239807,
      "rewards/forensics_tiou_reward": 0.31919676065444946,
      "step": 1719
    },
    {
      "completion_length": 34.5625,
      "epoch": 3.5989528795811516,
      "grad_norm": 23.189861297607422,
      "kl": 8.546875,
      "learning_rate": 9.905660377358491e-08,
      "loss": 0.6797,
      "reward": 0.5356512665748596,
      "reward_std": 0.16816465556621552,
      "reward_std_all": 0.22529639303684235,
      "rewards-max/forensics_sinkhorn_reward": 0.9871087372303009,
      "rewards-max/forensics_strict_edge_reward": 0.9565287232398987,
      "rewards-max/forensics_tiou_reward": 0.9925973117351532,
      "rewards/forensics_sinkhorn_reward": 0.5033323019742966,
      "rewards/forensics_strict_edge_reward": 0.3370363563299179,
      "rewards/forensics_tiou_reward": 0.3850948214530945,
      "step": 1720
    },
    {
      "completion_length": 42.203125,
      "epoch": 3.601047120418848,
      "grad_norm": 103.75092315673828,
      "kl": 9.265625,
      "learning_rate": 9.853249475890984e-08,
      "loss": 0.5363,
      "reward": 0.5952848494052887,
      "reward_std": 0.2009451612830162,
      "reward_std_all": 0.20930754393339157,
      "rewards-max/forensics_sinkhorn_reward": 0.9881324172019958,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9966206848621368,
      "rewards/forensics_sinkhorn_reward": 0.517509713768959,
      "rewards/forensics_strict_edge_reward": 0.3119743764400482,
      "rewards/forensics_tiou_reward": 0.3866370618343353,
      "step": 1721
    },
    {
      "completion_length": 27.53125,
      "epoch": 3.6031413612565446,
      "grad_norm": 22.339860916137695,
      "kl": 6.1875,
      "learning_rate": 9.80083857442348e-08,
      "loss": 0.4114,
      "reward": 0.5898831188678741,
      "reward_std": 0.21299558877944946,
      "reward_std_all": 0.22378955781459808,
      "rewards-max/forensics_sinkhorn_reward": 0.9369302093982697,
      "rewards-max/forensics_strict_edge_reward": 0.8659839034080505,
      "rewards-max/forensics_tiou_reward": 0.8895665407180786,
      "rewards/forensics_sinkhorn_reward": 0.49688534438610077,
      "rewards/forensics_strict_edge_reward": 0.24575982987880707,
      "rewards/forensics_tiou_reward": 0.3234163075685501,
      "step": 1722
    },
    {
      "completion_length": 19.953125,
      "epoch": 3.6052356020942407,
      "grad_norm": 40.73135757446289,
      "kl": 4.765625,
      "learning_rate": 9.748427672955974e-08,
      "loss": 0.1698,
      "reward": 0.538361132144928,
      "reward_std": 0.23153268545866013,
      "reward_std_all": 0.2556409388780594,
      "rewards-max/forensics_sinkhorn_reward": 0.9428891539573669,
      "rewards-max/forensics_strict_edge_reward": 0.9228302836418152,
      "rewards-max/forensics_tiou_reward": 0.9062298536300659,
      "rewards/forensics_sinkhorn_reward": 0.43953005969524384,
      "rewards/forensics_strict_edge_reward": 0.22598521411418915,
      "rewards/forensics_tiou_reward": 0.23452119529247284,
      "step": 1723
    },
    {
      "completion_length": 24.3125,
      "epoch": 3.607329842931937,
      "grad_norm": 26.232044219970703,
      "kl": 8.78125,
      "learning_rate": 9.696016771488469e-08,
      "loss": 0.5784,
      "reward": 0.4947018474340439,
      "reward_std": 0.18143506348133087,
      "reward_std_all": 0.23846060037612915,
      "rewards-max/forensics_sinkhorn_reward": 0.9589651226997375,
      "rewards-max/forensics_strict_edge_reward": 0.9151731729507446,
      "rewards-max/forensics_tiou_reward": 0.9351638853549957,
      "rewards/forensics_sinkhorn_reward": 0.3963058590888977,
      "rewards/forensics_strict_edge_reward": 0.1968492791056633,
      "rewards/forensics_tiou_reward": 0.23671156913042068,
      "step": 1724
    },
    {
      "completion_length": 20.1875,
      "epoch": 3.6094240837696336,
      "grad_norm": 22.91903305053711,
      "kl": 5.6875,
      "learning_rate": 9.643605870020964e-08,
      "loss": 0.4066,
      "reward": 0.5588570684194565,
      "reward_std": 0.19131533801555634,
      "reward_std_all": 0.21518050134181976,
      "rewards-max/forensics_sinkhorn_reward": 0.9827891290187836,
      "rewards-max/forensics_strict_edge_reward": 0.895060271024704,
      "rewards-max/forensics_tiou_reward": 0.9841136336326599,
      "rewards/forensics_sinkhorn_reward": 0.449104443192482,
      "rewards/forensics_strict_edge_reward": 0.20974916964769363,
      "rewards/forensics_tiou_reward": 0.3078465014696121,
      "step": 1725
    },
    {
      "completion_length": 19.921875,
      "epoch": 3.6115183246073297,
      "grad_norm": 28.093700408935547,
      "kl": 4.109375,
      "learning_rate": 9.59119496855346e-08,
      "loss": 0.336,
      "reward": 0.5837653577327728,
      "reward_std": 0.21044595539569855,
      "reward_std_all": 0.2250095009803772,
      "rewards-max/forensics_sinkhorn_reward": 0.9878159165382385,
      "rewards-max/forensics_strict_edge_reward": 0.9677534997463226,
      "rewards-max/forensics_tiou_reward": 0.995381087064743,
      "rewards/forensics_sinkhorn_reward": 0.505012571811676,
      "rewards/forensics_strict_edge_reward": 0.2983078211545944,
      "rewards/forensics_tiou_reward": 0.385626882314682,
      "step": 1726
    },
    {
      "completion_length": 34.0,
      "epoch": 3.613612565445026,
      "grad_norm": 21.65562629699707,
      "kl": 5.046875,
      "learning_rate": 9.538784067085953e-08,
      "loss": 0.4781,
      "reward": 0.5985127091407776,
      "reward_std": 0.18965209275484085,
      "reward_std_all": 0.20320876687765121,
      "rewards-max/forensics_sinkhorn_reward": 0.988130509853363,
      "rewards-max/forensics_strict_edge_reward": 0.9574736058712006,
      "rewards-max/forensics_tiou_reward": 0.996610164642334,
      "rewards/forensics_sinkhorn_reward": 0.5474107265472412,
      "rewards/forensics_strict_edge_reward": 0.30273111164569855,
      "rewards/forensics_tiou_reward": 0.3994111716747284,
      "step": 1727
    },
    {
      "completion_length": 18.90625,
      "epoch": 3.6157068062827227,
      "grad_norm": 22.565052032470703,
      "kl": 6.046875,
      "learning_rate": 9.486373165618448e-08,
      "loss": 0.355,
      "reward": 0.5881697535514832,
      "reward_std": 0.22852341830730438,
      "reward_std_all": 0.23317450284957886,
      "rewards-max/forensics_sinkhorn_reward": 0.9889025986194611,
      "rewards-max/forensics_strict_edge_reward": 0.9890114367008209,
      "rewards-max/forensics_tiou_reward": 0.999653548002243,
      "rewards/forensics_sinkhorn_reward": 0.5136913359165192,
      "rewards/forensics_strict_edge_reward": 0.2554330974817276,
      "rewards/forensics_tiou_reward": 0.357363685965538,
      "step": 1728
    },
    {
      "completion_length": 27.28125,
      "epoch": 3.6178010471204187,
      "grad_norm": 22.719051361083984,
      "kl": 7.296875,
      "learning_rate": 9.433962264150943e-08,
      "loss": 0.5044,
      "reward": 0.5564228296279907,
      "reward_std": 0.20072776079177856,
      "reward_std_all": 0.23036281019449234,
      "rewards-max/forensics_sinkhorn_reward": 0.9882354736328125,
      "rewards-max/forensics_strict_edge_reward": 0.9574736058712006,
      "rewards-max/forensics_tiou_reward": 0.9970348179340363,
      "rewards/forensics_sinkhorn_reward": 0.4531928449869156,
      "rewards/forensics_strict_edge_reward": 0.3219892233610153,
      "rewards/forensics_tiou_reward": 0.34770452976226807,
      "step": 1729
    },
    {
      "completion_length": 37.375,
      "epoch": 3.619895287958115,
      "grad_norm": 18.286396026611328,
      "kl": 5.625,
      "learning_rate": 9.381551362683438e-08,
      "loss": 0.3913,
      "reward": 0.6000283062458038,
      "reward_std": 0.24155502766370773,
      "reward_std_all": 0.23155277967453003,
      "rewards-max/forensics_sinkhorn_reward": 0.9354244470596313,
      "rewards-max/forensics_strict_edge_reward": 0.8663162291049957,
      "rewards-max/forensics_tiou_reward": 0.8853345811367035,
      "rewards/forensics_sinkhorn_reward": 0.5026236027479172,
      "rewards/forensics_strict_edge_reward": 0.2986768037080765,
      "rewards/forensics_tiou_reward": 0.34819337725639343,
      "step": 1730
    },
    {
      "completion_length": 23.53125,
      "epoch": 3.6219895287958117,
      "grad_norm": 21.356895446777344,
      "kl": 5.265625,
      "learning_rate": 9.329140461215932e-08,
      "loss": 0.2546,
      "reward": 0.6158246994018555,
      "reward_std": 0.13702701777219772,
      "reward_std_all": 0.1548403650522232,
      "rewards-max/forensics_sinkhorn_reward": 0.9711302816867828,
      "rewards-max/forensics_strict_edge_reward": 0.9151731729507446,
      "rewards-max/forensics_tiou_reward": 0.9597443640232086,
      "rewards/forensics_sinkhorn_reward": 0.5359412729740143,
      "rewards/forensics_strict_edge_reward": 0.2860470712184906,
      "rewards/forensics_tiou_reward": 0.36865483969449997,
      "step": 1731
    },
    {
      "completion_length": 25.21875,
      "epoch": 3.6240837696335078,
      "grad_norm": 21.94268035888672,
      "kl": 3.7890625,
      "learning_rate": 9.276729559748427e-08,
      "loss": 0.3078,
      "reward": 0.61634361743927,
      "reward_std": 0.188568115234375,
      "reward_std_all": 0.18753398954868317,
      "rewards-max/forensics_sinkhorn_reward": 0.988667756319046,
      "rewards-max/forensics_strict_edge_reward": 0.9890114367008209,
      "rewards-max/forensics_tiou_reward": 0.9987285435199738,
      "rewards/forensics_sinkhorn_reward": 0.5955592691898346,
      "rewards/forensics_strict_edge_reward": 0.39361824095249176,
      "rewards/forensics_tiou_reward": 0.4602361172437668,
      "step": 1732
    },
    {
      "completion_length": 21.171875,
      "epoch": 3.6261780104712042,
      "grad_norm": 43.80987548828125,
      "kl": 4.75,
      "learning_rate": 9.224318658280921e-08,
      "loss": 0.2765,
      "reward": 0.6424001455307007,
      "reward_std": 0.17166107892990112,
      "reward_std_all": 0.17152433842420578,
      "rewards-max/forensics_sinkhorn_reward": 0.9512152373790741,
      "rewards-max/forensics_strict_edge_reward": 0.8144323527812958,
      "rewards-max/forensics_tiou_reward": 0.9126836061477661,
      "rewards/forensics_sinkhorn_reward": 0.48495641350746155,
      "rewards/forensics_strict_edge_reward": 0.2067200019955635,
      "rewards/forensics_tiou_reward": 0.29533782601356506,
      "step": 1733
    },
    {
      "completion_length": 26.3125,
      "epoch": 3.6282722513089007,
      "grad_norm": 25.290315628051758,
      "kl": 3.8515625,
      "learning_rate": 9.171907756813417e-08,
      "loss": 0.2537,
      "reward": 0.6494320034980774,
      "reward_std": 0.11858486570417881,
      "reward_std_all": 0.13288768753409386,
      "rewards-max/forensics_sinkhorn_reward": 0.9639429748058319,
      "rewards-max/forensics_strict_edge_reward": 0.9208705723285675,
      "rewards-max/forensics_tiou_reward": 0.9427303373813629,
      "rewards/forensics_sinkhorn_reward": 0.582745224237442,
      "rewards/forensics_strict_edge_reward": 0.30757587403059006,
      "rewards/forensics_tiou_reward": 0.39084582030773163,
      "step": 1734
    },
    {
      "completion_length": 46.953125,
      "epoch": 3.630366492146597,
      "grad_norm": 23.879030227661133,
      "kl": 7.578125,
      "learning_rate": 9.11949685534591e-08,
      "loss": 0.5035,
      "reward": 0.5566798150539398,
      "reward_std": 0.1933821365237236,
      "reward_std_all": 0.2258262112736702,
      "rewards-max/forensics_sinkhorn_reward": 0.9719939827919006,
      "rewards-max/forensics_strict_edge_reward": 0.8876297473907471,
      "rewards-max/forensics_tiou_reward": 0.9603660702705383,
      "rewards/forensics_sinkhorn_reward": 0.44799038767814636,
      "rewards/forensics_strict_edge_reward": 0.2529499679803848,
      "rewards/forensics_tiou_reward": 0.3166920840740204,
      "step": 1735
    },
    {
      "completion_length": 26.234375,
      "epoch": 3.6324607329842933,
      "grad_norm": 29.264436721801758,
      "kl": 5.46875,
      "learning_rate": 9.067085953878406e-08,
      "loss": 0.3639,
      "reward": 0.5637800693511963,
      "reward_std": 0.20461224019527435,
      "reward_std_all": 0.22632043063640594,
      "rewards-max/forensics_sinkhorn_reward": 0.988244354724884,
      "rewards-max/forensics_strict_edge_reward": 0.9782643616199493,
      "rewards-max/forensics_tiou_reward": 0.9970588088035583,
      "rewards/forensics_sinkhorn_reward": 0.5105387568473816,
      "rewards/forensics_strict_edge_reward": 0.3321128338575363,
      "rewards/forensics_tiou_reward": 0.4240506589412689,
      "step": 1736
    },
    {
      "completion_length": 20.484375,
      "epoch": 3.6345549738219898,
      "grad_norm": 24.90482521057129,
      "kl": 5.359375,
      "learning_rate": 9.014675052410901e-08,
      "loss": 0.3323,
      "reward": 0.5903975963592529,
      "reward_std": 0.14862951636314392,
      "reward_std_all": 0.18167869746685028,
      "rewards-max/forensics_sinkhorn_reward": 0.9848531484603882,
      "rewards-max/forensics_strict_edge_reward": 0.946017861366272,
      "rewards-max/forensics_tiou_reward": 0.9895603954792023,
      "rewards/forensics_sinkhorn_reward": 0.4788040965795517,
      "rewards/forensics_strict_edge_reward": 0.27418845146894455,
      "rewards/forensics_tiou_reward": 0.3028121441602707,
      "step": 1737
    },
    {
      "completion_length": 24.90625,
      "epoch": 3.636649214659686,
      "grad_norm": 25.491058349609375,
      "kl": 7.3125,
      "learning_rate": 8.962264150943395e-08,
      "loss": 0.3786,
      "reward": 0.577189177274704,
      "reward_std": 0.241899773478508,
      "reward_std_all": 0.2329416498541832,
      "rewards-max/forensics_sinkhorn_reward": 0.9866408705711365,
      "rewards-max/forensics_strict_edge_reward": 0.947419673204422,
      "rewards-max/forensics_tiou_reward": 0.9907836019992828,
      "rewards/forensics_sinkhorn_reward": 0.48081856966018677,
      "rewards/forensics_strict_edge_reward": 0.28052498400211334,
      "rewards/forensics_tiou_reward": 0.3699968457221985,
      "step": 1738
    },
    {
      "completion_length": 26.34375,
      "epoch": 3.6387434554973823,
      "grad_norm": 21.030319213867188,
      "kl": 4.5390625,
      "learning_rate": 8.90985324947589e-08,
      "loss": 0.2925,
      "reward": 0.635608434677124,
      "reward_std": 0.2080601304769516,
      "reward_std_all": 0.19317525625228882,
      "rewards-max/forensics_sinkhorn_reward": 0.962211936712265,
      "rewards-max/forensics_strict_edge_reward": 0.895060271024704,
      "rewards-max/forensics_tiou_reward": 0.937301516532898,
      "rewards/forensics_sinkhorn_reward": 0.544820636510849,
      "rewards/forensics_strict_edge_reward": 0.2848494350910187,
      "rewards/forensics_tiou_reward": 0.39214491844177246,
      "step": 1739
    },
    {
      "completion_length": 36.96875,
      "epoch": 3.640837696335079,
      "grad_norm": 21.459630966186523,
      "kl": 7.640625,
      "learning_rate": 8.857442348008386e-08,
      "loss": 0.5694,
      "reward": 0.5492558181285858,
      "reward_std": 0.22573326528072357,
      "reward_std_all": 0.24500712752342224,
      "rewards-max/forensics_sinkhorn_reward": 0.9704427719116211,
      "rewards-max/forensics_strict_edge_reward": 0.8286563754081726,
      "rewards-max/forensics_tiou_reward": 0.9561555087566376,
      "rewards/forensics_sinkhorn_reward": 0.47259825468063354,
      "rewards/forensics_strict_edge_reward": 0.3038403019309044,
      "rewards/forensics_tiou_reward": 0.334986574947834,
      "step": 1740
    },
    {
      "completion_length": 19.1875,
      "epoch": 3.642931937172775,
      "grad_norm": 22.37051010131836,
      "kl": 6.21875,
      "learning_rate": 8.805031446540879e-08,
      "loss": 0.3991,
      "reward": 0.659743458032608,
      "reward_std": 0.15230296552181244,
      "reward_std_all": 0.14759700000286102,
      "rewards-max/forensics_sinkhorn_reward": 0.9871049523353577,
      "rewards-max/forensics_strict_edge_reward": 0.947419673204422,
      "rewards-max/forensics_tiou_reward": 0.9925966262817383,
      "rewards/forensics_sinkhorn_reward": 0.5857542753219604,
      "rewards/forensics_strict_edge_reward": 0.3587659150362015,
      "rewards/forensics_tiou_reward": 0.4248303025960922,
      "step": 1741
    },
    {
      "completion_length": 18.640625,
      "epoch": 3.6450261780104714,
      "grad_norm": 28.105125427246094,
      "kl": 4.453125,
      "learning_rate": 8.752620545073375e-08,
      "loss": 0.2615,
      "reward": 0.5970031023025513,
      "reward_std": 0.19185570627450943,
      "reward_std_all": 0.20181499421596527,
      "rewards-max/forensics_sinkhorn_reward": 0.9883732497692108,
      "rewards-max/forensics_strict_edge_reward": 0.9782643616199493,
      "rewards-max/forensics_tiou_reward": 0.9975698590278625,
      "rewards/forensics_sinkhorn_reward": 0.5928076803684235,
      "rewards/forensics_strict_edge_reward": 0.4105461984872818,
      "rewards/forensics_tiou_reward": 0.5006733387708664,
      "step": 1742
    },
    {
      "completion_length": 21.640625,
      "epoch": 3.6471204188481674,
      "grad_norm": 20.62841796875,
      "kl": 5.203125,
      "learning_rate": 8.70020964360587e-08,
      "loss": 0.4412,
      "reward": 0.5776120126247406,
      "reward_std": 0.24085114896297455,
      "reward_std_all": 0.2356591671705246,
      "rewards-max/forensics_sinkhorn_reward": 0.988304853439331,
      "rewards-max/forensics_strict_edge_reward": 0.9782643616199493,
      "rewards-max/forensics_tiou_reward": 0.9972972869873047,
      "rewards/forensics_sinkhorn_reward": 0.4943259358406067,
      "rewards/forensics_strict_edge_reward": 0.32062795758247375,
      "rewards/forensics_tiou_reward": 0.3471035957336426,
      "step": 1743
    },
    {
      "completion_length": 21.0625,
      "epoch": 3.649214659685864,
      "grad_norm": 30.07959747314453,
      "kl": 4.1953125,
      "learning_rate": 8.647798742138364e-08,
      "loss": 0.1781,
      "reward": 0.6136917769908905,
      "reward_std": 0.21800623834133148,
      "reward_std_all": 0.20930102467536926,
      "rewards-max/forensics_sinkhorn_reward": 0.9882354736328125,
      "rewards-max/forensics_strict_edge_reward": 0.9890114367008209,
      "rewards-max/forensics_tiou_reward": 0.9970542788505554,
      "rewards/forensics_sinkhorn_reward": 0.5738796889781952,
      "rewards/forensics_strict_edge_reward": 0.3064079135656357,
      "rewards/forensics_tiou_reward": 0.41539081931114197,
      "step": 1744
    },
    {
      "completion_length": 30.046875,
      "epoch": 3.65130890052356,
      "grad_norm": 20.621173858642578,
      "kl": 4.7734375,
      "learning_rate": 8.595387840670858e-08,
      "loss": 0.2099,
      "reward": 0.6155524849891663,
      "reward_std": 0.2233266904950142,
      "reward_std_all": 0.210187129676342,
      "rewards-max/forensics_sinkhorn_reward": 0.9876276552677155,
      "rewards-max/forensics_strict_edge_reward": 0.9567649364471436,
      "rewards-max/forensics_tiou_reward": 0.9946280419826508,
      "rewards/forensics_sinkhorn_reward": 0.5236673057079315,
      "rewards/forensics_strict_edge_reward": 0.2820591628551483,
      "rewards/forensics_tiou_reward": 0.3528927117586136,
      "step": 1745
    },
    {
      "completion_length": 25.078125,
      "epoch": 3.6534031413612564,
      "grad_norm": 20.1473331451416,
      "kl": 6.796875,
      "learning_rate": 8.542976939203354e-08,
      "loss": 0.4916,
      "reward": 0.5904960632324219,
      "reward_std": 0.20175301656126976,
      "reward_std_all": 0.21146802604198456,
      "rewards-max/forensics_sinkhorn_reward": 0.9888220727443695,
      "rewards-max/forensics_strict_edge_reward": 0.9782643616199493,
      "rewards-max/forensics_tiou_reward": 0.9993362128734589,
      "rewards/forensics_sinkhorn_reward": 0.5109134465456009,
      "rewards/forensics_strict_edge_reward": 0.3407670110464096,
      "rewards/forensics_tiou_reward": 0.37960463762283325,
      "step": 1746
    },
    {
      "completion_length": 29.640625,
      "epoch": 3.655497382198953,
      "grad_norm": 53.57568359375,
      "kl": 7.609375,
      "learning_rate": 8.490566037735848e-08,
      "loss": 0.5609,
      "reward": 0.6170370578765869,
      "reward_std": 0.230433389544487,
      "reward_std_all": 0.21817928552627563,
      "rewards-max/forensics_sinkhorn_reward": 0.9851446747779846,
      "rewards-max/forensics_strict_edge_reward": 0.9185642302036285,
      "rewards-max/forensics_tiou_reward": 0.9849632680416107,
      "rewards/forensics_sinkhorn_reward": 0.5289536118507385,
      "rewards/forensics_strict_edge_reward": 0.2917696535587311,
      "rewards/forensics_tiou_reward": 0.348163366317749,
      "step": 1747
    },
    {
      "completion_length": 26.015625,
      "epoch": 3.657591623036649,
      "grad_norm": 15.871936798095703,
      "kl": 4.0703125,
      "learning_rate": 8.438155136268343e-08,
      "loss": 0.2815,
      "reward": 0.6559686958789825,
      "reward_std": 0.1413043588399887,
      "reward_std_all": 0.15340978652238846,
      "rewards-max/forensics_sinkhorn_reward": 0.9854800701141357,
      "rewards-max/forensics_strict_edge_reward": 0.9169811904430389,
      "rewards-max/forensics_tiou_reward": 0.9862417876720428,
      "rewards/forensics_sinkhorn_reward": 0.5789285600185394,
      "rewards/forensics_strict_edge_reward": 0.3730996698141098,
      "rewards/forensics_tiou_reward": 0.4413737505674362,
      "step": 1748
    },
    {
      "completion_length": 21.796875,
      "epoch": 3.6596858638743455,
      "grad_norm": 21.753583908081055,
      "kl": 5.6640625,
      "learning_rate": 8.385744234800839e-08,
      "loss": 0.3897,
      "reward": 0.5613342225551605,
      "reward_std": 0.21257218718528748,
      "reward_std_all": 0.22897201776504517,
      "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": 0.5196954011917114,
      "rewards/forensics_strict_edge_reward": 0.3985339105129242,
      "rewards/forensics_tiou_reward": 0.4111773669719696,
      "step": 1749
    },
    {
      "completion_length": 43.25,
      "epoch": 3.661780104712042,
      "grad_norm": 27.04751205444336,
      "kl": 7.90625,
      "learning_rate": 8.333333333333333e-08,
      "loss": 0.5227,
      "reward": 0.5251158773899078,
      "reward_std": 0.2705310508608818,
      "reward_std_all": 0.27469947189092636,
      "rewards-max/forensics_sinkhorn_reward": 0.9562138020992279,
      "rewards-max/forensics_strict_edge_reward": 0.7889257967472076,
      "rewards-max/forensics_tiou_reward": 0.927093893289566,
      "rewards/forensics_sinkhorn_reward": 0.4477214515209198,
      "rewards/forensics_strict_edge_reward": 0.22039448469877243,
      "rewards/forensics_tiou_reward": 0.3120448887348175,
      "step": 1750
    },
    {
      "completion_length": 26.5,
      "epoch": 3.663874345549738,
      "grad_norm": 21.10282325744629,
      "kl": 4.2421875,
      "learning_rate": 8.280922431865828e-08,
      "loss": 0.2351,
      "reward": 0.6685277223587036,
      "reward_std": 0.14691969379782677,
      "reward_std_all": 0.14322879165410995,
      "rewards-max/forensics_sinkhorn_reward": 0.9885415136814117,
      "rewards-max/forensics_strict_edge_reward": 0.9890114367008209,
      "rewards-max/forensics_tiou_reward": 0.9982300996780396,
      "rewards/forensics_sinkhorn_reward": 0.6124399304389954,
      "rewards/forensics_strict_edge_reward": 0.40583181381225586,
      "rewards/forensics_tiou_reward": 0.49112503230571747,
      "step": 1751
    },
    {
      "completion_length": 42.21875,
      "epoch": 3.6659685863874345,
      "grad_norm": 19.232261657714844,
      "kl": 6.609375,
      "learning_rate": 8.228511530398323e-08,
      "loss": 0.4248,
      "reward": 0.5145001709461212,
      "reward_std": 0.22511166334152222,
      "reward_std_all": 0.26079992949962616,
      "rewards-max/forensics_sinkhorn_reward": 0.9877969920635223,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9953016042709351,
      "rewards/forensics_sinkhorn_reward": 0.47614723443984985,
      "rewards/forensics_strict_edge_reward": 0.36939048767089844,
      "rewards/forensics_tiou_reward": 0.3946707397699356,
      "step": 1752
    },
    {
      "completion_length": 21.984375,
      "epoch": 3.668062827225131,
      "grad_norm": 17.561960220336914,
      "kl": 4.609375,
      "learning_rate": 8.176100628930816e-08,
      "loss": 0.3959,
      "reward": 0.6223605573177338,
      "reward_std": 0.20921602100133896,
      "reward_std_all": 0.19950637221336365,
      "rewards-max/forensics_sinkhorn_reward": 0.9884152412414551,
      "rewards-max/forensics_strict_edge_reward": 0.9782643616199493,
      "rewards-max/forensics_tiou_reward": 0.9977349936962128,
      "rewards/forensics_sinkhorn_reward": 0.5763274431228638,
      "rewards/forensics_strict_edge_reward": 0.4288497716188431,
      "rewards/forensics_tiou_reward": 0.4673706740140915,
      "step": 1753
    },
    {
      "completion_length": 23.015625,
      "epoch": 3.670157068062827,
      "grad_norm": 22.885744094848633,
      "kl": 6.734375,
      "learning_rate": 8.123689727463312e-08,
      "loss": 0.553,
      "reward": 0.6124846339225769,
      "reward_std": 0.2185903638601303,
      "reward_std_all": 0.20615434646606445,
      "rewards-max/forensics_sinkhorn_reward": 0.988304853439331,
      "rewards-max/forensics_strict_edge_reward": 0.9782643616199493,
      "rewards-max/forensics_tiou_reward": 0.9972972869873047,
      "rewards/forensics_sinkhorn_reward": 0.5943664610385895,
      "rewards/forensics_strict_edge_reward": 0.37130922079086304,
      "rewards/forensics_tiou_reward": 0.48791128396987915,
      "step": 1754
    },
    {
      "completion_length": 23.671875,
      "epoch": 3.6722513089005235,
      "grad_norm": 25.490610122680664,
      "kl": 7.203125,
      "learning_rate": 8.071278825995808e-08,
      "loss": 0.5707,
      "reward": 0.5424854159355164,
      "reward_std": 0.22083129733800888,
      "reward_std_all": 0.24452796578407288,
      "rewards-max/forensics_sinkhorn_reward": 0.988667756319046,
      "rewards-max/forensics_strict_edge_reward": 0.9890114367008209,
      "rewards-max/forensics_tiou_reward": 0.9987285435199738,
      "rewards/forensics_sinkhorn_reward": 0.508337676525116,
      "rewards/forensics_strict_edge_reward": 0.36117081344127655,
      "rewards/forensics_tiou_reward": 0.39591944217681885,
      "step": 1755
    },
    {
      "completion_length": 25.6875,
      "epoch": 3.67434554973822,
      "grad_norm": 34.29875946044922,
      "kl": 8.96875,
      "learning_rate": 8.018867924528301e-08,
      "loss": 0.6055,
      "reward": 0.5271000117063522,
      "reward_std": 0.2229510247707367,
      "reward_std_all": 0.25425102561712265,
      "rewards-max/forensics_sinkhorn_reward": 0.9714537560939789,
      "rewards-max/forensics_strict_edge_reward": 0.946017861366272,
      "rewards-max/forensics_tiou_reward": 0.961248517036438,
      "rewards/forensics_sinkhorn_reward": 0.4314439743757248,
      "rewards/forensics_strict_edge_reward": 0.27454956620931625,
      "rewards/forensics_tiou_reward": 0.32928767800331116,
      "step": 1756
    },
    {
      "completion_length": 32.4375,
      "epoch": 3.676439790575916,
      "grad_norm": 23.981403350830078,
      "kl": 7.65625,
      "learning_rate": 7.966457023060797e-08,
      "loss": 0.6447,
      "reward": 0.53092460334301,
      "reward_std": 0.22873899340629578,
      "reward_std_all": 0.2590811848640442,
      "rewards-max/forensics_sinkhorn_reward": 0.9876793026924133,
      "rewards-max/forensics_strict_edge_reward": 0.9574736058712006,
      "rewards-max/forensics_tiou_reward": 0.9948453903198242,
      "rewards/forensics_sinkhorn_reward": 0.4012404680252075,
      "rewards/forensics_strict_edge_reward": 0.2567555755376816,
      "rewards/forensics_tiou_reward": 0.29234015196561813,
      "step": 1757
    },
    {
      "completion_length": 32.359375,
      "epoch": 3.6785340314136126,
      "grad_norm": 22.468942642211914,
      "kl": 9.421875,
      "learning_rate": 7.914046121593292e-08,
      "loss": 0.7571,
      "reward": 0.42668241262435913,
      "reward_std": 0.21928876638412476,
      "reward_std_all": 0.29144658148288727,
      "rewards-max/forensics_sinkhorn_reward": 0.980424553155899,
      "rewards-max/forensics_strict_edge_reward": 0.9464850425720215,
      "rewards-max/forensics_tiou_reward": 0.9800094962120056,
      "rewards/forensics_sinkhorn_reward": 0.35312533378601074,
      "rewards/forensics_strict_edge_reward": 0.2131391391158104,
      "rewards/forensics_tiou_reward": 0.24131463468074799,
      "step": 1758
    },
    {
      "completion_length": 28.34375,
      "epoch": 3.680628272251309,
      "grad_norm": 18.991758346557617,
      "kl": 6.90625,
      "learning_rate": 7.861635220125786e-08,
      "loss": 0.4017,
      "reward": 0.584882527589798,
      "reward_std": 0.19377953559160233,
      "reward_std_all": 0.1959124356508255,
      "rewards-max/forensics_sinkhorn_reward": 0.964068591594696,
      "rewards-max/forensics_strict_edge_reward": 0.8760378360748291,
      "rewards-max/forensics_tiou_reward": 0.9416686296463013,
      "rewards/forensics_sinkhorn_reward": 0.48510994017124176,
      "rewards/forensics_strict_edge_reward": 0.27336616814136505,
      "rewards/forensics_tiou_reward": 0.31995536386966705,
      "step": 1759
    },
    {
      "completion_length": 20.90625,
      "epoch": 3.682722513089005,
      "grad_norm": 25.991065979003906,
      "kl": 4.53125,
      "learning_rate": 7.80922431865828e-08,
      "loss": 0.4875,
      "reward": 0.55789715051651,
      "reward_std": 0.21902921795845032,
      "reward_std_all": 0.23276035487651825,
      "rewards-max/forensics_sinkhorn_reward": 0.9229670763015747,
      "rewards-max/forensics_strict_edge_reward": 0.8404377698898315,
      "rewards-max/forensics_tiou_reward": 0.8587411344051361,
      "rewards/forensics_sinkhorn_reward": 0.5168235599994659,
      "rewards/forensics_strict_edge_reward": 0.3493465334177017,
      "rewards/forensics_tiou_reward": 0.3876699209213257,
      "step": 1760
    },
    {
      "completion_length": 43.734375,
      "epoch": 3.6848167539267016,
      "grad_norm": 31.035499572753906,
      "kl": 8.921875,
      "learning_rate": 7.756813417190777e-08,
      "loss": 0.4762,
      "reward": 0.49368272721767426,
      "reward_std": 0.24217084050178528,
      "reward_std_all": 0.27600499987602234,
      "rewards-max/forensics_sinkhorn_reward": 0.9887329638004303,
      "rewards-max/forensics_strict_edge_reward": 0.9890114367008209,
      "rewards-max/forensics_tiou_reward": 0.9989847838878632,
      "rewards/forensics_sinkhorn_reward": 0.42597126960754395,
      "rewards/forensics_strict_edge_reward": 0.2504790872335434,
      "rewards/forensics_tiou_reward": 0.3090815991163254,
      "step": 1761
    },
    {
      "completion_length": 28.484375,
      "epoch": 3.686910994764398,
      "grad_norm": 23.608884811401367,
      "kl": 6.4609375,
      "learning_rate": 7.70440251572327e-08,
      "loss": 0.4058,
      "reward": 0.5226945132017136,
      "reward_std": 0.28148752450942993,
      "reward_std_all": 0.27918781340122223,
      "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": 0.4187559187412262,
      "rewards/forensics_strict_edge_reward": 0.27896735072135925,
      "rewards/forensics_tiou_reward": 0.2967821806669235,
      "step": 1762
    },
    {
      "completion_length": 20.125,
      "epoch": 3.689005235602094,
      "grad_norm": 20.052181243896484,
      "kl": 6.359375,
      "learning_rate": 7.651991614255766e-08,
      "loss": 0.3832,
      "reward": 0.5950315892696381,
      "reward_std": 0.20016871392726898,
      "reward_std_all": 0.20212514698505402,
      "rewards-max/forensics_sinkhorn_reward": 0.9860047698020935,
      "rewards-max/forensics_strict_edge_reward": 0.9890114367008209,
      "rewards-max/forensics_tiou_reward": 0.9916351139545441,
      "rewards/forensics_sinkhorn_reward": 0.5260817259550095,
      "rewards/forensics_strict_edge_reward": 0.2988099008798599,
      "rewards/forensics_tiou_reward": 0.35143841803073883,
      "step": 1763
    },
    {
      "completion_length": 44.921875,
      "epoch": 3.6910994764397906,
      "grad_norm": 20.62824249267578,
      "kl": 10.28125,
      "learning_rate": 7.59958071278826e-08,
      "loss": 0.7544,
      "reward": 0.4194677025079727,
      "reward_std": 0.24188415706157684,
      "reward_std_all": 0.3132585287094116,
      "rewards-max/forensics_sinkhorn_reward": 0.9875489175319672,
      "rewards-max/forensics_strict_edge_reward": 0.9677534997463226,
      "rewards-max/forensics_tiou_reward": 0.9943343102931976,
      "rewards/forensics_sinkhorn_reward": 0.3763253837823868,
      "rewards/forensics_strict_edge_reward": 0.19424916431307793,
      "rewards/forensics_tiou_reward": 0.2798178642988205,
      "step": 1764
    },
    {
      "completion_length": 31.1875,
      "epoch": 3.693193717277487,
      "grad_norm": 25.041858673095703,
      "kl": 5.703125,
      "learning_rate": 7.547169811320755e-08,
      "loss": 0.4724,
      "reward": 0.5583369433879852,
      "reward_std": 0.22918828576803207,
      "reward_std_all": 0.2422003298997879,
      "rewards-max/forensics_sinkhorn_reward": 0.986704409122467,
      "rewards-max/forensics_strict_edge_reward": 0.9574736058712006,
      "rewards-max/forensics_tiou_reward": 0.9910315573215485,
      "rewards/forensics_sinkhorn_reward": 0.4870437979698181,
      "rewards/forensics_strict_edge_reward": 0.3213995546102524,
      "rewards/forensics_tiou_reward": 0.33291755616664886,
      "step": 1765
    },
    {
      "completion_length": 27.25,
      "epoch": 3.695287958115183,
      "grad_norm": 20.28083038330078,
      "kl": 7.0625,
      "learning_rate": 7.494758909853249e-08,
      "loss": 0.4832,
      "reward": 0.6255361139774323,
      "reward_std": 0.19472557306289673,
      "reward_std_all": 0.19226212799549103,
      "rewards-max/forensics_sinkhorn_reward": 0.9683472514152527,
      "rewards-max/forensics_strict_edge_reward": 0.9677534997463226,
      "rewards-max/forensics_tiou_reward": 0.9566953480243683,
      "rewards/forensics_sinkhorn_reward": 0.5606129467487335,
      "rewards/forensics_strict_edge_reward": 0.3773125261068344,
      "rewards/forensics_tiou_reward": 0.4182412922382355,
      "step": 1766
    },
    {
      "completion_length": 19.0625,
      "epoch": 3.6973821989528797,
      "grad_norm": 32.83574676513672,
      "kl": 6.71875,
      "learning_rate": 7.442348008385745e-08,
      "loss": 0.4514,
      "reward": 0.5906563401222229,
      "reward_std": 0.14233703166246414,
      "reward_std_all": 0.18341855704784393,
      "rewards-max/forensics_sinkhorn_reward": 0.9883511364459991,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.997479647397995,
      "rewards/forensics_sinkhorn_reward": 0.5399577021598816,
      "rewards/forensics_strict_edge_reward": 0.31980976462364197,
      "rewards/forensics_tiou_reward": 0.40222832560539246,
      "step": 1767
    },
    {
      "completion_length": 37.671875,
      "epoch": 3.699476439790576,
      "grad_norm": 24.724451065063477,
      "kl": 6.046875,
      "learning_rate": 7.389937106918238e-08,
      "loss": 0.5088,
      "reward": 0.5530036091804504,
      "reward_std": 0.27429842948913574,
      "reward_std_all": 0.2639140337705612,
      "rewards-max/forensics_sinkhorn_reward": 0.9875489175319672,
      "rewards-max/forensics_strict_edge_reward": 0.9677534997463226,
      "rewards-max/forensics_tiou_reward": 0.9943343102931976,
      "rewards/forensics_sinkhorn_reward": 0.45137055218219757,
      "rewards/forensics_strict_edge_reward": 0.3226606845855713,
      "rewards/forensics_tiou_reward": 0.35925404727458954,
      "step": 1768
    },
    {
      "completion_length": 22.71875,
      "epoch": 3.701570680628272,
      "grad_norm": 30.341184616088867,
      "kl": 7.984375,
      "learning_rate": 7.337526205450734e-08,
      "loss": 0.4769,
      "reward": 0.5207183659076691,
      "reward_std": 0.25423943996429443,
      "reward_std_all": 0.27092939615249634,
      "rewards-max/forensics_sinkhorn_reward": 0.9880331456661224,
      "rewards-max/forensics_strict_edge_reward": 0.9782643616199493,
      "rewards-max/forensics_tiou_reward": 0.99622642993927,
      "rewards/forensics_sinkhorn_reward": 0.48211225867271423,
      "rewards/forensics_strict_edge_reward": 0.2607472836971283,
      "rewards/forensics_tiou_reward": 0.3538648933172226,
      "step": 1769
    },
    {
      "completion_length": 20.03125,
      "epoch": 3.7036649214659687,
      "grad_norm": 18.764162063598633,
      "kl": 6.46875,
      "learning_rate": 7.285115303983227e-08,
      "loss": 0.3131,
      "reward": 0.5831568539142609,
      "reward_std": 0.21115900576114655,
      "reward_std_all": 0.21997766196727753,
      "rewards-max/forensics_sinkhorn_reward": 0.9392624795436859,
      "rewards-max/forensics_strict_edge_reward": 0.7680098712444305,
      "rewards-max/forensics_tiou_reward": 0.8908001184463501,
      "rewards/forensics_sinkhorn_reward": 0.47908806800842285,
      "rewards/forensics_strict_edge_reward": 0.21062061935663223,
      "rewards/forensics_tiou_reward": 0.25756654143333435,
      "step": 1770
    },
    {
      "completion_length": 25.25,
      "epoch": 3.7057591623036648,
      "grad_norm": 21.500831604003906,
      "kl": 6.28125,
      "learning_rate": 7.232704402515723e-08,
      "loss": 0.4088,
      "reward": 0.5874623954296112,
      "reward_std": 0.16273797303438187,
      "reward_std_all": 0.18297288566827774,
      "rewards-max/forensics_sinkhorn_reward": 0.9880160391330719,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9961601197719574,
      "rewards/forensics_sinkhorn_reward": 0.5017149448394775,
      "rewards/forensics_strict_edge_reward": 0.30062349140644073,
      "rewards/forensics_tiou_reward": 0.34112922847270966,
      "step": 1771
    },
    {
      "completion_length": 33.609375,
      "epoch": 3.7078534031413612,
      "grad_norm": 25.1392822265625,
      "kl": 5.890625,
      "learning_rate": 7.180293501048218e-08,
      "loss": 0.3424,
      "reward": 0.5849984884262085,
      "reward_std": 0.16756881028413773,
      "reward_std_all": 0.195146806538105,
      "rewards-max/forensics_sinkhorn_reward": 0.9878387749195099,
      "rewards-max/forensics_strict_edge_reward": 0.9890114367008209,
      "rewards-max/forensics_tiou_reward": 0.9954707324504852,
      "rewards/forensics_sinkhorn_reward": 0.4751870334148407,
      "rewards/forensics_strict_edge_reward": 0.2761406898498535,
      "rewards/forensics_tiou_reward": 0.3366349935531616,
      "step": 1772
    },
    {
      "completion_length": 28.421875,
      "epoch": 3.7099476439790577,
      "grad_norm": 19.475723266601562,
      "kl": 4.359375,
      "learning_rate": 7.127882599580712e-08,
      "loss": 0.2995,
      "reward": 0.6177034378051758,
      "reward_std": 0.1992356777191162,
      "reward_std_all": 0.206417478621006,
      "rewards-max/forensics_sinkhorn_reward": 0.9817823469638824,
      "rewards-max/forensics_strict_edge_reward": 0.9158510267734528,
      "rewards-max/forensics_tiou_reward": 0.981356680393219,
      "rewards/forensics_sinkhorn_reward": 0.5429576635360718,
      "rewards/forensics_strict_edge_reward": 0.26331160962581635,
      "rewards/forensics_tiou_reward": 0.38608071208000183,
      "step": 1773
    },
    {
      "completion_length": 39.234375,
      "epoch": 3.712041884816754,
      "grad_norm": 20.641921997070312,
      "kl": 4.9375,
      "learning_rate": 7.075471698113207e-08,
      "loss": 0.3399,
      "reward": 0.5562272667884827,
      "reward_std": 0.1386469006538391,
      "reward_std_all": 0.19126781821250916,
      "rewards-max/forensics_sinkhorn_reward": 0.9774099290370941,
      "rewards-max/forensics_strict_edge_reward": 0.8805811703205109,
      "rewards-max/forensics_tiou_reward": 0.9746624529361725,
      "rewards/forensics_sinkhorn_reward": 0.5124103724956512,
      "rewards/forensics_strict_edge_reward": 0.23394392430782318,
      "rewards/forensics_tiou_reward": 0.34299759566783905,
      "step": 1774
    },
    {
      "completion_length": 24.3125,
      "epoch": 3.7141361256544503,
      "grad_norm": 22.486207962036133,
      "kl": 4.6171875,
      "learning_rate": 7.023060796645703e-08,
      "loss": 0.3083,
      "reward": 0.5473546087741852,
      "reward_std": 0.16397309303283691,
      "reward_std_all": 0.1975022554397583,
      "rewards-max/forensics_sinkhorn_reward": 0.9886867105960846,
      "rewards-max/forensics_strict_edge_reward": 0.9890114367008209,
      "rewards-max/forensics_tiou_reward": 0.9988023936748505,
      "rewards/forensics_sinkhorn_reward": 0.5080623626708984,
      "rewards/forensics_strict_edge_reward": 0.2911505252122879,
      "rewards/forensics_tiou_reward": 0.36893779039382935,
      "step": 1775
    },
    {
      "completion_length": 21.234375,
      "epoch": 3.7162303664921463,
      "grad_norm": 23.60224723815918,
      "kl": 5.265625,
      "learning_rate": 6.970649895178196e-08,
      "loss": 0.4537,
      "reward": 0.6437169909477234,
      "reward_std": 0.11369097232818604,
      "reward_std_all": 0.14613644778728485,
      "rewards-max/forensics_sinkhorn_reward": 0.9877418279647827,
      "rewards-max/forensics_strict_edge_reward": 0.9565287232398987,
      "rewards-max/forensics_tiou_reward": 0.9950778186321259,
      "rewards/forensics_sinkhorn_reward": 0.6280157268047333,
      "rewards/forensics_strict_edge_reward": 0.45836885273456573,
      "rewards/forensics_tiou_reward": 0.5183259546756744,
      "step": 1776
    },
    {
      "completion_length": 25.71875,
      "epoch": 3.718324607329843,
      "grad_norm": 29.369937896728516,
      "kl": 6.09375,
      "learning_rate": 6.918238993710692e-08,
      "loss": 0.3589,
      "reward": 0.5196272432804108,
      "reward_std": 0.25065896660089493,
      "reward_std_all": 0.275155670940876,
      "rewards-max/forensics_sinkhorn_reward": 0.9851446747779846,
      "rewards-max/forensics_strict_edge_reward": 0.9093653857707977,
      "rewards-max/forensics_tiou_reward": 0.9849632680416107,
      "rewards/forensics_sinkhorn_reward": 0.43198058009147644,
      "rewards/forensics_strict_edge_reward": 0.28707778453826904,
      "rewards/forensics_tiou_reward": 0.3270390033721924,
      "step": 1777
    },
    {
      "completion_length": 25.59375,
      "epoch": 3.7204188481675393,
      "grad_norm": 18.05295181274414,
      "kl": 5.078125,
      "learning_rate": 6.865828092243186e-08,
      "loss": 0.4182,
      "reward": 0.618443101644516,
      "reward_std": 0.16983914747834206,
      "reward_std_all": 0.18729405850172043,
      "rewards-max/forensics_sinkhorn_reward": 0.9868248105049133,
      "rewards-max/forensics_strict_edge_reward": 0.9364311099052429,
      "rewards-max/forensics_tiou_reward": 0.9914863705635071,
      "rewards/forensics_sinkhorn_reward": 0.5715588629245758,
      "rewards/forensics_strict_edge_reward": 0.3559187948703766,
      "rewards/forensics_tiou_reward": 0.46660324931144714,
      "step": 1778
    },
    {
      "completion_length": 25.5,
      "epoch": 3.7225130890052354,
      "grad_norm": 22.802894592285156,
      "kl": 3.4453125,
      "learning_rate": 6.813417190775681e-08,
      "loss": 0.2846,
      "reward": 0.6011505722999573,
      "reward_std": 0.19871524721384048,
      "reward_std_all": 0.20937379449605942,
      "rewards-max/forensics_sinkhorn_reward": 0.937270998954773,
      "rewards-max/forensics_strict_edge_reward": 0.8241921961307526,
      "rewards-max/forensics_tiou_reward": 0.8839805722236633,
      "rewards/forensics_sinkhorn_reward": 0.5220827162265778,
      "rewards/forensics_strict_edge_reward": 0.2961608022451401,
      "rewards/forensics_tiou_reward": 0.3687156066298485,
      "step": 1779
    },
    {
      "completion_length": 20.234375,
      "epoch": 3.724607329842932,
      "grad_norm": 21.40165901184082,
      "kl": 5.6328125,
      "learning_rate": 6.761006289308176e-08,
      "loss": 0.3506,
      "reward": 0.5504064857959747,
      "reward_std": 0.27789634466171265,
      "reward_std_all": 0.26687904447317123,
      "rewards-max/forensics_sinkhorn_reward": 0.9879094958305359,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9957433342933655,
      "rewards/forensics_sinkhorn_reward": 0.4955558180809021,
      "rewards/forensics_strict_edge_reward": 0.2808454930782318,
      "rewards/forensics_tiou_reward": 0.34082406759262085,
      "step": 1780
    },
    {
      "completion_length": 21.40625,
      "epoch": 3.7267015706806284,
      "grad_norm": 24.017107009887695,
      "kl": 4.2890625,
      "learning_rate": 6.708595387840671e-08,
      "loss": 0.2374,
      "reward": 0.6185569167137146,
      "reward_std": 0.2027641236782074,
      "reward_std_all": 0.19534802436828613,
      "rewards-max/forensics_sinkhorn_reward": 0.9882987141609192,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9972739219665527,
      "rewards/forensics_sinkhorn_reward": 0.535985141992569,
      "rewards/forensics_strict_edge_reward": 0.31638260185718536,
      "rewards/forensics_tiou_reward": 0.39009563624858856,
      "step": 1781
    },
    {
      "completion_length": 21.921875,
      "epoch": 3.7287958115183244,
      "grad_norm": 23.121225357055664,
      "kl": 5.953125,
      "learning_rate": 6.656184486373165e-08,
      "loss": 0.502,
      "reward": 0.5713642537593842,
      "reward_std": 0.1725659966468811,
      "reward_std_all": 0.19902333617210388,
      "rewards-max/forensics_sinkhorn_reward": 0.9878166019916534,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9953798651695251,
      "rewards/forensics_sinkhorn_reward": 0.5095853507518768,
      "rewards/forensics_strict_edge_reward": 0.2765713632106781,
      "rewards/forensics_tiou_reward": 0.3398419916629791,
      "step": 1782
    },
    {
      "completion_length": 20.734375,
      "epoch": 3.730890052356021,
      "grad_norm": 22.274049758911133,
      "kl": 3.328125,
      "learning_rate": 6.60377358490566e-08,
      "loss": 0.1555,
      "reward": 0.6688251197338104,
      "reward_std": 0.1781110242009163,
      "reward_std_all": 0.16395726799964905,
      "rewards-max/forensics_sinkhorn_reward": 0.9682352840900421,
      "rewards-max/forensics_strict_edge_reward": 0.878633052110672,
      "rewards-max/forensics_tiou_reward": 0.9496214389801025,
      "rewards/forensics_sinkhorn_reward": 0.5380000919103622,
      "rewards/forensics_strict_edge_reward": 0.24243207275867462,
      "rewards/forensics_tiou_reward": 0.312200702726841,
      "step": 1783
    },
    {
      "completion_length": 30.765625,
      "epoch": 3.7329842931937174,
      "grad_norm": 15.862608909606934,
      "kl": 5.796875,
      "learning_rate": 6.551362683438155e-08,
      "loss": 0.4453,
      "reward": 0.5347908139228821,
      "reward_std": 0.19520803540945053,
      "reward_std_all": 0.22876673936843872,
      "rewards-max/forensics_sinkhorn_reward": 0.8160980343818665,
      "rewards-max/forensics_strict_edge_reward": 0.6862104088068008,
      "rewards-max/forensics_tiou_reward": 0.7706710994243622,
      "rewards/forensics_sinkhorn_reward": 0.4313426911830902,
      "rewards/forensics_strict_edge_reward": 0.20354986190795898,
      "rewards/forensics_tiou_reward": 0.2735586017370224,
      "step": 1784
    },
    {
      "completion_length": 23.140625,
      "epoch": 3.7350785340314134,
      "grad_norm": 22.026634216308594,
      "kl": 7.21875,
      "learning_rate": 6.49895178197065e-08,
      "loss": 0.564,
      "reward": 0.5621809661388397,
      "reward_std": 0.1912170797586441,
      "reward_std_all": 0.20902152359485626,
      "rewards-max/forensics_sinkhorn_reward": 0.9435729682445526,
      "rewards-max/forensics_strict_edge_reward": 0.8340697288513184,
      "rewards-max/forensics_tiou_reward": 0.9022110402584076,
      "rewards/forensics_sinkhorn_reward": 0.4156412333250046,
      "rewards/forensics_strict_edge_reward": 0.1757269650697708,
      "rewards/forensics_tiou_reward": 0.19494953751564026,
      "step": 1785
    },
    {
      "completion_length": 20.265625,
      "epoch": 3.73717277486911,
      "grad_norm": 19.61614227294922,
      "kl": 5.296875,
      "learning_rate": 6.446540880503144e-08,
      "loss": 0.4364,
      "reward": 0.5747633278369904,
      "reward_std": 0.1233869194984436,
      "reward_std_all": 0.1742476522922516,
      "rewards-max/forensics_sinkhorn_reward": 0.9875966012477875,
      "rewards-max/forensics_strict_edge_reward": 0.9565287232398987,
      "rewards-max/forensics_tiou_reward": 0.9945054948329926,
      "rewards/forensics_sinkhorn_reward": 0.4745819419622421,
      "rewards/forensics_strict_edge_reward": 0.21416616439819336,
      "rewards/forensics_tiou_reward": 0.2657540291547775,
      "step": 1786
    },
    {
      "completion_length": 26.046875,
      "epoch": 3.7392670157068064,
      "grad_norm": 29.934749603271484,
      "kl": 7.09375,
      "learning_rate": 6.39412997903564e-08,
      "loss": 0.4581,
      "reward": 0.5473044365644455,
      "reward_std": 0.21960388123989105,
      "reward_std_all": 0.23813223838806152,
      "rewards-max/forensics_sinkhorn_reward": 0.9862746596336365,
      "rewards-max/forensics_strict_edge_reward": 0.9252271056175232,
      "rewards-max/forensics_tiou_reward": 0.9893244504928589,
      "rewards/forensics_sinkhorn_reward": 0.5117336362600327,
      "rewards/forensics_strict_edge_reward": 0.2794208526611328,
      "rewards/forensics_tiou_reward": 0.37749066948890686,
      "step": 1787
    },
    {
      "completion_length": 20.71875,
      "epoch": 3.7413612565445025,
      "grad_norm": 16.816761016845703,
      "kl": 4.765625,
      "learning_rate": 6.341719077568133e-08,
      "loss": 0.2159,
      "reward": 0.5972591042518616,
      "reward_std": 0.21283162385225296,
      "reward_std_all": 0.20814298838377,
      "rewards-max/forensics_sinkhorn_reward": 0.9883071184158325,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9973088800907135,
      "rewards/forensics_sinkhorn_reward": 0.5386016666889191,
      "rewards/forensics_strict_edge_reward": 0.29572926461696625,
      "rewards/forensics_tiou_reward": 0.3707769513130188,
      "step": 1788
    },
    {
      "completion_length": 29.453125,
      "epoch": 3.743455497382199,
      "grad_norm": 22.781036376953125,
      "kl": 3.625,
      "learning_rate": 6.289308176100629e-08,
      "loss": 0.24,
      "reward": 0.5881798267364502,
      "reward_std": 0.25392942130565643,
      "reward_std_all": 0.24368174374103546,
      "rewards-max/forensics_sinkhorn_reward": 0.9180969595909119,
      "rewards-max/forensics_strict_edge_reward": 0.7963844537734985,
      "rewards-max/forensics_tiou_reward": 0.8429471850395203,
      "rewards/forensics_sinkhorn_reward": 0.46393679082393646,
      "rewards/forensics_strict_edge_reward": 0.18654385209083557,
      "rewards/forensics_tiou_reward": 0.26060159504413605,
      "step": 1789
    },
    {
      "completion_length": 25.640625,
      "epoch": 3.7455497382198955,
      "grad_norm": 17.220487594604492,
      "kl": 6.671875,
      "learning_rate": 6.236897274633124e-08,
      "loss": 0.4477,
      "reward": 0.579381912946701,
      "reward_std": 0.19933335483074188,
      "reward_std_all": 0.21947626769542694,
      "rewards-max/forensics_sinkhorn_reward": 0.9878827035427094,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9956355094909668,
      "rewards/forensics_sinkhorn_reward": 0.5295875668525696,
      "rewards/forensics_strict_edge_reward": 0.2884945720434189,
      "rewards/forensics_tiou_reward": 0.4148884415626526,
      "step": 1790
    },
    {
      "completion_length": 26.53125,
      "epoch": 3.7476439790575915,
      "grad_norm": 27.856863021850586,
      "kl": 6.359375,
      "learning_rate": 6.184486373165618e-08,
      "loss": 0.4279,
      "reward": 0.5496664643287659,
      "reward_std": 0.20933745801448822,
      "reward_std_all": 0.2382996827363968,
      "rewards-max/forensics_sinkhorn_reward": 0.9870903789997101,
      "rewards-max/forensics_strict_edge_reward": 0.946017861366272,
      "rewards-max/forensics_tiou_reward": 0.9925104975700378,
      "rewards/forensics_sinkhorn_reward": 0.4741565138101578,
      "rewards/forensics_strict_edge_reward": 0.23849504441022873,
      "rewards/forensics_tiou_reward": 0.32093454897403717,
      "step": 1791
    },
    {
      "completion_length": 17.96875,
      "epoch": 3.749738219895288,
      "grad_norm": 22.871850967407227,
      "kl": 3.4453125,
      "learning_rate": 6.132075471698113e-08,
      "loss": 0.2404,
      "reward": 0.6920868754386902,
      "reward_std": 0.1134088933467865,
      "reward_std_all": 0.11897475272417068,
      "rewards-max/forensics_sinkhorn_reward": 0.9875458478927612,
      "rewards-max/forensics_strict_edge_reward": 0.9565287232398987,
      "rewards-max/forensics_tiou_reward": 0.9943076968193054,
      "rewards/forensics_sinkhorn_reward": 0.5782342851161957,
      "rewards/forensics_strict_edge_reward": 0.3004691004753113,
      "rewards/forensics_tiou_reward": 0.37198685109615326,
      "step": 1792
    },
    {
      "completion_length": 20.453125,
      "epoch": 3.7518324607329845,
      "grad_norm": 23.326440811157227,
      "kl": 5.984375,
      "learning_rate": 6.079664570230607e-08,
      "loss": 0.3823,
      "reward": 0.5982887148857117,
      "reward_std": 0.22708884626626968,
      "reward_std_all": 0.2188577800989151,
      "rewards-max/forensics_sinkhorn_reward": 0.9876037538051605,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9945367574691772,
      "rewards/forensics_sinkhorn_reward": 0.5186588615179062,
      "rewards/forensics_strict_edge_reward": 0.3213690370321274,
      "rewards/forensics_tiou_reward": 0.37929366528987885,
      "step": 1793
    },
    {
      "completion_length": 18.40625,
      "epoch": 3.7539267015706805,
      "grad_norm": 25.077550888061523,
      "kl": 3.8828125,
      "learning_rate": 6.027253668763103e-08,
      "loss": 0.3335,
      "reward": 0.6292828023433685,
      "reward_std": 0.20964831113815308,
      "reward_std_all": 0.19582954794168472,
      "rewards-max/forensics_sinkhorn_reward": 0.9832165241241455,
      "rewards-max/forensics_strict_edge_reward": 0.7510668337345123,
      "rewards-max/forensics_tiou_reward": 0.9775182008743286,
      "rewards/forensics_sinkhorn_reward": 0.579787403345108,
      "rewards/forensics_strict_edge_reward": 0.258015975356102,
      "rewards/forensics_tiou_reward": 0.4425918757915497,
      "step": 1794
    },
    {
      "completion_length": 17.65625,
      "epoch": 3.756020942408377,
      "grad_norm": 19.26595115661621,
      "kl": 4.84375,
      "learning_rate": 5.974842767295598e-08,
      "loss": 0.3602,
      "reward": 0.6081814467906952,
      "reward_std": 0.13262731581926346,
      "reward_std_all": 0.16680748760700226,
      "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": 0.5854213237762451,
      "rewards/forensics_strict_edge_reward": 0.43384869396686554,
      "rewards/forensics_tiou_reward": 0.48690538108348846,
      "step": 1795
    },
    {
      "completion_length": 21.0,
      "epoch": 3.7581151832460735,
      "grad_norm": 19.21015739440918,
      "kl": 4.921875,
      "learning_rate": 5.9224318658280916e-08,
      "loss": 0.4151,
      "reward": 0.5945072770118713,
      "reward_std": 0.22074225544929504,
      "reward_std_all": 0.21493395417928696,
      "rewards-max/forensics_sinkhorn_reward": 0.988498866558075,
      "rewards-max/forensics_strict_edge_reward": 0.9890114367008209,
      "rewards-max/forensics_tiou_reward": 0.9980638921260834,
      "rewards/forensics_sinkhorn_reward": 0.5321060419082642,
      "rewards/forensics_strict_edge_reward": 0.33748549222946167,
      "rewards/forensics_tiou_reward": 0.4104557931423187,
      "step": 1796
    },
    {
      "completion_length": 20.28125,
      "epoch": 3.7602094240837696,
      "grad_norm": 18.913185119628906,
      "kl": 4.7890625,
      "learning_rate": 5.870020964360587e-08,
      "loss": 0.3283,
      "reward": 0.6684934496879578,
      "reward_std": 0.16052550077438354,
      "reward_std_all": 0.15425041317939758,
      "rewards-max/forensics_sinkhorn_reward": 0.9881954789161682,
      "rewards-max/forensics_strict_edge_reward": 0.9567649364471436,
      "rewards-max/forensics_tiou_reward": 0.9968660771846771,
      "rewards/forensics_sinkhorn_reward": 0.5916962027549744,
      "rewards/forensics_strict_edge_reward": 0.3902607411146164,
      "rewards/forensics_tiou_reward": 0.40718992054462433,
      "step": 1797
    },
    {
      "completion_length": 29.8125,
      "epoch": 3.762303664921466,
      "grad_norm": 17.418445587158203,
      "kl": 7.875,
      "learning_rate": 5.8176100628930814e-08,
      "loss": 0.5503,
      "reward": 0.5505430102348328,
      "reward_std": 0.2329903095960617,
      "reward_std_all": 0.24620169401168823,
      "rewards-max/forensics_sinkhorn_reward": 0.9798587560653687,
      "rewards-max/forensics_strict_edge_reward": 0.9469369649887085,
      "rewards-max/forensics_tiou_reward": 0.9785918295383453,
      "rewards/forensics_sinkhorn_reward": 0.4925382137298584,
      "rewards/forensics_strict_edge_reward": 0.32391001284122467,
      "rewards/forensics_tiou_reward": 0.41412757337093353,
      "step": 1798
    },
    {
      "completion_length": 24.5625,
      "epoch": 3.7643979057591626,
      "grad_norm": 18.780319213867188,
      "kl": 6.75,
      "learning_rate": 5.765199161425576e-08,
      "loss": 0.3879,
      "reward": 0.5722264647483826,
      "reward_std": 0.18538084626197815,
      "reward_std_all": 0.19816402345895767,
      "rewards-max/forensics_sinkhorn_reward": 0.9883574843406677,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9975068867206573,
      "rewards/forensics_sinkhorn_reward": 0.5034130811691284,
      "rewards/forensics_strict_edge_reward": 0.2668910622596741,
      "rewards/forensics_tiou_reward": 0.3349453806877136,
      "step": 1799
    },
    {
      "completion_length": 35.859375,
      "epoch": 3.7664921465968586,
      "grad_norm": 28.378263473510742,
      "kl": 9.015625,
      "learning_rate": 5.7127882599580705e-08,
      "loss": 0.5979,
      "reward": 0.5741127133369446,
      "reward_std": 0.18744755536317825,
      "reward_std_all": 0.2147923707962036,
      "rewards-max/forensics_sinkhorn_reward": 0.959796130657196,
      "rewards-max/forensics_strict_edge_reward": 0.8365300297737122,
      "rewards-max/forensics_tiou_reward": 0.934424877166748,
      "rewards/forensics_sinkhorn_reward": 0.5073524564504623,
      "rewards/forensics_strict_edge_reward": 0.24730289727449417,
      "rewards/forensics_tiou_reward": 0.3320920765399933,
      "step": 1800
    },
    {
      "completion_length": 27.328125,
      "epoch": 3.768586387434555,
      "grad_norm": 17.486082077026367,
      "kl": 5.59375,
      "learning_rate": 5.660377358490566e-08,
      "loss": 0.3299,
      "reward": 0.5707335770130157,
      "reward_std": 0.20971793681383133,
      "reward_std_all": 0.22164412587881088,
      "rewards-max/forensics_sinkhorn_reward": 0.9133136570453644,
      "rewards-max/forensics_strict_edge_reward": 0.8764696419239044,
      "rewards-max/forensics_tiou_reward": 0.8438248634338379,
      "rewards/forensics_sinkhorn_reward": 0.45113658905029297,
      "rewards/forensics_strict_edge_reward": 0.2595050632953644,
      "rewards/forensics_tiou_reward": 0.30729904025793076,
      "step": 1801
    },
    {
      "completion_length": 20.28125,
      "epoch": 3.770680628272251,
      "grad_norm": 41.358280181884766,
      "kl": 6.609375,
      "learning_rate": 5.60796645702306e-08,
      "loss": 0.5106,
      "reward": 0.5122940987348557,
      "reward_std": 0.23331766575574875,
      "reward_std_all": 0.2721925377845764,
      "rewards-max/forensics_sinkhorn_reward": 0.9267206192016602,
      "rewards-max/forensics_strict_edge_reward": 0.7218027263879776,
      "rewards-max/forensics_tiou_reward": 0.8671224415302277,
      "rewards/forensics_sinkhorn_reward": 0.42749086022377014,
      "rewards/forensics_strict_edge_reward": 0.19784490019083023,
      "rewards/forensics_tiou_reward": 0.2553166523575783,
      "step": 1802
    },
    {
      "completion_length": 27.046875,
      "epoch": 3.7727748691099476,
      "grad_norm": 29.91301727294922,
      "kl": 7.5625,
      "learning_rate": 5.555555555555555e-08,
      "loss": 0.5759,
      "reward": 0.5035581290721893,
      "reward_std": 0.2493961900472641,
      "reward_std_all": 0.2691706717014313,
      "rewards-max/forensics_sinkhorn_reward": 0.967460960149765,
      "rewards-max/forensics_strict_edge_reward": 0.820955753326416,
      "rewards-max/forensics_tiou_reward": 0.9495742619037628,
      "rewards/forensics_sinkhorn_reward": 0.4371861517429352,
      "rewards/forensics_strict_edge_reward": 0.21977361291646957,
      "rewards/forensics_tiou_reward": 0.29937825351953506,
      "step": 1803
    },
    {
      "completion_length": 24.484375,
      "epoch": 3.774869109947644,
      "grad_norm": 20.739952087402344,
      "kl": 6.953125,
      "learning_rate": 5.50314465408805e-08,
      "loss": 0.3639,
      "reward": 0.5566272437572479,
      "reward_std": 0.22565095126628876,
      "reward_std_all": 0.24128293246030807,
      "rewards-max/forensics_sinkhorn_reward": 0.9883837699890137,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9976096451282501,
      "rewards/forensics_sinkhorn_reward": 0.4866640418767929,
      "rewards/forensics_strict_edge_reward": 0.25044646114110947,
      "rewards/forensics_tiou_reward": 0.3138458579778671,
      "step": 1804
    },
    {
      "completion_length": 22.140625,
      "epoch": 3.77696335078534,
      "grad_norm": 20.31114387512207,
      "kl": 3.53125,
      "learning_rate": 5.4507337526205445e-08,
      "loss": 0.1898,
      "reward": 0.6431401968002319,
      "reward_std": 0.15480569377541542,
      "reward_std_all": 0.1632513925433159,
      "rewards-max/forensics_sinkhorn_reward": 0.9436967372894287,
      "rewards-max/forensics_strict_edge_reward": 0.8263171315193176,
      "rewards-max/forensics_tiou_reward": 0.9005010426044464,
      "rewards/forensics_sinkhorn_reward": 0.5998029112815857,
      "rewards/forensics_strict_edge_reward": 0.2971177250146866,
      "rewards/forensics_tiou_reward": 0.426571249961853,
      "step": 1805
    },
    {
      "completion_length": 28.40625,
      "epoch": 3.7790575916230367,
      "grad_norm": 32.905879974365234,
      "kl": 6.4375,
      "learning_rate": 5.398322851153039e-08,
      "loss": 0.4582,
      "reward": 0.5725522637367249,
      "reward_std": 0.27586743235588074,
      "reward_std_all": 0.2645619139075279,
      "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": 0.4937063753604889,
      "rewards/forensics_strict_edge_reward": 0.3611036241054535,
      "rewards/forensics_tiou_reward": 0.3877669721841812,
      "step": 1806
    },
    {
      "completion_length": 36.734375,
      "epoch": 3.7811518324607327,
      "grad_norm": 60.79046630859375,
      "kl": 9.34375,
      "learning_rate": 5.345911949685534e-08,
      "loss": 0.6468,
      "reward": 0.5032312721014023,
      "reward_std": 0.2752465158700943,
      "reward_std_all": 0.2888641059398651,
      "rewards-max/forensics_sinkhorn_reward": 0.9871806204319,
      "rewards-max/forensics_strict_edge_reward": 0.9357379674911499,
      "rewards-max/forensics_tiou_reward": 0.9928747415542603,
      "rewards/forensics_sinkhorn_reward": 0.43840619921684265,
      "rewards/forensics_strict_edge_reward": 0.265093557536602,
      "rewards/forensics_tiou_reward": 0.2934044152498245,
      "step": 1807
    },
    {
      "completion_length": 20.96875,
      "epoch": 3.783246073298429,
      "grad_norm": 33.52851104736328,
      "kl": 8.40625,
      "learning_rate": 5.293501048218029e-08,
      "loss": 0.6121,
      "reward": 0.5531487464904785,
      "reward_std": 0.23448015004396439,
      "reward_std_all": 0.24595507979393005,
      "rewards-max/forensics_sinkhorn_reward": 0.916558176279068,
      "rewards-max/forensics_strict_edge_reward": 0.759757399559021,
      "rewards-max/forensics_tiou_reward": 0.8471813797950745,
      "rewards/forensics_sinkhorn_reward": 0.46602286398410797,
      "rewards/forensics_strict_edge_reward": 0.2327992171049118,
      "rewards/forensics_tiou_reward": 0.2990102171897888,
      "step": 1808
    },
    {
      "completion_length": 20.9375,
      "epoch": 3.7853403141361257,
      "grad_norm": 19.707256317138672,
      "kl": 3.9296875,
      "learning_rate": 5.2410901467505234e-08,
      "loss": 0.2259,
      "reward": 0.6167621910572052,
      "reward_std": 0.11970358341932297,
      "reward_std_all": 0.1525009348988533,
      "rewards-max/forensics_sinkhorn_reward": 0.966215044260025,
      "rewards-max/forensics_strict_edge_reward": 0.8771188855171204,
      "rewards-max/forensics_tiou_reward": 0.9485105276107788,
      "rewards/forensics_sinkhorn_reward": 0.48579275608062744,
      "rewards/forensics_strict_edge_reward": 0.2666364908218384,
      "rewards/forensics_tiou_reward": 0.3264206051826477,
      "step": 1809
    },
    {
      "completion_length": 41.046875,
      "epoch": 3.7874345549738218,
      "grad_norm": 25.789278030395508,
      "kl": 7.0234375,
      "learning_rate": 5.1886792452830186e-08,
      "loss": 0.425,
      "reward": 0.6291938126087189,
      "reward_std": 0.19553469121456146,
      "reward_std_all": 0.1868457905948162,
      "rewards-max/forensics_sinkhorn_reward": 0.9810271561145782,
      "rewards-max/forensics_strict_edge_reward": 0.9256840348243713,
      "rewards-max/forensics_tiou_reward": 0.9811415672302246,
      "rewards/forensics_sinkhorn_reward": 0.5642510056495667,
      "rewards/forensics_strict_edge_reward": 0.34248319268226624,
      "rewards/forensics_tiou_reward": 0.4204930514097214,
      "step": 1810
    },
    {
      "completion_length": 32.703125,
      "epoch": 3.7895287958115182,
      "grad_norm": 29.42921257019043,
      "kl": 8.265625,
      "learning_rate": 5.136268343815513e-08,
      "loss": 0.6667,
      "reward": 0.5302508473396301,
      "reward_std": 0.2785811275243759,
      "reward_std_all": 0.2886543869972229,
      "rewards-max/forensics_sinkhorn_reward": 0.9886008203029633,
      "rewards-max/forensics_strict_edge_reward": 0.9782643616199493,
      "rewards-max/forensics_tiou_reward": 0.9984650909900665,
      "rewards/forensics_sinkhorn_reward": 0.4912779927253723,
      "rewards/forensics_strict_edge_reward": 0.3490273803472519,
      "rewards/forensics_tiou_reward": 0.3928501456975937,
      "step": 1811
    },
    {
      "completion_length": 25.984375,
      "epoch": 3.7916230366492147,
      "grad_norm": 24.12607192993164,
      "kl": 7.5625,
      "learning_rate": 5.083857442348008e-08,
      "loss": 0.5382,
      "reward": 0.5227432250976562,
      "reward_std": 0.20181114971637726,
      "reward_std_all": 0.246914304792881,
      "rewards-max/forensics_sinkhorn_reward": 0.9848361909389496,
      "rewards-max/forensics_strict_edge_reward": 0.8663162291049957,
      "rewards-max/forensics_tiou_reward": 0.9837678670883179,
      "rewards/forensics_sinkhorn_reward": 0.41196706891059875,
      "rewards/forensics_strict_edge_reward": 0.24538998305797577,
      "rewards/forensics_tiou_reward": 0.2843107283115387,
      "step": 1812
    },
    {
      "completion_length": 21.15625,
      "epoch": 3.793717277486911,
      "grad_norm": 34.795570373535156,
      "kl": 7.828125,
      "learning_rate": 5.031446540880503e-08,
      "loss": 0.4249,
      "reward": 0.5293150246143341,
      "reward_std": 0.2582940608263016,
      "reward_std_all": 0.26654990017414093,
      "rewards-max/forensics_sinkhorn_reward": 0.9586863219738007,
      "rewards-max/forensics_strict_edge_reward": 0.7923295497894287,
      "rewards-max/forensics_tiou_reward": 0.9228885173797607,
      "rewards/forensics_sinkhorn_reward": 0.43784022331237793,
      "rewards/forensics_strict_edge_reward": 0.23122096061706543,
      "rewards/forensics_tiou_reward": 0.27663353085517883,
      "step": 1813
    },
    {
      "completion_length": 30.0,
      "epoch": 3.7958115183246073,
      "grad_norm": 32.756771087646484,
      "kl": 10.3125,
      "learning_rate": 4.9790356394129975e-08,
      "loss": 0.646,
      "reward": 0.4374682754278183,
      "reward_std": 0.26019221544265747,
      "reward_std_all": 0.313495397567749,
      "rewards-max/forensics_sinkhorn_reward": 0.9877510070800781,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9951144158840179,
      "rewards/forensics_sinkhorn_reward": 0.3782891631126404,
      "rewards/forensics_strict_edge_reward": 0.27051933109760284,
      "rewards/forensics_tiou_reward": 0.30615170300006866,
      "step": 1814
    },
    {
      "completion_length": 25.421875,
      "epoch": 3.7979057591623038,
      "grad_norm": 40.483612060546875,
      "kl": 11.53125,
      "learning_rate": 4.926624737945492e-08,
      "loss": 0.8796,
      "reward": 0.4472910314798355,
      "reward_std": 0.2901631146669388,
      "reward_std_all": 0.3212096691131592,
      "rewards-max/forensics_sinkhorn_reward": 0.8417973518371582,
      "rewards-max/forensics_strict_edge_reward": 0.833502858877182,
      "rewards-max/forensics_tiou_reward": 0.7680944204330444,
      "rewards/forensics_sinkhorn_reward": 0.35765980184078217,
      "rewards/forensics_strict_edge_reward": 0.21456210315227509,
      "rewards/forensics_tiou_reward": 0.2523522302508354,
      "step": 1815
    },
    {
      "completion_length": 22.9375,
      "epoch": 3.8,
      "grad_norm": 24.154884338378906,
      "kl": 6.84375,
      "learning_rate": 4.874213836477987e-08,
      "loss": 0.4189,
      "reward": 0.5477376580238342,
      "reward_std": 0.22210103273391724,
      "reward_std_all": 0.2321755215525627,
      "rewards-max/forensics_sinkhorn_reward": 0.97929248213768,
      "rewards-max/forensics_strict_edge_reward": 0.9256840348243713,
      "rewards-max/forensics_tiou_reward": 0.9738196730613708,
      "rewards/forensics_sinkhorn_reward": 0.45048701763153076,
      "rewards/forensics_strict_edge_reward": 0.19807149469852448,
      "rewards/forensics_tiou_reward": 0.2732287794351578,
      "step": 1816
    },
    {
      "completion_length": 34.703125,
      "epoch": 3.8020942408376963,
      "grad_norm": 21.486244201660156,
      "kl": 7.171875,
      "learning_rate": 4.821802935010482e-08,
      "loss": 0.4757,
      "reward": 0.5474644601345062,
      "reward_std": 0.19520363956689835,
      "reward_std_all": 0.22454943507909775,
      "rewards-max/forensics_sinkhorn_reward": 0.98000168800354,
      "rewards-max/forensics_strict_edge_reward": 0.8893801271915436,
      "rewards-max/forensics_tiou_reward": 0.9787635505199432,
      "rewards/forensics_sinkhorn_reward": 0.4751523733139038,
      "rewards/forensics_strict_edge_reward": 0.22548379004001617,
      "rewards/forensics_tiou_reward": 0.3372609317302704,
      "step": 1817
    },
    {
      "completion_length": 34.09375,
      "epoch": 3.804188481675393,
      "grad_norm": 24.25095558166504,
      "kl": 6.28125,
      "learning_rate": 4.769392033542976e-08,
      "loss": 0.3128,
      "reward": 0.5192039012908936,
      "reward_std": 0.2520381510257721,
      "reward_std_all": 0.27870437502861023,
      "rewards-max/forensics_sinkhorn_reward": 0.9528099596500397,
      "rewards-max/forensics_strict_edge_reward": 0.8426913917064667,
      "rewards-max/forensics_tiou_reward": 0.9194069504737854,
      "rewards/forensics_sinkhorn_reward": 0.42083844542503357,
      "rewards/forensics_strict_edge_reward": 0.23690278828144073,
      "rewards/forensics_tiou_reward": 0.27707376331090927,
      "step": 1818
    },
    {
      "completion_length": 31.828125,
      "epoch": 3.806282722513089,
      "grad_norm": 21.859268188476562,
      "kl": 5.75,
      "learning_rate": 4.7169811320754715e-08,
      "loss": 0.4758,
      "reward": 0.5502626895904541,
      "reward_std": 0.1966383159160614,
      "reward_std_all": 0.21845926344394684,
      "rewards-max/forensics_sinkhorn_reward": 0.9875927567481995,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9944918155670166,
      "rewards/forensics_sinkhorn_reward": 0.48412829637527466,
      "rewards/forensics_strict_edge_reward": 0.2922279164195061,
      "rewards/forensics_tiou_reward": 0.3371056616306305,
      "step": 1819
    },
    {
      "completion_length": 17.234375,
      "epoch": 3.8083769633507853,
      "grad_norm": 23.829524993896484,
      "kl": 3.2734375,
      "learning_rate": 4.664570230607966e-08,
      "loss": 0.2262,
      "reward": 0.5881212651729584,
      "reward_std": 0.17754840850830078,
      "reward_std_all": 0.1898588091135025,
      "rewards-max/forensics_sinkhorn_reward": 0.9882289469242096,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.99700066447258,
      "rewards/forensics_sinkhorn_reward": 0.5791516304016113,
      "rewards/forensics_strict_edge_reward": 0.33166953921318054,
      "rewards/forensics_tiou_reward": 0.4058738797903061,
      "step": 1820
    },
    {
      "completion_length": 35.734375,
      "epoch": 3.810471204188482,
      "grad_norm": 16.953678131103516,
      "kl": 6.4375,
      "learning_rate": 4.6121593291404606e-08,
      "loss": 0.4926,
      "reward": 0.5961742401123047,
      "reward_std": 0.23633398115634918,
      "reward_std_all": 0.22630999982357025,
      "rewards-max/forensics_sinkhorn_reward": 0.9863391518592834,
      "rewards-max/forensics_strict_edge_reward": 0.8663162291049957,
      "rewards-max/forensics_tiou_reward": 0.989552229642868,
      "rewards/forensics_sinkhorn_reward": 0.5335197299718857,
      "rewards/forensics_strict_edge_reward": 0.35052913427352905,
      "rewards/forensics_tiou_reward": 0.4124152511358261,
      "step": 1821
    },
    {
      "completion_length": 20.9375,
      "epoch": 3.812565445026178,
      "grad_norm": 22.29597282409668,
      "kl": 5.671875,
      "learning_rate": 4.559748427672955e-08,
      "loss": 0.3545,
      "reward": 0.5839493870735168,
      "reward_std": 0.15825724601745605,
      "reward_std_all": 0.18068166077136993,
      "rewards-max/forensics_sinkhorn_reward": 0.9877331852912903,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9950549900531769,
      "rewards/forensics_sinkhorn_reward": 0.5096631646156311,
      "rewards/forensics_strict_edge_reward": 0.24446579813957214,
      "rewards/forensics_tiou_reward": 0.29534414410591125,
      "step": 1822
    },
    {
      "completion_length": 27.703125,
      "epoch": 3.8146596858638744,
      "grad_norm": 21.978939056396484,
      "kl": 8.90625,
      "learning_rate": 4.5073375262054504e-08,
      "loss": 0.6795,
      "reward": 0.45945146679878235,
      "reward_std": 0.20894326269626617,
      "reward_std_all": 0.2694377303123474,
      "rewards-max/forensics_sinkhorn_reward": 0.986802339553833,
      "rewards-max/forensics_strict_edge_reward": 0.9048932790756226,
      "rewards-max/forensics_tiou_reward": 0.9913983047008514,
      "rewards/forensics_sinkhorn_reward": 0.4089456647634506,
      "rewards/forensics_strict_edge_reward": 0.2198338359594345,
      "rewards/forensics_tiou_reward": 0.2739057093858719,
      "step": 1823
    },
    {
      "completion_length": 29.84375,
      "epoch": 3.816753926701571,
      "grad_norm": 18.673526763916016,
      "kl": 4.1015625,
      "learning_rate": 4.454926624737945e-08,
      "loss": 0.3601,
      "reward": 0.5984428524971008,
      "reward_std": 0.2500247061252594,
      "reward_std_all": 0.23272330313920975,
      "rewards-max/forensics_sinkhorn_reward": 0.9887801706790924,
      "rewards-max/forensics_strict_edge_reward": 0.9890114367008209,
      "rewards-max/forensics_tiou_reward": 0.9991711676120758,
      "rewards/forensics_sinkhorn_reward": 0.5541201531887054,
      "rewards/forensics_strict_edge_reward": 0.3823910653591156,
      "rewards/forensics_tiou_reward": 0.4592988044023514,
      "step": 1824
    },
    {
      "completion_length": 29.171875,
      "epoch": 3.818848167539267,
      "grad_norm": 20.74547004699707,
      "kl": 5.109375,
      "learning_rate": 4.4025157232704395e-08,
      "loss": 0.2645,
      "reward": 0.5545481443405151,
      "reward_std": 0.17080021649599075,
      "reward_std_all": 0.20605815201997757,
      "rewards-max/forensics_sinkhorn_reward": 0.9309726655483246,
      "rewards-max/forensics_strict_edge_reward": 0.8189329206943512,
      "rewards-max/forensics_tiou_reward": 0.8691102564334869,
      "rewards/forensics_sinkhorn_reward": 0.4254339933395386,
      "rewards/forensics_strict_edge_reward": 0.2024853527545929,
      "rewards/forensics_tiou_reward": 0.2555917948484421,
      "step": 1825
    },
    {
      "completion_length": 34.8125,
      "epoch": 3.8209424083769634,
      "grad_norm": 25.209980010986328,
      "kl": 6.296875,
      "learning_rate": 4.350104821802935e-08,
      "loss": 0.3889,
      "reward": 0.6121728420257568,
      "reward_std": 0.20296670496463776,
      "reward_std_all": 0.20650973170995712,
      "rewards-max/forensics_sinkhorn_reward": 0.9874367415904999,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9938755035400391,
      "rewards/forensics_sinkhorn_reward": 0.5509625971317291,
      "rewards/forensics_strict_edge_reward": 0.3505086302757263,
      "rewards/forensics_tiou_reward": 0.40545332431793213,
      "step": 1826
    },
    {
      "completion_length": 23.359375,
      "epoch": 3.82303664921466,
      "grad_norm": 25.09739112854004,
      "kl": 5.0234375,
      "learning_rate": 4.297693920335429e-08,
      "loss": 0.4278,
      "reward": 0.5781063735485077,
      "reward_std": 0.2127043828368187,
      "reward_std_all": 0.22393463551998138,
      "rewards-max/forensics_sinkhorn_reward": 0.9877086877822876,
      "rewards-max/forensics_strict_edge_reward": 0.9464850425720215,
      "rewards-max/forensics_tiou_reward": 0.9949476718902588,
      "rewards/forensics_sinkhorn_reward": 0.4805179536342621,
      "rewards/forensics_strict_edge_reward": 0.25636371970176697,
      "rewards/forensics_tiou_reward": 0.2939559668302536,
      "step": 1827
    },
    {
      "completion_length": 22.875,
      "epoch": 3.825130890052356,
      "grad_norm": 31.83321762084961,
      "kl": 8.640625,
      "learning_rate": 4.245283018867924e-08,
      "loss": 0.5246,
      "reward": 0.4983961284160614,
      "reward_std": 0.24879620969295502,
      "reward_std_all": 0.27267536520957947,
      "rewards-max/forensics_sinkhorn_reward": 0.9457014203071594,
      "rewards-max/forensics_strict_edge_reward": 0.9062341153621674,
      "rewards-max/forensics_tiou_reward": 0.9077557325363159,
      "rewards/forensics_sinkhorn_reward": 0.40632855892181396,
      "rewards/forensics_strict_edge_reward": 0.1981658786535263,
      "rewards/forensics_tiou_reward": 0.28038348257541656,
      "step": 1828
    },
    {
      "completion_length": 28.328125,
      "epoch": 3.8272251308900525,
      "grad_norm": 21.84300994873047,
      "kl": 5.84375,
      "learning_rate": 4.1928721174004196e-08,
      "loss": 0.3787,
      "reward": 0.5853494107723236,
      "reward_std": 0.11420051008462906,
      "reward_std_all": 0.1648285761475563,
      "rewards-max/forensics_sinkhorn_reward": 0.9875428080558777,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9943039119243622,
      "rewards/forensics_sinkhorn_reward": 0.536570131778717,
      "rewards/forensics_strict_edge_reward": 0.3517751395702362,
      "rewards/forensics_tiou_reward": 0.43152396380901337,
      "step": 1829
    },
    {
      "completion_length": 17.828125,
      "epoch": 3.829319371727749,
      "grad_norm": 18.989395141601562,
      "kl": 4.65625,
      "learning_rate": 4.140461215932914e-08,
      "loss": 0.288,
      "reward": 0.6152416169643402,
      "reward_std": 0.1853502094745636,
      "reward_std_all": 0.18815256655216217,
      "rewards-max/forensics_sinkhorn_reward": 0.9857898652553558,
      "rewards-max/forensics_strict_edge_reward": 0.9567649364471436,
      "rewards-max/forensics_tiou_reward": 0.9874569177627563,
      "rewards/forensics_sinkhorn_reward": 0.5472762882709503,
      "rewards/forensics_strict_edge_reward": 0.30835703015327454,
      "rewards/forensics_tiou_reward": 0.3617287427186966,
      "step": 1830
    },
    {
      "completion_length": 25.96875,
      "epoch": 3.831413612565445,
      "grad_norm": 195.0552978515625,
      "kl": 7.4375,
      "learning_rate": 4.088050314465408e-08,
      "loss": 0.3846,
      "reward": 0.6511929035186768,
      "reward_std": 0.16740570217370987,
      "reward_std_all": 0.16123245283961296,
      "rewards-max/forensics_sinkhorn_reward": 0.947878360748291,
      "rewards-max/forensics_strict_edge_reward": 0.9012884795665741,
      "rewards-max/forensics_tiou_reward": 0.9105305671691895,
      "rewards/forensics_sinkhorn_reward": 0.5545666813850403,
      "rewards/forensics_strict_edge_reward": 0.2857239693403244,
      "rewards/forensics_tiou_reward": 0.35616037249565125,
      "step": 1831
    },
    {
      "completion_length": 21.203125,
      "epoch": 3.8335078534031415,
      "grad_norm": 20.025550842285156,
      "kl": 6.328125,
      "learning_rate": 4.035639412997904e-08,
      "loss": 0.506,
      "reward": 0.578503280878067,
      "reward_std": 0.19841941446065903,
      "reward_std_all": 0.2113209068775177,
      "rewards-max/forensics_sinkhorn_reward": 0.9810770452022552,
      "rewards-max/forensics_strict_edge_reward": 0.8260191679000854,
      "rewards-max/forensics_tiou_reward": 0.9781225323677063,
      "rewards/forensics_sinkhorn_reward": 0.46656176447868347,
      "rewards/forensics_strict_edge_reward": 0.27084001153707504,
      "rewards/forensics_tiou_reward": 0.3589716702699661,
      "step": 1832
    },
    {
      "completion_length": 45.25,
      "epoch": 3.8356020942408375,
      "grad_norm": 15.625092506408691,
      "kl": 6.625,
      "learning_rate": 3.9832285115303985e-08,
      "loss": 0.4979,
      "reward": 0.524996817111969,
      "reward_std": 0.21820053458213806,
      "reward_std_all": 0.24910133332014084,
      "rewards-max/forensics_sinkhorn_reward": 0.9883938431739807,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9976493716239929,
      "rewards/forensics_sinkhorn_reward": 0.4657290279865265,
      "rewards/forensics_strict_edge_reward": 0.24539855867624283,
      "rewards/forensics_tiou_reward": 0.3381481021642685,
      "step": 1833
    },
    {
      "completion_length": 29.46875,
      "epoch": 3.837696335078534,
      "grad_norm": 22.601987838745117,
      "kl": 6.78125,
      "learning_rate": 3.930817610062893e-08,
      "loss": 0.472,
      "reward": 0.5211318135261536,
      "reward_std": 0.2571875751018524,
      "reward_std_all": 0.2724307179450989,
      "rewards-max/forensics_sinkhorn_reward": 0.7462013065814972,
      "rewards-max/forensics_strict_edge_reward": 0.5314938277006149,
      "rewards-max/forensics_tiou_reward": 0.6288444697856903,
      "rewards/forensics_sinkhorn_reward": 0.37644144892692566,
      "rewards/forensics_strict_edge_reward": 0.1481776423752308,
      "rewards/forensics_tiou_reward": 0.18991898000240326,
      "step": 1834
    },
    {
      "completion_length": 36.5625,
      "epoch": 3.8397905759162305,
      "grad_norm": 26.089792251586914,
      "kl": 7.328125,
      "learning_rate": 3.878406708595388e-08,
      "loss": 0.3624,
      "reward": 0.5800411105155945,
      "reward_std": 0.18837056308984756,
      "reward_std_all": 0.20346103608608246,
      "rewards-max/forensics_sinkhorn_reward": 0.9679705500602722,
      "rewards-max/forensics_strict_edge_reward": 0.8394034504890442,
      "rewards-max/forensics_tiou_reward": 0.9509396553039551,
      "rewards/forensics_sinkhorn_reward": 0.46521060168743134,
      "rewards/forensics_strict_edge_reward": 0.23079101741313934,
      "rewards/forensics_tiou_reward": 0.2880490720272064,
      "step": 1835
    },
    {
      "completion_length": 24.921875,
      "epoch": 3.8418848167539266,
      "grad_norm": 22.19576072692871,
      "kl": 5.328125,
      "learning_rate": 3.825995807127883e-08,
      "loss": 0.3149,
      "reward": 0.5814551711082458,
      "reward_std": 0.2693806141614914,
      "reward_std_all": 0.2511882483959198,
      "rewards-max/forensics_sinkhorn_reward": 0.988244354724884,
      "rewards-max/forensics_strict_edge_reward": 0.9782643616199493,
      "rewards-max/forensics_tiou_reward": 0.9970588088035583,
      "rewards/forensics_sinkhorn_reward": 0.49488840997219086,
      "rewards/forensics_strict_edge_reward": 0.31716322898864746,
      "rewards/forensics_tiou_reward": 0.338767945766449,
      "step": 1836
    },
    {
      "completion_length": 29.859375,
      "epoch": 3.843979057591623,
      "grad_norm": 25.90485954284668,
      "kl": 6.171875,
      "learning_rate": 3.7735849056603774e-08,
      "loss": 0.356,
      "reward": 0.5250571966171265,
      "reward_std": 0.13327400758862495,
      "reward_std_all": 0.20429972559213638,
      "rewards-max/forensics_sinkhorn_reward": 0.9776778817176819,
      "rewards-max/forensics_strict_edge_reward": 0.8612285256385803,
      "rewards-max/forensics_tiou_reward": 0.9662461280822754,
      "rewards/forensics_sinkhorn_reward": 0.4291945993900299,
      "rewards/forensics_strict_edge_reward": 0.21155215799808502,
      "rewards/forensics_tiou_reward": 0.2549184784293175,
      "step": 1837
    },
    {
      "completion_length": 31.953125,
      "epoch": 3.846073298429319,
      "grad_norm": 31.661571502685547,
      "kl": 4.671875,
      "learning_rate": 3.7211740041928726e-08,
      "loss": 0.52,
      "reward": 0.5497925579547882,
      "reward_std": 0.18389804661273956,
      "reward_std_all": 0.21928270906209946,
      "rewards-max/forensics_sinkhorn_reward": 0.9872990250587463,
      "rewards-max/forensics_strict_edge_reward": 0.9464850425720215,
      "rewards-max/forensics_tiou_reward": 0.9933451414108276,
      "rewards/forensics_sinkhorn_reward": 0.5141242295503616,
      "rewards/forensics_strict_edge_reward": 0.31373709440231323,
      "rewards/forensics_tiou_reward": 0.40382443368434906,
      "step": 1838
    },
    {
      "completion_length": 32.703125,
      "epoch": 3.8481675392670156,
      "grad_norm": 20.09375,
      "kl": 6.234375,
      "learning_rate": 3.668763102725367e-08,
      "loss": 0.4006,
      "reward": 0.5526422262191772,
      "reward_std": 0.19148749858140945,
      "reward_std_all": 0.21445250511169434,
      "rewards-max/forensics_sinkhorn_reward": 0.986855536699295,
      "rewards-max/forensics_strict_edge_reward": 0.9567649364471436,
      "rewards-max/forensics_tiou_reward": 0.9916079938411713,
      "rewards/forensics_sinkhorn_reward": 0.5017650127410889,
      "rewards/forensics_strict_edge_reward": 0.2876756489276886,
      "rewards/forensics_tiou_reward": 0.3705475330352783,
      "step": 1839
    },
    {
      "completion_length": 21.234375,
      "epoch": 3.850261780104712,
      "grad_norm": 21.501840591430664,
      "kl": 3.4921875,
      "learning_rate": 3.6163522012578617e-08,
      "loss": 0.3622,
      "reward": 0.6382952928543091,
      "reward_std": 0.20464620739221573,
      "reward_std_all": 0.19236447662115097,
      "rewards-max/forensics_sinkhorn_reward": 0.9884001314640045,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9976727366447449,
      "rewards/forensics_sinkhorn_reward": 0.5635253190994263,
      "rewards/forensics_strict_edge_reward": 0.37291643023490906,
      "rewards/forensics_tiou_reward": 0.419069305062294,
      "step": 1840
    },
    {
      "completion_length": 17.21875,
      "epoch": 3.852356020942408,
      "grad_norm": 18.180450439453125,
      "kl": 4.75,
      "learning_rate": 3.563941299790356e-08,
      "loss": 0.3545,
      "reward": 0.6261177361011505,
      "reward_std": 0.1852731555700302,
      "reward_std_all": 0.1829582378268242,
      "rewards-max/forensics_sinkhorn_reward": 0.987770289182663,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9951901733875275,
      "rewards/forensics_sinkhorn_reward": 0.5955473184585571,
      "rewards/forensics_strict_edge_reward": 0.4137522578239441,
      "rewards/forensics_tiou_reward": 0.46245843172073364,
      "step": 1841
    },
    {
      "completion_length": 47.484375,
      "epoch": 3.8544502617801046,
      "grad_norm": 23.69032859802246,
      "kl": 8.25,
      "learning_rate": 3.5115303983228514e-08,
      "loss": 0.6277,
      "reward": 0.5075753778219223,
      "reward_std": 0.2731904983520508,
      "reward_std_all": 0.2948223054409027,
      "rewards-max/forensics_sinkhorn_reward": 0.9882822632789612,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.997208297252655,
      "rewards/forensics_sinkhorn_reward": 0.42419418692588806,
      "rewards/forensics_strict_edge_reward": 0.23994430154561996,
      "rewards/forensics_tiou_reward": 0.2856132313609123,
      "step": 1842
    },
    {
      "completion_length": 18.921875,
      "epoch": 3.856544502617801,
      "grad_norm": 21.412872314453125,
      "kl": 7.484375,
      "learning_rate": 3.459119496855346e-08,
      "loss": 0.4624,
      "reward": 0.486928254365921,
      "reward_std": 0.19919582456350327,
      "reward_std_all": 0.25753556936979294,
      "rewards-max/forensics_sinkhorn_reward": 0.9887009263038635,
      "rewards-max/forensics_strict_edge_reward": 0.9890114367008209,
      "rewards-max/forensics_tiou_reward": 0.9988591074943542,
      "rewards/forensics_sinkhorn_reward": 0.40624628961086273,
      "rewards/forensics_strict_edge_reward": 0.2806173637509346,
      "rewards/forensics_tiou_reward": 0.30473100394010544,
      "step": 1843
    },
    {
      "completion_length": 26.765625,
      "epoch": 3.858638743455497,
      "grad_norm": 25.506324768066406,
      "kl": 7.375,
      "learning_rate": 3.4067085953878405e-08,
      "loss": 0.6159,
      "reward": 0.5085052400827408,
      "reward_std": 0.2412906214594841,
      "reward_std_all": 0.2567247301340103,
      "rewards-max/forensics_sinkhorn_reward": 0.9872896671295166,
      "rewards-max/forensics_strict_edge_reward": 0.8503920137882233,
      "rewards-max/forensics_tiou_reward": 0.9933192431926727,
      "rewards/forensics_sinkhorn_reward": 0.48488500714302063,
      "rewards/forensics_strict_edge_reward": 0.32247380912303925,
      "rewards/forensics_tiou_reward": 0.39673884212970734,
      "step": 1844
    },
    {
      "completion_length": 17.65625,
      "epoch": 3.8607329842931937,
      "grad_norm": 24.33000373840332,
      "kl": 2.5,
      "learning_rate": 3.354297693920336e-08,
      "loss": 0.15,
      "reward": 0.6094002723693848,
      "reward_std": 0.12464343011379242,
      "reward_std_all": 0.15346024557948112,
      "rewards-max/forensics_sinkhorn_reward": 0.9873878359794617,
      "rewards-max/forensics_strict_edge_reward": 0.9464850425720215,
      "rewards-max/forensics_tiou_reward": 0.9937100410461426,
      "rewards/forensics_sinkhorn_reward": 0.5608570277690887,
      "rewards/forensics_strict_edge_reward": 0.3393825590610504,
      "rewards/forensics_tiou_reward": 0.38411808013916016,
      "step": 1845
    },
    {
      "completion_length": 42.109375,
      "epoch": 3.86282722513089,
      "grad_norm": 27.040231704711914,
      "kl": 6.0625,
      "learning_rate": 3.30188679245283e-08,
      "loss": 0.3767,
      "reward": 0.50955730676651,
      "reward_std": 0.11945106089115143,
      "reward_std_all": 0.20057765394449234,
      "rewards-max/forensics_sinkhorn_reward": 0.9700213670730591,
      "rewards-max/forensics_strict_edge_reward": 0.8460287153720856,
      "rewards-max/forensics_tiou_reward": 0.9540785253047943,
      "rewards/forensics_sinkhorn_reward": 0.42830735445022583,
      "rewards/forensics_strict_edge_reward": 0.1972695291042328,
      "rewards/forensics_tiou_reward": 0.27674566209316254,
      "step": 1846
    },
    {
      "completion_length": 18.015625,
      "epoch": 3.864921465968586,
      "grad_norm": 19.030344009399414,
      "kl": 4.78125,
      "learning_rate": 3.249475890985325e-08,
      "loss": 0.3927,
      "reward": 0.6506190001964569,
      "reward_std": 0.1484149694442749,
      "reward_std_all": 0.1608889326453209,
      "rewards-max/forensics_sinkhorn_reward": 0.9878363907337189,
      "rewards-max/forensics_strict_edge_reward": 0.946017861366272,
      "rewards-max/forensics_tiou_reward": 0.9954580962657928,
      "rewards/forensics_sinkhorn_reward": 0.6188590824604034,
      "rewards/forensics_strict_edge_reward": 0.36336976289749146,
      "rewards/forensics_tiou_reward": 0.5043412744998932,
      "step": 1847
    },
    {
      "completion_length": 19.890625,
      "epoch": 3.8670157068062827,
      "grad_norm": 30.094144821166992,
      "kl": 5.21875,
      "learning_rate": 3.19706498951782e-08,
      "loss": 0.4602,
      "reward": 0.568023681640625,
      "reward_std": 0.24290695786476135,
      "reward_std_all": 0.2388342097401619,
      "rewards-max/forensics_sinkhorn_reward": 0.9882270991802216,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9969930946826935,
      "rewards/forensics_sinkhorn_reward": 0.5115030854940414,
      "rewards/forensics_strict_edge_reward": 0.27878229320049286,
      "rewards/forensics_tiou_reward": 0.3342348635196686,
      "step": 1848
    },
    {
      "completion_length": 28.296875,
      "epoch": 3.869109947643979,
      "grad_norm": 21.209508895874023,
      "kl": 4.921875,
      "learning_rate": 3.1446540880503146e-08,
      "loss": 0.3041,
      "reward": 0.5870365500450134,
      "reward_std": 0.22355277836322784,
      "reward_std_all": 0.22168687731027603,
      "rewards-max/forensics_sinkhorn_reward": 0.9875442981719971,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.994306743144989,
      "rewards/forensics_sinkhorn_reward": 0.4934036433696747,
      "rewards/forensics_strict_edge_reward": 0.2526855617761612,
      "rewards/forensics_tiou_reward": 0.2856108397245407,
      "step": 1849
    },
    {
      "completion_length": 35.890625,
      "epoch": 3.8712041884816752,
      "grad_norm": 19.085966110229492,
      "kl": 7.078125,
      "learning_rate": 3.092243186582809e-08,
      "loss": 0.5644,
      "reward": 0.4237353056669235,
      "reward_std": 0.21236602216959,
      "reward_std_all": 0.2880350202322006,
      "rewards-max/forensics_sinkhorn_reward": 0.8530999422073364,
      "rewards-max/forensics_strict_edge_reward": 0.7157100737094879,
      "rewards-max/forensics_tiou_reward": 0.7266627103090286,
      "rewards/forensics_sinkhorn_reward": 0.3331228047609329,
      "rewards/forensics_strict_edge_reward": 0.20895598083734512,
      "rewards/forensics_tiou_reward": 0.22246555611491203,
      "step": 1850
    },
    {
      "completion_length": 23.828125,
      "epoch": 3.8732984293193717,
      "grad_norm": 28.73847770690918,
      "kl": 6.6875,
      "learning_rate": 3.0398322851153037e-08,
      "loss": 0.3839,
      "reward": 0.6007985770702362,
      "reward_std": 0.1872042715549469,
      "reward_std_all": 0.1991293802857399,
      "rewards-max/forensics_sinkhorn_reward": 0.9635065495967865,
      "rewards-max/forensics_strict_edge_reward": 0.9375866651535034,
      "rewards-max/forensics_tiou_reward": 0.9438543915748596,
      "rewards/forensics_sinkhorn_reward": 0.5489038228988647,
      "rewards/forensics_strict_edge_reward": 0.26992592215538025,
      "rewards/forensics_tiou_reward": 0.3798524886369705,
      "step": 1851
    },
    {
      "completion_length": 35.3125,
      "epoch": 3.8753926701570682,
      "grad_norm": 23.99011993408203,
      "kl": 6.484375,
      "learning_rate": 2.987421383647799e-08,
      "loss": 0.5134,
      "reward": 0.4043479263782501,
      "reward_std": 0.15544294565916061,
      "reward_std_all": 0.2673018127679825,
      "rewards-max/forensics_sinkhorn_reward": 0.9226210117340088,
      "rewards-max/forensics_strict_edge_reward": 0.8472771048545837,
      "rewards-max/forensics_tiou_reward": 0.8585478067398071,
      "rewards/forensics_sinkhorn_reward": 0.3348144143819809,
      "rewards/forensics_strict_edge_reward": 0.13798971846699715,
      "rewards/forensics_tiou_reward": 0.1780708283185959,
      "step": 1852
    },
    {
      "completion_length": 26.453125,
      "epoch": 3.8774869109947643,
      "grad_norm": 25.93553352355957,
      "kl": 7.4375,
      "learning_rate": 2.9350104821802934e-08,
      "loss": 0.49,
      "reward": 0.5244150012731552,
      "reward_std": 0.2689859867095947,
      "reward_std_all": 0.2768199294805527,
      "rewards-max/forensics_sinkhorn_reward": 0.9306873381137848,
      "rewards-max/forensics_strict_edge_reward": 0.8983768224716187,
      "rewards-max/forensics_tiou_reward": 0.8768741190433502,
      "rewards/forensics_sinkhorn_reward": 0.4337642937898636,
      "rewards/forensics_strict_edge_reward": 0.2407420575618744,
      "rewards/forensics_tiou_reward": 0.27860577404499054,
      "step": 1853
    },
    {
      "completion_length": 23.703125,
      "epoch": 3.8795811518324608,
      "grad_norm": 22.841012954711914,
      "kl": 7.5625,
      "learning_rate": 2.882599580712788e-08,
      "loss": 0.4839,
      "reward": 0.47450777888298035,
      "reward_std": 0.2842307388782501,
      "reward_std_all": 0.29988472908735275,
      "rewards-max/forensics_sinkhorn_reward": 0.9878372848033905,
      "rewards-max/forensics_strict_edge_reward": 0.9782643616199493,
      "rewards-max/forensics_tiou_reward": 0.9954545497894287,
      "rewards/forensics_sinkhorn_reward": 0.38245370984077454,
      "rewards/forensics_strict_edge_reward": 0.2184135913848877,
      "rewards/forensics_tiou_reward": 0.25599078834056854,
      "step": 1854
    },
    {
      "completion_length": 25.34375,
      "epoch": 3.8816753926701573,
      "grad_norm": 21.99659538269043,
      "kl": 7.078125,
      "learning_rate": 2.830188679245283e-08,
      "loss": 0.5267,
      "reward": 0.44308678805828094,
      "reward_std": 0.1739766038954258,
      "reward_std_all": 0.26300421357154846,
      "rewards-max/forensics_sinkhorn_reward": 0.890295535326004,
      "rewards-max/forensics_strict_edge_reward": 0.7210069298744202,
      "rewards-max/forensics_tiou_reward": 0.7982760369777679,
      "rewards/forensics_sinkhorn_reward": 0.30073027312755585,
      "rewards/forensics_strict_edge_reward": 0.1534200981259346,
      "rewards/forensics_tiou_reward": 0.2037145346403122,
      "step": 1855
    },
    {
      "completion_length": 31.15625,
      "epoch": 3.8837696335078533,
      "grad_norm": 18.337923049926758,
      "kl": 6.21875,
      "learning_rate": 2.7777777777777774e-08,
      "loss": 0.453,
      "reward": 0.6274366080760956,
      "reward_std": 0.22073618322610855,
      "reward_std_all": 0.204256109893322,
      "rewards-max/forensics_sinkhorn_reward": 0.9783043265342712,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9757847785949707,
      "rewards/forensics_sinkhorn_reward": 0.5509793162345886,
      "rewards/forensics_strict_edge_reward": 0.34334704279899597,
      "rewards/forensics_tiou_reward": 0.43634751439094543,
      "step": 1856
    },
    {
      "completion_length": 33.8125,
      "epoch": 3.88586387434555,
      "grad_norm": 24.68975067138672,
      "kl": 6.0,
      "learning_rate": 2.7253668763102723e-08,
      "loss": 0.4482,
      "reward": 0.5600214004516602,
      "reward_std": 0.31411874294281006,
      "reward_std_all": 0.29010987281799316,
      "rewards-max/forensics_sinkhorn_reward": 0.9490771293640137,
      "rewards-max/forensics_strict_edge_reward": 0.7763782441616058,
      "rewards-max/forensics_tiou_reward": 0.9073380529880524,
      "rewards/forensics_sinkhorn_reward": 0.4671690911054611,
      "rewards/forensics_strict_edge_reward": 0.322610542178154,
      "rewards/forensics_tiou_reward": 0.3452528715133667,
      "step": 1857
    },
    {
      "completion_length": 23.6875,
      "epoch": 3.8879581151832463,
      "grad_norm": 27.00141143798828,
      "kl": 4.265625,
      "learning_rate": 2.672955974842767e-08,
      "loss": 0.1573,
      "reward": 0.5858855247497559,
      "reward_std": 0.2099476233124733,
      "reward_std_all": 0.21613821387290955,
      "rewards-max/forensics_sinkhorn_reward": 0.9888400435447693,
      "rewards-max/forensics_strict_edge_reward": 0.9890114367008209,
      "rewards-max/forensics_tiou_reward": 0.9994070529937744,
      "rewards/forensics_sinkhorn_reward": 0.5269393920898438,
      "rewards/forensics_strict_edge_reward": 0.2142922729253769,
      "rewards/forensics_tiou_reward": 0.31276507675647736,
      "step": 1858
    },
    {
      "completion_length": 18.9375,
      "epoch": 3.8900523560209423,
      "grad_norm": 38.75638198852539,
      "kl": 4.9375,
      "learning_rate": 2.6205450733752617e-08,
      "loss": 0.2004,
      "reward": 0.6330494284629822,
      "reward_std": 0.1726638227701187,
      "reward_std_all": 0.16929355263710022,
      "rewards-max/forensics_sinkhorn_reward": 0.9694110453128815,
      "rewards-max/forensics_strict_edge_reward": 0.8134243786334991,
      "rewards-max/forensics_tiou_reward": 0.9543076753616333,
      "rewards/forensics_sinkhorn_reward": 0.55124831199646,
      "rewards/forensics_strict_edge_reward": 0.3581918925046921,
      "rewards/forensics_tiou_reward": 0.4133676141500473,
      "step": 1859
    },
    {
      "completion_length": 25.859375,
      "epoch": 3.892146596858639,
      "grad_norm": 35.383419036865234,
      "kl": 9.75,
      "learning_rate": 2.5681341719077566e-08,
      "loss": 0.6649,
      "reward": 0.500551849603653,
      "reward_std": 0.2729734778404236,
      "reward_std_all": 0.2985362112522125,
      "rewards-max/forensics_sinkhorn_reward": 0.9447607696056366,
      "rewards-max/forensics_strict_edge_reward": 0.8097340762615204,
      "rewards-max/forensics_tiou_reward": 0.8974641561508179,
      "rewards/forensics_sinkhorn_reward": 0.414853036403656,
      "rewards/forensics_strict_edge_reward": 0.2277526557445526,
      "rewards/forensics_tiou_reward": 0.28199100494384766,
      "step": 1860
    },
    {
      "completion_length": 20.890625,
      "epoch": 3.8942408376963353,
      "grad_norm": 21.37909698486328,
      "kl": 4.1796875,
      "learning_rate": 2.5157232704402515e-08,
      "loss": 0.2034,
      "reward": 0.6224722266197205,
      "reward_std": 0.14435675740242004,
      "reward_std_all": 0.16501232236623764,
      "rewards-max/forensics_sinkhorn_reward": 0.9884403049945831,
      "rewards-max/forensics_strict_edge_reward": 0.9565287232398987,
      "rewards-max/forensics_tiou_reward": 0.9978318810462952,
      "rewards/forensics_sinkhorn_reward": 0.5856914520263672,
      "rewards/forensics_strict_edge_reward": 0.27532222867012024,
      "rewards/forensics_tiou_reward": 0.36158591508865356,
      "step": 1861
    },
    {
      "completion_length": 28.84375,
      "epoch": 3.8963350785340314,
      "grad_norm": 18.18075942993164,
      "kl": 7.03125,
      "learning_rate": 2.463312368972746e-08,
      "loss": 0.5679,
      "reward": 0.44671057164669037,
      "reward_std": 0.1427634134888649,
      "reward_std_all": 0.23868580162525177,
      "rewards-max/forensics_sinkhorn_reward": 0.9881628155708313,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9967386722564697,
      "rewards/forensics_sinkhorn_reward": 0.34114882349967957,
      "rewards/forensics_strict_edge_reward": 0.14399903267621994,
      "rewards/forensics_tiou_reward": 0.1558733843266964,
      "step": 1862
    },
    {
      "completion_length": 32.953125,
      "epoch": 3.898429319371728,
      "grad_norm": 28.413352966308594,
      "kl": 9.53125,
      "learning_rate": 2.410901467505241e-08,
      "loss": 0.5242,
      "reward": 0.49864719808101654,
      "reward_std": 0.21319957077503204,
      "reward_std_all": 0.2586014121770859,
      "rewards-max/forensics_sinkhorn_reward": 0.9741587936878204,
      "rewards-max/forensics_strict_edge_reward": 0.9464850425720215,
      "rewards-max/forensics_tiou_reward": 0.9664222598075867,
      "rewards/forensics_sinkhorn_reward": 0.43584127724170685,
      "rewards/forensics_strict_edge_reward": 0.24217426776885986,
      "rewards/forensics_tiou_reward": 0.3030737489461899,
      "step": 1863
    },
    {
      "completion_length": 38.21875,
      "epoch": 3.900523560209424,
      "grad_norm": 18.895713806152344,
      "kl": 7.21875,
      "learning_rate": 2.3584905660377358e-08,
      "loss": 0.5448,
      "reward": 0.542049914598465,
      "reward_std": 0.2678103744983673,
      "reward_std_all": 0.2586952894926071,
      "rewards-max/forensics_sinkhorn_reward": 0.9871575236320496,
      "rewards-max/forensics_strict_edge_reward": 0.9364311099052429,
      "rewards-max/forensics_tiou_reward": 0.9927938878536224,
      "rewards/forensics_sinkhorn_reward": 0.4846869707107544,
      "rewards/forensics_strict_edge_reward": 0.27407801151275635,
      "rewards/forensics_tiou_reward": 0.38467833399772644,
      "step": 1864
    },
    {
      "completion_length": 35.6875,
      "epoch": 3.9026178010471204,
      "grad_norm": 21.1685791015625,
      "kl": 6.984375,
      "learning_rate": 2.3060796645702303e-08,
      "loss": 0.3764,
      "reward": 0.5405806601047516,
      "reward_std": 0.09347415715456009,
      "reward_std_all": 0.17085903510451317,
      "rewards-max/forensics_sinkhorn_reward": 0.9886906445026398,
      "rewards-max/forensics_strict_edge_reward": 0.9890114367008209,
      "rewards-max/forensics_tiou_reward": 0.9988186657428741,
      "rewards/forensics_sinkhorn_reward": 0.49758070707321167,
      "rewards/forensics_strict_edge_reward": 0.3321885019540787,
      "rewards/forensics_tiou_reward": 0.36894533038139343,
      "step": 1865
    },
    {
      "completion_length": 19.625,
      "epoch": 3.904712041884817,
      "grad_norm": 25.986644744873047,
      "kl": 4.28125,
      "learning_rate": 2.2536687631027252e-08,
      "loss": 0.2054,
      "reward": 0.6404168903827667,
      "reward_std": 0.17917539924383163,
      "reward_std_all": 0.17472966015338898,
      "rewards-max/forensics_sinkhorn_reward": 0.9840045869350433,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.988228052854538,
      "rewards/forensics_sinkhorn_reward": 0.5419698655605316,
      "rewards/forensics_strict_edge_reward": 0.30210747569799423,
      "rewards/forensics_tiou_reward": 0.3634142130613327,
      "step": 1866
    },
    {
      "completion_length": 31.9375,
      "epoch": 3.906806282722513,
      "grad_norm": 22.763307571411133,
      "kl": 6.453125,
      "learning_rate": 2.2012578616352197e-08,
      "loss": 0.512,
      "reward": 0.5071883499622345,
      "reward_std": 0.2607125788927078,
      "reward_std_all": 0.28606972098350525,
      "rewards-max/forensics_sinkhorn_reward": 0.9879663288593292,
      "rewards-max/forensics_strict_edge_reward": 0.9565287232398987,
      "rewards-max/forensics_tiou_reward": 0.9959681332111359,
      "rewards/forensics_sinkhorn_reward": 0.37899404764175415,
      "rewards/forensics_strict_edge_reward": 0.224588081240654,
      "rewards/forensics_tiou_reward": 0.24177926033735275,
      "step": 1867
    },
    {
      "completion_length": 29.8125,
      "epoch": 3.9089005235602095,
      "grad_norm": 27.30563735961914,
      "kl": 4.984375,
      "learning_rate": 2.1488469601677146e-08,
      "loss": 0.1264,
      "reward": 0.5669774115085602,
      "reward_std": 0.25182202458381653,
      "reward_std_all": 0.25592969357967377,
      "rewards-max/forensics_sinkhorn_reward": 0.963926374912262,
      "rewards-max/forensics_strict_edge_reward": 0.9003686904907227,
      "rewards-max/forensics_tiou_reward": 0.9460557401180267,
      "rewards/forensics_sinkhorn_reward": 0.4376492500305176,
      "rewards/forensics_strict_edge_reward": 0.27664585411548615,
      "rewards/forensics_tiou_reward": 0.3101600408554077,
      "step": 1868
    },
    {
      "completion_length": 22.984375,
      "epoch": 3.9109947643979055,
      "grad_norm": 38.40194320678711,
      "kl": 6.75,
      "learning_rate": 2.0964360587002098e-08,
      "loss": 0.4726,
      "reward": 0.5725868344306946,
      "reward_std": 0.24385957419872284,
      "reward_std_all": 0.2466517984867096,
      "rewards-max/forensics_sinkhorn_reward": 0.9875603318214417,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9943721294403076,
      "rewards/forensics_sinkhorn_reward": 0.4686416983604431,
      "rewards/forensics_strict_edge_reward": 0.23891621083021164,
      "rewards/forensics_tiou_reward": 0.3449976146221161,
      "step": 1869
    },
    {
      "completion_length": 28.40625,
      "epoch": 3.913089005235602,
      "grad_norm": 23.18557357788086,
      "kl": 4.3828125,
      "learning_rate": 2.044025157232704e-08,
      "loss": 0.2626,
      "reward": 0.5971418023109436,
      "reward_std": 0.2248978391289711,
      "reward_std_all": 0.21827685087919235,
      "rewards-max/forensics_sinkhorn_reward": 0.9858456552028656,
      "rewards-max/forensics_strict_edge_reward": 0.946017861366272,
      "rewards-max/forensics_tiou_reward": 0.9876428842544556,
      "rewards/forensics_sinkhorn_reward": 0.4798605144023895,
      "rewards/forensics_strict_edge_reward": 0.2804648131132126,
      "rewards/forensics_tiou_reward": 0.3361595571041107,
      "step": 1870
    },
    {
      "completion_length": 18.921875,
      "epoch": 3.9151832460732985,
      "grad_norm": 20.91241455078125,
      "kl": 4.3046875,
      "learning_rate": 1.9916142557651992e-08,
      "loss": 0.3373,
      "reward": 0.5672168135643005,
      "reward_std": 0.2107977494597435,
      "reward_std_all": 0.2308533936738968,
      "rewards-max/forensics_sinkhorn_reward": 0.9886357188224792,
      "rewards-max/forensics_strict_edge_reward": 0.9890114367008209,
      "rewards-max/forensics_tiou_reward": 0.9986014068126678,
      "rewards/forensics_sinkhorn_reward": 0.539618119597435,
      "rewards/forensics_strict_edge_reward": 0.35098792612552643,
      "rewards/forensics_tiou_reward": 0.3804469555616379,
      "step": 1871
    },
    {
      "completion_length": 20.28125,
      "epoch": 3.9172774869109945,
      "grad_norm": 19.209915161132812,
      "kl": 5.8125,
      "learning_rate": 1.939203354297694e-08,
      "loss": 0.3605,
      "reward": 0.6426180005073547,
      "reward_std": 0.18311257660388947,
      "reward_std_all": 0.17631139606237411,
      "rewards-max/forensics_sinkhorn_reward": 0.9792592227458954,
      "rewards-max/forensics_strict_edge_reward": 0.9048932790756226,
      "rewards-max/forensics_tiou_reward": 0.973223865032196,
      "rewards/forensics_sinkhorn_reward": 0.5761706829071045,
      "rewards/forensics_strict_edge_reward": 0.3072715401649475,
      "rewards/forensics_tiou_reward": 0.40062132477760315,
      "step": 1872
    },
    {
      "completion_length": 28.46875,
      "epoch": 3.919371727748691,
      "grad_norm": 32.304954528808594,
      "kl": 7.328125,
      "learning_rate": 1.8867924528301887e-08,
      "loss": 0.3918,
      "reward": 0.5123961865901947,
      "reward_std": 0.26487183570861816,
      "reward_std_all": 0.2766475975513458,
      "rewards-max/forensics_sinkhorn_reward": 0.9802125990390778,
      "rewards-max/forensics_strict_edge_reward": 0.8426913917064667,
      "rewards-max/forensics_tiou_reward": 0.9799035489559174,
      "rewards/forensics_sinkhorn_reward": 0.45366500318050385,
      "rewards/forensics_strict_edge_reward": 0.28063300251960754,
      "rewards/forensics_tiou_reward": 0.3326915130019188,
      "step": 1873
    },
    {
      "completion_length": 28.515625,
      "epoch": 3.9214659685863875,
      "grad_norm": 16.654836654663086,
      "kl": 7.671875,
      "learning_rate": 1.8343815513626836e-08,
      "loss": 0.6116,
      "reward": 0.493488073348999,
      "reward_std": 0.27665670216083527,
      "reward_std_all": 0.29217013716697693,
      "rewards-max/forensics_sinkhorn_reward": 0.9875406622886658,
      "rewards-max/forensics_strict_edge_reward": 0.9782643616199493,
      "rewards-max/forensics_tiou_reward": 0.9942857027053833,
      "rewards/forensics_sinkhorn_reward": 0.34167417883872986,
      "rewards/forensics_strict_edge_reward": 0.1834753006696701,
      "rewards/forensics_tiou_reward": 0.20121244341135025,
      "step": 1874
    },
    {
      "completion_length": 19.921875,
      "epoch": 3.9235602094240836,
      "grad_norm": 22.215171813964844,
      "kl": 3.5390625,
      "learning_rate": 1.781970649895178e-08,
      "loss": 0.1266,
      "reward": 0.6634328961372375,
      "reward_std": 0.12769482284784317,
      "reward_std_all": 0.13640817254781723,
      "rewards-max/forensics_sinkhorn_reward": 0.8996360898017883,
      "rewards-max/forensics_strict_edge_reward": 0.7630731761455536,
      "rewards-max/forensics_tiou_reward": 0.8013538718223572,
      "rewards/forensics_sinkhorn_reward": 0.5520793497562408,
      "rewards/forensics_strict_edge_reward": 0.29750271886587143,
      "rewards/forensics_tiou_reward": 0.37230604887008667,
      "step": 1875
    },
    {
      "completion_length": 21.9375,
      "epoch": 3.92565445026178,
      "grad_norm": 17.660966873168945,
      "kl": 3.890625,
      "learning_rate": 1.729559748427673e-08,
      "loss": 0.1445,
      "reward": 0.5620156228542328,
      "reward_std": 0.24074269086122513,
      "reward_std_all": 0.23795729875564575,
      "rewards-max/forensics_sinkhorn_reward": 0.9836609363555908,
      "rewards-max/forensics_strict_edge_reward": 0.9101436734199524,
      "rewards-max/forensics_tiou_reward": 0.9791324138641357,
      "rewards/forensics_sinkhorn_reward": 0.46532511711120605,
      "rewards/forensics_strict_edge_reward": 0.21280773729085922,
      "rewards/forensics_tiou_reward": 0.2523004412651062,
      "step": 1876
    },
    {
      "completion_length": 25.4375,
      "epoch": 3.9277486910994766,
      "grad_norm": 22.935138702392578,
      "kl": 4.375,
      "learning_rate": 1.677148846960168e-08,
      "loss": 0.3844,
      "reward": 0.6376174688339233,
      "reward_std": 0.1714591309428215,
      "reward_std_all": 0.17872436344623566,
      "rewards-max/forensics_sinkhorn_reward": 0.9889025986194611,
      "rewards-max/forensics_strict_edge_reward": 0.9890114367008209,
      "rewards-max/forensics_tiou_reward": 0.999653548002243,
      "rewards/forensics_sinkhorn_reward": 0.5484334826469421,
      "rewards/forensics_strict_edge_reward": 0.36541762948036194,
      "rewards/forensics_tiou_reward": 0.39607855677604675,
      "step": 1877
    },
    {
      "completion_length": 21.671875,
      "epoch": 3.9298429319371726,
      "grad_norm": 28.875375747680664,
      "kl": 4.984375,
      "learning_rate": 1.6247379454926624e-08,
      "loss": 0.4092,
      "reward": 0.5268574059009552,
      "reward_std": 0.1753738671541214,
      "reward_std_all": 0.22534413635730743,
      "rewards-max/forensics_sinkhorn_reward": 0.9879591763019562,
      "rewards-max/forensics_strict_edge_reward": 0.9782643616199493,
      "rewards-max/forensics_tiou_reward": 0.9959432184696198,
      "rewards/forensics_sinkhorn_reward": 0.42906729876995087,
      "rewards/forensics_strict_edge_reward": 0.2612188011407852,
      "rewards/forensics_tiou_reward": 0.26481547951698303,
      "step": 1878
    },
    {
      "completion_length": 30.234375,
      "epoch": 3.931937172774869,
      "grad_norm": 21.389814376831055,
      "kl": 5.484375,
      "learning_rate": 1.5723270440251573e-08,
      "loss": 0.3615,
      "reward": 0.598529726266861,
      "reward_std": 0.1814073771238327,
      "reward_std_all": 0.19837962836027145,
      "rewards-max/forensics_sinkhorn_reward": 0.9536857008934021,
      "rewards-max/forensics_strict_edge_reward": 0.8025174736976624,
      "rewards-max/forensics_tiou_reward": 0.9216893017292023,
      "rewards/forensics_sinkhorn_reward": 0.4949701577425003,
      "rewards/forensics_strict_edge_reward": 0.2573937475681305,
      "rewards/forensics_tiou_reward": 0.30738605558872223,
      "step": 1879
    },
    {
      "completion_length": 47.65625,
      "epoch": 3.9340314136125656,
      "grad_norm": 94.86246490478516,
      "kl": 7.609375,
      "learning_rate": 1.5199161425576518e-08,
      "loss": 0.4601,
      "reward": 0.5776300728321075,
      "reward_std": 0.23877838253974915,
      "reward_std_all": 0.23202116042375565,
      "rewards-max/forensics_sinkhorn_reward": 0.9838489890098572,
      "rewards-max/forensics_strict_edge_reward": 0.8893801271915436,
      "rewards-max/forensics_tiou_reward": 0.9799183309078217,
      "rewards/forensics_sinkhorn_reward": 0.5050821006298065,
      "rewards/forensics_strict_edge_reward": 0.29039037227630615,
      "rewards/forensics_tiou_reward": 0.3681522607803345,
      "step": 1880
    },
    {
      "completion_length": 22.515625,
      "epoch": 3.9361256544502616,
      "grad_norm": 22.6379337310791,
      "kl": 5.328125,
      "learning_rate": 1.4675052410901467e-08,
      "loss": 0.3241,
      "reward": 0.5957888960838318,
      "reward_std": 0.19750383496284485,
      "reward_std_all": 0.2119368612766266,
      "rewards-max/forensics_sinkhorn_reward": 0.9884797632694244,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9979881048202515,
      "rewards/forensics_sinkhorn_reward": 0.5700135827064514,
      "rewards/forensics_strict_edge_reward": 0.3381711393594742,
      "rewards/forensics_tiou_reward": 0.4265446662902832,
      "step": 1881
    },
    {
      "completion_length": 19.84375,
      "epoch": 3.938219895287958,
      "grad_norm": 25.196495056152344,
      "kl": 6.578125,
      "learning_rate": 1.4150943396226414e-08,
      "loss": 0.4756,
      "reward": 0.5561995506286621,
      "reward_std": 0.28665582835674286,
      "reward_std_all": 0.2689785212278366,
      "rewards-max/forensics_sinkhorn_reward": 0.9821612536907196,
      "rewards-max/forensics_strict_edge_reward": 0.9364311099052429,
      "rewards-max/forensics_tiou_reward": 0.9832797646522522,
      "rewards/forensics_sinkhorn_reward": 0.5201554745435715,
      "rewards/forensics_strict_edge_reward": 0.33315008878707886,
      "rewards/forensics_tiou_reward": 0.4092137813568115,
      "step": 1882
    },
    {
      "completion_length": 20.21875,
      "epoch": 3.9403141361256546,
      "grad_norm": 22.469879150390625,
      "kl": 6.484375,
      "learning_rate": 1.3626834381551361e-08,
      "loss": 0.2915,
      "reward": 0.622058629989624,
      "reward_std": 0.16562283039093018,
      "reward_std_all": 0.17355628311634064,
      "rewards-max/forensics_sinkhorn_reward": 0.9885846078395844,
      "rewards-max/forensics_strict_edge_reward": 0.9890114367008209,
      "rewards-max/forensics_tiou_reward": 0.9984000027179718,
      "rewards/forensics_sinkhorn_reward": 0.5335289239883423,
      "rewards/forensics_strict_edge_reward": 0.3011942505836487,
      "rewards/forensics_tiou_reward": 0.3571077138185501,
      "step": 1883
    },
    {
      "completion_length": 28.09375,
      "epoch": 3.9424083769633507,
      "grad_norm": 27.503204345703125,
      "kl": 6.046875,
      "learning_rate": 1.3102725366876308e-08,
      "loss": 0.438,
      "reward": 0.529743492603302,
      "reward_std": 0.19903849810361862,
      "reward_std_all": 0.23190948367118835,
      "rewards-max/forensics_sinkhorn_reward": 0.9879547953605652,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9959179162979126,
      "rewards/forensics_sinkhorn_reward": 0.46624840795993805,
      "rewards/forensics_strict_edge_reward": 0.23036599159240723,
      "rewards/forensics_tiou_reward": 0.28508107364177704,
      "step": 1884
    },
    {
      "completion_length": 26.578125,
      "epoch": 3.944502617801047,
      "grad_norm": 47.81047058105469,
      "kl": 7.078125,
      "learning_rate": 1.2578616352201257e-08,
      "loss": 0.4826,
      "reward": 0.5634951889514923,
      "reward_std": 0.19551245868206024,
      "reward_std_all": 0.22659121453762054,
      "rewards-max/forensics_sinkhorn_reward": 0.9871882200241089,
      "rewards-max/forensics_strict_edge_reward": 0.9464850425720215,
      "rewards-max/forensics_tiou_reward": 0.9929197728633881,
      "rewards/forensics_sinkhorn_reward": 0.5008018612861633,
      "rewards/forensics_strict_edge_reward": 0.30700305104255676,
      "rewards/forensics_tiou_reward": 0.3988145440816879,
      "step": 1885
    },
    {
      "completion_length": 22.203125,
      "epoch": 3.9465968586387437,
      "grad_norm": 19.01552391052246,
      "kl": 4.90625,
      "learning_rate": 1.2054507337526204e-08,
      "loss": 0.3985,
      "reward": 0.5952498316764832,
      "reward_std": 0.24793097376823425,
      "reward_std_all": 0.2320147231221199,
      "rewards-max/forensics_sinkhorn_reward": 0.9881614446640015,
      "rewards-max/forensics_strict_edge_reward": 0.9782643616199493,
      "rewards-max/forensics_tiou_reward": 0.9967373609542847,
      "rewards/forensics_sinkhorn_reward": 0.5398025810718536,
      "rewards/forensics_strict_edge_reward": 0.34273573756217957,
      "rewards/forensics_tiou_reward": 0.42372801899909973,
      "step": 1886
    },
    {
      "completion_length": 26.4375,
      "epoch": 3.9486910994764397,
      "grad_norm": 18.907243728637695,
      "kl": 6.046875,
      "learning_rate": 1.1530398322851152e-08,
      "loss": 0.4816,
      "reward": 0.5050951242446899,
      "reward_std": 0.288012832403183,
      "reward_std_all": 0.29294632375240326,
      "rewards-max/forensics_sinkhorn_reward": 0.9095018208026886,
      "rewards-max/forensics_strict_edge_reward": 0.7822710573673248,
      "rewards-max/forensics_tiou_reward": 0.8326546251773834,
      "rewards/forensics_sinkhorn_reward": 0.3800855576992035,
      "rewards/forensics_strict_edge_reward": 0.19237396121025085,
      "rewards/forensics_tiou_reward": 0.24573662877082825,
      "step": 1887
    },
    {
      "completion_length": 19.296875,
      "epoch": 3.950785340314136,
      "grad_norm": 30.479745864868164,
      "kl": 5.96875,
      "learning_rate": 1.1006289308176099e-08,
      "loss": 0.3751,
      "reward": 0.649750292301178,
      "reward_std": 0.1857641637325287,
      "reward_std_all": 0.1733611859381199,
      "rewards-max/forensics_sinkhorn_reward": 0.9878876507282257,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9956553280353546,
      "rewards/forensics_sinkhorn_reward": 0.5797626674175262,
      "rewards/forensics_strict_edge_reward": 0.27684423327445984,
      "rewards/forensics_tiou_reward": 0.4282132238149643,
      "step": 1888
    },
    {
      "completion_length": 23.0,
      "epoch": 3.9528795811518327,
      "grad_norm": 22.261343002319336,
      "kl": 6.125,
      "learning_rate": 1.0482180293501049e-08,
      "loss": 0.4414,
      "reward": 0.48334380984306335,
      "reward_std": 0.24124589562416077,
      "reward_std_all": 0.2770577669143677,
      "rewards-max/forensics_sinkhorn_reward": 0.9869504272937775,
      "rewards-max/forensics_strict_edge_reward": 0.9464850425720215,
      "rewards-max/forensics_tiou_reward": 0.9919785559177399,
      "rewards/forensics_sinkhorn_reward": 0.3769061267375946,
      "rewards/forensics_strict_edge_reward": 0.19444527477025986,
      "rewards/forensics_tiou_reward": 0.26337939500808716,
      "step": 1889
    },
    {
      "completion_length": 22.53125,
      "epoch": 3.9549738219895287,
      "grad_norm": 19.665889739990234,
      "kl": 6.859375,
      "learning_rate": 9.958071278825996e-09,
      "loss": 0.5818,
      "reward": 0.510688990354538,
      "reward_std": 0.29392483830451965,
      "reward_std_all": 0.2975669503211975,
      "rewards-max/forensics_sinkhorn_reward": 0.9755937159061432,
      "rewards-max/forensics_strict_edge_reward": 0.9158510267734528,
      "rewards-max/forensics_tiou_reward": 0.9689986705780029,
      "rewards/forensics_sinkhorn_reward": 0.4332965612411499,
      "rewards/forensics_strict_edge_reward": 0.217100128531456,
      "rewards/forensics_tiou_reward": 0.3388185352087021,
      "step": 1890
    },
    {
      "completion_length": 27.234375,
      "epoch": 3.9570680628272252,
      "grad_norm": 24.173398971557617,
      "kl": 5.6875,
      "learning_rate": 9.433962264150943e-09,
      "loss": 0.4623,
      "reward": 0.49394120275974274,
      "reward_std": 0.19775784015655518,
      "reward_std_all": 0.2507791444659233,
      "rewards-max/forensics_sinkhorn_reward": 0.9872593581676483,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9931934177875519,
      "rewards/forensics_sinkhorn_reward": 0.4240962266921997,
      "rewards/forensics_strict_edge_reward": 0.2882799506187439,
      "rewards/forensics_tiou_reward": 0.3036268353462219,
      "step": 1891
    },
    {
      "completion_length": 23.328125,
      "epoch": 3.9591623036649217,
      "grad_norm": 18.384647369384766,
      "kl": 5.859375,
      "learning_rate": 8.90985324947589e-09,
      "loss": 0.377,
      "reward": 0.5556670874357224,
      "reward_std": 0.19019761681556702,
      "reward_std_all": 0.2135079726576805,
      "rewards-max/forensics_sinkhorn_reward": 0.9839485883712769,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9881227612495422,
      "rewards/forensics_sinkhorn_reward": 0.44836241006851196,
      "rewards/forensics_strict_edge_reward": 0.23281222581863403,
      "rewards/forensics_tiou_reward": 0.3001093864440918,
      "step": 1892
    },
    {
      "completion_length": 18.078125,
      "epoch": 3.9612565445026178,
      "grad_norm": 21.711639404296875,
      "kl": 5.84375,
      "learning_rate": 8.38574423480084e-09,
      "loss": 0.4011,
      "reward": 0.6332387924194336,
      "reward_std": 0.1515570878982544,
      "reward_std_all": 0.15478067100048065,
      "rewards-max/forensics_sinkhorn_reward": 0.9805915653705597,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9811823070049286,
      "rewards/forensics_sinkhorn_reward": 0.5466973781585693,
      "rewards/forensics_strict_edge_reward": 0.22721236944198608,
      "rewards/forensics_tiou_reward": 0.36946359276771545,
      "step": 1893
    },
    {
      "completion_length": 27.984375,
      "epoch": 3.9633507853403143,
      "grad_norm": 24.581787109375,
      "kl": 4.921875,
      "learning_rate": 7.861635220125786e-09,
      "loss": 0.3376,
      "reward": 0.5577138364315033,
      "reward_std": 0.2502119839191437,
      "reward_std_all": 0.24376599490642548,
      "rewards-max/forensics_sinkhorn_reward": 0.988649994134903,
      "rewards-max/forensics_strict_edge_reward": 0.9890114367008209,
      "rewards-max/forensics_tiou_reward": 0.9986577033996582,
      "rewards/forensics_sinkhorn_reward": 0.48962780833244324,
      "rewards/forensics_strict_edge_reward": 0.3061116933822632,
      "rewards/forensics_tiou_reward": 0.3460584878921509,
      "step": 1894
    },
    {
      "completion_length": 36.96875,
      "epoch": 3.9654450261780103,
      "grad_norm": 32.230506896972656,
      "kl": 6.734375,
      "learning_rate": 7.3375262054507336e-09,
      "loss": 0.6174,
      "reward": 0.45667514204978943,
      "reward_std": 0.2703697234392166,
      "reward_std_all": 0.3047057092189789,
      "rewards-max/forensics_sinkhorn_reward": 0.9666618704795837,
      "rewards-max/forensics_strict_edge_reward": 0.8187307715415955,
      "rewards-max/forensics_tiou_reward": 0.9513137638568878,
      "rewards/forensics_sinkhorn_reward": 0.34994323551654816,
      "rewards/forensics_strict_edge_reward": 0.17068196833133698,
      "rewards/forensics_tiou_reward": 0.23253022134304047,
      "step": 1895
    },
    {
      "completion_length": 19.640625,
      "epoch": 3.967539267015707,
      "grad_norm": 21.708656311035156,
      "kl": 6.390625,
      "learning_rate": 6.813417190775681e-09,
      "loss": 0.4956,
      "reward": 0.5614305436611176,
      "reward_std": 0.20798378437757492,
      "reward_std_all": 0.22861575335264206,
      "rewards-max/forensics_sinkhorn_reward": 0.9884298741817474,
      "rewards-max/forensics_strict_edge_reward": 0.9574736058712006,
      "rewards-max/forensics_tiou_reward": 0.9977900683879852,
      "rewards/forensics_sinkhorn_reward": 0.4815230071544647,
      "rewards/forensics_strict_edge_reward": 0.26779307425022125,
      "rewards/forensics_tiou_reward": 0.3474702686071396,
      "step": 1896
    },
    {
      "completion_length": 26.59375,
      "epoch": 3.9696335078534033,
      "grad_norm": 24.599504470825195,
      "kl": 3.953125,
      "learning_rate": 6.289308176100629e-09,
      "loss": 0.4628,
      "reward": 0.5636695623397827,
      "reward_std": 0.17693553864955902,
      "reward_std_all": 0.2078835666179657,
      "rewards-max/forensics_sinkhorn_reward": 0.9806813895702362,
      "rewards-max/forensics_strict_edge_reward": 0.9169811904430389,
      "rewards-max/forensics_tiou_reward": 0.9677842557430267,
      "rewards/forensics_sinkhorn_reward": 0.4758302569389343,
      "rewards/forensics_strict_edge_reward": 0.3016284108161926,
      "rewards/forensics_tiou_reward": 0.3453926742076874,
      "step": 1897
    },
    {
      "completion_length": 23.609375,
      "epoch": 3.9717277486910993,
      "grad_norm": 21.663761138916016,
      "kl": 7.4375,
      "learning_rate": 5.765199161425576e-09,
      "loss": 0.5799,
      "reward": 0.48775285482406616,
      "reward_std": 0.21885435283184052,
      "reward_std_all": 0.2656361162662506,
      "rewards-max/forensics_sinkhorn_reward": 0.9874145090579987,
      "rewards-max/forensics_strict_edge_reward": 0.9256840348243713,
      "rewards-max/forensics_tiou_reward": 0.9937892854213715,
      "rewards/forensics_sinkhorn_reward": 0.4368875324726105,
      "rewards/forensics_strict_edge_reward": 0.3142163157463074,
      "rewards/forensics_tiou_reward": 0.3224542886018753,
      "step": 1898
    },
    {
      "completion_length": 29.703125,
      "epoch": 3.973821989528796,
      "grad_norm": 22.538923263549805,
      "kl": 5.640625,
      "learning_rate": 5.2410901467505246e-09,
      "loss": 0.3905,
      "reward": 0.5576947331428528,
      "reward_std": 0.32407140731811523,
      "reward_std_all": 0.29707635939121246,
      "rewards-max/forensics_sinkhorn_reward": 0.9842846989631653,
      "rewards-max/forensics_strict_edge_reward": 0.8876297473907471,
      "rewards-max/forensics_tiou_reward": 0.9815734326839447,
      "rewards/forensics_sinkhorn_reward": 0.4897071570158005,
      "rewards/forensics_strict_edge_reward": 0.23870719224214554,
      "rewards/forensics_tiou_reward": 0.3390989601612091,
      "step": 1899
    },
    {
      "completion_length": 27.3125,
      "epoch": 3.975916230366492,
      "grad_norm": 24.420162200927734,
      "kl": 5.40625,
      "learning_rate": 4.716981132075472e-09,
      "loss": 0.416,
      "reward": 0.5901806950569153,
      "reward_std": 0.1843787059187889,
      "reward_std_all": 0.19324129819869995,
      "rewards-max/forensics_sinkhorn_reward": 0.9876298010349274,
      "rewards-max/forensics_strict_edge_reward": 0.9265981018543243,
      "rewards-max/forensics_tiou_reward": 0.9946494102478027,
      "rewards/forensics_sinkhorn_reward": 0.5557761192321777,
      "rewards/forensics_strict_edge_reward": 0.3466952443122864,
      "rewards/forensics_tiou_reward": 0.4366607218980789,
      "step": 1900
    },
    {
      "completion_length": 31.21875,
      "epoch": 3.9780104712041884,
      "grad_norm": 23.62542724609375,
      "kl": 5.90625,
      "learning_rate": 4.19287211740042e-09,
      "loss": 0.5094,
      "reward": 0.5837951600551605,
      "reward_std": 0.2631876468658447,
      "reward_std_all": 0.24640195071697235,
      "rewards-max/forensics_sinkhorn_reward": 0.983471006155014,
      "rewards-max/forensics_strict_edge_reward": 0.947419673204422,
      "rewards-max/forensics_tiou_reward": 0.9872363805770874,
      "rewards/forensics_sinkhorn_reward": 0.5670086741447449,
      "rewards/forensics_strict_edge_reward": 0.34946949779987335,
      "rewards/forensics_tiou_reward": 0.46433955430984497,
      "step": 1901
    },
    {
      "completion_length": 28.03125,
      "epoch": 3.980104712041885,
      "grad_norm": 27.950443267822266,
      "kl": 7.578125,
      "learning_rate": 3.6687631027253668e-09,
      "loss": 0.4855,
      "reward": 0.6212238967418671,
      "reward_std": 0.2161627784371376,
      "reward_std_all": 0.20310140401124954,
      "rewards-max/forensics_sinkhorn_reward": 0.9882152080535889,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9969440698623657,
      "rewards/forensics_sinkhorn_reward": 0.6094460785388947,
      "rewards/forensics_strict_edge_reward": 0.3873473107814789,
      "rewards/forensics_tiou_reward": 0.5005029290914536,
      "step": 1902
    },
    {
      "completion_length": 20.28125,
      "epoch": 3.982198952879581,
      "grad_norm": 25.028627395629883,
      "kl": 7.46875,
      "learning_rate": 3.1446540880503143e-09,
      "loss": 0.5874,
      "reward": 0.535571426153183,
      "reward_std": 0.1937660202383995,
      "reward_std_all": 0.23022030293941498,
      "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": 0.453365296125412,
      "rewards/forensics_strict_edge_reward": 0.3180364519357681,
      "rewards/forensics_tiou_reward": 0.34717150032520294,
      "step": 1903
    },
    {
      "completion_length": 25.390625,
      "epoch": 3.9842931937172774,
      "grad_norm": 22.189651489257812,
      "kl": 5.765625,
      "learning_rate": 2.6205450733752623e-09,
      "loss": 0.297,
      "reward": 0.5660213828086853,
      "reward_std": 0.20415912568569183,
      "reward_std_all": 0.2200780287384987,
      "rewards-max/forensics_sinkhorn_reward": 0.9884341657161713,
      "rewards-max/forensics_strict_edge_reward": 0.9890114367008209,
      "rewards-max/forensics_tiou_reward": 0.9978094100952148,
      "rewards/forensics_sinkhorn_reward": 0.4628138691186905,
      "rewards/forensics_strict_edge_reward": 0.23555082082748413,
      "rewards/forensics_tiou_reward": 0.31190936267375946,
      "step": 1904
    },
    {
      "completion_length": 32.28125,
      "epoch": 3.986387434554974,
      "grad_norm": 25.618738174438477,
      "kl": 5.75,
      "learning_rate": 2.09643605870021e-09,
      "loss": 0.4169,
      "reward": 0.5468049049377441,
      "reward_std": 0.24393530189990997,
      "reward_std_all": 0.2536883056163788,
      "rewards-max/forensics_sinkhorn_reward": 0.9876976013183594,
      "rewards-max/forensics_strict_edge_reward": 0.9565287232398987,
      "rewards-max/forensics_tiou_reward": 0.9949045479297638,
      "rewards/forensics_sinkhorn_reward": 0.4585418850183487,
      "rewards/forensics_strict_edge_reward": 0.26377787441015244,
      "rewards/forensics_tiou_reward": 0.31583593785762787,
      "step": 1905
    },
    {
      "completion_length": 24.828125,
      "epoch": 3.98848167539267,
      "grad_norm": 22.933956146240234,
      "kl": 7.6875,
      "learning_rate": 1.5723270440251572e-09,
      "loss": 0.4768,
      "reward": 0.5325536727905273,
      "reward_std": 0.19237864017486572,
      "reward_std_all": 0.2323058918118477,
      "rewards-max/forensics_sinkhorn_reward": 0.9669904410839081,
      "rewards-max/forensics_strict_edge_reward": 0.8854433596134186,
      "rewards-max/forensics_tiou_reward": 0.9495017528533936,
      "rewards/forensics_sinkhorn_reward": 0.4361439049243927,
      "rewards/forensics_strict_edge_reward": 0.19687888026237488,
      "rewards/forensics_tiou_reward": 0.3013337701559067,
      "step": 1906
    },
    {
      "completion_length": 30.4375,
      "epoch": 3.9905759162303664,
      "grad_norm": 129.66734313964844,
      "kl": 12.0625,
      "learning_rate": 1.048218029350105e-09,
      "loss": 0.7758,
      "reward": 0.5430958271026611,
      "reward_std": 0.2241186499595642,
      "reward_std_all": 0.25318196415901184,
      "rewards-max/forensics_sinkhorn_reward": 0.9736675918102264,
      "rewards-max/forensics_strict_edge_reward": 0.8854433596134186,
      "rewards-max/forensics_tiou_reward": 0.9594757854938507,
      "rewards/forensics_sinkhorn_reward": 0.4956912696361542,
      "rewards/forensics_strict_edge_reward": 0.3296704739332199,
      "rewards/forensics_tiou_reward": 0.3705131411552429,
      "step": 1907
    },
    {
      "completion_length": 22.28125,
      "epoch": 3.992670157068063,
      "grad_norm": 23.276134490966797,
      "kl": 6.828125,
      "learning_rate": 5.241090146750525e-10,
      "loss": 0.6158,
      "reward": 0.5164562612771988,
      "reward_std": 0.20255695283412933,
      "reward_std_all": 0.24187926948070526,
      "rewards-max/forensics_sinkhorn_reward": 0.9457266628742218,
      "rewards-max/forensics_strict_edge_reward": 0.8060585260391235,
      "rewards-max/forensics_tiou_reward": 0.9027236998081207,
      "rewards/forensics_sinkhorn_reward": 0.41617022454738617,
      "rewards/forensics_strict_edge_reward": 0.22074200958013535,
      "rewards/forensics_tiou_reward": 0.26140350103378296,
      "step": 1908
    }
  ],
  "logging_steps": 1.0,
  "max_steps": 1908,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 4,
  "save_steps": 1908,
  "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
}