{
  "best_global_step": null,
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 3.9906790945406128,
  "eval_steps": 500,
  "global_step": 1500,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "completion_length": 28.390625,
      "epoch": 0.002663115845539281,
      "grad_norm": 9.29894733428955,
      "kl": 0.0,
      "learning_rate": 1e-06,
      "loss": 0.0741,
      "reward": 0.4086398333311081,
      "reward_std": 0.26997560262680054,
      "reward_std_all": 0.31346680223941803,
      "rewards-max/forensics_sinkhorn_reward": 0.9830864369869232,
      "rewards-max/forensics_strict_edge_reward": 0.8805811703205109,
      "rewards-max/forensics_tiou_reward": 0.9770284295082092,
      "rewards/forensics_sinkhorn_reward": 0.3095322549343109,
      "rewards/forensics_strict_edge_reward": 0.10735158994793892,
      "rewards/forensics_tiou_reward": 0.13784009218215942,
      "step": 1
    },
    {
      "completion_length": 17.671875,
      "epoch": 0.005326231691078562,
      "grad_norm": 11.407604217529297,
      "kl": 0.001178741455078125,
      "learning_rate": 9.993333333333333e-07,
      "loss": 0.1655,
      "reward": 0.5042315423488617,
      "reward_std": 0.22098541259765625,
      "reward_std_all": 0.24766144156455994,
      "rewards-max/forensics_sinkhorn_reward": 0.9778173565864563,
      "rewards-max/forensics_strict_edge_reward": 0.771587461233139,
      "rewards-max/forensics_tiou_reward": 0.9568407833576202,
      "rewards/forensics_sinkhorn_reward": 0.4045754075050354,
      "rewards/forensics_strict_edge_reward": 0.13213112950325012,
      "rewards/forensics_tiou_reward": 0.18539617955684662,
      "step": 2
    },
    {
      "completion_length": 24.953125,
      "epoch": 0.007989347536617843,
      "grad_norm": 10.11056137084961,
      "kl": 0.0051422119140625,
      "learning_rate": 9.986666666666667e-07,
      "loss": 0.1049,
      "reward": 0.46321001648902893,
      "reward_std": 0.23060308396816254,
      "reward_std_all": 0.2748890593647957,
      "rewards-max/forensics_sinkhorn_reward": 0.8398863673210144,
      "rewards-max/forensics_strict_edge_reward": 0.7023755609989166,
      "rewards-max/forensics_tiou_reward": 0.6887232661247253,
      "rewards/forensics_sinkhorn_reward": 0.35593292117118835,
      "rewards/forensics_strict_edge_reward": 0.0785924680531025,
      "rewards/forensics_tiou_reward": 0.13195902109146118,
      "step": 3
    },
    {
      "completion_length": 19.34375,
      "epoch": 0.010652463382157125,
      "grad_norm": 11.115805625915527,
      "kl": 0.0071868896484375,
      "learning_rate": 9.98e-07,
      "loss": 0.0655,
      "reward": 0.46068069338798523,
      "reward_std": 0.17519566416740417,
      "reward_std_all": 0.24144987761974335,
      "rewards-max/forensics_sinkhorn_reward": 0.779057651758194,
      "rewards-max/forensics_strict_edge_reward": 0.40168769285082817,
      "rewards-max/forensics_tiou_reward": 0.5777097791433334,
      "rewards/forensics_sinkhorn_reward": 0.3283158242702484,
      "rewards/forensics_strict_edge_reward": 0.03145818295888603,
      "rewards/forensics_tiou_reward": 0.0599687360227108,
      "step": 4
    },
    {
      "completion_length": 16.390625,
      "epoch": 0.013315579227696404,
      "grad_norm": 10.53939151763916,
      "kl": 0.0090179443359375,
      "learning_rate": 9.973333333333332e-07,
      "loss": 0.0381,
      "reward": 0.5218296647071838,
      "reward_std": 0.2377341389656067,
      "reward_std_all": 0.24891460686922073,
      "rewards-max/forensics_sinkhorn_reward": 0.9834356307983398,
      "rewards-max/forensics_strict_edge_reward": 0.8025174736976624,
      "rewards-max/forensics_tiou_reward": 0.9781380295753479,
      "rewards/forensics_sinkhorn_reward": 0.4103679358959198,
      "rewards/forensics_strict_edge_reward": 0.12051121145486832,
      "rewards/forensics_tiou_reward": 0.1821586713194847,
      "step": 5
    },
    {
      "completion_length": 17.953125,
      "epoch": 0.015978695073235686,
      "grad_norm": 7.783426761627197,
      "kl": 0.02490234375,
      "learning_rate": 9.966666666666667e-07,
      "loss": 0.1638,
      "reward": 0.4452627897262573,
      "reward_std": 0.2299928218126297,
      "reward_std_all": 0.2711813300848007,
      "rewards-max/forensics_sinkhorn_reward": 0.9787635803222656,
      "rewards-max/forensics_strict_edge_reward": 0.7514884769916534,
      "rewards-max/forensics_tiou_reward": 0.960509717464447,
      "rewards/forensics_sinkhorn_reward": 0.3471182584762573,
      "rewards/forensics_strict_edge_reward": 0.0744602344930172,
      "rewards/forensics_tiou_reward": 0.127113476395607,
      "step": 6
    },
    {
      "completion_length": 19.609375,
      "epoch": 0.018641810918774968,
      "grad_norm": 8.041932106018066,
      "kl": 0.020751953125,
      "learning_rate": 9.959999999999999e-07,
      "loss": 0.1299,
      "reward": 0.5155395865440369,
      "reward_std": 0.17948758602142334,
      "reward_std_all": 0.2170572653412819,
      "rewards-max/forensics_sinkhorn_reward": 0.9875972867012024,
      "rewards-max/forensics_strict_edge_reward": 0.9265981018543243,
      "rewards-max/forensics_tiou_reward": 0.9945169389247894,
      "rewards/forensics_sinkhorn_reward": 0.4073195159435272,
      "rewards/forensics_strict_edge_reward": 0.10304785519838333,
      "rewards/forensics_tiou_reward": 0.1436779648065567,
      "step": 7
    },
    {
      "completion_length": 21.484375,
      "epoch": 0.02130492676431425,
      "grad_norm": 6.8961052894592285,
      "kl": 0.0509033203125,
      "learning_rate": 9.953333333333332e-07,
      "loss": -0.0084,
      "reward": 0.5199030190706253,
      "reward_std": 0.1826605424284935,
      "reward_std_all": 0.21658708155155182,
      "rewards-max/forensics_sinkhorn_reward": 0.9845453202724457,
      "rewards-max/forensics_strict_edge_reward": 0.87538942694664,
      "rewards-max/forensics_tiou_reward": 0.9824892580509186,
      "rewards/forensics_sinkhorn_reward": 0.3933893144130707,
      "rewards/forensics_strict_edge_reward": 0.06631532125174999,
      "rewards/forensics_tiou_reward": 0.12760280817747116,
      "step": 8
    },
    {
      "completion_length": 17.140625,
      "epoch": 0.023968042609853527,
      "grad_norm": 9.21143913269043,
      "kl": 0.02154541015625,
      "learning_rate": 9.946666666666666e-07,
      "loss": 0.0016,
      "reward": 0.5432040095329285,
      "reward_std": 0.17328565567731857,
      "reward_std_all": 0.20720140635967255,
      "rewards-max/forensics_sinkhorn_reward": 0.984515368938446,
      "rewards-max/forensics_strict_edge_reward": 0.9048932790756226,
      "rewards-max/forensics_tiou_reward": 0.9823990762233734,
      "rewards/forensics_sinkhorn_reward": 0.44202756881713867,
      "rewards/forensics_strict_edge_reward": 0.13317842781543732,
      "rewards/forensics_tiou_reward": 0.20309347659349442,
      "step": 9
    },
    {
      "completion_length": 15.703125,
      "epoch": 0.02663115845539281,
      "grad_norm": 7.609653949737549,
      "kl": 0.05078125,
      "learning_rate": 9.94e-07,
      "loss": 0.1474,
      "reward": 0.5336731374263763,
      "reward_std": 0.19808345288038254,
      "reward_std_all": 0.21776142716407776,
      "rewards-max/forensics_sinkhorn_reward": 0.9733379185199738,
      "rewards-max/forensics_strict_edge_reward": 0.6643738597631454,
      "rewards-max/forensics_tiou_reward": 0.9402096569538116,
      "rewards/forensics_sinkhorn_reward": 0.4341726750135422,
      "rewards/forensics_strict_edge_reward": 0.0814566072076559,
      "rewards/forensics_tiou_reward": 0.18328393250703812,
      "step": 10
    },
    {
      "completion_length": 17.09375,
      "epoch": 0.02929427430093209,
      "grad_norm": 7.67330265045166,
      "kl": 0.0537109375,
      "learning_rate": 9.933333333333333e-07,
      "loss": 0.005,
      "reward": 0.5450533628463745,
      "reward_std": 0.1856384500861168,
      "reward_std_all": 0.21416932344436646,
      "rewards-max/forensics_sinkhorn_reward": 0.9775280952453613,
      "rewards-max/forensics_strict_edge_reward": 0.7321087718009949,
      "rewards-max/forensics_tiou_reward": 0.9556748270988464,
      "rewards/forensics_sinkhorn_reward": 0.44775256514549255,
      "rewards/forensics_strict_edge_reward": 0.11067913100123405,
      "rewards/forensics_tiou_reward": 0.17874787747859955,
      "step": 11
    },
    {
      "completion_length": 17.0625,
      "epoch": 0.03195739014647137,
      "grad_norm": 6.808846473693848,
      "kl": 0.0819091796875,
      "learning_rate": 9.926666666666666e-07,
      "loss": -0.051,
      "reward": 0.594014585018158,
      "reward_std": 0.12121974304318428,
      "reward_std_all": 0.15896928310394287,
      "rewards-max/forensics_sinkhorn_reward": 0.9811396300792694,
      "rewards-max/forensics_strict_edge_reward": 0.856150895357132,
      "rewards-max/forensics_tiou_reward": 0.9693707227706909,
      "rewards/forensics_sinkhorn_reward": 0.5096725523471832,
      "rewards/forensics_strict_edge_reward": 0.1126391664147377,
      "rewards/forensics_tiou_reward": 0.20488902926445007,
      "step": 12
    },
    {
      "completion_length": 14.984375,
      "epoch": 0.03462050599201065,
      "grad_norm": 8.500831604003906,
      "kl": 0.161376953125,
      "learning_rate": 9.92e-07,
      "loss": 0.0547,
      "reward": 0.5691072642803192,
      "reward_std": 0.1260806955397129,
      "reward_std_all": 0.16554372757673264,
      "rewards-max/forensics_sinkhorn_reward": 0.8476171493530273,
      "rewards-max/forensics_strict_edge_reward": 0.46170242317020893,
      "rewards-max/forensics_tiou_reward": 0.6833895891904831,
      "rewards/forensics_sinkhorn_reward": 0.46117326617240906,
      "rewards/forensics_strict_edge_reward": 0.058489357761573046,
      "rewards/forensics_tiou_reward": 0.12379583716392517,
      "step": 13
    },
    {
      "completion_length": 20.3125,
      "epoch": 0.037283621837549935,
      "grad_norm": 9.45209789276123,
      "kl": 0.118408203125,
      "learning_rate": 9.913333333333333e-07,
      "loss": -0.0966,
      "reward": 0.5817526280879974,
      "reward_std": 0.1479472517967224,
      "reward_std_all": 0.17589227855205536,
      "rewards-max/forensics_sinkhorn_reward": 0.9862816631793976,
      "rewards-max/forensics_strict_edge_reward": 0.6886534988880157,
      "rewards-max/forensics_tiou_reward": 0.9893633425235748,
      "rewards/forensics_sinkhorn_reward": 0.4617370218038559,
      "rewards/forensics_strict_edge_reward": 0.05362704582512379,
      "rewards/forensics_tiou_reward": 0.14887719601392746,
      "step": 14
    },
    {
      "completion_length": 18.390625,
      "epoch": 0.03994673768308921,
      "grad_norm": 5.719789028167725,
      "kl": 0.0340576171875,
      "learning_rate": 9.906666666666667e-07,
      "loss": -0.0174,
      "reward": 0.5899978876113892,
      "reward_std": 0.1373818814754486,
      "reward_std_all": 0.1708868443965912,
      "rewards-max/forensics_sinkhorn_reward": 0.9570292532444,
      "rewards-max/forensics_strict_edge_reward": 0.4223303720355034,
      "rewards-max/forensics_tiou_reward": 0.8797259032726288,
      "rewards/forensics_sinkhorn_reward": 0.4846782088279724,
      "rewards/forensics_strict_edge_reward": 0.04261419177055359,
      "rewards/forensics_tiou_reward": 0.14913446083664894,
      "step": 15
    },
    {
      "completion_length": 18.546875,
      "epoch": 0.0426098535286285,
      "grad_norm": 7.66411018371582,
      "kl": 0.09619140625,
      "learning_rate": 9.9e-07,
      "loss": 0.1099,
      "reward": 0.5428480207920074,
      "reward_std": 0.15787605941295624,
      "reward_std_all": 0.20047824084758759,
      "rewards-max/forensics_sinkhorn_reward": 0.9835960865020752,
      "rewards-max/forensics_strict_edge_reward": 0.856150895357132,
      "rewards-max/forensics_tiou_reward": 0.978856235742569,
      "rewards/forensics_sinkhorn_reward": 0.4248253256082535,
      "rewards/forensics_strict_edge_reward": 0.08928800374269485,
      "rewards/forensics_tiou_reward": 0.12562625855207443,
      "step": 16
    },
    {
      "completion_length": 14.984375,
      "epoch": 0.045272969374167776,
      "grad_norm": 8.69349193572998,
      "kl": 0.09912109375,
      "learning_rate": 9.893333333333332e-07,
      "loss": 0.1319,
      "reward": 0.6043495535850525,
      "reward_std": 0.156044602394104,
      "reward_std_all": 0.17739573866128922,
      "rewards-max/forensics_sinkhorn_reward": 0.9792869091033936,
      "rewards-max/forensics_strict_edge_reward": 0.8279296159744263,
      "rewards-max/forensics_tiou_reward": 0.9620246887207031,
      "rewards/forensics_sinkhorn_reward": 0.5241909325122833,
      "rewards/forensics_strict_edge_reward": 0.11916235461831093,
      "rewards/forensics_tiou_reward": 0.24396689236164093,
      "step": 17
    },
    {
      "completion_length": 18.453125,
      "epoch": 0.047936085219707054,
      "grad_norm": 6.6025590896606445,
      "kl": 0.102294921875,
      "learning_rate": 9.886666666666665e-07,
      "loss": 0.1703,
      "reward": 0.5729065537452698,
      "reward_std": 0.13117582350969315,
      "reward_std_all": 0.1705784574151039,
      "rewards-max/forensics_sinkhorn_reward": 0.9862249493598938,
      "rewards-max/forensics_strict_edge_reward": 0.7187243551015854,
      "rewards-max/forensics_tiou_reward": 0.9891485571861267,
      "rewards/forensics_sinkhorn_reward": 0.48259446024894714,
      "rewards/forensics_strict_edge_reward": 0.11970033496618271,
      "rewards/forensics_tiou_reward": 0.20366672426462173,
      "step": 18
    },
    {
      "completion_length": 17.453125,
      "epoch": 0.05059920106524634,
      "grad_norm": 7.298027992248535,
      "kl": 0.068603515625,
      "learning_rate": 9.88e-07,
      "loss": 0.0477,
      "reward": 0.5728516280651093,
      "reward_std": 0.15258736163377762,
      "reward_std_all": 0.18551771342754364,
      "rewards-max/forensics_sinkhorn_reward": 0.977234810590744,
      "rewards-max/forensics_strict_edge_reward": 0.7507607042789459,
      "rewards-max/forensics_tiou_reward": 0.9542609751224518,
      "rewards/forensics_sinkhorn_reward": 0.44103506207466125,
      "rewards/forensics_strict_edge_reward": 0.10773840546607971,
      "rewards/forensics_tiou_reward": 0.17716073244810104,
      "step": 19
    },
    {
      "completion_length": 26.609375,
      "epoch": 0.05326231691078562,
      "grad_norm": 8.791863441467285,
      "kl": 0.0780029296875,
      "learning_rate": 9.873333333333333e-07,
      "loss": -0.0277,
      "reward": 0.5452322363853455,
      "reward_std": 0.19527539610862732,
      "reward_std_all": 0.21627026051282883,
      "rewards-max/forensics_sinkhorn_reward": 0.9848517775535583,
      "rewards-max/forensics_strict_edge_reward": 0.8863177299499512,
      "rewards-max/forensics_tiou_reward": 0.9838092029094696,
      "rewards/forensics_sinkhorn_reward": 0.4295450448989868,
      "rewards/forensics_strict_edge_reward": 0.09235870093107224,
      "rewards/forensics_tiou_reward": 0.17999781668186188,
      "step": 20
    },
    {
      "completion_length": 15.78125,
      "epoch": 0.0559254327563249,
      "grad_norm": 5.4900078773498535,
      "kl": 0.06591796875,
      "learning_rate": 9.866666666666666e-07,
      "loss": -0.0445,
      "reward": 0.5980218946933746,
      "reward_std": 0.1348695494234562,
      "reward_std_all": 0.1631237417459488,
      "rewards-max/forensics_sinkhorn_reward": 0.8541437685489655,
      "rewards-max/forensics_strict_edge_reward": 0.4760446548461914,
      "rewards-max/forensics_tiou_reward": 0.6878986954689026,
      "rewards/forensics_sinkhorn_reward": 0.47453394532203674,
      "rewards/forensics_strict_edge_reward": 0.06288716103881598,
      "rewards/forensics_tiou_reward": 0.14452947303652763,
      "step": 21
    },
    {
      "completion_length": 30.25,
      "epoch": 0.05858854860186418,
      "grad_norm": 8.303208351135254,
      "kl": 0.103515625,
      "learning_rate": 9.86e-07,
      "loss": 0.0634,
      "reward": 0.5474997013807297,
      "reward_std": 0.2218603491783142,
      "reward_std_all": 0.23214779049158096,
      "rewards-max/forensics_sinkhorn_reward": 0.9847613573074341,
      "rewards-max/forensics_strict_edge_reward": 0.6149742305278778,
      "rewards-max/forensics_tiou_reward": 0.9834027886390686,
      "rewards/forensics_sinkhorn_reward": 0.4740086793899536,
      "rewards/forensics_strict_edge_reward": 0.11860586702823639,
      "rewards/forensics_tiou_reward": 0.24515466019511223,
      "step": 22
    },
    {
      "completion_length": 18.078125,
      "epoch": 0.06125166444740346,
      "grad_norm": 2190.119873046875,
      "kl": 39.27197265625,
      "learning_rate": 9.853333333333333e-07,
      "loss": 1.676,
      "reward": 0.5256841629743576,
      "reward_std": 0.16766901314258575,
      "reward_std_all": 0.21390779316425323,
      "rewards-max/forensics_sinkhorn_reward": 0.8484093248844147,
      "rewards-max/forensics_strict_edge_reward": 0.506361348554492,
      "rewards-max/forensics_tiou_reward": 0.7874421179294586,
      "rewards/forensics_sinkhorn_reward": 0.41348688304424286,
      "rewards/forensics_strict_edge_reward": 0.07989034539787099,
      "rewards/forensics_tiou_reward": 0.15140294283628464,
      "step": 23
    },
    {
      "completion_length": 20.71875,
      "epoch": 0.06391478029294274,
      "grad_norm": 65.95122528076172,
      "kl": 1.46484375,
      "learning_rate": 9.846666666666667e-07,
      "loss": -0.0063,
      "reward": 0.5664458870887756,
      "reward_std": 0.23702630400657654,
      "reward_std_all": 0.231249138712883,
      "rewards-max/forensics_sinkhorn_reward": 0.9649512767791748,
      "rewards-max/forensics_strict_edge_reward": 0.6340309977531433,
      "rewards-max/forensics_tiou_reward": 0.9083749651908875,
      "rewards/forensics_sinkhorn_reward": 0.42730847001075745,
      "rewards/forensics_strict_edge_reward": 0.10669314488768578,
      "rewards/forensics_tiou_reward": 0.19024378806352615,
      "step": 24
    },
    {
      "completion_length": 18.734375,
      "epoch": 0.06657789613848203,
      "grad_norm": 6.479513168334961,
      "kl": 0.19677734375,
      "learning_rate": 9.84e-07,
      "loss": 0.0572,
      "reward": 0.5782392621040344,
      "reward_std": 0.18927911669015884,
      "reward_std_all": 0.19922681897878647,
      "rewards-max/forensics_sinkhorn_reward": 0.9842584133148193,
      "rewards-max/forensics_strict_edge_reward": 0.8426913917064667,
      "rewards-max/forensics_tiou_reward": 0.9815333783626556,
      "rewards/forensics_sinkhorn_reward": 0.4665991961956024,
      "rewards/forensics_strict_edge_reward": 0.09222216531634331,
      "rewards/forensics_tiou_reward": 0.20285607874393463,
      "step": 25
    },
    {
      "completion_length": 20.0,
      "epoch": 0.0692410119840213,
      "grad_norm": 6.174808025360107,
      "kl": 0.1134033203125,
      "learning_rate": 9.833333333333332e-07,
      "loss": 0.118,
      "reward": 0.5551607012748718,
      "reward_std": 0.19269869476556778,
      "reward_std_all": 0.22274817526340485,
      "rewards-max/forensics_sinkhorn_reward": 0.935801774263382,
      "rewards-max/forensics_strict_edge_reward": 0.6513393223285675,
      "rewards-max/forensics_tiou_reward": 0.8660163581371307,
      "rewards/forensics_sinkhorn_reward": 0.4470242112874985,
      "rewards/forensics_strict_edge_reward": 0.09175489097833633,
      "rewards/forensics_tiou_reward": 0.1824454367160797,
      "step": 26
    },
    {
      "completion_length": 14.703125,
      "epoch": 0.07190412782956059,
      "grad_norm": 14.41097640991211,
      "kl": 0.350341796875,
      "learning_rate": 9.826666666666667e-07,
      "loss": -0.1163,
      "reward": 0.6030694246292114,
      "reward_std": 0.12141401320695877,
      "reward_std_all": 0.1512317806482315,
      "rewards-max/forensics_sinkhorn_reward": 0.817534863948822,
      "rewards-max/forensics_strict_edge_reward": 0.5664886832237244,
      "rewards-max/forensics_tiou_reward": 0.6542260646820068,
      "rewards/forensics_sinkhorn_reward": 0.4777335524559021,
      "rewards/forensics_strict_edge_reward": 0.05566445551812649,
      "rewards/forensics_tiou_reward": 0.10650096088647842,
      "step": 27
    },
    {
      "completion_length": 19.796875,
      "epoch": 0.07456724367509987,
      "grad_norm": 5.736626625061035,
      "kl": 0.171142578125,
      "learning_rate": 9.819999999999999e-07,
      "loss": 0.0066,
      "reward": 0.5663455724716187,
      "reward_std": 0.15034855157136917,
      "reward_std_all": 0.18235668540000916,
      "rewards-max/forensics_sinkhorn_reward": 0.9826925098896027,
      "rewards-max/forensics_strict_edge_reward": 0.8593232333660126,
      "rewards-max/forensics_tiou_reward": 0.9754151701927185,
      "rewards/forensics_sinkhorn_reward": 0.4464074969291687,
      "rewards/forensics_strict_edge_reward": 0.07941146939992905,
      "rewards/forensics_tiou_reward": 0.1662084236741066,
      "step": 28
    },
    {
      "completion_length": 19.578125,
      "epoch": 0.07723035952063914,
      "grad_norm": 10.922316551208496,
      "kl": 0.30810546875,
      "learning_rate": 9.813333333333332e-07,
      "loss": -0.1021,
      "reward": 0.540280669927597,
      "reward_std": 0.11480604857206345,
      "reward_std_all": 0.17880165576934814,
      "rewards-max/forensics_sinkhorn_reward": 0.8204984664916992,
      "rewards-max/forensics_strict_edge_reward": 0.5664832070469856,
      "rewards-max/forensics_tiou_reward": 0.6611384302377701,
      "rewards/forensics_sinkhorn_reward": 0.41748087108135223,
      "rewards/forensics_strict_edge_reward": 0.03509402275085449,
      "rewards/forensics_tiou_reward": 0.06969364732503891,
      "step": 29
    },
    {
      "completion_length": 16.9375,
      "epoch": 0.07989347536617843,
      "grad_norm": 10.992301940917969,
      "kl": 0.3681640625,
      "learning_rate": 9.806666666666666e-07,
      "loss": 0.1008,
      "reward": 0.5475892424583435,
      "reward_std": 0.1473541222512722,
      "reward_std_all": 0.1890474334359169,
      "rewards-max/forensics_sinkhorn_reward": 0.9834068417549133,
      "rewards-max/forensics_strict_edge_reward": 0.9075756669044495,
      "rewards-max/forensics_tiou_reward": 0.9782311022281647,
      "rewards/forensics_sinkhorn_reward": 0.4477708488702774,
      "rewards/forensics_strict_edge_reward": 0.10528717190027237,
      "rewards/forensics_tiou_reward": 0.1576320081949234,
      "step": 30
    },
    {
      "completion_length": 14.796875,
      "epoch": 0.08255659121171771,
      "grad_norm": 9.813607215881348,
      "kl": 0.131103515625,
      "learning_rate": 9.8e-07,
      "loss": 0.03,
      "reward": 0.5701765716075897,
      "reward_std": 0.1786087453365326,
      "reward_std_all": 0.19464410841464996,
      "rewards-max/forensics_sinkhorn_reward": 0.7982874512672424,
      "rewards-max/forensics_strict_edge_reward": 0.2799733653664589,
      "rewards-max/forensics_tiou_reward": 0.5974356681108475,
      "rewards/forensics_sinkhorn_reward": 0.432478129863739,
      "rewards/forensics_strict_edge_reward": 0.02321901172399521,
      "rewards/forensics_tiou_reward": 0.10521948337554932,
      "step": 31
    },
    {
      "completion_length": 14.703125,
      "epoch": 0.085219707057257,
      "grad_norm": 11.04999828338623,
      "kl": 0.296142578125,
      "learning_rate": 9.793333333333333e-07,
      "loss": 0.1877,
      "reward": 0.6117554903030396,
      "reward_std": 0.1322607472538948,
      "reward_std_all": 0.15357864648103714,
      "rewards-max/forensics_sinkhorn_reward": 0.9759586751461029,
      "rewards-max/forensics_strict_edge_reward": 0.8039028942584991,
      "rewards-max/forensics_tiou_reward": 0.9495926797389984,
      "rewards/forensics_sinkhorn_reward": 0.505751371383667,
      "rewards/forensics_strict_edge_reward": 0.10793902724981308,
      "rewards/forensics_tiou_reward": 0.17293140292167664,
      "step": 32
    },
    {
      "completion_length": 16.8125,
      "epoch": 0.08788282290279627,
      "grad_norm": 18.60615348815918,
      "kl": 0.892333984375,
      "learning_rate": 9.786666666666666e-07,
      "loss": 0.0138,
      "reward": 0.5228902995586395,
      "reward_std": 0.11979753524065018,
      "reward_std_all": 0.191026970744133,
      "rewards-max/forensics_sinkhorn_reward": 0.8717826008796692,
      "rewards-max/forensics_strict_edge_reward": 0.7783617675304413,
      "rewards-max/forensics_tiou_reward": 0.7647337913513184,
      "rewards/forensics_sinkhorn_reward": 0.4031356871128082,
      "rewards/forensics_strict_edge_reward": 0.08310506120324135,
      "rewards/forensics_tiou_reward": 0.11253763362765312,
      "step": 33
    },
    {
      "completion_length": 16.734375,
      "epoch": 0.09054593874833555,
      "grad_norm": 7.8803534507751465,
      "kl": 0.093994140625,
      "learning_rate": 9.78e-07,
      "loss": 0.066,
      "reward": 0.6216742098331451,
      "reward_std": 0.13689129799604416,
      "reward_std_all": 0.15666251629590988,
      "rewards-max/forensics_sinkhorn_reward": 0.9815513491630554,
      "rewards-max/forensics_strict_edge_reward": 0.820955753326416,
      "rewards-max/forensics_tiou_reward": 0.9710539281368256,
      "rewards/forensics_sinkhorn_reward": 0.4804246723651886,
      "rewards/forensics_strict_edge_reward": 0.10335594788193703,
      "rewards/forensics_tiou_reward": 0.14536026120185852,
      "step": 34
    },
    {
      "completion_length": 15.859375,
      "epoch": 0.09320905459387484,
      "grad_norm": 13.341156005859375,
      "kl": 0.14453125,
      "learning_rate": 9.773333333333333e-07,
      "loss": 0.134,
      "reward": 0.5303187966346741,
      "reward_std": 0.13473792374134064,
      "reward_std_all": 0.189279705286026,
      "rewards-max/forensics_sinkhorn_reward": 0.9737036228179932,
      "rewards-max/forensics_strict_edge_reward": 0.7466395199298859,
      "rewards-max/forensics_tiou_reward": 0.9398985207080841,
      "rewards/forensics_sinkhorn_reward": 0.4045257568359375,
      "rewards/forensics_strict_edge_reward": 0.06906820833683014,
      "rewards/forensics_tiou_reward": 0.1101025864481926,
      "step": 35
    },
    {
      "completion_length": 16.234375,
      "epoch": 0.09587217043941411,
      "grad_norm": 7.105246067047119,
      "kl": 0.12939453125,
      "learning_rate": 9.766666666666667e-07,
      "loss": -0.0145,
      "reward": 0.5484877526760101,
      "reward_std": 0.1974329650402069,
      "reward_std_all": 0.21806228905916214,
      "rewards-max/forensics_sinkhorn_reward": 0.9754278659820557,
      "rewards-max/forensics_strict_edge_reward": 0.7540734112262726,
      "rewards-max/forensics_tiou_reward": 0.9477841258049011,
      "rewards/forensics_sinkhorn_reward": 0.4274825006723404,
      "rewards/forensics_strict_edge_reward": 0.0570477731525898,
      "rewards/forensics_tiou_reward": 0.1244678646326065,
      "step": 36
    },
    {
      "completion_length": 17.265625,
      "epoch": 0.0985352862849534,
      "grad_norm": 7.880741596221924,
      "kl": 0.2236328125,
      "learning_rate": 9.759999999999998e-07,
      "loss": -0.0275,
      "reward": 0.595670223236084,
      "reward_std": 0.16836226731538773,
      "reward_std_all": 0.1770133599638939,
      "rewards-max/forensics_sinkhorn_reward": 0.9408888816833496,
      "rewards-max/forensics_strict_edge_reward": 0.684678316116333,
      "rewards-max/forensics_tiou_reward": 0.8849721252918243,
      "rewards/forensics_sinkhorn_reward": 0.47457699477672577,
      "rewards/forensics_strict_edge_reward": 0.0731218010187149,
      "rewards/forensics_tiou_reward": 0.16579332202672958,
      "step": 37
    },
    {
      "completion_length": 17.4375,
      "epoch": 0.10119840213049268,
      "grad_norm": 5.864415645599365,
      "kl": 0.101318359375,
      "learning_rate": 9.753333333333334e-07,
      "loss": -0.0289,
      "reward": 0.6076243817806244,
      "reward_std": 0.16427645832300186,
      "reward_std_all": 0.1757229119539261,
      "rewards-max/forensics_sinkhorn_reward": 0.9850141108036041,
      "rewards-max/forensics_strict_edge_reward": 0.8957232534885406,
      "rewards-max/forensics_tiou_reward": 0.9843927323818207,
      "rewards/forensics_sinkhorn_reward": 0.5115897357463837,
      "rewards/forensics_strict_edge_reward": 0.148214440792799,
      "rewards/forensics_tiou_reward": 0.2388000786304474,
      "step": 38
    },
    {
      "completion_length": 16.71875,
      "epoch": 0.10386151797603196,
      "grad_norm": 7.960866451263428,
      "kl": 0.1376953125,
      "learning_rate": 9.746666666666666e-07,
      "loss": -0.1157,
      "reward": 0.5995885133743286,
      "reward_std": 0.13927236571907997,
      "reward_std_all": 0.16964086145162582,
      "rewards-max/forensics_sinkhorn_reward": 0.9867011606693268,
      "rewards-max/forensics_strict_edge_reward": 0.9355069994926453,
      "rewards-max/forensics_tiou_reward": 0.9909894168376923,
      "rewards/forensics_sinkhorn_reward": 0.5123098343610764,
      "rewards/forensics_strict_edge_reward": 0.13997873663902283,
      "rewards/forensics_tiou_reward": 0.24005284905433655,
      "step": 39
    },
    {
      "completion_length": 18.78125,
      "epoch": 0.10652463382157124,
      "grad_norm": 6.372307300567627,
      "kl": 0.139404296875,
      "learning_rate": 9.74e-07,
      "loss": 0.1319,
      "reward": 0.5468582510948181,
      "reward_std": 0.1400267854332924,
      "reward_std_all": 0.18642199784517288,
      "rewards-max/forensics_sinkhorn_reward": 0.9867039322853088,
      "rewards-max/forensics_strict_edge_reward": 0.8863177299499512,
      "rewards-max/forensics_tiou_reward": 0.9910095930099487,
      "rewards/forensics_sinkhorn_reward": 0.4197252243757248,
      "rewards/forensics_strict_edge_reward": 0.10549904778599739,
      "rewards/forensics_tiou_reward": 0.15031521022319794,
      "step": 40
    },
    {
      "completion_length": 22.796875,
      "epoch": 0.10918774966711052,
      "grad_norm": 8.567546844482422,
      "kl": 0.375,
      "learning_rate": 9.733333333333333e-07,
      "loss": 0.0046,
      "reward": 0.5222525000572205,
      "reward_std": 0.21904606372117996,
      "reward_std_all": 0.23882505297660828,
      "rewards-max/forensics_sinkhorn_reward": 0.9675409197807312,
      "rewards-max/forensics_strict_edge_reward": 0.5755094587802887,
      "rewards-max/forensics_tiou_reward": 0.918191134929657,
      "rewards/forensics_sinkhorn_reward": 0.40554192662239075,
      "rewards/forensics_strict_edge_reward": 0.054424453526735306,
      "rewards/forensics_tiou_reward": 0.15749064087867737,
      "step": 41
    },
    {
      "completion_length": 20.171875,
      "epoch": 0.1118508655126498,
      "grad_norm": 13.328232765197754,
      "kl": 0.42822265625,
      "learning_rate": 9.726666666666666e-07,
      "loss": 0.1434,
      "reward": 0.5134193599224091,
      "reward_std": 0.21779754757881165,
      "reward_std_all": 0.24307453632354736,
      "rewards-max/forensics_sinkhorn_reward": 0.8932319283485413,
      "rewards-max/forensics_strict_edge_reward": 0.5530991777777672,
      "rewards-max/forensics_tiou_reward": 0.784642219543457,
      "rewards/forensics_sinkhorn_reward": 0.3780360668897629,
      "rewards/forensics_strict_edge_reward": 0.05664666183292866,
      "rewards/forensics_tiou_reward": 0.1416548304259777,
      "step": 42
    },
    {
      "completion_length": 17.9375,
      "epoch": 0.11451398135818908,
      "grad_norm": 6.969388008117676,
      "kl": 0.17626953125,
      "learning_rate": 9.72e-07,
      "loss": 0.0854,
      "reward": 0.5935494899749756,
      "reward_std": 0.18043086677789688,
      "reward_std_all": 0.18818584084510803,
      "rewards-max/forensics_sinkhorn_reward": 0.9431585669517517,
      "rewards-max/forensics_strict_edge_reward": 0.6515887677669525,
      "rewards-max/forensics_tiou_reward": 0.8886676132678986,
      "rewards/forensics_sinkhorn_reward": 0.4605676233768463,
      "rewards/forensics_strict_edge_reward": 0.08754175156354904,
      "rewards/forensics_tiou_reward": 0.16131430119276047,
      "step": 43
    },
    {
      "completion_length": 17.3125,
      "epoch": 0.11717709720372836,
      "grad_norm": 6.951797962188721,
      "kl": 0.20849609375,
      "learning_rate": 9.713333333333333e-07,
      "loss": 0.1608,
      "reward": 0.5216935873031616,
      "reward_std": 0.1942300945520401,
      "reward_std_all": 0.22627532482147217,
      "rewards-max/forensics_sinkhorn_reward": 0.9241812229156494,
      "rewards-max/forensics_strict_edge_reward": 0.6574696898460388,
      "rewards-max/forensics_tiou_reward": 0.7735942900180817,
      "rewards/forensics_sinkhorn_reward": 0.4040602445602417,
      "rewards/forensics_strict_edge_reward": 0.09021345525979996,
      "rewards/forensics_tiou_reward": 0.1513233706355095,
      "step": 44
    },
    {
      "completion_length": 16.34375,
      "epoch": 0.11984021304926765,
      "grad_norm": 28.40485954284668,
      "kl": 1.59765625,
      "learning_rate": 9.706666666666667e-07,
      "loss": 0.1029,
      "reward": 0.590306967496872,
      "reward_std": 0.16696462780237198,
      "reward_std_all": 0.18497779965400696,
      "rewards-max/forensics_sinkhorn_reward": 0.9790017604827881,
      "rewards-max/forensics_strict_edge_reward": 0.713913232088089,
      "rewards-max/forensics_tiou_reward": 0.9606557190418243,
      "rewards/forensics_sinkhorn_reward": 0.4826332926750183,
      "rewards/forensics_strict_edge_reward": 0.125225268304348,
      "rewards/forensics_tiou_reward": 0.20677559077739716,
      "step": 45
    },
    {
      "completion_length": 16.21875,
      "epoch": 0.12250332889480692,
      "grad_norm": 12.2609281539917,
      "kl": 0.5263671875,
      "learning_rate": 9.7e-07,
      "loss": 0.1501,
      "reward": 0.5670871436595917,
      "reward_std": 0.16359052062034607,
      "reward_std_all": 0.18785345554351807,
      "rewards-max/forensics_sinkhorn_reward": 0.923775851726532,
      "rewards-max/forensics_strict_edge_reward": 0.6632815897464752,
      "rewards-max/forensics_tiou_reward": 0.8460819721221924,
      "rewards/forensics_sinkhorn_reward": 0.43150076270103455,
      "rewards/forensics_strict_edge_reward": 0.062235256657004356,
      "rewards/forensics_tiou_reward": 0.126576766371727,
      "step": 46
    },
    {
      "completion_length": 16.953125,
      "epoch": 0.12516644474034622,
      "grad_norm": 8.887773513793945,
      "kl": 0.127685546875,
      "learning_rate": 9.693333333333334e-07,
      "loss": 0.1126,
      "reward": 0.615827888250351,
      "reward_std": 0.16168732196092606,
      "reward_std_all": 0.1740669682621956,
      "rewards-max/forensics_sinkhorn_reward": 0.9823220074176788,
      "rewards-max/forensics_strict_edge_reward": 0.820955753326416,
      "rewards-max/forensics_tiou_reward": 0.9737347960472107,
      "rewards/forensics_sinkhorn_reward": 0.523596853017807,
      "rewards/forensics_strict_edge_reward": 0.11729126423597336,
      "rewards/forensics_tiou_reward": 0.2683228775858879,
      "step": 47
    },
    {
      "completion_length": 16.734375,
      "epoch": 0.1278295605858855,
      "grad_norm": 7.0622711181640625,
      "kl": 0.12548828125,
      "learning_rate": 9.686666666666667e-07,
      "loss": 0.1447,
      "reward": 0.5864030122756958,
      "reward_std": 0.16022051870822906,
      "reward_std_all": 0.18041496723890305,
      "rewards-max/forensics_sinkhorn_reward": 0.981318324804306,
      "rewards-max/forensics_strict_edge_reward": 0.8260191679000854,
      "rewards-max/forensics_tiou_reward": 0.9700842201709747,
      "rewards/forensics_sinkhorn_reward": 0.4828140139579773,
      "rewards/forensics_strict_edge_reward": 0.09837046638131142,
      "rewards/forensics_tiou_reward": 0.2054804265499115,
      "step": 48
    },
    {
      "completion_length": 18.890625,
      "epoch": 0.13049267643142476,
      "grad_norm": 7.65331506729126,
      "kl": 0.31103515625,
      "learning_rate": 9.679999999999999e-07,
      "loss": 0.056,
      "reward": 0.5871792137622833,
      "reward_std": 0.1723039373755455,
      "reward_std_all": 0.19412129372358322,
      "rewards-max/forensics_sinkhorn_reward": 0.9819386303424835,
      "rewards-max/forensics_strict_edge_reward": 0.820955753326416,
      "rewards-max/forensics_tiou_reward": 0.9725805222988129,
      "rewards/forensics_sinkhorn_reward": 0.4903780519962311,
      "rewards/forensics_strict_edge_reward": 0.12154188752174377,
      "rewards/forensics_tiou_reward": 0.2634575068950653,
      "step": 49
    },
    {
      "completion_length": 17.125,
      "epoch": 0.13315579227696406,
      "grad_norm": 6.401210308074951,
      "kl": 0.27783203125,
      "learning_rate": 9.673333333333332e-07,
      "loss": 0.151,
      "reward": 0.5435433387756348,
      "reward_std": 0.14310751855373383,
      "reward_std_all": 0.19150426238775253,
      "rewards-max/forensics_sinkhorn_reward": 0.8701649606227875,
      "rewards-max/forensics_strict_edge_reward": 0.3072603642940521,
      "rewards-max/forensics_tiou_reward": 0.705231249332428,
      "rewards/forensics_sinkhorn_reward": 0.3988594263792038,
      "rewards/forensics_strict_edge_reward": 0.03463464695960283,
      "rewards/forensics_tiou_reward": 0.10984815657138824,
      "step": 50
    },
    {
      "completion_length": 14.765625,
      "epoch": 0.13581890812250333,
      "grad_norm": 9.284012794494629,
      "kl": 0.44677734375,
      "learning_rate": 9.666666666666666e-07,
      "loss": 0.1663,
      "reward": 0.6433711349964142,
      "reward_std": 0.09647273644804955,
      "reward_std_all": 0.12366404756903648,
      "rewards-max/forensics_sinkhorn_reward": 0.9501650929450989,
      "rewards-max/forensics_strict_edge_reward": 0.5485865473747253,
      "rewards-max/forensics_tiou_reward": 0.8592914640903473,
      "rewards/forensics_sinkhorn_reward": 0.542092889547348,
      "rewards/forensics_strict_edge_reward": 0.07395035214722157,
      "rewards/forensics_tiou_reward": 0.16930300742387772,
      "step": 51
    },
    {
      "completion_length": 14.53125,
      "epoch": 0.1384820239680426,
      "grad_norm": 15.053791046142578,
      "kl": 0.59716796875,
      "learning_rate": 9.66e-07,
      "loss": 0.0517,
      "reward": 0.682532012462616,
      "reward_std": 0.07362105324864388,
      "reward_std_all": 0.0970255397260189,
      "rewards-max/forensics_sinkhorn_reward": 0.9871909618377686,
      "rewards-max/forensics_strict_edge_reward": 0.9375866651535034,
      "rewards-max/forensics_tiou_reward": 0.9929329454898834,
      "rewards/forensics_sinkhorn_reward": 0.6107509732246399,
      "rewards/forensics_strict_edge_reward": 0.10403987020254135,
      "rewards/forensics_tiou_reward": 0.24074533581733704,
      "step": 52
    },
    {
      "completion_length": 14.796875,
      "epoch": 0.1411451398135819,
      "grad_norm": 13.389613151550293,
      "kl": 0.9521484375,
      "learning_rate": 9.653333333333333e-07,
      "loss": 0.0072,
      "reward": 0.6256192326545715,
      "reward_std": 0.10905022919178009,
      "reward_std_all": 0.13433687388896942,
      "rewards-max/forensics_sinkhorn_reward": 0.9789450466632843,
      "rewards-max/forensics_strict_edge_reward": 0.8365300297737122,
      "rewards-max/forensics_tiou_reward": 0.96109938621521,
      "rewards/forensics_sinkhorn_reward": 0.5062458515167236,
      "rewards/forensics_strict_edge_reward": 0.08128317818045616,
      "rewards/forensics_tiou_reward": 0.15189103782176971,
      "step": 53
    },
    {
      "completion_length": 15.609375,
      "epoch": 0.14380825565912117,
      "grad_norm": 9.74947452545166,
      "kl": 0.18798828125,
      "learning_rate": 9.646666666666666e-07,
      "loss": -0.0434,
      "reward": 0.6040829718112946,
      "reward_std": 0.08999990299344063,
      "reward_std_all": 0.1389729082584381,
      "rewards-max/forensics_sinkhorn_reward": 0.9850082099437714,
      "rewards-max/forensics_strict_edge_reward": 0.8876297473907471,
      "rewards-max/forensics_tiou_reward": 0.9843829274177551,
      "rewards/forensics_sinkhorn_reward": 0.5163671672344208,
      "rewards/forensics_strict_edge_reward": 0.10008637234568596,
      "rewards/forensics_tiou_reward": 0.18917445093393326,
      "step": 54
    },
    {
      "completion_length": 15.09375,
      "epoch": 0.14647137150466044,
      "grad_norm": 7.782678127288818,
      "kl": 0.28564453125,
      "learning_rate": 9.64e-07,
      "loss": -0.2008,
      "reward": 0.6494394242763519,
      "reward_std": 0.1314561367034912,
      "reward_std_all": 0.14319942146539688,
      "rewards-max/forensics_sinkhorn_reward": 0.9842129945755005,
      "rewards-max/forensics_strict_edge_reward": 0.7328133881092072,
      "rewards-max/forensics_tiou_reward": 0.9812710583209991,
      "rewards/forensics_sinkhorn_reward": 0.5563969910144806,
      "rewards/forensics_strict_edge_reward": 0.07152920588850975,
      "rewards/forensics_tiou_reward": 0.22843392193317413,
      "step": 55
    },
    {
      "completion_length": 14.765625,
      "epoch": 0.14913448735019974,
      "grad_norm": 7.960399627685547,
      "kl": 0.244873046875,
      "learning_rate": 9.633333333333334e-07,
      "loss": 0.0058,
      "reward": 0.6927297413349152,
      "reward_std": 0.08602303639054298,
      "reward_std_all": 0.09857727214694023,
      "rewards-max/forensics_sinkhorn_reward": 0.9875590801239014,
      "rewards-max/forensics_strict_edge_reward": 0.9151731729507446,
      "rewards-max/forensics_tiou_reward": 0.994357705116272,
      "rewards/forensics_sinkhorn_reward": 0.6163599789142609,
      "rewards/forensics_strict_edge_reward": 0.16296356916427612,
      "rewards/forensics_tiou_reward": 0.3045654296875,
      "step": 56
    },
    {
      "completion_length": 15.78125,
      "epoch": 0.151797603195739,
      "grad_norm": 9.974654197692871,
      "kl": 0.3505859375,
      "learning_rate": 9.626666666666667e-07,
      "loss": -0.0101,
      "reward": 0.6191849708557129,
      "reward_std": 0.12513402849435806,
      "reward_std_all": 0.14628427475690842,
      "rewards-max/forensics_sinkhorn_reward": 0.9816524088382721,
      "rewards-max/forensics_strict_edge_reward": 0.6748847216367722,
      "rewards-max/forensics_tiou_reward": 0.9714205265045166,
      "rewards/forensics_sinkhorn_reward": 0.5275104641914368,
      "rewards/forensics_strict_edge_reward": 0.08063454739749432,
      "rewards/forensics_tiou_reward": 0.18131878972053528,
      "step": 57
    },
    {
      "completion_length": 15.5,
      "epoch": 0.15446071904127828,
      "grad_norm": 15.311497688293457,
      "kl": 0.6201171875,
      "learning_rate": 9.619999999999999e-07,
      "loss": -0.1188,
      "reward": 0.6332015693187714,
      "reward_std": 0.088595200330019,
      "reward_std_all": 0.1255846954882145,
      "rewards-max/forensics_sinkhorn_reward": 0.8795382678508759,
      "rewards-max/forensics_strict_edge_reward": 0.2837844006717205,
      "rewards-max/forensics_tiou_reward": 0.7433047592639923,
      "rewards/forensics_sinkhorn_reward": 0.5023183673620224,
      "rewards/forensics_strict_edge_reward": 0.024685984244570136,
      "rewards/forensics_tiou_reward": 0.11485756561160088,
      "step": 58
    },
    {
      "completion_length": 15.3125,
      "epoch": 0.15712383488681758,
      "grad_norm": 12.637654304504395,
      "kl": 0.26953125,
      "learning_rate": 9.613333333333334e-07,
      "loss": -0.0202,
      "reward": 0.6587575376033783,
      "reward_std": 0.07419601827859879,
      "reward_std_all": 0.10798169299960136,
      "rewards-max/forensics_sinkhorn_reward": 0.8976172506809235,
      "rewards-max/forensics_strict_edge_reward": 0.6011910140514374,
      "rewards-max/forensics_tiou_reward": 0.7803120017051697,
      "rewards/forensics_sinkhorn_reward": 0.5486320555210114,
      "rewards/forensics_strict_edge_reward": 0.05579410865902901,
      "rewards/forensics_tiou_reward": 0.14793265610933304,
      "step": 59
    },
    {
      "completion_length": 15.625,
      "epoch": 0.15978695073235685,
      "grad_norm": 8.668984413146973,
      "kl": 0.1904296875,
      "learning_rate": 9.606666666666666e-07,
      "loss": -0.0425,
      "reward": 0.6790395081043243,
      "reward_std": 0.09612300246953964,
      "reward_std_all": 0.1099388599395752,
      "rewards-max/forensics_sinkhorn_reward": 0.9821555316448212,
      "rewards-max/forensics_strict_edge_reward": 0.842301219701767,
      "rewards-max/forensics_tiou_reward": 0.9730860888957977,
      "rewards/forensics_sinkhorn_reward": 0.6098041534423828,
      "rewards/forensics_strict_edge_reward": 0.09692244231700897,
      "rewards/forensics_tiou_reward": 0.2949248254299164,
      "step": 60
    },
    {
      "completion_length": 14.96875,
      "epoch": 0.16245006657789615,
      "grad_norm": 9.447555541992188,
      "kl": 0.318359375,
      "learning_rate": 9.6e-07,
      "loss": -0.1107,
      "reward": 0.6249408423900604,
      "reward_std": 0.09911797195672989,
      "reward_std_all": 0.13421860337257385,
      "rewards-max/forensics_sinkhorn_reward": 0.8463257849216461,
      "rewards-max/forensics_strict_edge_reward": 0.5633035525679588,
      "rewards-max/forensics_tiou_reward": 0.6864325553178787,
      "rewards/forensics_sinkhorn_reward": 0.5068614482879639,
      "rewards/forensics_strict_edge_reward": 0.05439484678208828,
      "rewards/forensics_tiou_reward": 0.13346058502793312,
      "step": 61
    },
    {
      "completion_length": 17.265625,
      "epoch": 0.16511318242343542,
      "grad_norm": 12.929052352905273,
      "kl": 0.8056640625,
      "learning_rate": 9.593333333333333e-07,
      "loss": 0.0416,
      "reward": 0.6101242899894714,
      "reward_std": 0.1761699914932251,
      "reward_std_all": 0.18550800532102585,
      "rewards-max/forensics_sinkhorn_reward": 0.8223551511764526,
      "rewards-max/forensics_strict_edge_reward": 0.46859800070524216,
      "rewards-max/forensics_tiou_reward": 0.6681092828512192,
      "rewards/forensics_sinkhorn_reward": 0.46719251573085785,
      "rewards/forensics_strict_edge_reward": 0.06776568852365017,
      "rewards/forensics_tiou_reward": 0.1660056635737419,
      "step": 62
    },
    {
      "completion_length": 16.1875,
      "epoch": 0.1677762982689747,
      "grad_norm": 9.125450134277344,
      "kl": 0.376953125,
      "learning_rate": 9.586666666666666e-07,
      "loss": 0.0828,
      "reward": 0.5569111853837967,
      "reward_std": 0.1362248659133911,
      "reward_std_all": 0.18047942221164703,
      "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": 0.4300268292427063,
      "rewards/forensics_strict_edge_reward": 0.10160654410719872,
      "rewards/forensics_tiou_reward": 0.1588968113064766,
      "step": 63
    },
    {
      "completion_length": 17.140625,
      "epoch": 0.170439414114514,
      "grad_norm": 9.564756393432617,
      "kl": 0.4111328125,
      "learning_rate": 9.58e-07,
      "loss": 0.0323,
      "reward": 0.5752001702785492,
      "reward_std": 0.18216128647327423,
      "reward_std_all": 0.19847062975168228,
      "rewards-max/forensics_sinkhorn_reward": 0.9850542843341827,
      "rewards-max/forensics_strict_edge_reward": 0.8612285256385803,
      "rewards-max/forensics_tiou_reward": 0.9844889640808105,
      "rewards/forensics_sinkhorn_reward": 0.47334979474544525,
      "rewards/forensics_strict_edge_reward": 0.11966240406036377,
      "rewards/forensics_tiou_reward": 0.18192044645547867,
      "step": 64
    },
    {
      "completion_length": 17.75,
      "epoch": 0.17310252996005326,
      "grad_norm": 9.720016479492188,
      "kl": 0.232421875,
      "learning_rate": 9.573333333333333e-07,
      "loss": 0.1074,
      "reward": 0.561612606048584,
      "reward_std": 0.2016088217496872,
      "reward_std_all": 0.20991574227809906,
      "rewards-max/forensics_sinkhorn_reward": 0.9882064163684845,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.996913492679596,
      "rewards/forensics_sinkhorn_reward": 0.4672520309686661,
      "rewards/forensics_strict_edge_reward": 0.12937962263822556,
      "rewards/forensics_tiou_reward": 0.2298465520143509,
      "step": 65
    },
    {
      "completion_length": 16.046875,
      "epoch": 0.17576564580559254,
      "grad_norm": 10.890335083007812,
      "kl": 0.55078125,
      "learning_rate": 9.566666666666667e-07,
      "loss": 0.1759,
      "reward": 0.6295726001262665,
      "reward_std": 0.16222135722637177,
      "reward_std_all": 0.16631349176168442,
      "rewards-max/forensics_sinkhorn_reward": 0.980494350194931,
      "rewards-max/forensics_strict_edge_reward": 0.8507176637649536,
      "rewards-max/forensics_tiou_reward": 0.9669423401355743,
      "rewards/forensics_sinkhorn_reward": 0.5002126693725586,
      "rewards/forensics_strict_edge_reward": 0.1639283262193203,
      "rewards/forensics_tiou_reward": 0.24954397976398468,
      "step": 66
    },
    {
      "completion_length": 17.765625,
      "epoch": 0.17842876165113183,
      "grad_norm": 6.965580463409424,
      "kl": 0.24365234375,
      "learning_rate": 9.559999999999998e-07,
      "loss": 0.144,
      "reward": 0.574868232011795,
      "reward_std": 0.12513920664787292,
      "reward_std_all": 0.17245642095804214,
      "rewards-max/forensics_sinkhorn_reward": 0.9512659907341003,
      "rewards-max/forensics_strict_edge_reward": 0.6342070251703262,
      "rewards-max/forensics_tiou_reward": 0.8648374974727631,
      "rewards/forensics_sinkhorn_reward": 0.4472798705101013,
      "rewards/forensics_strict_edge_reward": 0.07770641893148422,
      "rewards/forensics_tiou_reward": 0.14751378446817398,
      "step": 67
    },
    {
      "completion_length": 17.234375,
      "epoch": 0.1810918774966711,
      "grad_norm": 10.115572929382324,
      "kl": 0.419921875,
      "learning_rate": 9.553333333333334e-07,
      "loss": 0.2022,
      "reward": 0.5680648386478424,
      "reward_std": 0.18759088218212128,
      "reward_std_all": 0.20617134869098663,
      "rewards-max/forensics_sinkhorn_reward": 0.9803624749183655,
      "rewards-max/forensics_strict_edge_reward": 0.8137359023094177,
      "rewards-max/forensics_tiou_reward": 0.9663463234901428,
      "rewards/forensics_sinkhorn_reward": 0.4711231291294098,
      "rewards/forensics_strict_edge_reward": 0.14136986434459686,
      "rewards/forensics_tiou_reward": 0.24035810679197311,
      "step": 68
    },
    {
      "completion_length": 19.9375,
      "epoch": 0.18375499334221038,
      "grad_norm": 8.366336822509766,
      "kl": 0.533203125,
      "learning_rate": 9.546666666666665e-07,
      "loss": 0.2749,
      "reward": 0.5938333570957184,
      "reward_std": 0.19164197891950607,
      "reward_std_all": 0.19194843620061874,
      "rewards-max/forensics_sinkhorn_reward": 0.982760488986969,
      "rewards-max/forensics_strict_edge_reward": 0.8279296159744263,
      "rewards-max/forensics_tiou_reward": 0.9756384789943695,
      "rewards/forensics_sinkhorn_reward": 0.4849932938814163,
      "rewards/forensics_strict_edge_reward": 0.15175433456897736,
      "rewards/forensics_tiou_reward": 0.25135813653469086,
      "step": 69
    },
    {
      "completion_length": 18.640625,
      "epoch": 0.18641810918774968,
      "grad_norm": 8.97227954864502,
      "kl": 0.25048828125,
      "learning_rate": 9.539999999999999e-07,
      "loss": 0.0993,
      "reward": 0.5837817788124084,
      "reward_std": 0.16490665823221207,
      "reward_std_all": 0.18471302092075348,
      "rewards-max/forensics_sinkhorn_reward": 0.9860824346542358,
      "rewards-max/forensics_strict_edge_reward": 0.8404377698898315,
      "rewards-max/forensics_tiou_reward": 0.9885647892951965,
      "rewards/forensics_sinkhorn_reward": 0.47388219833374023,
      "rewards/forensics_strict_edge_reward": 0.11588184535503387,
      "rewards/forensics_tiou_reward": 0.20341524481773376,
      "step": 70
    },
    {
      "completion_length": 18.171875,
      "epoch": 0.18908122503328895,
      "grad_norm": 7.401217460632324,
      "kl": 0.30908203125,
      "learning_rate": 9.533333333333333e-07,
      "loss": 0.1776,
      "reward": 0.5715860426425934,
      "reward_std": 0.1289229542016983,
      "reward_std_all": 0.1720452904701233,
      "rewards-max/forensics_sinkhorn_reward": 0.9827758073806763,
      "rewards-max/forensics_strict_edge_reward": 0.8771188855171204,
      "rewards-max/forensics_tiou_reward": 0.9755169451236725,
      "rewards/forensics_sinkhorn_reward": 0.46416573226451874,
      "rewards/forensics_strict_edge_reward": 0.12165477871894836,
      "rewards/forensics_tiou_reward": 0.20573541522026062,
      "step": 71
    },
    {
      "completion_length": 16.9375,
      "epoch": 0.19174434087882822,
      "grad_norm": 8.470230102539062,
      "kl": 0.3671875,
      "learning_rate": 9.526666666666666e-07,
      "loss": -0.0039,
      "reward": 0.6258170902729034,
      "reward_std": 0.11892053857445717,
      "reward_std_all": 0.14293593168258667,
      "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": 0.5043872892856598,
      "rewards/forensics_strict_edge_reward": 0.13774226605892181,
      "rewards/forensics_tiou_reward": 0.20650185644626617,
      "step": 72
    },
    {
      "completion_length": 15.375,
      "epoch": 0.19440745672436752,
      "grad_norm": 10.845209121704102,
      "kl": 0.2880859375,
      "learning_rate": 9.52e-07,
      "loss": 0.0067,
      "reward": 0.6222835779190063,
      "reward_std": 0.12980585917830467,
      "reward_std_all": 0.1480036899447441,
      "rewards-max/forensics_sinkhorn_reward": 0.980263739824295,
      "rewards-max/forensics_strict_edge_reward": 0.7026884853839874,
      "rewards-max/forensics_tiou_reward": 0.9660679399967194,
      "rewards/forensics_sinkhorn_reward": 0.5489517450332642,
      "rewards/forensics_strict_edge_reward": 0.11403521709144115,
      "rewards/forensics_tiou_reward": 0.23566800355911255,
      "step": 73
    },
    {
      "completion_length": 16.53125,
      "epoch": 0.1970705725699068,
      "grad_norm": 10.50134563446045,
      "kl": 0.330078125,
      "learning_rate": 9.513333333333333e-07,
      "loss": -0.0588,
      "reward": 0.5972116589546204,
      "reward_std": 0.11568301543593407,
      "reward_std_all": 0.15601830929517746,
      "rewards-max/forensics_sinkhorn_reward": 0.9840137660503387,
      "rewards-max/forensics_strict_edge_reward": 0.9048932790756226,
      "rewards-max/forensics_tiou_reward": 0.9803884029388428,
      "rewards/forensics_sinkhorn_reward": 0.4691968411207199,
      "rewards/forensics_strict_edge_reward": 0.09534106217324734,
      "rewards/forensics_tiou_reward": 0.13267318159341812,
      "step": 74
    },
    {
      "completion_length": 16.484375,
      "epoch": 0.19973368841544606,
      "grad_norm": 8.384505271911621,
      "kl": 0.37109375,
      "learning_rate": 9.506666666666667e-07,
      "loss": 0.0783,
      "reward": 0.5403240025043488,
      "reward_std": 0.1439751759171486,
      "reward_std_all": 0.19397584348917007,
      "rewards-max/forensics_sinkhorn_reward": 0.9539699554443359,
      "rewards-max/forensics_strict_edge_reward": 0.6064688861370087,
      "rewards-max/forensics_tiou_reward": 0.9152258932590485,
      "rewards/forensics_sinkhorn_reward": 0.41369351744651794,
      "rewards/forensics_strict_edge_reward": 0.07960711792111397,
      "rewards/forensics_tiou_reward": 0.11440018564462662,
      "step": 75
    },
    {
      "completion_length": 15.359375,
      "epoch": 0.20239680426098536,
      "grad_norm": 7.975419998168945,
      "kl": 0.75,
      "learning_rate": 9.499999999999999e-07,
      "loss": 0.0341,
      "reward": 0.6158767938613892,
      "reward_std": 0.1380389928817749,
      "reward_std_all": 0.15441491454839706,
      "rewards-max/forensics_sinkhorn_reward": 0.9696409702301025,
      "rewards-max/forensics_strict_edge_reward": 0.562614694237709,
      "rewards-max/forensics_tiou_reward": 0.9261110126972198,
      "rewards/forensics_sinkhorn_reward": 0.5231785774230957,
      "rewards/forensics_strict_edge_reward": 0.11907242424786091,
      "rewards/forensics_tiou_reward": 0.20338140800595284,
      "step": 76
    },
    {
      "completion_length": 15.0625,
      "epoch": 0.20505992010652463,
      "grad_norm": 7.076149940490723,
      "kl": 0.2607421875,
      "learning_rate": 9.493333333333334e-07,
      "loss": 0.0236,
      "reward": 0.6462406814098358,
      "reward_std": 0.06703417003154755,
      "reward_std_all": 0.11378135159611702,
      "rewards-max/forensics_sinkhorn_reward": 0.9671061336994171,
      "rewards-max/forensics_strict_edge_reward": 0.7849180996417999,
      "rewards-max/forensics_tiou_reward": 0.9147506654262543,
      "rewards/forensics_sinkhorn_reward": 0.5782259106636047,
      "rewards/forensics_strict_edge_reward": 0.15080903843045235,
      "rewards/forensics_tiou_reward": 0.27225636690855026,
      "step": 77
    },
    {
      "completion_length": 15.71875,
      "epoch": 0.20772303595206393,
      "grad_norm": 9.996088027954102,
      "kl": 0.388671875,
      "learning_rate": 9.486666666666666e-07,
      "loss": 0.0109,
      "reward": 0.5830234587192535,
      "reward_std": 0.14772459864616394,
      "reward_std_all": 0.18079302459955215,
      "rewards-max/forensics_sinkhorn_reward": 0.9860033690929413,
      "rewards-max/forensics_strict_edge_reward": 0.9062341153621674,
      "rewards-max/forensics_tiou_reward": 0.9882274270057678,
      "rewards/forensics_sinkhorn_reward": 0.49044477939605713,
      "rewards/forensics_strict_edge_reward": 0.12115197256207466,
      "rewards/forensics_tiou_reward": 0.19791854545474052,
      "step": 78
    },
    {
      "completion_length": 14.984375,
      "epoch": 0.2103861517976032,
      "grad_norm": 12.08537483215332,
      "kl": 0.3896484375,
      "learning_rate": 9.479999999999999e-07,
      "loss": -0.0906,
      "reward": 0.613667368888855,
      "reward_std": 0.130758598446846,
      "reward_std_all": 0.155406653881073,
      "rewards-max/forensics_sinkhorn_reward": 0.9857164323329926,
      "rewards-max/forensics_strict_edge_reward": 0.8968285918235779,
      "rewards-max/forensics_tiou_reward": 0.9871504604816437,
      "rewards/forensics_sinkhorn_reward": 0.49936023354530334,
      "rewards/forensics_strict_edge_reward": 0.07956559211015701,
      "rewards/forensics_tiou_reward": 0.1476401947438717,
      "step": 79
    },
    {
      "completion_length": 15.265625,
      "epoch": 0.21304926764314247,
      "grad_norm": 10.611842155456543,
      "kl": 0.837890625,
      "learning_rate": 9.473333333333333e-07,
      "loss": 0.0309,
      "reward": 0.593157947063446,
      "reward_std": 0.13391278311610222,
      "reward_std_all": 0.16419422253966331,
      "rewards-max/forensics_sinkhorn_reward": 0.9250591397285461,
      "rewards-max/forensics_strict_edge_reward": 0.6924179792404175,
      "rewards-max/forensics_tiou_reward": 0.8597679734230042,
      "rewards/forensics_sinkhorn_reward": 0.4938914477825165,
      "rewards/forensics_strict_edge_reward": 0.07966161146759987,
      "rewards/forensics_tiou_reward": 0.15537148714065552,
      "step": 80
    },
    {
      "completion_length": 15.65625,
      "epoch": 0.21571238348868177,
      "grad_norm": 10.44372844696045,
      "kl": 0.5673828125,
      "learning_rate": 9.466666666666666e-07,
      "loss": 0.1394,
      "reward": 0.6932105720043182,
      "reward_std": 0.08044762164354324,
      "reward_std_all": 0.09573153033852577,
      "rewards-max/forensics_sinkhorn_reward": 0.9879333972930908,
      "rewards-max/forensics_strict_edge_reward": 0.947419673204422,
      "rewards-max/forensics_tiou_reward": 0.9958420097827911,
      "rewards/forensics_sinkhorn_reward": 0.6570630371570587,
      "rewards/forensics_strict_edge_reward": 0.22021085768938065,
      "rewards/forensics_tiou_reward": 0.3848250210285187,
      "step": 81
    },
    {
      "completion_length": 15.328125,
      "epoch": 0.21837549933422104,
      "grad_norm": 8.801751136779785,
      "kl": 0.41015625,
      "learning_rate": 9.459999999999999e-07,
      "loss": 0.0537,
      "reward": 0.6720805466175079,
      "reward_std": 0.11793665587902069,
      "reward_std_all": 0.12319478765130043,
      "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": 0.6232078373432159,
      "rewards/forensics_strict_edge_reward": 0.23219331726431847,
      "rewards/forensics_tiou_reward": 0.35826291143894196,
      "step": 82
    },
    {
      "completion_length": 18.71875,
      "epoch": 0.2210386151797603,
      "grad_norm": 8.83098316192627,
      "kl": 0.638671875,
      "learning_rate": 9.453333333333333e-07,
      "loss": 0.0937,
      "reward": 0.6465869545936584,
      "reward_std": 0.14208325743675232,
      "reward_std_all": 0.14835554361343384,
      "rewards-max/forensics_sinkhorn_reward": 0.9845632016658783,
      "rewards-max/forensics_strict_edge_reward": 0.8863177299499512,
      "rewards-max/forensics_tiou_reward": 0.9826216697692871,
      "rewards/forensics_sinkhorn_reward": 0.5666826367378235,
      "rewards/forensics_strict_edge_reward": 0.21892252564430237,
      "rewards/forensics_tiou_reward": 0.375044047832489,
      "step": 83
    },
    {
      "completion_length": 17.34375,
      "epoch": 0.2237017310252996,
      "grad_norm": 10.704774856567383,
      "kl": 0.640625,
      "learning_rate": 9.446666666666666e-07,
      "loss": 0.2332,
      "reward": 0.597716897726059,
      "reward_std": 0.1504768803715706,
      "reward_std_all": 0.17680078744888306,
      "rewards-max/forensics_sinkhorn_reward": 0.9776792228221893,
      "rewards-max/forensics_strict_edge_reward": 0.8157392740249634,
      "rewards-max/forensics_tiou_reward": 0.9563564658164978,
      "rewards/forensics_sinkhorn_reward": 0.49156060814857483,
      "rewards/forensics_strict_edge_reward": 0.11761555075645447,
      "rewards/forensics_tiou_reward": 0.2268727794289589,
      "step": 84
    },
    {
      "completion_length": 16.59375,
      "epoch": 0.22636484687083888,
      "grad_norm": 9.186814308166504,
      "kl": 0.431640625,
      "learning_rate": 9.439999999999999e-07,
      "loss": 0.0717,
      "reward": 0.6462385952472687,
      "reward_std": 0.1399202086031437,
      "reward_std_all": 0.14749878644943237,
      "rewards-max/forensics_sinkhorn_reward": 0.9883498847484589,
      "rewards-max/forensics_strict_edge_reward": 0.9890114367008209,
      "rewards-max/forensics_tiou_reward": 0.9974779486656189,
      "rewards/forensics_sinkhorn_reward": 0.5591698735952377,
      "rewards/forensics_strict_edge_reward": 0.13592233881354332,
      "rewards/forensics_tiou_reward": 0.3024324029684067,
      "step": 85
    },
    {
      "completion_length": 17.890625,
      "epoch": 0.22902796271637815,
      "grad_norm": 8.298967361450195,
      "kl": 0.36328125,
      "learning_rate": 9.433333333333333e-07,
      "loss": 0.0725,
      "reward": 0.6525886058807373,
      "reward_std": 0.13599811494350433,
      "reward_std_all": 0.14329908415675163,
      "rewards-max/forensics_sinkhorn_reward": 0.9878159165382385,
      "rewards-max/forensics_strict_edge_reward": 0.9574736058712006,
      "rewards-max/forensics_tiou_reward": 0.995381087064743,
      "rewards/forensics_sinkhorn_reward": 0.5290963649749756,
      "rewards/forensics_strict_edge_reward": 0.14791415631771088,
      "rewards/forensics_tiou_reward": 0.251541867852211,
      "step": 86
    },
    {
      "completion_length": 16.765625,
      "epoch": 0.23169107856191745,
      "grad_norm": 9.865011215209961,
      "kl": 0.40234375,
      "learning_rate": 9.426666666666666e-07,
      "loss": 0.0707,
      "reward": 0.5910792350769043,
      "reward_std": 0.1547044739127159,
      "reward_std_all": 0.1774439439177513,
      "rewards-max/forensics_sinkhorn_reward": 0.9877195656299591,
      "rewards-max/forensics_strict_edge_reward": 0.946017861366272,
      "rewards-max/forensics_tiou_reward": 0.9949924945831299,
      "rewards/forensics_sinkhorn_reward": 0.4995725154876709,
      "rewards/forensics_strict_edge_reward": 0.1523890718817711,
      "rewards/forensics_tiou_reward": 0.24182948470115662,
      "step": 87
    },
    {
      "completion_length": 17.015625,
      "epoch": 0.23435419440745672,
      "grad_norm": 9.068342208862305,
      "kl": 0.69140625,
      "learning_rate": 9.419999999999999e-07,
      "loss": 0.1122,
      "reward": 0.5819195210933685,
      "reward_std": 0.16976121813058853,
      "reward_std_all": 0.18351253867149353,
      "rewards-max/forensics_sinkhorn_reward": 0.8471198678016663,
      "rewards-max/forensics_strict_edge_reward": 0.5422598570585251,
      "rewards-max/forensics_tiou_reward": 0.6978879421949387,
      "rewards/forensics_sinkhorn_reward": 0.43328234553337097,
      "rewards/forensics_strict_edge_reward": 0.05599803477525711,
      "rewards/forensics_tiou_reward": 0.1290995255112648,
      "step": 88
    },
    {
      "completion_length": 17.203125,
      "epoch": 0.237017310252996,
      "grad_norm": 9.691319465637207,
      "kl": 0.4765625,
      "learning_rate": 9.413333333333333e-07,
      "loss": 0.0839,
      "reward": 0.6102991700172424,
      "reward_std": 0.12339653819799423,
      "reward_std_all": 0.15438274294137955,
      "rewards-max/forensics_sinkhorn_reward": 0.9868887662887573,
      "rewards-max/forensics_strict_edge_reward": 0.9364311099052429,
      "rewards-max/forensics_tiou_reward": 0.9917485415935516,
      "rewards/forensics_sinkhorn_reward": 0.5158287584781647,
      "rewards/forensics_strict_edge_reward": 0.1162291057407856,
      "rewards/forensics_tiou_reward": 0.21201354265213013,
      "step": 89
    },
    {
      "completion_length": 16.3125,
      "epoch": 0.2396804260985353,
      "grad_norm": 13.777551651000977,
      "kl": 0.572265625,
      "learning_rate": 9.406666666666666e-07,
      "loss": 0.0524,
      "reward": 0.6013910472393036,
      "reward_std": 0.1230844110250473,
      "reward_std_all": 0.160539910197258,
      "rewards-max/forensics_sinkhorn_reward": 0.9870546758174896,
      "rewards-max/forensics_strict_edge_reward": 0.9357379674911499,
      "rewards-max/forensics_tiou_reward": 0.9923851490020752,
      "rewards/forensics_sinkhorn_reward": 0.5646244883537292,
      "rewards/forensics_strict_edge_reward": 0.2490309774875641,
      "rewards/forensics_tiou_reward": 0.3500634431838989,
      "step": 90
    },
    {
      "completion_length": 15.9375,
      "epoch": 0.24234354194407456,
      "grad_norm": 9.857560157775879,
      "kl": 0.8642578125,
      "learning_rate": 9.399999999999999e-07,
      "loss": -0.1152,
      "reward": 0.6411105692386627,
      "reward_std": 0.12510047480463982,
      "reward_std_all": 0.14477981626987457,
      "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": 0.5576821863651276,
      "rewards/forensics_strict_edge_reward": 0.16623761504888535,
      "rewards/forensics_tiou_reward": 0.28172076493501663,
      "step": 91
    },
    {
      "completion_length": 15.203125,
      "epoch": 0.24500665778961384,
      "grad_norm": 7.678003311157227,
      "kl": 0.3193359375,
      "learning_rate": 9.393333333333334e-07,
      "loss": 0.1849,
      "reward": 0.6829651296138763,
      "reward_std": 0.11584979295730591,
      "reward_std_all": 0.12197250872850418,
      "rewards-max/forensics_sinkhorn_reward": 0.9876191020011902,
      "rewards-max/forensics_strict_edge_reward": 0.9464850425720215,
      "rewards-max/forensics_tiou_reward": 0.9945956468582153,
      "rewards/forensics_sinkhorn_reward": 0.6209813952445984,
      "rewards/forensics_strict_edge_reward": 0.19313669204711914,
      "rewards/forensics_tiou_reward": 0.3370053470134735,
      "step": 92
    },
    {
      "completion_length": 16.609375,
      "epoch": 0.24766977363515313,
      "grad_norm": 8.043535232543945,
      "kl": 0.4677734375,
      "learning_rate": 9.386666666666666e-07,
      "loss": -0.0244,
      "reward": 0.6181093752384186,
      "reward_std": 0.1586940661072731,
      "reward_std_all": 0.17213702201843262,
      "rewards-max/forensics_sinkhorn_reward": 0.9052269756793976,
      "rewards-max/forensics_strict_edge_reward": 0.6798607707023621,
      "rewards-max/forensics_tiou_reward": 0.820443868637085,
      "rewards/forensics_sinkhorn_reward": 0.5092495977878571,
      "rewards/forensics_strict_edge_reward": 0.10814061015844345,
      "rewards/forensics_tiou_reward": 0.20610299706459045,
      "step": 93
    },
    {
      "completion_length": 15.40625,
      "epoch": 0.25033288948069243,
      "grad_norm": 11.66938591003418,
      "kl": 0.6240234375,
      "learning_rate": 9.379999999999998e-07,
      "loss": 0.0758,
      "reward": 0.6051214933395386,
      "reward_std": 0.10005970299243927,
      "reward_std_all": 0.14117446541786194,
      "rewards-max/forensics_sinkhorn_reward": 0.9755345582962036,
      "rewards-max/forensics_strict_edge_reward": 0.7463583946228027,
      "rewards-max/forensics_tiou_reward": 0.9479533135890961,
      "rewards/forensics_sinkhorn_reward": 0.5050567835569382,
      "rewards/forensics_strict_edge_reward": 0.1002696193754673,
      "rewards/forensics_tiou_reward": 0.17720702290534973,
      "step": 94
    },
    {
      "completion_length": 17.390625,
      "epoch": 0.2529960053262317,
      "grad_norm": 8.989799499511719,
      "kl": 0.2158203125,
      "learning_rate": 9.373333333333333e-07,
      "loss": 0.0025,
      "reward": 0.6023695766925812,
      "reward_std": 0.15675510838627815,
      "reward_std_all": 0.17157506942749023,
      "rewards-max/forensics_sinkhorn_reward": 0.9880018830299377,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9961049854755402,
      "rewards/forensics_sinkhorn_reward": 0.5155875384807587,
      "rewards/forensics_strict_edge_reward": 0.1968102604150772,
      "rewards/forensics_tiou_reward": 0.26109375804662704,
      "step": 95
    },
    {
      "completion_length": 16.234375,
      "epoch": 0.255659121171771,
      "grad_norm": 7.09065580368042,
      "kl": 0.3916015625,
      "learning_rate": 9.366666666666666e-07,
      "loss": 0.0143,
      "reward": 0.6417196691036224,
      "reward_std": 0.13279350101947784,
      "reward_std_all": 0.14592380821704865,
      "rewards-max/forensics_sinkhorn_reward": 0.9881395697593689,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9966491162776947,
      "rewards/forensics_sinkhorn_reward": 0.5213043391704559,
      "rewards/forensics_strict_edge_reward": 0.1403966061770916,
      "rewards/forensics_tiou_reward": 0.22797031700611115,
      "step": 96
    },
    {
      "completion_length": 16.046875,
      "epoch": 0.2583222370173103,
      "grad_norm": 8.366538047790527,
      "kl": 0.2490234375,
      "learning_rate": 9.36e-07,
      "loss": 0.0863,
      "reward": 0.5752991437911987,
      "reward_std": 0.1307762935757637,
      "reward_std_all": 0.16716231405735016,
      "rewards-max/forensics_sinkhorn_reward": 0.9135870337486267,
      "rewards-max/forensics_strict_edge_reward": 0.4803053140640259,
      "rewards-max/forensics_tiou_reward": 0.7996685206890106,
      "rewards/forensics_sinkhorn_reward": 0.4366791546344757,
      "rewards/forensics_strict_edge_reward": 0.05896664969623089,
      "rewards/forensics_tiou_reward": 0.1010749563574791,
      "step": 97
    },
    {
      "completion_length": 17.09375,
      "epoch": 0.2609853528628495,
      "grad_norm": 7.034094333648682,
      "kl": 0.39111328125,
      "learning_rate": 9.353333333333333e-07,
      "loss": -0.082,
      "reward": 0.5895669758319855,
      "reward_std": 0.15045391768217087,
      "reward_std_all": 0.17538152635097504,
      "rewards-max/forensics_sinkhorn_reward": 0.987076073884964,
      "rewards-max/forensics_strict_edge_reward": 0.9567649364471436,
      "rewards-max/forensics_tiou_reward": 0.9924558699131012,
      "rewards/forensics_sinkhorn_reward": 0.47005486488342285,
      "rewards/forensics_strict_edge_reward": 0.10074158757925034,
      "rewards/forensics_tiou_reward": 0.16197418421506882,
      "step": 98
    },
    {
      "completion_length": 15.65625,
      "epoch": 0.2636484687083888,
      "grad_norm": 11.049232482910156,
      "kl": 0.73046875,
      "learning_rate": 9.346666666666666e-07,
      "loss": 0.0235,
      "reward": 0.5999670624732971,
      "reward_std": 0.15632694214582443,
      "reward_std_all": 0.1729525700211525,
      "rewards-max/forensics_sinkhorn_reward": 0.9831275641918182,
      "rewards-max/forensics_strict_edge_reward": 0.9169811904430389,
      "rewards-max/forensics_tiou_reward": 0.980034351348877,
      "rewards/forensics_sinkhorn_reward": 0.47533948719501495,
      "rewards/forensics_strict_edge_reward": 0.15377023816108704,
      "rewards/forensics_tiou_reward": 0.22139472514390945,
      "step": 99
    },
    {
      "completion_length": 16.234375,
      "epoch": 0.2663115845539281,
      "grad_norm": 8.98658561706543,
      "kl": 0.4306640625,
      "learning_rate": 9.34e-07,
      "loss": 0.1524,
      "reward": 0.5599134266376495,
      "reward_std": 0.13433481752872467,
      "reward_std_all": 0.18188871443271637,
      "rewards-max/forensics_sinkhorn_reward": 0.8796332478523254,
      "rewards-max/forensics_strict_edge_reward": 0.7291640937328339,
      "rewards-max/forensics_tiou_reward": 0.7698028087615967,
      "rewards/forensics_sinkhorn_reward": 0.47206422686576843,
      "rewards/forensics_strict_edge_reward": 0.17106961458921432,
      "rewards/forensics_tiou_reward": 0.2150200977921486,
      "step": 100
    },
    {
      "completion_length": 15.734375,
      "epoch": 0.26897470039946736,
      "grad_norm": 7.820175647735596,
      "kl": 0.453125,
      "learning_rate": 9.333333333333333e-07,
      "loss": 0.1248,
      "reward": 0.6039802134037018,
      "reward_std": 0.1242724284529686,
      "reward_std_all": 0.1527751311659813,
      "rewards-max/forensics_sinkhorn_reward": 0.9452349841594696,
      "rewards-max/forensics_strict_edge_reward": 0.7805719971656799,
      "rewards-max/forensics_tiou_reward": 0.9004514515399933,
      "rewards/forensics_sinkhorn_reward": 0.4905038923025131,
      "rewards/forensics_strict_edge_reward": 0.11436634510755539,
      "rewards/forensics_tiou_reward": 0.16726529598236084,
      "step": 101
    },
    {
      "completion_length": 17.953125,
      "epoch": 0.27163781624500666,
      "grad_norm": 6.7588605880737305,
      "kl": 0.5830078125,
      "learning_rate": 9.326666666666666e-07,
      "loss": -0.0154,
      "reward": 0.57574462890625,
      "reward_std": 0.15409158915281296,
      "reward_std_all": 0.18292644619941711,
      "rewards-max/forensics_sinkhorn_reward": 0.8888445794582367,
      "rewards-max/forensics_strict_edge_reward": 0.376499280333519,
      "rewards-max/forensics_tiou_reward": 0.769353598356247,
      "rewards/forensics_sinkhorn_reward": 0.4188723564147949,
      "rewards/forensics_strict_edge_reward": 0.06034293957054615,
      "rewards/forensics_tiou_reward": 0.15115190297365189,
      "step": 102
    },
    {
      "completion_length": 17.359375,
      "epoch": 0.27430093209054596,
      "grad_norm": 9.33104133605957,
      "kl": 0.466796875,
      "learning_rate": 9.32e-07,
      "loss": -0.1161,
      "reward": 0.6039147973060608,
      "reward_std": 0.18056053668260574,
      "reward_std_all": 0.18577715754508972,
      "rewards-max/forensics_sinkhorn_reward": 0.9684061408042908,
      "rewards-max/forensics_strict_edge_reward": 0.7209588885307312,
      "rewards-max/forensics_tiou_reward": 0.9382394850254059,
      "rewards/forensics_sinkhorn_reward": 0.4508805274963379,
      "rewards/forensics_strict_edge_reward": 0.13070529326796532,
      "rewards/forensics_tiou_reward": 0.2025028094649315,
      "step": 103
    },
    {
      "completion_length": 18.453125,
      "epoch": 0.2769640479360852,
      "grad_norm": 7.291207790374756,
      "kl": 0.4091796875,
      "learning_rate": 9.313333333333333e-07,
      "loss": 0.037,
      "reward": 0.5556281805038452,
      "reward_std": 0.13227901607751846,
      "reward_std_all": 0.18088820576667786,
      "rewards-max/forensics_sinkhorn_reward": 0.9047788977622986,
      "rewards-max/forensics_strict_edge_reward": 0.5966909527778625,
      "rewards-max/forensics_tiou_reward": 0.8068059980869293,
      "rewards/forensics_sinkhorn_reward": 0.407061904668808,
      "rewards/forensics_strict_edge_reward": 0.07858781516551971,
      "rewards/forensics_tiou_reward": 0.14847459644079208,
      "step": 104
    },
    {
      "completion_length": 17.9375,
      "epoch": 0.2796271637816245,
      "grad_norm": 8.732421875,
      "kl": 0.6962890625,
      "learning_rate": 9.306666666666666e-07,
      "loss": 0.2655,
      "reward": 0.5657001733779907,
      "reward_std": 0.20498760044574738,
      "reward_std_all": 0.2177470177412033,
      "rewards-max/forensics_sinkhorn_reward": 0.9860347211360931,
      "rewards-max/forensics_strict_edge_reward": 0.8157392740249634,
      "rewards-max/forensics_tiou_reward": 0.9884039461612701,
      "rewards/forensics_sinkhorn_reward": 0.47987446188926697,
      "rewards/forensics_strict_edge_reward": 0.1994321569800377,
      "rewards/forensics_tiou_reward": 0.26580366492271423,
      "step": 105
    },
    {
      "completion_length": 18.203125,
      "epoch": 0.2822902796271638,
      "grad_norm": 9.636371612548828,
      "kl": 0.53125,
      "learning_rate": 9.3e-07,
      "loss": 0.1179,
      "reward": 0.5295453369617462,
      "reward_std": 0.17435332387685776,
      "reward_std_all": 0.21606046706438065,
      "rewards-max/forensics_sinkhorn_reward": 0.9837221205234528,
      "rewards-max/forensics_strict_edge_reward": 0.842301219701767,
      "rewards-max/forensics_tiou_reward": 0.9793721139431,
      "rewards/forensics_sinkhorn_reward": 0.4316035956144333,
      "rewards/forensics_strict_edge_reward": 0.08096777647733688,
      "rewards/forensics_tiou_reward": 0.14790047705173492,
      "step": 106
    },
    {
      "completion_length": 20.796875,
      "epoch": 0.28495339547270304,
      "grad_norm": 9.569670677185059,
      "kl": 0.443359375,
      "learning_rate": 9.293333333333333e-07,
      "loss": 0.0518,
      "reward": 0.5442227125167847,
      "reward_std": 0.25175468623638153,
      "reward_std_all": 0.24865922331809998,
      "rewards-max/forensics_sinkhorn_reward": 0.9830925166606903,
      "rewards-max/forensics_strict_edge_reward": 0.6886534988880157,
      "rewards-max/forensics_tiou_reward": 0.9768911302089691,
      "rewards/forensics_sinkhorn_reward": 0.40823474526405334,
      "rewards/forensics_strict_edge_reward": 0.10185801982879639,
      "rewards/forensics_tiou_reward": 0.21336808055639267,
      "step": 107
    },
    {
      "completion_length": 19.40625,
      "epoch": 0.28761651131824234,
      "grad_norm": 11.378337860107422,
      "kl": 0.6923828125,
      "learning_rate": 9.286666666666666e-07,
      "loss": 0.1353,
      "reward": 0.5899629890918732,
      "reward_std": 0.18972714990377426,
      "reward_std_all": 0.1962791234254837,
      "rewards-max/forensics_sinkhorn_reward": 0.9592091739177704,
      "rewards-max/forensics_strict_edge_reward": 0.8668389916419983,
      "rewards-max/forensics_tiou_reward": 0.9315100312232971,
      "rewards/forensics_sinkhorn_reward": 0.4588228911161423,
      "rewards/forensics_strict_edge_reward": 0.1259186528623104,
      "rewards/forensics_tiou_reward": 0.22804319858551025,
      "step": 108
    },
    {
      "completion_length": 19.0,
      "epoch": 0.29027962716378164,
      "grad_norm": 7.989408493041992,
      "kl": 0.404296875,
      "learning_rate": 9.28e-07,
      "loss": 0.1148,
      "reward": 0.6061216294765472,
      "reward_std": 0.18449459969997406,
      "reward_std_all": 0.19044339656829834,
      "rewards-max/forensics_sinkhorn_reward": 0.9866780340671539,
      "rewards-max/forensics_strict_edge_reward": 0.9364311099052429,
      "rewards-max/forensics_tiou_reward": 0.9908868074417114,
      "rewards/forensics_sinkhorn_reward": 0.5010799467563629,
      "rewards/forensics_strict_edge_reward": 0.1873023360967636,
      "rewards/forensics_tiou_reward": 0.2578626275062561,
      "step": 109
    },
    {
      "completion_length": 21.046875,
      "epoch": 0.2929427430093209,
      "grad_norm": 8.399534225463867,
      "kl": 0.513671875,
      "learning_rate": 9.273333333333333e-07,
      "loss": 0.123,
      "reward": 0.5343140363693237,
      "reward_std": 0.16298867389559746,
      "reward_std_all": 0.2049630731344223,
      "rewards-max/forensics_sinkhorn_reward": 0.9867397248744965,
      "rewards-max/forensics_strict_edge_reward": 0.7955476641654968,
      "rewards-max/forensics_tiou_reward": 0.9911520481109619,
      "rewards/forensics_sinkhorn_reward": 0.4034288227558136,
      "rewards/forensics_strict_edge_reward": 0.1171361431479454,
      "rewards/forensics_tiou_reward": 0.2033735066652298,
      "step": 110
    },
    {
      "completion_length": 20.15625,
      "epoch": 0.2956058588548602,
      "grad_norm": 7.520458221435547,
      "kl": 0.568359375,
      "learning_rate": 9.266666666666665e-07,
      "loss": -0.042,
      "reward": 0.5854456722736359,
      "reward_std": 0.1473926603794098,
      "reward_std_all": 0.17660768330097198,
      "rewards-max/forensics_sinkhorn_reward": 0.8555800318717957,
      "rewards-max/forensics_strict_edge_reward": 0.5525899678468704,
      "rewards-max/forensics_tiou_reward": 0.6891380101442337,
      "rewards/forensics_sinkhorn_reward": 0.4199218600988388,
      "rewards/forensics_strict_edge_reward": 0.08226514048874378,
      "rewards/forensics_tiou_reward": 0.16610895842313766,
      "step": 111
    },
    {
      "completion_length": 19.15625,
      "epoch": 0.2982689747003995,
      "grad_norm": 10.942309379577637,
      "kl": 0.68359375,
      "learning_rate": 9.26e-07,
      "loss": 0.0806,
      "reward": 0.5755053162574768,
      "reward_std": 0.15103767812252045,
      "reward_std_all": 0.18100295215845108,
      "rewards-max/forensics_sinkhorn_reward": 0.9832496345043182,
      "rewards-max/forensics_strict_edge_reward": 0.8528120815753937,
      "rewards-max/forensics_tiou_reward": 0.97755166888237,
      "rewards/forensics_sinkhorn_reward": 0.4300125539302826,
      "rewards/forensics_strict_edge_reward": 0.14736858010292053,
      "rewards/forensics_tiou_reward": 0.20793328434228897,
      "step": 112
    },
    {
      "completion_length": 19.203125,
      "epoch": 0.3009320905459387,
      "grad_norm": 7.592473030090332,
      "kl": 0.31640625,
      "learning_rate": 9.253333333333333e-07,
      "loss": 0.0185,
      "reward": 0.6109958589076996,
      "reward_std": 0.14552468061447144,
      "reward_std_all": 0.16090687364339828,
      "rewards-max/forensics_sinkhorn_reward": 0.8596497476100922,
      "rewards-max/forensics_strict_edge_reward": 0.6738131046295166,
      "rewards-max/forensics_tiou_reward": 0.7326201796531677,
      "rewards/forensics_sinkhorn_reward": 0.5082330107688904,
      "rewards/forensics_strict_edge_reward": 0.1466316059231758,
      "rewards/forensics_tiou_reward": 0.259017713367939,
      "step": 113
    },
    {
      "completion_length": 18.9375,
      "epoch": 0.303595206391478,
      "grad_norm": 7.853825092315674,
      "kl": 0.6416015625,
      "learning_rate": 9.246666666666666e-07,
      "loss": 0.229,
      "reward": 0.5899736285209656,
      "reward_std": 0.10757698118686676,
      "reward_std_all": 0.1569807156920433,
      "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": 0.47256940603256226,
      "rewards/forensics_strict_edge_reward": 0.10604921728372574,
      "rewards/forensics_tiou_reward": 0.18971937894821167,
      "step": 114
    },
    {
      "completion_length": 16.65625,
      "epoch": 0.3062583222370173,
      "grad_norm": 10.133501052856445,
      "kl": 0.623046875,
      "learning_rate": 9.24e-07,
      "loss": -0.0055,
      "reward": 0.619108259677887,
      "reward_std": 0.168376587331295,
      "reward_std_all": 0.17237535119056702,
      "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": 0.5050268471240997,
      "rewards/forensics_strict_edge_reward": 0.11818640306591988,
      "rewards/forensics_tiou_reward": 0.23845131695270538,
      "step": 115
    },
    {
      "completion_length": 15.75,
      "epoch": 0.30892143808255657,
      "grad_norm": 10.603941917419434,
      "kl": 0.845703125,
      "learning_rate": 9.233333333333333e-07,
      "loss": 0.0342,
      "reward": 0.6264451146125793,
      "reward_std": 0.11719994246959686,
      "reward_std_all": 0.14284973591566086,
      "rewards-max/forensics_sinkhorn_reward": 0.981780081987381,
      "rewards-max/forensics_strict_edge_reward": 0.8635591566562653,
      "rewards-max/forensics_tiou_reward": 0.971604734659195,
      "rewards/forensics_sinkhorn_reward": 0.5341146886348724,
      "rewards/forensics_strict_edge_reward": 0.1279963105916977,
      "rewards/forensics_tiou_reward": 0.2108498439192772,
      "step": 116
    },
    {
      "completion_length": 15.828125,
      "epoch": 0.31158455392809586,
      "grad_norm": 7.917219161987305,
      "kl": 0.876953125,
      "learning_rate": 9.226666666666666e-07,
      "loss": -0.0086,
      "reward": 0.6114667356014252,
      "reward_std": 0.11536996066570282,
      "reward_std_all": 0.15149085223674774,
      "rewards-max/forensics_sinkhorn_reward": 0.8616735339164734,
      "rewards-max/forensics_strict_edge_reward": 0.4401264041662216,
      "rewards-max/forensics_tiou_reward": 0.7265892922878265,
      "rewards/forensics_sinkhorn_reward": 0.4957630783319473,
      "rewards/forensics_strict_edge_reward": 0.07763054594397545,
      "rewards/forensics_tiou_reward": 0.17281464487314224,
      "step": 117
    },
    {
      "completion_length": 16.1875,
      "epoch": 0.31424766977363516,
      "grad_norm": 9.764839172363281,
      "kl": 0.5244140625,
      "learning_rate": 9.22e-07,
      "loss": 0.0944,
      "reward": 0.5989775061607361,
      "reward_std": 0.1583404466509819,
      "reward_std_all": 0.17502474784851074,
      "rewards-max/forensics_sinkhorn_reward": 0.9185696542263031,
      "rewards-max/forensics_strict_edge_reward": 0.5309127271175385,
      "rewards-max/forensics_tiou_reward": 0.8399055302143097,
      "rewards/forensics_sinkhorn_reward": 0.4747755527496338,
      "rewards/forensics_strict_edge_reward": 0.06541861966252327,
      "rewards/forensics_tiou_reward": 0.19598383456468582,
      "step": 118
    },
    {
      "completion_length": 16.734375,
      "epoch": 0.3169107856191744,
      "grad_norm": 8.356746673583984,
      "kl": 0.404296875,
      "learning_rate": 9.213333333333333e-07,
      "loss": -0.0203,
      "reward": 0.6009339094161987,
      "reward_std": 0.1517573967576027,
      "reward_std_all": 0.17161925137043,
      "rewards-max/forensics_sinkhorn_reward": 0.9803187251091003,
      "rewards-max/forensics_strict_edge_reward": 0.8389894068241119,
      "rewards-max/forensics_tiou_reward": 0.9662131667137146,
      "rewards/forensics_sinkhorn_reward": 0.4820146858692169,
      "rewards/forensics_strict_edge_reward": 0.09480001032352448,
      "rewards/forensics_tiou_reward": 0.18854708969593048,
      "step": 119
    },
    {
      "completion_length": 17.875,
      "epoch": 0.3195739014647137,
      "grad_norm": 7.765913486480713,
      "kl": 0.384765625,
      "learning_rate": 9.206666666666666e-07,
      "loss": 0.0575,
      "reward": 0.5898056030273438,
      "reward_std": 0.10726142302155495,
      "reward_std_all": 0.15520106256008148,
      "rewards-max/forensics_sinkhorn_reward": 0.9878958761692047,
      "rewards-max/forensics_strict_edge_reward": 0.9151731729507446,
      "rewards-max/forensics_tiou_reward": 0.9956873059272766,
      "rewards/forensics_sinkhorn_reward": 0.4824589192867279,
      "rewards/forensics_strict_edge_reward": 0.1385655179619789,
      "rewards/forensics_tiou_reward": 0.1956210397183895,
      "step": 120
    },
    {
      "completion_length": 16.65625,
      "epoch": 0.322237017310253,
      "grad_norm": 9.72462272644043,
      "kl": 0.5302734375,
      "learning_rate": 9.2e-07,
      "loss": 0.2125,
      "reward": 0.5477150082588196,
      "reward_std": 0.15372920781373978,
      "reward_std_all": 0.19014763832092285,
      "rewards-max/forensics_sinkhorn_reward": 0.9517066478729248,
      "rewards-max/forensics_strict_edge_reward": 0.7014762461185455,
      "rewards-max/forensics_tiou_reward": 0.9079895913600922,
      "rewards/forensics_sinkhorn_reward": 0.42882443964481354,
      "rewards/forensics_strict_edge_reward": 0.09216184914112091,
      "rewards/forensics_tiou_reward": 0.1434936448931694,
      "step": 121
    },
    {
      "completion_length": 15.53125,
      "epoch": 0.3249001331557923,
      "grad_norm": 10.692328453063965,
      "kl": 0.654296875,
      "learning_rate": 9.193333333333333e-07,
      "loss": 0.0532,
      "reward": 0.6426339149475098,
      "reward_std": 0.13993729650974274,
      "reward_std_all": 0.1497473269701004,
      "rewards-max/forensics_sinkhorn_reward": 0.974577933549881,
      "rewards-max/forensics_strict_edge_reward": 0.687186524271965,
      "rewards-max/forensics_tiou_reward": 0.9443953633308411,
      "rewards/forensics_sinkhorn_reward": 0.5355427265167236,
      "rewards/forensics_strict_edge_reward": 0.14940496906638145,
      "rewards/forensics_tiou_reward": 0.23083191365003586,
      "step": 122
    },
    {
      "completion_length": 17.71875,
      "epoch": 0.32756324900133155,
      "grad_norm": 13.229665756225586,
      "kl": 1.197265625,
      "learning_rate": 9.186666666666666e-07,
      "loss": 0.1306,
      "reward": 0.5807770490646362,
      "reward_std": 0.1336458995938301,
      "reward_std_all": 0.1656293421983719,
      "rewards-max/forensics_sinkhorn_reward": 0.9010693430900574,
      "rewards-max/forensics_strict_edge_reward": 0.6649194061756134,
      "rewards-max/forensics_tiou_reward": 0.8090681731700897,
      "rewards/forensics_sinkhorn_reward": 0.4274047017097473,
      "rewards/forensics_strict_edge_reward": 0.06406990997493267,
      "rewards/forensics_tiou_reward": 0.11993379145860672,
      "step": 123
    },
    {
      "completion_length": 15.875,
      "epoch": 0.33022636484687085,
      "grad_norm": 8.637514114379883,
      "kl": 0.42431640625,
      "learning_rate": 9.18e-07,
      "loss": 0.164,
      "reward": 0.6269934773445129,
      "reward_std": 0.1281859502196312,
      "reward_std_all": 0.14679620787501335,
      "rewards-max/forensics_sinkhorn_reward": 0.987324982881546,
      "rewards-max/forensics_strict_edge_reward": 0.9565287232398987,
      "rewards-max/forensics_tiou_reward": 0.9934432208538055,
      "rewards/forensics_sinkhorn_reward": 0.5168432146310806,
      "rewards/forensics_strict_edge_reward": 0.16061414033174515,
      "rewards/forensics_tiou_reward": 0.23638136684894562,
      "step": 124
    },
    {
      "completion_length": 18.453125,
      "epoch": 0.33288948069241014,
      "grad_norm": 8.772561073303223,
      "kl": 0.5322265625,
      "learning_rate": 9.173333333333333e-07,
      "loss": 0.1247,
      "reward": 0.5645075440406799,
      "reward_std": 0.20972448587417603,
      "reward_std_all": 0.22194283455610275,
      "rewards-max/forensics_sinkhorn_reward": 0.8497481942176819,
      "rewards-max/forensics_strict_edge_reward": 0.38988301157951355,
      "rewards-max/forensics_tiou_reward": 0.6991802453994751,
      "rewards/forensics_sinkhorn_reward": 0.41504649817943573,
      "rewards/forensics_strict_edge_reward": 0.07949817925691605,
      "rewards/forensics_tiou_reward": 0.17153233289718628,
      "step": 125
    },
    {
      "completion_length": 17.671875,
      "epoch": 0.3355525965379494,
      "grad_norm": 13.942354202270508,
      "kl": 0.92578125,
      "learning_rate": 9.166666666666665e-07,
      "loss": 0.1185,
      "reward": 0.6211750209331512,
      "reward_std": 0.19128531962633133,
      "reward_std_all": 0.18473981320858002,
      "rewards-max/forensics_sinkhorn_reward": 0.8480602204799652,
      "rewards-max/forensics_strict_edge_reward": 0.8096016347408295,
      "rewards-max/forensics_tiou_reward": 0.7055127620697021,
      "rewards/forensics_sinkhorn_reward": 0.48219457268714905,
      "rewards/forensics_strict_edge_reward": 0.1609758958220482,
      "rewards/forensics_tiou_reward": 0.2471287027001381,
      "step": 126
    },
    {
      "completion_length": 16.234375,
      "epoch": 0.3382157123834887,
      "grad_norm": 9.191665649414062,
      "kl": 0.693359375,
      "learning_rate": 9.16e-07,
      "loss": 0.1336,
      "reward": 0.611801028251648,
      "reward_std": 0.105014618486166,
      "reward_std_all": 0.14797919988632202,
      "rewards-max/forensics_sinkhorn_reward": 0.9868468344211578,
      "rewards-max/forensics_strict_edge_reward": 0.9357379674911499,
      "rewards-max/forensics_tiou_reward": 0.9915505349636078,
      "rewards/forensics_sinkhorn_reward": 0.5212115049362183,
      "rewards/forensics_strict_edge_reward": 0.18089134991168976,
      "rewards/forensics_tiou_reward": 0.2607821524143219,
      "step": 127
    },
    {
      "completion_length": 17.046875,
      "epoch": 0.340878828229028,
      "grad_norm": 18.406267166137695,
      "kl": 1.12890625,
      "learning_rate": 9.153333333333332e-07,
      "loss": 0.231,
      "reward": 0.5939473211765289,
      "reward_std": 0.14601558446884155,
      "reward_std_all": 0.17344050854444504,
      "rewards-max/forensics_sinkhorn_reward": 0.9865024983882904,
      "rewards-max/forensics_strict_edge_reward": 0.8983768224716187,
      "rewards-max/forensics_tiou_reward": 0.9902258217334747,
      "rewards/forensics_sinkhorn_reward": 0.4985857754945755,
      "rewards/forensics_strict_edge_reward": 0.16259410604834557,
      "rewards/forensics_tiou_reward": 0.2607378140091896,
      "step": 128
    },
    {
      "completion_length": 18.5625,
      "epoch": 0.34354194407456723,
      "grad_norm": 7.401249408721924,
      "kl": 0.6123046875,
      "learning_rate": 9.146666666666666e-07,
      "loss": 0.0858,
      "reward": 0.6305528879165649,
      "reward_std": 0.1698264628648758,
      "reward_std_all": 0.1745184063911438,
      "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": 0.5194219797849655,
      "rewards/forensics_strict_edge_reward": 0.15488074719905853,
      "rewards/forensics_tiou_reward": 0.27005791664123535,
      "step": 129
    },
    {
      "completion_length": 18.890625,
      "epoch": 0.34620505992010653,
      "grad_norm": 11.490357398986816,
      "kl": 0.927734375,
      "learning_rate": 9.14e-07,
      "loss": -0.0075,
      "reward": 0.6121405065059662,
      "reward_std": 0.1556042581796646,
      "reward_std_all": 0.1652020514011383,
      "rewards-max/forensics_sinkhorn_reward": 0.9822929501533508,
      "rewards-max/forensics_strict_edge_reward": 0.7833328545093536,
      "rewards-max/forensics_tiou_reward": 0.9737075865268707,
      "rewards/forensics_sinkhorn_reward": 0.4984126538038254,
      "rewards/forensics_strict_edge_reward": 0.08636795729398727,
      "rewards/forensics_tiou_reward": 0.17130615562200546,
      "step": 130
    },
    {
      "completion_length": 17.96875,
      "epoch": 0.3488681757656458,
      "grad_norm": 11.168463706970215,
      "kl": 0.677734375,
      "learning_rate": 9.133333333333333e-07,
      "loss": 0.1077,
      "reward": 0.6184626817703247,
      "reward_std": 0.15379439294338226,
      "reward_std_all": 0.1675512045621872,
      "rewards-max/forensics_sinkhorn_reward": 0.9215633273124695,
      "rewards-max/forensics_strict_edge_reward": 0.6470251679420471,
      "rewards-max/forensics_tiou_reward": 0.8512333035469055,
      "rewards/forensics_sinkhorn_reward": 0.45349636673927307,
      "rewards/forensics_strict_edge_reward": 0.13579514995217323,
      "rewards/forensics_tiou_reward": 0.19381796568632126,
      "step": 131
    },
    {
      "completion_length": 16.328125,
      "epoch": 0.35153129161118507,
      "grad_norm": 12.0525484085083,
      "kl": 0.47265625,
      "learning_rate": 9.126666666666666e-07,
      "loss": 0.1766,
      "reward": 0.6202570199966431,
      "reward_std": 0.13065993413329124,
      "reward_std_all": 0.15730560570955276,
      "rewards-max/forensics_sinkhorn_reward": 0.985043078660965,
      "rewards-max/forensics_strict_edge_reward": 0.8771188855171204,
      "rewards-max/forensics_tiou_reward": 0.9845057725906372,
      "rewards/forensics_sinkhorn_reward": 0.5517975687980652,
      "rewards/forensics_strict_edge_reward": 0.19712130725383759,
      "rewards/forensics_tiou_reward": 0.28945500403642654,
      "step": 132
    },
    {
      "completion_length": 18.171875,
      "epoch": 0.35419440745672437,
      "grad_norm": 11.293721199035645,
      "kl": 0.97265625,
      "learning_rate": 9.12e-07,
      "loss": 0.1179,
      "reward": 0.6540302336215973,
      "reward_std": 0.15447025001049042,
      "reward_std_all": 0.15286847203969955,
      "rewards-max/forensics_sinkhorn_reward": 0.9863657355308533,
      "rewards-max/forensics_strict_edge_reward": 0.8893801271915436,
      "rewards-max/forensics_tiou_reward": 0.9896954596042633,
      "rewards/forensics_sinkhorn_reward": 0.5574354231357574,
      "rewards/forensics_strict_edge_reward": 0.21845519542694092,
      "rewards/forensics_tiou_reward": 0.3082035183906555,
      "step": 133
    },
    {
      "completion_length": 19.796875,
      "epoch": 0.35685752330226367,
      "grad_norm": 8.251396179199219,
      "kl": 0.4951171875,
      "learning_rate": 9.113333333333333e-07,
      "loss": 0.0957,
      "reward": 0.5870692431926727,
      "reward_std": 0.19480377435684204,
      "reward_std_all": 0.20069590955972672,
      "rewards-max/forensics_sinkhorn_reward": 0.966911107301712,
      "rewards-max/forensics_strict_edge_reward": 0.5652900487184525,
      "rewards-max/forensics_tiou_reward": 0.9489380419254303,
      "rewards/forensics_sinkhorn_reward": 0.4256502091884613,
      "rewards/forensics_strict_edge_reward": 0.0866863876581192,
      "rewards/forensics_tiou_reward": 0.2006370723247528,
      "step": 134
    },
    {
      "completion_length": 16.421875,
      "epoch": 0.3595206391478029,
      "grad_norm": 10.865056991577148,
      "kl": 0.7607421875,
      "learning_rate": 9.106666666666666e-07,
      "loss": 0.1503,
      "reward": 0.6171325743198395,
      "reward_std": 0.15488478541374207,
      "reward_std_all": 0.16376054286956787,
      "rewards-max/forensics_sinkhorn_reward": 0.9877389967441559,
      "rewards-max/forensics_strict_edge_reward": 0.9169811904430389,
      "rewards-max/forensics_tiou_reward": 0.9950739145278931,
      "rewards/forensics_sinkhorn_reward": 0.537779688835144,
      "rewards/forensics_strict_edge_reward": 0.2116386741399765,
      "rewards/forensics_tiou_reward": 0.31479479372501373,
      "step": 135
    },
    {
      "completion_length": 18.65625,
      "epoch": 0.3621837549933422,
      "grad_norm": 11.849557876586914,
      "kl": 0.71484375,
      "learning_rate": 9.1e-07,
      "loss": 0.0537,
      "reward": 0.6608247756958008,
      "reward_std": 0.14795905351638794,
      "reward_std_all": 0.146497905254364,
      "rewards-max/forensics_sinkhorn_reward": 0.9886516034603119,
      "rewards-max/forensics_strict_edge_reward": 0.9890114367008209,
      "rewards-max/forensics_tiou_reward": 0.9986648857593536,
      "rewards/forensics_sinkhorn_reward": 0.5504065454006195,
      "rewards/forensics_strict_edge_reward": 0.21732406318187714,
      "rewards/forensics_tiou_reward": 0.32184383273124695,
      "step": 136
    },
    {
      "completion_length": 19.078125,
      "epoch": 0.3648468708388815,
      "grad_norm": 9.685832977294922,
      "kl": 0.912109375,
      "learning_rate": 9.093333333333333e-07,
      "loss": 0.1746,
      "reward": 0.5948807597160339,
      "reward_std": 0.19514260441064835,
      "reward_std_all": 0.19635102897882462,
      "rewards-max/forensics_sinkhorn_reward": 0.9190512299537659,
      "rewards-max/forensics_strict_edge_reward": 0.7391544282436371,
      "rewards-max/forensics_tiou_reward": 0.8239635229110718,
      "rewards/forensics_sinkhorn_reward": 0.41468407213687897,
      "rewards/forensics_strict_edge_reward": 0.1333642117679119,
      "rewards/forensics_tiou_reward": 0.19742940366268158,
      "step": 137
    },
    {
      "completion_length": 15.53125,
      "epoch": 0.36750998668442075,
      "grad_norm": 10.101276397705078,
      "kl": 0.5341796875,
      "learning_rate": 9.086666666666666e-07,
      "loss": 0.1153,
      "reward": 0.6543624997138977,
      "reward_std": 0.16709847748279572,
      "reward_std_all": 0.16118331253528595,
      "rewards-max/forensics_sinkhorn_reward": 0.9874891936779022,
      "rewards-max/forensics_strict_edge_reward": 0.9075756669044495,
      "rewards-max/forensics_tiou_reward": 0.9940910041332245,
      "rewards/forensics_sinkhorn_reward": 0.5473982393741608,
      "rewards/forensics_strict_edge_reward": 0.20893698185682297,
      "rewards/forensics_tiou_reward": 0.2796097621321678,
      "step": 138
    },
    {
      "completion_length": 19.109375,
      "epoch": 0.37017310252996005,
      "grad_norm": 8.84184741973877,
      "kl": 0.732421875,
      "learning_rate": 9.08e-07,
      "loss": 0.1959,
      "reward": 0.577239066362381,
      "reward_std": 0.15800420194864273,
      "reward_std_all": 0.18663863837718964,
      "rewards-max/forensics_sinkhorn_reward": 0.9760152995586395,
      "rewards-max/forensics_strict_edge_reward": 0.7224751859903336,
      "rewards-max/forensics_tiou_reward": 0.9500325918197632,
      "rewards/forensics_sinkhorn_reward": 0.4420379102230072,
      "rewards/forensics_strict_edge_reward": 0.10747440904378891,
      "rewards/forensics_tiou_reward": 0.1933157593011856,
      "step": 139
    },
    {
      "completion_length": 19.59375,
      "epoch": 0.37283621837549935,
      "grad_norm": 11.06269645690918,
      "kl": 1.009765625,
      "learning_rate": 9.073333333333333e-07,
      "loss": 0.172,
      "reward": 0.5686860084533691,
      "reward_std": 0.19631150364875793,
      "reward_std_all": 0.21272170543670654,
      "rewards-max/forensics_sinkhorn_reward": 0.9621439874172211,
      "rewards-max/forensics_strict_edge_reward": 0.8593232333660126,
      "rewards-max/forensics_tiou_reward": 0.9390334486961365,
      "rewards/forensics_sinkhorn_reward": 0.430966317653656,
      "rewards/forensics_strict_edge_reward": 0.10315623506903648,
      "rewards/forensics_tiou_reward": 0.16504255682229996,
      "step": 140
    },
    {
      "completion_length": 15.546875,
      "epoch": 0.3754993342210386,
      "grad_norm": 13.227934837341309,
      "kl": 0.853515625,
      "learning_rate": 9.066666666666665e-07,
      "loss": 0.086,
      "reward": 0.6173641085624695,
      "reward_std": 0.15630901232361794,
      "reward_std_all": 0.16470056772232056,
      "rewards-max/forensics_sinkhorn_reward": 0.9881953299045563,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9968656599521637,
      "rewards/forensics_sinkhorn_reward": 0.523371547460556,
      "rewards/forensics_strict_edge_reward": 0.17846593260765076,
      "rewards/forensics_tiou_reward": 0.27311596274375916,
      "step": 141
    },
    {
      "completion_length": 19.140625,
      "epoch": 0.3781624500665779,
      "grad_norm": 8.937053680419922,
      "kl": 0.681640625,
      "learning_rate": 9.06e-07,
      "loss": 0.1801,
      "reward": 0.6088562607765198,
      "reward_std": 0.21294791996479034,
      "reward_std_all": 0.20458849519491196,
      "rewards-max/forensics_sinkhorn_reward": 0.9844389855861664,
      "rewards-max/forensics_strict_edge_reward": 0.8850063383579254,
      "rewards-max/forensics_tiou_reward": 0.9820950031280518,
      "rewards/forensics_sinkhorn_reward": 0.4912750720977783,
      "rewards/forensics_strict_edge_reward": 0.19559206068515778,
      "rewards/forensics_tiou_reward": 0.28198929131031036,
      "step": 142
    },
    {
      "completion_length": 15.390625,
      "epoch": 0.3808255659121172,
      "grad_norm": 12.466411590576172,
      "kl": 1.02734375,
      "learning_rate": 9.053333333333332e-07,
      "loss": 0.2653,
      "reward": 0.6728435456752777,
      "reward_std": 0.1437450349330902,
      "reward_std_all": 0.1383320316672325,
      "rewards-max/forensics_sinkhorn_reward": 0.987404078245163,
      "rewards-max/forensics_strict_edge_reward": 0.9565287232398987,
      "rewards-max/forensics_tiou_reward": 0.9937525689601898,
      "rewards/forensics_sinkhorn_reward": 0.5829667448997498,
      "rewards/forensics_strict_edge_reward": 0.2706212028861046,
      "rewards/forensics_tiou_reward": 0.3527804911136627,
      "step": 143
    },
    {
      "completion_length": 18.140625,
      "epoch": 0.38348868175765644,
      "grad_norm": 9.386006355285645,
      "kl": 0.701171875,
      "learning_rate": 9.046666666666666e-07,
      "loss": 0.1747,
      "reward": 0.611629992723465,
      "reward_std": 0.18621128052473068,
      "reward_std_all": 0.1860935017466545,
      "rewards-max/forensics_sinkhorn_reward": 0.9552089869976044,
      "rewards-max/forensics_strict_edge_reward": 0.5316388309001923,
      "rewards-max/forensics_tiou_reward": 0.9168148636817932,
      "rewards/forensics_sinkhorn_reward": 0.48132118582725525,
      "rewards/forensics_strict_edge_reward": 0.13193807750940323,
      "rewards/forensics_tiou_reward": 0.23624424636363983,
      "step": 144
    },
    {
      "completion_length": 18.03125,
      "epoch": 0.38615179760319573,
      "grad_norm": 12.756590843200684,
      "kl": 1.109375,
      "learning_rate": 9.039999999999999e-07,
      "loss": 0.0131,
      "reward": 0.610366940498352,
      "reward_std": 0.16584182530641556,
      "reward_std_all": 0.17779995501041412,
      "rewards-max/forensics_sinkhorn_reward": 0.9876776933670044,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9948250353336334,
      "rewards/forensics_sinkhorn_reward": 0.5031227171421051,
      "rewards/forensics_strict_edge_reward": 0.17421066015958786,
      "rewards/forensics_tiou_reward": 0.24405347555875778,
      "step": 145
    },
    {
      "completion_length": 16.0,
      "epoch": 0.38881491344873503,
      "grad_norm": 11.894669532775879,
      "kl": 0.798828125,
      "learning_rate": 9.033333333333333e-07,
      "loss": 0.0396,
      "reward": 0.65998175740242,
      "reward_std": 0.11536695435643196,
      "reward_std_all": 0.1268751621246338,
      "rewards-max/forensics_sinkhorn_reward": 0.9827696681022644,
      "rewards-max/forensics_strict_edge_reward": 0.9565287232398987,
      "rewards-max/forensics_tiou_reward": 0.984478622674942,
      "rewards/forensics_sinkhorn_reward": 0.5438460409641266,
      "rewards/forensics_strict_edge_reward": 0.18395588546991348,
      "rewards/forensics_tiou_reward": 0.24572952836751938,
      "step": 146
    },
    {
      "completion_length": 16.296875,
      "epoch": 0.3914780292942743,
      "grad_norm": 13.032052993774414,
      "kl": 1.265625,
      "learning_rate": 9.026666666666665e-07,
      "loss": 0.1613,
      "reward": 0.634424239397049,
      "reward_std": 0.10681340470910072,
      "reward_std_all": 0.13388223573565483,
      "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": 0.5514074563980103,
      "rewards/forensics_strict_edge_reward": 0.19617781043052673,
      "rewards/forensics_tiou_reward": 0.2953443229198456,
      "step": 147
    },
    {
      "completion_length": 19.53125,
      "epoch": 0.3941411451398136,
      "grad_norm": 9.067607879638672,
      "kl": 0.765625,
      "learning_rate": 9.02e-07,
      "loss": 0.1607,
      "reward": 0.6390457451343536,
      "reward_std": 0.14341934770345688,
      "reward_std_all": 0.15232424065470695,
      "rewards-max/forensics_sinkhorn_reward": 0.9310151934623718,
      "rewards-max/forensics_strict_edge_reward": 0.7791877388954163,
      "rewards-max/forensics_tiou_reward": 0.8701453804969788,
      "rewards/forensics_sinkhorn_reward": 0.5033949464559555,
      "rewards/forensics_strict_edge_reward": 0.26651300489902496,
      "rewards/forensics_tiou_reward": 0.3238460123538971,
      "step": 148
    },
    {
      "completion_length": 19.34375,
      "epoch": 0.3968042609853529,
      "grad_norm": 10.68254566192627,
      "kl": 0.63671875,
      "learning_rate": 9.013333333333333e-07,
      "loss": 0.1864,
      "reward": 0.5966029465198517,
      "reward_std": 0.12855324149131775,
      "reward_std_all": 0.1611739844083786,
      "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": 0.47554272413253784,
      "rewards/forensics_strict_edge_reward": 0.19219446927309036,
      "rewards/forensics_tiou_reward": 0.269194096326828,
      "step": 149
    },
    {
      "completion_length": 17.390625,
      "epoch": 0.3994673768308921,
      "grad_norm": 12.780250549316406,
      "kl": 0.712890625,
      "learning_rate": 9.006666666666666e-07,
      "loss": 0.2818,
      "reward": 0.6693755686283112,
      "reward_std": 0.09151444584131241,
      "reward_std_all": 0.10907264798879623,
      "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": 0.5718741714954376,
      "rewards/forensics_strict_edge_reward": 0.20701617747545242,
      "rewards/forensics_tiou_reward": 0.31703953444957733,
      "step": 150
    },
    {
      "completion_length": 18.625,
      "epoch": 0.4021304926764314,
      "grad_norm": 10.040881156921387,
      "kl": 0.67578125,
      "learning_rate": 9e-07,
      "loss": 0.0708,
      "reward": 0.6432290375232697,
      "reward_std": 0.11693793535232544,
      "reward_std_all": 0.1355224996805191,
      "rewards-max/forensics_sinkhorn_reward": 0.9880679547786713,
      "rewards-max/forensics_strict_edge_reward": 0.9677534997463226,
      "rewards-max/forensics_tiou_reward": 0.996363639831543,
      "rewards/forensics_sinkhorn_reward": 0.5653112232685089,
      "rewards/forensics_strict_edge_reward": 0.18285630643367767,
      "rewards/forensics_tiou_reward": 0.29547637701034546,
      "step": 151
    },
    {
      "completion_length": 18.734375,
      "epoch": 0.4047936085219707,
      "grad_norm": 14.608660697937012,
      "kl": 1.166015625,
      "learning_rate": 8.993333333333333e-07,
      "loss": 0.1175,
      "reward": 0.6569816768169403,
      "reward_std": 0.14663168787956238,
      "reward_std_all": 0.1468176692724228,
      "rewards-max/forensics_sinkhorn_reward": 0.9841423332691193,
      "rewards-max/forensics_strict_edge_reward": 0.8915697336196899,
      "rewards-max/forensics_tiou_reward": 0.9810851216316223,
      "rewards/forensics_sinkhorn_reward": 0.5281596481800079,
      "rewards/forensics_strict_edge_reward": 0.21203437447547913,
      "rewards/forensics_tiou_reward": 0.33333931863307953,
      "step": 152
    },
    {
      "completion_length": 21.609375,
      "epoch": 0.40745672436750996,
      "grad_norm": 7.318052291870117,
      "kl": 0.787109375,
      "learning_rate": 8.986666666666666e-07,
      "loss": 0.1259,
      "reward": 0.6338475048542023,
      "reward_std": 0.19117175042629242,
      "reward_std_all": 0.18538255244493484,
      "rewards-max/forensics_sinkhorn_reward": 0.9882158637046814,
      "rewards-max/forensics_strict_edge_reward": 0.9265981018543243,
      "rewards-max/forensics_tiou_reward": 0.9969489872455597,
      "rewards/forensics_sinkhorn_reward": 0.505912721157074,
      "rewards/forensics_strict_edge_reward": 0.18250516802072525,
      "rewards/forensics_tiou_reward": 0.24361714720726013,
      "step": 153
    },
    {
      "completion_length": 16.375,
      "epoch": 0.41011984021304926,
      "grad_norm": 13.255621910095215,
      "kl": 0.5234375,
      "learning_rate": 8.98e-07,
      "loss": 0.0345,
      "reward": 0.6390739977359772,
      "reward_std": 0.10263493284583092,
      "reward_std_all": 0.13265854120254517,
      "rewards-max/forensics_sinkhorn_reward": 0.986718475818634,
      "rewards-max/forensics_strict_edge_reward": 0.87538942694664,
      "rewards-max/forensics_tiou_reward": 0.9910454750061035,
      "rewards/forensics_sinkhorn_reward": 0.529874250292778,
      "rewards/forensics_strict_edge_reward": 0.1380084529519081,
      "rewards/forensics_tiou_reward": 0.2133231908082962,
      "step": 154
    },
    {
      "completion_length": 16.859375,
      "epoch": 0.41278295605858856,
      "grad_norm": 12.115199089050293,
      "kl": 0.87109375,
      "learning_rate": 8.973333333333333e-07,
      "loss": 0.0389,
      "reward": 0.6638091802597046,
      "reward_std": 0.14143352210521698,
      "reward_std_all": 0.14032064378261566,
      "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": 0.5492680966854095,
      "rewards/forensics_strict_edge_reward": 0.2006857842206955,
      "rewards/forensics_tiou_reward": 0.25590141117572784,
      "step": 155
    },
    {
      "completion_length": 16.515625,
      "epoch": 0.41544607190412786,
      "grad_norm": 11.642334938049316,
      "kl": 0.984375,
      "learning_rate": 8.966666666666666e-07,
      "loss": 0.0074,
      "reward": 0.687284380197525,
      "reward_std": 0.08348850160837173,
      "reward_std_all": 0.1051802933216095,
      "rewards-max/forensics_sinkhorn_reward": 0.9686859846115112,
      "rewards-max/forensics_strict_edge_reward": 0.8719756007194519,
      "rewards-max/forensics_tiou_reward": 0.9546608328819275,
      "rewards/forensics_sinkhorn_reward": 0.583355039358139,
      "rewards/forensics_strict_edge_reward": 0.21107475459575653,
      "rewards/forensics_tiou_reward": 0.28400805592536926,
      "step": 156
    },
    {
      "completion_length": 17.1875,
      "epoch": 0.4181091877496671,
      "grad_norm": 10.854626655578613,
      "kl": 1.013671875,
      "learning_rate": 8.96e-07,
      "loss": -0.0009,
      "reward": 0.5970537662506104,
      "reward_std": 0.14197160303592682,
      "reward_std_all": 0.16897539794445038,
      "rewards-max/forensics_sinkhorn_reward": 0.8292636871337891,
      "rewards-max/forensics_strict_edge_reward": 0.7078047692775726,
      "rewards-max/forensics_tiou_reward": 0.6788863986730576,
      "rewards/forensics_sinkhorn_reward": 0.47954779863357544,
      "rewards/forensics_strict_edge_reward": 0.13940194621682167,
      "rewards/forensics_tiou_reward": 0.19457513839006424,
      "step": 157
    },
    {
      "completion_length": 21.96875,
      "epoch": 0.4207723035952064,
      "grad_norm": 10.695926666259766,
      "kl": 1.15625,
      "learning_rate": 8.953333333333332e-07,
      "loss": 0.0882,
      "reward": 0.6619275808334351,
      "reward_std": 0.12822424620389938,
      "reward_std_all": 0.13525881990790367,
      "rewards-max/forensics_sinkhorn_reward": 0.9243486225605011,
      "rewards-max/forensics_strict_edge_reward": 0.7457199990749359,
      "rewards-max/forensics_tiou_reward": 0.8585122525691986,
      "rewards/forensics_sinkhorn_reward": 0.4948897361755371,
      "rewards/forensics_strict_edge_reward": 0.16533119603991508,
      "rewards/forensics_tiou_reward": 0.2431422919034958,
      "step": 158
    },
    {
      "completion_length": 16.46875,
      "epoch": 0.4234354194407457,
      "grad_norm": 10.715901374816895,
      "kl": 0.962890625,
      "learning_rate": 8.946666666666667e-07,
      "loss": 0.1369,
      "reward": 0.6546039879322052,
      "reward_std": 0.11721326783299446,
      "reward_std_all": 0.13611837476491928,
      "rewards-max/forensics_sinkhorn_reward": 0.9849734604358673,
      "rewards-max/forensics_strict_edge_reward": 0.9357379674911499,
      "rewards-max/forensics_tiou_reward": 0.9842817187309265,
      "rewards/forensics_sinkhorn_reward": 0.586457371711731,
      "rewards/forensics_strict_edge_reward": 0.1920478343963623,
      "rewards/forensics_tiou_reward": 0.3264385163784027,
      "step": 159
    },
    {
      "completion_length": 17.765625,
      "epoch": 0.42609853528628494,
      "grad_norm": 9.478907585144043,
      "kl": 0.5693359375,
      "learning_rate": 8.939999999999999e-07,
      "loss": 0.1476,
      "reward": 0.6015924215316772,
      "reward_std": 0.1243014857172966,
      "reward_std_all": 0.15988831222057343,
      "rewards-max/forensics_sinkhorn_reward": 0.9871755838394165,
      "rewards-max/forensics_strict_edge_reward": 0.946017861366272,
      "rewards-max/forensics_tiou_reward": 0.9928461313247681,
      "rewards/forensics_sinkhorn_reward": 0.5139853656291962,
      "rewards/forensics_strict_edge_reward": 0.19517754763364792,
      "rewards/forensics_tiou_reward": 0.2614968866109848,
      "step": 160
    },
    {
      "completion_length": 17.375,
      "epoch": 0.42876165113182424,
      "grad_norm": 12.976717948913574,
      "kl": 0.5556640625,
      "learning_rate": 8.933333333333333e-07,
      "loss": 0.0096,
      "reward": 0.6413801908493042,
      "reward_std": 0.1412414312362671,
      "reward_std_all": 0.1515086442232132,
      "rewards-max/forensics_sinkhorn_reward": 0.9397855997085571,
      "rewards-max/forensics_strict_edge_reward": 0.7389469742774963,
      "rewards-max/forensics_tiou_reward": 0.8924753963947296,
      "rewards/forensics_sinkhorn_reward": 0.5236174315214157,
      "rewards/forensics_strict_edge_reward": 0.1634555086493492,
      "rewards/forensics_tiou_reward": 0.23718616366386414,
      "step": 161
    },
    {
      "completion_length": 17.28125,
      "epoch": 0.43142476697736354,
      "grad_norm": 7.391456604003906,
      "kl": 0.52734375,
      "learning_rate": 8.926666666666666e-07,
      "loss": 0.0636,
      "reward": 0.6241860389709473,
      "reward_std": 0.08136438950896263,
      "reward_std_all": 0.1323472335934639,
      "rewards-max/forensics_sinkhorn_reward": 0.9828758239746094,
      "rewards-max/forensics_strict_edge_reward": 0.759757399559021,
      "rewards-max/forensics_tiou_reward": 0.9760767221450806,
      "rewards/forensics_sinkhorn_reward": 0.5115838795900345,
      "rewards/forensics_strict_edge_reward": 0.11605716869235039,
      "rewards/forensics_tiou_reward": 0.20922739803791046,
      "step": 162
    },
    {
      "completion_length": 16.59375,
      "epoch": 0.4340878828229028,
      "grad_norm": 8.92918872833252,
      "kl": 0.634765625,
      "learning_rate": 8.92e-07,
      "loss": 0.0786,
      "reward": 0.6441163718700409,
      "reward_std": 0.12973789125680923,
      "reward_std_all": 0.1444825455546379,
      "rewards-max/forensics_sinkhorn_reward": 0.9840749800205231,
      "rewards-max/forensics_strict_edge_reward": 0.8189329206943512,
      "rewards-max/forensics_tiou_reward": 0.9806761145591736,
      "rewards/forensics_sinkhorn_reward": 0.5236266553401947,
      "rewards/forensics_strict_edge_reward": 0.1754131317138672,
      "rewards/forensics_tiou_reward": 0.26739057898521423,
      "step": 163
    },
    {
      "completion_length": 17.671875,
      "epoch": 0.4367509986684421,
      "grad_norm": 8.834824562072754,
      "kl": 0.7490234375,
      "learning_rate": 8.913333333333332e-07,
      "loss": 0.0226,
      "reward": 0.5934005379676819,
      "reward_std": 0.1641082689166069,
      "reward_std_all": 0.18429335951805115,
      "rewards-max/forensics_sinkhorn_reward": 0.9139218330383301,
      "rewards-max/forensics_strict_edge_reward": 0.7321087718009949,
      "rewards-max/forensics_tiou_reward": 0.836888998746872,
      "rewards/forensics_sinkhorn_reward": 0.45029835402965546,
      "rewards/forensics_strict_edge_reward": 0.14641280472278595,
      "rewards/forensics_tiou_reward": 0.192075215280056,
      "step": 164
    },
    {
      "completion_length": 16.671875,
      "epoch": 0.4394141145139814,
      "grad_norm": 9.177244186401367,
      "kl": 0.80859375,
      "learning_rate": 8.906666666666667e-07,
      "loss": 0.1752,
      "reward": 0.5922981202602386,
      "reward_std": 0.16736627370119095,
      "reward_std_all": 0.18697555363178253,
      "rewards-max/forensics_sinkhorn_reward": 0.9866192638874054,
      "rewards-max/forensics_strict_edge_reward": 0.9364311099052429,
      "rewards-max/forensics_tiou_reward": 0.9906830489635468,
      "rewards/forensics_sinkhorn_reward": 0.511687844991684,
      "rewards/forensics_strict_edge_reward": 0.20964086055755615,
      "rewards/forensics_tiou_reward": 0.2968127503991127,
      "step": 165
    },
    {
      "completion_length": 17.140625,
      "epoch": 0.4420772303595206,
      "grad_norm": 10.855785369873047,
      "kl": 0.5068359375,
      "learning_rate": 8.9e-07,
      "loss": 0.0904,
      "reward": 0.6434586942195892,
      "reward_std": 0.14070183038711548,
      "reward_std_all": 0.14448942989110947,
      "rewards-max/forensics_sinkhorn_reward": 0.9873698353767395,
      "rewards-max/forensics_strict_edge_reward": 0.9565287232398987,
      "rewards-max/forensics_tiou_reward": 0.9936116635799408,
      "rewards/forensics_sinkhorn_reward": 0.5864962935447693,
      "rewards/forensics_strict_edge_reward": 0.24594639986753464,
      "rewards/forensics_tiou_reward": 0.36951233446598053,
      "step": 166
    },
    {
      "completion_length": 15.828125,
      "epoch": 0.4447403462050599,
      "grad_norm": 10.205177307128906,
      "kl": 0.5078125,
      "learning_rate": 8.893333333333333e-07,
      "loss": 0.1048,
      "reward": 0.6525287926197052,
      "reward_std": 0.149662833660841,
      "reward_std_all": 0.1512243002653122,
      "rewards-max/forensics_sinkhorn_reward": 0.9871321320533752,
      "rewards-max/forensics_strict_edge_reward": 0.9265981018543243,
      "rewards-max/forensics_tiou_reward": 0.9926958084106445,
      "rewards/forensics_sinkhorn_reward": 0.5564461946487427,
      "rewards/forensics_strict_edge_reward": 0.2169121354818344,
      "rewards/forensics_tiou_reward": 0.28790727257728577,
      "step": 167
    },
    {
      "completion_length": 19.96875,
      "epoch": 0.4474034620505992,
      "grad_norm": 10.571441650390625,
      "kl": 0.755859375,
      "learning_rate": 8.886666666666667e-07,
      "loss": 0.2128,
      "reward": 0.6419511437416077,
      "reward_std": 0.1827227845788002,
      "reward_std_all": 0.1765345185995102,
      "rewards-max/forensics_sinkhorn_reward": 0.9143511652946472,
      "rewards-max/forensics_strict_edge_reward": 0.5868593752384186,
      "rewards-max/forensics_tiou_reward": 0.8300602436065674,
      "rewards/forensics_sinkhorn_reward": 0.5226549357175827,
      "rewards/forensics_strict_edge_reward": 0.13758041709661484,
      "rewards/forensics_tiou_reward": 0.24301570653915405,
      "step": 168
    },
    {
      "completion_length": 16.765625,
      "epoch": 0.45006657789613846,
      "grad_norm": 11.521516799926758,
      "kl": 0.689453125,
      "learning_rate": 8.88e-07,
      "loss": 0.1969,
      "reward": 0.6422907114028931,
      "reward_std": 0.12418733164668083,
      "reward_std_all": 0.1413561999797821,
      "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": 0.5347932577133179,
      "rewards/forensics_strict_edge_reward": 0.2399127185344696,
      "rewards/forensics_tiou_reward": 0.2851094529032707,
      "step": 169
    },
    {
      "completion_length": 20.921875,
      "epoch": 0.45272969374167776,
      "grad_norm": 8.08419418334961,
      "kl": 0.681640625,
      "learning_rate": 8.873333333333333e-07,
      "loss": 0.0788,
      "reward": 0.578255832195282,
      "reward_std": 0.19356747716665268,
      "reward_std_all": 0.200973279774189,
      "rewards-max/forensics_sinkhorn_reward": 0.8873681128025055,
      "rewards-max/forensics_strict_edge_reward": 0.5277700275182724,
      "rewards-max/forensics_tiou_reward": 0.7795004546642303,
      "rewards/forensics_sinkhorn_reward": 0.3731842488050461,
      "rewards/forensics_strict_edge_reward": 0.10780490189790726,
      "rewards/forensics_tiou_reward": 0.17692920565605164,
      "step": 170
    },
    {
      "completion_length": 16.671875,
      "epoch": 0.45539280958721706,
      "grad_norm": 10.212274551391602,
      "kl": 1.3359375,
      "learning_rate": 8.866666666666667e-07,
      "loss": 0.1723,
      "reward": 0.6043387949466705,
      "reward_std": 0.1693563163280487,
      "reward_std_all": 0.18295595049858093,
      "rewards-max/forensics_sinkhorn_reward": 0.9808246791362762,
      "rewards-max/forensics_strict_edge_reward": 0.8528120815753937,
      "rewards-max/forensics_tiou_reward": 0.9678453207015991,
      "rewards/forensics_sinkhorn_reward": 0.49662622809410095,
      "rewards/forensics_strict_edge_reward": 0.164702907204628,
      "rewards/forensics_tiou_reward": 0.2385300099849701,
      "step": 171
    },
    {
      "completion_length": 20.40625,
      "epoch": 0.4580559254327563,
      "grad_norm": 9.866545677185059,
      "kl": 0.677734375,
      "learning_rate": 8.86e-07,
      "loss": 0.1527,
      "reward": 0.6349680721759796,
      "reward_std": 0.15000338852405548,
      "reward_std_all": 0.16108186542987823,
      "rewards-max/forensics_sinkhorn_reward": 0.9872516095638275,
      "rewards-max/forensics_strict_edge_reward": 0.946017861366272,
      "rewards-max/forensics_tiou_reward": 0.9931583106517792,
      "rewards/forensics_sinkhorn_reward": 0.517552062869072,
      "rewards/forensics_strict_edge_reward": 0.26499226689338684,
      "rewards/forensics_tiou_reward": 0.34756170213222504,
      "step": 172
    },
    {
      "completion_length": 18.96875,
      "epoch": 0.4607190412782956,
      "grad_norm": 12.227088928222656,
      "kl": 1.017578125,
      "learning_rate": 8.853333333333332e-07,
      "loss": 0.2398,
      "reward": 0.5639435648918152,
      "reward_std": 0.17983874678611755,
      "reward_std_all": 0.2075464129447937,
      "rewards-max/forensics_sinkhorn_reward": 0.9608168303966522,
      "rewards-max/forensics_strict_edge_reward": 0.7793509364128113,
      "rewards-max/forensics_tiou_reward": 0.9171469509601593,
      "rewards/forensics_sinkhorn_reward": 0.44148804247379303,
      "rewards/forensics_strict_edge_reward": 0.1527269296348095,
      "rewards/forensics_tiou_reward": 0.24641678482294083,
      "step": 173
    },
    {
      "completion_length": 21.203125,
      "epoch": 0.4633821571238349,
      "grad_norm": 14.917133331298828,
      "kl": 0.73828125,
      "learning_rate": 8.846666666666667e-07,
      "loss": 0.1521,
      "reward": 0.584745466709137,
      "reward_std": 0.17143654823303223,
      "reward_std_all": 0.19231490790843964,
      "rewards-max/forensics_sinkhorn_reward": 0.9869163036346436,
      "rewards-max/forensics_strict_edge_reward": 0.9256840348243713,
      "rewards-max/forensics_tiou_reward": 0.9918412864208221,
      "rewards/forensics_sinkhorn_reward": 0.45531754195690155,
      "rewards/forensics_strict_edge_reward": 0.17894718796014786,
      "rewards/forensics_tiou_reward": 0.24016042798757553,
      "step": 174
    },
    {
      "completion_length": 21.046875,
      "epoch": 0.46604527296937415,
      "grad_norm": 8.023601531982422,
      "kl": 0.802734375,
      "learning_rate": 8.839999999999999e-07,
      "loss": 0.1161,
      "reward": 0.6148205697536469,
      "reward_std": 0.19362442195415497,
      "reward_std_all": 0.19404766708612442,
      "rewards-max/forensics_sinkhorn_reward": 0.9850809574127197,
      "rewards-max/forensics_strict_edge_reward": 0.8983768224716187,
      "rewards-max/forensics_tiou_reward": 0.9846814870834351,
      "rewards/forensics_sinkhorn_reward": 0.48998573422431946,
      "rewards/forensics_strict_edge_reward": 0.20014021545648575,
      "rewards/forensics_tiou_reward": 0.25915108621120453,
      "step": 175
    },
    {
      "completion_length": 18.6875,
      "epoch": 0.46870838881491345,
      "grad_norm": 15.224766731262207,
      "kl": 1.54296875,
      "learning_rate": 8.833333333333333e-07,
      "loss": 0.2604,
      "reward": 0.6423448324203491,
      "reward_std": 0.1598030850291252,
      "reward_std_all": 0.1597236916422844,
      "rewards-max/forensics_sinkhorn_reward": 0.9880439043045044,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9962712526321411,
      "rewards/forensics_sinkhorn_reward": 0.5767391920089722,
      "rewards/forensics_strict_edge_reward": 0.22710277140140533,
      "rewards/forensics_tiou_reward": 0.3329036682844162,
      "step": 176
    },
    {
      "completion_length": 17.1875,
      "epoch": 0.47137150466045274,
      "grad_norm": 13.046853065490723,
      "kl": 0.70703125,
      "learning_rate": 8.826666666666666e-07,
      "loss": -0.0819,
      "reward": 0.6493583619594574,
      "reward_std": 0.12563341483473778,
      "reward_std_all": 0.13864073902368546,
      "rewards-max/forensics_sinkhorn_reward": 0.9714928269386292,
      "rewards-max/forensics_strict_edge_reward": 0.8876297473907471,
      "rewards-max/forensics_tiou_reward": 0.9589625000953674,
      "rewards/forensics_sinkhorn_reward": 0.5332195162773132,
      "rewards/forensics_strict_edge_reward": 0.14585648104548454,
      "rewards/forensics_tiou_reward": 0.21151631325483322,
      "step": 177
    },
    {
      "completion_length": 20.40625,
      "epoch": 0.474034620505992,
      "grad_norm": 10.634424209594727,
      "kl": 0.873046875,
      "learning_rate": 8.82e-07,
      "loss": 0.121,
      "reward": 0.6069703996181488,
      "reward_std": 0.17810288816690445,
      "reward_std_all": 0.1835819035768509,
      "rewards-max/forensics_sinkhorn_reward": 0.9847692251205444,
      "rewards-max/forensics_strict_edge_reward": 0.8655564188957214,
      "rewards-max/forensics_tiou_reward": 0.9834348261356354,
      "rewards/forensics_sinkhorn_reward": 0.49458035826683044,
      "rewards/forensics_strict_edge_reward": 0.12278062850236893,
      "rewards/forensics_tiou_reward": 0.2611643522977829,
      "step": 178
    },
    {
      "completion_length": 18.578125,
      "epoch": 0.4766977363515313,
      "grad_norm": 9.233796119689941,
      "kl": 0.830078125,
      "learning_rate": 8.813333333333332e-07,
      "loss": 0.0922,
      "reward": 0.6434774100780487,
      "reward_std": 0.14450759440660477,
      "reward_std_all": 0.15135586261749268,
      "rewards-max/forensics_sinkhorn_reward": 0.9886386692523956,
      "rewards-max/forensics_strict_edge_reward": 0.9574736058712006,
      "rewards-max/forensics_tiou_reward": 0.998613029718399,
      "rewards/forensics_sinkhorn_reward": 0.5591737776994705,
      "rewards/forensics_strict_edge_reward": 0.17920802161097527,
      "rewards/forensics_tiou_reward": 0.30383580923080444,
      "step": 179
    },
    {
      "completion_length": 19.84375,
      "epoch": 0.4793608521970706,
      "grad_norm": 10.570542335510254,
      "kl": 0.861328125,
      "learning_rate": 8.806666666666667e-07,
      "loss": 0.0999,
      "reward": 0.6256847977638245,
      "reward_std": 0.14662301540374756,
      "reward_std_all": 0.1573890596628189,
      "rewards-max/forensics_sinkhorn_reward": 0.9653630554676056,
      "rewards-max/forensics_strict_edge_reward": 0.7197270095348358,
      "rewards-max/forensics_tiou_reward": 0.9465413093566895,
      "rewards/forensics_sinkhorn_reward": 0.47131767868995667,
      "rewards/forensics_strict_edge_reward": 0.11355791240930557,
      "rewards/forensics_tiou_reward": 0.23893168568611145,
      "step": 180
    },
    {
      "completion_length": 19.640625,
      "epoch": 0.48202396804260983,
      "grad_norm": 12.160025596618652,
      "kl": 0.685546875,
      "learning_rate": 8.799999999999999e-07,
      "loss": 0.3038,
      "reward": 0.6239978075027466,
      "reward_std": 0.16456590592861176,
      "reward_std_all": 0.16804582625627518,
      "rewards-max/forensics_sinkhorn_reward": 0.9871818125247955,
      "rewards-max/forensics_strict_edge_reward": 0.9357379674911499,
      "rewards-max/forensics_tiou_reward": 0.9928867518901825,
      "rewards/forensics_sinkhorn_reward": 0.5198829174041748,
      "rewards/forensics_strict_edge_reward": 0.2290375605225563,
      "rewards/forensics_tiou_reward": 0.30210956931114197,
      "step": 181
    },
    {
      "completion_length": 18.125,
      "epoch": 0.48468708388814913,
      "grad_norm": 12.559310913085938,
      "kl": 1.359375,
      "learning_rate": 8.793333333333333e-07,
      "loss": 0.088,
      "reward": 0.6387496292591095,
      "reward_std": 0.07632732763886452,
      "reward_std_all": 0.11621152609586716,
      "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": 0.5189328342676163,
      "rewards/forensics_strict_edge_reward": 0.20559890568256378,
      "rewards/forensics_tiou_reward": 0.24647167325019836,
      "step": 182
    },
    {
      "completion_length": 17.53125,
      "epoch": 0.4873501997336884,
      "grad_norm": 10.39967155456543,
      "kl": 0.796875,
      "learning_rate": 8.786666666666666e-07,
      "loss": 0.0729,
      "reward": 0.6531941890716553,
      "reward_std": 0.12030242756009102,
      "reward_std_all": 0.1367371529340744,
      "rewards-max/forensics_sinkhorn_reward": 0.9878359735012054,
      "rewards-max/forensics_strict_edge_reward": 0.927969753742218,
      "rewards-max/forensics_tiou_reward": 0.9954597353935242,
      "rewards/forensics_sinkhorn_reward": 0.5345275402069092,
      "rewards/forensics_strict_edge_reward": 0.20989970117807388,
      "rewards/forensics_tiou_reward": 0.28921838104724884,
      "step": 183
    },
    {
      "completion_length": 21.078125,
      "epoch": 0.49001331557922767,
      "grad_norm": 9.050863265991211,
      "kl": 1.0,
      "learning_rate": 8.78e-07,
      "loss": 0.2149,
      "reward": 0.6415248811244965,
      "reward_std": 0.165846049785614,
      "reward_std_all": 0.16178487241268158,
      "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": 0.4980335831642151,
      "rewards/forensics_strict_edge_reward": 0.22968459874391556,
      "rewards/forensics_tiou_reward": 0.35217659175395966,
      "step": 184
    },
    {
      "completion_length": 19.4375,
      "epoch": 0.49267643142476697,
      "grad_norm": 10.727019309997559,
      "kl": 0.583984375,
      "learning_rate": 8.773333333333332e-07,
      "loss": 0.1425,
      "reward": 0.6654395759105682,
      "reward_std": 0.12675615027546883,
      "reward_std_all": 0.13796084374189377,
      "rewards-max/forensics_sinkhorn_reward": 0.9868837893009186,
      "rewards-max/forensics_strict_edge_reward": 0.9256840348243713,
      "rewards-max/forensics_tiou_reward": 0.9916960895061493,
      "rewards/forensics_sinkhorn_reward": 0.5265544354915619,
      "rewards/forensics_strict_edge_reward": 0.2220413163304329,
      "rewards/forensics_tiou_reward": 0.32195834815502167,
      "step": 185
    },
    {
      "completion_length": 19.984375,
      "epoch": 0.49533954727030627,
      "grad_norm": 9.44768238067627,
      "kl": 0.8671875,
      "learning_rate": 8.766666666666667e-07,
      "loss": 0.2667,
      "reward": 0.5896729528903961,
      "reward_std": 0.1548006609082222,
      "reward_std_all": 0.18164435774087906,
      "rewards-max/forensics_sinkhorn_reward": 0.9825315475463867,
      "rewards-max/forensics_strict_edge_reward": 0.9075756669044495,
      "rewards-max/forensics_tiou_reward": 0.9748600721359253,
      "rewards/forensics_sinkhorn_reward": 0.48997753858566284,
      "rewards/forensics_strict_edge_reward": 0.21344192326068878,
      "rewards/forensics_tiou_reward": 0.283224418759346,
      "step": 186
    },
    {
      "completion_length": 17.671875,
      "epoch": 0.4980026631158455,
      "grad_norm": 9.084659576416016,
      "kl": 0.76171875,
      "learning_rate": 8.76e-07,
      "loss": 0.0523,
      "reward": 0.612575352191925,
      "reward_std": 0.16577491536736488,
      "reward_std_all": 0.17931636422872543,
      "rewards-max/forensics_sinkhorn_reward": 0.982068806886673,
      "rewards-max/forensics_strict_edge_reward": 0.8404377698898315,
      "rewards-max/forensics_tiou_reward": 0.9729256927967072,
      "rewards/forensics_sinkhorn_reward": 0.48658742010593414,
      "rewards/forensics_strict_edge_reward": 0.19233942031860352,
      "rewards/forensics_tiou_reward": 0.26415395736694336,
      "step": 187
    },
    {
      "completion_length": 16.125,
      "epoch": 0.5006657789613849,
      "grad_norm": 10.399298667907715,
      "kl": 0.841796875,
      "learning_rate": 8.753333333333332e-07,
      "loss": 0.1218,
      "reward": 0.677586704492569,
      "reward_std": 0.1263272501528263,
      "reward_std_all": 0.12781860679388046,
      "rewards-max/forensics_sinkhorn_reward": 0.9865729510784149,
      "rewards-max/forensics_strict_edge_reward": 0.946017861366272,
      "rewards-max/forensics_tiou_reward": 0.9904714822769165,
      "rewards/forensics_sinkhorn_reward": 0.5685397088527679,
      "rewards/forensics_strict_edge_reward": 0.25018787384033203,
      "rewards/forensics_tiou_reward": 0.3112432360649109,
      "step": 188
    },
    {
      "completion_length": 19.265625,
      "epoch": 0.5033288948069241,
      "grad_norm": 11.627152442932129,
      "kl": 1.18359375,
      "learning_rate": 8.746666666666667e-07,
      "loss": 0.2164,
      "reward": 0.6210099458694458,
      "reward_std": 0.14589587599039078,
      "reward_std_all": 0.15955137461423874,
      "rewards-max/forensics_sinkhorn_reward": 0.9742932021617889,
      "rewards-max/forensics_strict_edge_reward": 0.7355287075042725,
      "rewards-max/forensics_tiou_reward": 0.9585672616958618,
      "rewards/forensics_sinkhorn_reward": 0.48776106536388397,
      "rewards/forensics_strict_edge_reward": 0.1713550165295601,
      "rewards/forensics_tiou_reward": 0.2852849066257477,
      "step": 189
    },
    {
      "completion_length": 14.65625,
      "epoch": 0.5059920106524634,
      "grad_norm": 11.787837028503418,
      "kl": 0.5078125,
      "learning_rate": 8.739999999999999e-07,
      "loss": 0.1651,
      "reward": 0.6387689709663391,
      "reward_std": 0.14107439666986465,
      "reward_std_all": 0.15145844221115112,
      "rewards-max/forensics_sinkhorn_reward": 0.9876793026924133,
      "rewards-max/forensics_strict_edge_reward": 0.9677534997463226,
      "rewards-max/forensics_tiou_reward": 0.9948453903198242,
      "rewards/forensics_sinkhorn_reward": 0.5746563673019409,
      "rewards/forensics_strict_edge_reward": 0.24067625403404236,
      "rewards/forensics_tiou_reward": 0.3078151047229767,
      "step": 190
    },
    {
      "completion_length": 19.609375,
      "epoch": 0.5086551264980027,
      "grad_norm": 9.703954696655273,
      "kl": 0.806640625,
      "learning_rate": 8.733333333333333e-07,
      "loss": 0.1074,
      "reward": 0.5217445939779282,
      "reward_std": 0.14449632912874222,
      "reward_std_all": 0.19965167343616486,
      "rewards-max/forensics_sinkhorn_reward": 0.9402555823326111,
      "rewards-max/forensics_strict_edge_reward": 0.8389894068241119,
      "rewards-max/forensics_tiou_reward": 0.8914231956005096,
      "rewards/forensics_sinkhorn_reward": 0.3805054873228073,
      "rewards/forensics_strict_edge_reward": 0.14821873605251312,
      "rewards/forensics_tiou_reward": 0.1789923906326294,
      "step": 191
    },
    {
      "completion_length": 18.78125,
      "epoch": 0.511318242343542,
      "grad_norm": 9.277739524841309,
      "kl": 1.41015625,
      "learning_rate": 8.726666666666666e-07,
      "loss": 0.1635,
      "reward": 0.645701676607132,
      "reward_std": 0.13694143295288086,
      "reward_std_all": 0.14112430810928345,
      "rewards-max/forensics_sinkhorn_reward": 0.9882653057575226,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9971435070037842,
      "rewards/forensics_sinkhorn_reward": 0.5400737524032593,
      "rewards/forensics_strict_edge_reward": 0.22632156312465668,
      "rewards/forensics_tiou_reward": 0.3379313200712204,
      "step": 192
    },
    {
      "completion_length": 16.453125,
      "epoch": 0.5139813581890812,
      "grad_norm": 8.68982982635498,
      "kl": 0.51953125,
      "learning_rate": 8.72e-07,
      "loss": 0.118,
      "reward": 0.6644903719425201,
      "reward_std": 0.15854796022176743,
      "reward_std_all": 0.1532665193080902,
      "rewards-max/forensics_sinkhorn_reward": 0.986740916967392,
      "rewards-max/forensics_strict_edge_reward": 0.8771188855171204,
      "rewards-max/forensics_tiou_reward": 0.9911527633666992,
      "rewards/forensics_sinkhorn_reward": 0.5703174769878387,
      "rewards/forensics_strict_edge_reward": 0.17437417060136795,
      "rewards/forensics_tiou_reward": 0.3274838626384735,
      "step": 193
    },
    {
      "completion_length": 16.890625,
      "epoch": 0.5166444740346205,
      "grad_norm": 9.779521942138672,
      "kl": 0.5302734375,
      "learning_rate": 8.713333333333332e-07,
      "loss": 0.089,
      "reward": 0.5990574955940247,
      "reward_std": 0.08307796716690063,
      "reward_std_all": 0.13935154676437378,
      "rewards-max/forensics_sinkhorn_reward": 0.9869023859500885,
      "rewards-max/forensics_strict_edge_reward": 0.9464850425720215,
      "rewards-max/forensics_tiou_reward": 0.9917698204517365,
      "rewards/forensics_sinkhorn_reward": 0.4644373804330826,
      "rewards/forensics_strict_edge_reward": 0.16950875148177147,
      "rewards/forensics_tiou_reward": 0.2329237461090088,
      "step": 194
    },
    {
      "completion_length": 16.53125,
      "epoch": 0.5193075898801598,
      "grad_norm": 10.948802947998047,
      "kl": 0.4423828125,
      "learning_rate": 8.706666666666667e-07,
      "loss": 0.0362,
      "reward": 0.6183261573314667,
      "reward_std": 0.1306052953004837,
      "reward_std_all": 0.1494721919298172,
      "rewards-max/forensics_sinkhorn_reward": 0.9464926719665527,
      "rewards-max/forensics_strict_edge_reward": 0.7216397970914841,
      "rewards-max/forensics_tiou_reward": 0.9102432131767273,
      "rewards/forensics_sinkhorn_reward": 0.47344033420085907,
      "rewards/forensics_strict_edge_reward": 0.08535582944750786,
      "rewards/forensics_tiou_reward": 0.13890229538083076,
      "step": 195
    },
    {
      "completion_length": 16.390625,
      "epoch": 0.521970705725699,
      "grad_norm": 9.239683151245117,
      "kl": 0.865234375,
      "learning_rate": 8.699999999999999e-07,
      "loss": -0.0083,
      "reward": 0.6267433166503906,
      "reward_std": 0.07659706100821495,
      "reward_std_all": 0.12379956617951393,
      "rewards-max/forensics_sinkhorn_reward": 0.9392774403095245,
      "rewards-max/forensics_strict_edge_reward": 0.8025559782981873,
      "rewards-max/forensics_tiou_reward": 0.8885965049266815,
      "rewards/forensics_sinkhorn_reward": 0.48231393098831177,
      "rewards/forensics_strict_edge_reward": 0.1302010342478752,
      "rewards/forensics_tiou_reward": 0.16877015680074692,
      "step": 196
    },
    {
      "completion_length": 16.703125,
      "epoch": 0.5246338215712384,
      "grad_norm": 11.338549613952637,
      "kl": 1.064453125,
      "learning_rate": 8.693333333333333e-07,
      "loss": 0.0842,
      "reward": 0.6038511693477631,
      "reward_std": 0.08515523001551628,
      "reward_std_all": 0.13925977796316147,
      "rewards-max/forensics_sinkhorn_reward": 0.954600989818573,
      "rewards-max/forensics_strict_edge_reward": 0.6194965690374374,
      "rewards-max/forensics_tiou_reward": 0.9246389865875244,
      "rewards/forensics_sinkhorn_reward": 0.4724423289299011,
      "rewards/forensics_strict_edge_reward": 0.10376756638288498,
      "rewards/forensics_tiou_reward": 0.18323646485805511,
      "step": 197
    },
    {
      "completion_length": 19.875,
      "epoch": 0.5272969374167776,
      "grad_norm": 9.86098861694336,
      "kl": 0.62109375,
      "learning_rate": 8.686666666666666e-07,
      "loss": 0.1088,
      "reward": 0.6370527148246765,
      "reward_std": 0.10761985555291176,
      "reward_std_all": 0.13351907581090927,
      "rewards-max/forensics_sinkhorn_reward": 0.9858402907848358,
      "rewards-max/forensics_strict_edge_reward": 0.9252271056175232,
      "rewards-max/forensics_tiou_reward": 0.9876229763031006,
      "rewards/forensics_sinkhorn_reward": 0.519268274307251,
      "rewards/forensics_strict_edge_reward": 0.18585048615932465,
      "rewards/forensics_tiou_reward": 0.2545647770166397,
      "step": 198
    },
    {
      "completion_length": 21.625,
      "epoch": 0.5299600532623169,
      "grad_norm": 7.769320011138916,
      "kl": 0.986328125,
      "learning_rate": 8.68e-07,
      "loss": 0.1307,
      "reward": 0.6551156938076019,
      "reward_std": 0.14631858468055725,
      "reward_std_all": 0.14722658693790436,
      "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": 0.5696271061897278,
      "rewards/forensics_strict_edge_reward": 0.25257743149995804,
      "rewards/forensics_tiou_reward": 0.41168612241744995,
      "step": 199
    },
    {
      "completion_length": 19.953125,
      "epoch": 0.5326231691078562,
      "grad_norm": 12.094202041625977,
      "kl": 1.25390625,
      "learning_rate": 8.673333333333332e-07,
      "loss": 0.0402,
      "reward": 0.6098348200321198,
      "reward_std": 0.15363675728440285,
      "reward_std_all": 0.17237119376659393,
      "rewards-max/forensics_sinkhorn_reward": 0.9665297269821167,
      "rewards-max/forensics_strict_edge_reward": 0.7357586026191711,
      "rewards-max/forensics_tiou_reward": 0.9154556393623352,
      "rewards/forensics_sinkhorn_reward": 0.5138148367404938,
      "rewards/forensics_strict_edge_reward": 0.1838206723332405,
      "rewards/forensics_tiou_reward": 0.30694641917943954,
      "step": 200
    },
    {
      "completion_length": 18.046875,
      "epoch": 0.5352862849533955,
      "grad_norm": 9.566694259643555,
      "kl": 0.689453125,
      "learning_rate": 8.666666666666667e-07,
      "loss": 0.1505,
      "reward": 0.6366775631904602,
      "reward_std": 0.1532418131828308,
      "reward_std_all": 0.1568957008421421,
      "rewards-max/forensics_sinkhorn_reward": 0.9621348083019257,
      "rewards-max/forensics_strict_edge_reward": 0.7573865950107574,
      "rewards-max/forensics_tiou_reward": 0.9371521770954132,
      "rewards/forensics_sinkhorn_reward": 0.556957334280014,
      "rewards/forensics_strict_edge_reward": 0.22152663581073284,
      "rewards/forensics_tiou_reward": 0.33460962772369385,
      "step": 201
    },
    {
      "completion_length": 15.84375,
      "epoch": 0.5379494007989347,
      "grad_norm": 12.404425621032715,
      "kl": 0.63671875,
      "learning_rate": 8.659999999999999e-07,
      "loss": -0.0224,
      "reward": 0.6797004044055939,
      "reward_std": 0.07832690328359604,
      "reward_std_all": 0.10484405234456062,
      "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": 0.6044240295886993,
      "rewards/forensics_strict_edge_reward": 0.19663960486650467,
      "rewards/forensics_tiou_reward": 0.28410743176937103,
      "step": 202
    },
    {
      "completion_length": 18.15625,
      "epoch": 0.5406125166444741,
      "grad_norm": 8.530579566955566,
      "kl": 0.72265625,
      "learning_rate": 8.653333333333333e-07,
      "loss": 0.1193,
      "reward": 0.6540484726428986,
      "reward_std": 0.12178054824471474,
      "reward_std_all": 0.13421660661697388,
      "rewards-max/forensics_sinkhorn_reward": 0.988030344247818,
      "rewards-max/forensics_strict_edge_reward": 0.9364311099052429,
      "rewards-max/forensics_tiou_reward": 0.996220737695694,
      "rewards/forensics_sinkhorn_reward": 0.528099998831749,
      "rewards/forensics_strict_edge_reward": 0.19160325825214386,
      "rewards/forensics_tiou_reward": 0.28020551800727844,
      "step": 203
    },
    {
      "completion_length": 17.359375,
      "epoch": 0.5432756324900133,
      "grad_norm": 8.99083423614502,
      "kl": 1.1796875,
      "learning_rate": 8.646666666666667e-07,
      "loss": 0.2601,
      "reward": 0.6485846638679504,
      "reward_std": 0.11050675809383392,
      "reward_std_all": 0.1324235051870346,
      "rewards-max/forensics_sinkhorn_reward": 0.9869354367256165,
      "rewards-max/forensics_strict_edge_reward": 0.8957232534885406,
      "rewards-max/forensics_tiou_reward": 0.9919202923774719,
      "rewards/forensics_sinkhorn_reward": 0.5356681048870087,
      "rewards/forensics_strict_edge_reward": 0.20340345799922943,
      "rewards/forensics_tiou_reward": 0.2725738435983658,
      "step": 204
    },
    {
      "completion_length": 19.828125,
      "epoch": 0.5459387483355526,
      "grad_norm": 9.556720733642578,
      "kl": 1.099609375,
      "learning_rate": 8.639999999999999e-07,
      "loss": 0.1334,
      "reward": 0.5988282859325409,
      "reward_std": 0.12088693305850029,
      "reward_std_all": 0.15759079158306122,
      "rewards-max/forensics_sinkhorn_reward": 0.9871483445167542,
      "rewards-max/forensics_strict_edge_reward": 0.9364311099052429,
      "rewards-max/forensics_tiou_reward": 0.992755800485611,
      "rewards/forensics_sinkhorn_reward": 0.5003054440021515,
      "rewards/forensics_strict_edge_reward": 0.12966763600707054,
      "rewards/forensics_tiou_reward": 0.2319471687078476,
      "step": 205
    },
    {
      "completion_length": 20.09375,
      "epoch": 0.5486018641810919,
      "grad_norm": 24.103010177612305,
      "kl": 2.041015625,
      "learning_rate": 8.633333333333333e-07,
      "loss": 0.254,
      "reward": 0.6317821145057678,
      "reward_std": 0.17155065387487411,
      "reward_std_all": 0.16911214590072632,
      "rewards-max/forensics_sinkhorn_reward": 0.9873689115047455,
      "rewards-max/forensics_strict_edge_reward": 0.9567649364471436,
      "rewards-max/forensics_tiou_reward": 0.993622899055481,
      "rewards/forensics_sinkhorn_reward": 0.4854564964771271,
      "rewards/forensics_strict_edge_reward": 0.23527168482542038,
      "rewards/forensics_tiou_reward": 0.30230407416820526,
      "step": 206
    },
    {
      "completion_length": 17.296875,
      "epoch": 0.5512649800266312,
      "grad_norm": 14.323897361755371,
      "kl": 1.55859375,
      "learning_rate": 8.626666666666666e-07,
      "loss": 0.0568,
      "reward": 0.6662566363811493,
      "reward_std": 0.09408537298440933,
      "reward_std_all": 0.11635085195302963,
      "rewards-max/forensics_sinkhorn_reward": 0.987255871295929,
      "rewards-max/forensics_strict_edge_reward": 0.9357379674911499,
      "rewards-max/forensics_tiou_reward": 0.9931744337081909,
      "rewards/forensics_sinkhorn_reward": 0.5625303387641907,
      "rewards/forensics_strict_edge_reward": 0.2608971521258354,
      "rewards/forensics_tiou_reward": 0.35368451476097107,
      "step": 207
    },
    {
      "completion_length": 19.4375,
      "epoch": 0.5539280958721704,
      "grad_norm": 8.874499320983887,
      "kl": 0.765625,
      "learning_rate": 8.62e-07,
      "loss": 0.34,
      "reward": 0.6560455560684204,
      "reward_std": 0.13031183555722237,
      "reward_std_all": 0.14199145883321762,
      "rewards-max/forensics_sinkhorn_reward": 0.9880021512508392,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9961079359054565,
      "rewards/forensics_sinkhorn_reward": 0.5350740998983383,
      "rewards/forensics_strict_edge_reward": 0.1960601657629013,
      "rewards/forensics_tiou_reward": 0.31763118505477905,
      "step": 208
    },
    {
      "completion_length": 20.640625,
      "epoch": 0.5565912117177098,
      "grad_norm": 9.490084648132324,
      "kl": 0.671875,
      "learning_rate": 8.613333333333332e-07,
      "loss": 0.1436,
      "reward": 0.6361706554889679,
      "reward_std": 0.16406850516796112,
      "reward_std_all": 0.16370709985494614,
      "rewards-max/forensics_sinkhorn_reward": 0.9880532920360565,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9963157176971436,
      "rewards/forensics_sinkhorn_reward": 0.5389240086078644,
      "rewards/forensics_strict_edge_reward": 0.2699917405843735,
      "rewards/forensics_tiou_reward": 0.36942258477211,
      "step": 209
    },
    {
      "completion_length": 21.03125,
      "epoch": 0.559254327563249,
      "grad_norm": 10.878541946411133,
      "kl": 1.173828125,
      "learning_rate": 8.606666666666667e-07,
      "loss": 0.3031,
      "reward": 0.6090760827064514,
      "reward_std": 0.11182515695691109,
      "reward_std_all": 0.14827969670295715,
      "rewards-max/forensics_sinkhorn_reward": 0.9871093332767487,
      "rewards-max/forensics_strict_edge_reward": 0.9364311099052429,
      "rewards-max/forensics_tiou_reward": 0.9926067888736725,
      "rewards/forensics_sinkhorn_reward": 0.49890854954719543,
      "rewards/forensics_strict_edge_reward": 0.25181347131729126,
      "rewards/forensics_tiou_reward": 0.30495692789554596,
      "step": 210
    },
    {
      "completion_length": 22.40625,
      "epoch": 0.5619174434087882,
      "grad_norm": 10.270121574401855,
      "kl": 1.125,
      "learning_rate": 8.599999999999999e-07,
      "loss": 0.1385,
      "reward": 0.6701552867889404,
      "reward_std": 0.13301103934645653,
      "reward_std_all": 0.136239156126976,
      "rewards-max/forensics_sinkhorn_reward": 0.9867915511131287,
      "rewards-max/forensics_strict_edge_reward": 0.9062341153621674,
      "rewards-max/forensics_tiou_reward": 0.991333544254303,
      "rewards/forensics_sinkhorn_reward": 0.5542615354061127,
      "rewards/forensics_strict_edge_reward": 0.26607614755630493,
      "rewards/forensics_tiou_reward": 0.3821098804473877,
      "step": 211
    },
    {
      "completion_length": 16.15625,
      "epoch": 0.5645805592543276,
      "grad_norm": 14.893415451049805,
      "kl": 1.166015625,
      "learning_rate": 8.593333333333333e-07,
      "loss": 0.082,
      "reward": 0.6627331674098969,
      "reward_std": 0.10663998499512672,
      "reward_std_all": 0.12179911509156227,
      "rewards-max/forensics_sinkhorn_reward": 0.9846440553665161,
      "rewards-max/forensics_strict_edge_reward": 0.8876297473907471,
      "rewards-max/forensics_tiou_reward": 0.9829782247543335,
      "rewards/forensics_sinkhorn_reward": 0.6051175892353058,
      "rewards/forensics_strict_edge_reward": 0.22559484839439392,
      "rewards/forensics_tiou_reward": 0.3215268552303314,
      "step": 212
    },
    {
      "completion_length": 17.546875,
      "epoch": 0.5672436750998668,
      "grad_norm": 12.490218162536621,
      "kl": 1.62890625,
      "learning_rate": 8.586666666666666e-07,
      "loss": 0.2032,
      "reward": 0.6590126156806946,
      "reward_std": 0.14292436093091965,
      "reward_std_all": 0.14287681132555008,
      "rewards-max/forensics_sinkhorn_reward": 0.9879377782344818,
      "rewards-max/forensics_strict_edge_reward": 0.8582656681537628,
      "rewards-max/forensics_tiou_reward": 0.9958506226539612,
      "rewards/forensics_sinkhorn_reward": 0.5318712592124939,
      "rewards/forensics_strict_edge_reward": 0.21719887852668762,
      "rewards/forensics_tiou_reward": 0.31614363193511963,
      "step": 213
    },
    {
      "completion_length": 19.90625,
      "epoch": 0.5699067909454061,
      "grad_norm": 12.677924156188965,
      "kl": 1.59765625,
      "learning_rate": 8.58e-07,
      "loss": 0.276,
      "reward": 0.6608815789222717,
      "reward_std": 0.1198483407497406,
      "reward_std_all": 0.1276378519833088,
      "rewards-max/forensics_sinkhorn_reward": 0.8374972939491272,
      "rewards-max/forensics_strict_edge_reward": 0.637529581785202,
      "rewards-max/forensics_tiou_reward": 0.7428064942359924,
      "rewards/forensics_sinkhorn_reward": 0.4790011942386627,
      "rewards/forensics_strict_edge_reward": 0.1712905615568161,
      "rewards/forensics_tiou_reward": 0.25323374569416046,
      "step": 214
    },
    {
      "completion_length": 20.859375,
      "epoch": 0.5725699067909454,
      "grad_norm": 14.79261302947998,
      "kl": 1.96484375,
      "learning_rate": 8.573333333333332e-07,
      "loss": 0.0925,
      "reward": 0.6095295250415802,
      "reward_std": 0.1513115055859089,
      "reward_std_all": 0.16808509081602097,
      "rewards-max/forensics_sinkhorn_reward": 0.9831533133983612,
      "rewards-max/forensics_strict_edge_reward": 0.6810412704944611,
      "rewards-max/forensics_tiou_reward": 0.9771145284175873,
      "rewards/forensics_sinkhorn_reward": 0.45882053673267365,
      "rewards/forensics_strict_edge_reward": 0.11721094697713852,
      "rewards/forensics_tiou_reward": 0.19774480164051056,
      "step": 215
    },
    {
      "completion_length": 20.75,
      "epoch": 0.5752330226364847,
      "grad_norm": 11.703033447265625,
      "kl": 0.7734375,
      "learning_rate": 8.566666666666667e-07,
      "loss": 0.1398,
      "reward": 0.5681907832622528,
      "reward_std": 0.13322781398892403,
      "reward_std_all": 0.17489925771951675,
      "rewards-max/forensics_sinkhorn_reward": 0.98715740442276,
      "rewards-max/forensics_strict_edge_reward": 0.9567649364471436,
      "rewards-max/forensics_tiou_reward": 0.9927772879600525,
      "rewards/forensics_sinkhorn_reward": 0.4633030295372009,
      "rewards/forensics_strict_edge_reward": 0.1546492502093315,
      "rewards/forensics_tiou_reward": 0.256673626601696,
      "step": 216
    },
    {
      "completion_length": 22.734375,
      "epoch": 0.5778961384820239,
      "grad_norm": 17.151147842407227,
      "kl": 2.5234375,
      "learning_rate": 8.559999999999999e-07,
      "loss": 0.4067,
      "reward": 0.6211174726486206,
      "reward_std": 0.14966114610433578,
      "reward_std_all": 0.16256655007600784,
      "rewards-max/forensics_sinkhorn_reward": 0.9400677680969238,
      "rewards-max/forensics_strict_edge_reward": 0.8326341509819031,
      "rewards-max/forensics_tiou_reward": 0.8908194005489349,
      "rewards/forensics_sinkhorn_reward": 0.4384097158908844,
      "rewards/forensics_strict_edge_reward": 0.18045508116483688,
      "rewards/forensics_tiou_reward": 0.23532556742429733,
      "step": 217
    },
    {
      "completion_length": 19.9375,
      "epoch": 0.5805592543275633,
      "grad_norm": 11.177781105041504,
      "kl": 1.013671875,
      "learning_rate": 8.553333333333333e-07,
      "loss": 0.2881,
      "reward": 0.5767864882946014,
      "reward_std": 0.18038802593946457,
      "reward_std_all": 0.1937507465481758,
      "rewards-max/forensics_sinkhorn_reward": 0.9102048575878143,
      "rewards-max/forensics_strict_edge_reward": 0.7401526570320129,
      "rewards-max/forensics_tiou_reward": 0.8279662430286407,
      "rewards/forensics_sinkhorn_reward": 0.44018134474754333,
      "rewards/forensics_strict_edge_reward": 0.15489569678902626,
      "rewards/forensics_tiou_reward": 0.2249480038881302,
      "step": 218
    },
    {
      "completion_length": 17.453125,
      "epoch": 0.5832223701731025,
      "grad_norm": 12.268922805786133,
      "kl": 0.783203125,
      "learning_rate": 8.546666666666666e-07,
      "loss": 0.0001,
      "reward": 0.7070095539093018,
      "reward_std": 0.12795548886060715,
      "reward_std_all": 0.11795772984623909,
      "rewards-max/forensics_sinkhorn_reward": 0.988641083240509,
      "rewards-max/forensics_strict_edge_reward": 0.9782643616199493,
      "rewards-max/forensics_tiou_reward": 0.9986235499382019,
      "rewards/forensics_sinkhorn_reward": 0.5907281637191772,
      "rewards/forensics_strict_edge_reward": 0.2797130197286606,
      "rewards/forensics_tiou_reward": 0.3839048445224762,
      "step": 219
    },
    {
      "completion_length": 19.890625,
      "epoch": 0.5858854860186418,
      "grad_norm": 9.949393272399902,
      "kl": 1.37890625,
      "learning_rate": 8.539999999999999e-07,
      "loss": 0.1184,
      "reward": 0.6164632141590118,
      "reward_std": 0.10843417793512344,
      "reward_std_all": 0.14546189457178116,
      "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": 0.528471976518631,
      "rewards/forensics_strict_edge_reward": 0.2615787088871002,
      "rewards/forensics_tiou_reward": 0.33009539544582367,
      "step": 220
    },
    {
      "completion_length": 18.734375,
      "epoch": 0.5885486018641811,
      "grad_norm": 10.168049812316895,
      "kl": 0.921875,
      "learning_rate": 8.533333333333334e-07,
      "loss": 0.0678,
      "reward": 0.6141278147697449,
      "reward_std": 0.13737154006958008,
      "reward_std_all": 0.15713176876306534,
      "rewards-max/forensics_sinkhorn_reward": 0.987847775220871,
      "rewards-max/forensics_strict_edge_reward": 0.8593232333660126,
      "rewards-max/forensics_tiou_reward": 0.9954955577850342,
      "rewards/forensics_sinkhorn_reward": 0.46162495017051697,
      "rewards/forensics_strict_edge_reward": 0.12528090924024582,
      "rewards/forensics_tiou_reward": 0.20101209729909897,
      "step": 221
    },
    {
      "completion_length": 14.421875,
      "epoch": 0.5912117177097204,
      "grad_norm": 10.200089454650879,
      "kl": 0.49609375,
      "learning_rate": 8.526666666666666e-07,
      "loss": 0.1954,
      "reward": 0.7079249322414398,
      "reward_std": 0.07375225983560085,
      "reward_std_all": 0.08614818193018436,
      "rewards-max/forensics_sinkhorn_reward": 0.9886585474014282,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9986919462680817,
      "rewards/forensics_sinkhorn_reward": 0.6827018857002258,
      "rewards/forensics_strict_edge_reward": 0.30834734439849854,
      "rewards/forensics_tiou_reward": 0.4294387847185135,
      "step": 222
    },
    {
      "completion_length": 17.46875,
      "epoch": 0.5938748335552596,
      "grad_norm": 11.10137939453125,
      "kl": 0.396484375,
      "learning_rate": 8.52e-07,
      "loss": 0.0037,
      "reward": 0.63167405128479,
      "reward_std": 0.1428026482462883,
      "reward_std_all": 0.16293345391750336,
      "rewards-max/forensics_sinkhorn_reward": 0.9877058565616608,
      "rewards-max/forensics_strict_edge_reward": 0.9567649364471436,
      "rewards-max/forensics_tiou_reward": 0.9949560165405273,
      "rewards/forensics_sinkhorn_reward": 0.5159839242696762,
      "rewards/forensics_strict_edge_reward": 0.11808008328080177,
      "rewards/forensics_tiou_reward": 0.17985066026449203,
      "step": 223
    },
    {
      "completion_length": 18.015625,
      "epoch": 0.596537949400799,
      "grad_norm": 10.245888710021973,
      "kl": 0.90234375,
      "learning_rate": 8.513333333333333e-07,
      "loss": 0.0382,
      "reward": 0.6742427945137024,
      "reward_std": 0.12132086232304573,
      "reward_std_all": 0.1307198405265808,
      "rewards-max/forensics_sinkhorn_reward": 0.9851446747779846,
      "rewards-max/forensics_strict_edge_reward": 0.9890114367008209,
      "rewards-max/forensics_tiou_reward": 0.9918771088123322,
      "rewards/forensics_sinkhorn_reward": 0.5766114890575409,
      "rewards/forensics_strict_edge_reward": 0.25942444801330566,
      "rewards/forensics_tiou_reward": 0.3280744254589081,
      "step": 224
    },
    {
      "completion_length": 16.59375,
      "epoch": 0.5992010652463382,
      "grad_norm": 8.345564842224121,
      "kl": 1.16796875,
      "learning_rate": 8.506666666666667e-07,
      "loss": 0.1347,
      "reward": 0.6371663808822632,
      "reward_std": 0.1366998478770256,
      "reward_std_all": 0.14960891753435135,
      "rewards-max/forensics_sinkhorn_reward": 0.9879142045974731,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9957631826400757,
      "rewards/forensics_sinkhorn_reward": 0.571506917476654,
      "rewards/forensics_strict_edge_reward": 0.2439384013414383,
      "rewards/forensics_tiou_reward": 0.30939795076847076,
      "step": 225
    },
    {
      "completion_length": 16.4375,
      "epoch": 0.6018641810918774,
      "grad_norm": 10.523458480834961,
      "kl": 1.27734375,
      "learning_rate": 8.499999999999999e-07,
      "loss": -0.0153,
      "reward": 0.6958262026309967,
      "reward_std": 0.07920945808291435,
      "reward_std_all": 0.09307018667459488,
      "rewards-max/forensics_sinkhorn_reward": 0.9874260127544403,
      "rewards-max/forensics_strict_edge_reward": 0.9565287232398987,
      "rewards-max/forensics_tiou_reward": 0.9938330054283142,
      "rewards/forensics_sinkhorn_reward": 0.5969948172569275,
      "rewards/forensics_strict_edge_reward": 0.2666887640953064,
      "rewards/forensics_tiou_reward": 0.31883250176906586,
      "step": 226
    },
    {
      "completion_length": 19.1875,
      "epoch": 0.6045272969374168,
      "grad_norm": 13.101468086242676,
      "kl": 1.064453125,
      "learning_rate": 8.493333333333334e-07,
      "loss": 0.2728,
      "reward": 0.6573914289474487,
      "reward_std": 0.12887034565210342,
      "reward_std_all": 0.1401432268321514,
      "rewards-max/forensics_sinkhorn_reward": 0.977112740278244,
      "rewards-max/forensics_strict_edge_reward": 0.8481724560260773,
      "rewards-max/forensics_tiou_reward": 0.9726738929748535,
      "rewards/forensics_sinkhorn_reward": 0.5582058727741241,
      "rewards/forensics_strict_edge_reward": 0.18641414865851402,
      "rewards/forensics_tiou_reward": 0.3306434154510498,
      "step": 227
    },
    {
      "completion_length": 18.203125,
      "epoch": 0.607190412782956,
      "grad_norm": 14.57637882232666,
      "kl": 1.92578125,
      "learning_rate": 8.486666666666666e-07,
      "loss": 0.1597,
      "reward": 0.6334865987300873,
      "reward_std": 0.10702790692448616,
      "reward_std_all": 0.1333995722234249,
      "rewards-max/forensics_sinkhorn_reward": 0.9875025451183319,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9941399693489075,
      "rewards/forensics_sinkhorn_reward": 0.4993070214986801,
      "rewards/forensics_strict_edge_reward": 0.18201889842748642,
      "rewards/forensics_tiou_reward": 0.22235283255577087,
      "step": 228
    },
    {
      "completion_length": 18.0,
      "epoch": 0.6098535286284953,
      "grad_norm": 12.75692367553711,
      "kl": 1.236328125,
      "learning_rate": 8.48e-07,
      "loss": 0.1831,
      "reward": 0.6518921256065369,
      "reward_std": 0.1749451830983162,
      "reward_std_all": 0.16931172460317612,
      "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": 0.574438214302063,
      "rewards/forensics_strict_edge_reward": 0.306166909635067,
      "rewards/forensics_tiou_reward": 0.3674478679895401,
      "step": 229
    },
    {
      "completion_length": 17.03125,
      "epoch": 0.6125166444740346,
      "grad_norm": 24.19082260131836,
      "kl": 2.8828125,
      "learning_rate": 8.473333333333333e-07,
      "loss": 0.0993,
      "reward": 0.5928244292736053,
      "reward_std": 0.1849762126803398,
      "reward_std_all": 0.19639745354652405,
      "rewards-max/forensics_sinkhorn_reward": 0.9124291241168976,
      "rewards-max/forensics_strict_edge_reward": 0.6451423764228821,
      "rewards-max/forensics_tiou_reward": 0.8270975053310394,
      "rewards/forensics_sinkhorn_reward": 0.449556902050972,
      "rewards/forensics_strict_edge_reward": 0.1011311449110508,
      "rewards/forensics_tiou_reward": 0.1894621178507805,
      "step": 230
    },
    {
      "completion_length": 17.140625,
      "epoch": 0.6151797603195739,
      "grad_norm": 11.886222839355469,
      "kl": 2.53125,
      "learning_rate": 8.466666666666667e-07,
      "loss": 0.4056,
      "reward": 0.6243040859699249,
      "reward_std": 0.13884440064430237,
      "reward_std_all": 0.15626217424869537,
      "rewards-max/forensics_sinkhorn_reward": 0.9866087436676025,
      "rewards-max/forensics_strict_edge_reward": 0.8983768224716187,
      "rewards-max/forensics_tiou_reward": 0.9906151592731476,
      "rewards/forensics_sinkhorn_reward": 0.5104982852935791,
      "rewards/forensics_strict_edge_reward": 0.17676566541194916,
      "rewards/forensics_tiou_reward": 0.2631656229496002,
      "step": 231
    },
    {
      "completion_length": 19.296875,
      "epoch": 0.6178428761651131,
      "grad_norm": 9.795589447021484,
      "kl": 1.59375,
      "learning_rate": 8.459999999999999e-07,
      "loss": 0.1902,
      "reward": 0.6084122061729431,
      "reward_std": 0.1454295814037323,
      "reward_std_all": 0.16337239742279053,
      "rewards-max/forensics_sinkhorn_reward": 0.9868934154510498,
      "rewards-max/forensics_strict_edge_reward": 0.947419673204422,
      "rewards-max/forensics_tiou_reward": 0.9917355477809906,
      "rewards/forensics_sinkhorn_reward": 0.4477853775024414,
      "rewards/forensics_strict_edge_reward": 0.18935545533895493,
      "rewards/forensics_tiou_reward": 0.24126877635717392,
      "step": 232
    },
    {
      "completion_length": 18.859375,
      "epoch": 0.6205059920106525,
      "grad_norm": 11.427024841308594,
      "kl": 1.1796875,
      "learning_rate": 8.453333333333334e-07,
      "loss": 0.148,
      "reward": 0.6623857915401459,
      "reward_std": 0.16425395011901855,
      "reward_std_all": 0.159023255109787,
      "rewards-max/forensics_sinkhorn_reward": 0.9860345423221588,
      "rewards-max/forensics_strict_edge_reward": 0.8578422963619232,
      "rewards-max/forensics_tiou_reward": 0.9883737862110138,
      "rewards/forensics_sinkhorn_reward": 0.513625979423523,
      "rewards/forensics_strict_edge_reward": 0.18052127957344055,
      "rewards/forensics_tiou_reward": 0.30980919301509857,
      "step": 233
    },
    {
      "completion_length": 18.921875,
      "epoch": 0.6231691078561917,
      "grad_norm": 11.538118362426758,
      "kl": 1.359375,
      "learning_rate": 8.446666666666666e-07,
      "loss": 0.1072,
      "reward": 0.670103520154953,
      "reward_std": 0.14735978841781616,
      "reward_std_all": 0.14542420208454132,
      "rewards-max/forensics_sinkhorn_reward": 0.9882099032402039,
      "rewards-max/forensics_strict_edge_reward": 0.9782643616199493,
      "rewards-max/forensics_tiou_reward": 0.9969230890274048,
      "rewards/forensics_sinkhorn_reward": 0.5754360854625702,
      "rewards/forensics_strict_edge_reward": 0.2633967846632004,
      "rewards/forensics_tiou_reward": 0.37140123546123505,
      "step": 234
    },
    {
      "completion_length": 21.640625,
      "epoch": 0.625832223701731,
      "grad_norm": 18.613130569458008,
      "kl": 1.66796875,
      "learning_rate": 8.439999999999999e-07,
      "loss": 0.195,
      "reward": 0.5467893481254578,
      "reward_std": 0.13115286082029343,
      "reward_std_all": 0.1813468188047409,
      "rewards-max/forensics_sinkhorn_reward": 0.9208278357982635,
      "rewards-max/forensics_strict_edge_reward": 0.7915336191654205,
      "rewards-max/forensics_tiou_reward": 0.855863630771637,
      "rewards/forensics_sinkhorn_reward": 0.3743527978658676,
      "rewards/forensics_strict_edge_reward": 0.084262914955616,
      "rewards/forensics_tiou_reward": 0.165571890771389,
      "step": 235
    },
    {
      "completion_length": 18.03125,
      "epoch": 0.6284953395472703,
      "grad_norm": 8.949464797973633,
      "kl": 1.3125,
      "learning_rate": 8.433333333333333e-07,
      "loss": 0.1223,
      "reward": 0.5911184549331665,
      "reward_std": 0.12568054348230362,
      "reward_std_all": 0.16176610440015793,
      "rewards-max/forensics_sinkhorn_reward": 0.9396195709705353,
      "rewards-max/forensics_strict_edge_reward": 0.7333561778068542,
      "rewards-max/forensics_tiou_reward": 0.8858632743358612,
      "rewards/forensics_sinkhorn_reward": 0.4640497863292694,
      "rewards/forensics_strict_edge_reward": 0.12022818997502327,
      "rewards/forensics_tiou_reward": 0.18955454230308533,
      "step": 236
    },
    {
      "completion_length": 21.671875,
      "epoch": 0.6311584553928096,
      "grad_norm": 10.913715362548828,
      "kl": 1.013671875,
      "learning_rate": 8.426666666666666e-07,
      "loss": 0.1754,
      "reward": 0.6402294337749481,
      "reward_std": 0.17390140891075134,
      "reward_std_all": 0.17294715344905853,
      "rewards-max/forensics_sinkhorn_reward": 0.9728107750415802,
      "rewards-max/forensics_strict_edge_reward": 0.7931119501590729,
      "rewards-max/forensics_tiou_reward": 0.9373224079608917,
      "rewards/forensics_sinkhorn_reward": 0.49147269129753113,
      "rewards/forensics_strict_edge_reward": 0.259955033659935,
      "rewards/forensics_tiou_reward": 0.3540656715631485,
      "step": 237
    },
    {
      "completion_length": 20.203125,
      "epoch": 0.6338215712383488,
      "grad_norm": 11.994964599609375,
      "kl": 1.6484375,
      "learning_rate": 8.419999999999999e-07,
      "loss": 0.2364,
      "reward": 0.6004427671432495,
      "reward_std": 0.1143111027777195,
      "reward_std_all": 0.15114732086658478,
      "rewards-max/forensics_sinkhorn_reward": 0.9803160429000854,
      "rewards-max/forensics_strict_edge_reward": 0.927969753742218,
      "rewards-max/forensics_tiou_reward": 0.9799806773662567,
      "rewards/forensics_sinkhorn_reward": 0.4311448633670807,
      "rewards/forensics_strict_edge_reward": 0.14427487552165985,
      "rewards/forensics_tiou_reward": 0.19458508491516113,
      "step": 238
    },
    {
      "completion_length": 19.390625,
      "epoch": 0.6364846870838882,
      "grad_norm": 9.637722969055176,
      "kl": 1.076171875,
      "learning_rate": 8.413333333333333e-07,
      "loss": 0.1223,
      "reward": 0.6231351792812347,
      "reward_std": 0.12977687641978264,
      "reward_std_all": 0.15396983921527863,
      "rewards-max/forensics_sinkhorn_reward": 0.9847711324691772,
      "rewards-max/forensics_strict_edge_reward": 0.895060271024704,
      "rewards-max/forensics_tiou_reward": 0.9834457039833069,
      "rewards/forensics_sinkhorn_reward": 0.46408510208129883,
      "rewards/forensics_strict_edge_reward": 0.14024358615279198,
      "rewards/forensics_tiou_reward": 0.2123372033238411,
      "step": 239
    },
    {
      "completion_length": 18.109375,
      "epoch": 0.6391478029294274,
      "grad_norm": 14.529332160949707,
      "kl": 1.298828125,
      "learning_rate": 8.406666666666667e-07,
      "loss": 0.2024,
      "reward": 0.5857728719711304,
      "reward_std": 0.13886075466871262,
      "reward_std_all": 0.17010345309972763,
      "rewards-max/forensics_sinkhorn_reward": 0.9847248792648315,
      "rewards-max/forensics_strict_edge_reward": 0.8854433596134186,
      "rewards-max/forensics_tiou_reward": 0.9831984341144562,
      "rewards/forensics_sinkhorn_reward": 0.4640509635210037,
      "rewards/forensics_strict_edge_reward": 0.13229933381080627,
      "rewards/forensics_tiou_reward": 0.17878912389278412,
      "step": 240
    },
    {
      "completion_length": 17.71875,
      "epoch": 0.6418109187749668,
      "grad_norm": 11.006660461425781,
      "kl": 0.611328125,
      "learning_rate": 8.399999999999999e-07,
      "loss": -0.0306,
      "reward": 0.6388898193836212,
      "reward_std": 0.11384809389710426,
      "reward_std_all": 0.13567928969860077,
      "rewards-max/forensics_sinkhorn_reward": 0.9885037541389465,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9980824589729309,
      "rewards/forensics_sinkhorn_reward": 0.524876743555069,
      "rewards/forensics_strict_edge_reward": 0.18156781047582626,
      "rewards/forensics_tiou_reward": 0.303423672914505,
      "step": 241
    },
    {
      "completion_length": 21.015625,
      "epoch": 0.644474034620506,
      "grad_norm": 8.778552055358887,
      "kl": 0.83203125,
      "learning_rate": 8.393333333333334e-07,
      "loss": 0.1148,
      "reward": 0.6512473523616791,
      "reward_std": 0.12435828149318695,
      "reward_std_all": 0.13934126496315002,
      "rewards-max/forensics_sinkhorn_reward": 0.9875309765338898,
      "rewards-max/forensics_strict_edge_reward": 0.9464850425720215,
      "rewards-max/forensics_tiou_reward": 0.9942469894886017,
      "rewards/forensics_sinkhorn_reward": 0.5570741146802902,
      "rewards/forensics_strict_edge_reward": 0.20401650294661522,
      "rewards/forensics_tiou_reward": 0.32920774817466736,
      "step": 242
    },
    {
      "completion_length": 18.34375,
      "epoch": 0.6471371504660453,
      "grad_norm": 10.595473289489746,
      "kl": 0.587890625,
      "learning_rate": 8.386666666666666e-07,
      "loss": 0.243,
      "reward": 0.5917151272296906,
      "reward_std": 0.1564973220229149,
      "reward_std_all": 0.1767892986536026,
      "rewards-max/forensics_sinkhorn_reward": 0.9826066792011261,
      "rewards-max/forensics_strict_edge_reward": 0.8771188855171204,
      "rewards-max/forensics_tiou_reward": 0.9750571250915527,
      "rewards/forensics_sinkhorn_reward": 0.4664898067712784,
      "rewards/forensics_strict_edge_reward": 0.12842118740081787,
      "rewards/forensics_tiou_reward": 0.23344844579696655,
      "step": 243
    },
    {
      "completion_length": 17.546875,
      "epoch": 0.6498002663115846,
      "grad_norm": 9.341554641723633,
      "kl": 0.74609375,
      "learning_rate": 8.38e-07,
      "loss": 0.1198,
      "reward": 0.6955329775810242,
      "reward_std": 0.1137361079454422,
      "reward_std_all": 0.1115681417286396,
      "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": 0.5869027376174927,
      "rewards/forensics_strict_edge_reward": 0.2854182869195938,
      "rewards/forensics_tiou_reward": 0.35931193828582764,
      "step": 244
    },
    {
      "completion_length": 20.03125,
      "epoch": 0.6524633821571239,
      "grad_norm": 11.449000358581543,
      "kl": 0.919921875,
      "learning_rate": 8.373333333333333e-07,
      "loss": 0.3061,
      "reward": 0.5833157598972321,
      "reward_std": 0.12220191210508347,
      "reward_std_all": 0.16154096275568008,
      "rewards-max/forensics_sinkhorn_reward": 0.9128322303295135,
      "rewards-max/forensics_strict_edge_reward": 0.7281628549098969,
      "rewards-max/forensics_tiou_reward": 0.8263081610202789,
      "rewards/forensics_sinkhorn_reward": 0.4128365218639374,
      "rewards/forensics_strict_edge_reward": 0.1290443316102028,
      "rewards/forensics_tiou_reward": 0.18443265557289124,
      "step": 245
    },
    {
      "completion_length": 17.71875,
      "epoch": 0.6551264980026631,
      "grad_norm": 8.10509967803955,
      "kl": 0.8193359375,
      "learning_rate": 8.366666666666667e-07,
      "loss": 0.0444,
      "reward": 0.6334274709224701,
      "reward_std": 0.14098723977804184,
      "reward_std_all": 0.1473420411348343,
      "rewards-max/forensics_sinkhorn_reward": 0.9865230023860931,
      "rewards-max/forensics_strict_edge_reward": 0.9265981018543243,
      "rewards-max/forensics_tiou_reward": 0.990313321352005,
      "rewards/forensics_sinkhorn_reward": 0.4793490022420883,
      "rewards/forensics_strict_edge_reward": 0.12488172203302383,
      "rewards/forensics_tiou_reward": 0.22013381123542786,
      "step": 246
    },
    {
      "completion_length": 17.25,
      "epoch": 0.6577896138482024,
      "grad_norm": 8.795705795288086,
      "kl": 0.9453125,
      "learning_rate": 8.359999999999999e-07,
      "loss": 0.0314,
      "reward": 0.7165410816669464,
      "reward_std": 0.1049015112221241,
      "reward_std_all": 0.10246388614177704,
      "rewards-max/forensics_sinkhorn_reward": 0.9799319505691528,
      "rewards-max/forensics_strict_edge_reward": 0.7017043828964233,
      "rewards-max/forensics_tiou_reward": 0.9648831486701965,
      "rewards/forensics_sinkhorn_reward": 0.6466827392578125,
      "rewards/forensics_strict_edge_reward": 0.17613929137587547,
      "rewards/forensics_tiou_reward": 0.37066662311553955,
      "step": 247
    },
    {
      "completion_length": 17.609375,
      "epoch": 0.6604527296937417,
      "grad_norm": 10.956794738769531,
      "kl": 0.916015625,
      "learning_rate": 8.353333333333334e-07,
      "loss": 0.1988,
      "reward": 0.59388667345047,
      "reward_std": 0.1571369171142578,
      "reward_std_all": 0.17459450662136078,
      "rewards-max/forensics_sinkhorn_reward": 0.9880453944206238,
      "rewards-max/forensics_strict_edge_reward": 0.9252271056175232,
      "rewards-max/forensics_tiou_reward": 0.9962753653526306,
      "rewards/forensics_sinkhorn_reward": 0.4714347571134567,
      "rewards/forensics_strict_edge_reward": 0.22181133925914764,
      "rewards/forensics_tiou_reward": 0.2588799446821213,
      "step": 248
    },
    {
      "completion_length": 20.15625,
      "epoch": 0.6631158455392809,
      "grad_norm": 12.951305389404297,
      "kl": 0.9296875,
      "learning_rate": 8.346666666666666e-07,
      "loss": 0.1873,
      "reward": 0.6139063537120819,
      "reward_std": 0.10715112090110779,
      "reward_std_all": 0.13963690027594566,
      "rewards-max/forensics_sinkhorn_reward": 0.9049948155879974,
      "rewards-max/forensics_strict_edge_reward": 0.5548860281705856,
      "rewards-max/forensics_tiou_reward": 0.8156220316886902,
      "rewards/forensics_sinkhorn_reward": 0.49365904927253723,
      "rewards/forensics_strict_edge_reward": 0.08961662463843822,
      "rewards/forensics_tiou_reward": 0.21892138570547104,
      "step": 249
    },
    {
      "completion_length": 17.9375,
      "epoch": 0.6657789613848203,
      "grad_norm": 10.381532669067383,
      "kl": 1.259765625,
      "learning_rate": 8.34e-07,
      "loss": 0.1114,
      "reward": 0.64147087931633,
      "reward_std": 0.13316918164491653,
      "reward_std_all": 0.14567561447620392,
      "rewards-max/forensics_sinkhorn_reward": 0.9872403740882874,
      "rewards-max/forensics_strict_edge_reward": 0.946017861366272,
      "rewards-max/forensics_tiou_reward": 0.9931014478206635,
      "rewards/forensics_sinkhorn_reward": 0.5672041773796082,
      "rewards/forensics_strict_edge_reward": 0.23819604516029358,
      "rewards/forensics_tiou_reward": 0.3893163800239563,
      "step": 250
    },
    {
      "completion_length": 17.640625,
      "epoch": 0.6684420772303595,
      "grad_norm": 12.400884628295898,
      "kl": 1.1328125,
      "learning_rate": 8.333333333333333e-07,
      "loss": 0.1285,
      "reward": 0.5778560042381287,
      "reward_std": 0.16376717388629913,
      "reward_std_all": 0.18374748528003693,
      "rewards-max/forensics_sinkhorn_reward": 0.9517349004745483,
      "rewards-max/forensics_strict_edge_reward": 0.810933917760849,
      "rewards-max/forensics_tiou_reward": 0.9133533239364624,
      "rewards/forensics_sinkhorn_reward": 0.4584234952926636,
      "rewards/forensics_strict_edge_reward": 0.09553761407732964,
      "rewards/forensics_tiou_reward": 0.15900654718279839,
      "step": 251
    },
    {
      "completion_length": 19.453125,
      "epoch": 0.6711051930758988,
      "grad_norm": 13.518980979919434,
      "kl": 1.82421875,
      "learning_rate": 8.326666666666666e-07,
      "loss": 0.2612,
      "reward": 0.6425739228725433,
      "reward_std": 0.18861062824726105,
      "reward_std_all": 0.17900878936052322,
      "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": 0.5276815295219421,
      "rewards/forensics_strict_edge_reward": 0.20544984936714172,
      "rewards/forensics_tiou_reward": 0.26573997735977173,
      "step": 252
    },
    {
      "completion_length": 21.765625,
      "epoch": 0.6737683089214381,
      "grad_norm": 10.04255199432373,
      "kl": 0.9140625,
      "learning_rate": 8.319999999999999e-07,
      "loss": 0.1696,
      "reward": 0.6242047548294067,
      "reward_std": 0.1551303043961525,
      "reward_std_all": 0.16669514030218124,
      "rewards-max/forensics_sinkhorn_reward": 0.9878852963447571,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9956481158733368,
      "rewards/forensics_sinkhorn_reward": 0.5158143043518066,
      "rewards/forensics_strict_edge_reward": 0.1652745008468628,
      "rewards/forensics_tiou_reward": 0.22092606127262115,
      "step": 253
    },
    {
      "completion_length": 20.4375,
      "epoch": 0.6764314247669774,
      "grad_norm": 10.177725791931152,
      "kl": 1.0703125,
      "learning_rate": 8.313333333333333e-07,
      "loss": 0.1283,
      "reward": 0.6282060146331787,
      "reward_std": 0.1662949025630951,
      "reward_std_all": 0.17222264409065247,
      "rewards-max/forensics_sinkhorn_reward": 0.9886268973350525,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9985666573047638,
      "rewards/forensics_sinkhorn_reward": 0.545713871717453,
      "rewards/forensics_strict_edge_reward": 0.2229587882757187,
      "rewards/forensics_tiou_reward": 0.36973538994789124,
      "step": 254
    },
    {
      "completion_length": 22.90625,
      "epoch": 0.6790945406125166,
      "grad_norm": 9.800183296203613,
      "kl": 1.33203125,
      "learning_rate": 8.306666666666666e-07,
      "loss": 0.3056,
      "reward": 0.5984651148319244,
      "reward_std": 0.12801440060138702,
      "reward_std_all": 0.15991907566785812,
      "rewards-max/forensics_sinkhorn_reward": 0.8672530651092529,
      "rewards-max/forensics_strict_edge_reward": 0.4018917679786682,
      "rewards-max/forensics_tiou_reward": 0.7354022264480591,
      "rewards/forensics_sinkhorn_reward": 0.41617414355278015,
      "rewards/forensics_strict_edge_reward": 0.10628153942525387,
      "rewards/forensics_tiou_reward": 0.1640942469239235,
      "step": 255
    },
    {
      "completion_length": 21.75,
      "epoch": 0.681757656458056,
      "grad_norm": 9.997414588928223,
      "kl": 0.935546875,
      "learning_rate": 8.299999999999999e-07,
      "loss": 0.1494,
      "reward": 0.6019060611724854,
      "reward_std": 0.1823408454656601,
      "reward_std_all": 0.18897689878940582,
      "rewards-max/forensics_sinkhorn_reward": 0.9862858057022095,
      "rewards-max/forensics_strict_edge_reward": 0.9357379674911499,
      "rewards-max/forensics_tiou_reward": 0.9893827438354492,
      "rewards/forensics_sinkhorn_reward": 0.4910180866718292,
      "rewards/forensics_strict_edge_reward": 0.24271979928016663,
      "rewards/forensics_tiou_reward": 0.299411378800869,
      "step": 256
    },
    {
      "completion_length": 23.921875,
      "epoch": 0.6844207723035952,
      "grad_norm": 10.02680778503418,
      "kl": 0.96484375,
      "learning_rate": 8.293333333333333e-07,
      "loss": 0.0096,
      "reward": 0.6207666099071503,
      "reward_std": 0.16048917174339294,
      "reward_std_all": 0.1676059141755104,
      "rewards-max/forensics_sinkhorn_reward": 0.9353019595146179,
      "rewards-max/forensics_strict_edge_reward": 0.7588363587856293,
      "rewards-max/forensics_tiou_reward": 0.8770423233509064,
      "rewards/forensics_sinkhorn_reward": 0.48054051399230957,
      "rewards/forensics_strict_edge_reward": 0.18088239058852196,
      "rewards/forensics_tiou_reward": 0.30403459817171097,
      "step": 257
    },
    {
      "completion_length": 23.28125,
      "epoch": 0.6870838881491345,
      "grad_norm": 10.792930603027344,
      "kl": 1.08984375,
      "learning_rate": 8.286666666666666e-07,
      "loss": 0.1829,
      "reward": 0.6547953188419342,
      "reward_std": 0.12980087846517563,
      "reward_std_all": 0.14100630581378937,
      "rewards-max/forensics_sinkhorn_reward": 0.9866690933704376,
      "rewards-max/forensics_strict_edge_reward": 0.9357379674911499,
      "rewards-max/forensics_tiou_reward": 0.9908645451068878,
      "rewards/forensics_sinkhorn_reward": 0.5098649561405182,
      "rewards/forensics_strict_edge_reward": 0.19246065244078636,
      "rewards/forensics_tiou_reward": 0.32352420687675476,
      "step": 258
    },
    {
      "completion_length": 19.734375,
      "epoch": 0.6897470039946738,
      "grad_norm": 12.832439422607422,
      "kl": 1.14453125,
      "learning_rate": 8.28e-07,
      "loss": 0.2437,
      "reward": 0.6156850159168243,
      "reward_std": 0.15757203474640846,
      "reward_std_all": 0.1678774356842041,
      "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": 0.4974201172590256,
      "rewards/forensics_strict_edge_reward": 0.1691896840929985,
      "rewards/forensics_tiou_reward": 0.23741137236356735,
      "step": 259
    },
    {
      "completion_length": 18.8125,
      "epoch": 0.6924101198402131,
      "grad_norm": 11.599274635314941,
      "kl": 0.880859375,
      "learning_rate": 8.273333333333333e-07,
      "loss": -0.0716,
      "reward": 0.6446437239646912,
      "reward_std": 0.09715518355369568,
      "reward_std_all": 0.12380192056298256,
      "rewards-max/forensics_sinkhorn_reward": 0.9462969303131104,
      "rewards-max/forensics_strict_edge_reward": 0.6770844608545303,
      "rewards-max/forensics_tiou_reward": 0.9026462435722351,
      "rewards/forensics_sinkhorn_reward": 0.5117746293544769,
      "rewards/forensics_strict_edge_reward": 0.11772942543029785,
      "rewards/forensics_tiou_reward": 0.19275955110788345,
      "step": 260
    },
    {
      "completion_length": 19.03125,
      "epoch": 0.6950732356857523,
      "grad_norm": 11.781136512756348,
      "kl": 0.7509765625,
      "learning_rate": 8.266666666666667e-07,
      "loss": 0.2296,
      "reward": 0.6254642009735107,
      "reward_std": 0.16372431814670563,
      "reward_std_all": 0.165555439889431,
      "rewards-max/forensics_sinkhorn_reward": 0.987960010766983,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.995940238237381,
      "rewards/forensics_sinkhorn_reward": 0.52689129114151,
      "rewards/forensics_strict_edge_reward": 0.20436805486679077,
      "rewards/forensics_tiou_reward": 0.3017624169588089,
      "step": 261
    },
    {
      "completion_length": 17.625,
      "epoch": 0.6977363515312917,
      "grad_norm": 10.924505233764648,
      "kl": 1.1328125,
      "learning_rate": 8.259999999999999e-07,
      "loss": 0.0,
      "reward": 0.6156396567821503,
      "reward_std": 0.13114379718899727,
      "reward_std_all": 0.1619926020503044,
      "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": 0.5206404626369476,
      "rewards/forensics_strict_edge_reward": 0.2189016342163086,
      "rewards/forensics_tiou_reward": 0.2951079308986664,
      "step": 262
    },
    {
      "completion_length": 18.265625,
      "epoch": 0.7003994673768309,
      "grad_norm": 10.99704647064209,
      "kl": 0.6484375,
      "learning_rate": 8.253333333333334e-07,
      "loss": -0.0575,
      "reward": 0.6714907884597778,
      "reward_std": 0.11850103735923767,
      "reward_std_all": 0.1244141049683094,
      "rewards-max/forensics_sinkhorn_reward": 0.9868721067905426,
      "rewards-max/forensics_strict_edge_reward": 0.9464850425720215,
      "rewards-max/forensics_tiou_reward": 0.9916790425777435,
      "rewards/forensics_sinkhorn_reward": 0.550674706697464,
      "rewards/forensics_strict_edge_reward": 0.20417774468660355,
      "rewards/forensics_tiou_reward": 0.25886668264865875,
      "step": 263
    },
    {
      "completion_length": 16.265625,
      "epoch": 0.7030625832223701,
      "grad_norm": 11.724828720092773,
      "kl": 0.9609375,
      "learning_rate": 8.246666666666666e-07,
      "loss": -0.0765,
      "reward": 0.6403245329856873,
      "reward_std": 0.09316987171769142,
      "reward_std_all": 0.12443014979362488,
      "rewards-max/forensics_sinkhorn_reward": 0.9809998571872711,
      "rewards-max/forensics_strict_edge_reward": 0.8101339638233185,
      "rewards-max/forensics_tiou_reward": 0.9688819944858551,
      "rewards/forensics_sinkhorn_reward": 0.5152706801891327,
      "rewards/forensics_strict_edge_reward": 0.11449472606182098,
      "rewards/forensics_tiou_reward": 0.17338109761476517,
      "step": 264
    },
    {
      "completion_length": 16.1875,
      "epoch": 0.7057256990679095,
      "grad_norm": 9.317560195922852,
      "kl": 0.86328125,
      "learning_rate": 8.24e-07,
      "loss": 0.1737,
      "reward": 0.6732669770717621,
      "reward_std": 0.11349227279424667,
      "reward_std_all": 0.11842994764447212,
      "rewards-max/forensics_sinkhorn_reward": 0.9880331456661224,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9962261021137238,
      "rewards/forensics_sinkhorn_reward": 0.5890284180641174,
      "rewards/forensics_strict_edge_reward": 0.2948397994041443,
      "rewards/forensics_tiou_reward": 0.3681633472442627,
      "step": 265
    },
    {
      "completion_length": 18.25,
      "epoch": 0.7083888149134487,
      "grad_norm": 14.146949768066406,
      "kl": 1.21875,
      "learning_rate": 8.233333333333333e-07,
      "loss": 0.1564,
      "reward": 0.6509624421596527,
      "reward_std": 0.21122895926237106,
      "reward_std_all": 0.19211506843566895,
      "rewards-max/forensics_sinkhorn_reward": 0.9702796041965485,
      "rewards-max/forensics_strict_edge_reward": 0.6971181482076645,
      "rewards-max/forensics_tiou_reward": 0.9481084942817688,
      "rewards/forensics_sinkhorn_reward": 0.5095986574888229,
      "rewards/forensics_strict_edge_reward": 0.20727860927581787,
      "rewards/forensics_tiou_reward": 0.2837982773780823,
      "step": 266
    },
    {
      "completion_length": 18.328125,
      "epoch": 0.711051930758988,
      "grad_norm": 15.205549240112305,
      "kl": 1.3203125,
      "learning_rate": 8.226666666666666e-07,
      "loss": 0.1989,
      "reward": 0.6707205176353455,
      "reward_std": 0.11750712245702744,
      "reward_std_all": 0.1236029751598835,
      "rewards-max/forensics_sinkhorn_reward": 0.9717632830142975,
      "rewards-max/forensics_strict_edge_reward": 0.7244633734226227,
      "rewards-max/forensics_tiou_reward": 0.9341122508049011,
      "rewards/forensics_sinkhorn_reward": 0.4925232380628586,
      "rewards/forensics_strict_edge_reward": 0.16598066687583923,
      "rewards/forensics_tiou_reward": 0.2831322252750397,
      "step": 267
    },
    {
      "completion_length": 18.515625,
      "epoch": 0.7137150466045273,
      "grad_norm": 20.409404754638672,
      "kl": 1.58203125,
      "learning_rate": 8.219999999999999e-07,
      "loss": 0.2829,
      "reward": 0.61915123462677,
      "reward_std": 0.11384455487132072,
      "reward_std_all": 0.1419857069849968,
      "rewards-max/forensics_sinkhorn_reward": 0.9880613088607788,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9963403344154358,
      "rewards/forensics_sinkhorn_reward": 0.5414691865444183,
      "rewards/forensics_strict_edge_reward": 0.18348965793848038,
      "rewards/forensics_tiou_reward": 0.27765771746635437,
      "step": 268
    },
    {
      "completion_length": 21.78125,
      "epoch": 0.7163781624500666,
      "grad_norm": 7.111173152923584,
      "kl": 1.1875,
      "learning_rate": 8.213333333333333e-07,
      "loss": 0.2563,
      "reward": 0.5867463648319244,
      "reward_std": 0.18161974102258682,
      "reward_std_all": 0.1939767673611641,
      "rewards-max/forensics_sinkhorn_reward": 0.87320876121521,
      "rewards-max/forensics_strict_edge_reward": 0.7110060602426529,
      "rewards-max/forensics_tiou_reward": 0.7705346345901489,
      "rewards/forensics_sinkhorn_reward": 0.44088172912597656,
      "rewards/forensics_strict_edge_reward": 0.19656608626246452,
      "rewards/forensics_tiou_reward": 0.27007581293582916,
      "step": 269
    },
    {
      "completion_length": 21.109375,
      "epoch": 0.7190412782956058,
      "grad_norm": 12.447610855102539,
      "kl": 1.173828125,
      "learning_rate": 8.206666666666666e-07,
      "loss": 0.1722,
      "reward": 0.616735190153122,
      "reward_std": 0.18052896484732628,
      "reward_std_all": 0.18013907596468925,
      "rewards-max/forensics_sinkhorn_reward": 0.9852924942970276,
      "rewards-max/forensics_strict_edge_reward": 0.8426913917064667,
      "rewards-max/forensics_tiou_reward": 0.9854304492473602,
      "rewards/forensics_sinkhorn_reward": 0.5055177062749863,
      "rewards/forensics_strict_edge_reward": 0.27824651449918747,
      "rewards/forensics_tiou_reward": 0.35995565354824066,
      "step": 270
    },
    {
      "completion_length": 17.40625,
      "epoch": 0.7217043941411452,
      "grad_norm": 11.11212158203125,
      "kl": 1.109375,
      "learning_rate": 8.199999999999999e-07,
      "loss": 0.1304,
      "reward": 0.6463896036148071,
      "reward_std": 0.09609570726752281,
      "reward_std_all": 0.12279828637838364,
      "rewards-max/forensics_sinkhorn_reward": 0.9876440167427063,
      "rewards-max/forensics_strict_edge_reward": 0.946017861366272,
      "rewards-max/forensics_tiou_reward": 0.9946986138820648,
      "rewards/forensics_sinkhorn_reward": 0.5370108783245087,
      "rewards/forensics_strict_edge_reward": 0.17413035780191422,
      "rewards/forensics_tiou_reward": 0.2541157901287079,
      "step": 271
    },
    {
      "completion_length": 21.453125,
      "epoch": 0.7243675099866844,
      "grad_norm": 9.801152229309082,
      "kl": 1.2734375,
      "learning_rate": 8.193333333333333e-07,
      "loss": 0.2111,
      "reward": 0.6649723649024963,
      "reward_std": 0.14137892425060272,
      "reward_std_all": 0.14420641586184502,
      "rewards-max/forensics_sinkhorn_reward": 0.988587498664856,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9984112977981567,
      "rewards/forensics_sinkhorn_reward": 0.500913992524147,
      "rewards/forensics_strict_edge_reward": 0.24987442791461945,
      "rewards/forensics_tiou_reward": 0.3393667936325073,
      "step": 272
    },
    {
      "completion_length": 21.09375,
      "epoch": 0.7270306258322237,
      "grad_norm": 10.21261215209961,
      "kl": 1.1796875,
      "learning_rate": 8.186666666666666e-07,
      "loss": 0.1265,
      "reward": 0.6052590012550354,
      "reward_std": 0.10099596530199051,
      "reward_std_all": 0.14540743827819824,
      "rewards-max/forensics_sinkhorn_reward": 0.980522871017456,
      "rewards-max/forensics_strict_edge_reward": 0.882964164018631,
      "rewards-max/forensics_tiou_reward": 0.9672222137451172,
      "rewards/forensics_sinkhorn_reward": 0.4642595797777176,
      "rewards/forensics_strict_edge_reward": 0.17065753787755966,
      "rewards/forensics_tiou_reward": 0.21671444177627563,
      "step": 273
    },
    {
      "completion_length": 20.59375,
      "epoch": 0.729693741677763,
      "grad_norm": 9.212467193603516,
      "kl": 0.9921875,
      "learning_rate": 8.179999999999999e-07,
      "loss": 0.1732,
      "reward": 0.648341178894043,
      "reward_std": 0.1323242336511612,
      "reward_std_all": 0.14255884662270546,
      "rewards-max/forensics_sinkhorn_reward": 0.9710784256458282,
      "rewards-max/forensics_strict_edge_reward": 0.8668389916419983,
      "rewards-max/forensics_tiou_reward": 0.9561679065227509,
      "rewards/forensics_sinkhorn_reward": 0.5147879123687744,
      "rewards/forensics_strict_edge_reward": 0.20611119270324707,
      "rewards/forensics_tiou_reward": 0.3150556981563568,
      "step": 274
    },
    {
      "completion_length": 19.34375,
      "epoch": 0.7323568575233023,
      "grad_norm": 10.71447467803955,
      "kl": 0.830078125,
      "learning_rate": 8.173333333333333e-07,
      "loss": 0.1314,
      "reward": 0.6378756165504456,
      "reward_std": 0.11839346215128899,
      "reward_std_all": 0.1381784901022911,
      "rewards-max/forensics_sinkhorn_reward": 0.9886178374290466,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9985312819480896,
      "rewards/forensics_sinkhorn_reward": 0.5648188889026642,
      "rewards/forensics_strict_edge_reward": 0.2625877782702446,
      "rewards/forensics_tiou_reward": 0.3630296289920807,
      "step": 275
    },
    {
      "completion_length": 20.40625,
      "epoch": 0.7350199733688415,
      "grad_norm": 11.143919944763184,
      "kl": 0.970703125,
      "learning_rate": 8.166666666666666e-07,
      "loss": 0.1702,
      "reward": 0.6788950562477112,
      "reward_std": 0.09178367629647255,
      "reward_std_all": 0.10875441879034042,
      "rewards-max/forensics_sinkhorn_reward": 0.986598938703537,
      "rewards-max/forensics_strict_edge_reward": 0.9062341153621674,
      "rewards-max/forensics_tiou_reward": 0.9905738830566406,
      "rewards/forensics_sinkhorn_reward": 0.5420042276382446,
      "rewards/forensics_strict_edge_reward": 0.24460971355438232,
      "rewards/forensics_tiou_reward": 0.36111506819725037,
      "step": 276
    },
    {
      "completion_length": 18.71875,
      "epoch": 0.7376830892143809,
      "grad_norm": 11.43387508392334,
      "kl": 0.9921875,
      "learning_rate": 8.159999999999999e-07,
      "loss": 0.0277,
      "reward": 0.66496342420578,
      "reward_std": 0.15910161286592484,
      "reward_std_all": 0.15172132849693298,
      "rewards-max/forensics_sinkhorn_reward": 0.9799095690250397,
      "rewards-max/forensics_strict_edge_reward": 0.8528120815753937,
      "rewards-max/forensics_tiou_reward": 0.9647552073001862,
      "rewards/forensics_sinkhorn_reward": 0.5760262608528137,
      "rewards/forensics_strict_edge_reward": 0.23720106482505798,
      "rewards/forensics_tiou_reward": 0.36343882977962494,
      "step": 277
    },
    {
      "completion_length": 16.5,
      "epoch": 0.7403462050599201,
      "grad_norm": 11.425178527832031,
      "kl": 1.39453125,
      "learning_rate": 8.153333333333334e-07,
      "loss": 0.2801,
      "reward": 0.6287667155265808,
      "reward_std": 0.1098383441567421,
      "reward_std_all": 0.1413917988538742,
      "rewards-max/forensics_sinkhorn_reward": 0.9866214692592621,
      "rewards-max/forensics_strict_edge_reward": 0.9357379674911499,
      "rewards-max/forensics_tiou_reward": 0.9906690716743469,
      "rewards/forensics_sinkhorn_reward": 0.5194458663463593,
      "rewards/forensics_strict_edge_reward": 0.1716613620519638,
      "rewards/forensics_tiou_reward": 0.25893235206604004,
      "step": 278
    },
    {
      "completion_length": 21.296875,
      "epoch": 0.7430093209054593,
      "grad_norm": 8.849178314208984,
      "kl": 1.3828125,
      "learning_rate": 8.146666666666666e-07,
      "loss": 0.2432,
      "reward": 0.6440408229827881,
      "reward_std": 0.19446424394845963,
      "reward_std_all": 0.18124204128980637,
      "rewards-max/forensics_sinkhorn_reward": 0.9877912700176239,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9952805042266846,
      "rewards/forensics_sinkhorn_reward": 0.5632826387882233,
      "rewards/forensics_strict_edge_reward": 0.31619399785995483,
      "rewards/forensics_tiou_reward": 0.41398388147354126,
      "step": 279
    },
    {
      "completion_length": 19.578125,
      "epoch": 0.7456724367509987,
      "grad_norm": 12.574585914611816,
      "kl": 0.7890625,
      "learning_rate": 8.14e-07,
      "loss": 0.1929,
      "reward": 0.5836258381605148,
      "reward_std": 0.14507989212870598,
      "reward_std_all": 0.1713840626180172,
      "rewards-max/forensics_sinkhorn_reward": 0.9872406125068665,
      "rewards-max/forensics_strict_edge_reward": 0.9574736058712006,
      "rewards-max/forensics_tiou_reward": 0.9931034445762634,
      "rewards/forensics_sinkhorn_reward": 0.47739486396312714,
      "rewards/forensics_strict_edge_reward": 0.15513306856155396,
      "rewards/forensics_tiou_reward": 0.29213741421699524,
      "step": 280
    },
    {
      "completion_length": 16.875,
      "epoch": 0.748335552596538,
      "grad_norm": 222.6031036376953,
      "kl": 3.904296875,
      "learning_rate": 8.133333333333333e-07,
      "loss": 0.2771,
      "reward": 0.640634298324585,
      "reward_std": 0.11599704623222351,
      "reward_std_all": 0.13921751827001572,
      "rewards-max/forensics_sinkhorn_reward": 0.9118989109992981,
      "rewards-max/forensics_strict_edge_reward": 0.713139146566391,
      "rewards-max/forensics_tiou_reward": 0.8328157067298889,
      "rewards/forensics_sinkhorn_reward": 0.5515846610069275,
      "rewards/forensics_strict_edge_reward": 0.17551380768418312,
      "rewards/forensics_tiou_reward": 0.24235066398978233,
      "step": 281
    },
    {
      "completion_length": 19.625,
      "epoch": 0.7509986684420772,
      "grad_norm": 9.11777400970459,
      "kl": 1.48828125,
      "learning_rate": 8.126666666666666e-07,
      "loss": 0.0661,
      "reward": 0.6860064268112183,
      "reward_std": 0.09595155715942383,
      "reward_std_all": 0.10356295853853226,
      "rewards-max/forensics_sinkhorn_reward": 0.9872406125068665,
      "rewards-max/forensics_strict_edge_reward": 0.9677534997463226,
      "rewards-max/forensics_tiou_reward": 0.9931034445762634,
      "rewards/forensics_sinkhorn_reward": 0.6158185601234436,
      "rewards/forensics_strict_edge_reward": 0.2547091916203499,
      "rewards/forensics_tiou_reward": 0.4006357043981552,
      "step": 282
    },
    {
      "completion_length": 19.984375,
      "epoch": 0.7536617842876165,
      "grad_norm": 11.515162467956543,
      "kl": 1.02734375,
      "learning_rate": 8.12e-07,
      "loss": 0.1199,
      "reward": 0.6074201166629791,
      "reward_std": 0.15224438533186913,
      "reward_std_all": 0.16715199500322342,
      "rewards-max/forensics_sinkhorn_reward": 0.9858666658401489,
      "rewards-max/forensics_strict_edge_reward": 0.8893801271915436,
      "rewards-max/forensics_tiou_reward": 0.9877598285675049,
      "rewards/forensics_sinkhorn_reward": 0.5234895348548889,
      "rewards/forensics_strict_edge_reward": 0.2188110202550888,
      "rewards/forensics_tiou_reward": 0.3604973703622818,
      "step": 283
    },
    {
      "completion_length": 18.734375,
      "epoch": 0.7563249001331558,
      "grad_norm": 12.560454368591309,
      "kl": 1.1640625,
      "learning_rate": 8.113333333333333e-07,
      "loss": 0.2046,
      "reward": 0.6133989691734314,
      "reward_std": 0.13377317786216736,
      "reward_std_all": 0.15882693976163864,
      "rewards-max/forensics_sinkhorn_reward": 0.9878761768341064,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9956259429454803,
      "rewards/forensics_sinkhorn_reward": 0.5390784740447998,
      "rewards/forensics_strict_edge_reward": 0.22566930949687958,
      "rewards/forensics_tiou_reward": 0.30675433576107025,
      "step": 284
    },
    {
      "completion_length": 18.828125,
      "epoch": 0.758988015978695,
      "grad_norm": 15.785697937011719,
      "kl": 1.2265625,
      "learning_rate": 8.106666666666666e-07,
      "loss": 0.3583,
      "reward": 0.6828575730323792,
      "reward_std": 0.11230028793215752,
      "reward_std_all": 0.12209295481443405,
      "rewards-max/forensics_sinkhorn_reward": 0.9864945709705353,
      "rewards-max/forensics_strict_edge_reward": 0.9265981018543243,
      "rewards-max/forensics_tiou_reward": 0.9901986122131348,
      "rewards/forensics_sinkhorn_reward": 0.5592545866966248,
      "rewards/forensics_strict_edge_reward": 0.18731823563575745,
      "rewards/forensics_tiou_reward": 0.3192148506641388,
      "step": 285
    },
    {
      "completion_length": 16.734375,
      "epoch": 0.7616511318242344,
      "grad_norm": 13.985852241516113,
      "kl": 1.1455078125,
      "learning_rate": 8.1e-07,
      "loss": 0.0712,
      "reward": 0.6312172710895538,
      "reward_std": 0.08494680747389793,
      "reward_std_all": 0.12558545544743538,
      "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": 0.5390727669000626,
      "rewards/forensics_strict_edge_reward": 0.2099587619304657,
      "rewards/forensics_tiou_reward": 0.2623070031404495,
      "step": 286
    },
    {
      "completion_length": 17.578125,
      "epoch": 0.7643142476697736,
      "grad_norm": 11.286441802978516,
      "kl": 1.4296875,
      "learning_rate": 8.093333333333333e-07,
      "loss": 0.1465,
      "reward": 0.6689505875110626,
      "reward_std": 0.12809867411851883,
      "reward_std_all": 0.13119173049926758,
      "rewards-max/forensics_sinkhorn_reward": 0.9875111877918243,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9941709339618683,
      "rewards/forensics_sinkhorn_reward": 0.5201993584632874,
      "rewards/forensics_strict_edge_reward": 0.2549012750387192,
      "rewards/forensics_tiou_reward": 0.294336661696434,
      "step": 287
    },
    {
      "completion_length": 19.265625,
      "epoch": 0.7669773635153129,
      "grad_norm": 11.699003219604492,
      "kl": 1.447265625,
      "learning_rate": 8.086666666666666e-07,
      "loss": 0.3505,
      "reward": 0.6510405838489532,
      "reward_std": 0.14848759770393372,
      "reward_std_all": 0.15184693038463593,
      "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": 0.5967718362808228,
      "rewards/forensics_strict_edge_reward": 0.31257903575897217,
      "rewards/forensics_tiou_reward": 0.4223356395959854,
      "step": 288
    },
    {
      "completion_length": 17.125,
      "epoch": 0.7696404793608522,
      "grad_norm": 12.164105415344238,
      "kl": 1.0859375,
      "learning_rate": 8.08e-07,
      "loss": 0.2197,
      "reward": 0.6080265641212463,
      "reward_std": 0.15664425864815712,
      "reward_std_all": 0.1741364449262619,
      "rewards-max/forensics_sinkhorn_reward": 0.9819956719875336,
      "rewards-max/forensics_strict_edge_reward": 0.8719756007194519,
      "rewards-max/forensics_tiou_reward": 0.9727922081947327,
      "rewards/forensics_sinkhorn_reward": 0.5233410000801086,
      "rewards/forensics_strict_edge_reward": 0.2217693291604519,
      "rewards/forensics_tiou_reward": 0.31666404008865356,
      "step": 289
    },
    {
      "completion_length": 17.421875,
      "epoch": 0.7723035952063915,
      "grad_norm": 14.122589111328125,
      "kl": 1.421875,
      "learning_rate": 8.073333333333333e-07,
      "loss": 0.1702,
      "reward": 0.6450920403003693,
      "reward_std": 0.13155421242117882,
      "reward_std_all": 0.1463315188884735,
      "rewards-max/forensics_sinkhorn_reward": 0.9875145554542542,
      "rewards-max/forensics_strict_edge_reward": 0.9565287232398987,
      "rewards-max/forensics_tiou_reward": 0.9941868782043457,
      "rewards/forensics_sinkhorn_reward": 0.5781508982181549,
      "rewards/forensics_strict_edge_reward": 0.2860706001520157,
      "rewards/forensics_tiou_reward": 0.3898417800664902,
      "step": 290
    },
    {
      "completion_length": 18.59375,
      "epoch": 0.7749667110519307,
      "grad_norm": 11.88532543182373,
      "kl": 1.7421875,
      "learning_rate": 8.066666666666666e-07,
      "loss": 0.2107,
      "reward": 0.660124659538269,
      "reward_std": 0.12083186209201813,
      "reward_std_all": 0.12710563838481903,
      "rewards-max/forensics_sinkhorn_reward": 0.9722632169723511,
      "rewards-max/forensics_strict_edge_reward": 0.8416292369365692,
      "rewards-max/forensics_tiou_reward": 0.9361904859542847,
      "rewards/forensics_sinkhorn_reward": 0.529259979724884,
      "rewards/forensics_strict_edge_reward": 0.2030164748430252,
      "rewards/forensics_tiou_reward": 0.27481040358543396,
      "step": 291
    },
    {
      "completion_length": 19.296875,
      "epoch": 0.7776298268974701,
      "grad_norm": 11.582927703857422,
      "kl": 1.515625,
      "learning_rate": 8.06e-07,
      "loss": 0.31,
      "reward": 0.6186321377754211,
      "reward_std": 0.15392021089792252,
      "reward_std_all": 0.167552188038826,
      "rewards-max/forensics_sinkhorn_reward": 0.9883430302143097,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9974490702152252,
      "rewards/forensics_sinkhorn_reward": 0.5773393511772156,
      "rewards/forensics_strict_edge_reward": 0.31072811782360077,
      "rewards/forensics_tiou_reward": 0.41680708527565,
      "step": 292
    },
    {
      "completion_length": 24.640625,
      "epoch": 0.7802929427430093,
      "grad_norm": 10.170701026916504,
      "kl": 1.3828125,
      "learning_rate": 8.053333333333333e-07,
      "loss": 0.3548,
      "reward": 0.6343289017677307,
      "reward_std": 0.13777702301740646,
      "reward_std_all": 0.14915581047534943,
      "rewards-max/forensics_sinkhorn_reward": 0.9796305894851685,
      "rewards-max/forensics_strict_edge_reward": 0.9357379674911499,
      "rewards-max/forensics_tiou_reward": 0.977931022644043,
      "rewards/forensics_sinkhorn_reward": 0.45570966601371765,
      "rewards/forensics_strict_edge_reward": 0.2502523809671402,
      "rewards/forensics_tiou_reward": 0.31752534210681915,
      "step": 293
    },
    {
      "completion_length": 19.671875,
      "epoch": 0.7829560585885486,
      "grad_norm": 14.830796241760254,
      "kl": 2.19140625,
      "learning_rate": 8.046666666666666e-07,
      "loss": 0.2434,
      "reward": 0.6416351199150085,
      "reward_std": 0.19349273294210434,
      "reward_std_all": 0.1867256909608841,
      "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": 0.545047253370285,
      "rewards/forensics_strict_edge_reward": 0.33225758373737335,
      "rewards/forensics_tiou_reward": 0.421848326921463,
      "step": 294
    },
    {
      "completion_length": 19.59375,
      "epoch": 0.7856191744340879,
      "grad_norm": 12.343267440795898,
      "kl": 1.35546875,
      "learning_rate": 8.04e-07,
      "loss": 0.2011,
      "reward": 0.6645878553390503,
      "reward_std": 0.1502804383635521,
      "reward_std_all": 0.14691200107336044,
      "rewards-max/forensics_sinkhorn_reward": 0.9767015278339386,
      "rewards-max/forensics_strict_edge_reward": 0.9265981018543243,
      "rewards-max/forensics_tiou_reward": 0.9710128307342529,
      "rewards/forensics_sinkhorn_reward": 0.5295840799808502,
      "rewards/forensics_strict_edge_reward": 0.2868490666151047,
      "rewards/forensics_tiou_reward": 0.34531643986701965,
      "step": 295
    },
    {
      "completion_length": 21.578125,
      "epoch": 0.7882822902796272,
      "grad_norm": 11.374067306518555,
      "kl": 1.44921875,
      "learning_rate": 8.033333333333333e-07,
      "loss": 0.2261,
      "reward": 0.6017534732818604,
      "reward_std": 0.17590152472257614,
      "reward_std_all": 0.18754343688488007,
      "rewards-max/forensics_sinkhorn_reward": 0.8597002923488617,
      "rewards-max/forensics_strict_edge_reward": 0.8000465035438538,
      "rewards-max/forensics_tiou_reward": 0.7483347058296204,
      "rewards/forensics_sinkhorn_reward": 0.45291203260421753,
      "rewards/forensics_strict_edge_reward": 0.23831724375486374,
      "rewards/forensics_tiou_reward": 0.2759634852409363,
      "step": 296
    },
    {
      "completion_length": 21.125,
      "epoch": 0.7909454061251664,
      "grad_norm": 27.23101806640625,
      "kl": 2.5,
      "learning_rate": 8.026666666666667e-07,
      "loss": 0.1516,
      "reward": 0.5882476270198822,
      "reward_std": 0.18325142562389374,
      "reward_std_all": 0.19176343828439713,
      "rewards-max/forensics_sinkhorn_reward": 0.955661416053772,
      "rewards-max/forensics_strict_edge_reward": 0.7400782704353333,
      "rewards-max/forensics_tiou_reward": 0.9249157309532166,
      "rewards/forensics_sinkhorn_reward": 0.4754004031419754,
      "rewards/forensics_strict_edge_reward": 0.12983134388923645,
      "rewards/forensics_tiou_reward": 0.2267499417066574,
      "step": 297
    },
    {
      "completion_length": 21.1875,
      "epoch": 0.7936085219707057,
      "grad_norm": 12.964068412780762,
      "kl": 1.65234375,
      "learning_rate": 8.02e-07,
      "loss": 0.2411,
      "reward": 0.5670634508132935,
      "reward_std": 0.18838468566536903,
      "reward_std_all": 0.20606641471385956,
      "rewards-max/forensics_sinkhorn_reward": 0.9756947755813599,
      "rewards-max/forensics_strict_edge_reward": 0.8241714537143707,
      "rewards-max/forensics_tiou_reward": 0.9490256607532501,
      "rewards/forensics_sinkhorn_reward": 0.4394993931055069,
      "rewards/forensics_strict_edge_reward": 0.12241263315081596,
      "rewards/forensics_tiou_reward": 0.2089875191450119,
      "step": 298
    },
    {
      "completion_length": 20.40625,
      "epoch": 0.796271637816245,
      "grad_norm": 13.351824760437012,
      "kl": 1.73046875,
      "learning_rate": 8.013333333333333e-07,
      "loss": 0.0926,
      "reward": 0.6312801241874695,
      "reward_std": 0.0987207405269146,
      "reward_std_all": 0.137701578438282,
      "rewards-max/forensics_sinkhorn_reward": 0.9838697612285614,
      "rewards-max/forensics_strict_edge_reward": 0.9364311099052429,
      "rewards-max/forensics_tiou_reward": 0.9859069287776947,
      "rewards/forensics_sinkhorn_reward": 0.5346105098724365,
      "rewards/forensics_strict_edge_reward": 0.25626757740974426,
      "rewards/forensics_tiou_reward": 0.3371104300022125,
      "step": 299
    },
    {
      "completion_length": 18.453125,
      "epoch": 0.7989347536617842,
      "grad_norm": 12.912697792053223,
      "kl": 1.609375,
      "learning_rate": 8.006666666666666e-07,
      "loss": 0.2374,
      "reward": 0.6608543992042542,
      "reward_std": 0.14243775606155396,
      "reward_std_all": 0.14205548912286758,
      "rewards-max/forensics_sinkhorn_reward": 0.9881232976913452,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9965846836566925,
      "rewards/forensics_sinkhorn_reward": 0.5604559183120728,
      "rewards/forensics_strict_edge_reward": 0.2815285921096802,
      "rewards/forensics_tiou_reward": 0.3426913768053055,
      "step": 300
    },
    {
      "completion_length": 20.140625,
      "epoch": 0.8015978695073236,
      "grad_norm": 1481.3287353515625,
      "kl": 23.359375,
      "learning_rate": 8e-07,
      "loss": 1.1044,
      "reward": 0.6555112898349762,
      "reward_std": 0.15178877860307693,
      "reward_std_all": 0.1482553631067276,
      "rewards-max/forensics_sinkhorn_reward": 0.9860453009605408,
      "rewards-max/forensics_strict_edge_reward": 0.8365300297737122,
      "rewards-max/forensics_tiou_reward": 0.988394558429718,
      "rewards/forensics_sinkhorn_reward": 0.5559071898460388,
      "rewards/forensics_strict_edge_reward": 0.2532515674829483,
      "rewards/forensics_tiou_reward": 0.3745286166667938,
      "step": 301
    },
    {
      "completion_length": 22.125,
      "epoch": 0.8042609853528628,
      "grad_norm": 24.696693420410156,
      "kl": 3.5546875,
      "learning_rate": 7.993333333333333e-07,
      "loss": 0.1789,
      "reward": 0.5799663662910461,
      "reward_std": 0.1459280028939247,
      "reward_std_all": 0.1788794919848442,
      "rewards-max/forensics_sinkhorn_reward": 0.9633235335350037,
      "rewards-max/forensics_strict_edge_reward": 0.6856317967176437,
      "rewards-max/forensics_tiou_reward": 0.9407934248447418,
      "rewards/forensics_sinkhorn_reward": 0.4555535167455673,
      "rewards/forensics_strict_edge_reward": 0.1197454147040844,
      "rewards/forensics_tiou_reward": 0.2044651061296463,
      "step": 302
    },
    {
      "completion_length": 20.046875,
      "epoch": 0.8069241011984021,
      "grad_norm": 12.7333402633667,
      "kl": 1.30859375,
      "learning_rate": 7.986666666666666e-07,
      "loss": 0.2814,
      "reward": 0.6493317186832428,
      "reward_std": 0.11048568040132523,
      "reward_std_all": 0.1289139911532402,
      "rewards-max/forensics_sinkhorn_reward": 0.9871033728122711,
      "rewards-max/forensics_strict_edge_reward": 0.9355069994926453,
      "rewards-max/forensics_tiou_reward": 0.9925629496574402,
      "rewards/forensics_sinkhorn_reward": 0.5271479785442352,
      "rewards/forensics_strict_edge_reward": 0.2338482066988945,
      "rewards/forensics_tiou_reward": 0.3230924904346466,
      "step": 303
    },
    {
      "completion_length": 20.703125,
      "epoch": 0.8095872170439414,
      "grad_norm": 11.896825790405273,
      "kl": 1.94140625,
      "learning_rate": 7.98e-07,
      "loss": 0.3314,
      "reward": 0.6068331897258759,
      "reward_std": 0.14266842603683472,
      "reward_std_all": 0.1668700873851776,
      "rewards-max/forensics_sinkhorn_reward": 0.9859626591205597,
      "rewards-max/forensics_strict_edge_reward": 0.9149472117424011,
      "rewards-max/forensics_tiou_reward": 0.9880807399749756,
      "rewards/forensics_sinkhorn_reward": 0.4810812771320343,
      "rewards/forensics_strict_edge_reward": 0.2040209248661995,
      "rewards/forensics_tiou_reward": 0.26515768468379974,
      "step": 304
    },
    {
      "completion_length": 20.3125,
      "epoch": 0.8122503328894807,
      "grad_norm": 11.520916938781738,
      "kl": 1.92578125,
      "learning_rate": 7.973333333333333e-07,
      "loss": 0.2127,
      "reward": 0.663652628660202,
      "reward_std": 0.13493932783603668,
      "reward_std_all": 0.13586333394050598,
      "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": 0.5705940127372742,
      "rewards/forensics_strict_edge_reward": 0.3332349956035614,
      "rewards/forensics_tiou_reward": 0.40181295573711395,
      "step": 305
    },
    {
      "completion_length": 20.4375,
      "epoch": 0.8149134487350199,
      "grad_norm": 11.61680793762207,
      "kl": 1.30078125,
      "learning_rate": 7.966666666666666e-07,
      "loss": 0.1062,
      "reward": 0.6583661437034607,
      "reward_std": 0.1678488776087761,
      "reward_std_all": 0.1621711626648903,
      "rewards-max/forensics_sinkhorn_reward": 0.9867400527000427,
      "rewards-max/forensics_strict_edge_reward": 0.9357379674911499,
      "rewards-max/forensics_tiou_reward": 0.991150826215744,
      "rewards/forensics_sinkhorn_reward": 0.5474845468997955,
      "rewards/forensics_strict_edge_reward": 0.19154537469148636,
      "rewards/forensics_tiou_reward": 0.3167360723018646,
      "step": 306
    },
    {
      "completion_length": 19.953125,
      "epoch": 0.8175765645805593,
      "grad_norm": 11.903273582458496,
      "kl": 1.46484375,
      "learning_rate": 7.96e-07,
      "loss": 0.1543,
      "reward": 0.5968060195446014,
      "reward_std": 0.15272284299135208,
      "reward_std_all": 0.18127433955669403,
      "rewards-max/forensics_sinkhorn_reward": 0.9857164025306702,
      "rewards-max/forensics_strict_edge_reward": 0.9062341153621674,
      "rewards-max/forensics_tiou_reward": 0.9871439039707184,
      "rewards/forensics_sinkhorn_reward": 0.5048179775476456,
      "rewards/forensics_strict_edge_reward": 0.1779671385884285,
      "rewards/forensics_tiou_reward": 0.29062266647815704,
      "step": 307
    },
    {
      "completion_length": 17.390625,
      "epoch": 0.8202396804260985,
      "grad_norm": 12.847716331481934,
      "kl": 1.69921875,
      "learning_rate": 7.953333333333333e-07,
      "loss": 0.1604,
      "reward": 0.5946975350379944,
      "reward_std": 0.17061685770750046,
      "reward_std_all": 0.18657372891902924,
      "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": 0.5300185978412628,
      "rewards/forensics_strict_edge_reward": 0.19795922189950943,
      "rewards/forensics_tiou_reward": 0.31431545317173004,
      "step": 308
    },
    {
      "completion_length": 17.265625,
      "epoch": 0.8229027962716379,
      "grad_norm": 14.237764358520508,
      "kl": 1.515625,
      "learning_rate": 7.946666666666666e-07,
      "loss": 0.0997,
      "reward": 0.6720875799655914,
      "reward_std": 0.10201545059680939,
      "reward_std_all": 0.11666955053806305,
      "rewards-max/forensics_sinkhorn_reward": 0.9882625639438629,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9971331357955933,
      "rewards/forensics_sinkhorn_reward": 0.6253772675991058,
      "rewards/forensics_strict_edge_reward": 0.3661041110754013,
      "rewards/forensics_tiou_reward": 0.4518081694841385,
      "step": 309
    },
    {
      "completion_length": 16.359375,
      "epoch": 0.8255659121171771,
      "grad_norm": 15.404681205749512,
      "kl": 2.74609375,
      "learning_rate": 7.94e-07,
      "loss": 0.3083,
      "reward": 0.6739621460437775,
      "reward_std": 0.11575056985020638,
      "reward_std_all": 0.12460322305560112,
      "rewards-max/forensics_sinkhorn_reward": 0.9876488745212555,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9947141408920288,
      "rewards/forensics_sinkhorn_reward": 0.6542766392230988,
      "rewards/forensics_strict_edge_reward": 0.29065361618995667,
      "rewards/forensics_tiou_reward": 0.4599416255950928,
      "step": 310
    },
    {
      "completion_length": 21.6875,
      "epoch": 0.8282290279627164,
      "grad_norm": 9.682944297790527,
      "kl": 1.53125,
      "learning_rate": 7.933333333333333e-07,
      "loss": 0.1465,
      "reward": 0.6570861339569092,
      "reward_std": 0.1344311609864235,
      "reward_std_all": 0.14330856502056122,
      "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": 0.5516262650489807,
      "rewards/forensics_strict_edge_reward": 0.23196908831596375,
      "rewards/forensics_tiou_reward": 0.34372737258672714,
      "step": 311
    },
    {
      "completion_length": 17.390625,
      "epoch": 0.8308921438082557,
      "grad_norm": 17.240427017211914,
      "kl": 1.78515625,
      "learning_rate": 7.926666666666666e-07,
      "loss": 0.1369,
      "reward": 0.6520290374755859,
      "reward_std": 0.13585468381643295,
      "reward_std_all": 0.14023428782820702,
      "rewards-max/forensics_sinkhorn_reward": 0.8851341903209686,
      "rewards-max/forensics_strict_edge_reward": 0.5552518665790558,
      "rewards-max/forensics_tiou_reward": 0.773091733455658,
      "rewards/forensics_sinkhorn_reward": 0.5030836462974548,
      "rewards/forensics_strict_edge_reward": 0.13569985330104828,
      "rewards/forensics_tiou_reward": 0.2419532760977745,
      "step": 312
    },
    {
      "completion_length": 18.15625,
      "epoch": 0.833555259653795,
      "grad_norm": 15.061391830444336,
      "kl": 1.71875,
      "learning_rate": 7.92e-07,
      "loss": 0.1827,
      "reward": 0.694728672504425,
      "reward_std": 0.11751728132367134,
      "reward_std_all": 0.11725055426359177,
      "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": 0.63373664021492,
      "rewards/forensics_strict_edge_reward": 0.33414238691329956,
      "rewards/forensics_tiou_reward": 0.4544195979833603,
      "step": 313
    },
    {
      "completion_length": 18.3125,
      "epoch": 0.8362183754993342,
      "grad_norm": 12.862442016601562,
      "kl": 1.111328125,
      "learning_rate": 7.913333333333332e-07,
      "loss": 0.1261,
      "reward": 0.6476929783821106,
      "reward_std": 0.1056831069290638,
      "reward_std_all": 0.12576057761907578,
      "rewards-max/forensics_sinkhorn_reward": 0.9858572483062744,
      "rewards-max/forensics_strict_edge_reward": 0.8205901980400085,
      "rewards-max/forensics_tiou_reward": 0.9877305328845978,
      "rewards/forensics_sinkhorn_reward": 0.5894222557544708,
      "rewards/forensics_strict_edge_reward": 0.21862389892339706,
      "rewards/forensics_tiou_reward": 0.34841417521238327,
      "step": 314
    },
    {
      "completion_length": 16.984375,
      "epoch": 0.8388814913448736,
      "grad_norm": 12.236846923828125,
      "kl": 1.16796875,
      "learning_rate": 7.906666666666666e-07,
      "loss": 0.2276,
      "reward": 0.6551576852798462,
      "reward_std": 0.12447256222367287,
      "reward_std_all": 0.13443556427955627,
      "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": 0.614607036113739,
      "rewards/forensics_strict_edge_reward": 0.33710336685180664,
      "rewards/forensics_tiou_reward": 0.39343075454235077,
      "step": 315
    },
    {
      "completion_length": 17.453125,
      "epoch": 0.8415446071904128,
      "grad_norm": 9.95130729675293,
      "kl": 1.490234375,
      "learning_rate": 7.9e-07,
      "loss": 0.1885,
      "reward": 0.6336425244808197,
      "reward_std": 0.1474345326423645,
      "reward_std_all": 0.15443845093250275,
      "rewards-max/forensics_sinkhorn_reward": 0.9874955117702484,
      "rewards-max/forensics_strict_edge_reward": 0.878633052110672,
      "rewards-max/forensics_tiou_reward": 0.9941156804561615,
      "rewards/forensics_sinkhorn_reward": 0.5292112529277802,
      "rewards/forensics_strict_edge_reward": 0.15965255349874496,
      "rewards/forensics_tiou_reward": 0.30990535765886307,
      "step": 316
    },
    {
      "completion_length": 18.703125,
      "epoch": 0.844207723035952,
      "grad_norm": 10.12855052947998,
      "kl": 1.01953125,
      "learning_rate": 7.893333333333333e-07,
      "loss": 0.0381,
      "reward": 0.6678285002708435,
      "reward_std": 0.14580336958169937,
      "reward_std_all": 0.14676880836486816,
      "rewards-max/forensics_sinkhorn_reward": 0.9878180623054504,
      "rewards-max/forensics_strict_edge_reward": 0.9567649364471436,
      "rewards-max/forensics_tiou_reward": 0.9953836798667908,
      "rewards/forensics_sinkhorn_reward": 0.5757104456424713,
      "rewards/forensics_strict_edge_reward": 0.28147005289793015,
      "rewards/forensics_tiou_reward": 0.3400588780641556,
      "step": 317
    },
    {
      "completion_length": 17.0,
      "epoch": 0.8468708388814914,
      "grad_norm": 13.697378158569336,
      "kl": 0.99609375,
      "learning_rate": 7.886666666666666e-07,
      "loss": 0.0964,
      "reward": 0.5940702259540558,
      "reward_std": 0.16000856459140778,
      "reward_std_all": 0.17727256566286087,
      "rewards-max/forensics_sinkhorn_reward": 0.984457403421402,
      "rewards-max/forensics_strict_edge_reward": 0.9093653857707977,
      "rewards-max/forensics_tiou_reward": 0.9823023080825806,
      "rewards/forensics_sinkhorn_reward": 0.49729979038238525,
      "rewards/forensics_strict_edge_reward": 0.14140323922038078,
      "rewards/forensics_tiou_reward": 0.20286939665675163,
      "step": 318
    },
    {
      "completion_length": 17.0,
      "epoch": 0.8495339547270306,
      "grad_norm": 12.389047622680664,
      "kl": 1.005859375,
      "learning_rate": 7.88e-07,
      "loss": 0.1593,
      "reward": 0.6488113701343536,
      "reward_std": 0.11509359627962112,
      "reward_std_all": 0.13335374742746353,
      "rewards-max/forensics_sinkhorn_reward": 0.9886340200901031,
      "rewards-max/forensics_strict_edge_reward": 0.9565287232398987,
      "rewards-max/forensics_tiou_reward": 0.9985951483249664,
      "rewards/forensics_sinkhorn_reward": 0.558227926492691,
      "rewards/forensics_strict_edge_reward": 0.23399759829044342,
      "rewards/forensics_tiou_reward": 0.34115344285964966,
      "step": 319
    },
    {
      "completion_length": 20.46875,
      "epoch": 0.8521970705725699,
      "grad_norm": 14.326817512512207,
      "kl": 1.17578125,
      "learning_rate": 7.873333333333333e-07,
      "loss": 0.2635,
      "reward": 0.6542325615882874,
      "reward_std": 0.1353820636868477,
      "reward_std_all": 0.14461755752563477,
      "rewards-max/forensics_sinkhorn_reward": 0.9880039393901825,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9961146712303162,
      "rewards/forensics_sinkhorn_reward": 0.5444271564483643,
      "rewards/forensics_strict_edge_reward": 0.24212660640478134,
      "rewards/forensics_tiou_reward": 0.3598214387893677,
      "step": 320
    },
    {
      "completion_length": 18.0,
      "epoch": 0.8548601864181092,
      "grad_norm": 16.359195709228516,
      "kl": 2.390625,
      "learning_rate": 7.866666666666666e-07,
      "loss": 0.2624,
      "reward": 0.6833318769931793,
      "reward_std": 0.08103765547275543,
      "reward_std_all": 0.10208843275904655,
      "rewards-max/forensics_sinkhorn_reward": 0.9886032044887543,
      "rewards-max/forensics_strict_edge_reward": 0.9782643616199493,
      "rewards-max/forensics_tiou_reward": 0.9984732866287231,
      "rewards/forensics_sinkhorn_reward": 0.6033858060836792,
      "rewards/forensics_strict_edge_reward": 0.3021548390388489,
      "rewards/forensics_tiou_reward": 0.3967333734035492,
      "step": 321
    },
    {
      "completion_length": 18.359375,
      "epoch": 0.8575233022636485,
      "grad_norm": 14.079505920410156,
      "kl": 2.03515625,
      "learning_rate": 7.86e-07,
      "loss": 0.1255,
      "reward": 0.6208773553371429,
      "reward_std": 0.13994338363409042,
      "reward_std_all": 0.1548539102077484,
      "rewards-max/forensics_sinkhorn_reward": 0.9647980332374573,
      "rewards-max/forensics_strict_edge_reward": 0.8078656196594238,
      "rewards-max/forensics_tiou_reward": 0.9418855011463165,
      "rewards/forensics_sinkhorn_reward": 0.4580666422843933,
      "rewards/forensics_strict_edge_reward": 0.12084084004163742,
      "rewards/forensics_tiou_reward": 0.17599821835756302,
      "step": 322
    },
    {
      "completion_length": 21.359375,
      "epoch": 0.8601864181091877,
      "grad_norm": 12.00942325592041,
      "kl": 2.0,
      "learning_rate": 7.853333333333333e-07,
      "loss": 0.1257,
      "reward": 0.6290614604949951,
      "reward_std": 0.11448922008275986,
      "reward_std_all": 0.14030347764492035,
      "rewards-max/forensics_sinkhorn_reward": 0.9880734086036682,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9963885247707367,
      "rewards/forensics_sinkhorn_reward": 0.5110504627227783,
      "rewards/forensics_strict_edge_reward": 0.19148986041545868,
      "rewards/forensics_tiou_reward": 0.26302866637706757,
      "step": 323
    },
    {
      "completion_length": 31.15625,
      "epoch": 0.8628495339547271,
      "grad_norm": 10.53917407989502,
      "kl": 1.58984375,
      "learning_rate": 7.846666666666666e-07,
      "loss": 0.2778,
      "reward": 0.6183070540428162,
      "reward_std": 0.14003052935004234,
      "reward_std_all": 0.1618386134505272,
      "rewards-max/forensics_sinkhorn_reward": 0.95914426445961,
      "rewards-max/forensics_strict_edge_reward": 0.9265981018543243,
      "rewards-max/forensics_tiou_reward": 0.9352794885635376,
      "rewards/forensics_sinkhorn_reward": 0.49174241721630096,
      "rewards/forensics_strict_edge_reward": 0.14565565437078476,
      "rewards/forensics_tiou_reward": 0.2678241729736328,
      "step": 324
    },
    {
      "completion_length": 36.34375,
      "epoch": 0.8655126498002663,
      "grad_norm": 13.76929759979248,
      "kl": 1.95703125,
      "learning_rate": 7.84e-07,
      "loss": 0.1548,
      "reward": 0.6073059737682343,
      "reward_std": 0.17711802572011948,
      "reward_std_all": 0.18162279576063156,
      "rewards-max/forensics_sinkhorn_reward": 0.8505918979644775,
      "rewards-max/forensics_strict_edge_reward": 0.6930834054946899,
      "rewards-max/forensics_tiou_reward": 0.6953283846378326,
      "rewards/forensics_sinkhorn_reward": 0.4334900230169296,
      "rewards/forensics_strict_edge_reward": 0.15028628706932068,
      "rewards/forensics_tiou_reward": 0.2425691783428192,
      "step": 325
    },
    {
      "completion_length": 37.578125,
      "epoch": 0.8681757656458056,
      "grad_norm": 11.660378456115723,
      "kl": 1.5625,
      "learning_rate": 7.833333333333333e-07,
      "loss": 0.25,
      "reward": 0.6094803810119629,
      "reward_std": 0.21211131662130356,
      "reward_std_all": 0.20443026721477509,
      "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": 0.5277030169963837,
      "rewards/forensics_strict_edge_reward": 0.26586297154426575,
      "rewards/forensics_tiou_reward": 0.3996732532978058,
      "step": 326
    },
    {
      "completion_length": 28.03125,
      "epoch": 0.8708388814913449,
      "grad_norm": 14.692164421081543,
      "kl": 1.328125,
      "learning_rate": 7.826666666666666e-07,
      "loss": -0.0453,
      "reward": 0.5596891939640045,
      "reward_std": 0.16619092971086502,
      "reward_std_all": 0.19412270933389664,
      "rewards-max/forensics_sinkhorn_reward": 0.9874920845031738,
      "rewards-max/forensics_strict_edge_reward": 0.9151731729507446,
      "rewards-max/forensics_tiou_reward": 0.9940988421440125,
      "rewards/forensics_sinkhorn_reward": 0.43153059482574463,
      "rewards/forensics_strict_edge_reward": 0.09755625948309898,
      "rewards/forensics_tiou_reward": 0.16397970169782639,
      "step": 327
    },
    {
      "completion_length": 28.09375,
      "epoch": 0.8735019973368842,
      "grad_norm": 10.844564437866211,
      "kl": 1.625,
      "learning_rate": 7.82e-07,
      "loss": 0.1267,
      "reward": 0.6354429125785828,
      "reward_std": 0.13964422047138214,
      "reward_std_all": 0.1497989371418953,
      "rewards-max/forensics_sinkhorn_reward": 0.9872882068157196,
      "rewards-max/forensics_strict_edge_reward": 0.9364311099052429,
      "rewards-max/forensics_tiou_reward": 0.9933082163333893,
      "rewards/forensics_sinkhorn_reward": 0.5099149346351624,
      "rewards/forensics_strict_edge_reward": 0.2558852806687355,
      "rewards/forensics_tiou_reward": 0.3430675119161606,
      "step": 328
    },
    {
      "completion_length": 19.5,
      "epoch": 0.8761651131824234,
      "grad_norm": 16.532543182373047,
      "kl": 1.41796875,
      "learning_rate": 7.813333333333332e-07,
      "loss": 0.152,
      "reward": 0.6270295679569244,
      "reward_std": 0.16265930980443954,
      "reward_std_all": 0.17050965130329132,
      "rewards-max/forensics_sinkhorn_reward": 0.9877539277076721,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9951256513595581,
      "rewards/forensics_sinkhorn_reward": 0.5225020945072174,
      "rewards/forensics_strict_edge_reward": 0.21932536363601685,
      "rewards/forensics_tiou_reward": 0.31643135845661163,
      "step": 329
    },
    {
      "completion_length": 27.6875,
      "epoch": 0.8788282290279628,
      "grad_norm": 22.340286254882812,
      "kl": 2.5703125,
      "learning_rate": 7.806666666666666e-07,
      "loss": 0.2962,
      "reward": 0.6064441800117493,
      "reward_std": 0.14956526458263397,
      "reward_std_all": 0.16695526987314224,
      "rewards-max/forensics_sinkhorn_reward": 0.9630076885223389,
      "rewards-max/forensics_strict_edge_reward": 0.9048932790756226,
      "rewards-max/forensics_tiou_reward": 0.940121203660965,
      "rewards/forensics_sinkhorn_reward": 0.4444987028837204,
      "rewards/forensics_strict_edge_reward": 0.13531027734279633,
      "rewards/forensics_tiou_reward": 0.17566094547510147,
      "step": 330
    },
    {
      "completion_length": 31.859375,
      "epoch": 0.881491344873502,
      "grad_norm": 14.763282775878906,
      "kl": 1.58203125,
      "learning_rate": 7.799999999999999e-07,
      "loss": 0.0069,
      "reward": 0.5811516344547272,
      "reward_std": 0.18044136464595795,
      "reward_std_all": 0.19217245280742645,
      "rewards-max/forensics_sinkhorn_reward": 0.9883315861225128,
      "rewards-max/forensics_strict_edge_reward": 0.9890114367008209,
      "rewards-max/forensics_tiou_reward": 0.9974026083946228,
      "rewards/forensics_sinkhorn_reward": 0.4603531062602997,
      "rewards/forensics_strict_edge_reward": 0.17139273136854172,
      "rewards/forensics_tiou_reward": 0.2577814757823944,
      "step": 331
    },
    {
      "completion_length": 34.0,
      "epoch": 0.8841544607190412,
      "grad_norm": 11.928691864013672,
      "kl": 1.40625,
      "learning_rate": 7.793333333333333e-07,
      "loss": 0.1966,
      "reward": 0.5660203099250793,
      "reward_std": 0.1622229963541031,
      "reward_std_all": 0.19388478249311447,
      "rewards-max/forensics_sinkhorn_reward": 0.9848247468471527,
      "rewards-max/forensics_strict_edge_reward": 0.8719756007194519,
      "rewards-max/forensics_tiou_reward": 0.9882729649543762,
      "rewards/forensics_sinkhorn_reward": 0.4199993908405304,
      "rewards/forensics_strict_edge_reward": 0.12640169262886047,
      "rewards/forensics_tiou_reward": 0.21659813076257706,
      "step": 332
    },
    {
      "completion_length": 30.203125,
      "epoch": 0.8868175765645806,
      "grad_norm": 10.42966365814209,
      "kl": 1.69140625,
      "learning_rate": 7.786666666666665e-07,
      "loss": 0.356,
      "reward": 0.5770660638809204,
      "reward_std": 0.17152512818574905,
      "reward_std_all": 0.1923098787665367,
      "rewards-max/forensics_sinkhorn_reward": 0.9886762201786041,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9987611472606659,
      "rewards/forensics_sinkhorn_reward": 0.42063771188259125,
      "rewards/forensics_strict_edge_reward": 0.18180086463689804,
      "rewards/forensics_tiou_reward": 0.25266363471746445,
      "step": 333
    },
    {
      "completion_length": 29.75,
      "epoch": 0.8894806924101198,
      "grad_norm": 9.822507858276367,
      "kl": 1.28515625,
      "learning_rate": 7.78e-07,
      "loss": 0.0843,
      "reward": 0.6628249883651733,
      "reward_std": 0.16327470541000366,
      "reward_std_all": 0.15402452647686005,
      "rewards-max/forensics_sinkhorn_reward": 0.9887341856956482,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9989897608757019,
      "rewards/forensics_sinkhorn_reward": 0.5401171445846558,
      "rewards/forensics_strict_edge_reward": 0.2507883533835411,
      "rewards/forensics_tiou_reward": 0.33218538761138916,
      "step": 334
    },
    {
      "completion_length": 24.984375,
      "epoch": 0.8921438082556591,
      "grad_norm": 10.269000053405762,
      "kl": 0.875,
      "learning_rate": 7.773333333333333e-07,
      "loss": 0.0809,
      "reward": 0.6146166324615479,
      "reward_std": 0.18310336023569107,
      "reward_std_all": 0.1853436529636383,
      "rewards-max/forensics_sinkhorn_reward": 0.9441607296466827,
      "rewards-max/forensics_strict_edge_reward": 0.7768298983573914,
      "rewards-max/forensics_tiou_reward": 0.9002257287502289,
      "rewards/forensics_sinkhorn_reward": 0.49898384511470795,
      "rewards/forensics_strict_edge_reward": 0.16514596343040466,
      "rewards/forensics_tiou_reward": 0.2575378268957138,
      "step": 335
    },
    {
      "completion_length": 20.046875,
      "epoch": 0.8948069241011984,
      "grad_norm": 11.953224182128906,
      "kl": 1.28125,
      "learning_rate": 7.766666666666666e-07,
      "loss": 0.1463,
      "reward": 0.6325398981571198,
      "reward_std": 0.17050766199827194,
      "reward_std_all": 0.16927333176136017,
      "rewards-max/forensics_sinkhorn_reward": 0.9880152940750122,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9961563944816589,
      "rewards/forensics_sinkhorn_reward": 0.5772543847560883,
      "rewards/forensics_strict_edge_reward": 0.23986175656318665,
      "rewards/forensics_tiou_reward": 0.3396870791912079,
      "step": 336
    },
    {
      "completion_length": 25.625,
      "epoch": 0.8974700399467377,
      "grad_norm": 10.513797760009766,
      "kl": 0.89453125,
      "learning_rate": 7.76e-07,
      "loss": 0.213,
      "reward": 0.6949993968009949,
      "reward_std": 0.1586611494421959,
      "reward_std_all": 0.14291438087821007,
      "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": 0.6186044216156006,
      "rewards/forensics_strict_edge_reward": 0.3169075548648834,
      "rewards/forensics_tiou_reward": 0.5006300508975983,
      "step": 337
    },
    {
      "completion_length": 22.28125,
      "epoch": 0.9001331557922769,
      "grad_norm": 10.79873275756836,
      "kl": 1.54296875,
      "learning_rate": 7.753333333333333e-07,
      "loss": 0.3154,
      "reward": 0.5918120443820953,
      "reward_std": 0.1790970265865326,
      "reward_std_all": 0.19357100874185562,
      "rewards-max/forensics_sinkhorn_reward": 0.8940604031085968,
      "rewards-max/forensics_strict_edge_reward": 0.7141821682453156,
      "rewards-max/forensics_tiou_reward": 0.7960971891880035,
      "rewards/forensics_sinkhorn_reward": 0.4429190307855606,
      "rewards/forensics_strict_edge_reward": 0.152765404433012,
      "rewards/forensics_tiou_reward": 0.23594408482313156,
      "step": 338
    },
    {
      "completion_length": 22.515625,
      "epoch": 0.9027962716378163,
      "grad_norm": 14.463412284851074,
      "kl": 1.19921875,
      "learning_rate": 7.746666666666666e-07,
      "loss": 0.1153,
      "reward": 0.6000052690505981,
      "reward_std": 0.23973800241947174,
      "reward_std_all": 0.2256094440817833,
      "rewards-max/forensics_sinkhorn_reward": 0.9871852993965149,
      "rewards-max/forensics_strict_edge_reward": 0.9355069994926453,
      "rewards-max/forensics_tiou_reward": 0.9928973913192749,
      "rewards/forensics_sinkhorn_reward": 0.5147406905889511,
      "rewards/forensics_strict_edge_reward": 0.2308691069483757,
      "rewards/forensics_tiou_reward": 0.3364398181438446,
      "step": 339
    },
    {
      "completion_length": 21.484375,
      "epoch": 0.9054593874833555,
      "grad_norm": 13.28968620300293,
      "kl": 1.90625,
      "learning_rate": 7.74e-07,
      "loss": 0.2991,
      "reward": 0.5701091885566711,
      "reward_std": 0.20604044198989868,
      "reward_std_all": 0.21329832822084427,
      "rewards-max/forensics_sinkhorn_reward": 0.9882831871509552,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9972139000892639,
      "rewards/forensics_sinkhorn_reward": 0.480665922164917,
      "rewards/forensics_strict_edge_reward": 0.23715578019618988,
      "rewards/forensics_tiou_reward": 0.290096215903759,
      "step": 340
    },
    {
      "completion_length": 17.0625,
      "epoch": 0.9081225033288948,
      "grad_norm": 10.455963134765625,
      "kl": 1.00390625,
      "learning_rate": 7.733333333333333e-07,
      "loss": 0.0637,
      "reward": 0.6486509442329407,
      "reward_std": 0.11378715559840202,
      "reward_std_all": 0.1295805349946022,
      "rewards-max/forensics_sinkhorn_reward": 0.9874802529811859,
      "rewards-max/forensics_strict_edge_reward": 0.947419673204422,
      "rewards-max/forensics_tiou_reward": 0.9940653443336487,
      "rewards/forensics_sinkhorn_reward": 0.5532456934452057,
      "rewards/forensics_strict_edge_reward": 0.2465239018201828,
      "rewards/forensics_tiou_reward": 0.3208405002951622,
      "step": 341
    },
    {
      "completion_length": 21.640625,
      "epoch": 0.9107856191744341,
      "grad_norm": 11.077491760253906,
      "kl": 1.0703125,
      "learning_rate": 7.726666666666666e-07,
      "loss": 0.3277,
      "reward": 0.653517484664917,
      "reward_std": 0.10461386293172836,
      "reward_std_all": 0.12594664841890335,
      "rewards-max/forensics_sinkhorn_reward": 0.9881936013698578,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9968593716621399,
      "rewards/forensics_sinkhorn_reward": 0.5538546442985535,
      "rewards/forensics_strict_edge_reward": 0.26656945794820786,
      "rewards/forensics_tiou_reward": 0.32343707978725433,
      "step": 342
    },
    {
      "completion_length": 19.796875,
      "epoch": 0.9134487350199734,
      "grad_norm": 15.668992042541504,
      "kl": 1.34375,
      "learning_rate": 7.72e-07,
      "loss": 0.2254,
      "reward": 0.6390257477760315,
      "reward_std": 0.11666558310389519,
      "reward_std_all": 0.1382339522242546,
      "rewards-max/forensics_sinkhorn_reward": 0.9877291023731232,
      "rewards-max/forensics_strict_edge_reward": 0.9565287232398987,
      "rewards-max/forensics_tiou_reward": 0.9950343072414398,
      "rewards/forensics_sinkhorn_reward": 0.543705940246582,
      "rewards/forensics_strict_edge_reward": 0.21549968421459198,
      "rewards/forensics_tiou_reward": 0.3327532857656479,
      "step": 343
    },
    {
      "completion_length": 30.21875,
      "epoch": 0.9161118508655126,
      "grad_norm": 14.753140449523926,
      "kl": 1.54296875,
      "learning_rate": 7.713333333333333e-07,
      "loss": 0.2552,
      "reward": 0.6546335220336914,
      "reward_std": 0.1589369848370552,
      "reward_std_all": 0.1573154181241989,
      "rewards-max/forensics_sinkhorn_reward": 0.9876930713653564,
      "rewards-max/forensics_strict_edge_reward": 0.9252271056175232,
      "rewards-max/forensics_tiou_reward": 0.9948891401290894,
      "rewards/forensics_sinkhorn_reward": 0.5536708235740662,
      "rewards/forensics_strict_edge_reward": 0.26147105544805527,
      "rewards/forensics_tiou_reward": 0.37223654985427856,
      "step": 344
    },
    {
      "completion_length": 39.953125,
      "epoch": 0.918774966711052,
      "grad_norm": 12.838464736938477,
      "kl": 2.0546875,
      "learning_rate": 7.706666666666667e-07,
      "loss": 0.2851,
      "reward": 0.581305593252182,
      "reward_std": 0.20289970189332962,
      "reward_std_all": 0.20640034973621368,
      "rewards-max/forensics_sinkhorn_reward": 0.9710777997970581,
      "rewards-max/forensics_strict_edge_reward": 0.8863177299499512,
      "rewards-max/forensics_tiou_reward": 0.9540283977985382,
      "rewards/forensics_sinkhorn_reward": 0.41785863041877747,
      "rewards/forensics_strict_edge_reward": 0.18459269404411316,
      "rewards/forensics_tiou_reward": 0.23300744593143463,
      "step": 345
    },
    {
      "completion_length": 42.109375,
      "epoch": 0.9214380825565912,
      "grad_norm": 11.995157241821289,
      "kl": 1.4765625,
      "learning_rate": 7.699999999999999e-07,
      "loss": 0.2247,
      "reward": 0.6351983547210693,
      "reward_std": 0.1286771111190319,
      "reward_std_all": 0.14847099781036377,
      "rewards-max/forensics_sinkhorn_reward": 0.9529398679733276,
      "rewards-max/forensics_strict_edge_reward": 0.8647556900978088,
      "rewards-max/forensics_tiou_reward": 0.9187125861644745,
      "rewards/forensics_sinkhorn_reward": 0.4780723452568054,
      "rewards/forensics_strict_edge_reward": 0.22522221505641937,
      "rewards/forensics_tiou_reward": 0.272127628326416,
      "step": 346
    },
    {
      "completion_length": 34.8125,
      "epoch": 0.9241011984021305,
      "grad_norm": 14.34437084197998,
      "kl": 2.09765625,
      "learning_rate": 7.693333333333333e-07,
      "loss": 0.2429,
      "reward": 0.6250233352184296,
      "reward_std": 0.17339543998241425,
      "reward_std_all": 0.174826942384243,
      "rewards-max/forensics_sinkhorn_reward": 0.9881556034088135,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9967087805271149,
      "rewards/forensics_sinkhorn_reward": 0.5037883967161179,
      "rewards/forensics_strict_edge_reward": 0.23143485188484192,
      "rewards/forensics_tiou_reward": 0.32827070355415344,
      "step": 347
    },
    {
      "completion_length": 27.203125,
      "epoch": 0.9267643142476698,
      "grad_norm": 12.603832244873047,
      "kl": 1.18359375,
      "learning_rate": 7.686666666666666e-07,
      "loss": 0.225,
      "reward": 0.6450299024581909,
      "reward_std": 0.14400669187307358,
      "reward_std_all": 0.15051186829805374,
      "rewards-max/forensics_sinkhorn_reward": 0.9878050684928894,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.995329737663269,
      "rewards/forensics_sinkhorn_reward": 0.5822740495204926,
      "rewards/forensics_strict_edge_reward": 0.26340483129024506,
      "rewards/forensics_tiou_reward": 0.3594297766685486,
      "step": 348
    },
    {
      "completion_length": 41.625,
      "epoch": 0.929427430093209,
      "grad_norm": 13.049034118652344,
      "kl": 1.6875,
      "learning_rate": 7.68e-07,
      "loss": 0.1167,
      "reward": 0.637482076883316,
      "reward_std": 0.18105753511190414,
      "reward_std_all": 0.17432522773742676,
      "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": 0.5547432601451874,
      "rewards/forensics_strict_edge_reward": 0.31048789620399475,
      "rewards/forensics_tiou_reward": 0.38970285654067993,
      "step": 349
    },
    {
      "completion_length": 36.609375,
      "epoch": 0.9320905459387483,
      "grad_norm": 141.2417449951172,
      "kl": 3.1796875,
      "learning_rate": 7.673333333333332e-07,
      "loss": 0.239,
      "reward": 0.6185206174850464,
      "reward_std": 0.17338596284389496,
      "reward_std_all": 0.17958158999681473,
      "rewards-max/forensics_sinkhorn_reward": 0.9737440347671509,
      "rewards-max/forensics_strict_edge_reward": 0.9256840348243713,
      "rewards-max/forensics_tiou_reward": 0.9635289013385773,
      "rewards/forensics_sinkhorn_reward": 0.5070932656526566,
      "rewards/forensics_strict_edge_reward": 0.20624786615371704,
      "rewards/forensics_tiou_reward": 0.30685707926750183,
      "step": 350
    },
    {
      "completion_length": 36.90625,
      "epoch": 0.9347536617842876,
      "grad_norm": 12.00897216796875,
      "kl": 2.34765625,
      "learning_rate": 7.666666666666667e-07,
      "loss": 0.2737,
      "reward": 0.6352086067199707,
      "reward_std": 0.12430338561534882,
      "reward_std_all": 0.14435850828886032,
      "rewards-max/forensics_sinkhorn_reward": 0.9786385595798492,
      "rewards-max/forensics_strict_edge_reward": 0.8876297473907471,
      "rewards-max/forensics_tiou_reward": 0.9756626486778259,
      "rewards/forensics_sinkhorn_reward": 0.4915127456188202,
      "rewards/forensics_strict_edge_reward": 0.25065475702285767,
      "rewards/forensics_tiou_reward": 0.30764710903167725,
      "step": 351
    },
    {
      "completion_length": 38.0,
      "epoch": 0.9374167776298269,
      "grad_norm": 14.573094367980957,
      "kl": 2.10546875,
      "learning_rate": 7.66e-07,
      "loss": 0.3487,
      "reward": 0.6359173953533173,
      "reward_std": 0.1652594804763794,
      "reward_std_all": 0.16730983555316925,
      "rewards-max/forensics_sinkhorn_reward": 0.9887357950210571,
      "rewards-max/forensics_strict_edge_reward": 0.9890114367008209,
      "rewards-max/forensics_tiou_reward": 0.9989964962005615,
      "rewards/forensics_sinkhorn_reward": 0.5398971438407898,
      "rewards/forensics_strict_edge_reward": 0.2657129839062691,
      "rewards/forensics_tiou_reward": 0.3659540265798569,
      "step": 352
    },
    {
      "completion_length": 37.8125,
      "epoch": 0.9400798934753661,
      "grad_norm": 10.836041450500488,
      "kl": 2.109375,
      "learning_rate": 7.653333333333333e-07,
      "loss": 0.1693,
      "reward": 0.6797682344913483,
      "reward_std": 0.1306898258626461,
      "reward_std_all": 0.13566581532359123,
      "rewards-max/forensics_sinkhorn_reward": 0.9870804250240326,
      "rewards-max/forensics_strict_edge_reward": 0.9464850425720215,
      "rewards-max/forensics_tiou_reward": 0.9924884736537933,
      "rewards/forensics_sinkhorn_reward": 0.6264715790748596,
      "rewards/forensics_strict_edge_reward": 0.35596850514411926,
      "rewards/forensics_tiou_reward": 0.46517089009284973,
      "step": 353
    },
    {
      "completion_length": 37.171875,
      "epoch": 0.9427430093209055,
      "grad_norm": 13.992439270019531,
      "kl": 1.9765625,
      "learning_rate": 7.646666666666667e-07,
      "loss": 0.3682,
      "reward": 0.6431562304496765,
      "reward_std": 0.18005873262882233,
      "reward_std_all": 0.1729484498500824,
      "rewards-max/forensics_sinkhorn_reward": 0.9887276291847229,
      "rewards-max/forensics_strict_edge_reward": 0.9782643616199493,
      "rewards-max/forensics_tiou_reward": 0.9989637434482574,
      "rewards/forensics_sinkhorn_reward": 0.582188069820404,
      "rewards/forensics_strict_edge_reward": 0.37122008204460144,
      "rewards/forensics_tiou_reward": 0.46336375176906586,
      "step": 354
    },
    {
      "completion_length": 33.203125,
      "epoch": 0.9454061251664447,
      "grad_norm": 14.620929718017578,
      "kl": 1.8203125,
      "learning_rate": 7.64e-07,
      "loss": 0.168,
      "reward": 0.6748265624046326,
      "reward_std": 0.14261573553085327,
      "reward_std_all": 0.14202411100268364,
      "rewards-max/forensics_sinkhorn_reward": 0.985675185918808,
      "rewards-max/forensics_strict_edge_reward": 0.8854433596134186,
      "rewards-max/forensics_tiou_reward": 0.9869342744350433,
      "rewards/forensics_sinkhorn_reward": 0.5178486257791519,
      "rewards/forensics_strict_edge_reward": 0.2735297977924347,
      "rewards/forensics_tiou_reward": 0.33420516550540924,
      "step": 355
    },
    {
      "completion_length": 47.46875,
      "epoch": 0.948069241011984,
      "grad_norm": 12.038981437683105,
      "kl": 1.9296875,
      "learning_rate": 7.633333333333333e-07,
      "loss": 0.1732,
      "reward": 0.6379469633102417,
      "reward_std": 0.17607014626264572,
      "reward_std_all": 0.17202632129192352,
      "rewards-max/forensics_sinkhorn_reward": 0.9798571467399597,
      "rewards-max/forensics_strict_edge_reward": 0.9075756669044495,
      "rewards-max/forensics_tiou_reward": 0.977696031332016,
      "rewards/forensics_sinkhorn_reward": 0.5155978351831436,
      "rewards/forensics_strict_edge_reward": 0.22258059680461884,
      "rewards/forensics_tiou_reward": 0.3030250146985054,
      "step": 356
    },
    {
      "completion_length": 21.0,
      "epoch": 0.9507323568575233,
      "grad_norm": 16.234298706054688,
      "kl": 1.40234375,
      "learning_rate": 7.626666666666667e-07,
      "loss": -0.049,
      "reward": 0.6333318650722504,
      "reward_std": 0.09295906126499176,
      "reward_std_all": 0.134429469704628,
      "rewards-max/forensics_sinkhorn_reward": 0.9867441654205322,
      "rewards-max/forensics_strict_edge_reward": 0.9048932790756226,
      "rewards-max/forensics_tiou_reward": 0.9911676645278931,
      "rewards/forensics_sinkhorn_reward": 0.5055013597011566,
      "rewards/forensics_strict_edge_reward": 0.15323954820632935,
      "rewards/forensics_tiou_reward": 0.21099799871444702,
      "step": 357
    },
    {
      "completion_length": 42.25,
      "epoch": 0.9533954727030626,
      "grad_norm": 14.884013175964355,
      "kl": 2.203125,
      "learning_rate": 7.62e-07,
      "loss": 0.3579,
      "reward": 0.6420764327049255,
      "reward_std": 0.1376148834824562,
      "reward_std_all": 0.1473623365163803,
      "rewards-max/forensics_sinkhorn_reward": 0.9880093634128571,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9961349964141846,
      "rewards/forensics_sinkhorn_reward": 0.542820617556572,
      "rewards/forensics_strict_edge_reward": 0.245211161673069,
      "rewards/forensics_tiou_reward": 0.3297993093729019,
      "step": 358
    },
    {
      "completion_length": 41.15625,
      "epoch": 0.9560585885486018,
      "grad_norm": 12.995977401733398,
      "kl": 2.1875,
      "learning_rate": 7.613333333333333e-07,
      "loss": 0.2554,
      "reward": 0.5724403262138367,
      "reward_std": 0.14851826429367065,
      "reward_std_all": 0.19031479209661484,
      "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": 0.45129184424877167,
      "rewards/forensics_strict_edge_reward": 0.2414257600903511,
      "rewards/forensics_tiou_reward": 0.319929800927639,
      "step": 359
    },
    {
      "completion_length": 24.546875,
      "epoch": 0.9587217043941412,
      "grad_norm": 13.037928581237793,
      "kl": 1.90234375,
      "learning_rate": 7.606666666666667e-07,
      "loss": 0.0592,
      "reward": 0.6840503215789795,
      "reward_std": 0.08383957669138908,
      "reward_std_all": 0.10162556543946266,
      "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": 0.6042293608188629,
      "rewards/forensics_strict_edge_reward": 0.3400847017765045,
      "rewards/forensics_tiou_reward": 0.4262760877609253,
      "step": 360
    },
    {
      "completion_length": 47.609375,
      "epoch": 0.9613848202396804,
      "grad_norm": 16.575641632080078,
      "kl": 2.6875,
      "learning_rate": 7.599999999999999e-07,
      "loss": 0.2405,
      "reward": 0.5837037265300751,
      "reward_std": 0.16420573741197586,
      "reward_std_all": 0.18530388176441193,
      "rewards-max/forensics_sinkhorn_reward": 0.9870399832725525,
      "rewards-max/forensics_strict_edge_reward": 0.946017861366272,
      "rewards-max/forensics_tiou_reward": 0.9923265874385834,
      "rewards/forensics_sinkhorn_reward": 0.44097912311553955,
      "rewards/forensics_strict_edge_reward": 0.20219416171312332,
      "rewards/forensics_tiou_reward": 0.2922985851764679,
      "step": 361
    },
    {
      "completion_length": 18.875,
      "epoch": 0.9640479360852197,
      "grad_norm": 14.568896293640137,
      "kl": 1.76953125,
      "learning_rate": 7.593333333333333e-07,
      "loss": 0.2534,
      "reward": 0.6389500200748444,
      "reward_std": 0.14289550483226776,
      "reward_std_all": 0.15709729492664337,
      "rewards-max/forensics_sinkhorn_reward": 0.9882118403911591,
      "rewards-max/forensics_strict_edge_reward": 0.9357379674911499,
      "rewards-max/forensics_tiou_reward": 0.996931403875351,
      "rewards/forensics_sinkhorn_reward": 0.5749056339263916,
      "rewards/forensics_strict_edge_reward": 0.24750924110412598,
      "rewards/forensics_tiou_reward": 0.35564881563186646,
      "step": 362
    },
    {
      "completion_length": 30.203125,
      "epoch": 0.966711051930759,
      "grad_norm": 13.10369873046875,
      "kl": 2.25,
      "learning_rate": 7.586666666666666e-07,
      "loss": 0.1101,
      "reward": 0.6243441700935364,
      "reward_std": 0.15991339832544327,
      "reward_std_all": 0.16538641601800919,
      "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": 0.49927370250225067,
      "rewards/forensics_strict_edge_reward": 0.19990836828947067,
      "rewards/forensics_tiou_reward": 0.27574022114276886,
      "step": 363
    },
    {
      "completion_length": 17.140625,
      "epoch": 0.9693741677762983,
      "grad_norm": 14.131290435791016,
      "kl": 2.265625,
      "learning_rate": 7.58e-07,
      "loss": 0.2959,
      "reward": 0.670084685087204,
      "reward_std": 0.09387709014117718,
      "reward_std_all": 0.11060265824198723,
      "rewards-max/forensics_sinkhorn_reward": 0.986841470003128,
      "rewards-max/forensics_strict_edge_reward": 0.9464850425720215,
      "rewards-max/forensics_tiou_reward": 0.9915540516376495,
      "rewards/forensics_sinkhorn_reward": 0.5472387373447418,
      "rewards/forensics_strict_edge_reward": 0.2842964828014374,
      "rewards/forensics_tiou_reward": 0.3267725110054016,
      "step": 364
    },
    {
      "completion_length": 23.890625,
      "epoch": 0.9720372836218375,
      "grad_norm": 14.288840293884277,
      "kl": 1.107421875,
      "learning_rate": 7.573333333333332e-07,
      "loss": 0.0998,
      "reward": 0.6915948987007141,
      "reward_std": 0.13370833545923233,
      "reward_std_all": 0.12543300539255142,
      "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": 0.6197196543216705,
      "rewards/forensics_strict_edge_reward": 0.32817189395427704,
      "rewards/forensics_tiou_reward": 0.39834432303905487,
      "step": 365
    },
    {
      "completion_length": 26.078125,
      "epoch": 0.9747003994673769,
      "grad_norm": 12.98089599609375,
      "kl": 1.41015625,
      "learning_rate": 7.566666666666667e-07,
      "loss": 0.0489,
      "reward": 0.6500694155693054,
      "reward_std": 0.15772487223148346,
      "reward_std_all": 0.16120609641075134,
      "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": 0.5783224105834961,
      "rewards/forensics_strict_edge_reward": 0.2650262862443924,
      "rewards/forensics_tiou_reward": 0.3776034265756607,
      "step": 366
    },
    {
      "completion_length": 23.96875,
      "epoch": 0.9773635153129161,
      "grad_norm": 14.620509147644043,
      "kl": 1.56640625,
      "learning_rate": 7.559999999999999e-07,
      "loss": 0.1594,
      "reward": 0.6234042346477509,
      "reward_std": 0.16361700743436813,
      "reward_std_all": 0.16931183636188507,
      "rewards-max/forensics_sinkhorn_reward": 0.9884293377399445,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9977881014347076,
      "rewards/forensics_sinkhorn_reward": 0.4973060339689255,
      "rewards/forensics_strict_edge_reward": 0.19931291043758392,
      "rewards/forensics_tiou_reward": 0.27433155477046967,
      "step": 367
    },
    {
      "completion_length": 23.03125,
      "epoch": 0.9800266311584553,
      "grad_norm": 16.819515228271484,
      "kl": 1.8359375,
      "learning_rate": 7.553333333333333e-07,
      "loss": 0.1465,
      "reward": 0.6284303367137909,
      "reward_std": 0.17617883533239365,
      "reward_std_all": 0.17722748965024948,
      "rewards-max/forensics_sinkhorn_reward": 0.9882227778434753,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9969750642776489,
      "rewards/forensics_sinkhorn_reward": 0.4823378175497055,
      "rewards/forensics_strict_edge_reward": 0.23069187253713608,
      "rewards/forensics_tiou_reward": 0.3297414183616638,
      "step": 368
    },
    {
      "completion_length": 19.109375,
      "epoch": 0.9826897470039947,
      "grad_norm": 13.03810977935791,
      "kl": 1.234375,
      "learning_rate": 7.546666666666666e-07,
      "loss": 0.1607,
      "reward": 0.6135475039482117,
      "reward_std": 0.18075992912054062,
      "reward_std_all": 0.18917487561702728,
      "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": 0.5517973005771637,
      "rewards/forensics_strict_edge_reward": 0.2688656449317932,
      "rewards/forensics_tiou_reward": 0.33933931589126587,
      "step": 369
    },
    {
      "completion_length": 38.34375,
      "epoch": 0.9853528628495339,
      "grad_norm": 9.2352294921875,
      "kl": 1.154296875,
      "learning_rate": 7.54e-07,
      "loss": 0.1266,
      "reward": 0.6361438930034637,
      "reward_std": 0.13771577179431915,
      "reward_std_all": 0.1527446210384369,
      "rewards-max/forensics_sinkhorn_reward": 0.9511792361736298,
      "rewards-max/forensics_strict_edge_reward": 0.878633052110672,
      "rewards-max/forensics_tiou_reward": 0.9165467321872711,
      "rewards/forensics_sinkhorn_reward": 0.5110063254833221,
      "rewards/forensics_strict_edge_reward": 0.2187589779496193,
      "rewards/forensics_tiou_reward": 0.3006128966808319,
      "step": 370
    },
    {
      "completion_length": 19.9375,
      "epoch": 0.9880159786950732,
      "grad_norm": 14.312871932983398,
      "kl": 1.5625,
      "learning_rate": 7.533333333333332e-07,
      "loss": 0.0368,
      "reward": 0.6606544256210327,
      "reward_std": 0.11053097993135452,
      "reward_std_all": 0.12494595721364021,
      "rewards-max/forensics_sinkhorn_reward": 0.9641591906547546,
      "rewards-max/forensics_strict_edge_reward": 0.7901110649108887,
      "rewards-max/forensics_tiou_reward": 0.9363806247711182,
      "rewards/forensics_sinkhorn_reward": 0.48512521386146545,
      "rewards/forensics_strict_edge_reward": 0.19403277337551117,
      "rewards/forensics_tiou_reward": 0.2736350893974304,
      "step": 371
    },
    {
      "completion_length": 21.484375,
      "epoch": 0.9906790945406125,
      "grad_norm": 16.874168395996094,
      "kl": 1.75,
      "learning_rate": 7.526666666666667e-07,
      "loss": 0.0713,
      "reward": 0.5610812902450562,
      "reward_std": 0.18544866889715195,
      "reward_std_all": 0.208399198949337,
      "rewards-max/forensics_sinkhorn_reward": 0.9155183732509613,
      "rewards-max/forensics_strict_edge_reward": 0.4818517863750458,
      "rewards-max/forensics_tiou_reward": 0.8351747691631317,
      "rewards/forensics_sinkhorn_reward": 0.42819325625896454,
      "rewards/forensics_strict_edge_reward": 0.08133745566010475,
      "rewards/forensics_tiou_reward": 0.16914964467287064,
      "step": 372
    },
    {
      "completion_length": 16.171875,
      "epoch": 0.9933422103861518,
      "grad_norm": 14.27674674987793,
      "kl": 1.203125,
      "learning_rate": 7.52e-07,
      "loss": 0.2261,
      "reward": 0.6663000285625458,
      "reward_std": 0.11152268946170807,
      "reward_std_all": 0.12719853222370148,
      "rewards-max/forensics_sinkhorn_reward": 0.9888186454772949,
      "rewards-max/forensics_strict_edge_reward": 0.9890114367008209,
      "rewards-max/forensics_tiou_reward": 0.9993227124214172,
      "rewards/forensics_sinkhorn_reward": 0.5883159041404724,
      "rewards/forensics_strict_edge_reward": 0.3106555640697479,
      "rewards/forensics_tiou_reward": 0.33735403418540955,
      "step": 373
    },
    {
      "completion_length": 27.40625,
      "epoch": 0.996005326231691,
      "grad_norm": 18.469324111938477,
      "kl": 2.69140625,
      "learning_rate": 7.513333333333333e-07,
      "loss": 0.1519,
      "reward": 0.5832435190677643,
      "reward_std": 0.16414517164230347,
      "reward_std_all": 0.18878942728042603,
      "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": 0.4747259020805359,
      "rewards/forensics_strict_edge_reward": 0.1406198963522911,
      "rewards/forensics_tiou_reward": 0.23429807275533676,
      "step": 374
    },
    {
      "completion_length": 37.0625,
      "epoch": 0.9986684420772304,
      "grad_norm": 12.539093017578125,
      "kl": 1.96484375,
      "learning_rate": 7.506666666666667e-07,
      "loss": 0.3846,
      "reward": 0.6155391931533813,
      "reward_std": 0.1591990813612938,
      "reward_std_all": 0.17113424837589264,
      "rewards-max/forensics_sinkhorn_reward": 0.9802211821079254,
      "rewards-max/forensics_strict_edge_reward": 0.9158510267734528,
      "rewards-max/forensics_tiou_reward": 0.9786399602890015,
      "rewards/forensics_sinkhorn_reward": 0.4901476502418518,
      "rewards/forensics_strict_edge_reward": 0.2789943739771843,
      "rewards/forensics_tiou_reward": 0.3718404099345207,
      "step": 375
    },
    {
      "completion_length": 14.0,
      "epoch": 1.0,
      "grad_norm": 9.60606575012207,
      "kl": 3.359375,
      "learning_rate": 7.5e-07,
      "loss": 0.0271,
      "reward": 0.275884747505188,
      "reward_std": 0.011409241706132889,
      "reward_std_all": 0.2725812792778015,
      "rewards-max/forensics_sinkhorn_reward": 0.15784211456775665,
      "rewards-max/forensics_strict_edge_reward": 8.738252788020873e-09,
      "rewards-max/forensics_tiou_reward": 0.0,
      "rewards/forensics_sinkhorn_reward": 0.15784211456775665,
      "rewards/forensics_strict_edge_reward": 8.738252788020873e-09,
      "rewards/forensics_tiou_reward": 0.0,
      "step": 376
    },
    {
      "completion_length": 24.640625,
      "epoch": 1.0026631158455392,
      "grad_norm": 13.659440994262695,
      "kl": 1.33984375,
      "learning_rate": 7.493333333333333e-07,
      "loss": 0.1297,
      "reward": 0.6194565296173096,
      "reward_std": 0.14207863807678223,
      "reward_std_all": 0.15693356096744537,
      "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": 0.5375341773033142,
      "rewards/forensics_strict_edge_reward": 0.24978545308113098,
      "rewards/forensics_tiou_reward": 0.2939014583826065,
      "step": 377
    },
    {
      "completion_length": 45.15625,
      "epoch": 1.0053262316910785,
      "grad_norm": 11.522696495056152,
      "kl": 2.15625,
      "learning_rate": 7.486666666666666e-07,
      "loss": 0.2017,
      "reward": 0.633982926607132,
      "reward_std": 0.1676870733499527,
      "reward_std_all": 0.16550111025571823,
      "rewards-max/forensics_sinkhorn_reward": 0.9883557260036469,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9974982738494873,
      "rewards/forensics_sinkhorn_reward": 0.5012494325637817,
      "rewards/forensics_strict_edge_reward": 0.24030259251594543,
      "rewards/forensics_tiou_reward": 0.34584926068782806,
      "step": 378
    },
    {
      "completion_length": 28.4375,
      "epoch": 1.007989347536618,
      "grad_norm": 13.158669471740723,
      "kl": 1.6875,
      "learning_rate": 7.48e-07,
      "loss": 0.1905,
      "reward": 0.6751351356506348,
      "reward_std": 0.10728530213236809,
      "reward_std_all": 0.11422979086637497,
      "rewards-max/forensics_sinkhorn_reward": 0.9887206256389618,
      "rewards-max/forensics_strict_edge_reward": 0.9782643616199493,
      "rewards-max/forensics_tiou_reward": 0.9989361763000488,
      "rewards/forensics_sinkhorn_reward": 0.5250541269779205,
      "rewards/forensics_strict_edge_reward": 0.19370216131210327,
      "rewards/forensics_tiou_reward": 0.26697391271591187,
      "step": 379
    },
    {
      "completion_length": 50.46875,
      "epoch": 1.0106524633821572,
      "grad_norm": 13.66200065612793,
      "kl": 1.41796875,
      "learning_rate": 7.473333333333332e-07,
      "loss": 0.2338,
      "reward": 0.6211765110492706,
      "reward_std": 0.19097796827554703,
      "reward_std_all": 0.19183041155338287,
      "rewards-max/forensics_sinkhorn_reward": 0.9879695773124695,
      "rewards-max/forensics_strict_edge_reward": 0.9464850425720215,
      "rewards-max/forensics_tiou_reward": 0.9959795773029327,
      "rewards/forensics_sinkhorn_reward": 0.5411514937877655,
      "rewards/forensics_strict_edge_reward": 0.25583742558956146,
      "rewards/forensics_tiou_reward": 0.3612328916788101,
      "step": 380
    },
    {
      "completion_length": 60.5625,
      "epoch": 1.0133155792276964,
      "grad_norm": 13.714598655700684,
      "kl": 2.65234375,
      "learning_rate": 7.466666666666667e-07,
      "loss": 0.3234,
      "reward": 0.5537661612033844,
      "reward_std": 0.18413280695676804,
      "reward_std_all": 0.21221456676721573,
      "rewards-max/forensics_sinkhorn_reward": 0.9881341457366943,
      "rewards-max/forensics_strict_edge_reward": 0.9567649364471436,
      "rewards-max/forensics_tiou_reward": 0.9966257512569427,
      "rewards/forensics_sinkhorn_reward": 0.4466826617717743,
      "rewards/forensics_strict_edge_reward": 0.2575877830386162,
      "rewards/forensics_tiou_reward": 0.3251955509185791,
      "step": 381
    },
    {
      "completion_length": 38.109375,
      "epoch": 1.0159786950732357,
      "grad_norm": 16.685497283935547,
      "kl": 1.7890625,
      "learning_rate": 7.459999999999999e-07,
      "loss": 0.0859,
      "reward": 0.6181921064853668,
      "reward_std": 0.16372597217559814,
      "reward_std_all": 0.1753498911857605,
      "rewards-max/forensics_sinkhorn_reward": 0.9870048463344574,
      "rewards-max/forensics_strict_edge_reward": 0.9565287232398987,
      "rewards-max/forensics_tiou_reward": 0.9921732544898987,
      "rewards/forensics_sinkhorn_reward": 0.5291153490543365,
      "rewards/forensics_strict_edge_reward": 0.1834724210202694,
      "rewards/forensics_tiou_reward": 0.30911944806575775,
      "step": 382
    },
    {
      "completion_length": 41.875,
      "epoch": 1.018641810918775,
      "grad_norm": 16.710002899169922,
      "kl": 2.45703125,
      "learning_rate": 7.453333333333333e-07,
      "loss": 0.1091,
      "reward": 0.5498314201831818,
      "reward_std": 0.1329837404191494,
      "reward_std_all": 0.18345024436712265,
      "rewards-max/forensics_sinkhorn_reward": 0.9831215441226959,
      "rewards-max/forensics_strict_edge_reward": 0.8876297473907471,
      "rewards-max/forensics_tiou_reward": 0.9770902395248413,
      "rewards/forensics_sinkhorn_reward": 0.4224015176296234,
      "rewards/forensics_strict_edge_reward": 0.14526336640119553,
      "rewards/forensics_tiou_reward": 0.22090430557727814,
      "step": 383
    },
    {
      "completion_length": 38.0,
      "epoch": 1.0213049267643142,
      "grad_norm": 13.936138153076172,
      "kl": 1.333984375,
      "learning_rate": 7.446666666666666e-07,
      "loss": 0.149,
      "reward": 0.6843096613883972,
      "reward_std": 0.10467086732387543,
      "reward_std_all": 0.11342974007129669,
      "rewards-max/forensics_sinkhorn_reward": 0.9832706451416016,
      "rewards-max/forensics_strict_edge_reward": 0.9062341153621674,
      "rewards-max/forensics_tiou_reward": 0.9777075946331024,
      "rewards/forensics_sinkhorn_reward": 0.6216910779476166,
      "rewards/forensics_strict_edge_reward": 0.41993725299835205,
      "rewards/forensics_tiou_reward": 0.5070509314537048,
      "step": 384
    },
    {
      "completion_length": 39.015625,
      "epoch": 1.0239680426098536,
      "grad_norm": 19.021739959716797,
      "kl": 1.83203125,
      "learning_rate": 7.44e-07,
      "loss": 0.1869,
      "reward": 0.6099677681922913,
      "reward_std": 0.148708313703537,
      "reward_std_all": 0.1660350039601326,
      "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": 0.508239358663559,
      "rewards/forensics_strict_edge_reward": 0.2686380073428154,
      "rewards/forensics_tiou_reward": 0.30851565301418304,
      "step": 385
    },
    {
      "completion_length": 34.796875,
      "epoch": 1.0266311584553929,
      "grad_norm": 66.38172912597656,
      "kl": 1.802734375,
      "learning_rate": 7.433333333333332e-07,
      "loss": 0.0834,
      "reward": 0.5982154607772827,
      "reward_std": 0.14091742783784866,
      "reward_std_all": 0.1698971688747406,
      "rewards-max/forensics_sinkhorn_reward": 0.9885906875133514,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9984238743782043,
      "rewards/forensics_sinkhorn_reward": 0.5263531059026718,
      "rewards/forensics_strict_edge_reward": 0.22819878160953522,
      "rewards/forensics_tiou_reward": 0.3008638024330139,
      "step": 386
    },
    {
      "completion_length": 40.109375,
      "epoch": 1.0292942743009321,
      "grad_norm": 12.750140190124512,
      "kl": 2.07421875,
      "learning_rate": 7.426666666666667e-07,
      "loss": 0.1981,
      "reward": 0.6100027859210968,
      "reward_std": 0.17797886580228806,
      "reward_std_all": 0.18445567041635513,
      "rewards-max/forensics_sinkhorn_reward": 0.962865948677063,
      "rewards-max/forensics_strict_edge_reward": 0.9265981018543243,
      "rewards-max/forensics_tiou_reward": 0.9419319331645966,
      "rewards/forensics_sinkhorn_reward": 0.48443226516246796,
      "rewards/forensics_strict_edge_reward": 0.2190898135304451,
      "rewards/forensics_tiou_reward": 0.2817985564470291,
      "step": 387
    },
    {
      "completion_length": 40.109375,
      "epoch": 1.0319573901464714,
      "grad_norm": 10.006753921508789,
      "kl": 1.66796875,
      "learning_rate": 7.42e-07,
      "loss": 0.2619,
      "reward": 0.623804360628128,
      "reward_std": 0.13788265734910965,
      "reward_std_all": 0.1623525694012642,
      "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": 0.5380292534828186,
      "rewards/forensics_strict_edge_reward": 0.30788809061050415,
      "rewards/forensics_tiou_reward": 0.37074458599090576,
      "step": 388
    },
    {
      "completion_length": 50.75,
      "epoch": 1.0346205059920106,
      "grad_norm": 9.789613723754883,
      "kl": 1.72265625,
      "learning_rate": 7.413333333333333e-07,
      "loss": 0.1751,
      "reward": 0.6095360219478607,
      "reward_std": 0.1662122644484043,
      "reward_std_all": 0.17737559974193573,
      "rewards-max/forensics_sinkhorn_reward": 0.948736310005188,
      "rewards-max/forensics_strict_edge_reward": 0.8241714537143707,
      "rewards-max/forensics_tiou_reward": 0.9074342548847198,
      "rewards/forensics_sinkhorn_reward": 0.4511258453130722,
      "rewards/forensics_strict_edge_reward": 0.23948287963867188,
      "rewards/forensics_tiou_reward": 0.270640604197979,
      "step": 389
    },
    {
      "completion_length": 26.46875,
      "epoch": 1.0372836218375499,
      "grad_norm": 10.93973159790039,
      "kl": 1.40234375,
      "learning_rate": 7.406666666666667e-07,
      "loss": -0.0006,
      "reward": 0.671053022146225,
      "reward_std": 0.07461097650229931,
      "reward_std_all": 0.1038702093064785,
      "rewards-max/forensics_sinkhorn_reward": 0.9880475103855133,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.99628546833992,
      "rewards/forensics_sinkhorn_reward": 0.6219115555286407,
      "rewards/forensics_strict_edge_reward": 0.3141191154718399,
      "rewards/forensics_tiou_reward": 0.4171757996082306,
      "step": 390
    },
    {
      "completion_length": 44.40625,
      "epoch": 1.0399467376830893,
      "grad_norm": 10.766021728515625,
      "kl": 1.41015625,
      "learning_rate": 7.4e-07,
      "loss": 0.0127,
      "reward": 0.6352200508117676,
      "reward_std": 0.20625723153352737,
      "reward_std_all": 0.1957220807671547,
      "rewards-max/forensics_sinkhorn_reward": 0.9886920750141144,
      "rewards-max/forensics_strict_edge_reward": 0.9890114367008209,
      "rewards-max/forensics_tiou_reward": 0.9988235235214233,
      "rewards/forensics_sinkhorn_reward": 0.556361198425293,
      "rewards/forensics_strict_edge_reward": 0.29341644048690796,
      "rewards/forensics_tiou_reward": 0.37782493233680725,
      "step": 391
    },
    {
      "completion_length": 38.109375,
      "epoch": 1.0426098535286286,
      "grad_norm": 67.58948516845703,
      "kl": 3.0625,
      "learning_rate": 7.393333333333333e-07,
      "loss": 0.2642,
      "reward": 0.6746686697006226,
      "reward_std": 0.14069080352783203,
      "reward_std_all": 0.13521941751241684,
      "rewards-max/forensics_sinkhorn_reward": 0.8801115155220032,
      "rewards-max/forensics_strict_edge_reward": 0.7567555606365204,
      "rewards-max/forensics_tiou_reward": 0.7729608714580536,
      "rewards/forensics_sinkhorn_reward": 0.5831291973590851,
      "rewards/forensics_strict_edge_reward": 0.3169972077012062,
      "rewards/forensics_tiou_reward": 0.3972482904791832,
      "step": 392
    },
    {
      "completion_length": 48.203125,
      "epoch": 1.0452729693741678,
      "grad_norm": 12.550647735595703,
      "kl": 1.23046875,
      "learning_rate": 7.386666666666666e-07,
      "loss": 0.2067,
      "reward": 0.612714946269989,
      "reward_std": 0.16216115653514862,
      "reward_std_all": 0.1734369769692421,
      "rewards-max/forensics_sinkhorn_reward": 0.988852322101593,
      "rewards-max/forensics_strict_edge_reward": 0.9890114367008209,
      "rewards-max/forensics_tiou_reward": 0.9994554817676544,
      "rewards/forensics_sinkhorn_reward": 0.5223832875490189,
      "rewards/forensics_strict_edge_reward": 0.2238856852054596,
      "rewards/forensics_tiou_reward": 0.3071898818016052,
      "step": 393
    },
    {
      "completion_length": 48.5625,
      "epoch": 1.047936085219707,
      "grad_norm": 12.129658699035645,
      "kl": 2.703125,
      "learning_rate": 7.38e-07,
      "loss": 0.2248,
      "reward": 0.5905730724334717,
      "reward_std": 0.12941927835345268,
      "reward_std_all": 0.1664983034133911,
      "rewards-max/forensics_sinkhorn_reward": 0.953524649143219,
      "rewards-max/forensics_strict_edge_reward": 0.756708562374115,
      "rewards-max/forensics_tiou_reward": 0.9234206974506378,
      "rewards/forensics_sinkhorn_reward": 0.44141335785388947,
      "rewards/forensics_strict_edge_reward": 0.18031621724367142,
      "rewards/forensics_tiou_reward": 0.25447704643011093,
      "step": 394
    },
    {
      "completion_length": 29.484375,
      "epoch": 1.0505992010652463,
      "grad_norm": 10.502400398254395,
      "kl": 1.19921875,
      "learning_rate": 7.373333333333332e-07,
      "loss": 0.2171,
      "reward": 0.6607470512390137,
      "reward_std": 0.13039285689592361,
      "reward_std_all": 0.13740822672843933,
      "rewards-max/forensics_sinkhorn_reward": 0.9865042865276337,
      "rewards-max/forensics_strict_edge_reward": 0.8445805907249451,
      "rewards-max/forensics_tiou_reward": 0.9902160465717316,
      "rewards/forensics_sinkhorn_reward": 0.538292407989502,
      "rewards/forensics_strict_edge_reward": 0.19287541508674622,
      "rewards/forensics_tiou_reward": 0.2981930673122406,
      "step": 395
    },
    {
      "completion_length": 42.671875,
      "epoch": 1.0532623169107855,
      "grad_norm": 11.92680549621582,
      "kl": 1.69921875,
      "learning_rate": 7.366666666666667e-07,
      "loss": 0.1868,
      "reward": 0.5709348618984222,
      "reward_std": 0.21179022639989853,
      "reward_std_all": 0.21865951269865036,
      "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": 0.4665306508541107,
      "rewards/forensics_strict_edge_reward": 0.23292004317045212,
      "rewards/forensics_tiou_reward": 0.2856416404247284,
      "step": 396
    },
    {
      "completion_length": 37.015625,
      "epoch": 1.055925432756325,
      "grad_norm": 14.557930946350098,
      "kl": 1.11328125,
      "learning_rate": 7.359999999999999e-07,
      "loss": 0.1733,
      "reward": 0.6001582145690918,
      "reward_std": 0.17845922708511353,
      "reward_std_all": 0.18601271510124207,
      "rewards-max/forensics_sinkhorn_reward": 0.9807189106941223,
      "rewards-max/forensics_strict_edge_reward": 0.7919384241104126,
      "rewards-max/forensics_tiou_reward": 0.9677550196647644,
      "rewards/forensics_sinkhorn_reward": 0.47147679328918457,
      "rewards/forensics_strict_edge_reward": 0.22729824483394623,
      "rewards/forensics_tiou_reward": 0.295203298330307,
      "step": 397
    },
    {
      "completion_length": 46.609375,
      "epoch": 1.0585885486018642,
      "grad_norm": 11.586661338806152,
      "kl": 1.33984375,
      "learning_rate": 7.353333333333333e-07,
      "loss": 0.2792,
      "reward": 0.5889334380626678,
      "reward_std": 0.17550025135278702,
      "reward_std_all": 0.1906842142343521,
      "rewards-max/forensics_sinkhorn_reward": 0.9883341193199158,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.997413694858551,
      "rewards/forensics_sinkhorn_reward": 0.5086580216884613,
      "rewards/forensics_strict_edge_reward": 0.2475077509880066,
      "rewards/forensics_tiou_reward": 0.34642714262008667,
      "step": 398
    },
    {
      "completion_length": 32.5625,
      "epoch": 1.0612516644474035,
      "grad_norm": 14.34847640991211,
      "kl": 1.2421875,
      "learning_rate": 7.346666666666666e-07,
      "loss": 0.1609,
      "reward": 0.6537268161773682,
      "reward_std": 0.12969976663589478,
      "reward_std_all": 0.13518142700195312,
      "rewards-max/forensics_sinkhorn_reward": 0.9885596334934235,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9983020126819611,
      "rewards/forensics_sinkhorn_reward": 0.6167830228805542,
      "rewards/forensics_strict_edge_reward": 0.3904983848333359,
      "rewards/forensics_tiou_reward": 0.4782448709011078,
      "step": 399
    },
    {
      "completion_length": 31.15625,
      "epoch": 1.0639147802929427,
      "grad_norm": 13.674090385437012,
      "kl": 1.3046875,
      "learning_rate": 7.34e-07,
      "loss": 0.1973,
      "reward": 0.6506361365318298,
      "reward_std": 0.1783541738986969,
      "reward_std_all": 0.1710706278681755,
      "rewards-max/forensics_sinkhorn_reward": 0.9883758425712585,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9975776672363281,
      "rewards/forensics_sinkhorn_reward": 0.5530337989330292,
      "rewards/forensics_strict_edge_reward": 0.24933501332998276,
      "rewards/forensics_tiou_reward": 0.373421385884285,
      "step": 400
    },
    {
      "completion_length": 46.921875,
      "epoch": 1.066577896138482,
      "grad_norm": 12.693418502807617,
      "kl": 2.0078125,
      "learning_rate": 7.333333333333332e-07,
      "loss": 0.2869,
      "reward": 0.5732671916484833,
      "reward_std": 0.1954464241862297,
      "reward_std_all": 0.2020634040236473,
      "rewards-max/forensics_sinkhorn_reward": 0.9881614446640015,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9967349469661713,
      "rewards/forensics_sinkhorn_reward": 0.44987401366233826,
      "rewards/forensics_strict_edge_reward": 0.252123698592186,
      "rewards/forensics_tiou_reward": 0.32098984718322754,
      "step": 401
    },
    {
      "completion_length": 38.828125,
      "epoch": 1.0692410119840212,
      "grad_norm": 13.160168647766113,
      "kl": 2.125,
      "learning_rate": 7.326666666666667e-07,
      "loss": 0.3451,
      "reward": 0.6404977440834045,
      "reward_std": 0.11097854375839233,
      "reward_std_all": 0.13552890717983246,
      "rewards-max/forensics_sinkhorn_reward": 0.9878295958042145,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9954413771629333,
      "rewards/forensics_sinkhorn_reward": 0.5661293864250183,
      "rewards/forensics_strict_edge_reward": 0.3368643522262573,
      "rewards/forensics_tiou_reward": 0.4183039218187332,
      "step": 402
    },
    {
      "completion_length": 50.890625,
      "epoch": 1.0719041278295607,
      "grad_norm": 13.86429214477539,
      "kl": 1.75,
      "learning_rate": 7.319999999999999e-07,
      "loss": 0.2638,
      "reward": 0.5957625806331635,
      "reward_std": 0.21495157480239868,
      "reward_std_all": 0.21058788895606995,
      "rewards-max/forensics_sinkhorn_reward": 0.9872143268585205,
      "rewards-max/forensics_strict_edge_reward": 0.9464850425720215,
      "rewards-max/forensics_tiou_reward": 0.9930148720741272,
      "rewards/forensics_sinkhorn_reward": 0.4763620048761368,
      "rewards/forensics_strict_edge_reward": 0.27636049687862396,
      "rewards/forensics_tiou_reward": 0.3267456293106079,
      "step": 403
    },
    {
      "completion_length": 50.984375,
      "epoch": 1.0745672436751,
      "grad_norm": 12.561297416687012,
      "kl": 2.8125,
      "learning_rate": 7.313333333333333e-07,
      "loss": 0.2454,
      "reward": 0.5994217693805695,
      "reward_std": 0.13762104511260986,
      "reward_std_all": 0.16090942919254303,
      "rewards-max/forensics_sinkhorn_reward": 0.9849111139774323,
      "rewards-max/forensics_strict_edge_reward": 0.8983768224716187,
      "rewards-max/forensics_tiou_reward": 0.9840428829193115,
      "rewards/forensics_sinkhorn_reward": 0.4765470027923584,
      "rewards/forensics_strict_edge_reward": 0.1963111162185669,
      "rewards/forensics_tiou_reward": 0.29291243851184845,
      "step": 404
    },
    {
      "completion_length": 27.859375,
      "epoch": 1.0772303595206392,
      "grad_norm": 18.94430923461914,
      "kl": 2.265625,
      "learning_rate": 7.306666666666666e-07,
      "loss": 0.1728,
      "reward": 0.6508549153804779,
      "reward_std": 0.12810871377587318,
      "reward_std_all": 0.14058778062462807,
      "rewards-max/forensics_sinkhorn_reward": 0.98770472407341,
      "rewards-max/forensics_strict_edge_reward": 0.9464850425720215,
      "rewards-max/forensics_tiou_reward": 0.9949316680431366,
      "rewards/forensics_sinkhorn_reward": 0.5220272243022919,
      "rewards/forensics_strict_edge_reward": 0.19597665965557098,
      "rewards/forensics_tiou_reward": 0.3322097063064575,
      "step": 405
    },
    {
      "completion_length": 31.71875,
      "epoch": 1.0798934753661784,
      "grad_norm": 16.204904556274414,
      "kl": 2.53125,
      "learning_rate": 7.3e-07,
      "loss": 0.2762,
      "reward": 0.5955568253993988,
      "reward_std": 0.19805214554071426,
      "reward_std_all": 0.2000308409333229,
      "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": 0.5075600743293762,
      "rewards/forensics_strict_edge_reward": 0.23425480723381042,
      "rewards/forensics_tiou_reward": 0.3345891088247299,
      "step": 406
    },
    {
      "completion_length": 35.296875,
      "epoch": 1.0825565912117177,
      "grad_norm": 14.020045280456543,
      "kl": 1.9296875,
      "learning_rate": 7.293333333333332e-07,
      "loss": 0.1822,
      "reward": 0.6522582471370697,
      "reward_std": 0.18000245839357376,
      "reward_std_all": 0.17300200462341309,
      "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": 0.5529786050319672,
      "rewards/forensics_strict_edge_reward": 0.31977610290050507,
      "rewards/forensics_tiou_reward": 0.39486244320869446,
      "step": 407
    },
    {
      "completion_length": 77.546875,
      "epoch": 1.085219707057257,
      "grad_norm": 14.678840637207031,
      "kl": 3.078125,
      "learning_rate": 7.286666666666666e-07,
      "loss": 0.3093,
      "reward": 0.5592623054981232,
      "reward_std": 0.21562756597995758,
      "reward_std_all": 0.22536179423332214,
      "rewards-max/forensics_sinkhorn_reward": 0.9853417575359344,
      "rewards-max/forensics_strict_edge_reward": 0.9256840348243713,
      "rewards-max/forensics_tiou_reward": 0.9856211543083191,
      "rewards/forensics_sinkhorn_reward": 0.48591500520706177,
      "rewards/forensics_strict_edge_reward": 0.21628740429878235,
      "rewards/forensics_tiou_reward": 0.3797522634267807,
      "step": 408
    },
    {
      "completion_length": 45.03125,
      "epoch": 1.0878828229027964,
      "grad_norm": 15.537835121154785,
      "kl": 2.296875,
      "learning_rate": 7.28e-07,
      "loss": 0.1533,
      "reward": 0.6005059480667114,
      "reward_std": 0.17677967995405197,
      "reward_std_all": 0.1889730840921402,
      "rewards-max/forensics_sinkhorn_reward": 0.9864748418331146,
      "rewards-max/forensics_strict_edge_reward": 0.9169811904430389,
      "rewards-max/forensics_tiou_reward": 0.9901218414306641,
      "rewards/forensics_sinkhorn_reward": 0.5341278314590454,
      "rewards/forensics_strict_edge_reward": 0.2707735151052475,
      "rewards/forensics_tiou_reward": 0.3577297702431679,
      "step": 409
    },
    {
      "completion_length": 55.5,
      "epoch": 1.0905459387483356,
      "grad_norm": 10.568811416625977,
      "kl": 2.0390625,
      "learning_rate": 7.273333333333333e-07,
      "loss": 0.1206,
      "reward": 0.6440264880657196,
      "reward_std": 0.1376543641090393,
      "reward_std_all": 0.1533677875995636,
      "rewards-max/forensics_sinkhorn_reward": 0.9862253069877625,
      "rewards-max/forensics_strict_edge_reward": 0.9169811904430389,
      "rewards-max/forensics_tiou_reward": 0.9891103208065033,
      "rewards/forensics_sinkhorn_reward": 0.534238874912262,
      "rewards/forensics_strict_edge_reward": 0.3343093991279602,
      "rewards/forensics_tiou_reward": 0.4292805343866348,
      "step": 410
    },
    {
      "completion_length": 47.28125,
      "epoch": 1.0932090545938749,
      "grad_norm": 14.979262351989746,
      "kl": 1.765625,
      "learning_rate": 7.266666666666667e-07,
      "loss": 0.1845,
      "reward": 0.6124578714370728,
      "reward_std": 0.15250475704669952,
      "reward_std_all": 0.1660831943154335,
      "rewards-max/forensics_sinkhorn_reward": 0.9872406125068665,
      "rewards-max/forensics_strict_edge_reward": 0.9890114367008209,
      "rewards-max/forensics_tiou_reward": 0.9961067736148834,
      "rewards/forensics_sinkhorn_reward": 0.4939365088939667,
      "rewards/forensics_strict_edge_reward": 0.2327774539589882,
      "rewards/forensics_tiou_reward": 0.3468943387269974,
      "step": 411
    },
    {
      "completion_length": 59.5,
      "epoch": 1.095872170439414,
      "grad_norm": 13.611330032348633,
      "kl": 1.5234375,
      "learning_rate": 7.259999999999999e-07,
      "loss": 0.1607,
      "reward": 0.645874559879303,
      "reward_std": 0.14393576234579086,
      "reward_std_all": 0.15238816663622856,
      "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": 0.5456182062625885,
      "rewards/forensics_strict_edge_reward": 0.2521693706512451,
      "rewards/forensics_tiou_reward": 0.342818021774292,
      "step": 412
    },
    {
      "completion_length": 47.875,
      "epoch": 1.0985352862849533,
      "grad_norm": 14.074601173400879,
      "kl": 1.546875,
      "learning_rate": 7.253333333333334e-07,
      "loss": 0.1323,
      "reward": 0.6354533433914185,
      "reward_std": 0.16420269012451172,
      "reward_std_all": 0.16381068900227547,
      "rewards-max/forensics_sinkhorn_reward": 0.9715285003185272,
      "rewards-max/forensics_strict_edge_reward": 0.927969753742218,
      "rewards-max/forensics_tiou_reward": 0.9600353240966797,
      "rewards/forensics_sinkhorn_reward": 0.5049098581075668,
      "rewards/forensics_strict_edge_reward": 0.22886922955513,
      "rewards/forensics_tiou_reward": 0.30513791739940643,
      "step": 413
    },
    {
      "completion_length": 56.03125,
      "epoch": 1.1011984021304926,
      "grad_norm": 11.639203071594238,
      "kl": 2.33203125,
      "learning_rate": 7.246666666666666e-07,
      "loss": 0.1541,
      "reward": 0.5845421850681305,
      "reward_std": 0.11609474942088127,
      "reward_std_all": 0.16248837113380432,
      "rewards-max/forensics_sinkhorn_reward": 0.9671626687049866,
      "rewards-max/forensics_strict_edge_reward": 0.9093653857707977,
      "rewards-max/forensics_tiou_reward": 0.9512390792369843,
      "rewards/forensics_sinkhorn_reward": 0.4571598172187805,
      "rewards/forensics_strict_edge_reward": 0.1958746537566185,
      "rewards/forensics_tiou_reward": 0.23332123458385468,
      "step": 414
    },
    {
      "completion_length": 93.53125,
      "epoch": 1.103861517976032,
      "grad_norm": 12.80567741394043,
      "kl": 2.546875,
      "learning_rate": 7.24e-07,
      "loss": 0.2284,
      "reward": 0.5759762525558472,
      "reward_std": 0.16624793410301208,
      "reward_std_all": 0.19833305478096008,
      "rewards-max/forensics_sinkhorn_reward": 0.8954167068004608,
      "rewards-max/forensics_strict_edge_reward": 0.5971959978342056,
      "rewards-max/forensics_tiou_reward": 0.7956817150115967,
      "rewards/forensics_sinkhorn_reward": 0.41193559765815735,
      "rewards/forensics_strict_edge_reward": 0.12989194877445698,
      "rewards/forensics_tiou_reward": 0.19809802994132042,
      "step": 415
    },
    {
      "completion_length": 53.9375,
      "epoch": 1.1065246338215713,
      "grad_norm": 14.629151344299316,
      "kl": 2.5234375,
      "learning_rate": 7.233333333333333e-07,
      "loss": 0.2749,
      "reward": 0.6128780543804169,
      "reward_std": 0.1470508575439453,
      "reward_std_all": 0.16690243780612946,
      "rewards-max/forensics_sinkhorn_reward": 0.9451567828655243,
      "rewards-max/forensics_strict_edge_reward": 0.8189329206943512,
      "rewards-max/forensics_tiou_reward": 0.9006680548191071,
      "rewards/forensics_sinkhorn_reward": 0.4528314471244812,
      "rewards/forensics_strict_edge_reward": 0.21874941885471344,
      "rewards/forensics_tiou_reward": 0.27073003351688385,
      "step": 416
    },
    {
      "completion_length": 51.453125,
      "epoch": 1.1091877496671105,
      "grad_norm": 16.2568359375,
      "kl": 2.21875,
      "learning_rate": 7.226666666666667e-07,
      "loss": 0.3034,
      "reward": 0.5865218639373779,
      "reward_std": 0.14895374327898026,
      "reward_std_all": 0.17968466132879257,
      "rewards-max/forensics_sinkhorn_reward": 0.9174893498420715,
      "rewards-max/forensics_strict_edge_reward": 0.7868767082691193,
      "rewards-max/forensics_tiou_reward": 0.8453840613365173,
      "rewards/forensics_sinkhorn_reward": 0.4735715836286545,
      "rewards/forensics_strict_edge_reward": 0.17408613115549088,
      "rewards/forensics_tiou_reward": 0.22624638676643372,
      "step": 417
    },
    {
      "completion_length": 61.53125,
      "epoch": 1.1118508655126498,
      "grad_norm": 18.048686981201172,
      "kl": 3.109375,
      "learning_rate": 7.219999999999999e-07,
      "loss": 0.162,
      "reward": 0.5730506479740143,
      "reward_std": 0.16674575209617615,
      "reward_std_all": 0.1916387379169464,
      "rewards-max/forensics_sinkhorn_reward": 0.988638162612915,
      "rewards-max/forensics_strict_edge_reward": 0.9677534997463226,
      "rewards-max/forensics_tiou_reward": 0.9986120760440826,
      "rewards/forensics_sinkhorn_reward": 0.4262191206216812,
      "rewards/forensics_strict_edge_reward": 0.22346658259630203,
      "rewards/forensics_tiou_reward": 0.2584105357527733,
      "step": 418
    },
    {
      "completion_length": 69.265625,
      "epoch": 1.114513981358189,
      "grad_norm": 11.418391227722168,
      "kl": 2.1015625,
      "learning_rate": 7.213333333333334e-07,
      "loss": 0.27,
      "reward": 0.6416674852371216,
      "reward_std": 0.18887577950954437,
      "reward_std_all": 0.1803673356771469,
      "rewards-max/forensics_sinkhorn_reward": 0.9874922931194305,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9940958619117737,
      "rewards/forensics_sinkhorn_reward": 0.5316872149705887,
      "rewards/forensics_strict_edge_reward": 0.29058779776096344,
      "rewards/forensics_tiou_reward": 0.37282368540763855,
      "step": 419
    },
    {
      "completion_length": 46.0625,
      "epoch": 1.1171770972037283,
      "grad_norm": 13.581111907958984,
      "kl": 1.984375,
      "learning_rate": 7.206666666666666e-07,
      "loss": 0.3843,
      "reward": 0.6199376285076141,
      "reward_std": 0.1873987391591072,
      "reward_std_all": 0.19026856124401093,
      "rewards-max/forensics_sinkhorn_reward": 0.9868562817573547,
      "rewards-max/forensics_strict_edge_reward": 0.9464850425720215,
      "rewards-max/forensics_tiou_reward": 0.9916170239448547,
      "rewards/forensics_sinkhorn_reward": 0.5204031765460968,
      "rewards/forensics_strict_edge_reward": 0.22721266746520996,
      "rewards/forensics_tiou_reward": 0.32141098380088806,
      "step": 420
    },
    {
      "completion_length": 37.125,
      "epoch": 1.1198402130492677,
      "grad_norm": 143.3419952392578,
      "kl": 3.5390625,
      "learning_rate": 7.2e-07,
      "loss": 0.3883,
      "reward": 0.6089823544025421,
      "reward_std": 0.1778993047773838,
      "reward_std_all": 0.17986883223056793,
      "rewards-max/forensics_sinkhorn_reward": 0.9881793558597565,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9968023896217346,
      "rewards/forensics_sinkhorn_reward": 0.5097860395908356,
      "rewards/forensics_strict_edge_reward": 0.2256680503487587,
      "rewards/forensics_tiou_reward": 0.3270663619041443,
      "step": 421
    },
    {
      "completion_length": 61.671875,
      "epoch": 1.122503328894807,
      "grad_norm": 19.121604919433594,
      "kl": 2.5078125,
      "learning_rate": 7.193333333333333e-07,
      "loss": 0.2185,
      "reward": 0.619976282119751,
      "reward_std": 0.1603824831545353,
      "reward_std_all": 0.16591260582208633,
      "rewards-max/forensics_sinkhorn_reward": 0.9884564280509949,
      "rewards-max/forensics_strict_edge_reward": 0.9782643616199493,
      "rewards-max/forensics_tiou_reward": 0.9978947341442108,
      "rewards/forensics_sinkhorn_reward": 0.48983344435691833,
      "rewards/forensics_strict_edge_reward": 0.23473037034273148,
      "rewards/forensics_tiou_reward": 0.3767266124486923,
      "step": 422
    },
    {
      "completion_length": 67.640625,
      "epoch": 1.1251664447403462,
      "grad_norm": 18.312122344970703,
      "kl": 2.8046875,
      "learning_rate": 7.186666666666667e-07,
      "loss": 0.3236,
      "reward": 0.5962550044059753,
      "reward_std": 0.1773710548877716,
      "reward_std_all": 0.18882445991039276,
      "rewards-max/forensics_sinkhorn_reward": 0.9041338264942169,
      "rewards-max/forensics_strict_edge_reward": 0.5548870265483856,
      "rewards-max/forensics_tiou_reward": 0.8197605013847351,
      "rewards/forensics_sinkhorn_reward": 0.4277244955301285,
      "rewards/forensics_strict_edge_reward": 0.15070711076259613,
      "rewards/forensics_tiou_reward": 0.22931137681007385,
      "step": 423
    },
    {
      "completion_length": 90.671875,
      "epoch": 1.1278295605858855,
      "grad_norm": 14.083884239196777,
      "kl": 2.421875,
      "learning_rate": 7.179999999999999e-07,
      "loss": 0.1412,
      "reward": 0.6320231258869171,
      "reward_std": 0.1761232241988182,
      "reward_std_all": 0.17393241450190544,
      "rewards-max/forensics_sinkhorn_reward": 0.983804851770401,
      "rewards-max/forensics_strict_edge_reward": 0.946017861366272,
      "rewards-max/forensics_tiou_reward": 0.9872918426990509,
      "rewards/forensics_sinkhorn_reward": 0.4825397878885269,
      "rewards/forensics_strict_edge_reward": 0.2056404873728752,
      "rewards/forensics_tiou_reward": 0.28354504704475403,
      "step": 424
    },
    {
      "completion_length": 77.96875,
      "epoch": 1.1304926764314247,
      "grad_norm": 16.746891021728516,
      "kl": 2.29296875,
      "learning_rate": 7.173333333333333e-07,
      "loss": 0.1529,
      "reward": 0.5364169478416443,
      "reward_std": 0.2751166522502899,
      "reward_std_all": 0.2718387246131897,
      "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": 0.4340692460536957,
      "rewards/forensics_strict_edge_reward": 0.18623968958854675,
      "rewards/forensics_tiou_reward": 0.26107312738895416,
      "step": 425
    },
    {
      "completion_length": 80.0,
      "epoch": 1.133155792276964,
      "grad_norm": 10.835363388061523,
      "kl": 2.375,
      "learning_rate": 7.166666666666667e-07,
      "loss": 0.1844,
      "reward": 0.6147922873497009,
      "reward_std": 0.19146698340773582,
      "reward_std_all": 0.18793664127588272,
      "rewards-max/forensics_sinkhorn_reward": 0.9856811463832855,
      "rewards-max/forensics_strict_edge_reward": 0.8876297473907471,
      "rewards-max/forensics_tiou_reward": 0.9870330095291138,
      "rewards/forensics_sinkhorn_reward": 0.4794367849826813,
      "rewards/forensics_strict_edge_reward": 0.21117042005062103,
      "rewards/forensics_tiou_reward": 0.31723247468471527,
      "step": 426
    },
    {
      "completion_length": 80.46875,
      "epoch": 1.1358189081225034,
      "grad_norm": 385.3005065917969,
      "kl": 5.40625,
      "learning_rate": 7.159999999999999e-07,
      "loss": 0.2767,
      "reward": 0.6429980397224426,
      "reward_std": 0.16316571831703186,
      "reward_std_all": 0.16116737574338913,
      "rewards-max/forensics_sinkhorn_reward": 0.9885415136814117,
      "rewards-max/forensics_strict_edge_reward": 0.9782643616199493,
      "rewards-max/forensics_tiou_reward": 0.9982300996780396,
      "rewards/forensics_sinkhorn_reward": 0.5381563007831573,
      "rewards/forensics_strict_edge_reward": 0.2504286617040634,
      "rewards/forensics_tiou_reward": 0.4026334136724472,
      "step": 427
    },
    {
      "completion_length": 78.546875,
      "epoch": 1.1384820239680427,
      "grad_norm": 12.002371788024902,
      "kl": 2.2109375,
      "learning_rate": 7.153333333333334e-07,
      "loss": 0.2694,
      "reward": 0.6180631518363953,
      "reward_std": 0.16239354759454727,
      "reward_std_all": 0.17086123675107956,
      "rewards-max/forensics_sinkhorn_reward": 0.9875685572624207,
      "rewards-max/forensics_strict_edge_reward": 0.9565287232398987,
      "rewards-max/forensics_tiou_reward": 0.9943965971469879,
      "rewards/forensics_sinkhorn_reward": 0.5336214005947113,
      "rewards/forensics_strict_edge_reward": 0.23090407997369766,
      "rewards/forensics_tiou_reward": 0.34464915096759796,
      "step": 428
    },
    {
      "completion_length": 97.78125,
      "epoch": 1.141145139813582,
      "grad_norm": 15.416975975036621,
      "kl": 2.4296875,
      "learning_rate": 7.146666666666666e-07,
      "loss": 0.2263,
      "reward": 0.616814911365509,
      "reward_std": 0.2258726954460144,
      "reward_std_all": 0.2135426104068756,
      "rewards-max/forensics_sinkhorn_reward": 0.9884631037712097,
      "rewards-max/forensics_strict_edge_reward": 0.9574736058712006,
      "rewards-max/forensics_tiou_reward": 0.9979231655597687,
      "rewards/forensics_sinkhorn_reward": 0.5423714816570282,
      "rewards/forensics_strict_edge_reward": 0.32990850508213043,
      "rewards/forensics_tiou_reward": 0.4273093342781067,
      "step": 429
    },
    {
      "completion_length": 60.875,
      "epoch": 1.1438082556591211,
      "grad_norm": 15.239991188049316,
      "kl": 2.91796875,
      "learning_rate": 7.14e-07,
      "loss": 0.2771,
      "reward": 0.5955387055873871,
      "reward_std": 0.17691928893327713,
      "reward_std_all": 0.19365140795707703,
      "rewards-max/forensics_sinkhorn_reward": 0.9873608350753784,
      "rewards-max/forensics_strict_edge_reward": 0.9464850425720215,
      "rewards-max/forensics_tiou_reward": 0.9935909807682037,
      "rewards/forensics_sinkhorn_reward": 0.44828304648399353,
      "rewards/forensics_strict_edge_reward": 0.20773577690124512,
      "rewards/forensics_tiou_reward": 0.2713035121560097,
      "step": 430
    },
    {
      "completion_length": 37.15625,
      "epoch": 1.1464713715046604,
      "grad_norm": 14.284860610961914,
      "kl": 1.5390625,
      "learning_rate": 7.133333333333333e-07,
      "loss": 0.1297,
      "reward": 0.607736200094223,
      "reward_std": 0.1591751128435135,
      "reward_std_all": 0.17593380063772202,
      "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": 0.5206788033246994,
      "rewards/forensics_strict_edge_reward": 0.22315441071987152,
      "rewards/forensics_tiou_reward": 0.2484150379896164,
      "step": 431
    },
    {
      "completion_length": 47.515625,
      "epoch": 1.1491344873501999,
      "grad_norm": 9.713846206665039,
      "kl": 1.29296875,
      "learning_rate": 7.126666666666667e-07,
      "loss": 0.1832,
      "reward": 0.6911019682884216,
      "reward_std": 0.09359833225607872,
      "reward_std_all": 0.10122167691588402,
      "rewards-max/forensics_sinkhorn_reward": 0.9879873096942902,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9960492849349976,
      "rewards/forensics_sinkhorn_reward": 0.6109708249568939,
      "rewards/forensics_strict_edge_reward": 0.36500564217567444,
      "rewards/forensics_tiou_reward": 0.43882201611995697,
      "step": 432
    },
    {
      "completion_length": 40.1875,
      "epoch": 1.151797603195739,
      "grad_norm": 11.716303825378418,
      "kl": 1.494140625,
      "learning_rate": 7.119999999999999e-07,
      "loss": 0.1936,
      "reward": 0.6353454887866974,
      "reward_std": 0.1766907423734665,
      "reward_std_all": 0.17609946429729462,
      "rewards-max/forensics_sinkhorn_reward": 0.9885231554508209,
      "rewards-max/forensics_strict_edge_reward": 0.9464850425720215,
      "rewards-max/forensics_tiou_reward": 0.9981577694416046,
      "rewards/forensics_sinkhorn_reward": 0.4899820536375046,
      "rewards/forensics_strict_edge_reward": 0.2343483567237854,
      "rewards/forensics_tiou_reward": 0.27499528229236603,
      "step": 433
    },
    {
      "completion_length": 63.0,
      "epoch": 1.1544607190412783,
      "grad_norm": 14.695828437805176,
      "kl": 1.72265625,
      "learning_rate": 7.113333333333334e-07,
      "loss": 0.1629,
      "reward": 0.6128577291965485,
      "reward_std": 0.10460885241627693,
      "reward_std_all": 0.14517025649547577,
      "rewards-max/forensics_sinkhorn_reward": 0.9874576926231384,
      "rewards-max/forensics_strict_edge_reward": 0.9357379674911499,
      "rewards-max/forensics_tiou_reward": 0.9939590096473694,
      "rewards/forensics_sinkhorn_reward": 0.5073689222335815,
      "rewards/forensics_strict_edge_reward": 0.15685506165027618,
      "rewards/forensics_tiou_reward": 0.22702836245298386,
      "step": 434
    },
    {
      "completion_length": 62.375,
      "epoch": 1.1571238348868176,
      "grad_norm": 10.847455978393555,
      "kl": 1.90234375,
      "learning_rate": 7.106666666666666e-07,
      "loss": 0.2234,
      "reward": 0.6533180773258209,
      "reward_std": 0.15233907848596573,
      "reward_std_all": 0.15837842971086502,
      "rewards-max/forensics_sinkhorn_reward": 0.9668512046337128,
      "rewards-max/forensics_strict_edge_reward": 0.7849180996417999,
      "rewards-max/forensics_tiou_reward": 0.9490052461624146,
      "rewards/forensics_sinkhorn_reward": 0.5853076279163361,
      "rewards/forensics_strict_edge_reward": 0.2233840525150299,
      "rewards/forensics_tiou_reward": 0.3676670491695404,
      "step": 435
    },
    {
      "completion_length": 75.6875,
      "epoch": 1.1597869507323568,
      "grad_norm": 12.723292350769043,
      "kl": 2.0703125,
      "learning_rate": 7.1e-07,
      "loss": 0.2807,
      "reward": 0.6055534183979034,
      "reward_std": 0.21458209306001663,
      "reward_std_all": 0.21425621211528778,
      "rewards-max/forensics_sinkhorn_reward": 0.9857568740844727,
      "rewards-max/forensics_strict_edge_reward": 0.9149472117424011,
      "rewards-max/forensics_tiou_reward": 0.9872725903987885,
      "rewards/forensics_sinkhorn_reward": 0.4826354682445526,
      "rewards/forensics_strict_edge_reward": 0.23071691393852234,
      "rewards/forensics_tiou_reward": 0.320293128490448,
      "step": 436
    },
    {
      "completion_length": 39.953125,
      "epoch": 1.162450066577896,
      "grad_norm": 11.636223793029785,
      "kl": 1.55078125,
      "learning_rate": 7.093333333333333e-07,
      "loss": 0.1018,
      "reward": 0.6901736259460449,
      "reward_std": 0.0938710868358612,
      "reward_std_all": 0.10210209339857101,
      "rewards-max/forensics_sinkhorn_reward": 0.9871693551540375,
      "rewards-max/forensics_strict_edge_reward": 0.9565287232398987,
      "rewards-max/forensics_tiou_reward": 0.9928304851055145,
      "rewards/forensics_sinkhorn_reward": 0.6074456572532654,
      "rewards/forensics_strict_edge_reward": 0.29473453760147095,
      "rewards/forensics_tiou_reward": 0.40951456129550934,
      "step": 437
    },
    {
      "completion_length": 58.078125,
      "epoch": 1.1651131824234353,
      "grad_norm": 15.61916446685791,
      "kl": 2.328125,
      "learning_rate": 7.086666666666667e-07,
      "loss": 0.386,
      "reward": 0.6338359713554382,
      "reward_std": 0.18089110776782036,
      "reward_std_all": 0.1786789447069168,
      "rewards-max/forensics_sinkhorn_reward": 0.9885382950305939,
      "rewards-max/forensics_strict_edge_reward": 0.9782643616199493,
      "rewards-max/forensics_tiou_reward": 0.9982190430164337,
      "rewards/forensics_sinkhorn_reward": 0.5568652600049973,
      "rewards/forensics_strict_edge_reward": 0.2693333774805069,
      "rewards/forensics_tiou_reward": 0.3682444542646408,
      "step": 438
    },
    {
      "completion_length": 105.765625,
      "epoch": 1.1677762982689748,
      "grad_norm": 12.511971473693848,
      "kl": 2.6640625,
      "learning_rate": 7.079999999999999e-07,
      "loss": 0.1021,
      "reward": 0.6209384202957153,
      "reward_std": 0.1774429902434349,
      "reward_std_all": 0.18109063059091568,
      "rewards-max/forensics_sinkhorn_reward": 0.9879733324050903,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9959903657436371,
      "rewards/forensics_sinkhorn_reward": 0.48391585052013397,
      "rewards/forensics_strict_edge_reward": 0.24171533435583115,
      "rewards/forensics_tiou_reward": 0.3287755995988846,
      "step": 439
    },
    {
      "completion_length": 65.53125,
      "epoch": 1.170439414114514,
      "grad_norm": 13.204354286193848,
      "kl": 1.98046875,
      "learning_rate": 7.073333333333333e-07,
      "loss": 0.1663,
      "reward": 0.6162531077861786,
      "reward_std": 0.20779452472925186,
      "reward_std_all": 0.20823932439088821,
      "rewards-max/forensics_sinkhorn_reward": 0.9463672637939453,
      "rewards-max/forensics_strict_edge_reward": 0.9003686904907227,
      "rewards-max/forensics_tiou_reward": 0.9071884751319885,
      "rewards/forensics_sinkhorn_reward": 0.5734341740608215,
      "rewards/forensics_strict_edge_reward": 0.282083697617054,
      "rewards/forensics_tiou_reward": 0.3918347507715225,
      "step": 440
    },
    {
      "completion_length": 92.28125,
      "epoch": 1.1731025299600533,
      "grad_norm": 13.819408416748047,
      "kl": 2.88671875,
      "learning_rate": 7.066666666666666e-07,
      "loss": 0.3524,
      "reward": 0.6044211089611053,
      "reward_std": 0.19289613515138626,
      "reward_std_all": 0.19754207879304886,
      "rewards-max/forensics_sinkhorn_reward": 0.9882830381393433,
      "rewards-max/forensics_strict_edge_reward": 0.9565287232398987,
      "rewards-max/forensics_tiou_reward": 0.9972112774848938,
      "rewards/forensics_sinkhorn_reward": 0.5177440643310547,
      "rewards/forensics_strict_edge_reward": 0.2519407942891121,
      "rewards/forensics_tiou_reward": 0.32147035002708435,
      "step": 441
    },
    {
      "completion_length": 118.25,
      "epoch": 1.1757656458055925,
      "grad_norm": 12.807854652404785,
      "kl": 3.3203125,
      "learning_rate": 7.059999999999999e-07,
      "loss": 0.2823,
      "reward": 0.5500967502593994,
      "reward_std": 0.21839138865470886,
      "reward_std_all": 0.23421575874090195,
      "rewards-max/forensics_sinkhorn_reward": 0.9635065495967865,
      "rewards-max/forensics_strict_edge_reward": 0.9003686904907227,
      "rewards-max/forensics_tiou_reward": 0.9420340061187744,
      "rewards/forensics_sinkhorn_reward": 0.41709013283252716,
      "rewards/forensics_strict_edge_reward": 0.17675668001174927,
      "rewards/forensics_tiou_reward": 0.2935793697834015,
      "step": 442
    },
    {
      "completion_length": 123.90625,
      "epoch": 1.1784287616511318,
      "grad_norm": 17.218637466430664,
      "kl": 3.8359375,
      "learning_rate": 7.053333333333333e-07,
      "loss": 0.3057,
      "reward": 0.5622024536132812,
      "reward_std": 0.2176959365606308,
      "reward_std_all": 0.22496342658996582,
      "rewards-max/forensics_sinkhorn_reward": 0.9263728559017181,
      "rewards-max/forensics_strict_edge_reward": 0.8760378360748291,
      "rewards-max/forensics_tiou_reward": 0.8681951761245728,
      "rewards/forensics_sinkhorn_reward": 0.4266226142644882,
      "rewards/forensics_strict_edge_reward": 0.1683100387454033,
      "rewards/forensics_tiou_reward": 0.27637913823127747,
      "step": 443
    },
    {
      "completion_length": 117.359375,
      "epoch": 1.1810918774966712,
      "grad_norm": 16.029579162597656,
      "kl": 3.796875,
      "learning_rate": 7.046666666666666e-07,
      "loss": 0.3572,
      "reward": 0.5293907523155212,
      "reward_std": 0.18209611624479294,
      "reward_std_all": 0.23072554171085358,
      "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": 0.4260355681180954,
      "rewards/forensics_strict_edge_reward": 0.23156391084194183,
      "rewards/forensics_tiou_reward": 0.3261127471923828,
      "step": 444
    },
    {
      "completion_length": 83.71875,
      "epoch": 1.1837549933422105,
      "grad_norm": 11.730932235717773,
      "kl": 3.0078125,
      "learning_rate": 7.04e-07,
      "loss": 0.3416,
      "reward": 0.6274255216121674,
      "reward_std": 0.18284276872873306,
      "reward_std_all": 0.18396703898906708,
      "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": 0.5085338056087494,
      "rewards/forensics_strict_edge_reward": 0.29374855756759644,
      "rewards/forensics_tiou_reward": 0.34626878798007965,
      "step": 445
    },
    {
      "completion_length": 129.65625,
      "epoch": 1.1864181091877497,
      "grad_norm": 17.39276885986328,
      "kl": 2.7421875,
      "learning_rate": 7.033333333333333e-07,
      "loss": 0.2379,
      "reward": 0.5734078288078308,
      "reward_std": 0.25232212245464325,
      "reward_std_all": 0.2402593120932579,
      "rewards-max/forensics_sinkhorn_reward": 0.9194470047950745,
      "rewards-max/forensics_strict_edge_reward": 0.7464974820613861,
      "rewards-max/forensics_tiou_reward": 0.8430760502815247,
      "rewards/forensics_sinkhorn_reward": 0.46046195924282074,
      "rewards/forensics_strict_edge_reward": 0.26610126346349716,
      "rewards/forensics_tiou_reward": 0.336942620575428,
      "step": 446
    },
    {
      "completion_length": 139.09375,
      "epoch": 1.189081225033289,
      "grad_norm": 21.747873306274414,
      "kl": 4.6328125,
      "learning_rate": 7.026666666666667e-07,
      "loss": 0.2727,
      "reward": 0.6011695265769958,
      "reward_std": 0.20952491462230682,
      "reward_std_all": 0.20795172452926636,
      "rewards-max/forensics_sinkhorn_reward": 0.9507943987846375,
      "rewards-max/forensics_strict_edge_reward": 0.805042564868927,
      "rewards-max/forensics_tiou_reward": 0.9159512519836426,
      "rewards/forensics_sinkhorn_reward": 0.39447934925556183,
      "rewards/forensics_strict_edge_reward": 0.22175993770360947,
      "rewards/forensics_tiou_reward": 0.25886841118335724,
      "step": 447
    },
    {
      "completion_length": 107.96875,
      "epoch": 1.1917443408788282,
      "grad_norm": 12.565279960632324,
      "kl": 2.75,
      "learning_rate": 7.019999999999999e-07,
      "loss": 0.1931,
      "reward": 0.5906471014022827,
      "reward_std": 0.17024538293480873,
      "reward_std_all": 0.18840879201889038,
      "rewards-max/forensics_sinkhorn_reward": 0.9759139716625214,
      "rewards-max/forensics_strict_edge_reward": 0.8968285918235779,
      "rewards-max/forensics_tiou_reward": 0.9676253795623779,
      "rewards/forensics_sinkhorn_reward": 0.47352224588394165,
      "rewards/forensics_strict_edge_reward": 0.21271932870149612,
      "rewards/forensics_tiou_reward": 0.2833402454853058,
      "step": 448
    },
    {
      "completion_length": 51.90625,
      "epoch": 1.1944074567243674,
      "grad_norm": 14.518829345703125,
      "kl": 2.1953125,
      "learning_rate": 7.013333333333334e-07,
      "loss": 0.3314,
      "reward": 0.6201006770133972,
      "reward_std": 0.15082254260778427,
      "reward_std_all": 0.1625179424881935,
      "rewards-max/forensics_sinkhorn_reward": 0.9882135987281799,
      "rewards-max/forensics_strict_edge_reward": 0.9567649364471436,
      "rewards-max/forensics_tiou_reward": 0.996939480304718,
      "rewards/forensics_sinkhorn_reward": 0.5214625895023346,
      "rewards/forensics_strict_edge_reward": 0.2599751055240631,
      "rewards/forensics_tiou_reward": 0.33820249140262604,
      "step": 449
    },
    {
      "completion_length": 55.09375,
      "epoch": 1.1970705725699067,
      "grad_norm": 11.051692008972168,
      "kl": 1.9921875,
      "learning_rate": 7.006666666666666e-07,
      "loss": 0.3072,
      "reward": 0.6442990303039551,
      "reward_std": 0.10085269436240196,
      "reward_std_all": 0.1322912536561489,
      "rewards-max/forensics_sinkhorn_reward": 0.9851022362709045,
      "rewards-max/forensics_strict_edge_reward": 0.878633052110672,
      "rewards-max/forensics_tiou_reward": 0.9847719073295593,
      "rewards/forensics_sinkhorn_reward": 0.5339045226573944,
      "rewards/forensics_strict_edge_reward": 0.2694644629955292,
      "rewards/forensics_tiou_reward": 0.40793202817440033,
      "step": 450
    },
    {
      "completion_length": 40.703125,
      "epoch": 1.1997336884154461,
      "grad_norm": 16.4238338470459,
      "kl": 1.62109375,
      "learning_rate": 7e-07,
      "loss": 0.3126,
      "reward": 0.610440582036972,
      "reward_std": 0.1596500128507614,
      "reward_std_all": 0.16994233429431915,
      "rewards-max/forensics_sinkhorn_reward": 0.9865093529224396,
      "rewards-max/forensics_strict_edge_reward": 0.946017861366272,
      "rewards-max/forensics_tiou_reward": 0.990228921175003,
      "rewards/forensics_sinkhorn_reward": 0.4961957186460495,
      "rewards/forensics_strict_edge_reward": 0.2217751294374466,
      "rewards/forensics_tiou_reward": 0.3189276158809662,
      "step": 451
    },
    {
      "completion_length": 52.453125,
      "epoch": 1.2023968042609854,
      "grad_norm": 13.217440605163574,
      "kl": 1.6953125,
      "learning_rate": 6.993333333333333e-07,
      "loss": 0.2429,
      "reward": 0.6259284317493439,
      "reward_std": 0.16467135399580002,
      "reward_std_all": 0.17656506597995758,
      "rewards-max/forensics_sinkhorn_reward": 0.972150981426239,
      "rewards-max/forensics_strict_edge_reward": 0.8317028284072876,
      "rewards-max/forensics_tiou_reward": 0.9617137312889099,
      "rewards/forensics_sinkhorn_reward": 0.520070880651474,
      "rewards/forensics_strict_edge_reward": 0.2519233003258705,
      "rewards/forensics_tiou_reward": 0.34872812032699585,
      "step": 452
    },
    {
      "completion_length": 70.5,
      "epoch": 1.2050599201065246,
      "grad_norm": 13.744230270385742,
      "kl": 2.03125,
      "learning_rate": 6.986666666666667e-07,
      "loss": 0.266,
      "reward": 0.6381182074546814,
      "reward_std": 0.1518922969698906,
      "reward_std_all": 0.15804024785757065,
      "rewards-max/forensics_sinkhorn_reward": 0.9852842390537262,
      "rewards-max/forensics_strict_edge_reward": 0.9062341153621674,
      "rewards-max/forensics_tiou_reward": 0.985476166009903,
      "rewards/forensics_sinkhorn_reward": 0.5609870254993439,
      "rewards/forensics_strict_edge_reward": 0.27160847187042236,
      "rewards/forensics_tiou_reward": 0.38709811866283417,
      "step": 453
    },
    {
      "completion_length": 61.515625,
      "epoch": 1.2077230359520639,
      "grad_norm": 16.42621612548828,
      "kl": 2.09375,
      "learning_rate": 6.979999999999999e-07,
      "loss": 0.1375,
      "reward": 0.6382008194923401,
      "reward_std": 0.15563572943210602,
      "reward_std_all": 0.15715231001377106,
      "rewards-max/forensics_sinkhorn_reward": 0.9872835278511047,
      "rewards-max/forensics_strict_edge_reward": 0.9567649364471436,
      "rewards-max/forensics_tiou_reward": 0.9932717978954315,
      "rewards/forensics_sinkhorn_reward": 0.5333816409111023,
      "rewards/forensics_strict_edge_reward": 0.26303645968437195,
      "rewards/forensics_tiou_reward": 0.36407721042633057,
      "step": 454
    },
    {
      "completion_length": 73.328125,
      "epoch": 1.2103861517976031,
      "grad_norm": 15.961087226867676,
      "kl": 2.5703125,
      "learning_rate": 6.973333333333333e-07,
      "loss": 0.2695,
      "reward": 0.6257885694503784,
      "reward_std": 0.17393244057893753,
      "reward_std_all": 0.17436377704143524,
      "rewards-max/forensics_sinkhorn_reward": 0.9878372848033905,
      "rewards-max/forensics_strict_edge_reward": 0.9574736058712006,
      "rewards-max/forensics_tiou_reward": 0.9954545497894287,
      "rewards/forensics_sinkhorn_reward": 0.40629175305366516,
      "rewards/forensics_strict_edge_reward": 0.19318212568759918,
      "rewards/forensics_tiou_reward": 0.2554192990064621,
      "step": 455
    },
    {
      "completion_length": 44.90625,
      "epoch": 1.2130492676431426,
      "grad_norm": 15.93571662902832,
      "kl": 2.359375,
      "learning_rate": 6.966666666666666e-07,
      "loss": 0.219,
      "reward": 0.6401305198669434,
      "reward_std": 0.13821490854024887,
      "reward_std_all": 0.1556740403175354,
      "rewards-max/forensics_sinkhorn_reward": 0.9826421439647675,
      "rewards-max/forensics_strict_edge_reward": 0.9782643616199493,
      "rewards-max/forensics_tiou_reward": 0.9859223365783691,
      "rewards/forensics_sinkhorn_reward": 0.5221519470214844,
      "rewards/forensics_strict_edge_reward": 0.27433352917432785,
      "rewards/forensics_tiou_reward": 0.35040393471717834,
      "step": 456
    },
    {
      "completion_length": 51.703125,
      "epoch": 1.2157123834886818,
      "grad_norm": 13.388586044311523,
      "kl": 1.51953125,
      "learning_rate": 6.959999999999999e-07,
      "loss": 0.1661,
      "reward": 0.6436467170715332,
      "reward_std": 0.13121672719717026,
      "reward_std_all": 0.1401197910308838,
      "rewards-max/forensics_sinkhorn_reward": 0.9671737551689148,
      "rewards-max/forensics_strict_edge_reward": 0.8283384442329407,
      "rewards-max/forensics_tiou_reward": 0.9481236934661865,
      "rewards/forensics_sinkhorn_reward": 0.5097014755010605,
      "rewards/forensics_strict_edge_reward": 0.2088611125946045,
      "rewards/forensics_tiou_reward": 0.31191858649253845,
      "step": 457
    },
    {
      "completion_length": 32.953125,
      "epoch": 1.218375499334221,
      "grad_norm": 14.421076774597168,
      "kl": 1.53515625,
      "learning_rate": 6.953333333333333e-07,
      "loss": 0.0185,
      "reward": 0.7017952799797058,
      "reward_std": 0.08331763371825218,
      "reward_std_all": 0.0976678729057312,
      "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": 0.6578730940818787,
      "rewards/forensics_strict_edge_reward": 0.38881051540374756,
      "rewards/forensics_tiou_reward": 0.4650212526321411,
      "step": 458
    },
    {
      "completion_length": 43.359375,
      "epoch": 1.2210386151797603,
      "grad_norm": 14.279685020446777,
      "kl": 1.71875,
      "learning_rate": 6.946666666666666e-07,
      "loss": 0.1699,
      "reward": 0.6580673158168793,
      "reward_std": 0.10698796808719635,
      "reward_std_all": 0.13314774632453918,
      "rewards-max/forensics_sinkhorn_reward": 0.986920177936554,
      "rewards-max/forensics_strict_edge_reward": 0.9565287232398987,
      "rewards-max/forensics_tiou_reward": 0.9918562769889832,
      "rewards/forensics_sinkhorn_reward": 0.5880575776100159,
      "rewards/forensics_strict_edge_reward": 0.2975512892007828,
      "rewards/forensics_tiou_reward": 0.3981723189353943,
      "step": 459
    },
    {
      "completion_length": 49.140625,
      "epoch": 1.2237017310252996,
      "grad_norm": 8.975218772888184,
      "kl": 1.4296875,
      "learning_rate": 6.939999999999999e-07,
      "loss": 0.2362,
      "reward": 0.6728416979312897,
      "reward_std": 0.12676405906677246,
      "reward_std_all": 0.13036732003092766,
      "rewards-max/forensics_sinkhorn_reward": 0.988553911447525,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.998279869556427,
      "rewards/forensics_sinkhorn_reward": 0.6049027144908905,
      "rewards/forensics_strict_edge_reward": 0.33534496277570724,
      "rewards/forensics_tiou_reward": 0.43756847083568573,
      "step": 460
    },
    {
      "completion_length": 55.25,
      "epoch": 1.2263648468708388,
      "grad_norm": 11.90216064453125,
      "kl": 1.984375,
      "learning_rate": 6.933333333333333e-07,
      "loss": 0.1127,
      "reward": 0.6391955316066742,
      "reward_std": 0.18868296593427658,
      "reward_std_all": 0.18124652653932571,
      "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": 0.5627966225147247,
      "rewards/forensics_strict_edge_reward": 0.3508862406015396,
      "rewards/forensics_tiou_reward": 0.42001643776893616,
      "step": 461
    },
    {
      "completion_length": 53.109375,
      "epoch": 1.229027962716378,
      "grad_norm": 15.20920181274414,
      "kl": 1.37109375,
      "learning_rate": 6.926666666666666e-07,
      "loss": -0.0368,
      "reward": 0.6307258009910583,
      "reward_std": 0.1525096520781517,
      "reward_std_all": 0.16715964674949646,
      "rewards-max/forensics_sinkhorn_reward": 0.9844855666160583,
      "rewards-max/forensics_strict_edge_reward": 0.8957232534885406,
      "rewards-max/forensics_tiou_reward": 0.9823301434516907,
      "rewards/forensics_sinkhorn_reward": 0.4638812094926834,
      "rewards/forensics_strict_edge_reward": 0.19589264690876007,
      "rewards/forensics_tiou_reward": 0.20467911660671234,
      "step": 462
    },
    {
      "completion_length": 45.21875,
      "epoch": 1.2316910785619175,
      "grad_norm": 14.358567237854004,
      "kl": 1.55859375,
      "learning_rate": 6.919999999999999e-07,
      "loss": 0.1202,
      "reward": 0.6487367153167725,
      "reward_std": 0.13690169155597687,
      "reward_std_all": 0.14764698594808578,
      "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": 0.5794048309326172,
      "rewards/forensics_strict_edge_reward": 0.2835984081029892,
      "rewards/forensics_tiou_reward": 0.36913082003593445,
      "step": 463
    },
    {
      "completion_length": 89.796875,
      "epoch": 1.2343541944074568,
      "grad_norm": 13.345666885375977,
      "kl": 2.9375,
      "learning_rate": 6.913333333333334e-07,
      "loss": 0.2308,
      "reward": 0.564824640750885,
      "reward_std": 0.18656307458877563,
      "reward_std_all": 0.20720917731523514,
      "rewards-max/forensics_sinkhorn_reward": 0.9881978034973145,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9968767464160919,
      "rewards/forensics_sinkhorn_reward": 0.4595658928155899,
      "rewards/forensics_strict_edge_reward": 0.23799226433038712,
      "rewards/forensics_tiou_reward": 0.28823545575141907,
      "step": 464
    },
    {
      "completion_length": 29.28125,
      "epoch": 1.237017310252996,
      "grad_norm": 13.845705032348633,
      "kl": 1.234375,
      "learning_rate": 6.906666666666666e-07,
      "loss": 0.1132,
      "reward": 0.6497142910957336,
      "reward_std": 0.08546645194292068,
      "reward_std_all": 0.11976001411676407,
      "rewards-max/forensics_sinkhorn_reward": 0.9869871437549591,
      "rewards-max/forensics_strict_edge_reward": 0.9565287232398987,
      "rewards-max/forensics_tiou_reward": 0.9921191334724426,
      "rewards/forensics_sinkhorn_reward": 0.6107996106147766,
      "rewards/forensics_strict_edge_reward": 0.2999550551176071,
      "rewards/forensics_tiou_reward": 0.3742707818746567,
      "step": 465
    },
    {
      "completion_length": 47.65625,
      "epoch": 1.2396804260985352,
      "grad_norm": 12.155194282531738,
      "kl": 1.578125,
      "learning_rate": 6.9e-07,
      "loss": 0.1205,
      "reward": 0.6864826381206512,
      "reward_std": 0.09796284884214401,
      "reward_std_all": 0.10952840372920036,
      "rewards-max/forensics_sinkhorn_reward": 0.9723988771438599,
      "rewards-max/forensics_strict_edge_reward": 0.8260076344013214,
      "rewards-max/forensics_tiou_reward": 0.9627370536327362,
      "rewards/forensics_sinkhorn_reward": 0.6490186154842377,
      "rewards/forensics_strict_edge_reward": 0.28823621571063995,
      "rewards/forensics_tiou_reward": 0.4437760263681412,
      "step": 466
    },
    {
      "completion_length": 38.078125,
      "epoch": 1.2423435419440745,
      "grad_norm": 18.99393653869629,
      "kl": 2.1015625,
      "learning_rate": 6.893333333333333e-07,
      "loss": 0.1403,
      "reward": 0.6455529928207397,
      "reward_std": 0.15205364674329758,
      "reward_std_all": 0.15734362602233887,
      "rewards-max/forensics_sinkhorn_reward": 0.9646717011928558,
      "rewards-max/forensics_strict_edge_reward": 0.820955753326416,
      "rewards-max/forensics_tiou_reward": 0.9441051483154297,
      "rewards/forensics_sinkhorn_reward": 0.4950805902481079,
      "rewards/forensics_strict_edge_reward": 0.2144915610551834,
      "rewards/forensics_tiou_reward": 0.30736882984638214,
      "step": 467
    },
    {
      "completion_length": 51.96875,
      "epoch": 1.245006657789614,
      "grad_norm": 12.271119117736816,
      "kl": 2.4921875,
      "learning_rate": 6.886666666666667e-07,
      "loss": 0.1355,
      "reward": 0.6603822410106659,
      "reward_std": 0.19351337105035782,
      "reward_std_all": 0.17941690236330032,
      "rewards-max/forensics_sinkhorn_reward": 0.988001823425293,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9961040019989014,
      "rewards/forensics_sinkhorn_reward": 0.5691172778606415,
      "rewards/forensics_strict_edge_reward": 0.29766644537448883,
      "rewards/forensics_tiou_reward": 0.3672724813222885,
      "step": 468
    },
    {
      "completion_length": 27.46875,
      "epoch": 1.2476697736351532,
      "grad_norm": 13.512432098388672,
      "kl": 1.490234375,
      "learning_rate": 6.879999999999999e-07,
      "loss": 0.0439,
      "reward": 0.6575465202331543,
      "reward_std": 0.0897398591041565,
      "reward_std_all": 0.11925304681062698,
      "rewards-max/forensics_sinkhorn_reward": 0.9863811731338501,
      "rewards-max/forensics_strict_edge_reward": 0.9464850425720215,
      "rewards-max/forensics_tiou_reward": 0.989750325679779,
      "rewards/forensics_sinkhorn_reward": 0.5520871579647064,
      "rewards/forensics_strict_edge_reward": 0.23509882390499115,
      "rewards/forensics_tiou_reward": 0.31551340222358704,
      "step": 469
    },
    {
      "completion_length": 18.6875,
      "epoch": 1.2503328894806924,
      "grad_norm": 19.585506439208984,
      "kl": 1.865234375,
      "learning_rate": 6.873333333333334e-07,
      "loss": 0.0954,
      "reward": 0.70463427901268,
      "reward_std": 0.07097096741199493,
      "reward_std_all": 0.08684558048844337,
      "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": 0.6257392168045044,
      "rewards/forensics_strict_edge_reward": 0.27476267516613007,
      "rewards/forensics_tiou_reward": 0.371172234416008,
      "step": 470
    },
    {
      "completion_length": 32.4375,
      "epoch": 1.2529960053262317,
      "grad_norm": 14.421772956848145,
      "kl": 1.1796875,
      "learning_rate": 6.866666666666666e-07,
      "loss": 0.1402,
      "reward": 0.6617764234542847,
      "reward_std": 0.1094740480184555,
      "reward_std_all": 0.12479937821626663,
      "rewards-max/forensics_sinkhorn_reward": 0.9879468381404877,
      "rewards-max/forensics_strict_edge_reward": 0.9567649364471436,
      "rewards-max/forensics_tiou_reward": 0.9958901107311249,
      "rewards/forensics_sinkhorn_reward": 0.5744840502738953,
      "rewards/forensics_strict_edge_reward": 0.24860791862010956,
      "rewards/forensics_tiou_reward": 0.3128000646829605,
      "step": 471
    },
    {
      "completion_length": 34.34375,
      "epoch": 1.255659121171771,
      "grad_norm": 14.635504722595215,
      "kl": 1.66796875,
      "learning_rate": 6.86e-07,
      "loss": 0.2264,
      "reward": 0.6374242305755615,
      "reward_std": 0.11347965896129608,
      "reward_std_all": 0.13421853631734848,
      "rewards-max/forensics_sinkhorn_reward": 0.9883308708667755,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9974074065685272,
      "rewards/forensics_sinkhorn_reward": 0.5592497289180756,
      "rewards/forensics_strict_edge_reward": 0.3250724896788597,
      "rewards/forensics_tiou_reward": 0.39658641815185547,
      "step": 472
    },
    {
      "completion_length": 60.828125,
      "epoch": 1.2583222370173104,
      "grad_norm": 12.833170890808105,
      "kl": 2.89453125,
      "learning_rate": 6.853333333333333e-07,
      "loss": 0.1896,
      "reward": 0.6842690706253052,
      "reward_std": 0.08031357638537884,
      "reward_std_all": 0.09942630305886269,
      "rewards-max/forensics_sinkhorn_reward": 0.9872010946273804,
      "rewards-max/forensics_strict_edge_reward": 0.9464850425720215,
      "rewards-max/forensics_tiou_reward": 0.9929494559764862,
      "rewards/forensics_sinkhorn_reward": 0.5248356461524963,
      "rewards/forensics_strict_edge_reward": 0.293636679649353,
      "rewards/forensics_tiou_reward": 0.406925767660141,
      "step": 473
    },
    {
      "completion_length": 60.171875,
      "epoch": 1.2609853528628494,
      "grad_norm": 14.780105590820312,
      "kl": 2.8125,
      "learning_rate": 6.846666666666666e-07,
      "loss": 0.2292,
      "reward": 0.6392298638820648,
      "reward_std": 0.11641916632652283,
      "reward_std_all": 0.13669348508119583,
      "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": 0.4954083561897278,
      "rewards/forensics_strict_edge_reward": 0.22175823897123337,
      "rewards/forensics_tiou_reward": 0.2817819267511368,
      "step": 474
    },
    {
      "completion_length": 23.9375,
      "epoch": 1.2636484687083889,
      "grad_norm": 12.671178817749023,
      "kl": 2.44140625,
      "learning_rate": 6.84e-07,
      "loss": 0.1369,
      "reward": 0.6083011031150818,
      "reward_std": 0.11647382751107216,
      "reward_std_all": 0.15290514007210732,
      "rewards-max/forensics_sinkhorn_reward": 0.9448783099651337,
      "rewards-max/forensics_strict_edge_reward": 0.7293311953544617,
      "rewards-max/forensics_tiou_reward": 0.8436964750289917,
      "rewards/forensics_sinkhorn_reward": 0.4762980341911316,
      "rewards/forensics_strict_edge_reward": 0.13919835537672043,
      "rewards/forensics_tiou_reward": 0.21231891959905624,
      "step": 475
    },
    {
      "completion_length": 25.796875,
      "epoch": 1.2663115845539281,
      "grad_norm": 19.507421493530273,
      "kl": 1.4921875,
      "learning_rate": 6.833333333333333e-07,
      "loss": 0.2155,
      "reward": 0.6321206986904144,
      "reward_std": 0.148398257791996,
      "reward_std_all": 0.152907133102417,
      "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": 0.5098665803670883,
      "rewards/forensics_strict_edge_reward": 0.2657437026500702,
      "rewards/forensics_tiou_reward": 0.31472310423851013,
      "step": 476
    },
    {
      "completion_length": 58.75,
      "epoch": 1.2689747003994674,
      "grad_norm": 13.774798393249512,
      "kl": 3.04296875,
      "learning_rate": 6.826666666666666e-07,
      "loss": 0.354,
      "reward": 0.5939799249172211,
      "reward_std": 0.10803965106606483,
      "reward_std_all": 0.15002643316984177,
      "rewards-max/forensics_sinkhorn_reward": 0.9381087124347687,
      "rewards-max/forensics_strict_edge_reward": 0.7942859828472137,
      "rewards-max/forensics_tiou_reward": 0.8848046064376831,
      "rewards/forensics_sinkhorn_reward": 0.4739614874124527,
      "rewards/forensics_strict_edge_reward": 0.16724486649036407,
      "rewards/forensics_tiou_reward": 0.25380854308605194,
      "step": 477
    },
    {
      "completion_length": 39.828125,
      "epoch": 1.2716378162450066,
      "grad_norm": 16.84663963317871,
      "kl": 2.1328125,
      "learning_rate": 6.82e-07,
      "loss": 0.1457,
      "reward": 0.6366118490695953,
      "reward_std": 0.12953703477978706,
      "reward_std_all": 0.15032808482646942,
      "rewards-max/forensics_sinkhorn_reward": 0.986466646194458,
      "rewards-max/forensics_strict_edge_reward": 0.905340164899826,
      "rewards-max/forensics_tiou_reward": 0.9900805056095123,
      "rewards/forensics_sinkhorn_reward": 0.5338239818811417,
      "rewards/forensics_strict_edge_reward": 0.28863099217414856,
      "rewards/forensics_tiou_reward": 0.3813973218202591,
      "step": 478
    },
    {
      "completion_length": 42.8125,
      "epoch": 1.2743009320905458,
      "grad_norm": 19.775094985961914,
      "kl": 2.5390625,
      "learning_rate": 6.813333333333333e-07,
      "loss": 0.3259,
      "reward": 0.6772223711013794,
      "reward_std": 0.1364685520529747,
      "reward_std_all": 0.1359773501753807,
      "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": 0.5437791347503662,
      "rewards/forensics_strict_edge_reward": 0.28246381878852844,
      "rewards/forensics_tiou_reward": 0.33430342376232147,
      "step": 479
    },
    {
      "completion_length": 25.25,
      "epoch": 1.2769640479360853,
      "grad_norm": 14.009489059448242,
      "kl": 1.609375,
      "learning_rate": 6.806666666666666e-07,
      "loss": 0.2042,
      "reward": 0.6770690381526947,
      "reward_std": 0.11792811751365662,
      "reward_std_all": 0.12254111096262932,
      "rewards-max/forensics_sinkhorn_reward": 0.9884249866008759,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9977706968784332,
      "rewards/forensics_sinkhorn_reward": 0.6188764870166779,
      "rewards/forensics_strict_edge_reward": 0.3739411532878876,
      "rewards/forensics_tiou_reward": 0.46294620633125305,
      "step": 480
    },
    {
      "completion_length": 28.546875,
      "epoch": 1.2796271637816246,
      "grad_norm": 15.69017219543457,
      "kl": 2.1484375,
      "learning_rate": 6.800000000000001e-07,
      "loss": 0.2497,
      "reward": 0.5922783017158508,
      "reward_std": 0.18236864358186722,
      "reward_std_all": 0.19556092470884323,
      "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": 0.4784369170665741,
      "rewards/forensics_strict_edge_reward": 0.23902563750743866,
      "rewards/forensics_tiou_reward": 0.3074413612484932,
      "step": 481
    },
    {
      "completion_length": 24.390625,
      "epoch": 1.2822902796271638,
      "grad_norm": 18.769149780273438,
      "kl": 2.625,
      "learning_rate": 6.793333333333333e-07,
      "loss": 0.2086,
      "reward": 0.6070393025875092,
      "reward_std": 0.11831973120570183,
      "reward_std_all": 0.1534021720290184,
      "rewards-max/forensics_sinkhorn_reward": 0.9846445024013519,
      "rewards-max/forensics_strict_edge_reward": 0.8957232534885406,
      "rewards-max/forensics_tiou_reward": 0.9829744696617126,
      "rewards/forensics_sinkhorn_reward": 0.49540452659130096,
      "rewards/forensics_strict_edge_reward": 0.18276552855968475,
      "rewards/forensics_tiou_reward": 0.23464839160442352,
      "step": 482
    },
    {
      "completion_length": 24.828125,
      "epoch": 1.284953395472703,
      "grad_norm": 16.77455711364746,
      "kl": 1.91015625,
      "learning_rate": 6.786666666666667e-07,
      "loss": 0.1597,
      "reward": 0.5982202887535095,
      "reward_std": 0.16914880275726318,
      "reward_std_all": 0.18684543669223785,
      "rewards-max/forensics_sinkhorn_reward": 0.919835090637207,
      "rewards-max/forensics_strict_edge_reward": 0.8404377698898315,
      "rewards-max/forensics_tiou_reward": 0.838429868221283,
      "rewards/forensics_sinkhorn_reward": 0.447390541434288,
      "rewards/forensics_strict_edge_reward": 0.15466448664665222,
      "rewards/forensics_tiou_reward": 0.2547112852334976,
      "step": 483
    },
    {
      "completion_length": 45.65625,
      "epoch": 1.2876165113182423,
      "grad_norm": 12.039488792419434,
      "kl": 2.49609375,
      "learning_rate": 6.78e-07,
      "loss": 0.3498,
      "reward": 0.6371270418167114,
      "reward_std": 0.1751609519124031,
      "reward_std_all": 0.17009346187114716,
      "rewards-max/forensics_sinkhorn_reward": 0.9863424599170685,
      "rewards-max/forensics_strict_edge_reward": 0.9357379674911499,
      "rewards-max/forensics_tiou_reward": 0.9895949959754944,
      "rewards/forensics_sinkhorn_reward": 0.5772760808467865,
      "rewards/forensics_strict_edge_reward": 0.3237070143222809,
      "rewards/forensics_tiou_reward": 0.4299708902835846,
      "step": 484
    },
    {
      "completion_length": 47.671875,
      "epoch": 1.2902796271637818,
      "grad_norm": 13.522422790527344,
      "kl": 2.734375,
      "learning_rate": 6.773333333333334e-07,
      "loss": 0.3641,
      "reward": 0.6152097880840302,
      "reward_std": 0.16997383534908295,
      "reward_std_all": 0.17549757659435272,
      "rewards-max/forensics_sinkhorn_reward": 0.988310307264328,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9973214566707611,
      "rewards/forensics_sinkhorn_reward": 0.477008193731308,
      "rewards/forensics_strict_edge_reward": 0.17143843322992325,
      "rewards/forensics_tiou_reward": 0.3019992411136627,
      "step": 485
    },
    {
      "completion_length": 54.8125,
      "epoch": 1.2929427430093208,
      "grad_norm": 17.59832000732422,
      "kl": 2.984375,
      "learning_rate": 6.766666666666666e-07,
      "loss": 0.1484,
      "reward": 0.6575726270675659,
      "reward_std": 0.15947488695383072,
      "reward_std_all": 0.15459990501403809,
      "rewards-max/forensics_sinkhorn_reward": 0.9881263673305511,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9965938627719879,
      "rewards/forensics_sinkhorn_reward": 0.5541902184486389,
      "rewards/forensics_strict_edge_reward": 0.2555494159460068,
      "rewards/forensics_tiou_reward": 0.3347215950489044,
      "step": 486
    },
    {
      "completion_length": 30.84375,
      "epoch": 1.2956058588548602,
      "grad_norm": 17.31908416748047,
      "kl": 2.015625,
      "learning_rate": 6.76e-07,
      "loss": 0.2001,
      "reward": 0.6580592095851898,
      "reward_std": 0.17049474269151688,
      "reward_std_all": 0.1646927371621132,
      "rewards-max/forensics_sinkhorn_reward": 0.9790380299091339,
      "rewards-max/forensics_strict_edge_reward": 0.8681221902370453,
      "rewards-max/forensics_tiou_reward": 0.9614251852035522,
      "rewards/forensics_sinkhorn_reward": 0.5403985381126404,
      "rewards/forensics_strict_edge_reward": 0.2338002473115921,
      "rewards/forensics_tiou_reward": 0.31028519570827484,
      "step": 487
    },
    {
      "completion_length": 33.609375,
      "epoch": 1.2982689747003995,
      "grad_norm": 13.433326721191406,
      "kl": 1.9765625,
      "learning_rate": 6.753333333333333e-07,
      "loss": 0.1335,
      "reward": 0.7147774398326874,
      "reward_std": 0.1196775734424591,
      "reward_std_all": 0.10981452837586403,
      "rewards-max/forensics_sinkhorn_reward": 0.9867049753665924,
      "rewards-max/forensics_strict_edge_reward": 0.8719756007194519,
      "rewards-max/forensics_tiou_reward": 0.991023451089859,
      "rewards/forensics_sinkhorn_reward": 0.6093851625919342,
      "rewards/forensics_strict_edge_reward": 0.3186717927455902,
      "rewards/forensics_tiou_reward": 0.44887909293174744,
      "step": 488
    },
    {
      "completion_length": 52.34375,
      "epoch": 1.3009320905459387,
      "grad_norm": 16.374330520629883,
      "kl": 3.3203125,
      "learning_rate": 6.746666666666666e-07,
      "loss": 0.3763,
      "reward": 0.6098498702049255,
      "reward_std": 0.17974908649921417,
      "reward_std_all": 0.18267416208982468,
      "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": 0.5091345608234406,
      "rewards/forensics_strict_edge_reward": 0.18631266057491302,
      "rewards/forensics_tiou_reward": 0.30125343799591064,
      "step": 489
    },
    {
      "completion_length": 34.9375,
      "epoch": 1.303595206391478,
      "grad_norm": 21.508834838867188,
      "kl": 2.5390625,
      "learning_rate": 6.74e-07,
      "loss": 0.1734,
      "reward": 0.6242431998252869,
      "reward_std": 0.1718711107969284,
      "reward_std_all": 0.17385099083185196,
      "rewards-max/forensics_sinkhorn_reward": 0.9864546954631805,
      "rewards-max/forensics_strict_edge_reward": 0.9075756669044495,
      "rewards-max/forensics_tiou_reward": 0.9900394976139069,
      "rewards/forensics_sinkhorn_reward": 0.5054062306880951,
      "rewards/forensics_strict_edge_reward": 0.22770945727825165,
      "rewards/forensics_tiou_reward": 0.2936161831021309,
      "step": 490
    },
    {
      "completion_length": 37.53125,
      "epoch": 1.3062583222370172,
      "grad_norm": 24.919904708862305,
      "kl": 2.4453125,
      "learning_rate": 6.733333333333333e-07,
      "loss": 0.3011,
      "reward": 0.6391192078590393,
      "reward_std": 0.1352522298693657,
      "reward_std_all": 0.1517002023756504,
      "rewards-max/forensics_sinkhorn_reward": 0.9875576794147491,
      "rewards-max/forensics_strict_edge_reward": 0.9464850425720215,
      "rewards-max/forensics_tiou_reward": 0.9943526387214661,
      "rewards/forensics_sinkhorn_reward": 0.5549740791320801,
      "rewards/forensics_strict_edge_reward": 0.23792549967765808,
      "rewards/forensics_tiou_reward": 0.3639565110206604,
      "step": 491
    },
    {
      "completion_length": 39.859375,
      "epoch": 1.3089214380825567,
      "grad_norm": 18.478130340576172,
      "kl": 3.7890625,
      "learning_rate": 6.726666666666666e-07,
      "loss": 0.4008,
      "reward": 0.6232428848743439,
      "reward_std": 0.16450679302215576,
      "reward_std_all": 0.1661878526210785,
      "rewards-max/forensics_sinkhorn_reward": 0.9866049289703369,
      "rewards-max/forensics_strict_edge_reward": 0.9364311099052429,
      "rewards-max/forensics_tiou_reward": 0.9906335771083832,
      "rewards/forensics_sinkhorn_reward": 0.5167107731103897,
      "rewards/forensics_strict_edge_reward": 0.22487182915210724,
      "rewards/forensics_tiou_reward": 0.3005746304988861,
      "step": 492
    },
    {
      "completion_length": 37.546875,
      "epoch": 1.311584553928096,
      "grad_norm": 554.6721801757812,
      "kl": 13.59375,
      "learning_rate": 6.72e-07,
      "loss": 0.7162,
      "reward": 0.6167462766170502,
      "reward_std": 0.17090298235416412,
      "reward_std_all": 0.1801791936159134,
      "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": 0.5439972877502441,
      "rewards/forensics_strict_edge_reward": 0.30563701689243317,
      "rewards/forensics_tiou_reward": 0.36969348788261414,
      "step": 493
    },
    {
      "completion_length": 51.265625,
      "epoch": 1.3142476697736352,
      "grad_norm": 256.8647155761719,
      "kl": 7.6015625,
      "learning_rate": 6.713333333333333e-07,
      "loss": 0.4914,
      "reward": 0.6456623673439026,
      "reward_std": 0.11140328273177147,
      "reward_std_all": 0.13805069029331207,
      "rewards-max/forensics_sinkhorn_reward": 0.8485006093978882,
      "rewards-max/forensics_strict_edge_reward": 0.7570546269416809,
      "rewards-max/forensics_tiou_reward": 0.77842777967453,
      "rewards/forensics_sinkhorn_reward": 0.5119979679584503,
      "rewards/forensics_strict_edge_reward": 0.22309735417366028,
      "rewards/forensics_tiou_reward": 0.3076297342777252,
      "step": 494
    },
    {
      "completion_length": 53.765625,
      "epoch": 1.3169107856191744,
      "grad_norm": 19.442800521850586,
      "kl": 2.40625,
      "learning_rate": 6.706666666666666e-07,
      "loss": 0.1796,
      "reward": 0.6555840373039246,
      "reward_std": 0.1605323851108551,
      "reward_std_all": 0.1590031310915947,
      "rewards-max/forensics_sinkhorn_reward": 0.9872041046619415,
      "rewards-max/forensics_strict_edge_reward": 0.9565287232398987,
      "rewards-max/forensics_tiou_reward": 0.9929637908935547,
      "rewards/forensics_sinkhorn_reward": 0.5601330548524857,
      "rewards/forensics_strict_edge_reward": 0.26953115314245224,
      "rewards/forensics_tiou_reward": 0.36024244129657745,
      "step": 495
    },
    {
      "completion_length": 49.84375,
      "epoch": 1.3195739014647137,
      "grad_norm": 14.685772895812988,
      "kl": 2.6015625,
      "learning_rate": 6.7e-07,
      "loss": 0.1511,
      "reward": 0.6980077624320984,
      "reward_std": 0.1062137708067894,
      "reward_std_all": 0.11234055459499359,
      "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": 0.5947092771530151,
      "rewards/forensics_strict_edge_reward": 0.30723103880882263,
      "rewards/forensics_tiou_reward": 0.36636345088481903,
      "step": 496
    },
    {
      "completion_length": 27.890625,
      "epoch": 1.3222370173102531,
      "grad_norm": 17.235097885131836,
      "kl": 2.4921875,
      "learning_rate": 6.693333333333333e-07,
      "loss": 0.2701,
      "reward": 0.6415930092334747,
      "reward_std": 0.11543777212500572,
      "reward_std_all": 0.13420869037508965,
      "rewards-max/forensics_sinkhorn_reward": 0.9240263998508453,
      "rewards-max/forensics_strict_edge_reward": 0.7695834338665009,
      "rewards-max/forensics_tiou_reward": 0.8527003228664398,
      "rewards/forensics_sinkhorn_reward": 0.5089254379272461,
      "rewards/forensics_strict_edge_reward": 0.15439125150442123,
      "rewards/forensics_tiou_reward": 0.2719103544950485,
      "step": 497
    },
    {
      "completion_length": 30.203125,
      "epoch": 1.3249001331557924,
      "grad_norm": 21.479589462280273,
      "kl": 1.7734375,
      "learning_rate": 6.686666666666666e-07,
      "loss": 0.1804,
      "reward": 0.6497882604598999,
      "reward_std": 0.15851349383592606,
      "reward_std_all": 0.15707503259181976,
      "rewards-max/forensics_sinkhorn_reward": 0.9869279861450195,
      "rewards-max/forensics_strict_edge_reward": 0.946017861366272,
      "rewards-max/forensics_tiou_reward": 0.9918887913227081,
      "rewards/forensics_sinkhorn_reward": 0.5506866872310638,
      "rewards/forensics_strict_edge_reward": 0.24212802946567535,
      "rewards/forensics_tiou_reward": 0.366303026676178,
      "step": 498
    },
    {
      "completion_length": 22.34375,
      "epoch": 1.3275632490013316,
      "grad_norm": 17.911422729492188,
      "kl": 2.2109375,
      "learning_rate": 6.68e-07,
      "loss": 0.207,
      "reward": 0.6649187207221985,
      "reward_std": 0.12704090401530266,
      "reward_std_all": 0.13452302664518356,
      "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": 0.592345118522644,
      "rewards/forensics_strict_edge_reward": 0.35919541120529175,
      "rewards/forensics_tiou_reward": 0.46749068796634674,
      "step": 499
    },
    {
      "completion_length": 30.328125,
      "epoch": 1.3302263648468708,
      "grad_norm": 33.801849365234375,
      "kl": 2.765625,
      "learning_rate": 6.673333333333334e-07,
      "loss": 0.2096,
      "reward": 0.6276578009128571,
      "reward_std": 0.14473547041416168,
      "reward_std_all": 0.1588125005364418,
      "rewards-max/forensics_sinkhorn_reward": 0.9853898286819458,
      "rewards-max/forensics_strict_edge_reward": 0.9149472117424011,
      "rewards-max/forensics_tiou_reward": 0.9858502149581909,
      "rewards/forensics_sinkhorn_reward": 0.5479318499565125,
      "rewards/forensics_strict_edge_reward": 0.24135363847017288,
      "rewards/forensics_tiou_reward": 0.3714081346988678,
      "step": 500
    },
    {
      "completion_length": 37.0625,
      "epoch": 1.33288948069241,
      "grad_norm": 33.14463424682617,
      "kl": 2.01171875,
      "learning_rate": 6.666666666666666e-07,
      "loss": 0.2946,
      "reward": 0.6713139712810516,
      "reward_std": 0.1362965740263462,
      "reward_std_all": 0.13764448463916779,
      "rewards-max/forensics_sinkhorn_reward": 0.9559515118598938,
      "rewards-max/forensics_strict_edge_reward": 0.7781787216663361,
      "rewards-max/forensics_tiou_reward": 0.92293781042099,
      "rewards/forensics_sinkhorn_reward": 0.5316062271595001,
      "rewards/forensics_strict_edge_reward": 0.21162399649620056,
      "rewards/forensics_tiou_reward": 0.30981919169425964,
      "step": 501
    },
    {
      "completion_length": 35.203125,
      "epoch": 1.3355525965379493,
      "grad_norm": 16.663637161254883,
      "kl": 2.71875,
      "learning_rate": 6.66e-07,
      "loss": 0.338,
      "reward": 0.6472161412239075,
      "reward_std": 0.16282547265291214,
      "reward_std_all": 0.158541701734066,
      "rewards-max/forensics_sinkhorn_reward": 0.9886972308158875,
      "rewards-max/forensics_strict_edge_reward": 0.9890114367008209,
      "rewards-max/forensics_tiou_reward": 0.9988439381122589,
      "rewards/forensics_sinkhorn_reward": 0.5677942335605621,
      "rewards/forensics_strict_edge_reward": 0.3123485893011093,
      "rewards/forensics_tiou_reward": 0.4313125014305115,
      "step": 502
    },
    {
      "completion_length": 30.328125,
      "epoch": 1.3382157123834886,
      "grad_norm": 14.338709831237793,
      "kl": 2.671875,
      "learning_rate": 6.653333333333333e-07,
      "loss": 0.2386,
      "reward": 0.7173877060413361,
      "reward_std": 0.0947239100933075,
      "reward_std_all": 0.09753822535276413,
      "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": 0.6483564972877502,
      "rewards/forensics_strict_edge_reward": 0.35547445714473724,
      "rewards/forensics_tiou_reward": 0.508013516664505,
      "step": 503
    },
    {
      "completion_length": 33.625,
      "epoch": 1.340878828229028,
      "grad_norm": 20.461578369140625,
      "kl": 2.1328125,
      "learning_rate": 6.646666666666666e-07,
      "loss": 0.1585,
      "reward": 0.6529845893383026,
      "reward_std": 0.11539902538061142,
      "reward_std_all": 0.13325908780097961,
      "rewards-max/forensics_sinkhorn_reward": 0.9870195090770721,
      "rewards-max/forensics_strict_edge_reward": 0.9464850425720215,
      "rewards-max/forensics_tiou_reward": 0.9922601282596588,
      "rewards/forensics_sinkhorn_reward": 0.5159441828727722,
      "rewards/forensics_strict_edge_reward": 0.2762112095952034,
      "rewards/forensics_tiou_reward": 0.3312133252620697,
      "step": 504
    },
    {
      "completion_length": 39.296875,
      "epoch": 1.3435419440745673,
      "grad_norm": 17.143922805786133,
      "kl": 2.66015625,
      "learning_rate": 6.64e-07,
      "loss": 0.1829,
      "reward": 0.5979593992233276,
      "reward_std": 0.18825633823871613,
      "reward_std_all": 0.19195911288261414,
      "rewards-max/forensics_sinkhorn_reward": 0.9801363945007324,
      "rewards-max/forensics_strict_edge_reward": 0.8854433596134186,
      "rewards-max/forensics_tiou_reward": 0.9713366031646729,
      "rewards/forensics_sinkhorn_reward": 0.4565812647342682,
      "rewards/forensics_strict_edge_reward": 0.20705808699131012,
      "rewards/forensics_tiou_reward": 0.29095640778541565,
      "step": 505
    },
    {
      "completion_length": 42.453125,
      "epoch": 1.3462050599201065,
      "grad_norm": 20.370153427124023,
      "kl": 3.515625,
      "learning_rate": 6.633333333333333e-07,
      "loss": 0.3185,
      "reward": 0.5393282473087311,
      "reward_std": 0.12247341871261597,
      "reward_std_all": 0.19228608906269073,
      "rewards-max/forensics_sinkhorn_reward": 0.9864809215068817,
      "rewards-max/forensics_strict_edge_reward": 0.9062341153621674,
      "rewards-max/forensics_tiou_reward": 0.9901244938373566,
      "rewards/forensics_sinkhorn_reward": 0.4209749102592468,
      "rewards/forensics_strict_edge_reward": 0.21878349781036377,
      "rewards/forensics_tiou_reward": 0.2866736501455307,
      "step": 506
    },
    {
      "completion_length": 38.375,
      "epoch": 1.3488681757656458,
      "grad_norm": 23.886363983154297,
      "kl": 2.828125,
      "learning_rate": 6.626666666666666e-07,
      "loss": 0.151,
      "reward": 0.6331767737865448,
      "reward_std": 0.13742900639772415,
      "reward_std_all": 0.15979208052158356,
      "rewards-max/forensics_sinkhorn_reward": 0.9464689195156097,
      "rewards-max/forensics_strict_edge_reward": 0.7958523333072662,
      "rewards-max/forensics_tiou_reward": 0.8976539969444275,
      "rewards/forensics_sinkhorn_reward": 0.4980965852737427,
      "rewards/forensics_strict_edge_reward": 0.20394467562437057,
      "rewards/forensics_tiou_reward": 0.2771461382508278,
      "step": 507
    },
    {
      "completion_length": 50.578125,
      "epoch": 1.351531291611185,
      "grad_norm": 17.666894912719727,
      "kl": 4.578125,
      "learning_rate": 6.62e-07,
      "loss": 0.5499,
      "reward": 0.5888722836971283,
      "reward_std": 0.1927151456475258,
      "reward_std_all": 0.2075052633881569,
      "rewards-max/forensics_sinkhorn_reward": 0.8838441669940948,
      "rewards-max/forensics_strict_edge_reward": 0.7848804593086243,
      "rewards-max/forensics_tiou_reward": 0.7762152850627899,
      "rewards/forensics_sinkhorn_reward": 0.4058941453695297,
      "rewards/forensics_strict_edge_reward": 0.2312951683998108,
      "rewards/forensics_tiou_reward": 0.2308315709233284,
      "step": 508
    },
    {
      "completion_length": 43.71875,
      "epoch": 1.3541944074567245,
      "grad_norm": 12.125568389892578,
      "kl": 2.765625,
      "learning_rate": 6.613333333333333e-07,
      "loss": 0.2736,
      "reward": 0.6895735859870911,
      "reward_std": 0.1477639004588127,
      "reward_std_all": 0.13639945536851883,
      "rewards-max/forensics_sinkhorn_reward": 0.9883725643157959,
      "rewards-max/forensics_strict_edge_reward": 0.9364311099052429,
      "rewards-max/forensics_tiou_reward": 0.9975649416446686,
      "rewards/forensics_sinkhorn_reward": 0.6191440522670746,
      "rewards/forensics_strict_edge_reward": 0.3724038600921631,
      "rewards/forensics_tiou_reward": 0.48672838509082794,
      "step": 509
    },
    {
      "completion_length": 41.546875,
      "epoch": 1.3568575233022637,
      "grad_norm": 16.170978546142578,
      "kl": 3.140625,
      "learning_rate": 6.606666666666666e-07,
      "loss": 0.4052,
      "reward": 0.6480288505554199,
      "reward_std": 0.15600528940558434,
      "reward_std_all": 0.156147763133049,
      "rewards-max/forensics_sinkhorn_reward": 0.9874343574047089,
      "rewards-max/forensics_strict_edge_reward": 0.9357379674911499,
      "rewards-max/forensics_tiou_reward": 0.9938729107379913,
      "rewards/forensics_sinkhorn_reward": 0.44380776584148407,
      "rewards/forensics_strict_edge_reward": 0.23229211568832397,
      "rewards/forensics_tiou_reward": 0.2604609429836273,
      "step": 510
    },
    {
      "completion_length": 32.859375,
      "epoch": 1.359520639147803,
      "grad_norm": 22.294740676879883,
      "kl": 3.1484375,
      "learning_rate": 6.6e-07,
      "loss": 0.3195,
      "reward": 0.6030214130878448,
      "reward_std": 0.18027479201555252,
      "reward_std_all": 0.19268791377544403,
      "rewards-max/forensics_sinkhorn_reward": 0.9573889672756195,
      "rewards-max/forensics_strict_edge_reward": 0.744240403175354,
      "rewards-max/forensics_tiou_reward": 0.9232922792434692,
      "rewards/forensics_sinkhorn_reward": 0.48912861943244934,
      "rewards/forensics_strict_edge_reward": 0.21931307762861252,
      "rewards/forensics_tiou_reward": 0.349091112613678,
      "step": 511
    },
    {
      "completion_length": 33.3125,
      "epoch": 1.3621837549933422,
      "grad_norm": 17.529922485351562,
      "kl": 2.9765625,
      "learning_rate": 6.593333333333333e-07,
      "loss": 0.2729,
      "reward": 0.6318251192569733,
      "reward_std": 0.15579528734087944,
      "reward_std_all": 0.16430038213729858,
      "rewards-max/forensics_sinkhorn_reward": 0.9610923528671265,
      "rewards-max/forensics_strict_edge_reward": 0.8528120815753937,
      "rewards-max/forensics_tiou_reward": 0.9337665438652039,
      "rewards/forensics_sinkhorn_reward": 0.4789305478334427,
      "rewards/forensics_strict_edge_reward": 0.15989713370800018,
      "rewards/forensics_tiou_reward": 0.29581092298030853,
      "step": 512
    },
    {
      "completion_length": 35.71875,
      "epoch": 1.3648468708388815,
      "grad_norm": 105.51627349853516,
      "kl": 5.2421875,
      "learning_rate": 6.586666666666666e-07,
      "loss": 0.3084,
      "reward": 0.615369975566864,
      "reward_std": 0.1570507287979126,
      "reward_std_all": 0.1699170395731926,
      "rewards-max/forensics_sinkhorn_reward": 0.9692876636981964,
      "rewards-max/forensics_strict_edge_reward": 0.9364311099052429,
      "rewards-max/forensics_tiou_reward": 0.9567928910255432,
      "rewards/forensics_sinkhorn_reward": 0.4930935502052307,
      "rewards/forensics_strict_edge_reward": 0.27813243865966797,
      "rewards/forensics_tiou_reward": 0.3320325165987015,
      "step": 513
    },
    {
      "completion_length": 33.015625,
      "epoch": 1.3675099866844207,
      "grad_norm": 14.84485912322998,
      "kl": 2.12890625,
      "learning_rate": 6.58e-07,
      "loss": 0.1991,
      "reward": 0.6624055206775665,
      "reward_std": 0.10661539435386658,
      "reward_std_all": 0.12240851297974586,
      "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": 0.5716575682163239,
      "rewards/forensics_strict_edge_reward": 0.2796303555369377,
      "rewards/forensics_tiou_reward": 0.32479795813560486,
      "step": 514
    },
    {
      "completion_length": 38.28125,
      "epoch": 1.37017310252996,
      "grad_norm": 13.849119186401367,
      "kl": 1.87890625,
      "learning_rate": 6.573333333333333e-07,
      "loss": 0.228,
      "reward": 0.6373591423034668,
      "reward_std": 0.13903021067380905,
      "reward_std_all": 0.15171949565410614,
      "rewards-max/forensics_sinkhorn_reward": 0.9815817773342133,
      "rewards-max/forensics_strict_edge_reward": 0.8168056011199951,
      "rewards-max/forensics_tiou_reward": 0.974054753780365,
      "rewards/forensics_sinkhorn_reward": 0.5581998825073242,
      "rewards/forensics_strict_edge_reward": 0.20205889642238617,
      "rewards/forensics_tiou_reward": 0.3303237706422806,
      "step": 515
    },
    {
      "completion_length": 47.296875,
      "epoch": 1.3728362183754994,
      "grad_norm": 17.993173599243164,
      "kl": 1.96484375,
      "learning_rate": 6.566666666666666e-07,
      "loss": 0.2849,
      "reward": 0.6168056130409241,
      "reward_std": 0.17489252984523773,
      "reward_std_all": 0.17802518606185913,
      "rewards-max/forensics_sinkhorn_reward": 0.9864528179168701,
      "rewards-max/forensics_strict_edge_reward": 0.9574736058712006,
      "rewards-max/forensics_tiou_reward": 0.9900000095367432,
      "rewards/forensics_sinkhorn_reward": 0.5436452925205231,
      "rewards/forensics_strict_edge_reward": 0.30096233636140823,
      "rewards/forensics_tiou_reward": 0.40534253418445587,
      "step": 516
    },
    {
      "completion_length": 30.734375,
      "epoch": 1.3754993342210386,
      "grad_norm": 19.696123123168945,
      "kl": 1.7578125,
      "learning_rate": 6.56e-07,
      "loss": 0.3138,
      "reward": 0.6324408948421478,
      "reward_std": 0.17028237134218216,
      "reward_std_all": 0.17729653418064117,
      "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": 0.6059313714504242,
      "rewards/forensics_strict_edge_reward": 0.391571581363678,
      "rewards/forensics_tiou_reward": 0.4629838317632675,
      "step": 517
    },
    {
      "completion_length": 40.671875,
      "epoch": 1.378162450066578,
      "grad_norm": 17.331501007080078,
      "kl": 2.6875,
      "learning_rate": 6.553333333333333e-07,
      "loss": 0.3977,
      "reward": 0.6247396171092987,
      "reward_std": 0.1458350047469139,
      "reward_std_all": 0.16529881954193115,
      "rewards-max/forensics_sinkhorn_reward": 0.9883979856967926,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9976645410060883,
      "rewards/forensics_sinkhorn_reward": 0.49270687997341156,
      "rewards/forensics_strict_edge_reward": 0.21707461774349213,
      "rewards/forensics_tiou_reward": 0.2765844240784645,
      "step": 518
    },
    {
      "completion_length": 32.515625,
      "epoch": 1.3808255659121171,
      "grad_norm": 27.433868408203125,
      "kl": 2.48046875,
      "learning_rate": 6.546666666666665e-07,
      "loss": 0.06,
      "reward": 0.6109887659549713,
      "reward_std": 0.15452676266431808,
      "reward_std_all": 0.16672559827566147,
      "rewards-max/forensics_sinkhorn_reward": 0.988095611333847,
      "rewards-max/forensics_strict_edge_reward": 0.9574736058712006,
      "rewards-max/forensics_tiou_reward": 0.9964788854122162,
      "rewards/forensics_sinkhorn_reward": 0.5076537728309631,
      "rewards/forensics_strict_edge_reward": 0.19586776196956635,
      "rewards/forensics_tiou_reward": 0.2704614847898483,
      "step": 519
    },
    {
      "completion_length": 41.96875,
      "epoch": 1.3834886817576564,
      "grad_norm": 16.317380905151367,
      "kl": 2.53125,
      "learning_rate": 6.54e-07,
      "loss": 0.126,
      "reward": 0.6180530190467834,
      "reward_std": 0.17647841572761536,
      "reward_std_all": 0.18411589413881302,
      "rewards-max/forensics_sinkhorn_reward": 0.9885382950305939,
      "rewards-max/forensics_strict_edge_reward": 0.9782643616199493,
      "rewards-max/forensics_tiou_reward": 0.9982190430164337,
      "rewards/forensics_sinkhorn_reward": 0.5415656864643097,
      "rewards/forensics_strict_edge_reward": 0.239515483379364,
      "rewards/forensics_tiou_reward": 0.3650038242340088,
      "step": 520
    },
    {
      "completion_length": 54.421875,
      "epoch": 1.3861517976031958,
      "grad_norm": 16.327293395996094,
      "kl": 3.8203125,
      "learning_rate": 6.533333333333333e-07,
      "loss": 0.2088,
      "reward": 0.6106253564357758,
      "reward_std": 0.1883920282125473,
      "reward_std_all": 0.19122957438230515,
      "rewards-max/forensics_sinkhorn_reward": 0.9871762096881866,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9928499460220337,
      "rewards/forensics_sinkhorn_reward": 0.49550947546958923,
      "rewards/forensics_strict_edge_reward": 0.30468258261680603,
      "rewards/forensics_tiou_reward": 0.37988926470279694,
      "step": 521
    },
    {
      "completion_length": 43.375,
      "epoch": 1.388814913448735,
      "grad_norm": 18.688779830932617,
      "kl": 3.28125,
      "learning_rate": 6.526666666666666e-07,
      "loss": 0.3917,
      "reward": 0.5965788960456848,
      "reward_std": 0.17074991017580032,
      "reward_std_all": 0.1842888817191124,
      "rewards-max/forensics_sinkhorn_reward": 0.9522983729839325,
      "rewards-max/forensics_strict_edge_reward": 0.9075756669044495,
      "rewards-max/forensics_tiou_reward": 0.9151363670825958,
      "rewards/forensics_sinkhorn_reward": 0.43988843262195587,
      "rewards/forensics_strict_edge_reward": 0.2360038459300995,
      "rewards/forensics_tiou_reward": 0.27984529733657837,
      "step": 522
    },
    {
      "completion_length": 37.5625,
      "epoch": 1.3914780292942743,
      "grad_norm": 17.974735260009766,
      "kl": 3.7734375,
      "learning_rate": 6.52e-07,
      "loss": 0.4068,
      "reward": 0.5760892629623413,
      "reward_std": 0.23229122906923294,
      "reward_std_all": 0.227483868598938,
      "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": 0.458483025431633,
      "rewards/forensics_strict_edge_reward": 0.29965151846408844,
      "rewards/forensics_tiou_reward": 0.31917789578437805,
      "step": 523
    },
    {
      "completion_length": 53.3125,
      "epoch": 1.3941411451398136,
      "grad_norm": 15.552987098693848,
      "kl": 2.71875,
      "learning_rate": 6.513333333333333e-07,
      "loss": 0.2692,
      "reward": 0.6102214157581329,
      "reward_std": 0.20669179409742355,
      "reward_std_all": 0.20258542150259018,
      "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": 0.5307602882385254,
      "rewards/forensics_strict_edge_reward": 0.3605061173439026,
      "rewards/forensics_tiou_reward": 0.4032416194677353,
      "step": 524
    },
    {
      "completion_length": 29.28125,
      "epoch": 1.3968042609853528,
      "grad_norm": 22.07461929321289,
      "kl": 2.5625,
      "learning_rate": 6.506666666666666e-07,
      "loss": 0.3987,
      "reward": 0.6009897589683533,
      "reward_std": 0.1665702536702156,
      "reward_std_all": 0.1804654374718666,
      "rewards-max/forensics_sinkhorn_reward": 0.9882973730564117,
      "rewards-max/forensics_strict_edge_reward": 0.9782643616199493,
      "rewards-max/forensics_tiou_reward": 0.9972714781761169,
      "rewards/forensics_sinkhorn_reward": 0.5075846612453461,
      "rewards/forensics_strict_edge_reward": 0.23506871610879898,
      "rewards/forensics_tiou_reward": 0.3099600672721863,
      "step": 525
    },
    {
      "completion_length": 33.28125,
      "epoch": 1.399467376830892,
      "grad_norm": 12.710677146911621,
      "kl": 1.87890625,
      "learning_rate": 6.5e-07,
      "loss": 0.2472,
      "reward": 0.6575872004032135,
      "reward_std": 0.12582386657595634,
      "reward_std_all": 0.13521318137645721,
      "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": 0.5803046524524689,
      "rewards/forensics_strict_edge_reward": 0.36859847605228424,
      "rewards/forensics_tiou_reward": 0.47078582644462585,
      "step": 526
    },
    {
      "completion_length": 26.78125,
      "epoch": 1.4021304926764313,
      "grad_norm": 16.25255012512207,
      "kl": 2.11328125,
      "learning_rate": 6.493333333333333e-07,
      "loss": 0.2866,
      "reward": 0.629540354013443,
      "reward_std": 0.13678744435310364,
      "reward_std_all": 0.15801063179969788,
      "rewards-max/forensics_sinkhorn_reward": 0.9879033863544464,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9957192242145538,
      "rewards/forensics_sinkhorn_reward": 0.5285200923681259,
      "rewards/forensics_strict_edge_reward": 0.22703854739665985,
      "rewards/forensics_tiou_reward": 0.33272387087345123,
      "step": 527
    },
    {
      "completion_length": 38.0625,
      "epoch": 1.4047936085219708,
      "grad_norm": 15.549341201782227,
      "kl": 2.6171875,
      "learning_rate": 6.486666666666666e-07,
      "loss": 0.2922,
      "reward": 0.5976117551326752,
      "reward_std": 0.15673194080591202,
      "reward_std_all": 0.1746029481291771,
      "rewards-max/forensics_sinkhorn_reward": 0.9874636232852936,
      "rewards-max/forensics_strict_edge_reward": 0.9256840348243713,
      "rewards-max/forensics_tiou_reward": 0.993991881608963,
      "rewards/forensics_sinkhorn_reward": 0.46311990916728973,
      "rewards/forensics_strict_edge_reward": 0.18167812377214432,
      "rewards/forensics_tiou_reward": 0.22960121929645538,
      "step": 528
    },
    {
      "completion_length": 40.859375,
      "epoch": 1.40745672436751,
      "grad_norm": 16.16999626159668,
      "kl": 2.6171875,
      "learning_rate": 6.48e-07,
      "loss": 0.3216,
      "reward": 0.6392709016799927,
      "reward_std": 0.19282016158103943,
      "reward_std_all": 0.18125753104686737,
      "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": 0.5363958328962326,
      "rewards/forensics_strict_edge_reward": 0.29555515944957733,
      "rewards/forensics_tiou_reward": 0.3558591604232788,
      "step": 529
    },
    {
      "completion_length": 33.78125,
      "epoch": 1.4101198402130493,
      "grad_norm": 18.54907989501953,
      "kl": 2.453125,
      "learning_rate": 6.473333333333333e-07,
      "loss": 0.2516,
      "reward": 0.6566784381866455,
      "reward_std": 0.13509279489517212,
      "reward_std_all": 0.14283861219882965,
      "rewards-max/forensics_sinkhorn_reward": 0.9884359538555145,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9978154301643372,
      "rewards/forensics_sinkhorn_reward": 0.5799121260643005,
      "rewards/forensics_strict_edge_reward": 0.3054138571023941,
      "rewards/forensics_tiou_reward": 0.39990636706352234,
      "step": 530
    },
    {
      "completion_length": 66.078125,
      "epoch": 1.4127829560585885,
      "grad_norm": 15.812477111816406,
      "kl": 3.3046875,
      "learning_rate": 6.466666666666666e-07,
      "loss": 0.2691,
      "reward": 0.642578661441803,
      "reward_std": 0.16361235827207565,
      "reward_std_all": 0.16405071318149567,
      "rewards-max/forensics_sinkhorn_reward": 0.9882243871688843,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9969837069511414,
      "rewards/forensics_sinkhorn_reward": 0.5626039505004883,
      "rewards/forensics_strict_edge_reward": 0.31241072714328766,
      "rewards/forensics_tiou_reward": 0.38109907507896423,
      "step": 531
    },
    {
      "completion_length": 37.0,
      "epoch": 1.4154460719041277,
      "grad_norm": 15.541542053222656,
      "kl": 2.01953125,
      "learning_rate": 6.46e-07,
      "loss": 0.1501,
      "reward": 0.6333237290382385,
      "reward_std": 0.15655987337231636,
      "reward_std_all": 0.1593022160232067,
      "rewards-max/forensics_sinkhorn_reward": 0.9880931079387665,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.996462494134903,
      "rewards/forensics_sinkhorn_reward": 0.5628612041473389,
      "rewards/forensics_strict_edge_reward": 0.22402117401361465,
      "rewards/forensics_tiou_reward": 0.3391023501753807,
      "step": 532
    },
    {
      "completion_length": 51.265625,
      "epoch": 1.4181091877496672,
      "grad_norm": 14.763690948486328,
      "kl": 2.39453125,
      "learning_rate": 6.453333333333333e-07,
      "loss": 0.2761,
      "reward": 0.6209534406661987,
      "reward_std": 0.15036515146493912,
      "reward_std_all": 0.1637982241809368,
      "rewards-max/forensics_sinkhorn_reward": 0.9886393845081329,
      "rewards-max/forensics_strict_edge_reward": 0.9782643616199493,
      "rewards-max/forensics_tiou_reward": 0.9986159205436707,
      "rewards/forensics_sinkhorn_reward": 0.590804249048233,
      "rewards/forensics_strict_edge_reward": 0.3290303647518158,
      "rewards/forensics_tiou_reward": 0.39015865325927734,
      "step": 533
    },
    {
      "completion_length": 49.59375,
      "epoch": 1.4207723035952065,
      "grad_norm": 21.261404037475586,
      "kl": 3.328125,
      "learning_rate": 6.446666666666666e-07,
      "loss": 0.3881,
      "reward": 0.613479346036911,
      "reward_std": 0.1749841868877411,
      "reward_std_all": 0.1821914240717888,
      "rewards-max/forensics_sinkhorn_reward": 0.9878709316253662,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9955896735191345,
      "rewards/forensics_sinkhorn_reward": 0.47194355726242065,
      "rewards/forensics_strict_edge_reward": 0.16371122747659683,
      "rewards/forensics_tiou_reward": 0.24907919019460678,
      "step": 534
    },
    {
      "completion_length": 38.734375,
      "epoch": 1.4234354194407457,
      "grad_norm": 16.336559295654297,
      "kl": 2.640625,
      "learning_rate": 6.44e-07,
      "loss": 0.1937,
      "reward": 0.5659171342849731,
      "reward_std": 0.1324193775653839,
      "reward_std_all": 0.17976199835538864,
      "rewards-max/forensics_sinkhorn_reward": 0.950211763381958,
      "rewards-max/forensics_strict_edge_reward": 0.7624560296535492,
      "rewards-max/forensics_tiou_reward": 0.9070352613925934,
      "rewards/forensics_sinkhorn_reward": 0.4804915189743042,
      "rewards/forensics_strict_edge_reward": 0.20979265868663788,
      "rewards/forensics_tiou_reward": 0.2707132250070572,
      "step": 535
    },
    {
      "completion_length": 28.859375,
      "epoch": 1.426098535286285,
      "grad_norm": 12.69938850402832,
      "kl": 1.91796875,
      "learning_rate": 6.433333333333332e-07,
      "loss": 0.1223,
      "reward": 0.625084400177002,
      "reward_std": 0.1452723816037178,
      "reward_std_all": 0.16246627271175385,
      "rewards-max/forensics_sinkhorn_reward": 0.9878261983394623,
      "rewards-max/forensics_strict_edge_reward": 0.9464850425720215,
      "rewards-max/forensics_tiou_reward": 0.9954107999801636,
      "rewards/forensics_sinkhorn_reward": 0.5755123496055603,
      "rewards/forensics_strict_edge_reward": 0.26543866842985153,
      "rewards/forensics_tiou_reward": 0.34690865874290466,
      "step": 536
    },
    {
      "completion_length": 79.046875,
      "epoch": 1.4287616511318242,
      "grad_norm": 16.333450317382812,
      "kl": 3.21875,
      "learning_rate": 6.426666666666667e-07,
      "loss": 0.3863,
      "reward": 0.618906170129776,
      "reward_std": 0.19423852860927582,
      "reward_std_all": 0.19298996031284332,
      "rewards-max/forensics_sinkhorn_reward": 0.9877889454364777,
      "rewards-max/forensics_strict_edge_reward": 0.9565287232398987,
      "rewards-max/forensics_tiou_reward": 0.9952700138092041,
      "rewards/forensics_sinkhorn_reward": 0.5326056182384491,
      "rewards/forensics_strict_edge_reward": 0.30421918630599976,
      "rewards/forensics_tiou_reward": 0.33962132781744003,
      "step": 537
    },
    {
      "completion_length": 34.0625,
      "epoch": 1.4314247669773636,
      "grad_norm": 18.888769149780273,
      "kl": 2.52734375,
      "learning_rate": 6.42e-07,
      "loss": 0.2725,
      "reward": 0.6788230240345001,
      "reward_std": 0.11231854557991028,
      "reward_std_all": 0.11766528710722923,
      "rewards-max/forensics_sinkhorn_reward": 0.9858444333076477,
      "rewards-max/forensics_strict_edge_reward": 0.9565287232398987,
      "rewards-max/forensics_tiou_reward": 0.9914174675941467,
      "rewards/forensics_sinkhorn_reward": 0.5497338175773621,
      "rewards/forensics_strict_edge_reward": 0.25458288192749023,
      "rewards/forensics_tiou_reward": 0.35075971484184265,
      "step": 538
    },
    {
      "completion_length": 56.609375,
      "epoch": 1.4340878828229027,
      "grad_norm": 18.81692123413086,
      "kl": 3.7421875,
      "learning_rate": 6.413333333333333e-07,
      "loss": 0.294,
      "reward": 0.6480548977851868,
      "reward_std": 0.16099592298269272,
      "reward_std_all": 0.1563466191291809,
      "rewards-max/forensics_sinkhorn_reward": 0.9653716087341309,
      "rewards-max/forensics_strict_edge_reward": 0.8663162291049957,
      "rewards-max/forensics_tiou_reward": 0.9512645602226257,
      "rewards/forensics_sinkhorn_reward": 0.510808914899826,
      "rewards/forensics_strict_edge_reward": 0.27843740582466125,
      "rewards/forensics_tiou_reward": 0.33277571201324463,
      "step": 539
    },
    {
      "completion_length": 37.3125,
      "epoch": 1.4367509986684421,
      "grad_norm": 17.66356658935547,
      "kl": 2.5078125,
      "learning_rate": 6.406666666666667e-07,
      "loss": 0.1644,
      "reward": 0.6809448897838593,
      "reward_std": 0.08492368459701538,
      "reward_std_all": 0.10342716425657272,
      "rewards-max/forensics_sinkhorn_reward": 0.988531768321991,
      "rewards-max/forensics_strict_edge_reward": 0.947419673204422,
      "rewards-max/forensics_tiou_reward": 0.9981916844844818,
      "rewards/forensics_sinkhorn_reward": 0.6036771237850189,
      "rewards/forensics_strict_edge_reward": 0.28786514699459076,
      "rewards/forensics_tiou_reward": 0.3448773920536041,
      "step": 540
    },
    {
      "completion_length": 85.21875,
      "epoch": 1.4394141145139814,
      "grad_norm": 17.38348960876465,
      "kl": 3.1015625,
      "learning_rate": 6.4e-07,
      "loss": 0.2143,
      "reward": 0.6641793847084045,
      "reward_std": 0.15615974552929401,
      "reward_std_all": 0.14918522536754608,
      "rewards-max/forensics_sinkhorn_reward": 0.9854156970977783,
      "rewards-max/forensics_strict_edge_reward": 0.9375866651535034,
      "rewards-max/forensics_tiou_reward": 0.9860146939754486,
      "rewards/forensics_sinkhorn_reward": 0.5737526416778564,
      "rewards/forensics_strict_edge_reward": 0.3447607681155205,
      "rewards/forensics_tiou_reward": 0.4361061304807663,
      "step": 541
    },
    {
      "completion_length": 61.328125,
      "epoch": 1.4420772303595206,
      "grad_norm": 19.50615882873535,
      "kl": 3.203125,
      "learning_rate": 6.393333333333333e-07,
      "loss": 0.2826,
      "reward": 0.5369046926498413,
      "reward_std": 0.13513819128274918,
      "reward_std_all": 0.19295214116573334,
      "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": 0.43162600696086884,
      "rewards/forensics_strict_edge_reward": 0.182152159512043,
      "rewards/forensics_tiou_reward": 0.24836517870426178,
      "step": 542
    },
    {
      "completion_length": 55.921875,
      "epoch": 1.4447403462050599,
      "grad_norm": 15.496696472167969,
      "kl": 2.8828125,
      "learning_rate": 6.386666666666667e-07,
      "loss": 0.2666,
      "reward": 0.696435272693634,
      "reward_std": 0.16721542924642563,
      "reward_std_all": 0.14905792474746704,
      "rewards-max/forensics_sinkhorn_reward": 0.9877035021781921,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9949272572994232,
      "rewards/forensics_sinkhorn_reward": 0.6533665657043457,
      "rewards/forensics_strict_edge_reward": 0.34465356171131134,
      "rewards/forensics_tiou_reward": 0.4889581501483917,
      "step": 543
    },
    {
      "completion_length": 54.109375,
      "epoch": 1.447403462050599,
      "grad_norm": 17.04315948486328,
      "kl": 2.7265625,
      "learning_rate": 6.38e-07,
      "loss": 0.3078,
      "reward": 0.6778846681118011,
      "reward_std": 0.10295921191573143,
      "reward_std_all": 0.11333711817860603,
      "rewards-max/forensics_sinkhorn_reward": 0.987535297870636,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9942653775215149,
      "rewards/forensics_sinkhorn_reward": 0.5197346210479736,
      "rewards/forensics_strict_edge_reward": 0.25404688715934753,
      "rewards/forensics_tiou_reward": 0.3483987897634506,
      "step": 544
    },
    {
      "completion_length": 58.875,
      "epoch": 1.4500665778961386,
      "grad_norm": 15.92469310760498,
      "kl": 2.5625,
      "learning_rate": 6.373333333333333e-07,
      "loss": 0.3021,
      "reward": 0.6633304357528687,
      "reward_std": 0.13827787339687347,
      "reward_std_all": 0.14397092163562775,
      "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": 0.5635411441326141,
      "rewards/forensics_strict_edge_reward": 0.2749190479516983,
      "rewards/forensics_tiou_reward": 0.3386520594358444,
      "step": 545
    },
    {
      "completion_length": 54.859375,
      "epoch": 1.4527296937416778,
      "grad_norm": 20.996200561523438,
      "kl": 2.8203125,
      "learning_rate": 6.366666666666667e-07,
      "loss": 0.194,
      "reward": 0.6322159469127655,
      "reward_std": 0.1519785299897194,
      "reward_std_all": 0.15675007551908493,
      "rewards-max/forensics_sinkhorn_reward": 0.9871451556682587,
      "rewards-max/forensics_strict_edge_reward": 0.946017861366272,
      "rewards-max/forensics_tiou_reward": 0.9927400946617126,
      "rewards/forensics_sinkhorn_reward": 0.4939803183078766,
      "rewards/forensics_strict_edge_reward": 0.23735610395669937,
      "rewards/forensics_tiou_reward": 0.31053851544857025,
      "step": 546
    },
    {
      "completion_length": 48.359375,
      "epoch": 1.455392809587217,
      "grad_norm": 16.599992752075195,
      "kl": 2.62890625,
      "learning_rate": 6.36e-07,
      "loss": 0.2255,
      "reward": 0.6500250101089478,
      "reward_std": 0.15722305327653885,
      "reward_std_all": 0.1584065705537796,
      "rewards-max/forensics_sinkhorn_reward": 0.9888587892055511,
      "rewards-max/forensics_strict_edge_reward": 0.9890114367008209,
      "rewards-max/forensics_tiou_reward": 0.9994809329509735,
      "rewards/forensics_sinkhorn_reward": 0.6041995882987976,
      "rewards/forensics_strict_edge_reward": 0.32236096262931824,
      "rewards/forensics_tiou_reward": 0.45142166316509247,
      "step": 547
    },
    {
      "completion_length": 37.1875,
      "epoch": 1.4580559254327563,
      "grad_norm": 27.04420280456543,
      "kl": 2.421875,
      "learning_rate": 6.353333333333333e-07,
      "loss": 0.2049,
      "reward": 0.6884208917617798,
      "reward_std": 0.11352280899882317,
      "reward_std_all": 0.11640800908207893,
      "rewards-max/forensics_sinkhorn_reward": 0.9658735692501068,
      "rewards-max/forensics_strict_edge_reward": 0.8668389916419983,
      "rewards-max/forensics_tiou_reward": 0.9499287009239197,
      "rewards/forensics_sinkhorn_reward": 0.56702721118927,
      "rewards/forensics_strict_edge_reward": 0.28392428159713745,
      "rewards/forensics_tiou_reward": 0.34848473966121674,
      "step": 548
    },
    {
      "completion_length": 56.390625,
      "epoch": 1.4607190412782955,
      "grad_norm": 21.44302749633789,
      "kl": 2.7734375,
      "learning_rate": 6.346666666666666e-07,
      "loss": 0.4151,
      "reward": 0.6357203722000122,
      "reward_std": 0.14653639495372772,
      "reward_std_all": 0.15365789830684662,
      "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": 0.5200541615486145,
      "rewards/forensics_strict_edge_reward": 0.25889377295970917,
      "rewards/forensics_tiou_reward": 0.32766225934028625,
      "step": 549
    },
    {
      "completion_length": 50.015625,
      "epoch": 1.463382157123835,
      "grad_norm": 17.27256965637207,
      "kl": 2.453125,
      "learning_rate": 6.34e-07,
      "loss": 0.3153,
      "reward": 0.6286438703536987,
      "reward_std": 0.19382549822330475,
      "reward_std_all": 0.18427307903766632,
      "rewards-max/forensics_sinkhorn_reward": 0.986892819404602,
      "rewards-max/forensics_strict_edge_reward": 0.9256840348243713,
      "rewards-max/forensics_tiou_reward": 0.9917616248130798,
      "rewards/forensics_sinkhorn_reward": 0.5201915502548218,
      "rewards/forensics_strict_edge_reward": 0.2722381502389908,
      "rewards/forensics_tiou_reward": 0.3671036511659622,
      "step": 550
    },
    {
      "completion_length": 27.09375,
      "epoch": 1.466045272969374,
      "grad_norm": 17.076400756835938,
      "kl": 1.62890625,
      "learning_rate": 6.333333333333332e-07,
      "loss": 0.1516,
      "reward": 0.5947076082229614,
      "reward_std": 0.15732884407043457,
      "reward_std_all": 0.18215970695018768,
      "rewards-max/forensics_sinkhorn_reward": 0.9880246818065643,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9961954653263092,
      "rewards/forensics_sinkhorn_reward": 0.508614182472229,
      "rewards/forensics_strict_edge_reward": 0.16648868843913078,
      "rewards/forensics_tiou_reward": 0.24136712402105331,
      "step": 551
    },
    {
      "completion_length": 42.828125,
      "epoch": 1.4687083888149135,
      "grad_norm": 18.250024795532227,
      "kl": 1.91796875,
      "learning_rate": 6.326666666666667e-07,
      "loss": 0.2737,
      "reward": 0.6992999911308289,
      "reward_std": 0.11192836612462997,
      "reward_std_all": 0.10864407569169998,
      "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": 0.6011699438095093,
      "rewards/forensics_strict_edge_reward": 0.31818898767232895,
      "rewards/forensics_tiou_reward": 0.38255202770233154,
      "step": 552
    },
    {
      "completion_length": 63.984375,
      "epoch": 1.4713715046604527,
      "grad_norm": 19.452573776245117,
      "kl": 2.59375,
      "learning_rate": 6.319999999999999e-07,
      "loss": 0.2121,
      "reward": 0.624436765909195,
      "reward_std": 0.16888634860515594,
      "reward_std_all": 0.17427454143762589,
      "rewards-max/forensics_sinkhorn_reward": 0.9881693422794342,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.996766209602356,
      "rewards/forensics_sinkhorn_reward": 0.5104593634605408,
      "rewards/forensics_strict_edge_reward": 0.24104610830545425,
      "rewards/forensics_tiou_reward": 0.32773807644844055,
      "step": 553
    },
    {
      "completion_length": 35.578125,
      "epoch": 1.474034620505992,
      "grad_norm": 17.697772979736328,
      "kl": 1.64453125,
      "learning_rate": 6.313333333333333e-07,
      "loss": 0.0479,
      "reward": 0.6448699831962585,
      "reward_std": 0.13478543981909752,
      "reward_std_all": 0.1465575397014618,
      "rewards-max/forensics_sinkhorn_reward": 0.9874885976314545,
      "rewards-max/forensics_strict_edge_reward": 0.946017861366272,
      "rewards-max/forensics_tiou_reward": 0.9940907657146454,
      "rewards/forensics_sinkhorn_reward": 0.5505607426166534,
      "rewards/forensics_strict_edge_reward": 0.16168253868818283,
      "rewards/forensics_tiou_reward": 0.26338230073451996,
      "step": 554
    },
    {
      "completion_length": 45.515625,
      "epoch": 1.4766977363515312,
      "grad_norm": 17.16213607788086,
      "kl": 1.94140625,
      "learning_rate": 6.306666666666666e-07,
      "loss": 0.3029,
      "reward": 0.6021584868431091,
      "reward_std": 0.15437154099345207,
      "reward_std_all": 0.17240840569138527,
      "rewards-max/forensics_sinkhorn_reward": 0.9878435432910919,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9954814314842224,
      "rewards/forensics_sinkhorn_reward": 0.48988623917102814,
      "rewards/forensics_strict_edge_reward": 0.19919785112142563,
      "rewards/forensics_tiou_reward": 0.27394843101501465,
      "step": 555
    },
    {
      "completion_length": 27.9375,
      "epoch": 1.4793608521970705,
      "grad_norm": 23.192846298217773,
      "kl": 1.703125,
      "learning_rate": 6.3e-07,
      "loss": 0.1179,
      "reward": 0.6631215512752533,
      "reward_std": 0.09221431612968445,
      "reward_std_all": 0.11491996049880981,
      "rewards-max/forensics_sinkhorn_reward": 0.9684731364250183,
      "rewards-max/forensics_strict_edge_reward": 0.8983768224716187,
      "rewards-max/forensics_tiou_reward": 0.9519271552562714,
      "rewards/forensics_sinkhorn_reward": 0.5687934160232544,
      "rewards/forensics_strict_edge_reward": 0.21660929918289185,
      "rewards/forensics_tiou_reward": 0.3117527365684509,
      "step": 556
    },
    {
      "completion_length": 37.109375,
      "epoch": 1.48202396804261,
      "grad_norm": 16.240434646606445,
      "kl": 2.26953125,
      "learning_rate": 6.293333333333333e-07,
      "loss": 0.2174,
      "reward": 0.6395964920520782,
      "reward_std": 0.14696619659662247,
      "reward_std_all": 0.15918277204036713,
      "rewards-max/forensics_sinkhorn_reward": 0.9707283973693848,
      "rewards-max/forensics_strict_edge_reward": 0.8445805907249451,
      "rewards-max/forensics_tiou_reward": 0.9572918713092804,
      "rewards/forensics_sinkhorn_reward": 0.541534423828125,
      "rewards/forensics_strict_edge_reward": 0.16881543397903442,
      "rewards/forensics_tiou_reward": 0.29285305738449097,
      "step": 557
    },
    {
      "completion_length": 48.078125,
      "epoch": 1.4846870838881492,
      "grad_norm": 15.728469848632812,
      "kl": 2.421875,
      "learning_rate": 6.286666666666667e-07,
      "loss": 0.3529,
      "reward": 0.6651875674724579,
      "reward_std": 0.1249389536678791,
      "reward_std_all": 0.13310247287154198,
      "rewards-max/forensics_sinkhorn_reward": 0.9870686233043671,
      "rewards-max/forensics_strict_edge_reward": 0.8968285918235779,
      "rewards-max/forensics_tiou_reward": 0.9924325942993164,
      "rewards/forensics_sinkhorn_reward": 0.5487731099128723,
      "rewards/forensics_strict_edge_reward": 0.24603316187858582,
      "rewards/forensics_tiou_reward": 0.30813083052635193,
      "step": 558
    },
    {
      "completion_length": 42.3125,
      "epoch": 1.4873501997336884,
      "grad_norm": 14.33868408203125,
      "kl": 2.8125,
      "learning_rate": 6.28e-07,
      "loss": 0.375,
      "reward": 0.6917630136013031,
      "reward_std": 0.08635996840894222,
      "reward_std_all": 0.10045216232538223,
      "rewards-max/forensics_sinkhorn_reward": 0.9878926873207092,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9956751465797424,
      "rewards/forensics_sinkhorn_reward": 0.6236525177955627,
      "rewards/forensics_strict_edge_reward": 0.45725639164447784,
      "rewards/forensics_tiou_reward": 0.5240758210420609,
      "step": 559
    },
    {
      "completion_length": 37.71875,
      "epoch": 1.4900133155792277,
      "grad_norm": 12.359735488891602,
      "kl": 2.6484375,
      "learning_rate": 6.273333333333333e-07,
      "loss": 0.2449,
      "reward": 0.663689523935318,
      "reward_std": 0.10899756103754044,
      "reward_std_all": 0.12332706898450851,
      "rewards-max/forensics_sinkhorn_reward": 0.9835458695888519,
      "rewards-max/forensics_strict_edge_reward": 0.9565287232398987,
      "rewards-max/forensics_tiou_reward": 0.985905647277832,
      "rewards/forensics_sinkhorn_reward": 0.634440541267395,
      "rewards/forensics_strict_edge_reward": 0.38327164947986603,
      "rewards/forensics_tiou_reward": 0.4925384521484375,
      "step": 560
    },
    {
      "completion_length": 45.265625,
      "epoch": 1.492676431424767,
      "grad_norm": 23.67380142211914,
      "kl": 2.546875,
      "learning_rate": 6.266666666666667e-07,
      "loss": 0.1798,
      "reward": 0.6374432146549225,
      "reward_std": 0.12403252720832825,
      "reward_std_all": 0.14818713814020157,
      "rewards-max/forensics_sinkhorn_reward": 0.9665913879871368,
      "rewards-max/forensics_strict_edge_reward": 0.8135445415973663,
      "rewards-max/forensics_tiou_reward": 0.94741490483284,
      "rewards/forensics_sinkhorn_reward": 0.5171256959438324,
      "rewards/forensics_strict_edge_reward": 0.2569374740123749,
      "rewards/forensics_tiou_reward": 0.33807186782360077,
      "step": 561
    },
    {
      "completion_length": 30.765625,
      "epoch": 1.4953395472703064,
      "grad_norm": 16.5458984375,
      "kl": 2.03125,
      "learning_rate": 6.26e-07,
      "loss": 0.1807,
      "reward": 0.7357144951820374,
      "reward_std": 0.07320205494761467,
      "reward_std_all": 0.07664249837398529,
      "rewards-max/forensics_sinkhorn_reward": 0.9878228902816772,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9953997731208801,
      "rewards/forensics_sinkhorn_reward": 0.6871303617954254,
      "rewards/forensics_strict_edge_reward": 0.47352659702301025,
      "rewards/forensics_tiou_reward": 0.540841817855835,
      "step": 562
    },
    {
      "completion_length": 36.484375,
      "epoch": 1.4980026631158454,
      "grad_norm": 17.152385711669922,
      "kl": 2.1875,
      "learning_rate": 6.253333333333333e-07,
      "loss": 0.2759,
      "reward": 0.6745966672897339,
      "reward_std": 0.14277107268571854,
      "reward_std_all": 0.1428937464952469,
      "rewards-max/forensics_sinkhorn_reward": 0.972606897354126,
      "rewards-max/forensics_strict_edge_reward": 0.9169811904430389,
      "rewards-max/forensics_tiou_reward": 0.9617118537425995,
      "rewards/forensics_sinkhorn_reward": 0.5983076393604279,
      "rewards/forensics_strict_edge_reward": 0.3350983262062073,
      "rewards/forensics_tiou_reward": 0.4348982125520706,
      "step": 563
    },
    {
      "completion_length": 84.078125,
      "epoch": 1.5006657789613849,
      "grad_norm": 14.37605094909668,
      "kl": 3.9921875,
      "learning_rate": 6.246666666666667e-07,
      "loss": 0.3488,
      "reward": 0.6731078922748566,
      "reward_std": 0.15350355952978134,
      "reward_std_all": 0.14719003811478615,
      "rewards-max/forensics_sinkhorn_reward": 0.9872417151927948,
      "rewards-max/forensics_strict_edge_reward": 0.8066537380218506,
      "rewards-max/forensics_tiou_reward": 0.9931315183639526,
      "rewards/forensics_sinkhorn_reward": 0.5824321806430817,
      "rewards/forensics_strict_edge_reward": 0.30469535291194916,
      "rewards/forensics_tiou_reward": 0.4237712621688843,
      "step": 564
    },
    {
      "completion_length": 45.25,
      "epoch": 1.503328894806924,
      "grad_norm": 15.503963470458984,
      "kl": 3.1328125,
      "learning_rate": 6.24e-07,
      "loss": 0.156,
      "reward": 0.6775383353233337,
      "reward_std": 0.10319233685731888,
      "reward_std_all": 0.11069789528846741,
      "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": 0.6013030707836151,
      "rewards/forensics_strict_edge_reward": 0.27780695259571075,
      "rewards/forensics_tiou_reward": 0.43413737416267395,
      "step": 565
    },
    {
      "completion_length": 69.40625,
      "epoch": 1.5059920106524634,
      "grad_norm": 17.47755241394043,
      "kl": 3.4609375,
      "learning_rate": 6.233333333333332e-07,
      "loss": 0.2676,
      "reward": 0.6657887399196625,
      "reward_std": 0.09232650324702263,
      "reward_std_all": 0.11688345298171043,
      "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": 0.6420161724090576,
      "rewards/forensics_strict_edge_reward": 0.3843877613544464,
      "rewards/forensics_tiou_reward": 0.5376295447349548,
      "step": 566
    },
    {
      "completion_length": 71.59375,
      "epoch": 1.5086551264980028,
      "grad_norm": 25.32414436340332,
      "kl": 4.96875,
      "learning_rate": 6.226666666666667e-07,
      "loss": 0.3832,
      "reward": 0.6185276210308075,
      "reward_std": 0.1721840351819992,
      "reward_std_all": 0.17874747514724731,
      "rewards-max/forensics_sinkhorn_reward": 0.9876960515975952,
      "rewards-max/forensics_strict_edge_reward": 0.9677534997463226,
      "rewards-max/forensics_tiou_reward": 0.9949109852313995,
      "rewards/forensics_sinkhorn_reward": 0.5225644409656525,
      "rewards/forensics_strict_edge_reward": 0.32198596000671387,
      "rewards/forensics_tiou_reward": 0.4194065034389496,
      "step": 567
    },
    {
      "completion_length": 73.25,
      "epoch": 1.5113182423435418,
      "grad_norm": 17.4803524017334,
      "kl": 3.984375,
      "learning_rate": 6.219999999999999e-07,
      "loss": 0.2669,
      "reward": 0.5634298622608185,
      "reward_std": 0.14285922795534134,
      "reward_std_all": 0.18687742203474045,
      "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": 0.4696045517921448,
      "rewards/forensics_strict_edge_reward": 0.215187668800354,
      "rewards/forensics_tiou_reward": 0.25503791868686676,
      "step": 568
    },
    {
      "completion_length": 56.953125,
      "epoch": 1.5139813581890813,
      "grad_norm": 16.594877243041992,
      "kl": 4.1015625,
      "learning_rate": 6.213333333333333e-07,
      "loss": 0.4265,
      "reward": 0.6624949872493744,
      "reward_std": 0.14400744438171387,
      "reward_std_all": 0.14311690628528595,
      "rewards-max/forensics_sinkhorn_reward": 0.9855078458786011,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9913082420825958,
      "rewards/forensics_sinkhorn_reward": 0.577239453792572,
      "rewards/forensics_strict_edge_reward": 0.3321339339017868,
      "rewards/forensics_tiou_reward": 0.41465751826763153,
      "step": 569
    },
    {
      "completion_length": 100.203125,
      "epoch": 1.5166444740346205,
      "grad_norm": 19.157438278198242,
      "kl": 4.09375,
      "learning_rate": 6.206666666666666e-07,
      "loss": 0.3535,
      "reward": 0.6271226704120636,
      "reward_std": 0.15533588826656342,
      "reward_std_all": 0.16836372017860413,
      "rewards-max/forensics_sinkhorn_reward": 0.9882164895534515,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9969494938850403,
      "rewards/forensics_sinkhorn_reward": 0.47668901085853577,
      "rewards/forensics_strict_edge_reward": 0.22555901110172272,
      "rewards/forensics_tiou_reward": 0.2632320150732994,
      "step": 570
    },
    {
      "completion_length": 79.265625,
      "epoch": 1.5193075898801598,
      "grad_norm": 20.593109130859375,
      "kl": 3.8359375,
      "learning_rate": 6.2e-07,
      "loss": 0.3299,
      "reward": 0.6435001790523529,
      "reward_std": 0.16978739947080612,
      "reward_std_all": 0.16857578605413437,
      "rewards-max/forensics_sinkhorn_reward": 0.9886162281036377,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9985252320766449,
      "rewards/forensics_sinkhorn_reward": 0.5928777456283569,
      "rewards/forensics_strict_edge_reward": 0.4496413469314575,
      "rewards/forensics_tiou_reward": 0.5177260786294937,
      "step": 571
    },
    {
      "completion_length": 38.96875,
      "epoch": 1.521970705725699,
      "grad_norm": 16.268362045288086,
      "kl": 2.58203125,
      "learning_rate": 6.193333333333332e-07,
      "loss": 0.2782,
      "reward": 0.6294092535972595,
      "reward_std": 0.13664758577942848,
      "reward_std_all": 0.1465875767171383,
      "rewards-max/forensics_sinkhorn_reward": 0.9883731007575989,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9975663125514984,
      "rewards/forensics_sinkhorn_reward": 0.5223825871944427,
      "rewards/forensics_strict_edge_reward": 0.3026476576924324,
      "rewards/forensics_tiou_reward": 0.3825496584177017,
      "step": 572
    },
    {
      "completion_length": 48.546875,
      "epoch": 1.5246338215712383,
      "grad_norm": 23.867067337036133,
      "kl": 2.4296875,
      "learning_rate": 6.186666666666667e-07,
      "loss": 0.1828,
      "reward": 0.6562061309814453,
      "reward_std": 0.10041225701570511,
      "reward_std_all": 0.12015161663293839,
      "rewards-max/forensics_sinkhorn_reward": 0.9879714548587799,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9959829151630402,
      "rewards/forensics_sinkhorn_reward": 0.6074094176292419,
      "rewards/forensics_strict_edge_reward": 0.37668710947036743,
      "rewards/forensics_tiou_reward": 0.4318414032459259,
      "step": 573
    },
    {
      "completion_length": 50.3125,
      "epoch": 1.5272969374167777,
      "grad_norm": 14.319981575012207,
      "kl": 2.703125,
      "learning_rate": 6.18e-07,
      "loss": 0.2937,
      "reward": 0.6077819466590881,
      "reward_std": 0.14593957364559174,
      "reward_std_all": 0.17439783364534378,
      "rewards-max/forensics_sinkhorn_reward": 0.9865784049034119,
      "rewards-max/forensics_strict_edge_reward": 0.9565287232398987,
      "rewards-max/forensics_tiou_reward": 0.9905086159706116,
      "rewards/forensics_sinkhorn_reward": 0.5149780064821243,
      "rewards/forensics_strict_edge_reward": 0.254137858748436,
      "rewards/forensics_tiou_reward": 0.35432593524456024,
      "step": 574
    },
    {
      "completion_length": 36.578125,
      "epoch": 1.5299600532623168,
      "grad_norm": 20.519384384155273,
      "kl": 2.234375,
      "learning_rate": 6.173333333333333e-07,
      "loss": 0.2687,
      "reward": 0.6610526740550995,
      "reward_std": 0.14971193671226501,
      "reward_std_all": 0.15074921771883965,
      "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": 0.5785317420959473,
      "rewards/forensics_strict_edge_reward": 0.2862190753221512,
      "rewards/forensics_tiou_reward": 0.3654133826494217,
      "step": 575
    },
    {
      "completion_length": 64.453125,
      "epoch": 1.5326231691078562,
      "grad_norm": 16.172359466552734,
      "kl": 3.0078125,
      "learning_rate": 6.166666666666667e-07,
      "loss": 0.2133,
      "reward": 0.6475973427295685,
      "reward_std": 0.1722138300538063,
      "reward_std_all": 0.16737978905439377,
      "rewards-max/forensics_sinkhorn_reward": 0.9792178869247437,
      "rewards-max/forensics_strict_edge_reward": 0.9375866651535034,
      "rewards-max/forensics_tiou_reward": 0.9798251092433929,
      "rewards/forensics_sinkhorn_reward": 0.5727631449699402,
      "rewards/forensics_strict_edge_reward": 0.29170088469982147,
      "rewards/forensics_tiou_reward": 0.387710839509964,
      "step": 576
    },
    {
      "completion_length": 33.015625,
      "epoch": 1.5352862849533955,
      "grad_norm": 16.4523868560791,
      "kl": 2.0234375,
      "learning_rate": 6.16e-07,
      "loss": 0.2429,
      "reward": 0.6815073490142822,
      "reward_std": 0.08830128610134125,
      "reward_std_all": 0.10570405423641205,
      "rewards-max/forensics_sinkhorn_reward": 0.9879443049430847,
      "rewards-max/forensics_strict_edge_reward": 0.9782643616199493,
      "rewards-max/forensics_tiou_reward": 0.995876282453537,
      "rewards/forensics_sinkhorn_reward": 0.6028188169002533,
      "rewards/forensics_strict_edge_reward": 0.33797116577625275,
      "rewards/forensics_tiou_reward": 0.4087231457233429,
      "step": 577
    },
    {
      "completion_length": 59.1875,
      "epoch": 1.5379494007989347,
      "grad_norm": 17.201871871948242,
      "kl": 3.1171875,
      "learning_rate": 6.153333333333333e-07,
      "loss": 0.315,
      "reward": 0.6025400161743164,
      "reward_std": 0.1648920252919197,
      "reward_std_all": 0.1834937110543251,
      "rewards-max/forensics_sinkhorn_reward": 0.9735012948513031,
      "rewards-max/forensics_strict_edge_reward": 0.9357379674911499,
      "rewards-max/forensics_tiou_reward": 0.9631001353263855,
      "rewards/forensics_sinkhorn_reward": 0.47549350559711456,
      "rewards/forensics_strict_edge_reward": 0.26282183825969696,
      "rewards/forensics_tiou_reward": 0.2891925275325775,
      "step": 578
    },
    {
      "completion_length": 40.828125,
      "epoch": 1.5406125166444742,
      "grad_norm": 18.496665954589844,
      "kl": 2.9921875,
      "learning_rate": 6.146666666666667e-07,
      "loss": 0.262,
      "reward": 0.640127569437027,
      "reward_std": 0.1195078082382679,
      "reward_std_all": 0.13663875311613083,
      "rewards-max/forensics_sinkhorn_reward": 0.9878483712673187,
      "rewards-max/forensics_strict_edge_reward": 0.946017861366272,
      "rewards-max/forensics_tiou_reward": 0.9955047070980072,
      "rewards/forensics_sinkhorn_reward": 0.5292008519172668,
      "rewards/forensics_strict_edge_reward": 0.29549068212509155,
      "rewards/forensics_tiou_reward": 0.34188327193260193,
      "step": 579
    },
    {
      "completion_length": 41.5625,
      "epoch": 1.5432756324900132,
      "grad_norm": 15.835911750793457,
      "kl": 2.8671875,
      "learning_rate": 6.14e-07,
      "loss": 0.2089,
      "reward": 0.6304699182510376,
      "reward_std": 0.1569325216114521,
      "reward_std_all": 0.16402924060821533,
      "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": 0.5279892235994339,
      "rewards/forensics_strict_edge_reward": 0.31925317645072937,
      "rewards/forensics_tiou_reward": 0.3992619216442108,
      "step": 580
    },
    {
      "completion_length": 47.375,
      "epoch": 1.5459387483355527,
      "grad_norm": 50.84869384765625,
      "kl": 3.78125,
      "learning_rate": 6.133333333333332e-07,
      "loss": 0.331,
      "reward": 0.6137099266052246,
      "reward_std": 0.1451316624879837,
      "reward_std_all": 0.16260015219449997,
      "rewards-max/forensics_sinkhorn_reward": 0.9872587025165558,
      "rewards-max/forensics_strict_edge_reward": 0.9565287232398987,
      "rewards-max/forensics_tiou_reward": 0.9931796491146088,
      "rewards/forensics_sinkhorn_reward": 0.4949438273906708,
      "rewards/forensics_strict_edge_reward": 0.21880730986595154,
      "rewards/forensics_tiou_reward": 0.29925622045993805,
      "step": 581
    },
    {
      "completion_length": 45.609375,
      "epoch": 1.548601864181092,
      "grad_norm": 16.169605255126953,
      "kl": 2.12890625,
      "learning_rate": 6.126666666666667e-07,
      "loss": 0.1367,
      "reward": 0.6850690841674805,
      "reward_std": 0.11454097367823124,
      "reward_std_all": 0.1163281761109829,
      "rewards-max/forensics_sinkhorn_reward": 0.9888060092926025,
      "rewards-max/forensics_strict_edge_reward": 0.9890114367008209,
      "rewards-max/forensics_tiou_reward": 0.9992727339267731,
      "rewards/forensics_sinkhorn_reward": 0.6064284294843674,
      "rewards/forensics_strict_edge_reward": 0.3784977197647095,
      "rewards/forensics_tiou_reward": 0.4594799727201462,
      "step": 582
    },
    {
      "completion_length": 37.546875,
      "epoch": 1.5512649800266312,
      "grad_norm": 19.372915267944336,
      "kl": 2.29296875,
      "learning_rate": 6.119999999999999e-07,
      "loss": 0.1507,
      "reward": 0.6252315640449524,
      "reward_std": 0.12591975927352905,
      "reward_std_all": 0.14384828507900238,
      "rewards-max/forensics_sinkhorn_reward": 0.9877195656299591,
      "rewards-max/forensics_strict_edge_reward": 0.9464850425720215,
      "rewards-max/forensics_tiou_reward": 0.9949924945831299,
      "rewards/forensics_sinkhorn_reward": 0.5148237943649292,
      "rewards/forensics_strict_edge_reward": 0.26975858956575394,
      "rewards/forensics_tiou_reward": 0.3295644670724869,
      "step": 583
    },
    {
      "completion_length": 53.828125,
      "epoch": 1.5539280958721704,
      "grad_norm": 15.468720436096191,
      "kl": 3.78125,
      "learning_rate": 6.113333333333333e-07,
      "loss": 0.2854,
      "reward": 0.5558182299137115,
      "reward_std": 0.14944558590650558,
      "reward_std_all": 0.19301874935626984,
      "rewards-max/forensics_sinkhorn_reward": 0.9879915118217468,
      "rewards-max/forensics_strict_edge_reward": 0.9265981018543243,
      "rewards-max/forensics_tiou_reward": 0.9960638880729675,
      "rewards/forensics_sinkhorn_reward": 0.43014027178287506,
      "rewards/forensics_strict_edge_reward": 0.13744625821709633,
      "rewards/forensics_tiou_reward": 0.20139048993587494,
      "step": 584
    },
    {
      "completion_length": 40.328125,
      "epoch": 1.5565912117177096,
      "grad_norm": 17.133150100708008,
      "kl": 2.2890625,
      "learning_rate": 6.106666666666666e-07,
      "loss": 0.2448,
      "reward": 0.6597895324230194,
      "reward_std": 0.14122072607278824,
      "reward_std_all": 0.14137136563658714,
      "rewards-max/forensics_sinkhorn_reward": 0.9827971458435059,
      "rewards-max/forensics_strict_edge_reward": 0.947419673204422,
      "rewards-max/forensics_tiou_reward": 0.9759072065353394,
      "rewards/forensics_sinkhorn_reward": 0.5813526213169098,
      "rewards/forensics_strict_edge_reward": 0.32662467658519745,
      "rewards/forensics_tiou_reward": 0.39537397027015686,
      "step": 585
    },
    {
      "completion_length": 37.890625,
      "epoch": 1.559254327563249,
      "grad_norm": 15.823859214782715,
      "kl": 2.9140625,
      "learning_rate": 6.1e-07,
      "loss": 0.3981,
      "reward": 0.6994917988777161,
      "reward_std": 0.12036854028701782,
      "reward_std_all": 0.11556968837976456,
      "rewards-max/forensics_sinkhorn_reward": 0.9879122078418732,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9957539141178131,
      "rewards/forensics_sinkhorn_reward": 0.5920388102531433,
      "rewards/forensics_strict_edge_reward": 0.39002446830272675,
      "rewards/forensics_tiou_reward": 0.4617071747779846,
      "step": 586
    },
    {
      "completion_length": 31.96875,
      "epoch": 1.5619174434087881,
      "grad_norm": 16.28374481201172,
      "kl": 2.6796875,
      "learning_rate": 6.093333333333332e-07,
      "loss": 0.3006,
      "reward": 0.6467210650444031,
      "reward_std": 0.15253207087516785,
      "reward_std_all": 0.158301442861557,
      "rewards-max/forensics_sinkhorn_reward": 0.9888987839221954,
      "rewards-max/forensics_strict_edge_reward": 0.9890114367008209,
      "rewards-max/forensics_tiou_reward": 0.9996385276317596,
      "rewards/forensics_sinkhorn_reward": 0.5304970145225525,
      "rewards/forensics_strict_edge_reward": 0.3035021871328354,
      "rewards/forensics_tiou_reward": 0.3616998940706253,
      "step": 587
    },
    {
      "completion_length": 33.46875,
      "epoch": 1.5645805592543276,
      "grad_norm": 15.580506324768066,
      "kl": 1.83203125,
      "learning_rate": 6.086666666666667e-07,
      "loss": 0.2678,
      "reward": 0.6499233841896057,
      "reward_std": 0.12474719434976578,
      "reward_std_all": 0.13662225753068924,
      "rewards-max/forensics_sinkhorn_reward": 0.9798154532909393,
      "rewards-max/forensics_strict_edge_reward": 0.9364311099052429,
      "rewards-max/forensics_tiou_reward": 0.9786007106304169,
      "rewards/forensics_sinkhorn_reward": 0.5009462535381317,
      "rewards/forensics_strict_edge_reward": 0.2554452568292618,
      "rewards/forensics_tiou_reward": 0.3441840708255768,
      "step": 588
    },
    {
      "completion_length": 33.625,
      "epoch": 1.5672436750998668,
      "grad_norm": 19.72934913635254,
      "kl": 2.4609375,
      "learning_rate": 6.079999999999999e-07,
      "loss": 0.2204,
      "reward": 0.6045872867107391,
      "reward_std": 0.15441004186868668,
      "reward_std_all": 0.17066217958927155,
      "rewards-max/forensics_sinkhorn_reward": 0.9704816341400146,
      "rewards-max/forensics_strict_edge_reward": 0.9185642302036285,
      "rewards-max/forensics_tiou_reward": 0.9606245160102844,
      "rewards/forensics_sinkhorn_reward": 0.5012324303388596,
      "rewards/forensics_strict_edge_reward": 0.28459782898426056,
      "rewards/forensics_tiou_reward": 0.33944281935691833,
      "step": 589
    },
    {
      "completion_length": 54.9375,
      "epoch": 1.569906790945406,
      "grad_norm": 21.03252410888672,
      "kl": 2.8359375,
      "learning_rate": 6.073333333333333e-07,
      "loss": 0.1569,
      "reward": 0.662571132183075,
      "reward_std": 0.15038076043128967,
      "reward_std_all": 0.15640617907047272,
      "rewards-max/forensics_sinkhorn_reward": 0.9632222056388855,
      "rewards-max/forensics_strict_edge_reward": 0.9256840348243713,
      "rewards-max/forensics_tiou_reward": 0.9425566494464874,
      "rewards/forensics_sinkhorn_reward": 0.5492235720157623,
      "rewards/forensics_strict_edge_reward": 0.24300213903188705,
      "rewards/forensics_tiou_reward": 0.323862761259079,
      "step": 590
    },
    {
      "completion_length": 30.328125,
      "epoch": 1.5725699067909455,
      "grad_norm": 22.354759216308594,
      "kl": 2.22265625,
      "learning_rate": 6.066666666666666e-07,
      "loss": 0.2082,
      "reward": 0.6302515864372253,
      "reward_std": 0.1564987301826477,
      "reward_std_all": 0.16369648277759552,
      "rewards-max/forensics_sinkhorn_reward": 0.9787095785140991,
      "rewards-max/forensics_strict_edge_reward": 0.820955753326416,
      "rewards-max/forensics_tiou_reward": 0.9602025151252747,
      "rewards/forensics_sinkhorn_reward": 0.5422372817993164,
      "rewards/forensics_strict_edge_reward": 0.28540176153182983,
      "rewards/forensics_tiou_reward": 0.3679434806108475,
      "step": 591
    },
    {
      "completion_length": 29.40625,
      "epoch": 1.5752330226364846,
      "grad_norm": 20.149215698242188,
      "kl": 2.703125,
      "learning_rate": 6.06e-07,
      "loss": 0.2077,
      "reward": 0.6556015312671661,
      "reward_std": 0.1304517649114132,
      "reward_std_all": 0.1353340707719326,
      "rewards-max/forensics_sinkhorn_reward": 0.9872126281261444,
      "rewards-max/forensics_strict_edge_reward": 0.9565287232398987,
      "rewards-max/forensics_tiou_reward": 0.9929980635643005,
      "rewards/forensics_sinkhorn_reward": 0.6037258505821228,
      "rewards/forensics_strict_edge_reward": 0.23702691495418549,
      "rewards/forensics_tiou_reward": 0.3832085281610489,
      "step": 592
    },
    {
      "completion_length": 45.375,
      "epoch": 1.577896138482024,
      "grad_norm": 2826.712646484375,
      "kl": 30.4453125,
      "learning_rate": 6.053333333333332e-07,
      "loss": 1.3715,
      "reward": 0.648202657699585,
      "reward_std": 0.1619516983628273,
      "reward_std_all": 0.16259896755218506,
      "rewards-max/forensics_sinkhorn_reward": 0.9767740070819855,
      "rewards-max/forensics_strict_edge_reward": 0.9169811904430389,
      "rewards-max/forensics_tiou_reward": 0.9709292650222778,
      "rewards/forensics_sinkhorn_reward": 0.5237701833248138,
      "rewards/forensics_strict_edge_reward": 0.2631691098213196,
      "rewards/forensics_tiou_reward": 0.325539231300354,
      "step": 593
    },
    {
      "completion_length": 43.96875,
      "epoch": 1.5805592543275633,
      "grad_norm": 20.479612350463867,
      "kl": 2.8046875,
      "learning_rate": 6.046666666666667e-07,
      "loss": 0.1432,
      "reward": 0.6176835298538208,
      "reward_std": 0.19967065006494522,
      "reward_std_all": 0.19664306938648224,
      "rewards-max/forensics_sinkhorn_reward": 0.9874386787414551,
      "rewards-max/forensics_strict_edge_reward": 0.9574736058712006,
      "rewards-max/forensics_tiou_reward": 0.9939025044441223,
      "rewards/forensics_sinkhorn_reward": 0.4930894672870636,
      "rewards/forensics_strict_edge_reward": 0.25167374312877655,
      "rewards/forensics_tiou_reward": 0.3295962065458298,
      "step": 594
    },
    {
      "completion_length": 22.125,
      "epoch": 1.5832223701731025,
      "grad_norm": 20.12954330444336,
      "kl": 1.671875,
      "learning_rate": 6.04e-07,
      "loss": 0.2389,
      "reward": 0.6750295758247375,
      "reward_std": 0.09299232810735703,
      "reward_std_all": 0.11168170720338821,
      "rewards-max/forensics_sinkhorn_reward": 0.9885500967502594,
      "rewards-max/forensics_strict_edge_reward": 0.9574736058712006,
      "rewards-max/forensics_tiou_reward": 0.9982653856277466,
      "rewards/forensics_sinkhorn_reward": 0.5742895007133484,
      "rewards/forensics_strict_edge_reward": 0.25801989436149597,
      "rewards/forensics_tiou_reward": 0.37438878417015076,
      "step": 595
    },
    {
      "completion_length": 45.046875,
      "epoch": 1.5858854860186418,
      "grad_norm": 23.474584579467773,
      "kl": 3.1484375,
      "learning_rate": 6.033333333333333e-07,
      "loss": 0.3266,
      "reward": 0.5600204914808273,
      "reward_std": 0.2015298679471016,
      "reward_std_all": 0.2113475203514099,
      "rewards-max/forensics_sinkhorn_reward": 0.8857329785823822,
      "rewards-max/forensics_strict_edge_reward": 0.6922992169857025,
      "rewards-max/forensics_tiou_reward": 0.789675235748291,
      "rewards/forensics_sinkhorn_reward": 0.4319286346435547,
      "rewards/forensics_strict_edge_reward": 0.20147845707833767,
      "rewards/forensics_tiou_reward": 0.24086213111877441,
      "step": 596
    },
    {
      "completion_length": 33.046875,
      "epoch": 1.588548601864181,
      "grad_norm": 22.945369720458984,
      "kl": 2.16015625,
      "learning_rate": 6.026666666666667e-07,
      "loss": 0.2333,
      "reward": 0.6504736542701721,
      "reward_std": 0.17574197053909302,
      "reward_std_all": 0.16889850795269012,
      "rewards-max/forensics_sinkhorn_reward": 0.9688535332679749,
      "rewards-max/forensics_strict_edge_reward": 0.9364311099052429,
      "rewards-max/forensics_tiou_reward": 0.9540936052799225,
      "rewards/forensics_sinkhorn_reward": 0.5790893137454987,
      "rewards/forensics_strict_edge_reward": 0.2526355907320976,
      "rewards/forensics_tiou_reward": 0.36902500689029694,
      "step": 597
    },
    {
      "completion_length": 36.953125,
      "epoch": 1.5912117177097205,
      "grad_norm": 15.674314498901367,
      "kl": 2.84375,
      "learning_rate": 6.019999999999999e-07,
      "loss": 0.2894,
      "reward": 0.6621705889701843,
      "reward_std": 0.15663420408964157,
      "reward_std_all": 0.14998702704906464,
      "rewards-max/forensics_sinkhorn_reward": 0.9833482205867767,
      "rewards-max/forensics_strict_edge_reward": 0.9677534997463226,
      "rewards-max/forensics_tiou_reward": 0.9872158765792847,
      "rewards/forensics_sinkhorn_reward": 0.5447741746902466,
      "rewards/forensics_strict_edge_reward": 0.2706420123577118,
      "rewards/forensics_tiou_reward": 0.3599953353404999,
      "step": 598
    },
    {
      "completion_length": 35.21875,
      "epoch": 1.5938748335552595,
      "grad_norm": 14.635952949523926,
      "kl": 4.3125,
      "learning_rate": 6.013333333333334e-07,
      "loss": 0.3068,
      "reward": 0.6506007015705109,
      "reward_std": 0.1401040479540825,
      "reward_std_all": 0.1423376016318798,
      "rewards-max/forensics_sinkhorn_reward": 0.9882201254367828,
      "rewards-max/forensics_strict_edge_reward": 0.9565287232398987,
      "rewards-max/forensics_tiou_reward": 0.996965765953064,
      "rewards/forensics_sinkhorn_reward": 0.5014900267124176,
      "rewards/forensics_strict_edge_reward": 0.2623475193977356,
      "rewards/forensics_tiou_reward": 0.3548155128955841,
      "step": 599
    },
    {
      "completion_length": 31.453125,
      "epoch": 1.596537949400799,
      "grad_norm": 22.112180709838867,
      "kl": 2.91796875,
      "learning_rate": 6.006666666666666e-07,
      "loss": 0.2531,
      "reward": 0.6422003209590912,
      "reward_std": 0.10811281949281693,
      "reward_std_all": 0.12777907773852348,
      "rewards-max/forensics_sinkhorn_reward": 0.987913578748703,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9957555532455444,
      "rewards/forensics_sinkhorn_reward": 0.5392201840877533,
      "rewards/forensics_strict_edge_reward": 0.2769535332918167,
      "rewards/forensics_tiou_reward": 0.34273040294647217,
      "step": 600
    },
    {
      "completion_length": 19.859375,
      "epoch": 1.5992010652463382,
      "grad_norm": 25.354543685913086,
      "kl": 1.92578125,
      "learning_rate": 6e-07,
      "loss": 0.2427,
      "reward": 0.6306756138801575,
      "reward_std": 0.1360093429684639,
      "reward_std_all": 0.16128142923116684,
      "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": 0.5188955366611481,
      "rewards/forensics_strict_edge_reward": 0.2541334331035614,
      "rewards/forensics_tiou_reward": 0.34208844602108,
      "step": 601
    },
    {
      "completion_length": 35.421875,
      "epoch": 1.6018641810918774,
      "grad_norm": 15.972890853881836,
      "kl": 2.6484375,
      "learning_rate": 5.993333333333333e-07,
      "loss": 0.2637,
      "reward": 0.6719177961349487,
      "reward_std": 0.10915541648864746,
      "reward_std_all": 0.12170394882559776,
      "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": 0.5969399213790894,
      "rewards/forensics_strict_edge_reward": 0.2961394041776657,
      "rewards/forensics_tiou_reward": 0.4077042490243912,
      "step": 602
    },
    {
      "completion_length": 45.953125,
      "epoch": 1.604527296937417,
      "grad_norm": 17.09100341796875,
      "kl": 3.328125,
      "learning_rate": 5.986666666666667e-07,
      "loss": 0.2845,
      "reward": 0.6571895778179169,
      "reward_std": 0.1486375778913498,
      "reward_std_all": 0.1466517522931099,
      "rewards-max/forensics_sinkhorn_reward": 0.9529833793640137,
      "rewards-max/forensics_strict_edge_reward": 0.8394034504890442,
      "rewards-max/forensics_tiou_reward": 0.9180996119976044,
      "rewards/forensics_sinkhorn_reward": 0.5043603032827377,
      "rewards/forensics_strict_edge_reward": 0.216631717979908,
      "rewards/forensics_tiou_reward": 0.30969278514385223,
      "step": 603
    },
    {
      "completion_length": 39.578125,
      "epoch": 1.607190412782956,
      "grad_norm": 15.773613929748535,
      "kl": 3.4140625,
      "learning_rate": 5.979999999999999e-07,
      "loss": 0.2468,
      "reward": 0.6390289962291718,
      "reward_std": 0.16228678077459335,
      "reward_std_all": 0.16186527907848358,
      "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": 0.546545684337616,
      "rewards/forensics_strict_edge_reward": 0.3183088153600693,
      "rewards/forensics_tiou_reward": 0.3912498950958252,
      "step": 604
    },
    {
      "completion_length": 43.75,
      "epoch": 1.6098535286284954,
      "grad_norm": 15.915019989013672,
      "kl": 3.421875,
      "learning_rate": 5.973333333333334e-07,
      "loss": 0.3151,
      "reward": 0.6634775996208191,
      "reward_std": 0.15051355585455894,
      "reward_std_all": 0.14443004876375198,
      "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": 0.532327875494957,
      "rewards/forensics_strict_edge_reward": 0.2561098635196686,
      "rewards/forensics_tiou_reward": 0.3314386308193207,
      "step": 605
    },
    {
      "completion_length": 30.125,
      "epoch": 1.6125166444740346,
      "grad_norm": 16.035106658935547,
      "kl": 2.25,
      "learning_rate": 5.966666666666666e-07,
      "loss": 0.3269,
      "reward": 0.6200393438339233,
      "reward_std": 0.16651159524917603,
      "reward_std_all": 0.17348656803369522,
      "rewards-max/forensics_sinkhorn_reward": 0.9881306886672974,
      "rewards-max/forensics_strict_edge_reward": 0.9565287232398987,
      "rewards-max/forensics_tiou_reward": 0.9966105818748474,
      "rewards/forensics_sinkhorn_reward": 0.5688267052173615,
      "rewards/forensics_strict_edge_reward": 0.2683243378996849,
      "rewards/forensics_tiou_reward": 0.38659417629241943,
      "step": 606
    },
    {
      "completion_length": 37.140625,
      "epoch": 1.6151797603195739,
      "grad_norm": 14.856562614440918,
      "kl": 2.390625,
      "learning_rate": 5.96e-07,
      "loss": 0.1101,
      "reward": 0.6124779582023621,
      "reward_std": 0.15532474219799042,
      "reward_std_all": 0.17699721455574036,
      "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": 0.51486936211586,
      "rewards/forensics_strict_edge_reward": 0.21406198292970657,
      "rewards/forensics_tiou_reward": 0.28507691621780396,
      "step": 607
    },
    {
      "completion_length": 42.375,
      "epoch": 1.6178428761651131,
      "grad_norm": 18.31036376953125,
      "kl": 2.7890625,
      "learning_rate": 5.953333333333333e-07,
      "loss": 0.2317,
      "reward": 0.5964098572731018,
      "reward_std": 0.21134275943040848,
      "reward_std_all": 0.20696932822465897,
      "rewards-max/forensics_sinkhorn_reward": 0.9870846569538116,
      "rewards-max/forensics_strict_edge_reward": 0.9158510267734528,
      "rewards-max/forensics_tiou_reward": 0.9948383569717407,
      "rewards/forensics_sinkhorn_reward": 0.5338874459266663,
      "rewards/forensics_strict_edge_reward": 0.2569338083267212,
      "rewards/forensics_tiou_reward": 0.3840198814868927,
      "step": 608
    },
    {
      "completion_length": 33.25,
      "epoch": 1.6205059920106524,
      "grad_norm": 17.487213134765625,
      "kl": 2.96875,
      "learning_rate": 5.946666666666667e-07,
      "loss": 0.2134,
      "reward": 0.6607612371444702,
      "reward_std": 0.14275925606489182,
      "reward_std_all": 0.1420978382229805,
      "rewards-max/forensics_sinkhorn_reward": 0.9876837432384491,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9948499798774719,
      "rewards/forensics_sinkhorn_reward": 0.5163339376449585,
      "rewards/forensics_strict_edge_reward": 0.2812301516532898,
      "rewards/forensics_tiou_reward": 0.3206954300403595,
      "step": 609
    },
    {
      "completion_length": 48.203125,
      "epoch": 1.6231691078561918,
      "grad_norm": 15.71886920928955,
      "kl": 2.640625,
      "learning_rate": 5.939999999999999e-07,
      "loss": 0.2435,
      "reward": 0.5872180461883545,
      "reward_std": 0.16242654621601105,
      "reward_std_all": 0.19200324267148972,
      "rewards-max/forensics_sinkhorn_reward": 0.9877856373786926,
      "rewards-max/forensics_strict_edge_reward": 0.946017861366272,
      "rewards-max/forensics_tiou_reward": 0.9952558875083923,
      "rewards/forensics_sinkhorn_reward": 0.48735475540161133,
      "rewards/forensics_strict_edge_reward": 0.21052823960781097,
      "rewards/forensics_tiou_reward": 0.26819488406181335,
      "step": 610
    },
    {
      "completion_length": 39.296875,
      "epoch": 1.6258322237017309,
      "grad_norm": 16.24832534790039,
      "kl": 2.0546875,
      "learning_rate": 5.933333333333334e-07,
      "loss": 0.2529,
      "reward": 0.6108747720718384,
      "reward_std": 0.18151363730430603,
      "reward_std_all": 0.18721606582403183,
      "rewards-max/forensics_sinkhorn_reward": 0.9745139479637146,
      "rewards-max/forensics_strict_edge_reward": 0.9357379674911499,
      "rewards-max/forensics_tiou_reward": 0.965562492609024,
      "rewards/forensics_sinkhorn_reward": 0.4918573945760727,
      "rewards/forensics_strict_edge_reward": 0.23314812779426575,
      "rewards/forensics_tiou_reward": 0.3184341788291931,
      "step": 611
    },
    {
      "completion_length": 43.6875,
      "epoch": 1.6284953395472703,
      "grad_norm": 14.343453407287598,
      "kl": 2.71875,
      "learning_rate": 5.926666666666667e-07,
      "loss": 0.2643,
      "reward": 0.6300223171710968,
      "reward_std": 0.1541246846318245,
      "reward_std_all": 0.16184159368276596,
      "rewards-max/forensics_sinkhorn_reward": 0.9880906939506531,
      "rewards-max/forensics_strict_edge_reward": 0.9075756669044495,
      "rewards-max/forensics_tiou_reward": 0.9964528679847717,
      "rewards/forensics_sinkhorn_reward": 0.5549005568027496,
      "rewards/forensics_strict_edge_reward": 0.32593192905187607,
      "rewards/forensics_tiou_reward": 0.4055894613265991,
      "step": 612
    },
    {
      "completion_length": 55.78125,
      "epoch": 1.6311584553928096,
      "grad_norm": 20.304706573486328,
      "kl": 3.1875,
      "learning_rate": 5.919999999999999e-07,
      "loss": 0.3284,
      "reward": 0.6202647984027863,
      "reward_std": 0.16830171644687653,
      "reward_std_all": 0.17325901985168457,
      "rewards-max/forensics_sinkhorn_reward": 0.9559612572193146,
      "rewards-max/forensics_strict_edge_reward": 0.8915697336196899,
      "rewards-max/forensics_tiou_reward": 0.9248145222663879,
      "rewards/forensics_sinkhorn_reward": 0.435102179646492,
      "rewards/forensics_strict_edge_reward": 0.2887043356895447,
      "rewards/forensics_tiou_reward": 0.3470088094472885,
      "step": 613
    },
    {
      "completion_length": 59.78125,
      "epoch": 1.6338215712383488,
      "grad_norm": 32.58317184448242,
      "kl": 4.5234375,
      "learning_rate": 5.913333333333334e-07,
      "loss": 0.3824,
      "reward": 0.5931976437568665,
      "reward_std": 0.1367039829492569,
      "reward_std_all": 0.17007966339588165,
      "rewards-max/forensics_sinkhorn_reward": 0.9702767729759216,
      "rewards-max/forensics_strict_edge_reward": 0.947419673204422,
      "rewards-max/forensics_tiou_reward": 0.9583612084388733,
      "rewards/forensics_sinkhorn_reward": 0.47154104709625244,
      "rewards/forensics_strict_edge_reward": 0.15183649584650993,
      "rewards/forensics_tiou_reward": 0.25682318955659866,
      "step": 614
    },
    {
      "completion_length": 33.71875,
      "epoch": 1.6364846870838883,
      "grad_norm": 20.542461395263672,
      "kl": 2.921875,
      "learning_rate": 5.906666666666666e-07,
      "loss": 0.3746,
      "reward": 0.6246823370456696,
      "reward_std": 0.1636653020977974,
      "reward_std_all": 0.17047279328107834,
      "rewards-max/forensics_sinkhorn_reward": 0.9881977140903473,
      "rewards-max/forensics_strict_edge_reward": 0.9375866651535034,
      "rewards-max/forensics_tiou_reward": 0.996874988079071,
      "rewards/forensics_sinkhorn_reward": 0.5390708446502686,
      "rewards/forensics_strict_edge_reward": 0.28960371017456055,
      "rewards/forensics_tiou_reward": 0.4020627290010452,
      "step": 615
    },
    {
      "completion_length": 42.21875,
      "epoch": 1.6391478029294273,
      "grad_norm": 17.065868377685547,
      "kl": 3.3671875,
      "learning_rate": 5.9e-07,
      "loss": 0.27,
      "reward": 0.6150676608085632,
      "reward_std": 0.18271507322788239,
      "reward_std_all": 0.18663956969976425,
      "rewards-max/forensics_sinkhorn_reward": 0.987453043460846,
      "rewards-max/forensics_strict_edge_reward": 0.9357379674911499,
      "rewards-max/forensics_tiou_reward": 0.9939400255680084,
      "rewards/forensics_sinkhorn_reward": 0.5096842795610428,
      "rewards/forensics_strict_edge_reward": 0.25001613795757294,
      "rewards/forensics_tiou_reward": 0.34012123942375183,
      "step": 616
    },
    {
      "completion_length": 41.796875,
      "epoch": 1.6418109187749668,
      "grad_norm": 17.747642517089844,
      "kl": 3.3125,
      "learning_rate": 5.893333333333333e-07,
      "loss": 0.4365,
      "reward": 0.6369523406028748,
      "reward_std": 0.1169331781566143,
      "reward_std_all": 0.1375332698225975,
      "rewards-max/forensics_sinkhorn_reward": 0.9880842566490173,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9964277446269989,
      "rewards/forensics_sinkhorn_reward": 0.5555354207754135,
      "rewards/forensics_strict_edge_reward": 0.33005285263061523,
      "rewards/forensics_tiou_reward": 0.41823621839284897,
      "step": 617
    },
    {
      "completion_length": 59.484375,
      "epoch": 1.644474034620506,
      "grad_norm": 18.57377815246582,
      "kl": 3.5625,
      "learning_rate": 5.886666666666667e-07,
      "loss": 0.2735,
      "reward": 0.6104133725166321,
      "reward_std": 0.18592070043087006,
      "reward_std_all": 0.18516644835472107,
      "rewards-max/forensics_sinkhorn_reward": 0.987281858921051,
      "rewards-max/forensics_strict_edge_reward": 0.9574736058712006,
      "rewards-max/forensics_tiou_reward": 0.993288666009903,
      "rewards/forensics_sinkhorn_reward": 0.48254290223121643,
      "rewards/forensics_strict_edge_reward": 0.2539125233888626,
      "rewards/forensics_tiou_reward": 0.31805136799812317,
      "step": 618
    },
    {
      "completion_length": 80.875,
      "epoch": 1.6471371504660453,
      "grad_norm": 18.04183006286621,
      "kl": 4.171875,
      "learning_rate": 5.879999999999999e-07,
      "loss": 0.2426,
      "reward": 0.5433299243450165,
      "reward_std": 0.1906009465456009,
      "reward_std_all": 0.21294737607240677,
      "rewards-max/forensics_sinkhorn_reward": 0.9813622236251831,
      "rewards-max/forensics_strict_edge_reward": 0.8876297473907471,
      "rewards-max/forensics_tiou_reward": 0.9812324643135071,
      "rewards/forensics_sinkhorn_reward": 0.39600592851638794,
      "rewards/forensics_strict_edge_reward": 0.14054463058710098,
      "rewards/forensics_tiou_reward": 0.19591303914785385,
      "step": 619
    },
    {
      "completion_length": 42.859375,
      "epoch": 1.6498002663115847,
      "grad_norm": 13.950801849365234,
      "kl": 3.359375,
      "learning_rate": 5.873333333333334e-07,
      "loss": 0.3482,
      "reward": 0.6627835035324097,
      "reward_std": 0.16650648415088654,
      "reward_std_all": 0.15929650515317917,
      "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": 0.6370183527469635,
      "rewards/forensics_strict_edge_reward": 0.4811255931854248,
      "rewards/forensics_tiou_reward": 0.5363069772720337,
      "step": 620
    },
    {
      "completion_length": 48.875,
      "epoch": 1.6524633821571237,
      "grad_norm": 17.939857482910156,
      "kl": 3.484375,
      "learning_rate": 5.866666666666666e-07,
      "loss": 0.3433,
      "reward": 0.6561447381973267,
      "reward_std": 0.13559356331825256,
      "reward_std_all": 0.13959958404302597,
      "rewards-max/forensics_sinkhorn_reward": 0.9868011176586151,
      "rewards-max/forensics_strict_edge_reward": 0.9565287232398987,
      "rewards-max/forensics_tiou_reward": 0.9913810789585114,
      "rewards/forensics_sinkhorn_reward": 0.5168827772140503,
      "rewards/forensics_strict_edge_reward": 0.2717808783054352,
      "rewards/forensics_tiou_reward": 0.33711887896060944,
      "step": 621
    },
    {
      "completion_length": 69.21875,
      "epoch": 1.6551264980026632,
      "grad_norm": 15.222232818603516,
      "kl": 2.796875,
      "learning_rate": 5.86e-07,
      "loss": 0.2043,
      "reward": 0.5967482030391693,
      "reward_std": 0.13615791499614716,
      "reward_std_all": 0.1719931736588478,
      "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": 0.5540515184402466,
      "rewards/forensics_strict_edge_reward": 0.2674316018819809,
      "rewards/forensics_tiou_reward": 0.3545258492231369,
      "step": 622
    },
    {
      "completion_length": 75.4375,
      "epoch": 1.6577896138482024,
      "grad_norm": 19.90459442138672,
      "kl": 4.3125,
      "learning_rate": 5.853333333333333e-07,
      "loss": 0.3649,
      "reward": 0.5529963970184326,
      "reward_std": 0.199722059071064,
      "reward_std_all": 0.21839488297700882,
      "rewards-max/forensics_sinkhorn_reward": 0.9877093732357025,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9949526190757751,
      "rewards/forensics_sinkhorn_reward": 0.4318563789129257,
      "rewards/forensics_strict_edge_reward": 0.12759708613157272,
      "rewards/forensics_tiou_reward": 0.20348675549030304,
      "step": 623
    },
    {
      "completion_length": 90.53125,
      "epoch": 1.6604527296937417,
      "grad_norm": 20.146446228027344,
      "kl": 4.78125,
      "learning_rate": 5.846666666666667e-07,
      "loss": 0.3575,
      "reward": 0.5906301736831665,
      "reward_std": 0.17938527464866638,
      "reward_std_all": 0.1907159835100174,
      "rewards-max/forensics_sinkhorn_reward": 0.9886096119880676,
      "rewards-max/forensics_strict_edge_reward": 0.9890114367008209,
      "rewards-max/forensics_tiou_reward": 0.9985063076019287,
      "rewards/forensics_sinkhorn_reward": 0.4633789211511612,
      "rewards/forensics_strict_edge_reward": 0.25424351543188095,
      "rewards/forensics_tiou_reward": 0.2957105040550232,
      "step": 624
    },
    {
      "completion_length": 79.875,
      "epoch": 1.663115845539281,
      "grad_norm": 34.928558349609375,
      "kl": 4.453125,
      "learning_rate": 5.839999999999999e-07,
      "loss": 0.2347,
      "reward": 0.619141697883606,
      "reward_std": 0.15621411800384521,
      "reward_std_all": 0.16699431836605072,
      "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": 0.5066752731800079,
      "rewards/forensics_strict_edge_reward": 0.2071370929479599,
      "rewards/forensics_tiou_reward": 0.2758113741874695,
      "step": 625
    },
    {
      "completion_length": 94.609375,
      "epoch": 1.6657789613848202,
      "grad_norm": 89.09476470947266,
      "kl": 5.9375,
      "learning_rate": 5.833333333333334e-07,
      "loss": 0.3669,
      "reward": 0.5682187974452972,
      "reward_std": 0.19947144389152527,
      "reward_std_all": 0.21672959625720978,
      "rewards-max/forensics_sinkhorn_reward": 0.9575115442276001,
      "rewards-max/forensics_strict_edge_reward": 0.8503920137882233,
      "rewards-max/forensics_tiou_reward": 0.9299918115139008,
      "rewards/forensics_sinkhorn_reward": 0.4561217874288559,
      "rewards/forensics_strict_edge_reward": 0.1514941304922104,
      "rewards/forensics_tiou_reward": 0.24852831661701202,
      "step": 626
    },
    {
      "completion_length": 78.34375,
      "epoch": 1.6684420772303596,
      "grad_norm": 21.92697525024414,
      "kl": 4.2734375,
      "learning_rate": 5.826666666666666e-07,
      "loss": 0.3443,
      "reward": 0.5600215494632721,
      "reward_std": 0.1485888436436653,
      "reward_std_all": 0.18734033405780792,
      "rewards-max/forensics_sinkhorn_reward": 0.9879736006259918,
      "rewards-max/forensics_strict_edge_reward": 0.9169811904430389,
      "rewards-max/forensics_tiou_reward": 0.9959948062896729,
      "rewards/forensics_sinkhorn_reward": 0.4573938697576523,
      "rewards/forensics_strict_edge_reward": 0.2557412460446358,
      "rewards/forensics_tiou_reward": 0.3241286873817444,
      "step": 627
    },
    {
      "completion_length": 47.859375,
      "epoch": 1.6711051930758987,
      "grad_norm": 23.936729431152344,
      "kl": 3.765625,
      "learning_rate": 5.819999999999999e-07,
      "loss": 0.2159,
      "reward": 0.5880112648010254,
      "reward_std": 0.182778961956501,
      "reward_std_all": 0.1932820826768875,
      "rewards-max/forensics_sinkhorn_reward": 0.9656203985214233,
      "rewards-max/forensics_strict_edge_reward": 0.8978325724601746,
      "rewards-max/forensics_tiou_reward": 0.947766125202179,
      "rewards/forensics_sinkhorn_reward": 0.46228599548339844,
      "rewards/forensics_strict_edge_reward": 0.16242458671331406,
      "rewards/forensics_tiou_reward": 0.2160961553454399,
      "step": 628
    },
    {
      "completion_length": 59.75,
      "epoch": 1.6737683089214381,
      "grad_norm": 24.127779006958008,
      "kl": 2.78125,
      "learning_rate": 5.813333333333334e-07,
      "loss": 0.3608,
      "reward": 0.595588356256485,
      "reward_std": 0.17252637445926666,
      "reward_std_all": 0.1915029212832451,
      "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": 0.5436141192913055,
      "rewards/forensics_strict_edge_reward": 0.3122437596321106,
      "rewards/forensics_tiou_reward": 0.38390108942985535,
      "step": 629
    },
    {
      "completion_length": 94.21875,
      "epoch": 1.6764314247669774,
      "grad_norm": 20.59085464477539,
      "kl": 4.1171875,
      "learning_rate": 5.806666666666666e-07,
      "loss": 0.323,
      "reward": 0.5917627811431885,
      "reward_std": 0.16756556183099747,
      "reward_std_all": 0.18273885548114777,
      "rewards-max/forensics_sinkhorn_reward": 0.9881721436977386,
      "rewards-max/forensics_strict_edge_reward": 0.9574736058712006,
      "rewards-max/forensics_tiou_reward": 0.9967741966247559,
      "rewards/forensics_sinkhorn_reward": 0.5165655761957169,
      "rewards/forensics_strict_edge_reward": 0.27503669261932373,
      "rewards/forensics_tiou_reward": 0.36214837431907654,
      "step": 630
    },
    {
      "completion_length": 65.0,
      "epoch": 1.6790945406125166,
      "grad_norm": 19.180984497070312,
      "kl": 4.0390625,
      "learning_rate": 5.8e-07,
      "loss": 0.3068,
      "reward": 0.6464358270168304,
      "reward_std": 0.12633325159549713,
      "reward_std_all": 0.14155803620815277,
      "rewards-max/forensics_sinkhorn_reward": 0.9868609011173248,
      "rewards-max/forensics_strict_edge_reward": 0.9565287232398987,
      "rewards-max/forensics_tiou_reward": 0.99162358045578,
      "rewards/forensics_sinkhorn_reward": 0.5486056506633759,
      "rewards/forensics_strict_edge_reward": 0.2929144501686096,
      "rewards/forensics_tiou_reward": 0.36896321177482605,
      "step": 631
    },
    {
      "completion_length": 54.765625,
      "epoch": 1.681757656458056,
      "grad_norm": 22.311420440673828,
      "kl": 4.7421875,
      "learning_rate": 5.793333333333333e-07,
      "loss": 0.2822,
      "reward": 0.6666466593742371,
      "reward_std": 0.12892209738492966,
      "reward_std_all": 0.13451702520251274,
      "rewards-max/forensics_sinkhorn_reward": 0.9876722693443298,
      "rewards-max/forensics_strict_edge_reward": 0.9565287232398987,
      "rewards-max/forensics_tiou_reward": 0.9948104023933411,
      "rewards/forensics_sinkhorn_reward": 0.566011905670166,
      "rewards/forensics_strict_edge_reward": 0.290418803691864,
      "rewards/forensics_tiou_reward": 0.33045394718647003,
      "step": 632
    },
    {
      "completion_length": 100.640625,
      "epoch": 1.684420772303595,
      "grad_norm": 23.19611930847168,
      "kl": 5.734375,
      "learning_rate": 5.786666666666667e-07,
      "loss": 0.3629,
      "reward": 0.5793574750423431,
      "reward_std": 0.16668283939361572,
      "reward_std_all": 0.18796612322330475,
      "rewards-max/forensics_sinkhorn_reward": 0.9745430052280426,
      "rewards-max/forensics_strict_edge_reward": 0.8553276658058167,
      "rewards-max/forensics_tiou_reward": 0.9662681221961975,
      "rewards/forensics_sinkhorn_reward": 0.441180020570755,
      "rewards/forensics_strict_edge_reward": 0.21568139642477036,
      "rewards/forensics_tiou_reward": 0.25568945705890656,
      "step": 633
    },
    {
      "completion_length": 116.5625,
      "epoch": 1.6870838881491346,
      "grad_norm": 26.696664810180664,
      "kl": 6.5625,
      "learning_rate": 5.779999999999999e-07,
      "loss": 0.4804,
      "reward": 0.553119421005249,
      "reward_std": 0.19310597330331802,
      "reward_std_all": 0.2162781059741974,
      "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": 0.4644234776496887,
      "rewards/forensics_strict_edge_reward": 0.28017423301935196,
      "rewards/forensics_tiou_reward": 0.330543577671051,
      "step": 634
    },
    {
      "completion_length": 47.609375,
      "epoch": 1.6897470039946738,
      "grad_norm": 20.741899490356445,
      "kl": 3.03125,
      "learning_rate": 5.773333333333334e-07,
      "loss": 0.228,
      "reward": 0.6190571188926697,
      "reward_std": 0.14299696311354637,
      "reward_std_all": 0.16582193225622177,
      "rewards-max/forensics_sinkhorn_reward": 0.9743378460407257,
      "rewards-max/forensics_strict_edge_reward": 0.946017861366272,
      "rewards-max/forensics_tiou_reward": 0.967724472284317,
      "rewards/forensics_sinkhorn_reward": 0.5031653791666031,
      "rewards/forensics_strict_edge_reward": 0.1905411332845688,
      "rewards/forensics_tiou_reward": 0.2503446042537689,
      "step": 635
    },
    {
      "completion_length": 74.84375,
      "epoch": 1.692410119840213,
      "grad_norm": 23.185150146484375,
      "kl": 3.6875,
      "learning_rate": 5.766666666666666e-07,
      "loss": 0.2039,
      "reward": 0.5702788233757019,
      "reward_std": 0.20992298424243927,
      "reward_std_all": 0.21629925817251205,
      "rewards-max/forensics_sinkhorn_reward": 0.9675994515419006,
      "rewards-max/forensics_strict_edge_reward": 0.7780638039112091,
      "rewards-max/forensics_tiou_reward": 0.9486190676689148,
      "rewards/forensics_sinkhorn_reward": 0.45423972606658936,
      "rewards/forensics_strict_edge_reward": 0.23711739480495453,
      "rewards/forensics_tiou_reward": 0.2734617292881012,
      "step": 636
    },
    {
      "completion_length": 81.890625,
      "epoch": 1.6950732356857523,
      "grad_norm": 22.798978805541992,
      "kl": 3.609375,
      "learning_rate": 5.76e-07,
      "loss": 0.3471,
      "reward": 0.6371674835681915,
      "reward_std": 0.13404865562915802,
      "reward_std_all": 0.1540798768401146,
      "rewards-max/forensics_sinkhorn_reward": 0.9883545637130737,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9974929988384247,
      "rewards/forensics_sinkhorn_reward": 0.5693584531545639,
      "rewards/forensics_strict_edge_reward": 0.3546871021389961,
      "rewards/forensics_tiou_reward": 0.46561624109745026,
      "step": 637
    },
    {
      "completion_length": 69.28125,
      "epoch": 1.6977363515312915,
      "grad_norm": 16.994441986083984,
      "kl": 2.8671875,
      "learning_rate": 5.753333333333333e-07,
      "loss": 0.2533,
      "reward": 0.6589725911617279,
      "reward_std": 0.1733374446630478,
      "reward_std_all": 0.16470935195684433,
      "rewards-max/forensics_sinkhorn_reward": 0.9831868410110474,
      "rewards-max/forensics_strict_edge_reward": 0.9185642302036285,
      "rewards-max/forensics_tiou_reward": 0.9865808188915253,
      "rewards/forensics_sinkhorn_reward": 0.6389689445495605,
      "rewards/forensics_strict_edge_reward": 0.3062685877084732,
      "rewards/forensics_tiou_reward": 0.4603550136089325,
      "step": 638
    },
    {
      "completion_length": 100.171875,
      "epoch": 1.700399467376831,
      "grad_norm": 25.789337158203125,
      "kl": 4.296875,
      "learning_rate": 5.746666666666667e-07,
      "loss": 0.4061,
      "reward": 0.571903795003891,
      "reward_std": 0.23800282925367355,
      "reward_std_all": 0.23617853969335556,
      "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": 0.4876565635204315,
      "rewards/forensics_strict_edge_reward": 0.31766851246356964,
      "rewards/forensics_tiou_reward": 0.36962921917438507,
      "step": 639
    },
    {
      "completion_length": 75.234375,
      "epoch": 1.70306258322237,
      "grad_norm": 29.77236557006836,
      "kl": 4.5625,
      "learning_rate": 5.739999999999999e-07,
      "loss": 0.3443,
      "reward": 0.5924988090991974,
      "reward_std": 0.19915048778057098,
      "reward_std_all": 0.19846491515636444,
      "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": 0.5289664566516876,
      "rewards/forensics_strict_edge_reward": 0.3082936108112335,
      "rewards/forensics_tiou_reward": 0.37824995815753937,
      "step": 640
    },
    {
      "completion_length": 99.015625,
      "epoch": 1.7057256990679095,
      "grad_norm": 19.040348052978516,
      "kl": 6.53125,
      "learning_rate": 5.733333333333334e-07,
      "loss": 0.4612,
      "reward": 0.5556139349937439,
      "reward_std": 0.17942263185977936,
      "reward_std_all": 0.2071082517504692,
      "rewards-max/forensics_sinkhorn_reward": 0.9882858097553253,
      "rewards-max/forensics_strict_edge_reward": 0.9003686904907227,
      "rewards-max/forensics_tiou_reward": 0.9972260892391205,
      "rewards/forensics_sinkhorn_reward": 0.4443208873271942,
      "rewards/forensics_strict_edge_reward": 0.21289057284593582,
      "rewards/forensics_tiou_reward": 0.26126574724912643,
      "step": 641
    },
    {
      "completion_length": 72.859375,
      "epoch": 1.7083888149134487,
      "grad_norm": 20.799198150634766,
      "kl": 3.8671875,
      "learning_rate": 5.726666666666666e-07,
      "loss": 0.3727,
      "reward": 0.6145166456699371,
      "reward_std": 0.1611272096633911,
      "reward_std_all": 0.16743112355470657,
      "rewards-max/forensics_sinkhorn_reward": 0.9658514857292175,
      "rewards-max/forensics_strict_edge_reward": 0.8668389916419983,
      "rewards-max/forensics_tiou_reward": 0.944225549697876,
      "rewards/forensics_sinkhorn_reward": 0.48208218812942505,
      "rewards/forensics_strict_edge_reward": 0.21469366550445557,
      "rewards/forensics_tiou_reward": 0.2595488280057907,
      "step": 642
    },
    {
      "completion_length": 97.421875,
      "epoch": 1.711051930758988,
      "grad_norm": 28.85892677307129,
      "kl": 7.078125,
      "learning_rate": 5.719999999999999e-07,
      "loss": 0.5238,
      "reward": 0.5905431807041168,
      "reward_std": 0.17163455486297607,
      "reward_std_all": 0.1872420683503151,
      "rewards-max/forensics_sinkhorn_reward": 0.982073575258255,
      "rewards-max/forensics_strict_edge_reward": 0.9567649364471436,
      "rewards-max/forensics_tiou_reward": 0.9835951030254364,
      "rewards/forensics_sinkhorn_reward": 0.49927787482738495,
      "rewards/forensics_strict_edge_reward": 0.2002086490392685,
      "rewards/forensics_tiou_reward": 0.25172458589076996,
      "step": 643
    },
    {
      "completion_length": 98.625,
      "epoch": 1.7137150466045274,
      "grad_norm": 21.297807693481445,
      "kl": 5.5234375,
      "learning_rate": 5.713333333333333e-07,
      "loss": 0.3191,
      "reward": 0.5762839913368225,
      "reward_std": 0.22313787043094635,
      "reward_std_all": 0.22010548412799835,
      "rewards-max/forensics_sinkhorn_reward": 0.9473733305931091,
      "rewards-max/forensics_strict_edge_reward": 0.8490433394908905,
      "rewards-max/forensics_tiou_reward": 0.9043567776679993,
      "rewards/forensics_sinkhorn_reward": 0.4407351016998291,
      "rewards/forensics_strict_edge_reward": 0.24114742875099182,
      "rewards/forensics_tiou_reward": 0.2574857324361801,
      "step": 644
    },
    {
      "completion_length": 85.15625,
      "epoch": 1.7163781624500665,
      "grad_norm": 17.6658935546875,
      "kl": 4.390625,
      "learning_rate": 5.706666666666666e-07,
      "loss": 0.3135,
      "reward": 0.5889107584953308,
      "reward_std": 0.1698397845029831,
      "reward_std_all": 0.1928594782948494,
      "rewards-max/forensics_sinkhorn_reward": 0.9880746304988861,
      "rewards-max/forensics_strict_edge_reward": 0.947419673204422,
      "rewards-max/forensics_tiou_reward": 0.9963898956775665,
      "rewards/forensics_sinkhorn_reward": 0.46657121181488037,
      "rewards/forensics_strict_edge_reward": 0.24541841447353363,
      "rewards/forensics_tiou_reward": 0.3033522218465805,
      "step": 645
    },
    {
      "completion_length": 92.640625,
      "epoch": 1.719041278295606,
      "grad_norm": 24.87786293029785,
      "kl": 5.140625,
      "learning_rate": 5.699999999999999e-07,
      "loss": 0.3326,
      "reward": 0.5731369853019714,
      "reward_std": 0.21303097903728485,
      "reward_std_all": 0.21698008477687836,
      "rewards-max/forensics_sinkhorn_reward": 0.9867880046367645,
      "rewards-max/forensics_strict_edge_reward": 0.946017861366272,
      "rewards-max/forensics_tiou_reward": 0.9913443028926849,
      "rewards/forensics_sinkhorn_reward": 0.4719807952642441,
      "rewards/forensics_strict_edge_reward": 0.24841149896383286,
      "rewards/forensics_tiou_reward": 0.3441021740436554,
      "step": 646
    },
    {
      "completion_length": 84.21875,
      "epoch": 1.7217043941411452,
      "grad_norm": 22.69014549255371,
      "kl": 4.390625,
      "learning_rate": 5.693333333333333e-07,
      "loss": 0.212,
      "reward": 0.5787681937217712,
      "reward_std": 0.19669310748577118,
      "reward_std_all": 0.20989015698432922,
      "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": 0.45973537862300873,
      "rewards/forensics_strict_edge_reward": 0.21324704587459564,
      "rewards/forensics_tiou_reward": 0.2567152827978134,
      "step": 647
    },
    {
      "completion_length": 91.875,
      "epoch": 1.7243675099866844,
      "grad_norm": 29.970252990722656,
      "kl": 5.59375,
      "learning_rate": 5.686666666666667e-07,
      "loss": 0.4278,
      "reward": 0.5790627002716064,
      "reward_std": 0.194742850959301,
      "reward_std_all": 0.20210304856300354,
      "rewards-max/forensics_sinkhorn_reward": 0.9875268340110779,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9942398071289062,
      "rewards/forensics_sinkhorn_reward": 0.4311724901199341,
      "rewards/forensics_strict_edge_reward": 0.22965508699417114,
      "rewards/forensics_tiou_reward": 0.27056537568569183,
      "step": 648
    },
    {
      "completion_length": 93.0,
      "epoch": 1.7270306258322237,
      "grad_norm": 17.068130493164062,
      "kl": 3.9453125,
      "learning_rate": 5.679999999999999e-07,
      "loss": 0.3147,
      "reward": 0.5840797424316406,
      "reward_std": 0.13352004066109657,
      "reward_std_all": 0.16904614120721817,
      "rewards-max/forensics_sinkhorn_reward": 0.9247474670410156,
      "rewards-max/forensics_strict_edge_reward": 0.8066537380218506,
      "rewards-max/forensics_tiou_reward": 0.860848993062973,
      "rewards/forensics_sinkhorn_reward": 0.464701309800148,
      "rewards/forensics_strict_edge_reward": 0.18886896595358849,
      "rewards/forensics_tiou_reward": 0.2645275592803955,
      "step": 649
    },
    {
      "completion_length": 94.15625,
      "epoch": 1.729693741677763,
      "grad_norm": 23.950489044189453,
      "kl": 4.3984375,
      "learning_rate": 5.673333333333334e-07,
      "loss": 0.2961,
      "reward": 0.6191582381725311,
      "reward_std": 0.1851096674799919,
      "reward_std_all": 0.1845288947224617,
      "rewards-max/forensics_sinkhorn_reward": 0.988221287727356,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9969703257083893,
      "rewards/forensics_sinkhorn_reward": 0.5459154844284058,
      "rewards/forensics_strict_edge_reward": 0.3050166815519333,
      "rewards/forensics_tiou_reward": 0.38422778248786926,
      "step": 650
    },
    {
      "completion_length": 63.890625,
      "epoch": 1.7323568575233024,
      "grad_norm": 21.650922775268555,
      "kl": 2.625,
      "learning_rate": 5.666666666666666e-07,
      "loss": 0.4316,
      "reward": 0.6171180307865143,
      "reward_std": 0.1350611373782158,
      "reward_std_all": 0.1579996682703495,
      "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": 0.5233129858970642,
      "rewards/forensics_strict_edge_reward": 0.29351168870925903,
      "rewards/forensics_tiou_reward": 0.33242543041706085,
      "step": 651
    },
    {
      "completion_length": 60.140625,
      "epoch": 1.7350199733688414,
      "grad_norm": 22.840667724609375,
      "kl": 2.91796875,
      "learning_rate": 5.66e-07,
      "loss": 0.215,
      "reward": 0.6646407842636108,
      "reward_std": 0.16435683518648148,
      "reward_std_all": 0.15414556860923767,
      "rewards-max/forensics_sinkhorn_reward": 0.9884631037712097,
      "rewards-max/forensics_strict_edge_reward": 0.9890114367008209,
      "rewards-max/forensics_tiou_reward": 0.9979231655597687,
      "rewards/forensics_sinkhorn_reward": 0.6212264597415924,
      "rewards/forensics_strict_edge_reward": 0.31511375308036804,
      "rewards/forensics_tiou_reward": 0.4352639764547348,
      "step": 652
    },
    {
      "completion_length": 48.21875,
      "epoch": 1.7376830892143809,
      "grad_norm": 24.117950439453125,
      "kl": 2.359375,
      "learning_rate": 5.653333333333333e-07,
      "loss": 0.2239,
      "reward": 0.6358687579631805,
      "reward_std": 0.14284931868314743,
      "reward_std_all": 0.1526876464486122,
      "rewards-max/forensics_sinkhorn_reward": 0.970725417137146,
      "rewards-max/forensics_strict_edge_reward": 0.8394034504890442,
      "rewards-max/forensics_tiou_reward": 0.9569722414016724,
      "rewards/forensics_sinkhorn_reward": 0.5224618315696716,
      "rewards/forensics_strict_edge_reward": 0.24461250752210617,
      "rewards/forensics_tiou_reward": 0.3353024125099182,
      "step": 653
    },
    {
      "completion_length": 52.34375,
      "epoch": 1.74034620505992,
      "grad_norm": 24.226518630981445,
      "kl": 2.6953125,
      "learning_rate": 5.646666666666667e-07,
      "loss": 0.2578,
      "reward": 0.611117959022522,
      "reward_std": 0.136025108397007,
      "reward_std_all": 0.15616129338741302,
      "rewards-max/forensics_sinkhorn_reward": 0.9785954654216766,
      "rewards-max/forensics_strict_edge_reward": 0.895060271024704,
      "rewards-max/forensics_tiou_reward": 0.9755697250366211,
      "rewards/forensics_sinkhorn_reward": 0.5035351067781448,
      "rewards/forensics_strict_edge_reward": 0.1705685630440712,
      "rewards/forensics_tiou_reward": 0.271036259829998,
      "step": 654
    },
    {
      "completion_length": 112.21875,
      "epoch": 1.7430093209054593,
      "grad_norm": 27.96807289123535,
      "kl": 4.359375,
      "learning_rate": 5.639999999999999e-07,
      "loss": 0.3325,
      "reward": 0.5642574429512024,
      "reward_std": 0.20633596181869507,
      "reward_std_all": 0.22150124609470367,
      "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": 0.462447851896286,
      "rewards/forensics_strict_edge_reward": 0.2511947751045227,
      "rewards/forensics_tiou_reward": 0.31663474440574646,
      "step": 655
    },
    {
      "completion_length": 83.0,
      "epoch": 1.7456724367509988,
      "grad_norm": 30.416906356811523,
      "kl": 3.4921875,
      "learning_rate": 5.633333333333334e-07,
      "loss": 0.3799,
      "reward": 0.6389767825603485,
      "reward_std": 0.16148443520069122,
      "reward_std_all": 0.16734151542186737,
      "rewards-max/forensics_sinkhorn_reward": 0.9869766533374786,
      "rewards-max/forensics_strict_edge_reward": 0.9574736058712006,
      "rewards-max/forensics_tiou_reward": 0.9920949935913086,
      "rewards/forensics_sinkhorn_reward": 0.5545684099197388,
      "rewards/forensics_strict_edge_reward": 0.35987621545791626,
      "rewards/forensics_tiou_reward": 0.4254497289657593,
      "step": 656
    },
    {
      "completion_length": 74.4375,
      "epoch": 1.7483355525965378,
      "grad_norm": 20.735559463500977,
      "kl": 3.90625,
      "learning_rate": 5.626666666666666e-07,
      "loss": 0.2954,
      "reward": 0.6442573070526123,
      "reward_std": 0.15755313634872437,
      "reward_std_all": 0.161922175437212,
      "rewards-max/forensics_sinkhorn_reward": 0.9875257909297943,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9942264258861542,
      "rewards/forensics_sinkhorn_reward": 0.5654718279838562,
      "rewards/forensics_strict_edge_reward": 0.25504304468631744,
      "rewards/forensics_tiou_reward": 0.36551444232463837,
      "step": 657
    },
    {
      "completion_length": 90.265625,
      "epoch": 1.7509986684420773,
      "grad_norm": 25.326866149902344,
      "kl": 4.109375,
      "learning_rate": 5.620000000000001e-07,
      "loss": 0.3813,
      "reward": 0.661529004573822,
      "reward_std": 0.1646396964788437,
      "reward_std_all": 0.15999487787485123,
      "rewards-max/forensics_sinkhorn_reward": 0.987803190946579,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9953226745128632,
      "rewards/forensics_sinkhorn_reward": 0.5750120282173157,
      "rewards/forensics_strict_edge_reward": 0.29661382734775543,
      "rewards/forensics_tiou_reward": 0.3988146334886551,
      "step": 658
    },
    {
      "completion_length": 85.984375,
      "epoch": 1.7536617842876165,
      "grad_norm": 37.297760009765625,
      "kl": 2.75,
      "learning_rate": 5.613333333333333e-07,
      "loss": 0.3946,
      "reward": 0.6094380915164948,
      "reward_std": 0.19143014401197433,
      "reward_std_all": 0.19369052350521088,
      "rewards-max/forensics_sinkhorn_reward": 0.9873386025428772,
      "rewards-max/forensics_strict_edge_reward": 0.9567649364471436,
      "rewards-max/forensics_tiou_reward": 0.9935054779052734,
      "rewards/forensics_sinkhorn_reward": 0.532794177532196,
      "rewards/forensics_strict_edge_reward": 0.25948671251535416,
      "rewards/forensics_tiou_reward": 0.3239016979932785,
      "step": 659
    },
    {
      "completion_length": 61.4375,
      "epoch": 1.7563249001331558,
      "grad_norm": 21.485599517822266,
      "kl": 2.26171875,
      "learning_rate": 5.606666666666666e-07,
      "loss": 0.2504,
      "reward": 0.6695951521396637,
      "reward_std": 0.13318747282028198,
      "reward_std_all": 0.13609173148870468,
      "rewards-max/forensics_sinkhorn_reward": 0.987332820892334,
      "rewards-max/forensics_strict_edge_reward": 0.9565287232398987,
      "rewards-max/forensics_tiou_reward": 0.9934732019901276,
      "rewards/forensics_sinkhorn_reward": 0.6140499711036682,
      "rewards/forensics_strict_edge_reward": 0.3648396134376526,
      "rewards/forensics_tiou_reward": 0.43219631910324097,
      "step": 660
    },
    {
      "completion_length": 80.171875,
      "epoch": 1.758988015978695,
      "grad_norm": 21.92559051513672,
      "kl": 3.90625,
      "learning_rate": 5.6e-07,
      "loss": 0.2765,
      "reward": 0.6905462145805359,
      "reward_std": 0.08963847160339355,
      "reward_std_all": 0.10630650818347931,
      "rewards-max/forensics_sinkhorn_reward": 0.9887852668762207,
      "rewards-max/forensics_strict_edge_reward": 0.9890114367008209,
      "rewards-max/forensics_tiou_reward": 0.9991912543773651,
      "rewards/forensics_sinkhorn_reward": 0.6267688870429993,
      "rewards/forensics_strict_edge_reward": 0.34666796028614044,
      "rewards/forensics_tiou_reward": 0.4541521519422531,
      "step": 661
    },
    {
      "completion_length": 86.609375,
      "epoch": 1.7616511318242343,
      "grad_norm": 29.67249298095703,
      "kl": 3.359375,
      "learning_rate": 5.593333333333333e-07,
      "loss": 0.2572,
      "reward": 0.6250225901603699,
      "reward_std": 0.1513775885105133,
      "reward_std_all": 0.16321935504674911,
      "rewards-max/forensics_sinkhorn_reward": 0.9880077242851257,
      "rewards-max/forensics_strict_edge_reward": 0.9567649364471436,
      "rewards-max/forensics_tiou_reward": 0.9961324334144592,
      "rewards/forensics_sinkhorn_reward": 0.5835935175418854,
      "rewards/forensics_strict_edge_reward": 0.3318110257387161,
      "rewards/forensics_tiou_reward": 0.40479832887649536,
      "step": 662
    },
    {
      "completion_length": 106.84375,
      "epoch": 1.7643142476697737,
      "grad_norm": 30.495182037353516,
      "kl": 5.265625,
      "learning_rate": 5.586666666666666e-07,
      "loss": 0.3011,
      "reward": 0.6058760285377502,
      "reward_std": 0.1664394959807396,
      "reward_std_all": 0.17708554863929749,
      "rewards-max/forensics_sinkhorn_reward": 0.8992635309696198,
      "rewards-max/forensics_strict_edge_reward": 0.546617180109024,
      "rewards-max/forensics_tiou_reward": 0.805398166179657,
      "rewards/forensics_sinkhorn_reward": 0.42562350630760193,
      "rewards/forensics_strict_edge_reward": 0.1276865154504776,
      "rewards/forensics_tiou_reward": 0.20004689693450928,
      "step": 663
    },
    {
      "completion_length": 114.234375,
      "epoch": 1.7669773635153128,
      "grad_norm": 30.438732147216797,
      "kl": 4.484375,
      "learning_rate": 5.58e-07,
      "loss": 0.3918,
      "reward": 0.571035236120224,
      "reward_std": 0.18652670085430145,
      "reward_std_all": 0.2050352394580841,
      "rewards-max/forensics_sinkhorn_reward": 0.9386257529258728,
      "rewards-max/forensics_strict_edge_reward": 0.6770374178886414,
      "rewards-max/forensics_tiou_reward": 0.8848837316036224,
      "rewards/forensics_sinkhorn_reward": 0.4249339699745178,
      "rewards/forensics_strict_edge_reward": 0.16019324213266373,
      "rewards/forensics_tiou_reward": 0.22806163132190704,
      "step": 664
    },
    {
      "completion_length": 127.921875,
      "epoch": 1.7696404793608522,
      "grad_norm": 59.271976470947266,
      "kl": 9.859375,
      "learning_rate": 5.573333333333333e-07,
      "loss": 0.6344,
      "reward": 0.5721608102321625,
      "reward_std": 0.16914702951908112,
      "reward_std_all": 0.19619305431842804,
      "rewards-max/forensics_sinkhorn_reward": 0.9463672637939453,
      "rewards-max/forensics_strict_edge_reward": 0.8351600170135498,
      "rewards-max/forensics_tiou_reward": 0.9085021615028381,
      "rewards/forensics_sinkhorn_reward": 0.4387080520391464,
      "rewards/forensics_strict_edge_reward": 0.18860554695129395,
      "rewards/forensics_tiou_reward": 0.2381054162979126,
      "step": 665
    },
    {
      "completion_length": 83.421875,
      "epoch": 1.7723035952063915,
      "grad_norm": 32.05586242675781,
      "kl": 4.90625,
      "learning_rate": 5.566666666666666e-07,
      "loss": 0.3416,
      "reward": 0.6258903443813324,
      "reward_std": 0.16743440926074982,
      "reward_std_all": 0.17051882296800613,
      "rewards-max/forensics_sinkhorn_reward": 0.979315996170044,
      "rewards-max/forensics_strict_edge_reward": 0.9185642302036285,
      "rewards-max/forensics_tiou_reward": 0.9626487493515015,
      "rewards/forensics_sinkhorn_reward": 0.5367337763309479,
      "rewards/forensics_strict_edge_reward": 0.2387727051973343,
      "rewards/forensics_tiou_reward": 0.36498701572418213,
      "step": 666
    },
    {
      "completion_length": 52.90625,
      "epoch": 1.7749667110519307,
      "grad_norm": 31.644670486450195,
      "kl": 4.4453125,
      "learning_rate": 5.560000000000001e-07,
      "loss": 0.2072,
      "reward": 0.6338592171669006,
      "reward_std": 0.10449495911598206,
      "reward_std_all": 0.13896557688713074,
      "rewards-max/forensics_sinkhorn_reward": 0.9858294725418091,
      "rewards-max/forensics_strict_edge_reward": 0.8320213258266449,
      "rewards-max/forensics_tiou_reward": 0.9875732660293579,
      "rewards/forensics_sinkhorn_reward": 0.43489834666252136,
      "rewards/forensics_strict_edge_reward": 0.23907669633626938,
      "rewards/forensics_tiou_reward": 0.3044498860836029,
      "step": 667
    },
    {
      "completion_length": 65.4375,
      "epoch": 1.7776298268974702,
      "grad_norm": 31.579824447631836,
      "kl": 4.1015625,
      "learning_rate": 5.553333333333333e-07,
      "loss": 0.4181,
      "reward": 0.6134923696517944,
      "reward_std": 0.16216973960399628,
      "reward_std_all": 0.1793106570839882,
      "rewards-max/forensics_sinkhorn_reward": 0.9871301352977753,
      "rewards-max/forensics_strict_edge_reward": 0.9565287232398987,
      "rewards-max/forensics_tiou_reward": 0.9926805794239044,
      "rewards/forensics_sinkhorn_reward": 0.5163599252700806,
      "rewards/forensics_strict_edge_reward": 0.24175520241260529,
      "rewards/forensics_tiou_reward": 0.36141304671764374,
      "step": 668
    },
    {
      "completion_length": 66.265625,
      "epoch": 1.7802929427430092,
      "grad_norm": 46.591529846191406,
      "kl": 6.640625,
      "learning_rate": 5.546666666666667e-07,
      "loss": 0.2655,
      "reward": 0.6302814483642578,
      "reward_std": 0.2001662775874138,
      "reward_std_all": 0.18852491676807404,
      "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": 0.5377312302589417,
      "rewards/forensics_strict_edge_reward": 0.2596266344189644,
      "rewards/forensics_tiou_reward": 0.3670671284198761,
      "step": 669
    },
    {
      "completion_length": 34.21875,
      "epoch": 1.7829560585885487,
      "grad_norm": 25.976633071899414,
      "kl": 2.9375,
      "learning_rate": 5.54e-07,
      "loss": 0.2896,
      "reward": 0.6923783123493195,
      "reward_std": 0.082172691822052,
      "reward_std_all": 0.09715474769473076,
      "rewards-max/forensics_sinkhorn_reward": 0.9822023510932922,
      "rewards-max/forensics_strict_edge_reward": 0.905340164899826,
      "rewards-max/forensics_tiou_reward": 0.9750041365623474,
      "rewards/forensics_sinkhorn_reward": 0.6204633712768555,
      "rewards/forensics_strict_edge_reward": 0.22872771322727203,
      "rewards/forensics_tiou_reward": 0.43528440594673157,
      "step": 670
    },
    {
      "completion_length": 42.8125,
      "epoch": 1.785619174434088,
      "grad_norm": 27.915067672729492,
      "kl": 3.1953125,
      "learning_rate": 5.533333333333334e-07,
      "loss": 0.3167,
      "reward": 0.692205011844635,
      "reward_std": 0.15305602550506592,
      "reward_std_all": 0.1421329528093338,
      "rewards-max/forensics_sinkhorn_reward": 0.9838415384292603,
      "rewards-max/forensics_strict_edge_reward": 0.9567649364471436,
      "rewards-max/forensics_tiou_reward": 0.9845732748508453,
      "rewards/forensics_sinkhorn_reward": 0.6030136346817017,
      "rewards/forensics_strict_edge_reward": 0.3495066463947296,
      "rewards/forensics_tiou_reward": 0.4458155930042267,
      "step": 671
    },
    {
      "completion_length": 82.046875,
      "epoch": 1.7882822902796272,
      "grad_norm": 37.89414978027344,
      "kl": 6.390625,
      "learning_rate": 5.526666666666666e-07,
      "loss": 0.3771,
      "reward": 0.5650507807731628,
      "reward_std": 0.17335473746061325,
      "reward_std_all": 0.19789042323827744,
      "rewards-max/forensics_sinkhorn_reward": 0.9554018378257751,
      "rewards-max/forensics_strict_edge_reward": 0.8279371559619904,
      "rewards-max/forensics_tiou_reward": 0.9231570363044739,
      "rewards/forensics_sinkhorn_reward": 0.46920429170131683,
      "rewards/forensics_strict_edge_reward": 0.20561201870441437,
      "rewards/forensics_tiou_reward": 0.30691296607255936,
      "step": 672
    },
    {
      "completion_length": 96.40625,
      "epoch": 1.7909454061251664,
      "grad_norm": 27.382654190063477,
      "kl": 5.09375,
      "learning_rate": 5.520000000000001e-07,
      "loss": 0.3591,
      "reward": 0.5617471039295197,
      "reward_std": 0.21748124808073044,
      "reward_std_all": 0.2194148749113083,
      "rewards-max/forensics_sinkhorn_reward": 0.8710795640945435,
      "rewards-max/forensics_strict_edge_reward": 0.7589501440525055,
      "rewards-max/forensics_tiou_reward": 0.7868222296237946,
      "rewards/forensics_sinkhorn_reward": 0.41950060427188873,
      "rewards/forensics_strict_edge_reward": 0.1540767252445221,
      "rewards/forensics_tiou_reward": 0.2162308171391487,
      "step": 673
    },
    {
      "completion_length": 84.25,
      "epoch": 1.7936085219707056,
      "grad_norm": 26.229238510131836,
      "kl": 4.53125,
      "learning_rate": 5.513333333333333e-07,
      "loss": 0.4713,
      "reward": 0.6249451041221619,
      "reward_std": 0.140531238168478,
      "reward_std_all": 0.15726809203624725,
      "rewards-max/forensics_sinkhorn_reward": 0.9882171154022217,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9969525933265686,
      "rewards/forensics_sinkhorn_reward": 0.49386924505233765,
      "rewards/forensics_strict_edge_reward": 0.2646089196205139,
      "rewards/forensics_tiou_reward": 0.29004138708114624,
      "step": 674
    },
    {
      "completion_length": 72.8125,
      "epoch": 1.796271637816245,
      "grad_norm": 40.349754333496094,
      "kl": 3.5078125,
      "learning_rate": 5.506666666666666e-07,
      "loss": 0.2465,
      "reward": 0.6220612227916718,
      "reward_std": 0.1967504397034645,
      "reward_std_all": 0.18933428823947906,
      "rewards-max/forensics_sinkhorn_reward": 0.9795429110527039,
      "rewards-max/forensics_strict_edge_reward": 0.7997121512889862,
      "rewards-max/forensics_tiou_reward": 0.9635067582130432,
      "rewards/forensics_sinkhorn_reward": 0.5349414646625519,
      "rewards/forensics_strict_edge_reward": 0.24455858767032623,
      "rewards/forensics_tiou_reward": 0.3509160131216049,
      "step": 675
    },
    {
      "completion_length": 63.109375,
      "epoch": 1.7989347536617841,
      "grad_norm": 49.23927688598633,
      "kl": 2.609375,
      "learning_rate": 5.5e-07,
      "loss": 0.3225,
      "reward": 0.6345078945159912,
      "reward_std": 0.1972692534327507,
      "reward_std_all": 0.18937132507562637,
      "rewards-max/forensics_sinkhorn_reward": 0.9826345443725586,
      "rewards-max/forensics_strict_edge_reward": 0.8635591566562653,
      "rewards-max/forensics_tiou_reward": 0.9752392768859863,
      "rewards/forensics_sinkhorn_reward": 0.5120875686407089,
      "rewards/forensics_strict_edge_reward": 0.21028289943933487,
      "rewards/forensics_tiou_reward": 0.321973517537117,
      "step": 676
    },
    {
      "completion_length": 50.015625,
      "epoch": 1.8015978695073236,
      "grad_norm": 31.285200119018555,
      "kl": 2.109375,
      "learning_rate": 5.493333333333333e-07,
      "loss": 0.023,
      "reward": 0.6178505718708038,
      "reward_std": 0.15889423340559006,
      "reward_std_all": 0.16973039507865906,
      "rewards-max/forensics_sinkhorn_reward": 0.9499067962169647,
      "rewards-max/forensics_strict_edge_reward": 0.8745477199554443,
      "rewards-max/forensics_tiou_reward": 0.9122219383716583,
      "rewards/forensics_sinkhorn_reward": 0.5085815638303757,
      "rewards/forensics_strict_edge_reward": 0.239549919962883,
      "rewards/forensics_tiou_reward": 0.333561435341835,
      "step": 677
    },
    {
      "completion_length": 43.75,
      "epoch": 1.8042609853528628,
      "grad_norm": 24.91514778137207,
      "kl": 3.1640625,
      "learning_rate": 5.486666666666666e-07,
      "loss": 0.1341,
      "reward": 0.6724369525909424,
      "reward_std": 0.10673670098185539,
      "reward_std_all": 0.11543980613350868,
      "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": 0.6043860912322998,
      "rewards/forensics_strict_edge_reward": 0.23943285644054413,
      "rewards/forensics_tiou_reward": 0.4252956658601761,
      "step": 678
    },
    {
      "completion_length": 88.296875,
      "epoch": 1.806924101198402,
      "grad_norm": 33.642250061035156,
      "kl": 5.484375,
      "learning_rate": 5.48e-07,
      "loss": 0.304,
      "reward": 0.5561302602291107,
      "reward_std": 0.1634591668844223,
      "reward_std_all": 0.20081401616334915,
      "rewards-max/forensics_sinkhorn_reward": 0.9457014203071594,
      "rewards-max/forensics_strict_edge_reward": 0.6653123497962952,
      "rewards-max/forensics_tiou_reward": 0.9057525992393494,
      "rewards/forensics_sinkhorn_reward": 0.4160641133785248,
      "rewards/forensics_strict_edge_reward": 0.155955508351326,
      "rewards/forensics_tiou_reward": 0.24194985628128052,
      "step": 679
    },
    {
      "completion_length": 77.078125,
      "epoch": 1.8095872170439415,
      "grad_norm": 66.37832641601562,
      "kl": 5.734375,
      "learning_rate": 5.473333333333333e-07,
      "loss": 0.259,
      "reward": 0.6362997591495514,
      "reward_std": 0.20780505239963531,
      "reward_std_all": 0.19233421981334686,
      "rewards-max/forensics_sinkhorn_reward": 0.9941143691539764,
      "rewards-max/forensics_strict_edge_reward": 0.9890114367008209,
      "rewards-max/forensics_tiou_reward": 0.9985063076019287,
      "rewards/forensics_sinkhorn_reward": 0.5885727405548096,
      "rewards/forensics_strict_edge_reward": 0.33566245436668396,
      "rewards/forensics_tiou_reward": 0.42940443754196167,
      "step": 680
    },
    {
      "completion_length": 55.140625,
      "epoch": 1.8122503328894806,
      "grad_norm": 29.50246238708496,
      "kl": 2.47265625,
      "learning_rate": 5.466666666666666e-07,
      "loss": 0.3347,
      "reward": 0.67172771692276,
      "reward_std": 0.13565169647336006,
      "reward_std_all": 0.13419527933001518,
      "rewards-max/forensics_sinkhorn_reward": 0.9880301058292389,
      "rewards-max/forensics_strict_edge_reward": 0.9357379674911499,
      "rewards-max/forensics_tiou_reward": 0.9962140619754791,
      "rewards/forensics_sinkhorn_reward": 0.5957875847816467,
      "rewards/forensics_strict_edge_reward": 0.34292715042829514,
      "rewards/forensics_tiou_reward": 0.4706491529941559,
      "step": 681
    },
    {
      "completion_length": 41.75,
      "epoch": 1.81491344873502,
      "grad_norm": 24.852622985839844,
      "kl": 2.125,
      "learning_rate": 5.46e-07,
      "loss": 0.1859,
      "reward": 0.6753262281417847,
      "reward_std": 0.1072017177939415,
      "reward_std_all": 0.11192336305975914,
      "rewards-max/forensics_sinkhorn_reward": 0.988865852355957,
      "rewards-max/forensics_strict_edge_reward": 0.9890114367008209,
      "rewards-max/forensics_tiou_reward": 0.9995085895061493,
      "rewards/forensics_sinkhorn_reward": 0.6273216009140015,
      "rewards/forensics_strict_edge_reward": 0.3383500203490257,
      "rewards/forensics_tiou_reward": 0.42428356409072876,
      "step": 682
    },
    {
      "completion_length": 82.640625,
      "epoch": 1.8175765645805593,
      "grad_norm": 27.69342041015625,
      "kl": 3.640625,
      "learning_rate": 5.453333333333333e-07,
      "loss": 0.2925,
      "reward": 0.5717438757419586,
      "reward_std": 0.24454450607299805,
      "reward_std_all": 0.23620149493217468,
      "rewards-max/forensics_sinkhorn_reward": 0.9867118000984192,
      "rewards-max/forensics_strict_edge_reward": 0.9357379674911499,
      "rewards-max/forensics_tiou_reward": 0.9910414814949036,
      "rewards/forensics_sinkhorn_reward": 0.5016414970159531,
      "rewards/forensics_strict_edge_reward": 0.2861889526247978,
      "rewards/forensics_tiou_reward": 0.34555380791425705,
      "step": 683
    },
    {
      "completion_length": 45.09375,
      "epoch": 1.8202396804260985,
      "grad_norm": 26.9279727935791,
      "kl": 2.9765625,
      "learning_rate": 5.446666666666666e-07,
      "loss": 0.2084,
      "reward": 0.6188948154449463,
      "reward_std": 0.17269207537174225,
      "reward_std_all": 0.1748637780547142,
      "rewards-max/forensics_sinkhorn_reward": 0.9882983863353729,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9972728490829468,
      "rewards/forensics_sinkhorn_reward": 0.5675461888313293,
      "rewards/forensics_strict_edge_reward": 0.27891805768013,
      "rewards/forensics_tiou_reward": 0.3862048536539078,
      "step": 684
    },
    {
      "completion_length": 39.0,
      "epoch": 1.822902796271638,
      "grad_norm": 27.136659622192383,
      "kl": 2.6875,
      "learning_rate": 5.44e-07,
      "loss": 0.2237,
      "reward": 0.6549174189567566,
      "reward_std": 0.13587229698896408,
      "reward_std_all": 0.14268581569194794,
      "rewards-max/forensics_sinkhorn_reward": 0.9887631833553314,
      "rewards-max/forensics_strict_edge_reward": 0.9890114367008209,
      "rewards-max/forensics_tiou_reward": 0.9991043508052826,
      "rewards/forensics_sinkhorn_reward": 0.6014597415924072,
      "rewards/forensics_strict_edge_reward": 0.3729814887046814,
      "rewards/forensics_tiou_reward": 0.42126308381557465,
      "step": 685
    },
    {
      "completion_length": 33.859375,
      "epoch": 1.825565912117177,
      "grad_norm": 26.399160385131836,
      "kl": 3.08203125,
      "learning_rate": 5.433333333333334e-07,
      "loss": 0.1452,
      "reward": 0.676629364490509,
      "reward_std": 0.1080956682562828,
      "reward_std_all": 0.1227850690484047,
      "rewards-max/forensics_sinkhorn_reward": 0.9677196443080902,
      "rewards-max/forensics_strict_edge_reward": 0.8854433596134186,
      "rewards-max/forensics_tiou_reward": 0.949390172958374,
      "rewards/forensics_sinkhorn_reward": 0.5985507667064667,
      "rewards/forensics_strict_edge_reward": 0.2537463307380676,
      "rewards/forensics_tiou_reward": 0.37844811379909515,
      "step": 686
    },
    {
      "completion_length": 71.953125,
      "epoch": 1.8282290279627165,
      "grad_norm": 48.451881408691406,
      "kl": 1.47265625,
      "learning_rate": 5.426666666666666e-07,
      "loss": 0.2555,
      "reward": 0.6894933581352234,
      "reward_std": 0.09727468341588974,
      "reward_std_all": 0.10793488100171089,
      "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": 0.6668821275234222,
      "rewards/forensics_strict_edge_reward": 0.30989356338977814,
      "rewards/forensics_tiou_reward": 0.44444742798805237,
      "step": 687
    },
    {
      "completion_length": 32.890625,
      "epoch": 1.8308921438082557,
      "grad_norm": 22.00465202331543,
      "kl": 2.0625,
      "learning_rate": 5.420000000000001e-07,
      "loss": 0.136,
      "reward": 0.6366138458251953,
      "reward_std": 0.15236076712608337,
      "reward_std_all": 0.1561770960688591,
      "rewards-max/forensics_sinkhorn_reward": 0.9875832200050354,
      "rewards-max/forensics_strict_edge_reward": 0.9256840348243713,
      "rewards-max/forensics_tiou_reward": 0.9944629073143005,
      "rewards/forensics_sinkhorn_reward": 0.5671504735946655,
      "rewards/forensics_strict_edge_reward": 0.19531550258398056,
      "rewards/forensics_tiou_reward": 0.32981260120868683,
      "step": 688
    },
    {
      "completion_length": 79.890625,
      "epoch": 1.833555259653795,
      "grad_norm": 29.72758674621582,
      "kl": 4.0703125,
      "learning_rate": 5.413333333333333e-07,
      "loss": 0.4004,
      "reward": 0.6272476613521576,
      "reward_std": 0.15066293627023697,
      "reward_std_all": 0.15818975120782852,
      "rewards-max/forensics_sinkhorn_reward": 0.9822770953178406,
      "rewards-max/forensics_strict_edge_reward": 0.856150895357132,
      "rewards-max/forensics_tiou_reward": 0.97372767329216,
      "rewards/forensics_sinkhorn_reward": 0.5430242121219635,
      "rewards/forensics_strict_edge_reward": 0.2634180188179016,
      "rewards/forensics_tiou_reward": 0.3301694393157959,
      "step": 689
    },
    {
      "completion_length": 23.5,
      "epoch": 1.8362183754993342,
      "grad_norm": 21.274629592895508,
      "kl": 1.52734375,
      "learning_rate": 5.406666666666666e-07,
      "loss": -0.0597,
      "reward": 0.6463572084903717,
      "reward_std": 0.09541600197553635,
      "reward_std_all": 0.12271428853273392,
      "rewards-max/forensics_sinkhorn_reward": 0.9704488217830658,
      "rewards-max/forensics_strict_edge_reward": 0.6414133310317993,
      "rewards-max/forensics_tiou_reward": 0.9523414671421051,
      "rewards/forensics_sinkhorn_reward": 0.5426154732704163,
      "rewards/forensics_strict_edge_reward": 0.13285596296191216,
      "rewards/forensics_tiou_reward": 0.24984583258628845,
      "step": 690
    },
    {
      "completion_length": 54.0625,
      "epoch": 1.8388814913448734,
      "grad_norm": 30.414932250976562,
      "kl": 3.4921875,
      "learning_rate": 5.4e-07,
      "loss": 0.2864,
      "reward": 0.6662309467792511,
      "reward_std": 0.1485448256134987,
      "reward_std_all": 0.1440310701727867,
      "rewards-max/forensics_sinkhorn_reward": 0.9878903031349182,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9956636428833008,
      "rewards/forensics_sinkhorn_reward": 0.5829978883266449,
      "rewards/forensics_strict_edge_reward": 0.3001203238964081,
      "rewards/forensics_tiou_reward": 0.3887038379907608,
      "step": 691
    },
    {
      "completion_length": 46.3125,
      "epoch": 1.841544607190413,
      "grad_norm": 43.11351013183594,
      "kl": 1.50390625,
      "learning_rate": 5.393333333333333e-07,
      "loss": -0.0049,
      "reward": 0.7098471224308014,
      "reward_std": 0.12031861022114754,
      "reward_std_all": 0.11138097941875458,
      "rewards-max/forensics_sinkhorn_reward": 0.9883022010326385,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9972887635231018,
      "rewards/forensics_sinkhorn_reward": 0.6338493525981903,
      "rewards/forensics_strict_edge_reward": 0.27429626882076263,
      "rewards/forensics_tiou_reward": 0.3929574489593506,
      "step": 692
    },
    {
      "completion_length": 33.609375,
      "epoch": 1.844207723035952,
      "grad_norm": 24.260520935058594,
      "kl": 1.68359375,
      "learning_rate": 5.386666666666666e-07,
      "loss": 0.1824,
      "reward": 0.6387393474578857,
      "reward_std": 0.1524009257555008,
      "reward_std_all": 0.16029317677021027,
      "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": 0.5655739605426788,
      "rewards/forensics_strict_edge_reward": 0.2458043247461319,
      "rewards/forensics_tiou_reward": 0.354518860578537,
      "step": 693
    },
    {
      "completion_length": 42.515625,
      "epoch": 1.8468708388814914,
      "grad_norm": 20.21907615661621,
      "kl": 2.3359375,
      "learning_rate": 5.38e-07,
      "loss": 0.3552,
      "reward": 0.6655708253383636,
      "reward_std": 0.1581723764538765,
      "reward_std_all": 0.151862733066082,
      "rewards-max/forensics_sinkhorn_reward": 0.970177948474884,
      "rewards-max/forensics_strict_edge_reward": 0.8137359023094177,
      "rewards-max/forensics_tiou_reward": 0.9582628309726715,
      "rewards/forensics_sinkhorn_reward": 0.5717344284057617,
      "rewards/forensics_strict_edge_reward": 0.2827051877975464,
      "rewards/forensics_tiou_reward": 0.4095841348171234,
      "step": 694
    },
    {
      "completion_length": 47.0,
      "epoch": 1.8495339547270306,
      "grad_norm": 30.66800308227539,
      "kl": 2.40625,
      "learning_rate": 5.373333333333333e-07,
      "loss": 0.2707,
      "reward": 0.6417126059532166,
      "reward_std": 0.16393236815929413,
      "reward_std_all": 0.1651304066181183,
      "rewards-max/forensics_sinkhorn_reward": 0.9869098663330078,
      "rewards-max/forensics_strict_edge_reward": 0.946017861366272,
      "rewards-max/forensics_tiou_reward": 0.9918096661567688,
      "rewards/forensics_sinkhorn_reward": 0.5208872556686401,
      "rewards/forensics_strict_edge_reward": 0.24394939839839935,
      "rewards/forensics_tiou_reward": 0.3053267300128937,
      "step": 695
    },
    {
      "completion_length": 48.671875,
      "epoch": 1.8521970705725699,
      "grad_norm": 20.609865188598633,
      "kl": 2.9296875,
      "learning_rate": 5.366666666666666e-07,
      "loss": 0.2183,
      "reward": 0.6071025133132935,
      "reward_std": 0.13553059101104736,
      "reward_std_all": 0.16156724840402603,
      "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": 0.5076164305210114,
      "rewards/forensics_strict_edge_reward": 0.31097544729709625,
      "rewards/forensics_tiou_reward": 0.3406530171632767,
      "step": 696
    },
    {
      "completion_length": 36.859375,
      "epoch": 1.8548601864181093,
      "grad_norm": 20.32253074645996,
      "kl": 1.3125,
      "learning_rate": 5.36e-07,
      "loss": 0.0897,
      "reward": 0.6357274651527405,
      "reward_std": 0.12367799133062363,
      "reward_std_all": 0.14157376438379288,
      "rewards-max/forensics_sinkhorn_reward": 0.9880246818065643,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9961954653263092,
      "rewards/forensics_sinkhorn_reward": 0.5196181535720825,
      "rewards/forensics_strict_edge_reward": 0.2158493623137474,
      "rewards/forensics_tiou_reward": 0.30591484904289246,
      "step": 697
    },
    {
      "completion_length": 52.109375,
      "epoch": 1.8575233022636484,
      "grad_norm": 80.15803527832031,
      "kl": 3.1171875,
      "learning_rate": 5.353333333333333e-07,
      "loss": 0.2942,
      "reward": 0.6697180867195129,
      "reward_std": 0.17230496555566788,
      "reward_std_all": 0.16126079484820366,
      "rewards-max/forensics_sinkhorn_reward": 0.9763586223125458,
      "rewards-max/forensics_strict_edge_reward": 0.9158510267734528,
      "rewards-max/forensics_tiou_reward": 0.9699416160583496,
      "rewards/forensics_sinkhorn_reward": 0.6128461062908173,
      "rewards/forensics_strict_edge_reward": 0.2840309292078018,
      "rewards/forensics_tiou_reward": 0.4226353317499161,
      "step": 698
    },
    {
      "completion_length": 46.078125,
      "epoch": 1.8601864181091878,
      "grad_norm": 24.33510971069336,
      "kl": 2.1015625,
      "learning_rate": 5.346666666666666e-07,
      "loss": 0.1851,
      "reward": 0.6289026141166687,
      "reward_std": 0.11345511674880981,
      "reward_std_all": 0.14059560373425484,
      "rewards-max/forensics_sinkhorn_reward": 0.9885784685611725,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.998375803232193,
      "rewards/forensics_sinkhorn_reward": 0.5778040289878845,
      "rewards/forensics_strict_edge_reward": 0.28763003647327423,
      "rewards/forensics_tiou_reward": 0.377290815114975,
      "step": 699
    },
    {
      "completion_length": 79.546875,
      "epoch": 1.862849533954727,
      "grad_norm": 68.59878540039062,
      "kl": 5.5625,
      "learning_rate": 5.34e-07,
      "loss": 0.2379,
      "reward": 0.5935084819793701,
      "reward_std": 0.1468762643635273,
      "reward_std_all": 0.1712067946791649,
      "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": 0.5257805287837982,
      "rewards/forensics_strict_edge_reward": 0.1806519255042076,
      "rewards/forensics_tiou_reward": 0.26915256679058075,
      "step": 700
    },
    {
      "completion_length": 45.90625,
      "epoch": 1.8655126498002663,
      "grad_norm": 46.3173828125,
      "kl": 1.97265625,
      "learning_rate": 5.333333333333333e-07,
      "loss": -0.0169,
      "reward": 0.6369271278381348,
      "reward_std": 0.16670002415776253,
      "reward_std_all": 0.16627980768680573,
      "rewards-max/forensics_sinkhorn_reward": 0.9885198771953583,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.998145580291748,
      "rewards/forensics_sinkhorn_reward": 0.5267376899719238,
      "rewards/forensics_strict_edge_reward": 0.23034243285655975,
      "rewards/forensics_tiou_reward": 0.2924065887928009,
      "step": 701
    },
    {
      "completion_length": 47.421875,
      "epoch": 1.8681757656458056,
      "grad_norm": 25.045692443847656,
      "kl": 1.921875,
      "learning_rate": 5.326666666666666e-07,
      "loss": 0.1648,
      "reward": 0.6525198519229889,
      "reward_std": 0.17157769948244095,
      "reward_std_all": 0.16829928010702133,
      "rewards-max/forensics_sinkhorn_reward": 0.9883822500705719,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9976022839546204,
      "rewards/forensics_sinkhorn_reward": 0.5795863270759583,
      "rewards/forensics_strict_edge_reward": 0.3568080961704254,
      "rewards/forensics_tiou_reward": 0.3899926394224167,
      "step": 702
    },
    {
      "completion_length": 36.359375,
      "epoch": 1.8708388814913448,
      "grad_norm": 21.697057723999023,
      "kl": 1.90234375,
      "learning_rate": 5.32e-07,
      "loss": 0.0683,
      "reward": 0.6838051676750183,
      "reward_std": 0.1279592216014862,
      "reward_std_all": 0.1285974681377411,
      "rewards-max/forensics_sinkhorn_reward": 0.9848660230636597,
      "rewards-max/forensics_strict_edge_reward": 0.9062341153621674,
      "rewards-max/forensics_tiou_reward": 0.9838278293609619,
      "rewards/forensics_sinkhorn_reward": 0.5894941687583923,
      "rewards/forensics_strict_edge_reward": 0.28745022416114807,
      "rewards/forensics_tiou_reward": 0.38491225242614746,
      "step": 703
    },
    {
      "completion_length": 44.875,
      "epoch": 1.8735019973368843,
      "grad_norm": 25.55888557434082,
      "kl": 1.5859375,
      "learning_rate": 5.313333333333333e-07,
      "loss": 0.2917,
      "reward": 0.6771851778030396,
      "reward_std": 0.09837005287408829,
      "reward_std_all": 0.11799706891179085,
      "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": 0.6252344846725464,
      "rewards/forensics_strict_edge_reward": 0.356486052274704,
      "rewards/forensics_tiou_reward": 0.4469718486070633,
      "step": 704
    },
    {
      "completion_length": 62.5,
      "epoch": 1.8761651131824233,
      "grad_norm": 39.06727981567383,
      "kl": 2.0625,
      "learning_rate": 5.306666666666665e-07,
      "loss": 0.2463,
      "reward": 0.5755553543567657,
      "reward_std": 0.20334666967391968,
      "reward_std_all": 0.2136496752500534,
      "rewards-max/forensics_sinkhorn_reward": 0.9675160646438599,
      "rewards-max/forensics_strict_edge_reward": 0.8760378360748291,
      "rewards-max/forensics_tiou_reward": 0.9499525427818298,
      "rewards/forensics_sinkhorn_reward": 0.4713890999555588,
      "rewards/forensics_strict_edge_reward": 0.22956736385822296,
      "rewards/forensics_tiou_reward": 0.285856232047081,
      "step": 705
    },
    {
      "completion_length": 37.078125,
      "epoch": 1.8788282290279628,
      "grad_norm": 22.090282440185547,
      "kl": 1.46484375,
      "learning_rate": 5.3e-07,
      "loss": 0.1729,
      "reward": 0.6544960737228394,
      "reward_std": 0.12908226251602173,
      "reward_std_all": 0.14113262295722961,
      "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": 0.6241821646690369,
      "rewards/forensics_strict_edge_reward": 0.32158952951431274,
      "rewards/forensics_tiou_reward": 0.4463944733142853,
      "step": 706
    },
    {
      "completion_length": 52.15625,
      "epoch": 1.881491344873502,
      "grad_norm": 35.718387603759766,
      "kl": 2.2109375,
      "learning_rate": 5.293333333333333e-07,
      "loss": -0.0769,
      "reward": 0.6627146303653717,
      "reward_std": 0.16925883293151855,
      "reward_std_all": 0.15904377400875092,
      "rewards-max/forensics_sinkhorn_reward": 0.9885916411876678,
      "rewards-max/forensics_strict_edge_reward": 0.9677534997463226,
      "rewards-max/forensics_tiou_reward": 0.9984289109706879,
      "rewards/forensics_sinkhorn_reward": 0.5589433014392853,
      "rewards/forensics_strict_edge_reward": 0.28847044706344604,
      "rewards/forensics_tiou_reward": 0.34515537321567535,
      "step": 707
    },
    {
      "completion_length": 90.53125,
      "epoch": 1.8841544607190412,
      "grad_norm": 56.40549087524414,
      "kl": 3.1171875,
      "learning_rate": 5.286666666666666e-07,
      "loss": 0.3089,
      "reward": 0.5879568457603455,
      "reward_std": 0.21885666996240616,
      "reward_std_all": 0.2176697924733162,
      "rewards-max/forensics_sinkhorn_reward": 0.9238874912261963,
      "rewards-max/forensics_strict_edge_reward": 0.820955753326416,
      "rewards-max/forensics_tiou_reward": 0.8639338612556458,
      "rewards/forensics_sinkhorn_reward": 0.4496573507785797,
      "rewards/forensics_strict_edge_reward": 0.26099294424057007,
      "rewards/forensics_tiou_reward": 0.29212307929992676,
      "step": 708
    },
    {
      "completion_length": 68.125,
      "epoch": 1.8868175765645807,
      "grad_norm": 29.73518943786621,
      "kl": 4.4921875,
      "learning_rate": 5.28e-07,
      "loss": 0.4548,
      "reward": 0.5840663313865662,
      "reward_std": 0.16098463535308838,
      "reward_std_all": 0.18501633405685425,
      "rewards-max/forensics_sinkhorn_reward": 0.9874432384967804,
      "rewards-max/forensics_strict_edge_reward": 0.9464850425720215,
      "rewards-max/forensics_tiou_reward": 0.9939104616641998,
      "rewards/forensics_sinkhorn_reward": 0.4469298869371414,
      "rewards/forensics_strict_edge_reward": 0.22121165692806244,
      "rewards/forensics_tiou_reward": 0.31184788048267365,
      "step": 709
    },
    {
      "completion_length": 56.0,
      "epoch": 1.8894806924101197,
      "grad_norm": 70.87339782714844,
      "kl": 4.0234375,
      "learning_rate": 5.273333333333333e-07,
      "loss": 0.2691,
      "reward": 0.6545572876930237,
      "reward_std": 0.1485196240246296,
      "reward_std_all": 0.14860276877880096,
      "rewards-max/forensics_sinkhorn_reward": 0.9879550933837891,
      "rewards-max/forensics_strict_edge_reward": 0.9565287232398987,
      "rewards-max/forensics_tiou_reward": 0.9959183931350708,
      "rewards/forensics_sinkhorn_reward": 0.5727883577346802,
      "rewards/forensics_strict_edge_reward": 0.25507161021232605,
      "rewards/forensics_tiou_reward": 0.39972123503685,
      "step": 710
    },
    {
      "completion_length": 41.484375,
      "epoch": 1.8921438082556592,
      "grad_norm": 27.276756286621094,
      "kl": 2.8984375,
      "learning_rate": 5.266666666666666e-07,
      "loss": 0.3353,
      "reward": 0.6802700757980347,
      "reward_std": 0.1293109655380249,
      "reward_std_all": 0.13154607824981213,
      "rewards-max/forensics_sinkhorn_reward": 0.988226056098938,
      "rewards-max/forensics_strict_edge_reward": 0.9565287232398987,
      "rewards-max/forensics_tiou_reward": 0.9969902336597443,
      "rewards/forensics_sinkhorn_reward": 0.5723005533218384,
      "rewards/forensics_strict_edge_reward": 0.3035081773996353,
      "rewards/forensics_tiou_reward": 0.3388879597187042,
      "step": 711
    },
    {
      "completion_length": 55.5,
      "epoch": 1.8948069241011984,
      "grad_norm": 25.315048217773438,
      "kl": 2.609375,
      "learning_rate": 5.26e-07,
      "loss": 0.2282,
      "reward": 0.6595699489116669,
      "reward_std": 0.1258708592504263,
      "reward_std_all": 0.13177268207073212,
      "rewards-max/forensics_sinkhorn_reward": 0.9847089052200317,
      "rewards-max/forensics_strict_edge_reward": 0.6919911503791809,
      "rewards-max/forensics_tiou_reward": 0.9832380414009094,
      "rewards/forensics_sinkhorn_reward": 0.552923858165741,
      "rewards/forensics_strict_edge_reward": 0.23398921638727188,
      "rewards/forensics_tiou_reward": 0.2956779897212982,
      "step": 712
    },
    {
      "completion_length": 61.53125,
      "epoch": 1.8974700399467377,
      "grad_norm": 46.66682052612305,
      "kl": 4.046875,
      "learning_rate": 5.253333333333333e-07,
      "loss": 0.1774,
      "reward": 0.7006336450576782,
      "reward_std": 0.11739598214626312,
      "reward_std_all": 0.11581088602542877,
      "rewards-max/forensics_sinkhorn_reward": 0.9830297231674194,
      "rewards-max/forensics_strict_edge_reward": 0.87538942694664,
      "rewards-max/forensics_tiou_reward": 0.9766571819782257,
      "rewards/forensics_sinkhorn_reward": 0.5550844073295593,
      "rewards/forensics_strict_edge_reward": 0.30129867792129517,
      "rewards/forensics_tiou_reward": 0.37370140850543976,
      "step": 713
    },
    {
      "completion_length": 63.140625,
      "epoch": 1.900133155792277,
      "grad_norm": 32.78196716308594,
      "kl": 2.7734375,
      "learning_rate": 5.246666666666666e-07,
      "loss": 0.1257,
      "reward": 0.6575621962547302,
      "reward_std": 0.12087713181972504,
      "reward_std_all": 0.13439856469631195,
      "rewards-max/forensics_sinkhorn_reward": 0.9855768382549286,
      "rewards-max/forensics_strict_edge_reward": 0.9252271056175232,
      "rewards-max/forensics_tiou_reward": 0.986606776714325,
      "rewards/forensics_sinkhorn_reward": 0.537207305431366,
      "rewards/forensics_strict_edge_reward": 0.21308740228414536,
      "rewards/forensics_tiou_reward": 0.30353517830371857,
      "step": 714
    },
    {
      "completion_length": 41.4375,
      "epoch": 1.9027962716378162,
      "grad_norm": 27.929115295410156,
      "kl": 2.1171875,
      "learning_rate": 5.24e-07,
      "loss": 0.2401,
      "reward": 0.6246780455112457,
      "reward_std": 0.18887508660554886,
      "reward_std_all": 0.1840968206524849,
      "rewards-max/forensics_sinkhorn_reward": 0.9881555736064911,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9967103004455566,
      "rewards/forensics_sinkhorn_reward": 0.5633405148983002,
      "rewards/forensics_strict_edge_reward": 0.2798166871070862,
      "rewards/forensics_tiou_reward": 0.3953695297241211,
      "step": 715
    },
    {
      "completion_length": 59.078125,
      "epoch": 1.9054593874833556,
      "grad_norm": 17.375181198120117,
      "kl": 2.23046875,
      "learning_rate": 5.233333333333333e-07,
      "loss": 0.2393,
      "reward": 0.6633973717689514,
      "reward_std": 0.11861763894557953,
      "reward_std_all": 0.13256515562534332,
      "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": 0.6073154211044312,
      "rewards/forensics_strict_edge_reward": 0.2824976444244385,
      "rewards/forensics_tiou_reward": 0.38502076268196106,
      "step": 716
    },
    {
      "completion_length": 48.21875,
      "epoch": 1.9081225033288947,
      "grad_norm": 26.239410400390625,
      "kl": 3.125,
      "learning_rate": 5.226666666666666e-07,
      "loss": 0.0266,
      "reward": 0.6285754144191742,
      "reward_std": 0.17385031282901764,
      "reward_std_all": 0.17236236482858658,
      "rewards-max/forensics_sinkhorn_reward": 0.9869152903556824,
      "rewards-max/forensics_strict_edge_reward": 0.9565287232398987,
      "rewards-max/forensics_tiou_reward": 0.9918202757835388,
      "rewards/forensics_sinkhorn_reward": 0.5361298322677612,
      "rewards/forensics_strict_edge_reward": 0.17605111747980118,
      "rewards/forensics_tiou_reward": 0.25815539062023163,
      "step": 717
    },
    {
      "completion_length": 62.765625,
      "epoch": 1.9107856191744341,
      "grad_norm": 26.851192474365234,
      "kl": 2.8671875,
      "learning_rate": 5.22e-07,
      "loss": 0.3549,
      "reward": 0.6376165151596069,
      "reward_std": 0.1513185128569603,
      "reward_std_all": 0.15619337558746338,
      "rewards-max/forensics_sinkhorn_reward": 0.9940548539161682,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9968055188655853,
      "rewards/forensics_sinkhorn_reward": 0.5883159935474396,
      "rewards/forensics_strict_edge_reward": 0.30927594006061554,
      "rewards/forensics_tiou_reward": 0.3862309604883194,
      "step": 718
    },
    {
      "completion_length": 25.484375,
      "epoch": 1.9134487350199734,
      "grad_norm": 23.064735412597656,
      "kl": 1.6015625,
      "learning_rate": 5.213333333333333e-07,
      "loss": 0.183,
      "reward": 0.7078275382518768,
      "reward_std": 0.09476778656244278,
      "reward_std_all": 0.09477631747722626,
      "rewards-max/forensics_sinkhorn_reward": 0.988288402557373,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9972343444824219,
      "rewards/forensics_sinkhorn_reward": 0.6772019267082214,
      "rewards/forensics_strict_edge_reward": 0.39349035918712616,
      "rewards/forensics_tiou_reward": 0.49303293228149414,
      "step": 719
    },
    {
      "completion_length": 47.65625,
      "epoch": 1.9161118508655126,
      "grad_norm": 32.11891555786133,
      "kl": 2.609375,
      "learning_rate": 5.206666666666667e-07,
      "loss": 0.1405,
      "reward": 0.6609460413455963,
      "reward_std": 0.10403326898813248,
      "reward_std_all": 0.1199919693171978,
      "rewards-max/forensics_sinkhorn_reward": 0.9658802449703217,
      "rewards-max/forensics_strict_edge_reward": 0.8320213258266449,
      "rewards-max/forensics_tiou_reward": 0.945559561252594,
      "rewards/forensics_sinkhorn_reward": 0.5624237060546875,
      "rewards/forensics_strict_edge_reward": 0.19215385615825653,
      "rewards/forensics_tiou_reward": 0.28378142416477203,
      "step": 720
    },
    {
      "completion_length": 58.09375,
      "epoch": 1.918774966711052,
      "grad_norm": 27.083112716674805,
      "kl": 2.3203125,
      "learning_rate": 5.2e-07,
      "loss": 0.1819,
      "reward": 0.5921537280082703,
      "reward_std": 0.16266664862632751,
      "reward_std_all": 0.18482082337141037,
      "rewards-max/forensics_sinkhorn_reward": 0.98875692486763,
      "rewards-max/forensics_strict_edge_reward": 0.9890114367008209,
      "rewards-max/forensics_tiou_reward": 0.9990796148777008,
      "rewards/forensics_sinkhorn_reward": 0.544861227273941,
      "rewards/forensics_strict_edge_reward": 0.28741903603076935,
      "rewards/forensics_tiou_reward": 0.34582212567329407,
      "step": 721
    },
    {
      "completion_length": 36.015625,
      "epoch": 1.921438082556591,
      "grad_norm": 17.86674690246582,
      "kl": 2.4140625,
      "learning_rate": 5.193333333333332e-07,
      "loss": 0.2975,
      "reward": 0.6514641046524048,
      "reward_std": 0.14579519256949425,
      "reward_std_all": 0.14422687888145447,
      "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": 0.5933828353881836,
      "rewards/forensics_strict_edge_reward": 0.3265812546014786,
      "rewards/forensics_tiou_reward": 0.4191742241382599,
      "step": 722
    },
    {
      "completion_length": 75.34375,
      "epoch": 1.9241011984021306,
      "grad_norm": 29.622583389282227,
      "kl": 3.06640625,
      "learning_rate": 5.186666666666667e-07,
      "loss": 0.2531,
      "reward": 0.6179235577583313,
      "reward_std": 0.14683447033166885,
      "reward_std_all": 0.1677369549870491,
      "rewards-max/forensics_sinkhorn_reward": 0.9675253033638,
      "rewards-max/forensics_strict_edge_reward": 0.8612285256385803,
      "rewards-max/forensics_tiou_reward": 0.9493789970874786,
      "rewards/forensics_sinkhorn_reward": 0.5115623772144318,
      "rewards/forensics_strict_edge_reward": 0.2522350400686264,
      "rewards/forensics_tiou_reward": 0.2964807376265526,
      "step": 723
    },
    {
      "completion_length": 39.828125,
      "epoch": 1.9267643142476698,
      "grad_norm": 28.701946258544922,
      "kl": 3.328125,
      "learning_rate": 5.18e-07,
      "loss": 0.3228,
      "reward": 0.6473485827445984,
      "reward_std": 0.09069964289665222,
      "reward_std_all": 0.1172054260969162,
      "rewards-max/forensics_sinkhorn_reward": 0.9873929917812347,
      "rewards-max/forensics_strict_edge_reward": 0.9567649364471436,
      "rewards-max/forensics_tiou_reward": 0.9937303960323334,
      "rewards/forensics_sinkhorn_reward": 0.5569533705711365,
      "rewards/forensics_strict_edge_reward": 0.306205190718174,
      "rewards/forensics_tiou_reward": 0.35807134956121445,
      "step": 724
    },
    {
      "completion_length": 78.796875,
      "epoch": 1.929427430093209,
      "grad_norm": 32.89089584350586,
      "kl": 5.0,
      "learning_rate": 5.173333333333333e-07,
      "loss": 0.3883,
      "reward": 0.5935137867927551,
      "reward_std": 0.15030266717076302,
      "reward_std_all": 0.182085819542408,
      "rewards-max/forensics_sinkhorn_reward": 0.9882099032402039,
      "rewards-max/forensics_strict_edge_reward": 0.9782643616199493,
      "rewards-max/forensics_tiou_reward": 0.9969230890274048,
      "rewards/forensics_sinkhorn_reward": 0.5415538400411606,
      "rewards/forensics_strict_edge_reward": 0.26409928500652313,
      "rewards/forensics_tiou_reward": 0.3272237926721573,
      "step": 725
    },
    {
      "completion_length": 37.671875,
      "epoch": 1.9320905459387483,
      "grad_norm": 20.582338333129883,
      "kl": 1.91796875,
      "learning_rate": 5.166666666666667e-07,
      "loss": 0.2691,
      "reward": 0.6894451975822449,
      "reward_std": 0.13889964669942856,
      "reward_std_all": 0.13160191476345062,
      "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": 0.6832585632801056,
      "rewards/forensics_strict_edge_reward": 0.47166475653648376,
      "rewards/forensics_tiou_reward": 0.5606518983840942,
      "step": 726
    },
    {
      "completion_length": 52.03125,
      "epoch": 1.9347536617842875,
      "grad_norm": 26.06234359741211,
      "kl": 3.546875,
      "learning_rate": 5.16e-07,
      "loss": 0.3867,
      "reward": 0.681125670671463,
      "reward_std": 0.1108199954032898,
      "reward_std_all": 0.11538686603307724,
      "rewards-max/forensics_sinkhorn_reward": 0.9749804735183716,
      "rewards-max/forensics_strict_edge_reward": 0.8968285918235779,
      "rewards-max/forensics_tiou_reward": 0.9672049880027771,
      "rewards/forensics_sinkhorn_reward": 0.576499804854393,
      "rewards/forensics_strict_edge_reward": 0.23770882934331894,
      "rewards/forensics_tiou_reward": 0.37842904031276703,
      "step": 727
    },
    {
      "completion_length": 30.5,
      "epoch": 1.937416777629827,
      "grad_norm": 20.002059936523438,
      "kl": 1.8203125,
      "learning_rate": 5.153333333333333e-07,
      "loss": 0.1585,
      "reward": 0.6983921825885773,
      "reward_std": 0.08244802616536617,
      "reward_std_all": 0.09227878414094448,
      "rewards-max/forensics_sinkhorn_reward": 0.9885026514530182,
      "rewards-max/forensics_strict_edge_reward": 0.9782643616199493,
      "rewards-max/forensics_tiou_reward": 0.9980769157409668,
      "rewards/forensics_sinkhorn_reward": 0.6198449432849884,
      "rewards/forensics_strict_edge_reward": 0.37996653467416763,
      "rewards/forensics_tiou_reward": 0.45125918090343475,
      "step": 728
    },
    {
      "completion_length": 50.71875,
      "epoch": 1.940079893475366,
      "grad_norm": 20.052553176879883,
      "kl": 1.537109375,
      "learning_rate": 5.146666666666667e-07,
      "loss": 0.1456,
      "reward": 0.6300046741962433,
      "reward_std": 0.15018897503614426,
      "reward_std_all": 0.16235822439193726,
      "rewards-max/forensics_sinkhorn_reward": 0.9881715476512909,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9967735707759857,
      "rewards/forensics_sinkhorn_reward": 0.5680526494979858,
      "rewards/forensics_strict_edge_reward": 0.2962234169244766,
      "rewards/forensics_tiou_reward": 0.3950353413820267,
      "step": 729
    },
    {
      "completion_length": 39.453125,
      "epoch": 1.9427430093209055,
      "grad_norm": 18.282556533813477,
      "kl": 3.171875,
      "learning_rate": 5.14e-07,
      "loss": 0.313,
      "reward": 0.6410057544708252,
      "reward_std": 0.1576690599322319,
      "reward_std_all": 0.16414742171764374,
      "rewards-max/forensics_sinkhorn_reward": 0.9873669445514679,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9936111867427826,
      "rewards/forensics_sinkhorn_reward": 0.5191115438938141,
      "rewards/forensics_strict_edge_reward": 0.2729024440050125,
      "rewards/forensics_tiou_reward": 0.34637486934661865,
      "step": 730
    },
    {
      "completion_length": 51.21875,
      "epoch": 1.9454061251664447,
      "grad_norm": 25.148292541503906,
      "kl": 3.3203125,
      "learning_rate": 5.133333333333333e-07,
      "loss": 0.1875,
      "reward": 0.6143451631069183,
      "reward_std": 0.12645325437188148,
      "reward_std_all": 0.15643837302923203,
      "rewards-max/forensics_sinkhorn_reward": 0.9847838580608368,
      "rewards-max/forensics_strict_edge_reward": 0.9169811904430389,
      "rewards-max/forensics_tiou_reward": 0.9835289120674133,
      "rewards/forensics_sinkhorn_reward": 0.4935046434402466,
      "rewards/forensics_strict_edge_reward": 0.26258694380521774,
      "rewards/forensics_tiou_reward": 0.3254559338092804,
      "step": 731
    },
    {
      "completion_length": 71.03125,
      "epoch": 1.948069241011984,
      "grad_norm": 50.7841911315918,
      "kl": 2.578125,
      "learning_rate": 5.126666666666667e-07,
      "loss": 0.2027,
      "reward": 0.675344705581665,
      "reward_std": 0.13177600130438805,
      "reward_std_all": 0.1386285237967968,
      "rewards-max/forensics_sinkhorn_reward": 0.9837428033351898,
      "rewards-max/forensics_strict_edge_reward": 0.9062341153621674,
      "rewards-max/forensics_tiou_reward": 0.9794853925704956,
      "rewards/forensics_sinkhorn_reward": 0.6114016771316528,
      "rewards/forensics_strict_edge_reward": 0.33634914457798004,
      "rewards/forensics_tiou_reward": 0.39419716596603394,
      "step": 732
    },
    {
      "completion_length": 46.671875,
      "epoch": 1.9507323568575234,
      "grad_norm": 47.963401794433594,
      "kl": 3.484375,
      "learning_rate": 5.12e-07,
      "loss": 0.2854,
      "reward": 0.6412860751152039,
      "reward_std": 0.1473146341741085,
      "reward_std_all": 0.15784921497106552,
      "rewards-max/forensics_sinkhorn_reward": 0.9879404008388519,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9958615303039551,
      "rewards/forensics_sinkhorn_reward": 0.4860243797302246,
      "rewards/forensics_strict_edge_reward": 0.22312135994434357,
      "rewards/forensics_tiou_reward": 0.2724275216460228,
      "step": 733
    },
    {
      "completion_length": 41.953125,
      "epoch": 1.9533954727030625,
      "grad_norm": 31.976821899414062,
      "kl": 3.6171875,
      "learning_rate": 5.113333333333333e-07,
      "loss": 0.3644,
      "reward": 0.6381317675113678,
      "reward_std": 0.1667315512895584,
      "reward_std_all": 0.16791589558124542,
      "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": 0.5859761834144592,
      "rewards/forensics_strict_edge_reward": 0.3005120977759361,
      "rewards/forensics_tiou_reward": 0.35337115824222565,
      "step": 734
    },
    {
      "completion_length": 51.796875,
      "epoch": 1.956058588548602,
      "grad_norm": 42.170860290527344,
      "kl": 3.1953125,
      "learning_rate": 5.106666666666667e-07,
      "loss": 0.2743,
      "reward": 0.6425076425075531,
      "reward_std": 0.134210754185915,
      "reward_std_all": 0.15005561709403992,
      "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": 0.5622187703847885,
      "rewards/forensics_strict_edge_reward": 0.2716982588171959,
      "rewards/forensics_tiou_reward": 0.3528400957584381,
      "step": 735
    },
    {
      "completion_length": 34.9375,
      "epoch": 1.9587217043941412,
      "grad_norm": 22.34850311279297,
      "kl": 2.5859375,
      "learning_rate": 5.1e-07,
      "loss": 0.1306,
      "reward": 0.6266738176345825,
      "reward_std": 0.17919722944498062,
      "reward_std_all": 0.17801424860954285,
      "rewards-max/forensics_sinkhorn_reward": 0.9878629148006439,
      "rewards-max/forensics_strict_edge_reward": 0.9677534997463226,
      "rewards-max/forensics_tiou_reward": 0.9955752193927765,
      "rewards/forensics_sinkhorn_reward": 0.5488005876541138,
      "rewards/forensics_strict_edge_reward": 0.3070945292711258,
      "rewards/forensics_tiou_reward": 0.37082166969776154,
      "step": 736
    },
    {
      "completion_length": 36.8125,
      "epoch": 1.9613848202396804,
      "grad_norm": 24.738061904907227,
      "kl": 3.4921875,
      "learning_rate": 5.093333333333332e-07,
      "loss": 0.2276,
      "reward": 0.6472560465335846,
      "reward_std": 0.16485853493213654,
      "reward_std_all": 0.1634371653199196,
      "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": 0.5300170332193375,
      "rewards/forensics_strict_edge_reward": 0.2534000724554062,
      "rewards/forensics_tiou_reward": 0.34997978806495667,
      "step": 737
    },
    {
      "completion_length": 47.546875,
      "epoch": 1.9640479360852197,
      "grad_norm": 28.00408172607422,
      "kl": 3.4140625,
      "learning_rate": 5.086666666666667e-07,
      "loss": 0.2666,
      "reward": 0.6194522380828857,
      "reward_std": 0.1019517220556736,
      "reward_std_all": 0.14165277779102325,
      "rewards-max/forensics_sinkhorn_reward": 0.9703923463821411,
      "rewards-max/forensics_strict_edge_reward": 0.8277941644191742,
      "rewards-max/forensics_tiou_reward": 0.9601457118988037,
      "rewards/forensics_sinkhorn_reward": 0.49718576669692993,
      "rewards/forensics_strict_edge_reward": 0.1681453175842762,
      "rewards/forensics_tiou_reward": 0.2801100164651871,
      "step": 738
    },
    {
      "completion_length": 52.1875,
      "epoch": 1.966711051930759,
      "grad_norm": 24.73984718322754,
      "kl": 5.296875,
      "learning_rate": 5.079999999999999e-07,
      "loss": 0.3524,
      "reward": 0.6827394068241119,
      "reward_std": 0.12226370722055435,
      "reward_std_all": 0.12208326905965805,
      "rewards-max/forensics_sinkhorn_reward": 0.9866061806678772,
      "rewards-max/forensics_strict_edge_reward": 0.927969753742218,
      "rewards-max/forensics_tiou_reward": 0.9906040132045746,
      "rewards/forensics_sinkhorn_reward": 0.5787769556045532,
      "rewards/forensics_strict_edge_reward": 0.24747316539287567,
      "rewards/forensics_tiou_reward": 0.37311123311519623,
      "step": 739
    },
    {
      "completion_length": 39.0,
      "epoch": 1.9693741677762984,
      "grad_norm": 27.936342239379883,
      "kl": 2.6875,
      "learning_rate": 5.073333333333333e-07,
      "loss": 0.2085,
      "reward": 0.6849390864372253,
      "reward_std": 0.1361275240778923,
      "reward_std_all": 0.1312607228755951,
      "rewards-max/forensics_sinkhorn_reward": 0.9879870414733887,
      "rewards-max/forensics_strict_edge_reward": 0.9464850425720215,
      "rewards-max/forensics_tiou_reward": 0.9960491955280304,
      "rewards/forensics_sinkhorn_reward": 0.6071517467498779,
      "rewards/forensics_strict_edge_reward": 0.30810898542404175,
      "rewards/forensics_tiou_reward": 0.39995482563972473,
      "step": 740
    },
    {
      "completion_length": 74.015625,
      "epoch": 1.9720372836218374,
      "grad_norm": 34.93455123901367,
      "kl": 3.2265625,
      "learning_rate": 5.066666666666667e-07,
      "loss": 0.4229,
      "reward": 0.5619617402553558,
      "reward_std": 0.2226625457406044,
      "reward_std_all": 0.2308761328458786,
      "rewards-max/forensics_sinkhorn_reward": 0.9569372534751892,
      "rewards-max/forensics_strict_edge_reward": 0.9142521619796753,
      "rewards-max/forensics_tiou_reward": 0.9308549463748932,
      "rewards/forensics_sinkhorn_reward": 0.49154236912727356,
      "rewards/forensics_strict_edge_reward": 0.3066743463277817,
      "rewards/forensics_tiou_reward": 0.36784976720809937,
      "step": 741
    },
    {
      "completion_length": 37.8125,
      "epoch": 1.9747003994673769,
      "grad_norm": 23.21198081970215,
      "kl": 2.5625,
      "learning_rate": 5.06e-07,
      "loss": 0.2185,
      "reward": 0.6340007185935974,
      "reward_std": 0.1331970915198326,
      "reward_std_all": 0.14704939350485802,
      "rewards-max/forensics_sinkhorn_reward": 0.9876636564731598,
      "rewards-max/forensics_strict_edge_reward": 0.9565287232398987,
      "rewards-max/forensics_tiou_reward": 0.9947704970836639,
      "rewards/forensics_sinkhorn_reward": 0.5044723153114319,
      "rewards/forensics_strict_edge_reward": 0.17045816779136658,
      "rewards/forensics_tiou_reward": 0.2963074892759323,
      "step": 742
    },
    {
      "completion_length": 27.046875,
      "epoch": 1.977363515312916,
      "grad_norm": 29.996000289916992,
      "kl": 1.4765625,
      "learning_rate": 5.053333333333333e-07,
      "loss": 0.1837,
      "reward": 0.6940320730209351,
      "reward_std": 0.14523354917764664,
      "reward_std_all": 0.1334780901670456,
      "rewards-max/forensics_sinkhorn_reward": 0.987968236207962,
      "rewards-max/forensics_strict_edge_reward": 0.9565287232398987,
      "rewards-max/forensics_tiou_reward": 0.9959765374660492,
      "rewards/forensics_sinkhorn_reward": 0.6406291127204895,
      "rewards/forensics_strict_edge_reward": 0.2766568064689636,
      "rewards/forensics_tiou_reward": 0.43942368030548096,
      "step": 743
    },
    {
      "completion_length": 45.671875,
      "epoch": 1.9800266311584553,
      "grad_norm": 32.973533630371094,
      "kl": 2.65625,
      "learning_rate": 5.046666666666667e-07,
      "loss": 0.3132,
      "reward": 0.6585588753223419,
      "reward_std": 0.10935039818286896,
      "reward_std_all": 0.12633396312594414,
      "rewards-max/forensics_sinkhorn_reward": 0.9614607095718384,
      "rewards-max/forensics_strict_edge_reward": 0.8983768224716187,
      "rewards-max/forensics_tiou_reward": 0.9415535926818848,
      "rewards/forensics_sinkhorn_reward": 0.5403444766998291,
      "rewards/forensics_strict_edge_reward": 0.24120599776506424,
      "rewards/forensics_tiou_reward": 0.2839593142271042,
      "step": 744
    },
    {
      "completion_length": 49.796875,
      "epoch": 1.9826897470039948,
      "grad_norm": 21.100692749023438,
      "kl": 3.328125,
      "learning_rate": 5.04e-07,
      "loss": 0.2389,
      "reward": 0.6394239366054535,
      "reward_std": 0.12827710062265396,
      "reward_std_all": 0.14167937636375427,
      "rewards-max/forensics_sinkhorn_reward": 0.9863003790378571,
      "rewards-max/forensics_strict_edge_reward": 0.9364311099052429,
      "rewards-max/forensics_tiou_reward": 0.989441305398941,
      "rewards/forensics_sinkhorn_reward": 0.5287648886442184,
      "rewards/forensics_strict_edge_reward": 0.25010138750076294,
      "rewards/forensics_tiou_reward": 0.3280162811279297,
      "step": 745
    },
    {
      "completion_length": 47.3125,
      "epoch": 1.9853528628495338,
      "grad_norm": 32.795536041259766,
      "kl": 3.6328125,
      "learning_rate": 5.033333333333333e-07,
      "loss": 0.2226,
      "reward": 0.6242912411689758,
      "reward_std": 0.16465827822685242,
      "reward_std_all": 0.17201651632785797,
      "rewards-max/forensics_sinkhorn_reward": 0.9887834191322327,
      "rewards-max/forensics_strict_edge_reward": 0.9782643616199493,
      "rewards-max/forensics_tiou_reward": 0.9991836845874786,
      "rewards/forensics_sinkhorn_reward": 0.5941336750984192,
      "rewards/forensics_strict_edge_reward": 0.3397732377052307,
      "rewards/forensics_tiou_reward": 0.4242032319307327,
      "step": 746
    },
    {
      "completion_length": 45.640625,
      "epoch": 1.9880159786950733,
      "grad_norm": 21.74892807006836,
      "kl": 3.25,
      "learning_rate": 5.026666666666667e-07,
      "loss": 0.2905,
      "reward": 0.6122450232505798,
      "reward_std": 0.15177834406495094,
      "reward_std_all": 0.1629420518875122,
      "rewards-max/forensics_sinkhorn_reward": 0.9698415100574493,
      "rewards-max/forensics_strict_edge_reward": 0.8771188855171204,
      "rewards-max/forensics_tiou_reward": 0.9518537819385529,
      "rewards/forensics_sinkhorn_reward": 0.481404572725296,
      "rewards/forensics_strict_edge_reward": 0.17548933252692223,
      "rewards/forensics_tiou_reward": 0.2551712691783905,
      "step": 747
    },
    {
      "completion_length": 29.828125,
      "epoch": 1.9906790945406125,
      "grad_norm": 101.47584533691406,
      "kl": 2.98828125,
      "learning_rate": 5.02e-07,
      "loss": 0.3249,
      "reward": 0.6608009040355682,
      "reward_std": 0.12826604396104813,
      "reward_std_all": 0.13173412904143333,
      "rewards-max/forensics_sinkhorn_reward": 0.9882018268108368,
      "rewards-max/forensics_strict_edge_reward": 0.9093653857707977,
      "rewards-max/forensics_tiou_reward": 0.99689120054245,
      "rewards/forensics_sinkhorn_reward": 0.5750077664852142,
      "rewards/forensics_strict_edge_reward": 0.30492012202739716,
      "rewards/forensics_tiou_reward": 0.3774305284023285,
      "step": 748
    },
    {
      "completion_length": 42.21875,
      "epoch": 1.9933422103861518,
      "grad_norm": 25.42799949645996,
      "kl": 2.46875,
      "learning_rate": 5.013333333333333e-07,
      "loss": 0.295,
      "reward": 0.5949202179908752,
      "reward_std": 0.18158456683158875,
      "reward_std_all": 0.196820467710495,
      "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": 0.535341888666153,
      "rewards/forensics_strict_edge_reward": 0.30843237042427063,
      "rewards/forensics_tiou_reward": 0.3528534173965454,
      "step": 749
    },
    {
      "completion_length": 61.765625,
      "epoch": 1.996005326231691,
      "grad_norm": 24.897438049316406,
      "kl": 3.29296875,
      "learning_rate": 5.006666666666667e-07,
      "loss": 0.3961,
      "reward": 0.5917333364486694,
      "reward_std": 0.2320188209414482,
      "reward_std_all": 0.23040282726287842,
      "rewards-max/forensics_sinkhorn_reward": 0.988649994134903,
      "rewards-max/forensics_strict_edge_reward": 0.9782643616199493,
      "rewards-max/forensics_tiou_reward": 0.9986577033996582,
      "rewards/forensics_sinkhorn_reward": 0.5191376805305481,
      "rewards/forensics_strict_edge_reward": 0.23830975592136383,
      "rewards/forensics_tiou_reward": 0.325795441865921,
      "step": 750
    },
    {
      "completion_length": 43.21875,
      "epoch": 1.9986684420772303,
      "grad_norm": 32.37567901611328,
      "kl": 3.390625,
      "learning_rate": 5e-07,
      "loss": 0.2952,
      "reward": 0.6109482944011688,
      "reward_std": 0.18456435203552246,
      "reward_std_all": 0.18471184372901917,
      "rewards-max/forensics_sinkhorn_reward": 0.9636115431785583,
      "rewards-max/forensics_strict_edge_reward": 0.8805811703205109,
      "rewards-max/forensics_tiou_reward": 0.9428074955940247,
      "rewards/forensics_sinkhorn_reward": 0.5415833592414856,
      "rewards/forensics_strict_edge_reward": 0.26407913863658905,
      "rewards/forensics_tiou_reward": 0.3308684974908829,
      "step": 751
    },
    {
      "completion_length": 30.0,
      "epoch": 2.0,
      "grad_norm": 13.77607536315918,
      "kl": 2.28125,
      "learning_rate": 4.993333333333333e-07,
      "loss": 0.0825,
      "reward": 0.5038851499557495,
      "reward_std": 0.1689513623714447,
      "reward_std_all": 0.17092128098011017,
      "rewards-max/forensics_sinkhorn_reward": 0.25564467906951904,
      "rewards-max/forensics_strict_edge_reward": 0.126069113612175,
      "rewards-max/forensics_tiou_reward": 0.006965788546949625,
      "rewards/forensics_sinkhorn_reward": 0.25564467906951904,
      "rewards/forensics_strict_edge_reward": 0.126069113612175,
      "rewards/forensics_tiou_reward": 0.006965788546949625,
      "step": 752
    },
    {
      "completion_length": 68.3125,
      "epoch": 2.0026631158455395,
      "grad_norm": 30.514083862304688,
      "kl": 3.3671875,
      "learning_rate": 4.986666666666666e-07,
      "loss": 0.2554,
      "reward": 0.6496777534484863,
      "reward_std": 0.18957962840795517,
      "reward_std_all": 0.1769595667719841,
      "rewards-max/forensics_sinkhorn_reward": 0.9853009283542633,
      "rewards-max/forensics_strict_edge_reward": 0.9169811904430389,
      "rewards-max/forensics_tiou_reward": 0.9855399429798126,
      "rewards/forensics_sinkhorn_reward": 0.5670635104179382,
      "rewards/forensics_strict_edge_reward": 0.25079867243766785,
      "rewards/forensics_tiou_reward": 0.38721564412117004,
      "step": 753
    },
    {
      "completion_length": 33.90625,
      "epoch": 2.0053262316910785,
      "grad_norm": 30.49146842956543,
      "kl": 2.046875,
      "learning_rate": 4.979999999999999e-07,
      "loss": 0.2561,
      "reward": 0.6527973413467407,
      "reward_std": 0.18611864745616913,
      "reward_std_all": 0.1727791726589203,
      "rewards-max/forensics_sinkhorn_reward": 0.9885083138942719,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9981001615524292,
      "rewards/forensics_sinkhorn_reward": 0.6433434188365936,
      "rewards/forensics_strict_edge_reward": 0.42020663619041443,
      "rewards/forensics_tiou_reward": 0.5109311044216156,
      "step": 754
    },
    {
      "completion_length": 52.34375,
      "epoch": 2.007989347536618,
      "grad_norm": 22.749187469482422,
      "kl": 2.6328125,
      "learning_rate": 4.973333333333333e-07,
      "loss": 0.3128,
      "reward": 0.5605665147304535,
      "reward_std": 0.16754847764968872,
      "reward_std_all": 0.20385070145130157,
      "rewards-max/forensics_sinkhorn_reward": 0.9873413741588593,
      "rewards-max/forensics_strict_edge_reward": 0.927969753742218,
      "rewards-max/forensics_tiou_reward": 0.9935216009616852,
      "rewards/forensics_sinkhorn_reward": 0.467722088098526,
      "rewards/forensics_strict_edge_reward": 0.19686312228441238,
      "rewards/forensics_tiou_reward": 0.24862562865018845,
      "step": 755
    },
    {
      "completion_length": 36.109375,
      "epoch": 2.010652463382157,
      "grad_norm": 23.619436264038086,
      "kl": 1.75390625,
      "learning_rate": 4.966666666666666e-07,
      "loss": 0.1006,
      "reward": 0.6412440538406372,
      "reward_std": 0.18564801663160324,
      "reward_std_all": 0.1764824539422989,
      "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": 0.580197811126709,
      "rewards/forensics_strict_edge_reward": 0.2576861083507538,
      "rewards/forensics_tiou_reward": 0.3800840377807617,
      "step": 756
    },
    {
      "completion_length": 48.0625,
      "epoch": 2.0133155792276964,
      "grad_norm": 25.263553619384766,
      "kl": 2.81640625,
      "learning_rate": 4.96e-07,
      "loss": 0.1481,
      "reward": 0.6648366451263428,
      "reward_std": 0.13763565197587013,
      "reward_std_all": 0.13979361951351166,
      "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": 0.6187108755111694,
      "rewards/forensics_strict_edge_reward": 0.3157292529940605,
      "rewards/forensics_tiou_reward": 0.41200103610754013,
      "step": 757
    },
    {
      "completion_length": 33.796875,
      "epoch": 2.015978695073236,
      "grad_norm": 23.278213500976562,
      "kl": 1.8984375,
      "learning_rate": 4.953333333333333e-07,
      "loss": 0.2692,
      "reward": 0.6531952619552612,
      "reward_std": 0.09381669014692307,
      "reward_std_all": 0.12627286463975906,
      "rewards-max/forensics_sinkhorn_reward": 0.9881378412246704,
      "rewards-max/forensics_strict_edge_reward": 0.9567649364471436,
      "rewards-max/forensics_tiou_reward": 0.9966418743133545,
      "rewards/forensics_sinkhorn_reward": 0.5878407955169678,
      "rewards/forensics_strict_edge_reward": 0.2667321115732193,
      "rewards/forensics_tiou_reward": 0.4058656394481659,
      "step": 758
    },
    {
      "completion_length": 40.078125,
      "epoch": 2.018641810918775,
      "grad_norm": 27.425289154052734,
      "kl": 2.75390625,
      "learning_rate": 4.946666666666666e-07,
      "loss": 0.3071,
      "reward": 0.592978835105896,
      "reward_std": 0.17356255650520325,
      "reward_std_all": 0.18855180591344833,
      "rewards-max/forensics_sinkhorn_reward": 0.9755070805549622,
      "rewards-max/forensics_strict_edge_reward": 0.8915697336196899,
      "rewards-max/forensics_tiou_reward": 0.9695107638835907,
      "rewards/forensics_sinkhorn_reward": 0.47666046023368835,
      "rewards/forensics_strict_edge_reward": 0.258505254983902,
      "rewards/forensics_tiou_reward": 0.35518157482147217,
      "step": 759
    },
    {
      "completion_length": 53.046875,
      "epoch": 2.0213049267643144,
      "grad_norm": 38.561553955078125,
      "kl": 3.9375,
      "learning_rate": 4.94e-07,
      "loss": 0.2626,
      "reward": 0.6102192103862762,
      "reward_std": 0.17784545570611954,
      "reward_std_all": 0.19049779325723648,
      "rewards-max/forensics_sinkhorn_reward": 0.9771125316619873,
      "rewards-max/forensics_strict_edge_reward": 0.8771188855171204,
      "rewards-max/forensics_tiou_reward": 0.9723034203052521,
      "rewards/forensics_sinkhorn_reward": 0.5441673994064331,
      "rewards/forensics_strict_edge_reward": 0.28799933195114136,
      "rewards/forensics_tiou_reward": 0.35297828912734985,
      "step": 760
    },
    {
      "completion_length": 26.296875,
      "epoch": 2.0239680426098534,
      "grad_norm": 32.976016998291016,
      "kl": 3.6015625,
      "learning_rate": 4.933333333333333e-07,
      "loss": 0.2863,
      "reward": 0.674024909734726,
      "reward_std": 0.12411553412675858,
      "reward_std_all": 0.12841470167040825,
      "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": 0.5939656496047974,
      "rewards/forensics_strict_edge_reward": 0.28264273703098297,
      "rewards/forensics_tiou_reward": 0.3792007863521576,
      "step": 761
    },
    {
      "completion_length": 30.96875,
      "epoch": 2.026631158455393,
      "grad_norm": 24.70726203918457,
      "kl": 2.2421875,
      "learning_rate": 4.926666666666667e-07,
      "loss": 0.1828,
      "reward": 0.623591423034668,
      "reward_std": 0.18173345178365707,
      "reward_std_all": 0.1826571673154831,
      "rewards-max/forensics_sinkhorn_reward": 0.9872008860111237,
      "rewards-max/forensics_strict_edge_reward": 0.9158510267734528,
      "rewards-max/forensics_tiou_reward": 0.9929547607898712,
      "rewards/forensics_sinkhorn_reward": 0.5234817564487457,
      "rewards/forensics_strict_edge_reward": 0.28798189759254456,
      "rewards/forensics_tiou_reward": 0.38522113859653473,
      "step": 762
    },
    {
      "completion_length": 44.46875,
      "epoch": 2.029294274300932,
      "grad_norm": 45.12177658081055,
      "kl": 3.7109375,
      "learning_rate": 4.92e-07,
      "loss": 0.2599,
      "reward": 0.6184248328208923,
      "reward_std": 0.1439514085650444,
      "reward_std_all": 0.16533558815717697,
      "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": 0.5446945130825043,
      "rewards/forensics_strict_edge_reward": 0.21803759038448334,
      "rewards/forensics_tiou_reward": 0.29613494873046875,
      "step": 763
    },
    {
      "completion_length": 29.984375,
      "epoch": 2.0319573901464714,
      "grad_norm": 21.57252311706543,
      "kl": 2.6875,
      "learning_rate": 4.913333333333334e-07,
      "loss": 0.1317,
      "reward": 0.6719738841056824,
      "reward_std": 0.09090804681181908,
      "reward_std_all": 0.11348244175314903,
      "rewards-max/forensics_sinkhorn_reward": 0.9886920750141144,
      "rewards-max/forensics_strict_edge_reward": 0.9890114367008209,
      "rewards-max/forensics_tiou_reward": 0.9988235235214233,
      "rewards/forensics_sinkhorn_reward": 0.649415910243988,
      "rewards/forensics_strict_edge_reward": 0.35294242203235626,
      "rewards/forensics_tiou_reward": 0.46215127408504486,
      "step": 764
    },
    {
      "completion_length": 34.796875,
      "epoch": 2.034620505992011,
      "grad_norm": 31.541650772094727,
      "kl": 2.7578125,
      "learning_rate": 4.906666666666666e-07,
      "loss": 0.2581,
      "reward": 0.6378946006298065,
      "reward_std": 0.14719976484775543,
      "reward_std_all": 0.1519964188337326,
      "rewards-max/forensics_sinkhorn_reward": 0.9802449643611908,
      "rewards-max/forensics_strict_edge_reward": 0.9256840348243713,
      "rewards-max/forensics_tiou_reward": 0.9789565205574036,
      "rewards/forensics_sinkhorn_reward": 0.4971639961004257,
      "rewards/forensics_strict_edge_reward": 0.20179372280836105,
      "rewards/forensics_tiou_reward": 0.2637169063091278,
      "step": 765
    },
    {
      "completion_length": 32.15625,
      "epoch": 2.03728362183755,
      "grad_norm": 30.67522621154785,
      "kl": 4.2421875,
      "learning_rate": 4.9e-07,
      "loss": 0.4674,
      "reward": 0.5784075558185577,
      "reward_std": 0.17919405549764633,
      "reward_std_all": 0.19741247594356537,
      "rewards-max/forensics_sinkhorn_reward": 0.9508887529373169,
      "rewards-max/forensics_strict_edge_reward": 0.7393108606338501,
      "rewards-max/forensics_tiou_reward": 0.9153552055358887,
      "rewards/forensics_sinkhorn_reward": 0.4878290295600891,
      "rewards/forensics_strict_edge_reward": 0.2564289793372154,
      "rewards/forensics_tiou_reward": 0.3533112555742264,
      "step": 766
    },
    {
      "completion_length": 25.0,
      "epoch": 2.0399467376830893,
      "grad_norm": 20.64589500427246,
      "kl": 3.046875,
      "learning_rate": 4.893333333333333e-07,
      "loss": 0.2438,
      "reward": 0.6463129818439484,
      "reward_std": 0.15443909168243408,
      "reward_std_all": 0.1593417003750801,
      "rewards-max/forensics_sinkhorn_reward": 0.9670555293560028,
      "rewards-max/forensics_strict_edge_reward": 0.830383837223053,
      "rewards-max/forensics_tiou_reward": 0.9472827911376953,
      "rewards/forensics_sinkhorn_reward": 0.5538673996925354,
      "rewards/forensics_strict_edge_reward": 0.2456035017967224,
      "rewards/forensics_tiou_reward": 0.3165811747312546,
      "step": 767
    },
    {
      "completion_length": 31.453125,
      "epoch": 2.0426098535286283,
      "grad_norm": 35.09823989868164,
      "kl": 2.3359375,
      "learning_rate": 4.886666666666667e-07,
      "loss": 0.2409,
      "reward": 0.6525544822216034,
      "reward_std": 0.16363690048456192,
      "reward_std_all": 0.16362857818603516,
      "rewards-max/forensics_sinkhorn_reward": 0.987930417060852,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9958254992961884,
      "rewards/forensics_sinkhorn_reward": 0.5792020261287689,
      "rewards/forensics_strict_edge_reward": 0.28543415665626526,
      "rewards/forensics_tiou_reward": 0.3656492531299591,
      "step": 768
    },
    {
      "completion_length": 37.6875,
      "epoch": 2.045272969374168,
      "grad_norm": 23.662418365478516,
      "kl": 3.8828125,
      "learning_rate": 4.879999999999999e-07,
      "loss": 0.3437,
      "reward": 0.6026284396648407,
      "reward_std": 0.16573835164308548,
      "reward_std_all": 0.18133395910263062,
      "rewards-max/forensics_sinkhorn_reward": 0.9839130938053131,
      "rewards-max/forensics_strict_edge_reward": 0.905340164899826,
      "rewards-max/forensics_tiou_reward": 0.9801206886768341,
      "rewards/forensics_sinkhorn_reward": 0.51863893866539,
      "rewards/forensics_strict_edge_reward": 0.22044990956783295,
      "rewards/forensics_tiou_reward": 0.2827776372432709,
      "step": 769
    },
    {
      "completion_length": 24.03125,
      "epoch": 2.0479360852197073,
      "grad_norm": 27.068525314331055,
      "kl": 2.625,
      "learning_rate": 4.873333333333333e-07,
      "loss": 0.3828,
      "reward": 0.669321596622467,
      "reward_std": 0.12947329133749008,
      "reward_std_all": 0.13354666531085968,
      "rewards-max/forensics_sinkhorn_reward": 0.9886316061019897,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9985857307910919,
      "rewards/forensics_sinkhorn_reward": 0.6284947395324707,
      "rewards/forensics_strict_edge_reward": 0.3173149526119232,
      "rewards/forensics_tiou_reward": 0.4630666673183441,
      "step": 770
    },
    {
      "completion_length": 48.015625,
      "epoch": 2.0505992010652463,
      "grad_norm": 38.68260955810547,
      "kl": 5.625,
      "learning_rate": 4.866666666666666e-07,
      "loss": 0.3975,
      "reward": 0.5490012168884277,
      "reward_std": 0.18371574580669403,
      "reward_std_all": 0.20635052025318146,
      "rewards-max/forensics_sinkhorn_reward": 0.9691063463687897,
      "rewards-max/forensics_strict_edge_reward": 0.8968285918235779,
      "rewards-max/forensics_tiou_reward": 0.9543443024158478,
      "rewards/forensics_sinkhorn_reward": 0.4428844153881073,
      "rewards/forensics_strict_edge_reward": 0.23751066625118256,
      "rewards/forensics_tiou_reward": 0.30345572531223297,
      "step": 771
    },
    {
      "completion_length": 34.796875,
      "epoch": 2.0532623169107858,
      "grad_norm": 32.19666290283203,
      "kl": 3.4609375,
      "learning_rate": 4.86e-07,
      "loss": 0.2989,
      "reward": 0.6975841224193573,
      "reward_std": 0.07412074133753777,
      "reward_std_all": 0.09315726533532143,
      "rewards-max/forensics_sinkhorn_reward": 0.9886549413204193,
      "rewards-max/forensics_strict_edge_reward": 0.9890114367008209,
      "rewards-max/forensics_tiou_reward": 0.9986781179904938,
      "rewards/forensics_sinkhorn_reward": 0.6369367837905884,
      "rewards/forensics_strict_edge_reward": 0.2928561493754387,
      "rewards/forensics_tiou_reward": 0.4417680948972702,
      "step": 772
    },
    {
      "completion_length": 39.0625,
      "epoch": 2.0559254327563248,
      "grad_norm": 29.126768112182617,
      "kl": 5.34375,
      "learning_rate": 4.853333333333333e-07,
      "loss": 0.4188,
      "reward": 0.657939225435257,
      "reward_std": 0.14226198941469193,
      "reward_std_all": 0.14772792160511017,
      "rewards-max/forensics_sinkhorn_reward": 0.9847933351993561,
      "rewards-max/forensics_strict_edge_reward": 0.9151731729507446,
      "rewards-max/forensics_tiou_reward": 0.9835384786128998,
      "rewards/forensics_sinkhorn_reward": 0.5774043500423431,
      "rewards/forensics_strict_edge_reward": 0.26671262830495834,
      "rewards/forensics_tiou_reward": 0.38163498044013977,
      "step": 773
    },
    {
      "completion_length": 62.59375,
      "epoch": 2.0585885486018642,
      "grad_norm": 23.610963821411133,
      "kl": 3.625,
      "learning_rate": 4.846666666666667e-07,
      "loss": 0.2312,
      "reward": 0.6921237707138062,
      "reward_std": 0.16371510922908783,
      "reward_std_all": 0.14713438600301743,
      "rewards-max/forensics_sinkhorn_reward": 0.9855024516582489,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9910093247890472,
      "rewards/forensics_sinkhorn_reward": 0.6093597710132599,
      "rewards/forensics_strict_edge_reward": 0.2929801940917969,
      "rewards/forensics_tiou_reward": 0.4652464687824249,
      "step": 774
    },
    {
      "completion_length": 37.1875,
      "epoch": 2.0612516644474033,
      "grad_norm": 22.203998565673828,
      "kl": 3.609375,
      "learning_rate": 4.839999999999999e-07,
      "loss": 0.2901,
      "reward": 0.6516745686531067,
      "reward_std": 0.17489446699619293,
      "reward_std_all": 0.16600334644317627,
      "rewards-max/forensics_sinkhorn_reward": 0.9848044514656067,
      "rewards-max/forensics_strict_edge_reward": 0.9169811904430389,
      "rewards-max/forensics_tiou_reward": 0.9836398065090179,
      "rewards/forensics_sinkhorn_reward": 0.545935869216919,
      "rewards/forensics_strict_edge_reward": 0.28251227736473083,
      "rewards/forensics_tiou_reward": 0.3802631348371506,
      "step": 775
    },
    {
      "completion_length": 46.0,
      "epoch": 2.0639147802929427,
      "grad_norm": 28.941604614257812,
      "kl": 3.8125,
      "learning_rate": 4.833333333333333e-07,
      "loss": 0.3167,
      "reward": 0.6763834357261658,
      "reward_std": 0.1856854259967804,
      "reward_std_all": 0.16693215072155,
      "rewards-max/forensics_sinkhorn_reward": 0.9856926798820496,
      "rewards-max/forensics_strict_edge_reward": 0.878633052110672,
      "rewards-max/forensics_tiou_reward": 0.9870691001415253,
      "rewards/forensics_sinkhorn_reward": 0.5637682974338531,
      "rewards/forensics_strict_edge_reward": 0.30150918662548065,
      "rewards/forensics_tiou_reward": 0.42633724212646484,
      "step": 776
    },
    {
      "completion_length": 61.484375,
      "epoch": 2.066577896138482,
      "grad_norm": 30.03438377380371,
      "kl": 4.859375,
      "learning_rate": 4.826666666666666e-07,
      "loss": 0.3844,
      "reward": 0.6552611887454987,
      "reward_std": 0.16122370958328247,
      "reward_std_all": 0.15437600016593933,
      "rewards-max/forensics_sinkhorn_reward": 0.9658338129520416,
      "rewards-max/forensics_strict_edge_reward": 0.8351600170135498,
      "rewards-max/forensics_tiou_reward": 0.9474215805530548,
      "rewards/forensics_sinkhorn_reward": 0.5549608469009399,
      "rewards/forensics_strict_edge_reward": 0.31132251769304276,
      "rewards/forensics_tiou_reward": 0.36295782774686813,
      "step": 777
    },
    {
      "completion_length": 35.9375,
      "epoch": 2.069241011984021,
      "grad_norm": 27.127065658569336,
      "kl": 3.40625,
      "learning_rate": 4.82e-07,
      "loss": 0.2278,
      "reward": 0.6052506864070892,
      "reward_std": 0.12878866121172905,
      "reward_std_all": 0.15284844487905502,
      "rewards-max/forensics_sinkhorn_reward": 0.9819740056991577,
      "rewards-max/forensics_strict_edge_reward": 0.8983768224716187,
      "rewards-max/forensics_tiou_reward": 0.9830670952796936,
      "rewards/forensics_sinkhorn_reward": 0.4525325298309326,
      "rewards/forensics_strict_edge_reward": 0.20769137889146805,
      "rewards/forensics_tiou_reward": 0.26873811334371567,
      "step": 778
    },
    {
      "completion_length": 62.0,
      "epoch": 2.0719041278295607,
      "grad_norm": 49.13288497924805,
      "kl": 4.046875,
      "learning_rate": 4.813333333333334e-07,
      "loss": 0.3395,
      "reward": 0.6428032517433167,
      "reward_std": 0.13345623761415482,
      "reward_std_all": 0.1448722407221794,
      "rewards-max/forensics_sinkhorn_reward": 0.9665073752403259,
      "rewards-max/forensics_strict_edge_reward": 0.8854433596134186,
      "rewards-max/forensics_tiou_reward": 0.9494971930980682,
      "rewards/forensics_sinkhorn_reward": 0.5803672075271606,
      "rewards/forensics_strict_edge_reward": 0.24930336326360703,
      "rewards/forensics_tiou_reward": 0.380446121096611,
      "step": 779
    },
    {
      "completion_length": 33.15625,
      "epoch": 2.0745672436750997,
      "grad_norm": 26.55333709716797,
      "kl": 2.056640625,
      "learning_rate": 4.806666666666667e-07,
      "loss": 0.2064,
      "reward": 0.6559590697288513,
      "reward_std": 0.13350456207990646,
      "reward_std_all": 0.13760241866111755,
      "rewards-max/forensics_sinkhorn_reward": 0.9801705479621887,
      "rewards-max/forensics_strict_edge_reward": 0.8612285256385803,
      "rewards-max/forensics_tiou_reward": 0.9652787446975708,
      "rewards/forensics_sinkhorn_reward": 0.5734929144382477,
      "rewards/forensics_strict_edge_reward": 0.29496487975120544,
      "rewards/forensics_tiou_reward": 0.39548012614250183,
      "step": 780
    },
    {
      "completion_length": 43.984375,
      "epoch": 2.077230359520639,
      "grad_norm": 37.10206985473633,
      "kl": 4.515625,
      "learning_rate": 4.8e-07,
      "loss": 0.3641,
      "reward": 0.6510493755340576,
      "reward_std": 0.11594703048467636,
      "reward_std_all": 0.13221780955791473,
      "rewards-max/forensics_sinkhorn_reward": 0.9869134426116943,
      "rewards-max/forensics_strict_edge_reward": 0.9075756669044495,
      "rewards-max/forensics_tiou_reward": 0.9918159246444702,
      "rewards/forensics_sinkhorn_reward": 0.5869451463222504,
      "rewards/forensics_strict_edge_reward": 0.26132745295763016,
      "rewards/forensics_tiou_reward": 0.4056216776371002,
      "step": 781
    },
    {
      "completion_length": 47.484375,
      "epoch": 2.0798934753661786,
      "grad_norm": 29.771167755126953,
      "kl": 4.546875,
      "learning_rate": 4.793333333333333e-07,
      "loss": 0.3453,
      "reward": 0.6763312518596649,
      "reward_std": 0.11967761069536209,
      "reward_std_all": 0.12617524713277817,
      "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": 0.5838534235954285,
      "rewards/forensics_strict_edge_reward": 0.392843633890152,
      "rewards/forensics_tiou_reward": 0.4365409016609192,
      "step": 782
    },
    {
      "completion_length": 35.390625,
      "epoch": 2.0825565912117177,
      "grad_norm": 20.387802124023438,
      "kl": 4.2421875,
      "learning_rate": 4.786666666666667e-07,
      "loss": 0.3075,
      "reward": 0.6999676823616028,
      "reward_std": 0.09372566640377045,
      "reward_std_all": 0.10078145563602448,
      "rewards-max/forensics_sinkhorn_reward": 0.9861380159854889,
      "rewards-max/forensics_strict_edge_reward": 0.9149472117424011,
      "rewards-max/forensics_tiou_reward": 0.9887892007827759,
      "rewards/forensics_sinkhorn_reward": 0.5795628130435944,
      "rewards/forensics_strict_edge_reward": 0.29778099060058594,
      "rewards/forensics_tiou_reward": 0.3875119686126709,
      "step": 783
    },
    {
      "completion_length": 41.34375,
      "epoch": 2.085219707057257,
      "grad_norm": 27.4530086517334,
      "kl": 3.3515625,
      "learning_rate": 4.779999999999999e-07,
      "loss": 0.2409,
      "reward": 0.673655778169632,
      "reward_std": 0.11898646131157875,
      "reward_std_all": 0.11962024867534637,
      "rewards-max/forensics_sinkhorn_reward": 0.9566881954669952,
      "rewards-max/forensics_strict_edge_reward": 0.9003686904907227,
      "rewards-max/forensics_tiou_reward": 0.929720401763916,
      "rewards/forensics_sinkhorn_reward": 0.599330872297287,
      "rewards/forensics_strict_edge_reward": 0.2526276409626007,
      "rewards/forensics_tiou_reward": 0.3999253809452057,
      "step": 784
    },
    {
      "completion_length": 47.0625,
      "epoch": 2.087882822902796,
      "grad_norm": 69.70988464355469,
      "kl": 7.109375,
      "learning_rate": 4.773333333333333e-07,
      "loss": 0.3375,
      "reward": 0.6757962703704834,
      "reward_std": 0.14020534604787827,
      "reward_std_all": 0.1371736116707325,
      "rewards-max/forensics_sinkhorn_reward": 0.9845986366271973,
      "rewards-max/forensics_strict_edge_reward": 0.9357379674911499,
      "rewards-max/forensics_tiou_reward": 0.9861028492450714,
      "rewards/forensics_sinkhorn_reward": 0.5686385035514832,
      "rewards/forensics_strict_edge_reward": 0.2420080006122589,
      "rewards/forensics_tiou_reward": 0.31831254065036774,
      "step": 785
    },
    {
      "completion_length": 53.28125,
      "epoch": 2.0905459387483356,
      "grad_norm": 26.856021881103516,
      "kl": 4.6875,
      "learning_rate": 4.7666666666666667e-07,
      "loss": 0.4026,
      "reward": 0.5894337594509125,
      "reward_std": 0.18585053086280823,
      "reward_std_all": 0.2048621103167534,
      "rewards-max/forensics_sinkhorn_reward": 0.9222550094127655,
      "rewards-max/forensics_strict_edge_reward": 0.8042719960212708,
      "rewards-max/forensics_tiou_reward": 0.8527688980102539,
      "rewards/forensics_sinkhorn_reward": 0.4555472582578659,
      "rewards/forensics_strict_edge_reward": 0.240117609500885,
      "rewards/forensics_tiou_reward": 0.26273781806230545,
      "step": 786
    },
    {
      "completion_length": 34.5,
      "epoch": 2.0932090545938746,
      "grad_norm": 26.734554290771484,
      "kl": 3.9921875,
      "learning_rate": 4.76e-07,
      "loss": 0.4509,
      "reward": 0.6471266150474548,
      "reward_std": 0.15403354167938232,
      "reward_std_all": 0.15977032482624054,
      "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": 0.6048359274864197,
      "rewards/forensics_strict_edge_reward": 0.3424806445837021,
      "rewards/forensics_tiou_reward": 0.44715696573257446,
      "step": 787
    },
    {
      "completion_length": 29.0,
      "epoch": 2.095872170439414,
      "grad_norm": 28.739198684692383,
      "kl": 2.3515625,
      "learning_rate": 4.7533333333333333e-07,
      "loss": 0.2102,
      "reward": 0.6972728669643402,
      "reward_std": 0.11860412359237671,
      "reward_std_all": 0.11890218406915665,
      "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": 0.673284500837326,
      "rewards/forensics_strict_edge_reward": 0.38385556638240814,
      "rewards/forensics_tiou_reward": 0.49486398696899414,
      "step": 788
    },
    {
      "completion_length": 41.078125,
      "epoch": 2.0985352862849536,
      "grad_norm": 30.842819213867188,
      "kl": 3.1015625,
      "learning_rate": 4.746666666666667e-07,
      "loss": 0.2858,
      "reward": 0.6740429103374481,
      "reward_std": 0.13059396669268608,
      "reward_std_all": 0.13835084438323975,
      "rewards-max/forensics_sinkhorn_reward": 0.9599169194698334,
      "rewards-max/forensics_strict_edge_reward": 0.8854433596134186,
      "rewards-max/forensics_tiou_reward": 0.9310558438301086,
      "rewards/forensics_sinkhorn_reward": 0.5762042701244354,
      "rewards/forensics_strict_edge_reward": 0.32556013762950897,
      "rewards/forensics_tiou_reward": 0.42722389101982117,
      "step": 789
    },
    {
      "completion_length": 32.703125,
      "epoch": 2.1011984021304926,
      "grad_norm": 23.736989974975586,
      "kl": 2.56640625,
      "learning_rate": 4.7399999999999993e-07,
      "loss": 0.232,
      "reward": 0.6783686876296997,
      "reward_std": 0.0981806218624115,
      "reward_std_all": 0.11181960068643093,
      "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": 0.6942228078842163,
      "rewards/forensics_strict_edge_reward": 0.4202595576643944,
      "rewards/forensics_tiou_reward": 0.5634882152080536,
      "step": 790
    },
    {
      "completion_length": 54.28125,
      "epoch": 2.103861517976032,
      "grad_norm": 33.70967483520508,
      "kl": 6.1875,
      "learning_rate": 4.733333333333333e-07,
      "loss": 0.4792,
      "reward": 0.6183059513568878,
      "reward_std": 0.114097960293293,
      "reward_std_all": 0.146245576441288,
      "rewards-max/forensics_sinkhorn_reward": 0.9557605385780334,
      "rewards-max/forensics_strict_edge_reward": 0.7578859627246857,
      "rewards-max/forensics_tiou_reward": 0.9246740639209747,
      "rewards/forensics_sinkhorn_reward": 0.5163745135068893,
      "rewards/forensics_strict_edge_reward": 0.1873169057071209,
      "rewards/forensics_tiou_reward": 0.2779709994792938,
      "step": 791
    },
    {
      "completion_length": 37.296875,
      "epoch": 2.106524633821571,
      "grad_norm": 28.522518157958984,
      "kl": 3.5390625,
      "learning_rate": 4.7266666666666664e-07,
      "loss": 0.2747,
      "reward": 0.6587496697902679,
      "reward_std": 0.17756882309913635,
      "reward_std_all": 0.1684650257229805,
      "rewards-max/forensics_sinkhorn_reward": 0.9885195195674896,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9981445372104645,
      "rewards/forensics_sinkhorn_reward": 0.5839757919311523,
      "rewards/forensics_strict_edge_reward": 0.31608644127845764,
      "rewards/forensics_tiou_reward": 0.38854311406612396,
      "step": 792
    },
    {
      "completion_length": 29.71875,
      "epoch": 2.1091877496671105,
      "grad_norm": 23.93050193786621,
      "kl": 2.6484375,
      "learning_rate": 4.7199999999999994e-07,
      "loss": 0.3084,
      "reward": 0.6594774723052979,
      "reward_std": 0.12594585120677948,
      "reward_std_all": 0.13943195343017578,
      "rewards-max/forensics_sinkhorn_reward": 0.987354576587677,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9935540854930878,
      "rewards/forensics_sinkhorn_reward": 0.6007187962532043,
      "rewards/forensics_strict_edge_reward": 0.2962906062602997,
      "rewards/forensics_tiou_reward": 0.4201660454273224,
      "step": 793
    },
    {
      "completion_length": 29.28125,
      "epoch": 2.11185086551265,
      "grad_norm": 28.608983993530273,
      "kl": 2.6953125,
      "learning_rate": 4.713333333333333e-07,
      "loss": 0.168,
      "reward": 0.6808654963970184,
      "reward_std": 0.11122480407357216,
      "reward_std_all": 0.12073159962892532,
      "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": 0.6077000796794891,
      "rewards/forensics_strict_edge_reward": 0.33971312642097473,
      "rewards/forensics_tiou_reward": 0.4218256324529648,
      "step": 794
    },
    {
      "completion_length": 24.515625,
      "epoch": 2.114513981358189,
      "grad_norm": 22.8480281829834,
      "kl": 2.890625,
      "learning_rate": 4.7066666666666665e-07,
      "loss": 0.2535,
      "reward": 0.6964963376522064,
      "reward_std": 0.10668999701738358,
      "reward_std_all": 0.10757662355899811,
      "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": 0.6622953116893768,
      "rewards/forensics_strict_edge_reward": 0.3700464814901352,
      "rewards/forensics_tiou_reward": 0.4455634653568268,
      "step": 795
    },
    {
      "completion_length": 23.9375,
      "epoch": 2.1171770972037285,
      "grad_norm": 26.518754959106445,
      "kl": 1.7421875,
      "learning_rate": 4.6999999999999995e-07,
      "loss": 0.2115,
      "reward": 0.6749805510044098,
      "reward_std": 0.0888555683195591,
      "reward_std_all": 0.107465710490942,
      "rewards-max/forensics_sinkhorn_reward": 0.9719913005828857,
      "rewards-max/forensics_strict_edge_reward": 0.9265981018543243,
      "rewards-max/forensics_tiou_reward": 0.9596181213855743,
      "rewards/forensics_sinkhorn_reward": 0.5793484449386597,
      "rewards/forensics_strict_edge_reward": 0.2544810473918915,
      "rewards/forensics_tiou_reward": 0.3436786085367203,
      "step": 796
    },
    {
      "completion_length": 18.65625,
      "epoch": 2.1198402130492675,
      "grad_norm": 20.28908348083496,
      "kl": 1.48046875,
      "learning_rate": 4.693333333333333e-07,
      "loss": -0.0544,
      "reward": 0.7167279422283173,
      "reward_std": 0.07055711187422276,
      "reward_std_all": 0.08038321137428284,
      "rewards-max/forensics_sinkhorn_reward": 0.9887278974056244,
      "rewards-max/forensics_strict_edge_reward": 0.9890114367008209,
      "rewards-max/forensics_tiou_reward": 0.9989653527736664,
      "rewards/forensics_sinkhorn_reward": 0.6662785112857819,
      "rewards/forensics_strict_edge_reward": 0.32149720191955566,
      "rewards/forensics_tiou_reward": 0.41038815677165985,
      "step": 797
    },
    {
      "completion_length": 37.171875,
      "epoch": 2.122503328894807,
      "grad_norm": 25.4979190826416,
      "kl": 4.515625,
      "learning_rate": 4.6866666666666665e-07,
      "loss": 0.2999,
      "reward": 0.6800713241100311,
      "reward_std": 0.09537180885672569,
      "reward_std_all": 0.10903766751289368,
      "rewards-max/forensics_sinkhorn_reward": 0.9883276224136353,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9973877966403961,
      "rewards/forensics_sinkhorn_reward": 0.591971218585968,
      "rewards/forensics_strict_edge_reward": 0.2808019518852234,
      "rewards/forensics_tiou_reward": 0.3738923817873001,
      "step": 798
    },
    {
      "completion_length": 32.109375,
      "epoch": 2.125166444740346,
      "grad_norm": 20.822071075439453,
      "kl": 3.5078125,
      "learning_rate": 4.68e-07,
      "loss": 0.2267,
      "reward": 0.6343954205513,
      "reward_std": 0.13811054080724716,
      "reward_std_all": 0.15068462491035461,
      "rewards-max/forensics_sinkhorn_reward": 0.9578686058521271,
      "rewards-max/forensics_strict_edge_reward": 0.7634172439575195,
      "rewards-max/forensics_tiou_reward": 0.9310773015022278,
      "rewards/forensics_sinkhorn_reward": 0.5232577174901962,
      "rewards/forensics_strict_edge_reward": 0.22673841565847397,
      "rewards/forensics_tiou_reward": 0.3157954663038254,
      "step": 799
    },
    {
      "completion_length": 25.71875,
      "epoch": 2.1278295605858855,
      "grad_norm": 31.16747283935547,
      "kl": 3.109375,
      "learning_rate": 4.673333333333333e-07,
      "loss": 0.1309,
      "reward": 0.6443753242492676,
      "reward_std": 0.12258681654930115,
      "reward_std_all": 0.13963574171066284,
      "rewards-max/forensics_sinkhorn_reward": 0.9875406622886658,
      "rewards-max/forensics_strict_edge_reward": 0.9574736058712006,
      "rewards-max/forensics_tiou_reward": 0.9942857027053833,
      "rewards/forensics_sinkhorn_reward": 0.5358629524707794,
      "rewards/forensics_strict_edge_reward": 0.2117958962917328,
      "rewards/forensics_tiou_reward": 0.28712207078933716,
      "step": 800
    },
    {
      "completion_length": 26.109375,
      "epoch": 2.130492676431425,
      "grad_norm": 29.69499397277832,
      "kl": 2.0390625,
      "learning_rate": 4.6666666666666666e-07,
      "loss": 0.1896,
      "reward": 0.5825720429420471,
      "reward_std": 0.16985304653644562,
      "reward_std_all": 0.19291133433580399,
      "rewards-max/forensics_sinkhorn_reward": 0.922171026468277,
      "rewards-max/forensics_strict_edge_reward": 0.7415142953395844,
      "rewards-max/forensics_tiou_reward": 0.8552540838718414,
      "rewards/forensics_sinkhorn_reward": 0.45681919157505035,
      "rewards/forensics_strict_edge_reward": 0.10734353214502335,
      "rewards/forensics_tiou_reward": 0.21758755296468735,
      "step": 801
    },
    {
      "completion_length": 30.03125,
      "epoch": 2.133155792276964,
      "grad_norm": 31.005901336669922,
      "kl": 2.796875,
      "learning_rate": 4.66e-07,
      "loss": 0.1208,
      "reward": 0.585804283618927,
      "reward_std": 0.13586373254656792,
      "reward_std_all": 0.1680128499865532,
      "rewards-max/forensics_sinkhorn_reward": 0.9884027242660522,
      "rewards-max/forensics_strict_edge_reward": 0.9364311099052429,
      "rewards-max/forensics_tiou_reward": 0.9976828098297119,
      "rewards/forensics_sinkhorn_reward": 0.43733730912208557,
      "rewards/forensics_strict_edge_reward": 0.17692594975233078,
      "rewards/forensics_tiou_reward": 0.2635592818260193,
      "step": 802
    },
    {
      "completion_length": 25.3125,
      "epoch": 2.1358189081225034,
      "grad_norm": 31.191377639770508,
      "kl": 2.6796875,
      "learning_rate": 4.653333333333333e-07,
      "loss": 0.1636,
      "reward": 0.652725100517273,
      "reward_std": 0.16302917152643204,
      "reward_std_all": 0.15662074834108353,
      "rewards-max/forensics_sinkhorn_reward": 0.987350583076477,
      "rewards-max/forensics_strict_edge_reward": 0.9364311099052429,
      "rewards-max/forensics_tiou_reward": 0.9935542345046997,
      "rewards/forensics_sinkhorn_reward": 0.5315213352441788,
      "rewards/forensics_strict_edge_reward": 0.30402354896068573,
      "rewards/forensics_tiou_reward": 0.3714078515768051,
      "step": 803
    },
    {
      "completion_length": 45.46875,
      "epoch": 2.1384820239680424,
      "grad_norm": 26.141429901123047,
      "kl": 4.4609375,
      "learning_rate": 4.6466666666666667e-07,
      "loss": 0.3619,
      "reward": 0.5771970748901367,
      "reward_std": 0.14582327008247375,
      "reward_std_all": 0.18462568521499634,
      "rewards-max/forensics_sinkhorn_reward": 0.9876213073730469,
      "rewards-max/forensics_strict_edge_reward": 0.946017861366272,
      "rewards-max/forensics_tiou_reward": 0.9946033954620361,
      "rewards/forensics_sinkhorn_reward": 0.5400440990924835,
      "rewards/forensics_strict_edge_reward": 0.26922277361154556,
      "rewards/forensics_tiou_reward": 0.3649958521127701,
      "step": 804
    },
    {
      "completion_length": 32.453125,
      "epoch": 2.141145139813582,
      "grad_norm": 23.65264320373535,
      "kl": 2.671875,
      "learning_rate": 4.64e-07,
      "loss": 0.2673,
      "reward": 0.643202543258667,
      "reward_std": 0.15598655864596367,
      "reward_std_all": 0.16079463809728622,
      "rewards-max/forensics_sinkhorn_reward": 0.9884536862373352,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9978849291801453,
      "rewards/forensics_sinkhorn_reward": 0.5601958632469177,
      "rewards/forensics_strict_edge_reward": 0.23771638423204422,
      "rewards/forensics_tiou_reward": 0.34931713342666626,
      "step": 805
    },
    {
      "completion_length": 32.609375,
      "epoch": 2.1438082556591214,
      "grad_norm": 29.27274513244629,
      "kl": 3.6015625,
      "learning_rate": 4.633333333333333e-07,
      "loss": 0.2847,
      "reward": 0.6971210539340973,
      "reward_std": 0.12327733635902405,
      "reward_std_all": 0.11959455162286758,
      "rewards-max/forensics_sinkhorn_reward": 0.9823708236217499,
      "rewards-max/forensics_strict_edge_reward": 0.8719756007194519,
      "rewards-max/forensics_tiou_reward": 0.9742265343666077,
      "rewards/forensics_sinkhorn_reward": 0.5764843821525574,
      "rewards/forensics_strict_edge_reward": 0.2826872169971466,
      "rewards/forensics_tiou_reward": 0.3791971206665039,
      "step": 806
    },
    {
      "completion_length": 42.859375,
      "epoch": 2.1464713715046604,
      "grad_norm": 26.720590591430664,
      "kl": 3.796875,
      "learning_rate": 4.6266666666666663e-07,
      "loss": 0.2361,
      "reward": 0.5924094915390015,
      "reward_std": 0.2185024544596672,
      "reward_std_all": 0.21678271889686584,
      "rewards-max/forensics_sinkhorn_reward": 0.9862692058086395,
      "rewards-max/forensics_strict_edge_reward": 0.8983768224716187,
      "rewards-max/forensics_tiou_reward": 0.9893202185630798,
      "rewards/forensics_sinkhorn_reward": 0.4828956127166748,
      "rewards/forensics_strict_edge_reward": 0.20342765003442764,
      "rewards/forensics_tiou_reward": 0.2818191200494766,
      "step": 807
    },
    {
      "completion_length": 30.921875,
      "epoch": 2.1491344873502,
      "grad_norm": 29.974716186523438,
      "kl": 3.5546875,
      "learning_rate": 4.62e-07,
      "loss": 0.4428,
      "reward": 0.7084571719169617,
      "reward_std": 0.11377416923642159,
      "reward_std_all": 0.1071232482790947,
      "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": 0.6318375468254089,
      "rewards/forensics_strict_edge_reward": 0.35391151905059814,
      "rewards/forensics_tiou_reward": 0.41670067608356476,
      "step": 808
    },
    {
      "completion_length": 28.359375,
      "epoch": 2.151797603195739,
      "grad_norm": 32.86748504638672,
      "kl": 2.71875,
      "learning_rate": 4.613333333333333e-07,
      "loss": 0.4085,
      "reward": 0.6678414642810822,
      "reward_std": 0.1475573666393757,
      "reward_std_all": 0.14131607115268707,
      "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": 0.5725373327732086,
      "rewards/forensics_strict_edge_reward": 0.3244134485721588,
      "rewards/forensics_tiou_reward": 0.38823190331459045,
      "step": 809
    },
    {
      "completion_length": 36.75,
      "epoch": 2.1544607190412783,
      "grad_norm": 26.40170669555664,
      "kl": 3.125,
      "learning_rate": 4.6066666666666664e-07,
      "loss": 0.2545,
      "reward": 0.6433923244476318,
      "reward_std": 0.18758125603199005,
      "reward_std_all": 0.1793155074119568,
      "rewards-max/forensics_sinkhorn_reward": 0.9877614378929138,
      "rewards-max/forensics_strict_edge_reward": 0.8983768224716187,
      "rewards-max/forensics_tiou_reward": 0.9951565861701965,
      "rewards/forensics_sinkhorn_reward": 0.5489312708377838,
      "rewards/forensics_strict_edge_reward": 0.19423818588256836,
      "rewards/forensics_tiou_reward": 0.31022560596466064,
      "step": 810
    },
    {
      "completion_length": 40.953125,
      "epoch": 2.157123834886818,
      "grad_norm": 26.320661544799805,
      "kl": 3.3515625,
      "learning_rate": 4.6e-07,
      "loss": 0.3136,
      "reward": 0.6318120658397675,
      "reward_std": 0.11799485981464386,
      "reward_std_all": 0.14390283823013306,
      "rewards-max/forensics_sinkhorn_reward": 0.9702434241771698,
      "rewards-max/forensics_strict_edge_reward": 0.8472771048545837,
      "rewards-max/forensics_tiou_reward": 0.957728385925293,
      "rewards/forensics_sinkhorn_reward": 0.5224010646343231,
      "rewards/forensics_strict_edge_reward": 0.18149272724986076,
      "rewards/forensics_tiou_reward": 0.29805727303028107,
      "step": 811
    },
    {
      "completion_length": 36.40625,
      "epoch": 2.159786950732357,
      "grad_norm": 31.25617027282715,
      "kl": 2.84375,
      "learning_rate": 4.593333333333333e-07,
      "loss": 0.4468,
      "reward": 0.6797612309455872,
      "reward_std": 0.16151318699121475,
      "reward_std_all": 0.1521337553858757,
      "rewards-max/forensics_sinkhorn_reward": 0.9855700731277466,
      "rewards-max/forensics_strict_edge_reward": 0.9062341153621674,
      "rewards-max/forensics_tiou_reward": 0.9865866899490356,
      "rewards/forensics_sinkhorn_reward": 0.5970145463943481,
      "rewards/forensics_strict_edge_reward": 0.3141673058271408,
      "rewards/forensics_tiou_reward": 0.46650031208992004,
      "step": 812
    },
    {
      "completion_length": 35.734375,
      "epoch": 2.1624500665778963,
      "grad_norm": 29.511962890625,
      "kl": 3.1328125,
      "learning_rate": 4.5866666666666664e-07,
      "loss": 0.2213,
      "reward": 0.5435240864753723,
      "reward_std": 0.19173654913902283,
      "reward_std_all": 0.21902623027563095,
      "rewards-max/forensics_sinkhorn_reward": 0.946214109659195,
      "rewards-max/forensics_strict_edge_reward": 0.8472771048545837,
      "rewards-max/forensics_tiou_reward": 0.9096941649913788,
      "rewards/forensics_sinkhorn_reward": 0.4013599157333374,
      "rewards/forensics_strict_edge_reward": 0.15878678113222122,
      "rewards/forensics_tiou_reward": 0.21035588532686234,
      "step": 813
    },
    {
      "completion_length": 21.1875,
      "epoch": 2.1651131824234353,
      "grad_norm": 22.188526153564453,
      "kl": 1.71484375,
      "learning_rate": 4.58e-07,
      "loss": 0.2358,
      "reward": 0.7182566523551941,
      "reward_std": 0.10534017160534859,
      "reward_std_all": 0.10003810375928879,
      "rewards-max/forensics_sinkhorn_reward": 0.988227903842926,
      "rewards-max/forensics_strict_edge_reward": 0.947419673204422,
      "rewards-max/forensics_tiou_reward": 0.996998518705368,
      "rewards/forensics_sinkhorn_reward": 0.6766448020935059,
      "rewards/forensics_strict_edge_reward": 0.38710784912109375,
      "rewards/forensics_tiou_reward": 0.5213191509246826,
      "step": 814
    },
    {
      "completion_length": 30.8125,
      "epoch": 2.1677762982689748,
      "grad_norm": 22.556453704833984,
      "kl": 2.4453125,
      "learning_rate": 4.573333333333333e-07,
      "loss": 0.3509,
      "reward": 0.6491260230541229,
      "reward_std": 0.1673537790775299,
      "reward_std_all": 0.16402390599250793,
      "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": 0.5515523850917816,
      "rewards/forensics_strict_edge_reward": 0.27210432291030884,
      "rewards/forensics_tiou_reward": 0.3412320613861084,
      "step": 815
    },
    {
      "completion_length": 31.03125,
      "epoch": 2.170439414114514,
      "grad_norm": 21.65371322631836,
      "kl": 2.9375,
      "learning_rate": 4.5666666666666665e-07,
      "loss": 0.2442,
      "reward": 0.6127443313598633,
      "reward_std": 0.1181599572300911,
      "reward_std_all": 0.14763160794973373,
      "rewards-max/forensics_sinkhorn_reward": 0.9453952312469482,
      "rewards-max/forensics_strict_edge_reward": 0.7988545596599579,
      "rewards-max/forensics_tiou_reward": 0.9005192816257477,
      "rewards/forensics_sinkhorn_reward": 0.4563642591238022,
      "rewards/forensics_strict_edge_reward": 0.16094540059566498,
      "rewards/forensics_tiou_reward": 0.22052615880966187,
      "step": 816
    },
    {
      "completion_length": 20.1875,
      "epoch": 2.1731025299600533,
      "grad_norm": 30.121448516845703,
      "kl": 1.77734375,
      "learning_rate": 4.56e-07,
      "loss": 0.205,
      "reward": 0.636335551738739,
      "reward_std": 0.17882947623729706,
      "reward_std_all": 0.17666317522525787,
      "rewards-max/forensics_sinkhorn_reward": 0.9886386692523956,
      "rewards-max/forensics_strict_edge_reward": 0.9890114367008209,
      "rewards-max/forensics_tiou_reward": 0.998613029718399,
      "rewards/forensics_sinkhorn_reward": 0.5496874004602432,
      "rewards/forensics_strict_edge_reward": 0.30350491404533386,
      "rewards/forensics_tiou_reward": 0.3832218647003174,
      "step": 817
    },
    {
      "completion_length": 21.96875,
      "epoch": 2.1757656458055927,
      "grad_norm": 33.081947326660156,
      "kl": 2.1875,
      "learning_rate": 4.553333333333333e-07,
      "loss": 0.1821,
      "reward": 0.6919282376766205,
      "reward_std": 0.11445295810699463,
      "reward_std_all": 0.1133425384759903,
      "rewards-max/forensics_sinkhorn_reward": 0.8886857628822327,
      "rewards-max/forensics_strict_edge_reward": 0.6698408722877502,
      "rewards-max/forensics_tiou_reward": 0.7821586430072784,
      "rewards/forensics_sinkhorn_reward": 0.5367883443832397,
      "rewards/forensics_strict_edge_reward": 0.23849420994520187,
      "rewards/forensics_tiou_reward": 0.3319675624370575,
      "step": 818
    },
    {
      "completion_length": 23.671875,
      "epoch": 2.1784287616511318,
      "grad_norm": 30.587324142456055,
      "kl": 2.3046875,
      "learning_rate": 4.5466666666666666e-07,
      "loss": 0.1071,
      "reward": 0.6286105811595917,
      "reward_std": 0.1616961695253849,
      "reward_std_all": 0.17034174501895905,
      "rewards-max/forensics_sinkhorn_reward": 0.987038642168045,
      "rewards-max/forensics_strict_edge_reward": 0.9574736058712006,
      "rewards-max/forensics_tiou_reward": 0.9923076927661896,
      "rewards/forensics_sinkhorn_reward": 0.5739792585372925,
      "rewards/forensics_strict_edge_reward": 0.25838105380535126,
      "rewards/forensics_tiou_reward": 0.348592147231102,
      "step": 819
    },
    {
      "completion_length": 36.0,
      "epoch": 2.181091877496671,
      "grad_norm": 25.66701316833496,
      "kl": 2.953125,
      "learning_rate": 4.54e-07,
      "loss": 0.3689,
      "reward": 0.618284285068512,
      "reward_std": 0.16444943100214005,
      "reward_std_all": 0.1764022558927536,
      "rewards-max/forensics_sinkhorn_reward": 0.9883950054645538,
      "rewards-max/forensics_strict_edge_reward": 0.9782643616199493,
      "rewards-max/forensics_tiou_reward": 0.9976553320884705,
      "rewards/forensics_sinkhorn_reward": 0.4873719811439514,
      "rewards/forensics_strict_edge_reward": 0.2024509683251381,
      "rewards/forensics_tiou_reward": 0.2874171584844589,
      "step": 820
    },
    {
      "completion_length": 34.0625,
      "epoch": 2.1837549933422102,
      "grad_norm": 33.434207916259766,
      "kl": 4.046875,
      "learning_rate": 4.5333333333333326e-07,
      "loss": 0.3243,
      "reward": 0.6349903047084808,
      "reward_std": 0.13694645464420319,
      "reward_std_all": 0.146533764898777,
      "rewards-max/forensics_sinkhorn_reward": 0.9863300025463104,
      "rewards-max/forensics_strict_edge_reward": 0.9464850425720215,
      "rewards-max/forensics_tiou_reward": 0.9913018345832825,
      "rewards/forensics_sinkhorn_reward": 0.5310021340847015,
      "rewards/forensics_strict_edge_reward": 0.24637336283922195,
      "rewards/forensics_tiou_reward": 0.34633301198482513,
      "step": 821
    },
    {
      "completion_length": 25.46875,
      "epoch": 2.1864181091877497,
      "grad_norm": 24.80237579345703,
      "kl": 3.203125,
      "learning_rate": 4.526666666666666e-07,
      "loss": 0.2997,
      "reward": 0.6606453657150269,
      "reward_std": 0.12332713603973389,
      "reward_std_all": 0.1349829137325287,
      "rewards-max/forensics_sinkhorn_reward": 0.987397313117981,
      "rewards-max/forensics_strict_edge_reward": 0.9567649364471436,
      "rewards-max/forensics_tiou_reward": 0.9937322735786438,
      "rewards/forensics_sinkhorn_reward": 0.6148218512535095,
      "rewards/forensics_strict_edge_reward": 0.3978302776813507,
      "rewards/forensics_tiou_reward": 0.49169230461120605,
      "step": 822
    },
    {
      "completion_length": 27.390625,
      "epoch": 2.1890812250332887,
      "grad_norm": 29.49575424194336,
      "kl": 4.1875,
      "learning_rate": 4.5199999999999997e-07,
      "loss": 0.2398,
      "reward": 0.6195042133331299,
      "reward_std": 0.1658741980791092,
      "reward_std_all": 0.17023488134145737,
      "rewards-max/forensics_sinkhorn_reward": 0.9383405447006226,
      "rewards-max/forensics_strict_edge_reward": 0.8472771048545837,
      "rewards-max/forensics_tiou_reward": 0.8926426470279694,
      "rewards/forensics_sinkhorn_reward": 0.45320986211299896,
      "rewards/forensics_strict_edge_reward": 0.1861608698964119,
      "rewards/forensics_tiou_reward": 0.24050435423851013,
      "step": 823
    },
    {
      "completion_length": 27.78125,
      "epoch": 2.191744340878828,
      "grad_norm": 23.715608596801758,
      "kl": 2.9921875,
      "learning_rate": 4.5133333333333327e-07,
      "loss": 0.2182,
      "reward": 0.6751125752925873,
      "reward_std": 0.13256075233221054,
      "reward_std_all": 0.1382337361574173,
      "rewards-max/forensics_sinkhorn_reward": 0.9586948454380035,
      "rewards-max/forensics_strict_edge_reward": 0.8854433596134186,
      "rewards-max/forensics_tiou_reward": 0.9299778342247009,
      "rewards/forensics_sinkhorn_reward": 0.5436444580554962,
      "rewards/forensics_strict_edge_reward": 0.22094020247459412,
      "rewards/forensics_tiou_reward": 0.3623044490814209,
      "step": 824
    },
    {
      "completion_length": 28.984375,
      "epoch": 2.1944074567243677,
      "grad_norm": 27.088483810424805,
      "kl": 2.9296875,
      "learning_rate": 4.506666666666666e-07,
      "loss": 0.2192,
      "reward": 0.672929435968399,
      "reward_std": 0.08251015841960907,
      "reward_std_all": 0.10525844618678093,
      "rewards-max/forensics_sinkhorn_reward": 0.9757094979286194,
      "rewards-max/forensics_strict_edge_reward": 0.9252271056175232,
      "rewards-max/forensics_tiou_reward": 0.9677875339984894,
      "rewards/forensics_sinkhorn_reward": 0.5516067892313004,
      "rewards/forensics_strict_edge_reward": 0.2843591719865799,
      "rewards/forensics_tiou_reward": 0.35073070228099823,
      "step": 825
    },
    {
      "completion_length": 22.515625,
      "epoch": 2.1970705725699067,
      "grad_norm": 32.988319396972656,
      "kl": 3.64453125,
      "learning_rate": 4.5e-07,
      "loss": 0.3076,
      "reward": 0.6117990016937256,
      "reward_std": 0.15349466353654861,
      "reward_std_all": 0.17097963392734528,
      "rewards-max/forensics_sinkhorn_reward": 0.9848630726337433,
      "rewards-max/forensics_strict_edge_reward": 0.8503920137882233,
      "rewards-max/forensics_tiou_reward": 0.9838720262050629,
      "rewards/forensics_sinkhorn_reward": 0.5471566170454025,
      "rewards/forensics_strict_edge_reward": 0.26205119490623474,
      "rewards/forensics_tiou_reward": 0.37539011240005493,
      "step": 826
    },
    {
      "completion_length": 26.46875,
      "epoch": 2.199733688415446,
      "grad_norm": 29.039522171020508,
      "kl": 4.5859375,
      "learning_rate": 4.493333333333333e-07,
      "loss": 0.3771,
      "reward": 0.6441154181957245,
      "reward_std": 0.08943160250782967,
      "reward_std_all": 0.12068519368767738,
      "rewards-max/forensics_sinkhorn_reward": 0.9525250792503357,
      "rewards-max/forensics_strict_edge_reward": 0.8283384442329407,
      "rewards-max/forensics_tiou_reward": 0.9200779795646667,
      "rewards/forensics_sinkhorn_reward": 0.5281347930431366,
      "rewards/forensics_strict_edge_reward": 0.20447200536727905,
      "rewards/forensics_tiou_reward": 0.28817518055438995,
      "step": 827
    },
    {
      "completion_length": 19.828125,
      "epoch": 2.202396804260985,
      "grad_norm": 29.80126190185547,
      "kl": 2.625,
      "learning_rate": 4.4866666666666663e-07,
      "loss": 0.1339,
      "reward": 0.6954023241996765,
      "reward_std": 0.1081564910709858,
      "reward_std_all": 0.10722217336297035,
      "rewards-max/forensics_sinkhorn_reward": 0.9879208505153656,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9957861602306366,
      "rewards/forensics_sinkhorn_reward": 0.6478708386421204,
      "rewards/forensics_strict_edge_reward": 0.3470340967178345,
      "rewards/forensics_tiou_reward": 0.44746288657188416,
      "step": 828
    },
    {
      "completion_length": 25.03125,
      "epoch": 2.2050599201065246,
      "grad_norm": 30.233980178833008,
      "kl": 3.734375,
      "learning_rate": 4.48e-07,
      "loss": 0.1392,
      "reward": 0.6623899638652802,
      "reward_std": 0.11638433113694191,
      "reward_std_all": 0.12633153796195984,
      "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": 0.6087925434112549,
      "rewards/forensics_strict_edge_reward": 0.3231462836265564,
      "rewards/forensics_tiou_reward": 0.396513432264328,
      "step": 829
    },
    {
      "completion_length": 23.671875,
      "epoch": 2.207723035952064,
      "grad_norm": 30.423824310302734,
      "kl": 3.2578125,
      "learning_rate": 4.4733333333333334e-07,
      "loss": 0.2462,
      "reward": 0.6688658893108368,
      "reward_std": 0.11758901178836823,
      "reward_std_all": 0.13045461475849152,
      "rewards-max/forensics_sinkhorn_reward": 0.9878454208374023,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9954914748668671,
      "rewards/forensics_sinkhorn_reward": 0.5501430630683899,
      "rewards/forensics_strict_edge_reward": 0.30091601610183716,
      "rewards/forensics_tiou_reward": 0.35352614521980286,
      "step": 830
    },
    {
      "completion_length": 21.546875,
      "epoch": 2.210386151797603,
      "grad_norm": 27.362186431884766,
      "kl": 3.0390625,
      "learning_rate": 4.4666666666666664e-07,
      "loss": 0.2386,
      "reward": 0.6466683149337769,
      "reward_std": 0.189691461622715,
      "reward_std_all": 0.17818595468997955,
      "rewards-max/forensics_sinkhorn_reward": 0.9391380846500397,
      "rewards-max/forensics_strict_edge_reward": 0.8698340952396393,
      "rewards-max/forensics_tiou_reward": 0.8970016539096832,
      "rewards/forensics_sinkhorn_reward": 0.561563566327095,
      "rewards/forensics_strict_edge_reward": 0.29082632064819336,
      "rewards/forensics_tiou_reward": 0.36746522784233093,
      "step": 831
    },
    {
      "completion_length": 40.453125,
      "epoch": 2.2130492676431426,
      "grad_norm": 47.2363395690918,
      "kl": 4.296875,
      "learning_rate": 4.46e-07,
      "loss": 0.2875,
      "reward": 0.6347565948963165,
      "reward_std": 0.13776028901338577,
      "reward_std_all": 0.14405899494886398,
      "rewards-max/forensics_sinkhorn_reward": 0.9554042816162109,
      "rewards-max/forensics_strict_edge_reward": 0.9003686904907227,
      "rewards-max/forensics_tiou_reward": 0.9261157810688019,
      "rewards/forensics_sinkhorn_reward": 0.5222318321466446,
      "rewards/forensics_strict_edge_reward": 0.269869327545166,
      "rewards/forensics_tiou_reward": 0.3370586186647415,
      "step": 832
    },
    {
      "completion_length": 24.015625,
      "epoch": 2.2157123834886816,
      "grad_norm": 40.42284393310547,
      "kl": 3.06640625,
      "learning_rate": 4.4533333333333335e-07,
      "loss": 0.2454,
      "reward": 0.700216144323349,
      "reward_std": 0.09239360690116882,
      "reward_std_all": 0.09949296712875366,
      "rewards-max/forensics_sinkhorn_reward": 0.9882158637046814,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9969493746757507,
      "rewards/forensics_sinkhorn_reward": 0.663086473941803,
      "rewards/forensics_strict_edge_reward": 0.31491969525814056,
      "rewards/forensics_tiou_reward": 0.41486915946006775,
      "step": 833
    },
    {
      "completion_length": 34.78125,
      "epoch": 2.218375499334221,
      "grad_norm": 32.08162307739258,
      "kl": 3.4453125,
      "learning_rate": 4.4466666666666665e-07,
      "loss": 0.2144,
      "reward": 0.698486715555191,
      "reward_std": 0.08404466137290001,
      "reward_std_all": 0.0965236946940422,
      "rewards-max/forensics_sinkhorn_reward": 0.987462043762207,
      "rewards-max/forensics_strict_edge_reward": 0.9890114367008209,
      "rewards-max/forensics_tiou_reward": 0.9939940571784973,
      "rewards/forensics_sinkhorn_reward": 0.6581530570983887,
      "rewards/forensics_strict_edge_reward": 0.41395142674446106,
      "rewards/forensics_tiou_reward": 0.49227994680404663,
      "step": 834
    },
    {
      "completion_length": 17.625,
      "epoch": 2.2210386151797605,
      "grad_norm": 24.748634338378906,
      "kl": 1.75390625,
      "learning_rate": 4.44e-07,
      "loss": 0.1739,
      "reward": 0.5862432420253754,
      "reward_std": 0.13271033018827438,
      "reward_std_all": 0.1720382496714592,
      "rewards-max/forensics_sinkhorn_reward": 0.9727201759815216,
      "rewards-max/forensics_strict_edge_reward": 0.8394034504890442,
      "rewards-max/forensics_tiou_reward": 0.9380603134632111,
      "rewards/forensics_sinkhorn_reward": 0.5029833614826202,
      "rewards/forensics_strict_edge_reward": 0.15738338977098465,
      "rewards/forensics_tiou_reward": 0.24360130727291107,
      "step": 835
    },
    {
      "completion_length": 34.96875,
      "epoch": 2.2237017310252996,
      "grad_norm": 20.86745834350586,
      "kl": 3.9296875,
      "learning_rate": 4.4333333333333336e-07,
      "loss": 0.3157,
      "reward": 0.6629099249839783,
      "reward_std": 0.12427274882793427,
      "reward_std_all": 0.13040436059236526,
      "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": 0.6014955639839172,
      "rewards/forensics_strict_edge_reward": 0.36140431463718414,
      "rewards/forensics_tiou_reward": 0.4534744471311569,
      "step": 836
    },
    {
      "completion_length": 23.5625,
      "epoch": 2.226364846870839,
      "grad_norm": 25.495452880859375,
      "kl": 3.2890625,
      "learning_rate": 4.426666666666666e-07,
      "loss": 0.2266,
      "reward": 0.6789142489433289,
      "reward_std": 0.13629009202122688,
      "reward_std_all": 0.1339399330317974,
      "rewards-max/forensics_sinkhorn_reward": 0.9881978034973145,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9968753159046173,
      "rewards/forensics_sinkhorn_reward": 0.636989027261734,
      "rewards/forensics_strict_edge_reward": 0.3585182726383209,
      "rewards/forensics_tiou_reward": 0.49994614720344543,
      "step": 837
    },
    {
      "completion_length": 33.703125,
      "epoch": 2.229027962716378,
      "grad_norm": 34.23810958862305,
      "kl": 4.2421875,
      "learning_rate": 4.4199999999999996e-07,
      "loss": 0.2691,
      "reward": 0.6234856247901917,
      "reward_std": 0.15167709067463875,
      "reward_std_all": 0.16366977989673615,
      "rewards-max/forensics_sinkhorn_reward": 0.9575884938240051,
      "rewards-max/forensics_strict_edge_reward": 0.9003686904907227,
      "rewards-max/forensics_tiou_reward": 0.9261916279792786,
      "rewards/forensics_sinkhorn_reward": 0.48583923280239105,
      "rewards/forensics_strict_edge_reward": 0.179769366979599,
      "rewards/forensics_tiou_reward": 0.2524828314781189,
      "step": 838
    },
    {
      "completion_length": 19.875,
      "epoch": 2.2316910785619175,
      "grad_norm": 17.388965606689453,
      "kl": 1.76171875,
      "learning_rate": 4.413333333333333e-07,
      "loss": 0.0722,
      "reward": 0.7382660508155823,
      "reward_std": 0.0733090415596962,
      "reward_std_all": 0.07399609312415123,
      "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": 0.7155815958976746,
      "rewards/forensics_strict_edge_reward": 0.3459394574165344,
      "rewards/forensics_tiou_reward": 0.5160791128873825,
      "step": 839
    },
    {
      "completion_length": 36.671875,
      "epoch": 2.2343541944074565,
      "grad_norm": 27.72393035888672,
      "kl": 4.53125,
      "learning_rate": 4.406666666666666e-07,
      "loss": 0.2552,
      "reward": 0.5979745388031006,
      "reward_std": 0.17163436114788055,
      "reward_std_all": 0.191145122051239,
      "rewards-max/forensics_sinkhorn_reward": 0.9853282868862152,
      "rewards-max/forensics_strict_edge_reward": 0.8983768224716187,
      "rewards-max/forensics_tiou_reward": 0.9856572151184082,
      "rewards/forensics_sinkhorn_reward": 0.5158741474151611,
      "rewards/forensics_strict_edge_reward": 0.2123899906873703,
      "rewards/forensics_tiou_reward": 0.28785133361816406,
      "step": 840
    },
    {
      "completion_length": 44.125,
      "epoch": 2.237017310252996,
      "grad_norm": 30.047313690185547,
      "kl": 3.75,
      "learning_rate": 4.3999999999999997e-07,
      "loss": 0.3533,
      "reward": 0.6302254796028137,
      "reward_std": 0.1555010974407196,
      "reward_std_all": 0.16118201613426208,
      "rewards-max/forensics_sinkhorn_reward": 0.9833537638187408,
      "rewards-max/forensics_strict_edge_reward": 0.8472771048545837,
      "rewards-max/forensics_tiou_reward": 0.9780009388923645,
      "rewards/forensics_sinkhorn_reward": 0.5273246765136719,
      "rewards/forensics_strict_edge_reward": 0.2673915922641754,
      "rewards/forensics_tiou_reward": 0.3062259778380394,
      "step": 841
    },
    {
      "completion_length": 29.328125,
      "epoch": 2.2396804260985355,
      "grad_norm": 25.134918212890625,
      "kl": 2.9375,
      "learning_rate": 4.393333333333333e-07,
      "loss": -0.0356,
      "reward": 0.6550101041793823,
      "reward_std": 0.11304398626089096,
      "reward_std_all": 0.13029537349939346,
      "rewards-max/forensics_sinkhorn_reward": 0.9762793481349945,
      "rewards-max/forensics_strict_edge_reward": 0.9375866651535034,
      "rewards-max/forensics_tiou_reward": 0.9705699682235718,
      "rewards/forensics_sinkhorn_reward": 0.5499587655067444,
      "rewards/forensics_strict_edge_reward": 0.20591574162244797,
      "rewards/forensics_tiou_reward": 0.31023647636175156,
      "step": 842
    },
    {
      "completion_length": 35.125,
      "epoch": 2.2423435419440745,
      "grad_norm": 34.94190216064453,
      "kl": 3.7890625,
      "learning_rate": 4.386666666666666e-07,
      "loss": 0.1922,
      "reward": 0.6507952809333801,
      "reward_std": 0.12307050079107285,
      "reward_std_all": 0.13887407630681992,
      "rewards-max/forensics_sinkhorn_reward": 0.9873356521129608,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9934796690940857,
      "rewards/forensics_sinkhorn_reward": 0.5950548946857452,
      "rewards/forensics_strict_edge_reward": 0.2799709886312485,
      "rewards/forensics_tiou_reward": 0.35077185928821564,
      "step": 843
    },
    {
      "completion_length": 43.21875,
      "epoch": 2.245006657789614,
      "grad_norm": 33.23069763183594,
      "kl": 3.5546875,
      "learning_rate": 4.38e-07,
      "loss": 0.2209,
      "reward": 0.6386460065841675,
      "reward_std": 0.16785302758216858,
      "reward_std_all": 0.16376008093357086,
      "rewards-max/forensics_sinkhorn_reward": 0.9557054042816162,
      "rewards-max/forensics_strict_edge_reward": 0.8188318312168121,
      "rewards-max/forensics_tiou_reward": 0.9071450233459473,
      "rewards/forensics_sinkhorn_reward": 0.517352968454361,
      "rewards/forensics_strict_edge_reward": 0.17087342962622643,
      "rewards/forensics_tiou_reward": 0.2560035288333893,
      "step": 844
    },
    {
      "completion_length": 33.625,
      "epoch": 2.247669773635153,
      "grad_norm": 32.80221939086914,
      "kl": 3.8046875,
      "learning_rate": 4.3733333333333333e-07,
      "loss": 0.2667,
      "reward": 0.6473472714424133,
      "reward_std": 0.1323329135775566,
      "reward_std_all": 0.14527801424264908,
      "rewards-max/forensics_sinkhorn_reward": 0.9873856604099274,
      "rewards-max/forensics_strict_edge_reward": 0.9565287232398987,
      "rewards-max/forensics_tiou_reward": 0.9936806559562683,
      "rewards/forensics_sinkhorn_reward": 0.5450745224952698,
      "rewards/forensics_strict_edge_reward": 0.17814268916845322,
      "rewards/forensics_tiou_reward": 0.2679653987288475,
      "step": 845
    },
    {
      "completion_length": 25.921875,
      "epoch": 2.2503328894806924,
      "grad_norm": 64.86639404296875,
      "kl": 3.2265625,
      "learning_rate": 4.3666666666666663e-07,
      "loss": 0.2891,
      "reward": 0.6310746073722839,
      "reward_std": 0.1394670307636261,
      "reward_std_all": 0.1570667177438736,
      "rewards-max/forensics_sinkhorn_reward": 0.9807964265346527,
      "rewards-max/forensics_strict_edge_reward": 0.8582656681537628,
      "rewards-max/forensics_tiou_reward": 0.9811122417449951,
      "rewards/forensics_sinkhorn_reward": 0.5552759766578674,
      "rewards/forensics_strict_edge_reward": 0.2747998759150505,
      "rewards/forensics_tiou_reward": 0.37959355115890503,
      "step": 846
    },
    {
      "completion_length": 28.625,
      "epoch": 2.2529960053262315,
      "grad_norm": 25.16942596435547,
      "kl": 1.890625,
      "learning_rate": 4.36e-07,
      "loss": 0.1174,
      "reward": 0.7090571820735931,
      "reward_std": 0.07664451189339161,
      "reward_std_all": 0.08874732628464699,
      "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": 0.6825566291809082,
      "rewards/forensics_strict_edge_reward": 0.4320950359106064,
      "rewards/forensics_tiou_reward": 0.5032483041286469,
      "step": 847
    },
    {
      "completion_length": 26.609375,
      "epoch": 2.255659121171771,
      "grad_norm": 28.17510986328125,
      "kl": 2.5625,
      "learning_rate": 4.3533333333333334e-07,
      "loss": 0.2311,
      "reward": 0.7105197012424469,
      "reward_std": 0.11353309452533722,
      "reward_std_all": 0.10957608744502068,
      "rewards-max/forensics_sinkhorn_reward": 0.9886146783828735,
      "rewards-max/forensics_strict_edge_reward": 0.9782643616199493,
      "rewards-max/forensics_tiou_reward": 0.9985185265541077,
      "rewards/forensics_sinkhorn_reward": 0.6498711109161377,
      "rewards/forensics_strict_edge_reward": 0.36545464396476746,
      "rewards/forensics_tiou_reward": 0.45719408988952637,
      "step": 848
    },
    {
      "completion_length": 27.21875,
      "epoch": 2.2583222370173104,
      "grad_norm": 26.063016891479492,
      "kl": 3.0234375,
      "learning_rate": 4.3466666666666664e-07,
      "loss": 0.3628,
      "reward": 0.6402622163295746,
      "reward_std": 0.17808210849761963,
      "reward_std_all": 0.17087748646736145,
      "rewards-max/forensics_sinkhorn_reward": 0.9654048681259155,
      "rewards-max/forensics_strict_edge_reward": 0.8957232534885406,
      "rewards-max/forensics_tiou_reward": 0.9464623332023621,
      "rewards/forensics_sinkhorn_reward": 0.5554248094558716,
      "rewards/forensics_strict_edge_reward": 0.24297744035720825,
      "rewards/forensics_tiou_reward": 0.33969825506210327,
      "step": 849
    },
    {
      "completion_length": 17.78125,
      "epoch": 2.2609853528628494,
      "grad_norm": 19.965299606323242,
      "kl": 2.03125,
      "learning_rate": 4.34e-07,
      "loss": 0.1397,
      "reward": 0.6682605147361755,
      "reward_std": 0.14954029768705368,
      "reward_std_all": 0.14531458169221878,
      "rewards-max/forensics_sinkhorn_reward": 0.9879739284515381,
      "rewards-max/forensics_strict_edge_reward": 0.9357379674911499,
      "rewards-max/forensics_tiou_reward": 0.9959945678710938,
      "rewards/forensics_sinkhorn_reward": 0.6191665828227997,
      "rewards/forensics_strict_edge_reward": 0.2711751013994217,
      "rewards/forensics_tiou_reward": 0.41777199506759644,
      "step": 850
    },
    {
      "completion_length": 22.21875,
      "epoch": 2.263648468708389,
      "grad_norm": 23.775379180908203,
      "kl": 1.53515625,
      "learning_rate": 4.3333333333333335e-07,
      "loss": 0.2696,
      "reward": 0.7082497477531433,
      "reward_std": 0.11880292743444443,
      "reward_std_all": 0.11086325347423553,
      "rewards-max/forensics_sinkhorn_reward": 0.9884355962276459,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9978136420249939,
      "rewards/forensics_sinkhorn_reward": 0.66757732629776,
      "rewards/forensics_strict_edge_reward": 0.3610992133617401,
      "rewards/forensics_tiou_reward": 0.47516879439353943,
      "step": 851
    },
    {
      "completion_length": 17.40625,
      "epoch": 2.266311584553928,
      "grad_norm": 18.035722732543945,
      "kl": 1.76953125,
      "learning_rate": 4.3266666666666665e-07,
      "loss": 0.2878,
      "reward": 0.680350124835968,
      "reward_std": 0.12641474604606628,
      "reward_std_all": 0.1300584301352501,
      "rewards-max/forensics_sinkhorn_reward": 0.988226056098938,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.996988832950592,
      "rewards/forensics_sinkhorn_reward": 0.6788464486598969,
      "rewards/forensics_strict_edge_reward": 0.3904394656419754,
      "rewards/forensics_tiou_reward": 0.5235386490821838,
      "step": 852
    },
    {
      "completion_length": 20.34375,
      "epoch": 2.2689747003994674,
      "grad_norm": 27.181467056274414,
      "kl": 2.3671875,
      "learning_rate": 4.3199999999999995e-07,
      "loss": 0.2569,
      "reward": 0.6676255762577057,
      "reward_std": 0.09352175518870354,
      "reward_std_all": 0.11104657128453255,
      "rewards-max/forensics_sinkhorn_reward": 0.9876960515975952,
      "rewards-max/forensics_strict_edge_reward": 0.9677534997463226,
      "rewards-max/forensics_tiou_reward": 0.9949109852313995,
      "rewards/forensics_sinkhorn_reward": 0.6119755208492279,
      "rewards/forensics_strict_edge_reward": 0.29920540750026703,
      "rewards/forensics_tiou_reward": 0.41106662154197693,
      "step": 853
    },
    {
      "completion_length": 27.171875,
      "epoch": 2.271637816245007,
      "grad_norm": 19.563459396362305,
      "kl": 2.00390625,
      "learning_rate": 4.313333333333333e-07,
      "loss": 0.179,
      "reward": 0.6579760313034058,
      "reward_std": 0.12774638459086418,
      "reward_std_all": 0.14042095467448235,
      "rewards-max/forensics_sinkhorn_reward": 0.9863660931587219,
      "rewards-max/forensics_strict_edge_reward": 0.9565287232398987,
      "rewards-max/forensics_tiou_reward": 0.9896667897701263,
      "rewards/forensics_sinkhorn_reward": 0.5724797546863556,
      "rewards/forensics_strict_edge_reward": 0.2585682272911072,
      "rewards/forensics_tiou_reward": 0.34873613715171814,
      "step": 854
    },
    {
      "completion_length": 28.140625,
      "epoch": 2.274300932090546,
      "grad_norm": 269.23150634765625,
      "kl": 4.890625,
      "learning_rate": 4.306666666666666e-07,
      "loss": 0.2078,
      "reward": 0.6153479218482971,
      "reward_std": 0.13823644816875458,
      "reward_std_all": 0.1596870869398117,
      "rewards-max/forensics_sinkhorn_reward": 0.9753691554069519,
      "rewards-max/forensics_strict_edge_reward": 0.9677534997463226,
      "rewards-max/forensics_tiou_reward": 0.9702972173690796,
      "rewards/forensics_sinkhorn_reward": 0.5350764393806458,
      "rewards/forensics_strict_edge_reward": 0.22154468297958374,
      "rewards/forensics_tiou_reward": 0.2995371073484421,
      "step": 855
    },
    {
      "completion_length": 31.34375,
      "epoch": 2.2769640479360853,
      "grad_norm": 43.0960807800293,
      "kl": 5.5859375,
      "learning_rate": 4.2999999999999996e-07,
      "loss": 0.4341,
      "reward": 0.6274014115333557,
      "reward_std": 0.15480556339025497,
      "reward_std_all": 0.16874820739030838,
      "rewards-max/forensics_sinkhorn_reward": 0.9629502296447754,
      "rewards-max/forensics_strict_edge_reward": 0.9464850425720215,
      "rewards-max/forensics_tiou_reward": 0.9438620507717133,
      "rewards/forensics_sinkhorn_reward": 0.53665791451931,
      "rewards/forensics_strict_edge_reward": 0.20500922203063965,
      "rewards/forensics_tiou_reward": 0.3209277391433716,
      "step": 856
    },
    {
      "completion_length": 31.21875,
      "epoch": 2.2796271637816243,
      "grad_norm": 30.556886672973633,
      "kl": 2.9609375,
      "learning_rate": 4.293333333333333e-07,
      "loss": 0.277,
      "reward": 0.6896815001964569,
      "reward_std": 0.11956531554460526,
      "reward_std_all": 0.11920472234487534,
      "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": 0.6181986927986145,
      "rewards/forensics_strict_edge_reward": 0.32416196912527084,
      "rewards/forensics_tiou_reward": 0.4162246137857437,
      "step": 857
    },
    {
      "completion_length": 17.6875,
      "epoch": 2.282290279627164,
      "grad_norm": 22.438749313354492,
      "kl": 1.79296875,
      "learning_rate": 4.286666666666666e-07,
      "loss": 0.0482,
      "reward": 0.7077706456184387,
      "reward_std": 0.07134695909917355,
      "reward_std_all": 0.08694489300251007,
      "rewards-max/forensics_sinkhorn_reward": 0.9873977303504944,
      "rewards-max/forensics_strict_edge_reward": 0.9464850425720215,
      "rewards-max/forensics_tiou_reward": 0.9937230348587036,
      "rewards/forensics_sinkhorn_reward": 0.6318618357181549,
      "rewards/forensics_strict_edge_reward": 0.2785920798778534,
      "rewards/forensics_tiou_reward": 0.3905397653579712,
      "step": 858
    },
    {
      "completion_length": 22.046875,
      "epoch": 2.2849533954727033,
      "grad_norm": 20.987733840942383,
      "kl": 2.7421875,
      "learning_rate": 4.2799999999999997e-07,
      "loss": 0.1956,
      "reward": 0.68685382604599,
      "reward_std": 0.0920569896697998,
      "reward_std_all": 0.10063251852989197,
      "rewards-max/forensics_sinkhorn_reward": 0.9724439680576324,
      "rewards-max/forensics_strict_edge_reward": 0.8503920137882233,
      "rewards-max/forensics_tiou_reward": 0.9625844359397888,
      "rewards/forensics_sinkhorn_reward": 0.5882272720336914,
      "rewards/forensics_strict_edge_reward": 0.27095410227775574,
      "rewards/forensics_tiou_reward": 0.3704669624567032,
      "step": 859
    },
    {
      "completion_length": 36.8125,
      "epoch": 2.2876165113182423,
      "grad_norm": 24.618999481201172,
      "kl": 3.0390625,
      "learning_rate": 4.273333333333333e-07,
      "loss": 0.2097,
      "reward": 0.66480752825737,
      "reward_std": 0.16604524850845337,
      "reward_std_all": 0.15815885365009308,
      "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": 0.5830102562904358,
      "rewards/forensics_strict_edge_reward": 0.3310246765613556,
      "rewards/forensics_tiou_reward": 0.38349005579948425,
      "step": 860
    },
    {
      "completion_length": 25.03125,
      "epoch": 2.2902796271637818,
      "grad_norm": 21.283803939819336,
      "kl": 2.0859375,
      "learning_rate": 4.266666666666667e-07,
      "loss": 0.272,
      "reward": 0.6103509366512299,
      "reward_std": 0.13710173964500427,
      "reward_std_all": 0.15978015214204788,
      "rewards-max/forensics_sinkhorn_reward": 0.8967372477054596,
      "rewards-max/forensics_strict_edge_reward": 0.6000466495752335,
      "rewards-max/forensics_tiou_reward": 0.8036980628967285,
      "rewards/forensics_sinkhorn_reward": 0.5097799003124237,
      "rewards/forensics_strict_edge_reward": 0.12411263398826122,
      "rewards/forensics_tiou_reward": 0.20918677002191544,
      "step": 861
    },
    {
      "completion_length": 41.0,
      "epoch": 2.2929427430093208,
      "grad_norm": 31.79214859008789,
      "kl": 4.3515625,
      "learning_rate": 4.26e-07,
      "loss": 0.3249,
      "reward": 0.6593041121959686,
      "reward_std": 0.14582610875368118,
      "reward_std_all": 0.14891337975859642,
      "rewards-max/forensics_sinkhorn_reward": 0.9849430322647095,
      "rewards-max/forensics_strict_edge_reward": 0.9062341153621674,
      "rewards-max/forensics_tiou_reward": 0.9841366112232208,
      "rewards/forensics_sinkhorn_reward": 0.5913235247135162,
      "rewards/forensics_strict_edge_reward": 0.27960216999053955,
      "rewards/forensics_tiou_reward": 0.4203207939863205,
      "step": 862
    },
    {
      "completion_length": 27.890625,
      "epoch": 2.2956058588548602,
      "grad_norm": 24.999929428100586,
      "kl": 1.9140625,
      "learning_rate": 4.2533333333333333e-07,
      "loss": 0.1011,
      "reward": 0.6839781403541565,
      "reward_std": 0.08070646226406097,
      "reward_std_all": 0.1021052822470665,
      "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": 0.6672015488147736,
      "rewards/forensics_strict_edge_reward": 0.3798512816429138,
      "rewards/forensics_tiou_reward": 0.4578704535961151,
      "step": 863
    },
    {
      "completion_length": 29.28125,
      "epoch": 2.2982689747003997,
      "grad_norm": 27.070758819580078,
      "kl": 3.171875,
      "learning_rate": 4.246666666666667e-07,
      "loss": 0.1479,
      "reward": 0.6403219103813171,
      "reward_std": 0.18782538920640945,
      "reward_std_all": 0.1803816705942154,
      "rewards-max/forensics_sinkhorn_reward": 0.9873009622097015,
      "rewards-max/forensics_strict_edge_reward": 0.9169811904430389,
      "rewards-max/forensics_tiou_reward": 0.9933567345142365,
      "rewards/forensics_sinkhorn_reward": 0.5507213473320007,
      "rewards/forensics_strict_edge_reward": 0.2677273750305176,
      "rewards/forensics_tiou_reward": 0.33890342712402344,
      "step": 864
    },
    {
      "completion_length": 39.546875,
      "epoch": 2.3009320905459387,
      "grad_norm": 32.6789665222168,
      "kl": 3.1640625,
      "learning_rate": 4.24e-07,
      "loss": 0.2581,
      "reward": 0.6909035444259644,
      "reward_std": 0.11411182582378387,
      "reward_std_all": 0.11443711444735527,
      "rewards-max/forensics_sinkhorn_reward": 0.9884685277938843,
      "rewards-max/forensics_strict_edge_reward": 0.9782643616199493,
      "rewards-max/forensics_tiou_reward": 0.9979445040225983,
      "rewards/forensics_sinkhorn_reward": 0.6369277536869049,
      "rewards/forensics_strict_edge_reward": 0.3366446942090988,
      "rewards/forensics_tiou_reward": 0.4624953866004944,
      "step": 865
    },
    {
      "completion_length": 37.859375,
      "epoch": 2.303595206391478,
      "grad_norm": 40.515953063964844,
      "kl": 4.8359375,
      "learning_rate": 4.2333333333333334e-07,
      "loss": 0.2118,
      "reward": 0.6061131358146667,
      "reward_std": 0.1395835131406784,
      "reward_std_all": 0.16228705644607544,
      "rewards-max/forensics_sinkhorn_reward": 0.9327533543109894,
      "rewards-max/forensics_strict_edge_reward": 0.8134243786334991,
      "rewards-max/forensics_tiou_reward": 0.877223014831543,
      "rewards/forensics_sinkhorn_reward": 0.4651040732860565,
      "rewards/forensics_strict_edge_reward": 0.1845252625644207,
      "rewards/forensics_tiou_reward": 0.22710565477609634,
      "step": 866
    },
    {
      "completion_length": 21.390625,
      "epoch": 2.306258322237017,
      "grad_norm": 22.470033645629883,
      "kl": 2.6015625,
      "learning_rate": 4.226666666666667e-07,
      "loss": 0.1327,
      "reward": 0.6839390397071838,
      "reward_std": 0.064471535384655,
      "reward_std_all": 0.09703749045729637,
      "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": 0.6296457052230835,
      "rewards/forensics_strict_edge_reward": 0.37704411149024963,
      "rewards/forensics_tiou_reward": 0.431628093123436,
      "step": 867
    },
    {
      "completion_length": 35.0,
      "epoch": 2.3089214380825567,
      "grad_norm": 29.14173698425293,
      "kl": 2.97265625,
      "learning_rate": 4.2199999999999994e-07,
      "loss": 0.2193,
      "reward": 0.5993453860282898,
      "reward_std": 0.14138033986091614,
      "reward_std_all": 0.1734602302312851,
      "rewards-max/forensics_sinkhorn_reward": 0.9878576695919037,
      "rewards-max/forensics_strict_edge_reward": 0.9464850425720215,
      "rewards-max/forensics_tiou_reward": 0.9955397546291351,
      "rewards/forensics_sinkhorn_reward": 0.5406084954738617,
      "rewards/forensics_strict_edge_reward": 0.27752798795700073,
      "rewards/forensics_tiou_reward": 0.36494696140289307,
      "step": 868
    },
    {
      "completion_length": 30.515625,
      "epoch": 2.3115845539280957,
      "grad_norm": 25.995651245117188,
      "kl": 2.3828125,
      "learning_rate": 4.213333333333333e-07,
      "loss": 0.349,
      "reward": 0.632398247718811,
      "reward_std": 0.18017824739217758,
      "reward_std_all": 0.17571429163217545,
      "rewards-max/forensics_sinkhorn_reward": 0.987236887216568,
      "rewards-max/forensics_strict_edge_reward": 0.9464850425720215,
      "rewards-max/forensics_tiou_reward": 0.9931053221225739,
      "rewards/forensics_sinkhorn_reward": 0.5533011853694916,
      "rewards/forensics_strict_edge_reward": 0.27856696397066116,
      "rewards/forensics_tiou_reward": 0.3817533552646637,
      "step": 869
    },
    {
      "completion_length": 39.65625,
      "epoch": 2.314247669773635,
      "grad_norm": 31.378467559814453,
      "kl": 2.09765625,
      "learning_rate": 4.2066666666666665e-07,
      "loss": 0.0606,
      "reward": 0.6367171406745911,
      "reward_std": 0.12368538230657578,
      "reward_std_all": 0.13996662199497223,
      "rewards-max/forensics_sinkhorn_reward": 0.9809441566467285,
      "rewards-max/forensics_strict_edge_reward": 0.8612285256385803,
      "rewards-max/forensics_tiou_reward": 0.980350136756897,
      "rewards/forensics_sinkhorn_reward": 0.5650399923324585,
      "rewards/forensics_strict_edge_reward": 0.19034607708454132,
      "rewards/forensics_tiou_reward": 0.3204425871372223,
      "step": 870
    },
    {
      "completion_length": 22.0625,
      "epoch": 2.316910785619174,
      "grad_norm": 38.1732177734375,
      "kl": 3.140625,
      "learning_rate": 4.1999999999999995e-07,
      "loss": 0.2964,
      "reward": 0.6751050353050232,
      "reward_std": 0.12850604206323624,
      "reward_std_all": 0.13520226627588272,
      "rewards-max/forensics_sinkhorn_reward": 0.9885506629943848,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9982669651508331,
      "rewards/forensics_sinkhorn_reward": 0.6449163556098938,
      "rewards/forensics_strict_edge_reward": 0.3350995033979416,
      "rewards/forensics_tiou_reward": 0.45864687860012054,
      "step": 871
    },
    {
      "completion_length": 38.96875,
      "epoch": 2.3195739014647137,
      "grad_norm": 56.80735778808594,
      "kl": 3.375,
      "learning_rate": 4.193333333333333e-07,
      "loss": 0.2018,
      "reward": 0.6103840470314026,
      "reward_std": 0.18113695085048676,
      "reward_std_all": 0.183663472533226,
      "rewards-max/forensics_sinkhorn_reward": 0.986941933631897,
      "rewards-max/forensics_strict_edge_reward": 0.905340164899826,
      "rewards-max/forensics_tiou_reward": 0.9919477701187134,
      "rewards/forensics_sinkhorn_reward": 0.4907679557800293,
      "rewards/forensics_strict_edge_reward": 0.17770208418369293,
      "rewards/forensics_tiou_reward": 0.30068647861480713,
      "step": 872
    },
    {
      "completion_length": 21.0625,
      "epoch": 2.322237017310253,
      "grad_norm": 25.424461364746094,
      "kl": 3.12890625,
      "learning_rate": 4.1866666666666666e-07,
      "loss": 0.2775,
      "reward": 0.6700517535209656,
      "reward_std": 0.16730932891368866,
      "reward_std_all": 0.15489185601472855,
      "rewards-max/forensics_sinkhorn_reward": 0.9882753789424896,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9971828460693359,
      "rewards/forensics_sinkhorn_reward": 0.661500871181488,
      "rewards/forensics_strict_edge_reward": 0.4344271123409271,
      "rewards/forensics_tiou_reward": 0.5484100878238678,
      "step": 873
    },
    {
      "completion_length": 25.171875,
      "epoch": 2.324900133155792,
      "grad_norm": 22.48539161682129,
      "kl": 2.8046875,
      "learning_rate": 4.1799999999999996e-07,
      "loss": 0.2565,
      "reward": 0.645166277885437,
      "reward_std": 0.17170051485300064,
      "reward_std_all": 0.1642768532037735,
      "rewards-max/forensics_sinkhorn_reward": 0.9880655705928802,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9963542520999908,
      "rewards/forensics_sinkhorn_reward": 0.5294876992702484,
      "rewards/forensics_strict_edge_reward": 0.2726801335811615,
      "rewards/forensics_tiou_reward": 0.3259914517402649,
      "step": 874
    },
    {
      "completion_length": 17.140625,
      "epoch": 2.3275632490013316,
      "grad_norm": 17.476490020751953,
      "kl": 2.2421875,
      "learning_rate": 4.173333333333333e-07,
      "loss": 0.052,
      "reward": 0.7207053303718567,
      "reward_std": 0.06302211247384548,
      "reward_std_all": 0.0779777392745018,
      "rewards-max/forensics_sinkhorn_reward": 0.9886032044887543,
      "rewards-max/forensics_strict_edge_reward": 0.9782643616199493,
      "rewards-max/forensics_tiou_reward": 0.9984732866287231,
      "rewards/forensics_sinkhorn_reward": 0.6432828605175018,
      "rewards/forensics_strict_edge_reward": 0.32752394676208496,
      "rewards/forensics_tiou_reward": 0.4016318619251251,
      "step": 875
    },
    {
      "completion_length": 28.90625,
      "epoch": 2.3302263648468706,
      "grad_norm": 25.278247833251953,
      "kl": 3.625,
      "learning_rate": 4.1666666666666667e-07,
      "loss": 0.3658,
      "reward": 0.6395093202590942,
      "reward_std": 0.15233724378049374,
      "reward_std_all": 0.15825740061700344,
      "rewards-max/forensics_sinkhorn_reward": 0.964063286781311,
      "rewards-max/forensics_strict_edge_reward": 0.9093653857707977,
      "rewards-max/forensics_tiou_reward": 0.9447192549705505,
      "rewards/forensics_sinkhorn_reward": 0.5263715386390686,
      "rewards/forensics_strict_edge_reward": 0.30109912157058716,
      "rewards/forensics_tiou_reward": 0.39776381850242615,
      "step": 876
    },
    {
      "completion_length": 18.625,
      "epoch": 2.33288948069241,
      "grad_norm": 25.258010864257812,
      "kl": 2.484375,
      "learning_rate": 4.1599999999999997e-07,
      "loss": 0.1239,
      "reward": 0.6873098015785217,
      "reward_std": 0.08425910770893097,
      "reward_std_all": 0.10402751713991165,
      "rewards-max/forensics_sinkhorn_reward": 0.9885371029376984,
      "rewards-max/forensics_strict_edge_reward": 0.9464850425720215,
      "rewards-max/forensics_tiou_reward": 0.9982134103775024,
      "rewards/forensics_sinkhorn_reward": 0.6657900214195251,
      "rewards/forensics_strict_edge_reward": 0.24662239849567413,
      "rewards/forensics_tiou_reward": 0.4154353141784668,
      "step": 877
    },
    {
      "completion_length": 19.484375,
      "epoch": 2.3355525965379496,
      "grad_norm": 47.23659896850586,
      "kl": 2.56640625,
      "learning_rate": 4.153333333333333e-07,
      "loss": 0.2903,
      "reward": 0.6971471607685089,
      "reward_std": 0.1347133219242096,
      "reward_std_all": 0.12503629177808762,
      "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": 0.6637923419475555,
      "rewards/forensics_strict_edge_reward": 0.40048520267009735,
      "rewards/forensics_tiou_reward": 0.4877847731113434,
      "step": 878
    },
    {
      "completion_length": 26.671875,
      "epoch": 2.3382157123834886,
      "grad_norm": 31.024486541748047,
      "kl": 2.9609375,
      "learning_rate": 4.146666666666667e-07,
      "loss": 0.1124,
      "reward": 0.6633599102497101,
      "reward_std": 0.15481502562761307,
      "reward_std_all": 0.1504214107990265,
      "rewards-max/forensics_sinkhorn_reward": 0.9868031442165375,
      "rewards-max/forensics_strict_edge_reward": 0.946017861366272,
      "rewards-max/forensics_tiou_reward": 0.9913966655731201,
      "rewards/forensics_sinkhorn_reward": 0.5975911617279053,
      "rewards/forensics_strict_edge_reward": 0.2845180258154869,
      "rewards/forensics_tiou_reward": 0.3706626892089844,
      "step": 879
    },
    {
      "completion_length": 26.171875,
      "epoch": 2.340878828229028,
      "grad_norm": 32.85677719116211,
      "kl": 3.1640625,
      "learning_rate": 4.14e-07,
      "loss": 0.1566,
      "reward": 0.6502655744552612,
      "reward_std": 0.11080246418714523,
      "reward_std_all": 0.13225265964865685,
      "rewards-max/forensics_sinkhorn_reward": 0.9712030589580536,
      "rewards-max/forensics_strict_edge_reward": 0.8983768224716187,
      "rewards-max/forensics_tiou_reward": 0.9590673446655273,
      "rewards/forensics_sinkhorn_reward": 0.5514131039381027,
      "rewards/forensics_strict_edge_reward": 0.223140150308609,
      "rewards/forensics_tiou_reward": 0.28671200573444366,
      "step": 880
    },
    {
      "completion_length": 22.984375,
      "epoch": 2.343541944074567,
      "grad_norm": 31.846691131591797,
      "kl": 2.5703125,
      "learning_rate": 4.1333333333333333e-07,
      "loss": 0.1426,
      "reward": 0.6637210249900818,
      "reward_std": 0.15162569284439087,
      "reward_std_all": 0.14967913925647736,
      "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": 0.5489038228988647,
      "rewards/forensics_strict_edge_reward": 0.30638958513736725,
      "rewards/forensics_tiou_reward": 0.37141288816928864,
      "step": 881
    },
    {
      "completion_length": 18.53125,
      "epoch": 2.3462050599201065,
      "grad_norm": 19.843719482421875,
      "kl": 2.23828125,
      "learning_rate": 4.126666666666667e-07,
      "loss": 0.0906,
      "reward": 0.6540535092353821,
      "reward_std": 0.14155787229537964,
      "reward_std_all": 0.14997949451208115,
      "rewards-max/forensics_sinkhorn_reward": 0.9619917869567871,
      "rewards-max/forensics_strict_edge_reward": 0.9149472117424011,
      "rewards-max/forensics_tiou_reward": 0.9400631487369537,
      "rewards/forensics_sinkhorn_reward": 0.5139238983392715,
      "rewards/forensics_strict_edge_reward": 0.22874607145786285,
      "rewards/forensics_tiou_reward": 0.29605288803577423,
      "step": 882
    },
    {
      "completion_length": 18.21875,
      "epoch": 2.348868175765646,
      "grad_norm": 17.8708438873291,
      "kl": 2.31640625,
      "learning_rate": 4.12e-07,
      "loss": 0.1527,
      "reward": 0.6960020065307617,
      "reward_std": 0.09951813705265522,
      "reward_std_all": 0.10655693709850311,
      "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": 0.6627326607704163,
      "rewards/forensics_strict_edge_reward": 0.34284767508506775,
      "rewards/forensics_tiou_reward": 0.44641052186489105,
      "step": 883
    },
    {
      "completion_length": 21.9375,
      "epoch": 2.351531291611185,
      "grad_norm": 21.7729434967041,
      "kl": 1.9375,
      "learning_rate": 4.113333333333333e-07,
      "loss": 0.1563,
      "reward": 0.6724202632904053,
      "reward_std": 0.1065879575908184,
      "reward_std_all": 0.12144432961940765,
      "rewards-max/forensics_sinkhorn_reward": 0.988154798746109,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9967088103294373,
      "rewards/forensics_sinkhorn_reward": 0.5719738304615021,
      "rewards/forensics_strict_edge_reward": 0.24642610549926758,
      "rewards/forensics_tiou_reward": 0.3129560798406601,
      "step": 884
    },
    {
      "completion_length": 20.34375,
      "epoch": 2.3541944074567245,
      "grad_norm": 28.136075973510742,
      "kl": 2.859375,
      "learning_rate": 4.1066666666666664e-07,
      "loss": 0.1939,
      "reward": 0.6781499683856964,
      "reward_std": 0.14686666429042816,
      "reward_std_all": 0.14213210344314575,
      "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": 0.5657176971435547,
      "rewards/forensics_strict_edge_reward": 0.2952980697154999,
      "rewards/forensics_tiou_reward": 0.3829662799835205,
      "step": 885
    },
    {
      "completion_length": 30.203125,
      "epoch": 2.3568575233022635,
      "grad_norm": 28.83791160583496,
      "kl": 3.4609375,
      "learning_rate": 4.0999999999999994e-07,
      "loss": 0.2995,
      "reward": 0.6530297696590424,
      "reward_std": 0.12193511798977852,
      "reward_std_all": 0.13353779166936874,
      "rewards-max/forensics_sinkhorn_reward": 0.9860560297966003,
      "rewards-max/forensics_strict_edge_reward": 0.9565287232398987,
      "rewards-max/forensics_tiou_reward": 0.9884353876113892,
      "rewards/forensics_sinkhorn_reward": 0.5615168213844299,
      "rewards/forensics_strict_edge_reward": 0.2731132060289383,
      "rewards/forensics_tiou_reward": 0.33511456847190857,
      "step": 886
    },
    {
      "completion_length": 21.671875,
      "epoch": 2.359520639147803,
      "grad_norm": 34.33538055419922,
      "kl": 3.5703125,
      "learning_rate": 4.093333333333333e-07,
      "loss": 0.3065,
      "reward": 0.5974854230880737,
      "reward_std": 0.13618813455104828,
      "reward_std_all": 0.16593722999095917,
      "rewards-max/forensics_sinkhorn_reward": 0.9762163460254669,
      "rewards-max/forensics_strict_edge_reward": 0.9169811904430389,
      "rewards-max/forensics_tiou_reward": 0.9721630811691284,
      "rewards/forensics_sinkhorn_reward": 0.509074792265892,
      "rewards/forensics_strict_edge_reward": 0.1899736225605011,
      "rewards/forensics_tiou_reward": 0.27125559747219086,
      "step": 887
    },
    {
      "completion_length": 18.734375,
      "epoch": 2.3621837549933424,
      "grad_norm": 29.00982093811035,
      "kl": 3.41796875,
      "learning_rate": 4.0866666666666665e-07,
      "loss": 0.3131,
      "reward": 0.6892650723457336,
      "reward_std": 0.1162199079990387,
      "reward_std_all": 0.11676089465618134,
      "rewards-max/forensics_sinkhorn_reward": 0.9858444333076477,
      "rewards-max/forensics_strict_edge_reward": 0.9357379674911499,
      "rewards-max/forensics_tiou_reward": 0.9876417219638824,
      "rewards/forensics_sinkhorn_reward": 0.5959826409816742,
      "rewards/forensics_strict_edge_reward": 0.32497939467430115,
      "rewards/forensics_tiou_reward": 0.4232472628355026,
      "step": 888
    },
    {
      "completion_length": 27.96875,
      "epoch": 2.3648468708388815,
      "grad_norm": 28.713167190551758,
      "kl": 3.03125,
      "learning_rate": 4.0799999999999995e-07,
      "loss": 0.1463,
      "reward": 0.6744483709335327,
      "reward_std": 0.13204225897789001,
      "reward_std_all": 0.12927661836147308,
      "rewards-max/forensics_sinkhorn_reward": 0.9879767000675201,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9960085451602936,
      "rewards/forensics_sinkhorn_reward": 0.6117845475673676,
      "rewards/forensics_strict_edge_reward": 0.29056449234485626,
      "rewards/forensics_tiou_reward": 0.4087847173213959,
      "step": 889
    },
    {
      "completion_length": 18.296875,
      "epoch": 2.367509986684421,
      "grad_norm": 19.95808219909668,
      "kl": 2.06640625,
      "learning_rate": 4.073333333333333e-07,
      "loss": 0.2339,
      "reward": 0.6049842536449432,
      "reward_std": 0.1476023681461811,
      "reward_std_all": 0.18067028373479843,
      "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": 0.5447501540184021,
      "rewards/forensics_strict_edge_reward": 0.3641393333673477,
      "rewards/forensics_tiou_reward": 0.3834247291088104,
      "step": 890
    },
    {
      "completion_length": 19.328125,
      "epoch": 2.37017310252996,
      "grad_norm": 25.651466369628906,
      "kl": 3.2734375,
      "learning_rate": 4.0666666666666666e-07,
      "loss": 0.3103,
      "reward": 0.6339250206947327,
      "reward_std": 0.11829004809260368,
      "reward_std_all": 0.14169377088546753,
      "rewards-max/forensics_sinkhorn_reward": 0.9881109297275543,
      "rewards-max/forensics_strict_edge_reward": 0.9357379674911499,
      "rewards-max/forensics_tiou_reward": 0.9965358674526215,
      "rewards/forensics_sinkhorn_reward": 0.565077394247055,
      "rewards/forensics_strict_edge_reward": 0.3065223842859268,
      "rewards/forensics_tiou_reward": 0.3726568967103958,
      "step": 891
    },
    {
      "completion_length": 21.671875,
      "epoch": 2.3728362183754994,
      "grad_norm": 23.885089874267578,
      "kl": 2.6875,
      "learning_rate": 4.06e-07,
      "loss": 0.2207,
      "reward": 0.6659747064113617,
      "reward_std": 0.15265366435050964,
      "reward_std_all": 0.14993800222873688,
      "rewards-max/forensics_sinkhorn_reward": 0.9882354736328125,
      "rewards-max/forensics_strict_edge_reward": 0.9782643616199493,
      "rewards-max/forensics_tiou_reward": 0.9970413744449615,
      "rewards/forensics_sinkhorn_reward": 0.6258174777030945,
      "rewards/forensics_strict_edge_reward": 0.30493099987506866,
      "rewards/forensics_tiou_reward": 0.4822522699832916,
      "step": 892
    },
    {
      "completion_length": 29.421875,
      "epoch": 2.3754993342210384,
      "grad_norm": 32.506771087646484,
      "kl": 4.65625,
      "learning_rate": 4.053333333333333e-07,
      "loss": 0.2286,
      "reward": 0.5971240103244781,
      "reward_std": 0.19926001131534576,
      "reward_std_all": 0.2008877843618393,
      "rewards-max/forensics_sinkhorn_reward": 0.9755093455314636,
      "rewards-max/forensics_strict_edge_reward": 0.8578422963619232,
      "rewards-max/forensics_tiou_reward": 0.9671612977981567,
      "rewards/forensics_sinkhorn_reward": 0.5167247951030731,
      "rewards/forensics_strict_edge_reward": 0.2559056133031845,
      "rewards/forensics_tiou_reward": 0.3405793607234955,
      "step": 893
    },
    {
      "completion_length": 19.71875,
      "epoch": 2.378162450066578,
      "grad_norm": 25.96889305114746,
      "kl": 2.0078125,
      "learning_rate": 4.0466666666666666e-07,
      "loss": 0.279,
      "reward": 0.6746009588241577,
      "reward_std": 0.17761658132076263,
      "reward_std_all": 0.16275933384895325,
      "rewards-max/forensics_sinkhorn_reward": 0.9784989655017853,
      "rewards-max/forensics_strict_edge_reward": 0.8805811703205109,
      "rewards-max/forensics_tiou_reward": 0.9763144552707672,
      "rewards/forensics_sinkhorn_reward": 0.6026564836502075,
      "rewards/forensics_strict_edge_reward": 0.30505600571632385,
      "rewards/forensics_tiou_reward": 0.3977281004190445,
      "step": 894
    },
    {
      "completion_length": 21.125,
      "epoch": 2.3808255659121174,
      "grad_norm": 26.800832748413086,
      "kl": 5.28125,
      "learning_rate": 4.04e-07,
      "loss": 0.4323,
      "reward": 0.6861839592456818,
      "reward_std": 0.13401498273015022,
      "reward_std_all": 0.12741585075855255,
      "rewards-max/forensics_sinkhorn_reward": 0.9877351224422455,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.995056688785553,
      "rewards/forensics_sinkhorn_reward": 0.6257510781288147,
      "rewards/forensics_strict_edge_reward": 0.338500052690506,
      "rewards/forensics_tiou_reward": 0.4876707196235657,
      "step": 895
    },
    {
      "completion_length": 23.125,
      "epoch": 2.3834886817576564,
      "grad_norm": 23.778541564941406,
      "kl": 4.1875,
      "learning_rate": 4.033333333333333e-07,
      "loss": 0.2551,
      "reward": 0.6420460641384125,
      "reward_std": 0.17480874806642532,
      "reward_std_all": 0.17071104049682617,
      "rewards-max/forensics_sinkhorn_reward": 0.9882274866104126,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9969936907291412,
      "rewards/forensics_sinkhorn_reward": 0.5717858374118805,
      "rewards/forensics_strict_edge_reward": 0.243522509932518,
      "rewards/forensics_tiou_reward": 0.3663324564695358,
      "step": 896
    },
    {
      "completion_length": 21.46875,
      "epoch": 2.386151797603196,
      "grad_norm": 32.251487731933594,
      "kl": 4.65625,
      "learning_rate": 4.0266666666666667e-07,
      "loss": 0.234,
      "reward": 0.6050671935081482,
      "reward_std": 0.1852743849158287,
      "reward_std_all": 0.1874101459980011,
      "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": 0.5409140586853027,
      "rewards/forensics_strict_edge_reward": 0.23370933532714844,
      "rewards/forensics_tiou_reward": 0.30515915155410767,
      "step": 897
    },
    {
      "completion_length": 28.859375,
      "epoch": 2.388814913448735,
      "grad_norm": 67.82549285888672,
      "kl": 4.0234375,
      "learning_rate": 4.02e-07,
      "loss": 0.3401,
      "reward": 0.58537757396698,
      "reward_std": 0.22632581740617752,
      "reward_std_all": 0.22252816706895828,
      "rewards-max/forensics_sinkhorn_reward": 0.9882349669933319,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.997024416923523,
      "rewards/forensics_sinkhorn_reward": 0.47558607161045074,
      "rewards/forensics_strict_edge_reward": 0.2555146962404251,
      "rewards/forensics_tiou_reward": 0.2827310264110565,
      "step": 898
    },
    {
      "completion_length": 27.21875,
      "epoch": 2.3914780292942743,
      "grad_norm": 30.88664436340332,
      "kl": 4.2734375,
      "learning_rate": 4.0133333333333333e-07,
      "loss": 0.2707,
      "reward": 0.6644379198551178,
      "reward_std": 0.1692204773426056,
      "reward_std_all": 0.162438303232193,
      "rewards-max/forensics_sinkhorn_reward": 0.9887466132640839,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9990387260913849,
      "rewards/forensics_sinkhorn_reward": 0.6219297349452972,
      "rewards/forensics_strict_edge_reward": 0.30362625420093536,
      "rewards/forensics_tiou_reward": 0.4459120184183121,
      "step": 899
    },
    {
      "completion_length": 16.015625,
      "epoch": 2.3941411451398134,
      "grad_norm": 23.15738296508789,
      "kl": 1.59765625,
      "learning_rate": 4.0066666666666663e-07,
      "loss": 0.1087,
      "reward": 0.6900676190853119,
      "reward_std": 0.08632035087794065,
      "reward_std_all": 0.10046042315661907,
      "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": 0.636760801076889,
      "rewards/forensics_strict_edge_reward": 0.2659456208348274,
      "rewards/forensics_tiou_reward": 0.38953885436058044,
      "step": 900
    },
    {
      "completion_length": 32.234375,
      "epoch": 2.396804260985353,
      "grad_norm": 62.81169891357422,
      "kl": 4.84375,
      "learning_rate": 4e-07,
      "loss": 0.3481,
      "reward": 0.6066671311855316,
      "reward_std": 0.18336224555969238,
      "reward_std_all": 0.1874041110277176,
      "rewards-max/forensics_sinkhorn_reward": 0.9884401857852936,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9978321194648743,
      "rewards/forensics_sinkhorn_reward": 0.5217698514461517,
      "rewards/forensics_strict_edge_reward": 0.27060848474502563,
      "rewards/forensics_tiou_reward": 0.36918923258781433,
      "step": 901
    },
    {
      "completion_length": 25.671875,
      "epoch": 2.3994673768308923,
      "grad_norm": 19.8176326751709,
      "kl": 2.703125,
      "learning_rate": 3.993333333333333e-07,
      "loss": 0.2385,
      "reward": 0.6678020358085632,
      "reward_std": 0.10112805664539337,
      "reward_std_all": 0.11802821233868599,
      "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": 0.560109943151474,
      "rewards/forensics_strict_edge_reward": 0.25216934084892273,
      "rewards/forensics_tiou_reward": 0.30264410376548767,
      "step": 902
    },
    {
      "completion_length": 19.4375,
      "epoch": 2.4021304926764313,
      "grad_norm": 29.4451847076416,
      "kl": 3.3828125,
      "learning_rate": 3.9866666666666664e-07,
      "loss": 0.3922,
      "reward": 0.6647375524044037,
      "reward_std": 0.18112413585186005,
      "reward_std_all": 0.16620182991027832,
      "rewards-max/forensics_sinkhorn_reward": 0.9884959161281586,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.998051255941391,
      "rewards/forensics_sinkhorn_reward": 0.5856864154338837,
      "rewards/forensics_strict_edge_reward": 0.2784605920314789,
      "rewards/forensics_tiou_reward": 0.3859473317861557,
      "step": 903
    },
    {
      "completion_length": 25.296875,
      "epoch": 2.4047936085219708,
      "grad_norm": 26.071550369262695,
      "kl": 3.46875,
      "learning_rate": 3.98e-07,
      "loss": 0.256,
      "reward": 0.5970422625541687,
      "reward_std": 0.16586223989725113,
      "reward_std_all": 0.18872510641813278,
      "rewards-max/forensics_sinkhorn_reward": 0.9512248635292053,
      "rewards-max/forensics_strict_edge_reward": 0.8876297473907471,
      "rewards-max/forensics_tiou_reward": 0.9168268144130707,
      "rewards/forensics_sinkhorn_reward": 0.5022826790809631,
      "rewards/forensics_strict_edge_reward": 0.205628402531147,
      "rewards/forensics_tiou_reward": 0.27643832564353943,
      "step": 904
    },
    {
      "completion_length": 47.59375,
      "epoch": 2.40745672436751,
      "grad_norm": 30.632966995239258,
      "kl": 5.20703125,
      "learning_rate": 3.973333333333333e-07,
      "loss": 0.3161,
      "reward": 0.6355545222759247,
      "reward_std": 0.1528870314359665,
      "reward_std_all": 0.16258373484015465,
      "rewards-max/forensics_sinkhorn_reward": 0.9671759903430939,
      "rewards-max/forensics_strict_edge_reward": 0.8983768224716187,
      "rewards-max/forensics_tiou_reward": 0.9488550424575806,
      "rewards/forensics_sinkhorn_reward": 0.518071860074997,
      "rewards/forensics_strict_edge_reward": 0.26915179193019867,
      "rewards/forensics_tiou_reward": 0.31345856934785843,
      "step": 905
    },
    {
      "completion_length": 27.296875,
      "epoch": 2.4101198402130493,
      "grad_norm": 41.52530288696289,
      "kl": 4.4140625,
      "learning_rate": 3.9666666666666665e-07,
      "loss": 0.4286,
      "reward": 0.6605657339096069,
      "reward_std": 0.16415146738290787,
      "reward_std_all": 0.15536778420209885,
      "rewards-max/forensics_sinkhorn_reward": 0.988330602645874,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.997399240732193,
      "rewards/forensics_sinkhorn_reward": 0.5642586648464203,
      "rewards/forensics_strict_edge_reward": 0.3422940969467163,
      "rewards/forensics_tiou_reward": 0.4019557237625122,
      "step": 906
    },
    {
      "completion_length": 25.484375,
      "epoch": 2.4127829560585887,
      "grad_norm": 26.988998413085938,
      "kl": 3.37109375,
      "learning_rate": 3.96e-07,
      "loss": 0.2602,
      "reward": 0.5890657305717468,
      "reward_std": 0.18603325635194778,
      "reward_std_all": 0.1948796659708023,
      "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": 0.5540212392807007,
      "rewards/forensics_strict_edge_reward": 0.2884986326098442,
      "rewards/forensics_tiou_reward": 0.3581410348415375,
      "step": 907
    },
    {
      "completion_length": 22.921875,
      "epoch": 2.4154460719041277,
      "grad_norm": 20.8009090423584,
      "kl": 1.6875,
      "learning_rate": 3.953333333333333e-07,
      "loss": -0.017,
      "reward": 0.6223446428775787,
      "reward_std": 0.158158078789711,
      "reward_std_all": 0.1707221120595932,
      "rewards-max/forensics_sinkhorn_reward": 0.9383887350559235,
      "rewards-max/forensics_strict_edge_reward": 0.7901000380516052,
      "rewards-max/forensics_tiou_reward": 0.8919376730918884,
      "rewards/forensics_sinkhorn_reward": 0.5146452039480209,
      "rewards/forensics_strict_edge_reward": 0.15110093355178833,
      "rewards/forensics_tiou_reward": 0.25469499081373215,
      "step": 908
    },
    {
      "completion_length": 18.375,
      "epoch": 2.418109187749667,
      "grad_norm": 18.641803741455078,
      "kl": 1.6015625,
      "learning_rate": 3.9466666666666665e-07,
      "loss": 0.2416,
      "reward": 0.6805898547172546,
      "reward_std": 0.13221698999404907,
      "reward_std_all": 0.13179239630699158,
      "rewards-max/forensics_sinkhorn_reward": 0.9861218333244324,
      "rewards-max/forensics_strict_edge_reward": 0.9169811904430389,
      "rewards-max/forensics_tiou_reward": 0.9887554347515106,
      "rewards/forensics_sinkhorn_reward": 0.5926696956157684,
      "rewards/forensics_strict_edge_reward": 0.29029615223407745,
      "rewards/forensics_tiou_reward": 0.3714483231306076,
      "step": 909
    },
    {
      "completion_length": 22.3125,
      "epoch": 2.4207723035952062,
      "grad_norm": 35.01204299926758,
      "kl": 1.751953125,
      "learning_rate": 3.94e-07,
      "loss": 0.2482,
      "reward": 0.6234636306762695,
      "reward_std": 0.17694517970085144,
      "reward_std_all": 0.1838075891137123,
      "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": 0.5701249539852142,
      "rewards/forensics_strict_edge_reward": 0.27475786209106445,
      "rewards/forensics_tiou_reward": 0.3699491173028946,
      "step": 910
    },
    {
      "completion_length": 31.859375,
      "epoch": 2.4234354194407457,
      "grad_norm": 22.208864212036133,
      "kl": 2.6796875,
      "learning_rate": 3.933333333333333e-07,
      "loss": 0.3491,
      "reward": 0.5682701170444489,
      "reward_std": 0.1951604261994362,
      "reward_std_all": 0.21897359192371368,
      "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": 0.4997349679470062,
      "rewards/forensics_strict_edge_reward": 0.23872075229883194,
      "rewards/forensics_tiou_reward": 0.2911631017923355,
      "step": 911
    },
    {
      "completion_length": 28.546875,
      "epoch": 2.426098535286285,
      "grad_norm": 24.176958084106445,
      "kl": 2.25,
      "learning_rate": 3.9266666666666666e-07,
      "loss": 0.2638,
      "reward": 0.6436959505081177,
      "reward_std": 0.12617265805602074,
      "reward_std_all": 0.14140843600034714,
      "rewards-max/forensics_sinkhorn_reward": 0.9848736822605133,
      "rewards-max/forensics_strict_edge_reward": 0.9062341153621674,
      "rewards-max/forensics_tiou_reward": 0.9838954508304596,
      "rewards/forensics_sinkhorn_reward": 0.5389971733093262,
      "rewards/forensics_strict_edge_reward": 0.264014296233654,
      "rewards/forensics_tiou_reward": 0.30920204520225525,
      "step": 912
    },
    {
      "completion_length": 35.890625,
      "epoch": 2.428761651131824,
      "grad_norm": 21.65803337097168,
      "kl": 4.390625,
      "learning_rate": 3.92e-07,
      "loss": 0.5201,
      "reward": 0.591519683599472,
      "reward_std": 0.12106315791606903,
      "reward_std_all": 0.16667042672634125,
      "rewards-max/forensics_sinkhorn_reward": 0.9885322749614716,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9981936812400818,
      "rewards/forensics_sinkhorn_reward": 0.5191356986761093,
      "rewards/forensics_strict_edge_reward": 0.21808252483606339,
      "rewards/forensics_tiou_reward": 0.2845825105905533,
      "step": 913
    },
    {
      "completion_length": 38.09375,
      "epoch": 2.4314247669773636,
      "grad_norm": 32.56275177001953,
      "kl": 3.6328125,
      "learning_rate": 3.913333333333333e-07,
      "loss": 0.2639,
      "reward": 0.5706286728382111,
      "reward_std": 0.25234346836805344,
      "reward_std_all": 0.24539127200841904,
      "rewards-max/forensics_sinkhorn_reward": 0.9461337029933929,
      "rewards-max/forensics_strict_edge_reward": 0.9169811904430389,
      "rewards-max/forensics_tiou_reward": 0.911078542470932,
      "rewards/forensics_sinkhorn_reward": 0.43026575446128845,
      "rewards/forensics_strict_edge_reward": 0.19747582077980042,
      "rewards/forensics_tiou_reward": 0.2240256816148758,
      "step": 914
    },
    {
      "completion_length": 48.234375,
      "epoch": 2.4340878828229027,
      "grad_norm": 24.69686508178711,
      "kl": 4.953125,
      "learning_rate": 3.906666666666666e-07,
      "loss": 0.4513,
      "reward": 0.5909353494644165,
      "reward_std": 0.22125592827796936,
      "reward_std_all": 0.21781647205352783,
      "rewards-max/forensics_sinkhorn_reward": 0.9883792102336884,
      "rewards-max/forensics_strict_edge_reward": 0.9890114367008209,
      "rewards-max/forensics_tiou_reward": 0.9975903630256653,
      "rewards/forensics_sinkhorn_reward": 0.539307177066803,
      "rewards/forensics_strict_edge_reward": 0.28940972685813904,
      "rewards/forensics_tiou_reward": 0.3635767251253128,
      "step": 915
    },
    {
      "completion_length": 40.65625,
      "epoch": 2.436750998668442,
      "grad_norm": 24.031208038330078,
      "kl": 2.8046875,
      "learning_rate": 3.8999999999999997e-07,
      "loss": 0.2186,
      "reward": 0.6480360627174377,
      "reward_std": 0.17823409289121628,
      "reward_std_all": 0.17239882797002792,
      "rewards-max/forensics_sinkhorn_reward": 0.9879830479621887,
      "rewards-max/forensics_strict_edge_reward": 0.9565287232398987,
      "rewards-max/forensics_tiou_reward": 0.9960333704948425,
      "rewards/forensics_sinkhorn_reward": 0.5810269713401794,
      "rewards/forensics_strict_edge_reward": 0.31171052902936935,
      "rewards/forensics_tiou_reward": 0.3766689598560333,
      "step": 916
    },
    {
      "completion_length": 39.515625,
      "epoch": 2.4394141145139816,
      "grad_norm": 29.01822853088379,
      "kl": 4.796875,
      "learning_rate": 3.8933333333333327e-07,
      "loss": 0.3614,
      "reward": 0.6297548413276672,
      "reward_std": 0.15922807157039642,
      "reward_std_all": 0.16361870616674423,
      "rewards-max/forensics_sinkhorn_reward": 0.9773847162723541,
      "rewards-max/forensics_strict_edge_reward": 0.9252271056175232,
      "rewards-max/forensics_tiou_reward": 0.9710353314876556,
      "rewards/forensics_sinkhorn_reward": 0.5323998928070068,
      "rewards/forensics_strict_edge_reward": 0.2770979106426239,
      "rewards/forensics_tiou_reward": 0.3611520081758499,
      "step": 917
    },
    {
      "completion_length": 43.078125,
      "epoch": 2.4420772303595206,
      "grad_norm": 67.79995727539062,
      "kl": 4.15625,
      "learning_rate": 3.8866666666666663e-07,
      "loss": 0.3093,
      "reward": 0.5526625514030457,
      "reward_std": 0.18216384947299957,
      "reward_std_all": 0.20969875901937485,
      "rewards-max/forensics_sinkhorn_reward": 0.9857239723205566,
      "rewards-max/forensics_strict_edge_reward": 0.9169811904430389,
      "rewards-max/forensics_tiou_reward": 0.9871290326118469,
      "rewards/forensics_sinkhorn_reward": 0.4686707705259323,
      "rewards/forensics_strict_edge_reward": 0.2764437198638916,
      "rewards/forensics_tiou_reward": 0.3332536220550537,
      "step": 918
    },
    {
      "completion_length": 57.484375,
      "epoch": 2.44474034620506,
      "grad_norm": 38.45307159423828,
      "kl": 6.6953125,
      "learning_rate": 3.88e-07,
      "loss": 0.4121,
      "reward": 0.6042460799217224,
      "reward_std": 0.20263086259365082,
      "reward_std_all": 0.20386292040348053,
      "rewards-max/forensics_sinkhorn_reward": 0.9586120247840881,
      "rewards-max/forensics_strict_edge_reward": 0.6834079623222351,
      "rewards-max/forensics_tiou_reward": 0.928041011095047,
      "rewards/forensics_sinkhorn_reward": 0.49619725346565247,
      "rewards/forensics_strict_edge_reward": 0.22689666971564293,
      "rewards/forensics_tiou_reward": 0.29205968976020813,
      "step": 919
    },
    {
      "completion_length": 40.359375,
      "epoch": 2.447403462050599,
      "grad_norm": 37.95024490356445,
      "kl": 5.890625,
      "learning_rate": 3.873333333333333e-07,
      "loss": 0.4414,
      "reward": 0.6037519574165344,
      "reward_std": 0.1861734390258789,
      "reward_std_all": 0.1907912865281105,
      "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": 0.5255110114812851,
      "rewards/forensics_strict_edge_reward": 0.24295973777770996,
      "rewards/forensics_tiou_reward": 0.3301210552453995,
      "step": 920
    },
    {
      "completion_length": 76.390625,
      "epoch": 2.4500665778961386,
      "grad_norm": 35.91698455810547,
      "kl": 6.6875,
      "learning_rate": 3.8666666666666664e-07,
      "loss": 0.5449,
      "reward": 0.5148061960935593,
      "reward_std": 0.2653368264436722,
      "reward_std_all": 0.2773177698254585,
      "rewards-max/forensics_sinkhorn_reward": 0.8676612675189972,
      "rewards-max/forensics_strict_edge_reward": 0.7667815089225769,
      "rewards-max/forensics_tiou_reward": 0.771830826997757,
      "rewards/forensics_sinkhorn_reward": 0.46202144026756287,
      "rewards/forensics_strict_edge_reward": 0.29524148255586624,
      "rewards/forensics_tiou_reward": 0.34123384952545166,
      "step": 921
    },
    {
      "completion_length": 46.015625,
      "epoch": 2.4527296937416776,
      "grad_norm": 26.576778411865234,
      "kl": 5.40625,
      "learning_rate": 3.86e-07,
      "loss": 0.3582,
      "reward": 0.5771330446004868,
      "reward_std": 0.23755022883415222,
      "reward_std_all": 0.24002136290073395,
      "rewards-max/forensics_sinkhorn_reward": 0.9292957484722137,
      "rewards-max/forensics_strict_edge_reward": 0.7951961755752563,
      "rewards-max/forensics_tiou_reward": 0.8754190504550934,
      "rewards/forensics_sinkhorn_reward": 0.5103367865085602,
      "rewards/forensics_strict_edge_reward": 0.29193340241909027,
      "rewards/forensics_tiou_reward": 0.37223872542381287,
      "step": 922
    },
    {
      "completion_length": 62.890625,
      "epoch": 2.455392809587217,
      "grad_norm": 18.266246795654297,
      "kl": 4.265625,
      "learning_rate": 3.8533333333333334e-07,
      "loss": 0.3081,
      "reward": 0.5656977295875549,
      "reward_std": 0.18209335207939148,
      "reward_std_all": 0.20955614000558853,
      "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": 0.5384808480739594,
      "rewards/forensics_strict_edge_reward": 0.28285130858421326,
      "rewards/forensics_tiou_reward": 0.36282962560653687,
      "step": 923
    },
    {
      "completion_length": 69.953125,
      "epoch": 2.458055925432756,
      "grad_norm": 40.57970428466797,
      "kl": 8.0,
      "learning_rate": 3.8466666666666664e-07,
      "loss": 0.5601,
      "reward": 0.56723752617836,
      "reward_std": 0.20192153751850128,
      "reward_std_all": 0.21955659985542297,
      "rewards-max/forensics_sinkhorn_reward": 0.9747021794319153,
      "rewards-max/forensics_strict_edge_reward": 0.8719756007194519,
      "rewards-max/forensics_tiou_reward": 0.9658767580986023,
      "rewards/forensics_sinkhorn_reward": 0.4932050108909607,
      "rewards/forensics_strict_edge_reward": 0.26542915403842926,
      "rewards/forensics_tiou_reward": 0.3519667088985443,
      "step": 924
    },
    {
      "completion_length": 68.8125,
      "epoch": 2.4607190412782955,
      "grad_norm": 25.298038482666016,
      "kl": 5.53125,
      "learning_rate": 3.84e-07,
      "loss": 0.5638,
      "reward": 0.5309375822544098,
      "reward_std": 0.14395035430788994,
      "reward_std_all": 0.2081482857465744,
      "rewards-max/forensics_sinkhorn_reward": 0.9690580666065216,
      "rewards-max/forensics_strict_edge_reward": 0.9357379674911499,
      "rewards-max/forensics_tiou_reward": 0.9555900394916534,
      "rewards/forensics_sinkhorn_reward": 0.4154244512319565,
      "rewards/forensics_strict_edge_reward": 0.20275183022022247,
      "rewards/forensics_tiou_reward": 0.23562131822109222,
      "step": 925
    },
    {
      "completion_length": 59.671875,
      "epoch": 2.463382157123835,
      "grad_norm": 22.208528518676758,
      "kl": 4.328125,
      "learning_rate": 3.8333333333333335e-07,
      "loss": 0.461,
      "reward": 0.5567042231559753,
      "reward_std": 0.2408178597688675,
      "reward_std_all": 0.23773526400327682,
      "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": 0.49146081507205963,
      "rewards/forensics_strict_edge_reward": 0.2621416598558426,
      "rewards/forensics_tiou_reward": 0.35369904339313507,
      "step": 926
    },
    {
      "completion_length": 62.125,
      "epoch": 2.466045272969374,
      "grad_norm": 28.52703285217285,
      "kl": 4.4453125,
      "learning_rate": 3.8266666666666665e-07,
      "loss": 0.3206,
      "reward": 0.6369433701038361,
      "reward_std": 0.15503834933042526,
      "reward_std_all": 0.16132812201976776,
      "rewards-max/forensics_sinkhorn_reward": 0.9404881596565247,
      "rewards-max/forensics_strict_edge_reward": 0.7588363587856293,
      "rewards-max/forensics_tiou_reward": 0.8889747858047485,
      "rewards/forensics_sinkhorn_reward": 0.5647776573896408,
      "rewards/forensics_strict_edge_reward": 0.27915505692362785,
      "rewards/forensics_tiou_reward": 0.37545932829380035,
      "step": 927
    },
    {
      "completion_length": 44.46875,
      "epoch": 2.4687083888149135,
      "grad_norm": 22.53175926208496,
      "kl": 3.375,
      "learning_rate": 3.82e-07,
      "loss": 0.3376,
      "reward": 0.5586856305599213,
      "reward_std": 0.1968165934085846,
      "reward_std_all": 0.21515922248363495,
      "rewards-max/forensics_sinkhorn_reward": 0.988672286272049,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9987458288669586,
      "rewards/forensics_sinkhorn_reward": 0.4957450330257416,
      "rewards/forensics_strict_edge_reward": 0.25759759545326233,
      "rewards/forensics_tiou_reward": 0.32335494458675385,
      "step": 928
    },
    {
      "completion_length": 28.71875,
      "epoch": 2.4713715046604525,
      "grad_norm": 24.43876838684082,
      "kl": 3.203125,
      "learning_rate": 3.8133333333333336e-07,
      "loss": 0.188,
      "reward": 0.566681444644928,
      "reward_std": 0.16780296340584755,
      "reward_std_all": 0.2005317285656929,
      "rewards-max/forensics_sinkhorn_reward": 0.9743717312812805,
      "rewards-max/forensics_strict_edge_reward": 0.8317028284072876,
      "rewards-max/forensics_tiou_reward": 0.9663669466972351,
      "rewards/forensics_sinkhorn_reward": 0.5028165727853775,
      "rewards/forensics_strict_edge_reward": 0.19275441765785217,
      "rewards/forensics_tiou_reward": 0.317626029253006,
      "step": 929
    },
    {
      "completion_length": 36.484375,
      "epoch": 2.474034620505992,
      "grad_norm": 24.622196197509766,
      "kl": 6.15625,
      "learning_rate": 3.8066666666666666e-07,
      "loss": 0.4897,
      "reward": 0.47117213904857635,
      "reward_std": 0.24189157038927078,
      "reward_std_all": 0.28667354583740234,
      "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": 0.44355255365371704,
      "rewards/forensics_strict_edge_reward": 0.30342909693717957,
      "rewards/forensics_tiou_reward": 0.36822161078453064,
      "step": 930
    },
    {
      "completion_length": 25.609375,
      "epoch": 2.4766977363515315,
      "grad_norm": 28.147754669189453,
      "kl": 5.1875,
      "learning_rate": 3.7999999999999996e-07,
      "loss": 0.5265,
      "reward": 0.5304201692342758,
      "reward_std": 0.16791581362485886,
      "reward_std_all": 0.21731537580490112,
      "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": 0.5077140033245087,
      "rewards/forensics_strict_edge_reward": 0.3559775650501251,
      "rewards/forensics_tiou_reward": 0.40732017159461975,
      "step": 931
    },
    {
      "completion_length": 26.234375,
      "epoch": 2.4793608521970705,
      "grad_norm": 22.25926971435547,
      "kl": 4.0078125,
      "learning_rate": 3.793333333333333e-07,
      "loss": 0.364,
      "reward": 0.5603245794773102,
      "reward_std": 0.19074852764606476,
      "reward_std_all": 0.2170441448688507,
      "rewards-max/forensics_sinkhorn_reward": 0.9887480437755585,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9990443587303162,
      "rewards/forensics_sinkhorn_reward": 0.5522575080394745,
      "rewards/forensics_strict_edge_reward": 0.3007581979036331,
      "rewards/forensics_tiou_reward": 0.4195190966129303,
      "step": 932
    },
    {
      "completion_length": 17.78125,
      "epoch": 2.48202396804261,
      "grad_norm": 22.04395866394043,
      "kl": 2.296875,
      "learning_rate": 3.786666666666666e-07,
      "loss": 0.1005,
      "reward": 0.6193685531616211,
      "reward_std": 0.1848411187529564,
      "reward_std_all": 0.18385371565818787,
      "rewards-max/forensics_sinkhorn_reward": 0.9875493347644806,
      "rewards-max/forensics_strict_edge_reward": 0.9464850425720215,
      "rewards-max/forensics_tiou_reward": 0.9943199455738068,
      "rewards/forensics_sinkhorn_reward": 0.6030300855636597,
      "rewards/forensics_strict_edge_reward": 0.3572253882884979,
      "rewards/forensics_tiou_reward": 0.4304203540086746,
      "step": 933
    },
    {
      "completion_length": 39.109375,
      "epoch": 2.484687083888149,
      "grad_norm": 23.85295295715332,
      "kl": 5.265625,
      "learning_rate": 3.7799999999999997e-07,
      "loss": 0.3699,
      "reward": 0.5091898739337921,
      "reward_std": 0.23356564342975616,
      "reward_std_all": 0.2593948245048523,
      "rewards-max/forensics_sinkhorn_reward": 0.9880107939243317,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9961384236812592,
      "rewards/forensics_sinkhorn_reward": 0.47507938742637634,
      "rewards/forensics_strict_edge_reward": 0.28157204389572144,
      "rewards/forensics_tiou_reward": 0.3451547771692276,
      "step": 934
    },
    {
      "completion_length": 18.296875,
      "epoch": 2.4873501997336884,
      "grad_norm": 25.82586097717285,
      "kl": 3.5234375,
      "learning_rate": 3.773333333333333e-07,
      "loss": 0.2447,
      "reward": 0.5893076658248901,
      "reward_std": 0.21634554117918015,
      "reward_std_all": 0.2102297767996788,
      "rewards-max/forensics_sinkhorn_reward": 0.9547480940818787,
      "rewards-max/forensics_strict_edge_reward": 0.8096016347408295,
      "rewards-max/forensics_tiou_reward": 0.9261235594749451,
      "rewards/forensics_sinkhorn_reward": 0.5105103552341461,
      "rewards/forensics_strict_edge_reward": 0.17293621599674225,
      "rewards/forensics_tiou_reward": 0.26879578083753586,
      "step": 935
    },
    {
      "completion_length": 30.390625,
      "epoch": 2.490013315579228,
      "grad_norm": 21.849699020385742,
      "kl": 4.15625,
      "learning_rate": 3.766666666666666e-07,
      "loss": 0.3724,
      "reward": 0.5441367924213409,
      "reward_std": 0.22209329903125763,
      "reward_std_all": 0.23881027102470398,
      "rewards-max/forensics_sinkhorn_reward": 0.9884404540061951,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9978328943252563,
      "rewards/forensics_sinkhorn_reward": 0.4752993881702423,
      "rewards/forensics_strict_edge_reward": 0.22979436814785004,
      "rewards/forensics_tiou_reward": 0.3216138184070587,
      "step": 936
    },
    {
      "completion_length": 21.53125,
      "epoch": 2.492676431424767,
      "grad_norm": 21.438894271850586,
      "kl": 2.66015625,
      "learning_rate": 3.76e-07,
      "loss": 0.1158,
      "reward": 0.6107883155345917,
      "reward_std": 0.17315946519374847,
      "reward_std_all": 0.18058056384325027,
      "rewards-max/forensics_sinkhorn_reward": 0.987438440322876,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9938834607601166,
      "rewards/forensics_sinkhorn_reward": 0.5448976755142212,
      "rewards/forensics_strict_edge_reward": 0.24082112312316895,
      "rewards/forensics_tiou_reward": 0.31717683374881744,
      "step": 937
    },
    {
      "completion_length": 35.15625,
      "epoch": 2.4953395472703064,
      "grad_norm": 52.834938049316406,
      "kl": 6.046875,
      "learning_rate": 3.7533333333333333e-07,
      "loss": 0.5384,
      "reward": 0.5948922336101532,
      "reward_std": 0.11017829179763794,
      "reward_std_all": 0.1523696482181549,
      "rewards-max/forensics_sinkhorn_reward": 0.9885404109954834,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.998225599527359,
      "rewards/forensics_sinkhorn_reward": 0.5432183146476746,
      "rewards/forensics_strict_edge_reward": 0.22507362812757492,
      "rewards/forensics_tiou_reward": 0.30332837998867035,
      "step": 938
    },
    {
      "completion_length": 17.875,
      "epoch": 2.4980026631158454,
      "grad_norm": 29.790435791015625,
      "kl": 3.5234375,
      "learning_rate": 3.7466666666666663e-07,
      "loss": 0.2677,
      "reward": 0.6437385678291321,
      "reward_std": 0.11823435872793198,
      "reward_std_all": 0.14256662502884865,
      "rewards-max/forensics_sinkhorn_reward": 0.9881000220775604,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9964928030967712,
      "rewards/forensics_sinkhorn_reward": 0.5859695971012115,
      "rewards/forensics_strict_edge_reward": 0.34237460792064667,
      "rewards/forensics_tiou_reward": 0.41467514634132385,
      "step": 939
    },
    {
      "completion_length": 18.28125,
      "epoch": 2.500665778961385,
      "grad_norm": 82.49300384521484,
      "kl": 4.890625,
      "learning_rate": 3.74e-07,
      "loss": 0.4683,
      "reward": 0.6168459355831146,
      "reward_std": 0.24370187520980835,
      "reward_std_all": 0.22647938132286072,
      "rewards-max/forensics_sinkhorn_reward": 0.9869348704814911,
      "rewards-max/forensics_strict_edge_reward": 0.9565287232398987,
      "rewards-max/forensics_tiou_reward": 0.9919067919254303,
      "rewards/forensics_sinkhorn_reward": 0.57502681016922,
      "rewards/forensics_strict_edge_reward": 0.33530159294605255,
      "rewards/forensics_tiou_reward": 0.404366135597229,
      "step": 940
    },
    {
      "completion_length": 17.484375,
      "epoch": 2.5033288948069243,
      "grad_norm": 20.467559814453125,
      "kl": 2.1171875,
      "learning_rate": 3.7333333333333334e-07,
      "loss": 0.2636,
      "reward": 0.7065185308456421,
      "reward_std": 0.07964266836643219,
      "reward_std_all": 0.09229400753974915,
      "rewards-max/forensics_sinkhorn_reward": 0.9887368381023407,
      "rewards-max/forensics_strict_edge_reward": 0.9782643616199493,
      "rewards-max/forensics_tiou_reward": 0.9990000128746033,
      "rewards/forensics_sinkhorn_reward": 0.6663984060287476,
      "rewards/forensics_strict_edge_reward": 0.3732883334159851,
      "rewards/forensics_tiou_reward": 0.446419432759285,
      "step": 941
    },
    {
      "completion_length": 17.046875,
      "epoch": 2.5059920106524634,
      "grad_norm": 20.245752334594727,
      "kl": 1.98828125,
      "learning_rate": 3.7266666666666664e-07,
      "loss": 0.1868,
      "reward": 0.6367592215538025,
      "reward_std": 0.21171176433563232,
      "reward_std_all": 0.20094013214111328,
      "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": 0.5840948224067688,
      "rewards/forensics_strict_edge_reward": 0.3261598199605942,
      "rewards/forensics_tiou_reward": 0.397887647151947,
      "step": 942
    },
    {
      "completion_length": 23.6875,
      "epoch": 2.508655126498003,
      "grad_norm": 21.8060245513916,
      "kl": 1.89453125,
      "learning_rate": 3.72e-07,
      "loss": 0.1881,
      "reward": 0.6488339900970459,
      "reward_std": 0.13073676265776157,
      "reward_std_all": 0.13916553929448128,
      "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": 0.5999372601509094,
      "rewards/forensics_strict_edge_reward": 0.30646683275699615,
      "rewards/forensics_tiou_reward": 0.4029051512479782,
      "step": 943
    },
    {
      "completion_length": 29.484375,
      "epoch": 2.511318242343542,
      "grad_norm": 23.692014694213867,
      "kl": 4.234375,
      "learning_rate": 3.7133333333333335e-07,
      "loss": 0.4189,
      "reward": 0.5725472569465637,
      "reward_std": 0.19587420672178268,
      "reward_std_all": 0.21135984361171722,
      "rewards-max/forensics_sinkhorn_reward": 0.9889077544212341,
      "rewards-max/forensics_strict_edge_reward": 0.9890114367008209,
      "rewards-max/forensics_tiou_reward": 0.9996739029884338,
      "rewards/forensics_sinkhorn_reward": 0.501986175775528,
      "rewards/forensics_strict_edge_reward": 0.2423192784190178,
      "rewards/forensics_tiou_reward": 0.2794026583433151,
      "step": 944
    },
    {
      "completion_length": 21.71875,
      "epoch": 2.5139813581890813,
      "grad_norm": 28.215473175048828,
      "kl": 4.46875,
      "learning_rate": 3.7066666666666665e-07,
      "loss": 0.3604,
      "reward": 0.5716056823730469,
      "reward_std": 0.11894253641366959,
      "reward_std_all": 0.17238663136959076,
      "rewards-max/forensics_sinkhorn_reward": 0.9880601763725281,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9963328838348389,
      "rewards/forensics_sinkhorn_reward": 0.48731188476085663,
      "rewards/forensics_strict_edge_reward": 0.20162633806467056,
      "rewards/forensics_tiou_reward": 0.29320070147514343,
      "step": 945
    },
    {
      "completion_length": 23.203125,
      "epoch": 2.5166444740346208,
      "grad_norm": 40.944637298583984,
      "kl": 3.1875,
      "learning_rate": 3.7e-07,
      "loss": 0.2785,
      "reward": 0.5491845607757568,
      "reward_std": 0.24765951186418533,
      "reward_std_all": 0.25134846568107605,
      "rewards-max/forensics_sinkhorn_reward": 0.9856645166873932,
      "rewards-max/forensics_strict_edge_reward": 0.9357379674911499,
      "rewards-max/forensics_tiou_reward": 0.9869360625743866,
      "rewards/forensics_sinkhorn_reward": 0.4857131838798523,
      "rewards/forensics_strict_edge_reward": 0.20441755652427673,
      "rewards/forensics_tiou_reward": 0.2959018051624298,
      "step": 946
    },
    {
      "completion_length": 30.546875,
      "epoch": 2.51930758988016,
      "grad_norm": 28.067203521728516,
      "kl": 5.4375,
      "learning_rate": 3.693333333333333e-07,
      "loss": 0.4505,
      "reward": 0.5556835532188416,
      "reward_std": 0.21838876605033875,
      "reward_std_all": 0.23281222581863403,
      "rewards-max/forensics_sinkhorn_reward": 0.987695723772049,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9949044585227966,
      "rewards/forensics_sinkhorn_reward": 0.5209127515554428,
      "rewards/forensics_strict_edge_reward": 0.2948678880929947,
      "rewards/forensics_tiou_reward": 0.3870292603969574,
      "step": 947
    },
    {
      "completion_length": 34.578125,
      "epoch": 2.521970705725699,
      "grad_norm": 40.105628967285156,
      "kl": 6.9609375,
      "learning_rate": 3.686666666666666e-07,
      "loss": 0.5066,
      "reward": 0.5349991321563721,
      "reward_std": 0.16971951723098755,
      "reward_std_all": 0.21346910297870636,
      "rewards-max/forensics_sinkhorn_reward": 0.98251873254776,
      "rewards-max/forensics_strict_edge_reward": 0.8876297473907471,
      "rewards-max/forensics_tiou_reward": 0.9846581816673279,
      "rewards/forensics_sinkhorn_reward": 0.45004554092884064,
      "rewards/forensics_strict_edge_reward": 0.1994594782590866,
      "rewards/forensics_tiou_reward": 0.22748913615942,
      "step": 948
    },
    {
      "completion_length": 21.28125,
      "epoch": 2.5246338215712383,
      "grad_norm": 36.9461555480957,
      "kl": 5.90625,
      "learning_rate": 3.6799999999999996e-07,
      "loss": 0.3511,
      "reward": 0.5347173511981964,
      "reward_std": 0.2260253056883812,
      "reward_std_all": 0.2557060867547989,
      "rewards-max/forensics_sinkhorn_reward": 0.9875690937042236,
      "rewards-max/forensics_strict_edge_reward": 0.9574736058712006,
      "rewards-max/forensics_tiou_reward": 0.9944134652614594,
      "rewards/forensics_sinkhorn_reward": 0.4453486502170563,
      "rewards/forensics_strict_edge_reward": 0.16712110489606857,
      "rewards/forensics_tiou_reward": 0.23761644959449768,
      "step": 949
    },
    {
      "completion_length": 36.515625,
      "epoch": 2.5272969374167777,
      "grad_norm": 33.453407287597656,
      "kl": 6.078125,
      "learning_rate": 3.673333333333333e-07,
      "loss": 0.5229,
      "reward": 0.49807214736938477,
      "reward_std": 0.2436646819114685,
      "reward_std_all": 0.2657776102423668,
      "rewards-max/forensics_sinkhorn_reward": 0.9596930146217346,
      "rewards-max/forensics_strict_edge_reward": 0.8168056011199951,
      "rewards-max/forensics_tiou_reward": 0.9313094317913055,
      "rewards/forensics_sinkhorn_reward": 0.41523127257823944,
      "rewards/forensics_strict_edge_reward": 0.23752889037132263,
      "rewards/forensics_tiou_reward": 0.29094937443733215,
      "step": 950
    },
    {
      "completion_length": 21.78125,
      "epoch": 2.5299600532623168,
      "grad_norm": 30.275108337402344,
      "kl": 4.6015625,
      "learning_rate": 3.666666666666666e-07,
      "loss": 0.3366,
      "reward": 0.5325668454170227,
      "reward_std": 0.2984773516654968,
      "reward_std_all": 0.2833731174468994,
      "rewards-max/forensics_sinkhorn_reward": 0.960192859172821,
      "rewards-max/forensics_strict_edge_reward": 0.8968285918235779,
      "rewards-max/forensics_tiou_reward": 0.9327815175056458,
      "rewards/forensics_sinkhorn_reward": 0.44966474175453186,
      "rewards/forensics_strict_edge_reward": 0.1915852352976799,
      "rewards/forensics_tiou_reward": 0.3023683726787567,
      "step": 951
    },
    {
      "completion_length": 22.359375,
      "epoch": 2.5326231691078562,
      "grad_norm": 35.774044036865234,
      "kl": 3.8828125,
      "learning_rate": 3.6599999999999997e-07,
      "loss": 0.3033,
      "reward": 0.6041437983512878,
      "reward_std": 0.2066655158996582,
      "reward_std_all": 0.2116822898387909,
      "rewards-max/forensics_sinkhorn_reward": 0.935512900352478,
      "rewards-max/forensics_strict_edge_reward": 0.7229099273681641,
      "rewards-max/forensics_tiou_reward": 0.8828776180744171,
      "rewards/forensics_sinkhorn_reward": 0.5014670342206955,
      "rewards/forensics_strict_edge_reward": 0.1785358414053917,
      "rewards/forensics_tiou_reward": 0.30611440539360046,
      "step": 952
    },
    {
      "completion_length": 24.234375,
      "epoch": 2.5352862849533953,
      "grad_norm": 40.10868453979492,
      "kl": 5.390625,
      "learning_rate": 3.653333333333333e-07,
      "loss": 0.4489,
      "reward": 0.5490453243255615,
      "reward_std": 0.21735715121030807,
      "reward_std_all": 0.22981811314821243,
      "rewards-max/forensics_sinkhorn_reward": 0.9754160344600677,
      "rewards-max/forensics_strict_edge_reward": 0.9265981018543243,
      "rewards-max/forensics_tiou_reward": 0.9676740169525146,
      "rewards/forensics_sinkhorn_reward": 0.44410571455955505,
      "rewards/forensics_strict_edge_reward": 0.13052087277173996,
      "rewards/forensics_tiou_reward": 0.1971358060836792,
      "step": 953
    },
    {
      "completion_length": 26.046875,
      "epoch": 2.5379494007989347,
      "grad_norm": 35.05307388305664,
      "kl": 4.703125,
      "learning_rate": 3.646666666666666e-07,
      "loss": 0.3945,
      "reward": 0.5862710475921631,
      "reward_std": 0.18051248043775558,
      "reward_std_all": 0.20349671691656113,
      "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": 0.5202924907207489,
      "rewards/forensics_strict_edge_reward": 0.32177048921585083,
      "rewards/forensics_tiou_reward": 0.3530210554599762,
      "step": 954
    },
    {
      "completion_length": 19.84375,
      "epoch": 2.540612516644474,
      "grad_norm": 38.38772201538086,
      "kl": 6.453125,
      "learning_rate": 3.64e-07,
      "loss": 0.402,
      "reward": 0.5497422218322754,
      "reward_std": 0.19421889632940292,
      "reward_std_all": 0.21890047937631607,
      "rewards-max/forensics_sinkhorn_reward": 0.9836057424545288,
      "rewards-max/forensics_strict_edge_reward": 0.9364311099052429,
      "rewards-max/forensics_tiou_reward": 0.9878661036491394,
      "rewards/forensics_sinkhorn_reward": 0.5019663870334625,
      "rewards/forensics_strict_edge_reward": 0.21979525685310364,
      "rewards/forensics_tiou_reward": 0.34608352184295654,
      "step": 955
    },
    {
      "completion_length": 31.828125,
      "epoch": 2.543275632490013,
      "grad_norm": 21.196081161499023,
      "kl": 4.33984375,
      "learning_rate": 3.6333333333333333e-07,
      "loss": 0.3818,
      "reward": 0.5664638578891754,
      "reward_std": 0.18303027749061584,
      "reward_std_all": 0.20428498089313507,
      "rewards-max/forensics_sinkhorn_reward": 0.9874984622001648,
      "rewards-max/forensics_strict_edge_reward": 0.9567649364471436,
      "rewards-max/forensics_tiou_reward": 0.9941394627094269,
      "rewards/forensics_sinkhorn_reward": 0.4665120989084244,
      "rewards/forensics_strict_edge_reward": 0.19966118037700653,
      "rewards/forensics_tiou_reward": 0.2555083855986595,
      "step": 956
    },
    {
      "completion_length": 19.859375,
      "epoch": 2.5459387483355527,
      "grad_norm": 30.94243812561035,
      "kl": 4.328125,
      "learning_rate": 3.626666666666667e-07,
      "loss": 0.2814,
      "reward": 0.6164324879646301,
      "reward_std": 0.20512518286705017,
      "reward_std_all": 0.19848480820655823,
      "rewards-max/forensics_sinkhorn_reward": 0.945970892906189,
      "rewards-max/forensics_strict_edge_reward": 0.8968285918235779,
      "rewards-max/forensics_tiou_reward": 0.9064688086509705,
      "rewards/forensics_sinkhorn_reward": 0.5101704597473145,
      "rewards/forensics_strict_edge_reward": 0.18862546980381012,
      "rewards/forensics_tiou_reward": 0.2704256922006607,
      "step": 957
    },
    {
      "completion_length": 29.359375,
      "epoch": 2.5486018641810917,
      "grad_norm": 20.644588470458984,
      "kl": 2.796875,
      "learning_rate": 3.62e-07,
      "loss": 0.2713,
      "reward": 0.6326611936092377,
      "reward_std": 0.22118155658245087,
      "reward_std_all": 0.20381803810596466,
      "rewards-max/forensics_sinkhorn_reward": 0.9655823111534119,
      "rewards-max/forensics_strict_edge_reward": 0.844808965921402,
      "rewards-max/forensics_tiou_reward": 0.9456263482570648,
      "rewards/forensics_sinkhorn_reward": 0.5714946687221527,
      "rewards/forensics_strict_edge_reward": 0.2850882038474083,
      "rewards/forensics_tiou_reward": 0.4160546511411667,
      "step": 958
    },
    {
      "completion_length": 21.796875,
      "epoch": 2.551264980026631,
      "grad_norm": 25.647052764892578,
      "kl": 2.4921875,
      "learning_rate": 3.6133333333333334e-07,
      "loss": 0.1587,
      "reward": 0.6591014862060547,
      "reward_std": 0.1202443428337574,
      "reward_std_all": 0.13206437230110168,
      "rewards-max/forensics_sinkhorn_reward": 0.9876318573951721,
      "rewards-max/forensics_strict_edge_reward": 0.9565287232398987,
      "rewards-max/forensics_tiou_reward": 0.9946528971195221,
      "rewards/forensics_sinkhorn_reward": 0.6132147908210754,
      "rewards/forensics_strict_edge_reward": 0.2719944939017296,
      "rewards/forensics_tiou_reward": 0.3862431198358536,
      "step": 959
    },
    {
      "completion_length": 17.671875,
      "epoch": 2.5539280958721706,
      "grad_norm": 24.485395431518555,
      "kl": 3.03125,
      "learning_rate": 3.606666666666667e-07,
      "loss": 0.325,
      "reward": 0.6088318228721619,
      "reward_std": 0.1391855850815773,
      "reward_std_all": 0.15672509372234344,
      "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": 0.5127325654029846,
      "rewards/forensics_strict_edge_reward": 0.2541870027780533,
      "rewards/forensics_tiou_reward": 0.29740745574235916,
      "step": 960
    },
    {
      "completion_length": 18.734375,
      "epoch": 2.5565912117177096,
      "grad_norm": 52.27822494506836,
      "kl": 4.859375,
      "learning_rate": 3.6e-07,
      "loss": 0.4253,
      "reward": 0.610726535320282,
      "reward_std": 0.2168993353843689,
      "reward_std_all": 0.21735984086990356,
      "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": 0.5610262155532837,
      "rewards/forensics_strict_edge_reward": 0.33145633339881897,
      "rewards/forensics_tiou_reward": 0.3824755400419235,
      "step": 961
    },
    {
      "completion_length": 17.703125,
      "epoch": 2.559254327563249,
      "grad_norm": 27.600610733032227,
      "kl": 2.49609375,
      "learning_rate": 3.5933333333333335e-07,
      "loss": 0.2994,
      "reward": 0.5435107946395874,
      "reward_std": 0.19766266644001007,
      "reward_std_all": 0.2277745008468628,
      "rewards-max/forensics_sinkhorn_reward": 0.9720718264579773,
      "rewards-max/forensics_strict_edge_reward": 0.878633052110672,
      "rewards-max/forensics_tiou_reward": 0.9620969295501709,
      "rewards/forensics_sinkhorn_reward": 0.4737657755613327,
      "rewards/forensics_strict_edge_reward": 0.20160068571567535,
      "rewards/forensics_tiou_reward": 0.23923617601394653,
      "step": 962
    },
    {
      "completion_length": 20.25,
      "epoch": 2.561917443408788,
      "grad_norm": 19.765649795532227,
      "kl": 3.76171875,
      "learning_rate": 3.5866666666666665e-07,
      "loss": 0.3217,
      "reward": 0.6916297376155853,
      "reward_std": 0.12267602980136871,
      "reward_std_all": 0.11902330443263054,
      "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": 0.6646040678024292,
      "rewards/forensics_strict_edge_reward": 0.4078572392463684,
      "rewards/forensics_tiou_reward": 0.5286478698253632,
      "step": 963
    },
    {
      "completion_length": 15.96875,
      "epoch": 2.5645805592543276,
      "grad_norm": 24.782453536987305,
      "kl": 2.88671875,
      "learning_rate": 3.5799999999999995e-07,
      "loss": 0.3546,
      "reward": 0.6497908234596252,
      "reward_std": 0.13989946246147156,
      "reward_std_all": 0.1515050008893013,
      "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": 0.6096497476100922,
      "rewards/forensics_strict_edge_reward": 0.3108268678188324,
      "rewards/forensics_tiou_reward": 0.4551720470190048,
      "step": 964
    },
    {
      "completion_length": 16.765625,
      "epoch": 2.567243675099867,
      "grad_norm": 24.02741241455078,
      "kl": 3.2578125,
      "learning_rate": 3.573333333333333e-07,
      "loss": 0.2883,
      "reward": 0.6161408722400665,
      "reward_std": 0.13259492069482803,
      "reward_std_all": 0.16938821971416473,
      "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": 0.5819390416145325,
      "rewards/forensics_strict_edge_reward": 0.27246759831905365,
      "rewards/forensics_tiou_reward": 0.3660464584827423,
      "step": 965
    },
    {
      "completion_length": 25.8125,
      "epoch": 2.569906790945406,
      "grad_norm": 33.0323486328125,
      "kl": 7.015625,
      "learning_rate": 3.5666666666666666e-07,
      "loss": 0.5066,
      "reward": 0.42575471103191376,
      "reward_std": 0.27791690826416016,
      "reward_std_all": 0.315621480345726,
      "rewards-max/forensics_sinkhorn_reward": 0.9638358354568481,
      "rewards-max/forensics_strict_edge_reward": 0.905340164899826,
      "rewards-max/forensics_tiou_reward": 0.9441481530666351,
      "rewards/forensics_sinkhorn_reward": 0.3469352573156357,
      "rewards/forensics_strict_edge_reward": 0.16114410758018494,
      "rewards/forensics_tiou_reward": 0.18646088987588882,
      "step": 966
    },
    {
      "completion_length": 22.703125,
      "epoch": 2.5725699067909455,
      "grad_norm": 21.554784774780273,
      "kl": 4.9375,
      "learning_rate": 3.5599999999999996e-07,
      "loss": 0.4259,
      "reward": 0.5120849311351776,
      "reward_std": 0.2278890609741211,
      "reward_std_all": 0.2611543759703636,
      "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": 0.45460571348667145,
      "rewards/forensics_strict_edge_reward": 0.27462561428546906,
      "rewards/forensics_tiou_reward": 0.3183351680636406,
      "step": 967
    },
    {
      "completion_length": 20.546875,
      "epoch": 2.5752330226364846,
      "grad_norm": 20.996837615966797,
      "kl": 4.203125,
      "learning_rate": 3.553333333333333e-07,
      "loss": 0.2951,
      "reward": 0.6346325278282166,
      "reward_std": 0.19919954985380173,
      "reward_std_all": 0.19258605688810349,
      "rewards-max/forensics_sinkhorn_reward": 0.9878768920898438,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9956101775169373,
      "rewards/forensics_sinkhorn_reward": 0.607667863368988,
      "rewards/forensics_strict_edge_reward": 0.3955223858356476,
      "rewards/forensics_tiou_reward": 0.43952377140522003,
      "step": 968
    },
    {
      "completion_length": 33.8125,
      "epoch": 2.577896138482024,
      "grad_norm": 37.113502502441406,
      "kl": 5.859375,
      "learning_rate": 3.5466666666666667e-07,
      "loss": 0.3554,
      "reward": 0.5848114192485809,
      "reward_std": 0.17468023300170898,
      "reward_std_all": 0.19165321439504623,
      "rewards-max/forensics_sinkhorn_reward": 0.9875998497009277,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9945271015167236,
      "rewards/forensics_sinkhorn_reward": 0.5296418368816376,
      "rewards/forensics_strict_edge_reward": 0.28171801567077637,
      "rewards/forensics_tiou_reward": 0.34164319932460785,
      "step": 969
    },
    {
      "completion_length": 17.359375,
      "epoch": 2.5805592543275635,
      "grad_norm": 21.898977279663086,
      "kl": 3.03125,
      "learning_rate": 3.5399999999999997e-07,
      "loss": 0.3261,
      "reward": 0.6345793306827545,
      "reward_std": 0.1501796916127205,
      "reward_std_all": 0.15657110512256622,
      "rewards-max/forensics_sinkhorn_reward": 0.988058477640152,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.99632927775383,
      "rewards/forensics_sinkhorn_reward": 0.6033391356468201,
      "rewards/forensics_strict_edge_reward": 0.3392440229654312,
      "rewards/forensics_tiou_reward": 0.41169871389865875,
      "step": 970
    },
    {
      "completion_length": 17.4375,
      "epoch": 2.5832223701731025,
      "grad_norm": 21.020505905151367,
      "kl": 2.2734375,
      "learning_rate": 3.533333333333333e-07,
      "loss": 0.2579,
      "reward": 0.6338824033737183,
      "reward_std": 0.18469306826591492,
      "reward_std_all": 0.18246211856603622,
      "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": 0.5819600820541382,
      "rewards/forensics_strict_edge_reward": 0.3511815369129181,
      "rewards/forensics_tiou_reward": 0.41612766683101654,
      "step": 971
    },
    {
      "completion_length": 21.21875,
      "epoch": 2.5858854860186415,
      "grad_norm": 22.744726181030273,
      "kl": 3.6875,
      "learning_rate": 3.526666666666667e-07,
      "loss": 0.3339,
      "reward": 0.5974108576774597,
      "reward_std": 0.21371180564165115,
      "reward_std_all": 0.20964641869068146,
      "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": 0.5138179659843445,
      "rewards/forensics_strict_edge_reward": 0.2501114010810852,
      "rewards/forensics_tiou_reward": 0.3122089356184006,
      "step": 972
    },
    {
      "completion_length": 22.59375,
      "epoch": 2.588548601864181,
      "grad_norm": 19.63388442993164,
      "kl": 2.5625,
      "learning_rate": 3.52e-07,
      "loss": 0.1028,
      "reward": 0.6728506982326508,
      "reward_std": 0.13135068491101265,
      "reward_std_all": 0.1341700181365013,
      "rewards-max/forensics_sinkhorn_reward": 0.953328400850296,
      "rewards-max/forensics_strict_edge_reward": 0.878633052110672,
      "rewards-max/forensics_tiou_reward": 0.9217082262039185,
      "rewards/forensics_sinkhorn_reward": 0.5569965541362762,
      "rewards/forensics_strict_edge_reward": 0.26173578947782516,
      "rewards/forensics_tiou_reward": 0.35704681277275085,
      "step": 973
    },
    {
      "completion_length": 14.984375,
      "epoch": 2.5912117177097205,
      "grad_norm": 19.36612892150879,
      "kl": 1.44921875,
      "learning_rate": 3.5133333333333333e-07,
      "loss": 0.0398,
      "reward": 0.696762353181839,
      "reward_std": 0.08964192494750023,
      "reward_std_all": 0.09811719879508018,
      "rewards-max/forensics_sinkhorn_reward": 0.9887004792690277,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9988567233085632,
      "rewards/forensics_sinkhorn_reward": 0.6601495444774628,
      "rewards/forensics_strict_edge_reward": 0.31733427941799164,
      "rewards/forensics_tiou_reward": 0.4181957244873047,
      "step": 974
    },
    {
      "completion_length": 17.40625,
      "epoch": 2.5938748335552595,
      "grad_norm": 24.840232849121094,
      "kl": 3.75390625,
      "learning_rate": 3.506666666666667e-07,
      "loss": 0.2468,
      "reward": 0.657531350851059,
      "reward_std": 0.15455202013254166,
      "reward_std_all": 0.15093543007969856,
      "rewards-max/forensics_sinkhorn_reward": 0.9839756488800049,
      "rewards-max/forensics_strict_edge_reward": 0.9151731729507446,
      "rewards-max/forensics_tiou_reward": 0.9802879989147186,
      "rewards/forensics_sinkhorn_reward": 0.597781628370285,
      "rewards/forensics_strict_edge_reward": 0.30009353160858154,
      "rewards/forensics_tiou_reward": 0.4039427638053894,
      "step": 975
    },
    {
      "completion_length": 19.34375,
      "epoch": 2.596537949400799,
      "grad_norm": 49.416839599609375,
      "kl": 3.99609375,
      "learning_rate": 3.5e-07,
      "loss": 0.3181,
      "reward": 0.594791829586029,
      "reward_std": 0.19125768914818764,
      "reward_std_all": 0.19546429067850113,
      "rewards-max/forensics_sinkhorn_reward": 0.9862732887268066,
      "rewards-max/forensics_strict_edge_reward": 0.8957232534885406,
      "rewards-max/forensics_tiou_reward": 0.9893298447132111,
      "rewards/forensics_sinkhorn_reward": 0.5189042091369629,
      "rewards/forensics_strict_edge_reward": 0.2552453950047493,
      "rewards/forensics_tiou_reward": 0.31917594373226166,
      "step": 976
    },
    {
      "completion_length": 18.5625,
      "epoch": 2.599201065246338,
      "grad_norm": 18.183815002441406,
      "kl": 1.8359375,
      "learning_rate": 3.4933333333333334e-07,
      "loss": 0.1634,
      "reward": 0.6333062946796417,
      "reward_std": 0.1481107473373413,
      "reward_std_all": 0.15943852066993713,
      "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": 0.5770970284938812,
      "rewards/forensics_strict_edge_reward": 0.30178482830524445,
      "rewards/forensics_tiou_reward": 0.3432397395372391,
      "step": 977
    },
    {
      "completion_length": 16.015625,
      "epoch": 2.6018641810918774,
      "grad_norm": 23.0022029876709,
      "kl": 2.5859375,
      "learning_rate": 3.4866666666666664e-07,
      "loss": 0.1737,
      "reward": 0.6648906171321869,
      "reward_std": 0.1315089501440525,
      "reward_std_all": 0.13929862901568413,
      "rewards-max/forensics_sinkhorn_reward": 0.9666005373001099,
      "rewards-max/forensics_strict_edge_reward": 0.927969753742218,
      "rewards-max/forensics_tiou_reward": 0.9508086740970612,
      "rewards/forensics_sinkhorn_reward": 0.6013728380203247,
      "rewards/forensics_strict_edge_reward": 0.312306746840477,
      "rewards/forensics_tiou_reward": 0.40779298543930054,
      "step": 978
    },
    {
      "completion_length": 24.828125,
      "epoch": 2.604527296937417,
      "grad_norm": 22.125408172607422,
      "kl": 2.10546875,
      "learning_rate": 3.4799999999999994e-07,
      "loss": 0.2048,
      "reward": 0.6634699106216431,
      "reward_std": 0.16970022022724152,
      "reward_std_all": 0.15961098670959473,
      "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": 0.6061393320560455,
      "rewards/forensics_strict_edge_reward": 0.3394125625491142,
      "rewards/forensics_tiou_reward": 0.42957332730293274,
      "step": 979
    },
    {
      "completion_length": 26.703125,
      "epoch": 2.607190412782956,
      "grad_norm": 20.584638595581055,
      "kl": 5.390625,
      "learning_rate": 3.473333333333333e-07,
      "loss": 0.355,
      "reward": 0.5249398648738861,
      "reward_std": 0.22131338715553284,
      "reward_std_all": 0.24203920364379883,
      "rewards-max/forensics_sinkhorn_reward": 0.9713530838489532,
      "rewards-max/forensics_strict_edge_reward": 0.9062341153621674,
      "rewards-max/forensics_tiou_reward": 0.9600236117839813,
      "rewards/forensics_sinkhorn_reward": 0.4554072320461273,
      "rewards/forensics_strict_edge_reward": 0.193511001765728,
      "rewards/forensics_tiou_reward": 0.28204815089702606,
      "step": 980
    },
    {
      "completion_length": 18.53125,
      "epoch": 2.6098535286284954,
      "grad_norm": 34.12763214111328,
      "kl": 6.609375,
      "learning_rate": 3.4666666666666665e-07,
      "loss": 0.4723,
      "reward": 0.4853019565343857,
      "reward_std": 0.2340550422668457,
      "reward_std_all": 0.26748617738485336,
      "rewards-max/forensics_sinkhorn_reward": 0.9876616597175598,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9947633147239685,
      "rewards/forensics_sinkhorn_reward": 0.4166666120290756,
      "rewards/forensics_strict_edge_reward": 0.2277207374572754,
      "rewards/forensics_tiou_reward": 0.28353241086006165,
      "step": 981
    },
    {
      "completion_length": 18.109375,
      "epoch": 2.6125166444740344,
      "grad_norm": 121.26547241210938,
      "kl": 12.671875,
      "learning_rate": 3.4599999999999995e-07,
      "loss": 0.7219,
      "reward": 0.5462482273578644,
      "reward_std": 0.1730096936225891,
      "reward_std_all": 0.2249566689133644,
      "rewards-max/forensics_sinkhorn_reward": 0.985908031463623,
      "rewards-max/forensics_strict_edge_reward": 0.9677534997463226,
      "rewards-max/forensics_tiou_reward": 0.9929154515266418,
      "rewards/forensics_sinkhorn_reward": 0.5586225390434265,
      "rewards/forensics_strict_edge_reward": 0.33762815594673157,
      "rewards/forensics_tiou_reward": 0.45764249563217163,
      "step": 982
    },
    {
      "completion_length": 25.234375,
      "epoch": 2.615179760319574,
      "grad_norm": 20.693490982055664,
      "kl": 4.453125,
      "learning_rate": 3.453333333333333e-07,
      "loss": 0.3194,
      "reward": 0.5075829327106476,
      "reward_std": 0.2929661050438881,
      "reward_std_all": 0.28788090497255325,
      "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": 0.42063668370246887,
      "rewards/forensics_strict_edge_reward": 0.23025570809841156,
      "rewards/forensics_tiou_reward": 0.27550919353961945,
      "step": 983
    },
    {
      "completion_length": 17.796875,
      "epoch": 2.6178428761651134,
      "grad_norm": 27.694711685180664,
      "kl": 3.40625,
      "learning_rate": 3.4466666666666666e-07,
      "loss": 0.2611,
      "reward": 0.627314418554306,
      "reward_std": 0.2195592075586319,
      "reward_std_all": 0.20426569879055023,
      "rewards-max/forensics_sinkhorn_reward": 0.9880991280078888,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9964862167835236,
      "rewards/forensics_sinkhorn_reward": 0.5496058762073517,
      "rewards/forensics_strict_edge_reward": 0.2979883700609207,
      "rewards/forensics_tiou_reward": 0.3592015951871872,
      "step": 984
    },
    {
      "completion_length": 17.96875,
      "epoch": 2.6205059920106524,
      "grad_norm": 16.965333938598633,
      "kl": 3.875,
      "learning_rate": 3.4399999999999996e-07,
      "loss": 0.3419,
      "reward": 0.558991476893425,
      "reward_std": 0.21718795597553253,
      "reward_std_all": 0.23177654296159744,
      "rewards-max/forensics_sinkhorn_reward": 0.987824946641922,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9954081475734711,
      "rewards/forensics_sinkhorn_reward": 0.5223733186721802,
      "rewards/forensics_strict_edge_reward": 0.33408544957637787,
      "rewards/forensics_tiou_reward": 0.41044069826602936,
      "step": 985
    },
    {
      "completion_length": 19.03125,
      "epoch": 2.623169107856192,
      "grad_norm": 25.341754913330078,
      "kl": 4.515625,
      "learning_rate": 3.433333333333333e-07,
      "loss": 0.3734,
      "reward": 0.5107545554637909,
      "reward_std": 0.26456932723522186,
      "reward_std_all": 0.2688908874988556,
      "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": 0.4268815219402313,
      "rewards/forensics_strict_edge_reward": 0.19265614449977875,
      "rewards/forensics_tiou_reward": 0.26185014843940735,
      "step": 986
    },
    {
      "completion_length": 16.65625,
      "epoch": 2.625832223701731,
      "grad_norm": 34.51016616821289,
      "kl": 4.6875,
      "learning_rate": 3.4266666666666666e-07,
      "loss": 0.251,
      "reward": 0.568341076374054,
      "reward_std": 0.17049263417720795,
      "reward_std_all": 0.2013198807835579,
      "rewards-max/forensics_sinkhorn_reward": 0.9693070650100708,
      "rewards-max/forensics_strict_edge_reward": 0.8968285918235779,
      "rewards-max/forensics_tiou_reward": 0.9542346298694611,
      "rewards/forensics_sinkhorn_reward": 0.4801778793334961,
      "rewards/forensics_strict_edge_reward": 0.19954492151737213,
      "rewards/forensics_tiou_reward": 0.28204162418842316,
      "step": 987
    },
    {
      "completion_length": 19.34375,
      "epoch": 2.6284953395472703,
      "grad_norm": 33.15574645996094,
      "kl": 5.4609375,
      "learning_rate": 3.42e-07,
      "loss": 0.4222,
      "reward": 0.5311962366104126,
      "reward_std": 0.22471386939287186,
      "reward_std_all": 0.2603512108325958,
      "rewards-max/forensics_sinkhorn_reward": 0.9769553542137146,
      "rewards-max/forensics_strict_edge_reward": 0.878633052110672,
      "rewards-max/forensics_tiou_reward": 0.9712379276752472,
      "rewards/forensics_sinkhorn_reward": 0.48534899950027466,
      "rewards/forensics_strict_edge_reward": 0.22568600624799728,
      "rewards/forensics_tiou_reward": 0.31213144212961197,
      "step": 988
    },
    {
      "completion_length": 18.859375,
      "epoch": 2.63115845539281,
      "grad_norm": 25.318965911865234,
      "kl": 4.96875,
      "learning_rate": 3.413333333333333e-07,
      "loss": 0.1972,
      "reward": 0.6164030134677887,
      "reward_std": 0.18240153789520264,
      "reward_std_all": 0.18603180348873138,
      "rewards-max/forensics_sinkhorn_reward": 0.9879285991191864,
      "rewards-max/forensics_strict_edge_reward": 0.9364311099052429,
      "rewards-max/forensics_tiou_reward": 0.9958139359951019,
      "rewards/forensics_sinkhorn_reward": 0.5594877004623413,
      "rewards/forensics_strict_edge_reward": 0.2924966812133789,
      "rewards/forensics_tiou_reward": 0.37354467809200287,
      "step": 989
    },
    {
      "completion_length": 17.78125,
      "epoch": 2.633821571238349,
      "grad_norm": 24.46566390991211,
      "kl": 2.9375,
      "learning_rate": 3.4066666666666667e-07,
      "loss": 0.2657,
      "reward": 0.5815741419792175,
      "reward_std": 0.17860906571149826,
      "reward_std_all": 0.19760727882385254,
      "rewards-max/forensics_sinkhorn_reward": 0.9860047698020935,
      "rewards-max/forensics_strict_edge_reward": 0.9185642302036285,
      "rewards-max/forensics_tiou_reward": 0.9882873594760895,
      "rewards/forensics_sinkhorn_reward": 0.5039478093385696,
      "rewards/forensics_strict_edge_reward": 0.19440697133541107,
      "rewards/forensics_tiou_reward": 0.28045927733182907,
      "step": 990
    },
    {
      "completion_length": 15.78125,
      "epoch": 2.6364846870838883,
      "grad_norm": 22.527597427368164,
      "kl": 1.76171875,
      "learning_rate": 3.4000000000000003e-07,
      "loss": 0.0465,
      "reward": 0.6394820511341095,
      "reward_std": 0.11293458193540573,
      "reward_std_all": 0.13776932656764984,
      "rewards-max/forensics_sinkhorn_reward": 0.949497789144516,
      "rewards-max/forensics_strict_edge_reward": 0.8490433394908905,
      "rewards-max/forensics_tiou_reward": 0.9122479557991028,
      "rewards/forensics_sinkhorn_reward": 0.502370148897171,
      "rewards/forensics_strict_edge_reward": 0.11909675225615501,
      "rewards/forensics_tiou_reward": 0.19182507693767548,
      "step": 991
    },
    {
      "completion_length": 17.6875,
      "epoch": 2.6391478029294273,
      "grad_norm": 21.07198143005371,
      "kl": 2.37890625,
      "learning_rate": 3.3933333333333333e-07,
      "loss": 0.173,
      "reward": 0.6560436189174652,
      "reward_std": 0.175118125975132,
      "reward_std_all": 0.1678263396024704,
      "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": 0.5897563695907593,
      "rewards/forensics_strict_edge_reward": 0.3136969655752182,
      "rewards/forensics_tiou_reward": 0.3632090091705322,
      "step": 992
    },
    {
      "completion_length": 17.671875,
      "epoch": 2.6418109187749668,
      "grad_norm": 18.55877113342285,
      "kl": 2.546875,
      "learning_rate": 3.386666666666667e-07,
      "loss": 0.2107,
      "reward": 0.6109883785247803,
      "reward_std": 0.22491052746772766,
      "reward_std_all": 0.2185748815536499,
      "rewards-max/forensics_sinkhorn_reward": 0.9884915947914124,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9980331659317017,
      "rewards/forensics_sinkhorn_reward": 0.5742727220058441,
      "rewards/forensics_strict_edge_reward": 0.318000927567482,
      "rewards/forensics_tiou_reward": 0.4255036413669586,
      "step": 993
    },
    {
      "completion_length": 16.875,
      "epoch": 2.6444740346205062,
      "grad_norm": 27.57160758972168,
      "kl": 2.328125,
      "learning_rate": 3.38e-07,
      "loss": 0.3105,
      "reward": 0.633724182844162,
      "reward_std": 0.15189456194639206,
      "reward_std_all": 0.15992669761180878,
      "rewards-max/forensics_sinkhorn_reward": 0.9883638322353363,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9975321292877197,
      "rewards/forensics_sinkhorn_reward": 0.5414018332958221,
      "rewards/forensics_strict_edge_reward": 0.2817170023918152,
      "rewards/forensics_tiou_reward": 0.3352614939212799,
      "step": 994
    },
    {
      "completion_length": 16.125,
      "epoch": 2.6471371504660453,
      "grad_norm": 21.002071380615234,
      "kl": 1.67578125,
      "learning_rate": 3.373333333333333e-07,
      "loss": 0.1536,
      "reward": 0.6356655657291412,
      "reward_std": 0.12788988277316093,
      "reward_std_all": 0.14069438725709915,
      "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": 0.5449620187282562,
      "rewards/forensics_strict_edge_reward": 0.25170014798641205,
      "rewards/forensics_tiou_reward": 0.2960161417722702,
      "step": 995
    },
    {
      "completion_length": 16.171875,
      "epoch": 2.6498002663115847,
      "grad_norm": 20.705810546875,
      "kl": 2.50390625,
      "learning_rate": 3.3666666666666664e-07,
      "loss": 0.152,
      "reward": 0.6318834722042084,
      "reward_std": 0.1389530785381794,
      "reward_std_all": 0.15134089440107346,
      "rewards-max/forensics_sinkhorn_reward": 0.9649860858917236,
      "rewards-max/forensics_strict_edge_reward": 0.9048932790756226,
      "rewards-max/forensics_tiou_reward": 0.9459350407123566,
      "rewards/forensics_sinkhorn_reward": 0.5512532889842987,
      "rewards/forensics_strict_edge_reward": 0.24191498756408691,
      "rewards/forensics_tiou_reward": 0.35750439763069153,
      "step": 996
    },
    {
      "completion_length": 18.65625,
      "epoch": 2.6524633821571237,
      "grad_norm": 21.19947052001953,
      "kl": 3.4375,
      "learning_rate": 3.36e-07,
      "loss": 0.2372,
      "reward": 0.5942523777484894,
      "reward_std": 0.17271852493286133,
      "reward_std_all": 0.1930708959698677,
      "rewards-max/forensics_sinkhorn_reward": 0.9743748903274536,
      "rewards-max/forensics_strict_edge_reward": 0.947419673204422,
      "rewards-max/forensics_tiou_reward": 0.9666002690792084,
      "rewards/forensics_sinkhorn_reward": 0.5123599618673325,
      "rewards/forensics_strict_edge_reward": 0.2058131992816925,
      "rewards/forensics_tiou_reward": 0.2603670209646225,
      "step": 997
    },
    {
      "completion_length": 16.1875,
      "epoch": 2.655126498002663,
      "grad_norm": 25.852903366088867,
      "kl": 2.67578125,
      "learning_rate": 3.353333333333333e-07,
      "loss": 0.2945,
      "reward": 0.6579087376594543,
      "reward_std": 0.12461335584521294,
      "reward_std_all": 0.1400173455476761,
      "rewards-max/forensics_sinkhorn_reward": 0.9883792102336884,
      "rewards-max/forensics_strict_edge_reward": 0.9890114367008209,
      "rewards-max/forensics_tiou_reward": 0.9975903630256653,
      "rewards/forensics_sinkhorn_reward": 0.6506794393062592,
      "rewards/forensics_strict_edge_reward": 0.4230991452932358,
      "rewards/forensics_tiou_reward": 0.5002899467945099,
      "step": 998
    },
    {
      "completion_length": 15.4375,
      "epoch": 2.6577896138482027,
      "grad_norm": 23.547916412353516,
      "kl": 2.921875,
      "learning_rate": 3.3466666666666665e-07,
      "loss": 0.2707,
      "reward": 0.6702880263328552,
      "reward_std": 0.13456280902028084,
      "reward_std_all": 0.13682377338409424,
      "rewards-max/forensics_sinkhorn_reward": 0.9886549413204193,
      "rewards-max/forensics_strict_edge_reward": 0.9890114367008209,
      "rewards-max/forensics_tiou_reward": 0.9986781179904938,
      "rewards/forensics_sinkhorn_reward": 0.6636289358139038,
      "rewards/forensics_strict_edge_reward": 0.38189779222011566,
      "rewards/forensics_tiou_reward": 0.47860652208328247,
      "step": 999
    },
    {
      "completion_length": 19.84375,
      "epoch": 2.6604527296937417,
      "grad_norm": 20.50514793395996,
      "kl": 5.28125,
      "learning_rate": 3.34e-07,
      "loss": 0.4381,
      "reward": 0.5595229268074036,
      "reward_std": 0.25608667731285095,
      "reward_std_all": 0.25778740644454956,
      "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": 0.520432859659195,
      "rewards/forensics_strict_edge_reward": 0.34305988252162933,
      "rewards/forensics_tiou_reward": 0.417823389172554,
      "step": 1000
    },
    {
      "completion_length": 18.109375,
      "epoch": 2.6631158455392807,
      "grad_norm": 40.09211730957031,
      "kl": 8.4375,
      "learning_rate": 3.333333333333333e-07,
      "loss": 0.5568,
      "reward": 0.5281719863414764,
      "reward_std": 0.28320586681365967,
      "reward_std_all": 0.2887382060289383,
      "rewards-max/forensics_sinkhorn_reward": 0.9877223968505859,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9950033128261566,
      "rewards/forensics_sinkhorn_reward": 0.48127396404743195,
      "rewards/forensics_strict_edge_reward": 0.2940015122294426,
      "rewards/forensics_tiou_reward": 0.3621193766593933,
      "step": 1001
    },
    {
      "completion_length": 17.375,
      "epoch": 2.66577896138482,
      "grad_norm": 19.508996963500977,
      "kl": 3.5625,
      "learning_rate": 3.3266666666666665e-07,
      "loss": 0.44,
      "reward": 0.5649068057537079,
      "reward_std": 0.19556701183319092,
      "reward_std_all": 0.21832312643527985,
      "rewards-max/forensics_sinkhorn_reward": 0.9884622097015381,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9979189336299896,
      "rewards/forensics_sinkhorn_reward": 0.5400238931179047,
      "rewards/forensics_strict_edge_reward": 0.3248134106397629,
      "rewards/forensics_tiou_reward": 0.4029556214809418,
      "step": 1002
    },
    {
      "completion_length": 26.6875,
      "epoch": 2.6684420772303596,
      "grad_norm": 26.737688064575195,
      "kl": 7.5,
      "learning_rate": 3.32e-07,
      "loss": 0.4996,
      "reward": 0.5253798812627792,
      "reward_std": 0.24120549857616425,
      "reward_std_all": 0.26206814497709274,
      "rewards-max/forensics_sinkhorn_reward": 0.9882791042327881,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9971975684165955,
      "rewards/forensics_sinkhorn_reward": 0.4304018020629883,
      "rewards/forensics_strict_edge_reward": 0.13933081924915314,
      "rewards/forensics_tiou_reward": 0.2513537257909775,
      "step": 1003
    },
    {
      "completion_length": 16.875,
      "epoch": 2.6711051930758987,
      "grad_norm": 19.336904525756836,
      "kl": 4.546875,
      "learning_rate": 3.313333333333333e-07,
      "loss": 0.4145,
      "reward": 0.5868980288505554,
      "reward_std": 0.24242620170116425,
      "reward_std_all": 0.23827390372753143,
      "rewards-max/forensics_sinkhorn_reward": 0.9879181087017059,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9957726299762726,
      "rewards/forensics_sinkhorn_reward": 0.5684556365013123,
      "rewards/forensics_strict_edge_reward": 0.3183109164237976,
      "rewards/forensics_tiou_reward": 0.4547074884176254,
      "step": 1004
    },
    {
      "completion_length": 17.109375,
      "epoch": 2.673768308921438,
      "grad_norm": 27.079265594482422,
      "kl": 4.25,
      "learning_rate": 3.3066666666666666e-07,
      "loss": 0.3042,
      "reward": 0.5609697103500366,
      "reward_std": 0.1912839710712433,
      "reward_std_all": 0.22055093199014664,
      "rewards-max/forensics_sinkhorn_reward": 0.987817645072937,
      "rewards-max/forensics_strict_edge_reward": 0.9565287232398987,
      "rewards-max/forensics_tiou_reward": 0.9953833818435669,
      "rewards/forensics_sinkhorn_reward": 0.541763424873352,
      "rewards/forensics_strict_edge_reward": 0.294911190867424,
      "rewards/forensics_tiou_reward": 0.3511526882648468,
      "step": 1005
    },
    {
      "completion_length": 17.40625,
      "epoch": 2.676431424766977,
      "grad_norm": 23.503215789794922,
      "kl": 5.578125,
      "learning_rate": 3.3e-07,
      "loss": 0.383,
      "reward": 0.6172707676887512,
      "reward_std": 0.1631152331829071,
      "reward_std_all": 0.17310486733913422,
      "rewards-max/forensics_sinkhorn_reward": 0.9883091449737549,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.997315913438797,
      "rewards/forensics_sinkhorn_reward": 0.5902158319950104,
      "rewards/forensics_strict_edge_reward": 0.36411748826503754,
      "rewards/forensics_tiou_reward": 0.42987485229969025,
      "step": 1006
    },
    {
      "completion_length": 17.890625,
      "epoch": 2.6790945406125166,
      "grad_norm": 29.81661605834961,
      "kl": 5.65625,
      "learning_rate": 3.293333333333333e-07,
      "loss": 0.2931,
      "reward": 0.5626922249794006,
      "reward_std": 0.2728344649076462,
      "reward_std_all": 0.26335541903972626,
      "rewards-max/forensics_sinkhorn_reward": 0.9874255359172821,
      "rewards-max/forensics_strict_edge_reward": 0.9565287232398987,
      "rewards-max/forensics_tiou_reward": 0.9938311874866486,
      "rewards/forensics_sinkhorn_reward": 0.4941885769367218,
      "rewards/forensics_strict_edge_reward": 0.25381677597761154,
      "rewards/forensics_tiou_reward": 0.33296234905719757,
      "step": 1007
    },
    {
      "completion_length": 18.484375,
      "epoch": 2.681757656458056,
      "grad_norm": 45.73075866699219,
      "kl": 7.875,
      "learning_rate": 3.2866666666666667e-07,
      "loss": 0.4662,
      "reward": 0.6229707300662994,
      "reward_std": 0.20437614619731903,
      "reward_std_all": 0.19345825910568237,
      "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": 0.5997944176197052,
      "rewards/forensics_strict_edge_reward": 0.34532496333122253,
      "rewards/forensics_tiou_reward": 0.4425206780433655,
      "step": 1008
    },
    {
      "completion_length": 16.65625,
      "epoch": 2.684420772303595,
      "grad_norm": 18.927392959594727,
      "kl": 1.6015625,
      "learning_rate": 3.28e-07,
      "loss": -0.0197,
      "reward": 0.6424368619918823,
      "reward_std": 0.08866832032799721,
      "reward_std_all": 0.1228192001581192,
      "rewards-max/forensics_sinkhorn_reward": 0.9882029294967651,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9968973696231842,
      "rewards/forensics_sinkhorn_reward": 0.5898885875940323,
      "rewards/forensics_strict_edge_reward": 0.24630697071552277,
      "rewards/forensics_tiou_reward": 0.35544419288635254,
      "step": 1009
    },
    {
      "completion_length": 16.28125,
      "epoch": 2.6870838881491346,
      "grad_norm": 21.22551727294922,
      "kl": 2.734375,
      "learning_rate": 3.2733333333333327e-07,
      "loss": 0.2604,
      "reward": 0.5944193005561829,
      "reward_std": 0.20758849382400513,
      "reward_std_all": 0.208486407995224,
      "rewards-max/forensics_sinkhorn_reward": 0.9882006049156189,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9968890249729156,
      "rewards/forensics_sinkhorn_reward": 0.5612726211547852,
      "rewards/forensics_strict_edge_reward": 0.332470566034317,
      "rewards/forensics_tiou_reward": 0.3874972462654114,
      "step": 1010
    },
    {
      "completion_length": 16.84375,
      "epoch": 2.6897470039946736,
      "grad_norm": 19.366443634033203,
      "kl": 2.28515625,
      "learning_rate": 3.2666666666666663e-07,
      "loss": 0.2061,
      "reward": 0.7132680714130402,
      "reward_std": 0.09821391478180885,
      "reward_std_all": 0.09769555181264877,
      "rewards-max/forensics_sinkhorn_reward": 0.9847948253154755,
      "rewards-max/forensics_strict_edge_reward": 0.9782643616199493,
      "rewards-max/forensics_tiou_reward": 0.9904754757881165,
      "rewards/forensics_sinkhorn_reward": 0.6407710611820221,
      "rewards/forensics_strict_edge_reward": 0.36574114859104156,
      "rewards/forensics_tiou_reward": 0.48843321204185486,
      "step": 1011
    },
    {
      "completion_length": 16.234375,
      "epoch": 2.692410119840213,
      "grad_norm": 22.615217208862305,
      "kl": 1.455078125,
      "learning_rate": 3.26e-07,
      "loss": 0.117,
      "reward": 0.6381663084030151,
      "reward_std": 0.13378691673278809,
      "reward_std_all": 0.15127623826265335,
      "rewards-max/forensics_sinkhorn_reward": 0.9874637722969055,
      "rewards-max/forensics_strict_edge_reward": 0.9565287232398987,
      "rewards-max/forensics_tiou_reward": 0.9939818978309631,
      "rewards/forensics_sinkhorn_reward": 0.555330753326416,
      "rewards/forensics_strict_edge_reward": 0.2325618490576744,
      "rewards/forensics_tiou_reward": 0.30209725350141525,
      "step": 1012
    },
    {
      "completion_length": 19.421875,
      "epoch": 2.6950732356857525,
      "grad_norm": 21.25154685974121,
      "kl": 3.6484375,
      "learning_rate": 3.253333333333333e-07,
      "loss": 0.1972,
      "reward": 0.5624141097068787,
      "reward_std": 0.09482398442924023,
      "reward_std_all": 0.1584712266921997,
      "rewards-max/forensics_sinkhorn_reward": 0.9643326103687286,
      "rewards-max/forensics_strict_edge_reward": 0.8582656681537628,
      "rewards-max/forensics_tiou_reward": 0.9457189440727234,
      "rewards/forensics_sinkhorn_reward": 0.4688689708709717,
      "rewards/forensics_strict_edge_reward": 0.1922174207866192,
      "rewards/forensics_tiou_reward": 0.2799508273601532,
      "step": 1013
    },
    {
      "completion_length": 19.421875,
      "epoch": 2.6977363515312915,
      "grad_norm": 19.907394409179688,
      "kl": 3.890625,
      "learning_rate": 3.2466666666666664e-07,
      "loss": 0.3109,
      "reward": 0.5610590577125549,
      "reward_std": 0.238112673163414,
      "reward_std_all": 0.23418571054935455,
      "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": 0.48982518911361694,
      "rewards/forensics_strict_edge_reward": 0.25396785140037537,
      "rewards/forensics_tiou_reward": 0.34789566695690155,
      "step": 1014
    },
    {
      "completion_length": 16.078125,
      "epoch": 2.700399467376831,
      "grad_norm": 23.00149917602539,
      "kl": 3.7578125,
      "learning_rate": 3.24e-07,
      "loss": 0.2387,
      "reward": 0.5926361978054047,
      "reward_std": 0.17907850444316864,
      "reward_std_all": 0.20017977058887482,
      "rewards-max/forensics_sinkhorn_reward": 0.9878503680229187,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9955056309700012,
      "rewards/forensics_sinkhorn_reward": 0.5543049573898315,
      "rewards/forensics_strict_edge_reward": 0.29930077493190765,
      "rewards/forensics_tiou_reward": 0.38850070536136627,
      "step": 1015
    },
    {
      "completion_length": 17.03125,
      "epoch": 2.70306258322237,
      "grad_norm": 19.99647331237793,
      "kl": 4.171875,
      "learning_rate": 3.233333333333333e-07,
      "loss": 0.2201,
      "reward": 0.48316243290901184,
      "reward_std": 0.18898489326238632,
      "reward_std_all": 0.24505406618118286,
      "rewards-max/forensics_sinkhorn_reward": 0.9223892092704773,
      "rewards-max/forensics_strict_edge_reward": 0.7349729239940643,
      "rewards-max/forensics_tiou_reward": 0.8507684767246246,
      "rewards/forensics_sinkhorn_reward": 0.3866529017686844,
      "rewards/forensics_strict_edge_reward": 0.1374754086136818,
      "rewards/forensics_tiou_reward": 0.1652403585612774,
      "step": 1016
    },
    {
      "completion_length": 15.828125,
      "epoch": 2.7057256990679095,
      "grad_norm": 28.33995246887207,
      "kl": 2.55078125,
      "learning_rate": 3.2266666666666664e-07,
      "loss": 0.1437,
      "reward": 0.646063357591629,
      "reward_std": 0.13009704276919365,
      "reward_std_all": 0.14383096620440483,
      "rewards-max/forensics_sinkhorn_reward": 0.9881992936134338,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9968840181827545,
      "rewards/forensics_sinkhorn_reward": 0.5829273760318756,
      "rewards/forensics_strict_edge_reward": 0.2663102447986603,
      "rewards/forensics_tiou_reward": 0.37508074939250946,
      "step": 1017
    },
    {
      "completion_length": 21.828125,
      "epoch": 2.708388814913449,
      "grad_norm": 26.410757064819336,
      "kl": 6.671875,
      "learning_rate": 3.22e-07,
      "loss": 0.5512,
      "reward": 0.5179822891950607,
      "reward_std": 0.22889189422130585,
      "reward_std_all": 0.2568857967853546,
      "rewards-max/forensics_sinkhorn_reward": 0.9852629899978638,
      "rewards-max/forensics_strict_edge_reward": 0.8850063383579254,
      "rewards-max/forensics_tiou_reward": 0.985315591096878,
      "rewards/forensics_sinkhorn_reward": 0.40589991211891174,
      "rewards/forensics_strict_edge_reward": 0.15688928216695786,
      "rewards/forensics_tiou_reward": 0.23479049652814865,
      "step": 1018
    },
    {
      "completion_length": 26.90625,
      "epoch": 2.711051930758988,
      "grad_norm": 31.4025821685791,
      "kl": 5.546875,
      "learning_rate": 3.2133333333333335e-07,
      "loss": 0.3462,
      "reward": 0.594425767660141,
      "reward_std": 0.18498003482818604,
      "reward_std_all": 0.20351846516132355,
      "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": 0.5973374545574188,
      "rewards/forensics_strict_edge_reward": 0.37333518266677856,
      "rewards/forensics_tiou_reward": 0.4435143321752548,
      "step": 1019
    },
    {
      "completion_length": 17.28125,
      "epoch": 2.7137150466045274,
      "grad_norm": 27.555208206176758,
      "kl": 4.9296875,
      "learning_rate": 3.2066666666666665e-07,
      "loss": 0.3301,
      "reward": 0.5698009431362152,
      "reward_std": 0.13298543356359005,
      "reward_std_all": 0.18444612249732018,
      "rewards-max/forensics_sinkhorn_reward": 0.9866272509098053,
      "rewards-max/forensics_strict_edge_reward": 0.9464850425720215,
      "rewards-max/forensics_tiou_reward": 0.9907185435295105,
      "rewards/forensics_sinkhorn_reward": 0.5331015288829803,
      "rewards/forensics_strict_edge_reward": 0.3106937110424042,
      "rewards/forensics_tiou_reward": 0.3987128287553787,
      "step": 1020
    },
    {
      "completion_length": 19.984375,
      "epoch": 2.7163781624500665,
      "grad_norm": 31.448410034179688,
      "kl": 6.734375,
      "learning_rate": 3.2e-07,
      "loss": 0.4901,
      "reward": 0.5691415667533875,
      "reward_std": 0.17682337760925293,
      "reward_std_all": 0.2041592001914978,
      "rewards-max/forensics_sinkhorn_reward": 0.9887852668762207,
      "rewards-max/forensics_strict_edge_reward": 0.9890114367008209,
      "rewards-max/forensics_tiou_reward": 0.9991912543773651,
      "rewards/forensics_sinkhorn_reward": 0.4932214468717575,
      "rewards/forensics_strict_edge_reward": 0.25110626220703125,
      "rewards/forensics_tiou_reward": 0.3087974190711975,
      "step": 1021
    },
    {
      "completion_length": 18.09375,
      "epoch": 2.719041278295606,
      "grad_norm": 31.102602005004883,
      "kl": 4.046875,
      "learning_rate": 3.1933333333333336e-07,
      "loss": 0.4226,
      "reward": 0.6413381695747375,
      "reward_std": 0.21366830170154572,
      "reward_std_all": 0.197242833673954,
      "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": 0.6592726111412048,
      "rewards/forensics_strict_edge_reward": 0.4658883512020111,
      "rewards/forensics_tiou_reward": 0.5331798791885376,
      "step": 1022
    },
    {
      "completion_length": 17.359375,
      "epoch": 2.7217043941411454,
      "grad_norm": 27.9902400970459,
      "kl": 3.8515625,
      "learning_rate": 3.1866666666666666e-07,
      "loss": 0.3322,
      "reward": 0.606868177652359,
      "reward_std": 0.26731380820274353,
      "reward_std_all": 0.2453765869140625,
      "rewards-max/forensics_sinkhorn_reward": 0.9887278974056244,
      "rewards-max/forensics_strict_edge_reward": 0.9890114367008209,
      "rewards-max/forensics_tiou_reward": 0.9989653527736664,
      "rewards/forensics_sinkhorn_reward": 0.5588154196739197,
      "rewards/forensics_strict_edge_reward": 0.34020985662937164,
      "rewards/forensics_tiou_reward": 0.4058859050273895,
      "step": 1023
    },
    {
      "completion_length": 20.8125,
      "epoch": 2.7243675099866844,
      "grad_norm": 33.365989685058594,
      "kl": 5.984375,
      "learning_rate": 3.18e-07,
      "loss": 0.4235,
      "reward": 0.5381612181663513,
      "reward_std": 0.2317260429263115,
      "reward_std_all": 0.2427019327878952,
      "rewards-max/forensics_sinkhorn_reward": 0.9800669550895691,
      "rewards-max/forensics_strict_edge_reward": 0.8168056011199951,
      "rewards-max/forensics_tiou_reward": 0.9786489605903625,
      "rewards/forensics_sinkhorn_reward": 0.4747768044471741,
      "rewards/forensics_strict_edge_reward": 0.23193292319774628,
      "rewards/forensics_tiou_reward": 0.30865539610385895,
      "step": 1024
    },
    {
      "completion_length": 23.484375,
      "epoch": 2.7270306258322234,
      "grad_norm": 33.77571105957031,
      "kl": 7.8125,
      "learning_rate": 3.173333333333333e-07,
      "loss": 0.56,
      "reward": 0.4779488295316696,
      "reward_std": 0.24723006784915924,
      "reward_std_all": 0.28945451974868774,
      "rewards-max/forensics_sinkhorn_reward": 0.9854584038257599,
      "rewards-max/forensics_strict_edge_reward": 0.9169811904430389,
      "rewards-max/forensics_tiou_reward": 0.9861594438552856,
      "rewards/forensics_sinkhorn_reward": 0.43398863077163696,
      "rewards/forensics_strict_edge_reward": 0.24211931228637695,
      "rewards/forensics_tiou_reward": 0.3570721447467804,
      "step": 1025
    },
    {
      "completion_length": 18.171875,
      "epoch": 2.729693741677763,
      "grad_norm": 22.85603141784668,
      "kl": 4.1484375,
      "learning_rate": 3.166666666666666e-07,
      "loss": 0.4172,
      "reward": 0.474431574344635,
      "reward_std": 0.20139531046152115,
      "reward_std_all": 0.2589259445667267,
      "rewards-max/forensics_sinkhorn_reward": 0.9723470211029053,
      "rewards-max/forensics_strict_edge_reward": 0.927969753742218,
      "rewards-max/forensics_tiou_reward": 0.9605790078639984,
      "rewards/forensics_sinkhorn_reward": 0.41453468799591064,
      "rewards/forensics_strict_edge_reward": 0.18580512702465057,
      "rewards/forensics_tiou_reward": 0.25161606818437576,
      "step": 1026
    },
    {
      "completion_length": 23.5,
      "epoch": 2.7323568575233024,
      "grad_norm": 25.473846435546875,
      "kl": 8.0625,
      "learning_rate": 3.1599999999999997e-07,
      "loss": 0.5683,
      "reward": 0.4095972031354904,
      "reward_std": 0.265198677778244,
      "reward_std_all": 0.32273101806640625,
      "rewards-max/forensics_sinkhorn_reward": 0.934901088476181,
      "rewards-max/forensics_strict_edge_reward": 0.7346234917640686,
      "rewards-max/forensics_tiou_reward": 0.8718777000904083,
      "rewards/forensics_sinkhorn_reward": 0.3163386434316635,
      "rewards/forensics_strict_edge_reward": 0.14168531447649002,
      "rewards/forensics_tiou_reward": 0.21098360419273376,
      "step": 1027
    },
    {
      "completion_length": 17.578125,
      "epoch": 2.7350199733688414,
      "grad_norm": 20.933040618896484,
      "kl": 3.984375,
      "learning_rate": 3.153333333333333e-07,
      "loss": 0.3553,
      "reward": 0.5246810913085938,
      "reward_std": 0.2539473623037338,
      "reward_std_all": 0.26292969286441803,
      "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": 0.4800705164670944,
      "rewards/forensics_strict_edge_reward": 0.22833620756864548,
      "rewards/forensics_tiou_reward": 0.2926851660013199,
      "step": 1028
    },
    {
      "completion_length": 18.75,
      "epoch": 2.737683089214381,
      "grad_norm": 22.69399070739746,
      "kl": 3.203125,
      "learning_rate": 3.146666666666666e-07,
      "loss": 0.2525,
      "reward": 0.5791943371295929,
      "reward_std": 0.14243097603321075,
      "reward_std_all": 0.18507079035043716,
      "rewards-max/forensics_sinkhorn_reward": 0.9879797399044037,
      "rewards-max/forensics_strict_edge_reward": 0.9003686904907227,
      "rewards-max/forensics_tiou_reward": 0.9960238635540009,
      "rewards/forensics_sinkhorn_reward": 0.5124418139457703,
      "rewards/forensics_strict_edge_reward": 0.24135564267635345,
      "rewards/forensics_tiou_reward": 0.3441605567932129,
      "step": 1029
    },
    {
      "completion_length": 19.84375,
      "epoch": 2.74034620505992,
      "grad_norm": 19.62248992919922,
      "kl": 4.390625,
      "learning_rate": 3.14e-07,
      "loss": 0.2866,
      "reward": 0.5775823891162872,
      "reward_std": 0.17711667716503143,
      "reward_std_all": 0.20491226017475128,
      "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": 0.5535483360290527,
      "rewards/forensics_strict_edge_reward": 0.30377553403377533,
      "rewards/forensics_tiou_reward": 0.3995751291513443,
      "step": 1030
    },
    {
      "completion_length": 18.953125,
      "epoch": 2.7430093209054593,
      "grad_norm": 20.222274780273438,
      "kl": 3.71875,
      "learning_rate": 3.1333333333333333e-07,
      "loss": 0.3712,
      "reward": 0.5740367770195007,
      "reward_std": 0.2409101277589798,
      "reward_std_all": 0.23979908972978592,
      "rewards-max/forensics_sinkhorn_reward": 0.984705775976181,
      "rewards-max/forensics_strict_edge_reward": 0.7718352377414703,
      "rewards-max/forensics_tiou_reward": 0.9831589758396149,
      "rewards/forensics_sinkhorn_reward": 0.5267033278942108,
      "rewards/forensics_strict_edge_reward": 0.26501283049583435,
      "rewards/forensics_tiou_reward": 0.3957826644182205,
      "step": 1031
    },
    {
      "completion_length": 18.125,
      "epoch": 2.745672436750999,
      "grad_norm": 24.543127059936523,
      "kl": 4.5546875,
      "learning_rate": 3.1266666666666663e-07,
      "loss": 0.3298,
      "reward": 0.5632883906364441,
      "reward_std": 0.22920262068510056,
      "reward_std_all": 0.23859363794326782,
      "rewards-max/forensics_sinkhorn_reward": 0.9887688159942627,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9991262555122375,
      "rewards/forensics_sinkhorn_reward": 0.5360576808452606,
      "rewards/forensics_strict_edge_reward": 0.3438776582479477,
      "rewards/forensics_tiou_reward": 0.42918358743190765,
      "step": 1032
    },
    {
      "completion_length": 17.34375,
      "epoch": 2.748335552596538,
      "grad_norm": 19.921236038208008,
      "kl": 4.2578125,
      "learning_rate": 3.12e-07,
      "loss": 0.3212,
      "reward": 0.5946772396564484,
      "reward_std": 0.2245851755142212,
      "reward_std_all": 0.22207391262054443,
      "rewards-max/forensics_sinkhorn_reward": 0.9879234433174133,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9957938492298126,
      "rewards/forensics_sinkhorn_reward": 0.5173454880714417,
      "rewards/forensics_strict_edge_reward": 0.2583125904202461,
      "rewards/forensics_tiou_reward": 0.35030195116996765,
      "step": 1033
    },
    {
      "completion_length": 16.78125,
      "epoch": 2.7509986684420773,
      "grad_norm": 16.76108169555664,
      "kl": 4.1875,
      "learning_rate": 3.1133333333333334e-07,
      "loss": 0.3439,
      "reward": 0.6247385442256927,
      "reward_std": 0.20064878463745117,
      "reward_std_all": 0.1930585354566574,
      "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": 0.566340982913971,
      "rewards/forensics_strict_edge_reward": 0.3053276240825653,
      "rewards/forensics_tiou_reward": 0.3713800311088562,
      "step": 1034
    },
    {
      "completion_length": 17.359375,
      "epoch": 2.7536617842876163,
      "grad_norm": 32.89645004272461,
      "kl": 4.375,
      "learning_rate": 3.1066666666666664e-07,
      "loss": 0.2757,
      "reward": 0.6185849905014038,
      "reward_std": 0.17618072777986526,
      "reward_std_all": 0.1841094195842743,
      "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": 0.6050058007240295,
      "rewards/forensics_strict_edge_reward": 0.35306040942668915,
      "rewards/forensics_tiou_reward": 0.4302259385585785,
      "step": 1035
    },
    {
      "completion_length": 20.609375,
      "epoch": 2.756324900133156,
      "grad_norm": 22.926767349243164,
      "kl": 4.453125,
      "learning_rate": 3.1e-07,
      "loss": 0.3464,
      "reward": 0.6481260359287262,
      "reward_std": 0.18645568192005157,
      "reward_std_all": 0.17762109637260437,
      "rewards-max/forensics_sinkhorn_reward": 0.9878478050231934,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9955023229122162,
      "rewards/forensics_sinkhorn_reward": 0.5995114743709564,
      "rewards/forensics_strict_edge_reward": 0.321733683347702,
      "rewards/forensics_tiou_reward": 0.4415794312953949,
      "step": 1036
    },
    {
      "completion_length": 18.515625,
      "epoch": 2.7589880159786953,
      "grad_norm": 25.62255096435547,
      "kl": 5.6015625,
      "learning_rate": 3.0933333333333335e-07,
      "loss": 0.3454,
      "reward": 0.608783483505249,
      "reward_std": 0.14254872500896454,
      "reward_std_all": 0.1697683334350586,
      "rewards-max/forensics_sinkhorn_reward": 0.9751927256584167,
      "rewards-max/forensics_strict_edge_reward": 0.8663162291049957,
      "rewards-max/forensics_tiou_reward": 0.9677499532699585,
      "rewards/forensics_sinkhorn_reward": 0.48962435126304626,
      "rewards/forensics_strict_edge_reward": 0.21720634400844574,
      "rewards/forensics_tiou_reward": 0.3068820536136627,
      "step": 1037
    },
    {
      "completion_length": 18.390625,
      "epoch": 2.7616511318242343,
      "grad_norm": 19.456026077270508,
      "kl": 3.5859375,
      "learning_rate": 3.0866666666666665e-07,
      "loss": 0.2489,
      "reward": 0.6162535548210144,
      "reward_std": 0.16808956116437912,
      "reward_std_all": 0.17755427956581116,
      "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": 0.5560387074947357,
      "rewards/forensics_strict_edge_reward": 0.3070594370365143,
      "rewards/forensics_tiou_reward": 0.4172498881816864,
      "step": 1038
    },
    {
      "completion_length": 17.6875,
      "epoch": 2.7643142476697737,
      "grad_norm": 21.241819381713867,
      "kl": 5.76171875,
      "learning_rate": 3.08e-07,
      "loss": 0.4309,
      "reward": 0.5806322395801544,
      "reward_std": 0.14797765761613846,
      "reward_std_all": 0.18619034439325333,
      "rewards-max/forensics_sinkhorn_reward": 0.9839045703411102,
      "rewards-max/forensics_strict_edge_reward": 0.8317028284072876,
      "rewards-max/forensics_tiou_reward": 0.9801265597343445,
      "rewards/forensics_sinkhorn_reward": 0.5360687375068665,
      "rewards/forensics_strict_edge_reward": 0.2760425768792629,
      "rewards/forensics_tiou_reward": 0.4358649104833603,
      "step": 1039
    },
    {
      "completion_length": 16.53125,
      "epoch": 2.7669773635153128,
      "grad_norm": 30.177736282348633,
      "kl": 4.578125,
      "learning_rate": 3.0733333333333336e-07,
      "loss": 0.3467,
      "reward": 0.6259885132312775,
      "reward_std": 0.18634655326604843,
      "reward_std_all": 0.18074524402618408,
      "rewards-max/forensics_sinkhorn_reward": 0.9877533614635468,
      "rewards-max/forensics_strict_edge_reward": 0.9464850425720215,
      "rewards-max/forensics_tiou_reward": 0.9951232969760895,
      "rewards/forensics_sinkhorn_reward": 0.5849385857582092,
      "rewards/forensics_strict_edge_reward": 0.31497035920619965,
      "rewards/forensics_tiou_reward": 0.4430147558450699,
      "step": 1040
    },
    {
      "completion_length": 18.09375,
      "epoch": 2.7696404793608522,
      "grad_norm": 25.624404907226562,
      "kl": 4.296875,
      "learning_rate": 3.066666666666666e-07,
      "loss": 0.3699,
      "reward": 0.6125552952289581,
      "reward_std": 0.1783015877008438,
      "reward_std_all": 0.18453272432088852,
      "rewards-max/forensics_sinkhorn_reward": 0.9886096119880676,
      "rewards-max/forensics_strict_edge_reward": 0.9890114367008209,
      "rewards-max/forensics_tiou_reward": 0.9985063076019287,
      "rewards/forensics_sinkhorn_reward": 0.600178599357605,
      "rewards/forensics_strict_edge_reward": 0.35838043689727783,
      "rewards/forensics_tiou_reward": 0.4647342264652252,
      "step": 1041
    },
    {
      "completion_length": 16.625,
      "epoch": 2.7723035952063917,
      "grad_norm": 17.650278091430664,
      "kl": 3.75,
      "learning_rate": 3.0599999999999996e-07,
      "loss": 0.1736,
      "reward": 0.7198411822319031,
      "reward_std": 0.09946294873952866,
      "reward_std_all": 0.09416238218545914,
      "rewards-max/forensics_sinkhorn_reward": 0.9880732297897339,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9963840842247009,
      "rewards/forensics_sinkhorn_reward": 0.6515360772609711,
      "rewards/forensics_strict_edge_reward": 0.3291641101241112,
      "rewards/forensics_tiou_reward": 0.48513880372047424,
      "step": 1042
    },
    {
      "completion_length": 16.3125,
      "epoch": 2.7749667110519307,
      "grad_norm": 34.88309097290039,
      "kl": 5.0625,
      "learning_rate": 3.053333333333333e-07,
      "loss": 0.2719,
      "reward": 0.6058831810951233,
      "reward_std": 0.08757282793521881,
      "reward_std_all": 0.13905063271522522,
      "rewards-max/forensics_sinkhorn_reward": 0.988510012626648,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9981060028076172,
      "rewards/forensics_sinkhorn_reward": 0.5420735478401184,
      "rewards/forensics_strict_edge_reward": 0.21483305096626282,
      "rewards/forensics_tiou_reward": 0.28946173936128616,
      "step": 1043
    },
    {
      "completion_length": 19.125,
      "epoch": 2.77762982689747,
      "grad_norm": 28.804107666015625,
      "kl": 7.5,
      "learning_rate": 3.046666666666666e-07,
      "loss": 0.5174,
      "reward": 0.46791572868824005,
      "reward_std": 0.21149272471666336,
      "reward_std_all": 0.2683921456336975,
      "rewards-max/forensics_sinkhorn_reward": 0.9769952893257141,
      "rewards-max/forensics_strict_edge_reward": 0.8854433596134186,
      "rewards-max/forensics_tiou_reward": 0.9663679003715515,
      "rewards/forensics_sinkhorn_reward": 0.4156824052333832,
      "rewards/forensics_strict_edge_reward": 0.19362949579954147,
      "rewards/forensics_tiou_reward": 0.24532616138458252,
      "step": 1044
    },
    {
      "completion_length": 17.453125,
      "epoch": 2.780292942743009,
      "grad_norm": 38.448158264160156,
      "kl": 6.4375,
      "learning_rate": 3.0399999999999997e-07,
      "loss": 0.3789,
      "reward": 0.522388219833374,
      "reward_std": 0.24054160714149475,
      "reward_std_all": 0.2647530436515808,
      "rewards-max/forensics_sinkhorn_reward": 0.9819359183311462,
      "rewards-max/forensics_strict_edge_reward": 0.927969753742218,
      "rewards-max/forensics_tiou_reward": 0.9851674735546112,
      "rewards/forensics_sinkhorn_reward": 0.4430762827396393,
      "rewards/forensics_strict_edge_reward": 0.16236454248428345,
      "rewards/forensics_tiou_reward": 0.2646181285381317,
      "step": 1045
    },
    {
      "completion_length": 14.84375,
      "epoch": 2.7829560585885487,
      "grad_norm": 17.62283706665039,
      "kl": 3.4453125,
      "learning_rate": 3.033333333333333e-07,
      "loss": 0.1966,
      "reward": 0.7065711915493011,
      "reward_std": 0.10511132329702377,
      "reward_std_all": 0.10671216249465942,
      "rewards-max/forensics_sinkhorn_reward": 0.9887852668762207,
      "rewards-max/forensics_strict_edge_reward": 0.9890114367008209,
      "rewards-max/forensics_tiou_reward": 0.9991912543773651,
      "rewards/forensics_sinkhorn_reward": 0.7212761342525482,
      "rewards/forensics_strict_edge_reward": 0.46914826333522797,
      "rewards/forensics_tiou_reward": 0.5893741846084595,
      "step": 1046
    },
    {
      "completion_length": 20.015625,
      "epoch": 2.785619174434088,
      "grad_norm": 55.18415832519531,
      "kl": 8.390625,
      "learning_rate": 3.026666666666666e-07,
      "loss": 0.4282,
      "reward": 0.5848012566566467,
      "reward_std": 0.1646745428442955,
      "reward_std_all": 0.18884821236133575,
      "rewards-max/forensics_sinkhorn_reward": 0.9857366681098938,
      "rewards-max/forensics_strict_edge_reward": 0.947419673204422,
      "rewards-max/forensics_tiou_reward": 0.9872616827487946,
      "rewards/forensics_sinkhorn_reward": 0.5227250903844833,
      "rewards/forensics_strict_edge_reward": 0.29876627027988434,
      "rewards/forensics_tiou_reward": 0.37035782635211945,
      "step": 1047
    },
    {
      "completion_length": 17.421875,
      "epoch": 2.788282290279627,
      "grad_norm": 23.670061111450195,
      "kl": 3.734375,
      "learning_rate": 3.02e-07,
      "loss": 0.3855,
      "reward": 0.5362724363803864,
      "reward_std": 0.1910834237933159,
      "reward_std_all": 0.2237185761332512,
      "rewards-max/forensics_sinkhorn_reward": 0.9884076714515686,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9977030456066132,
      "rewards/forensics_sinkhorn_reward": 0.4877305030822754,
      "rewards/forensics_strict_edge_reward": 0.26731254905462265,
      "rewards/forensics_tiou_reward": 0.3254142254590988,
      "step": 1048
    },
    {
      "completion_length": 18.1875,
      "epoch": 2.790945406125166,
      "grad_norm": 28.419748306274414,
      "kl": 4.3046875,
      "learning_rate": 3.0133333333333333e-07,
      "loss": 0.342,
      "reward": 0.5259203612804413,
      "reward_std": 0.19851891696453094,
      "reward_std_all": 0.23334262520074844,
      "rewards-max/forensics_sinkhorn_reward": 0.8639156520366669,
      "rewards-max/forensics_strict_edge_reward": 0.652708426117897,
      "rewards-max/forensics_tiou_reward": 0.7411665916442871,
      "rewards/forensics_sinkhorn_reward": 0.43719974160194397,
      "rewards/forensics_strict_edge_reward": 0.16614633053541183,
      "rewards/forensics_tiou_reward": 0.2456807792186737,
      "step": 1049
    },
    {
      "completion_length": 16.8125,
      "epoch": 2.7936085219707056,
      "grad_norm": 19.475719451904297,
      "kl": 3.09375,
      "learning_rate": 3.006666666666667e-07,
      "loss": 0.1874,
      "reward": 0.6223748326301575,
      "reward_std": 0.1638835221529007,
      "reward_std_all": 0.17216582596302032,
      "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": 0.5594819486141205,
      "rewards/forensics_strict_edge_reward": 0.3061932474374771,
      "rewards/forensics_tiou_reward": 0.36982962489128113,
      "step": 1050
    },
    {
      "completion_length": 18.640625,
      "epoch": 2.796271637816245,
      "grad_norm": 23.31951332092285,
      "kl": 3.6640625,
      "learning_rate": 3e-07,
      "loss": 0.3004,
      "reward": 0.6255932152271271,
      "reward_std": 0.17302240431308746,
      "reward_std_all": 0.17927487194538116,
      "rewards-max/forensics_sinkhorn_reward": 0.9826216101646423,
      "rewards-max/forensics_strict_edge_reward": 0.9158510267734528,
      "rewards-max/forensics_tiou_reward": 0.9836993217468262,
      "rewards/forensics_sinkhorn_reward": 0.5213917791843414,
      "rewards/forensics_strict_edge_reward": 0.2726474404335022,
      "rewards/forensics_tiou_reward": 0.33694860339164734,
      "step": 1051
    },
    {
      "completion_length": 21.21875,
      "epoch": 2.798934753661784,
      "grad_norm": 26.956220626831055,
      "kl": 5.671875,
      "learning_rate": 2.9933333333333334e-07,
      "loss": 0.2615,
      "reward": 0.6278142035007477,
      "reward_std": 0.15938815847039223,
      "reward_std_all": 0.17690180987119675,
      "rewards-max/forensics_sinkhorn_reward": 0.973092257976532,
      "rewards-max/forensics_strict_edge_reward": 0.9093653857707977,
      "rewards-max/forensics_tiou_reward": 0.9629129767417908,
      "rewards/forensics_sinkhorn_reward": 0.5304822027683258,
      "rewards/forensics_strict_edge_reward": 0.21458111703395844,
      "rewards/forensics_tiou_reward": 0.37588824331760406,
      "step": 1052
    },
    {
      "completion_length": 17.453125,
      "epoch": 2.8015978695073236,
      "grad_norm": 20.991966247558594,
      "kl": 3.8203125,
      "learning_rate": 2.986666666666667e-07,
      "loss": 0.3036,
      "reward": 0.6403833329677582,
      "reward_std": 0.18854346126317978,
      "reward_std_all": 0.179203599691391,
      "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": 0.5640608370304108,
      "rewards/forensics_strict_edge_reward": 0.33058907091617584,
      "rewards/forensics_tiou_reward": 0.39024315774440765,
      "step": 1053
    },
    {
      "completion_length": 17.1875,
      "epoch": 2.8042609853528626,
      "grad_norm": 71.03580474853516,
      "kl": 4.2734375,
      "learning_rate": 2.98e-07,
      "loss": 0.2641,
      "reward": 0.5855014026165009,
      "reward_std": 0.17441540211439133,
      "reward_std_all": 0.1940215453505516,
      "rewards-max/forensics_sinkhorn_reward": 0.9882925748825073,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9972501695156097,
      "rewards/forensics_sinkhorn_reward": 0.5429301857948303,
      "rewards/forensics_strict_edge_reward": 0.3480624109506607,
      "rewards/forensics_tiou_reward": 0.3994545489549637,
      "step": 1054
    },
    {
      "completion_length": 16.5,
      "epoch": 2.806924101198402,
      "grad_norm": 15.286580085754395,
      "kl": 3.23828125,
      "learning_rate": 2.9733333333333335e-07,
      "loss": 0.3256,
      "reward": 0.63139408826828,
      "reward_std": 0.16099217534065247,
      "reward_std_all": 0.16545937955379486,
      "rewards-max/forensics_sinkhorn_reward": 0.9882822930812836,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.997212678194046,
      "rewards/forensics_sinkhorn_reward": 0.670420229434967,
      "rewards/forensics_strict_edge_reward": 0.41881927847862244,
      "rewards/forensics_tiou_reward": 0.5674024820327759,
      "step": 1055
    },
    {
      "completion_length": 17.296875,
      "epoch": 2.8095872170439415,
      "grad_norm": 22.17440414428711,
      "kl": 4.3671875,
      "learning_rate": 2.966666666666667e-07,
      "loss": 0.3257,
      "reward": 0.5991659462451935,
      "reward_std": 0.1775563284754753,
      "reward_std_all": 0.18587836623191833,
      "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": 0.5579221844673157,
      "rewards/forensics_strict_edge_reward": 0.3168019652366638,
      "rewards/forensics_tiou_reward": 0.37204650044441223,
      "step": 1056
    },
    {
      "completion_length": 20.765625,
      "epoch": 2.8122503328894806,
      "grad_norm": 170.66323852539062,
      "kl": 8.828125,
      "learning_rate": 2.9599999999999995e-07,
      "loss": 0.524,
      "reward": 0.5207710713148117,
      "reward_std": 0.2855720669031143,
      "reward_std_all": 0.2824272960424423,
      "rewards-max/forensics_sinkhorn_reward": 0.9282685220241547,
      "rewards-max/forensics_strict_edge_reward": 0.7406483888626099,
      "rewards-max/forensics_tiou_reward": 0.8672806918621063,
      "rewards/forensics_sinkhorn_reward": 0.4498952329158783,
      "rewards/forensics_strict_edge_reward": 0.1875481065362692,
      "rewards/forensics_tiou_reward": 0.2967991456389427,
      "step": 1057
    },
    {
      "completion_length": 17.125,
      "epoch": 2.81491344873502,
      "grad_norm": 18.829092025756836,
      "kl": 3.09375,
      "learning_rate": 2.953333333333333e-07,
      "loss": 0.331,
      "reward": 0.61943119764328,
      "reward_std": 0.14122141152620316,
      "reward_std_all": 0.16133089363574982,
      "rewards-max/forensics_sinkhorn_reward": 0.9878641068935394,
      "rewards-max/forensics_strict_edge_reward": 0.9567649364471436,
      "rewards-max/forensics_tiou_reward": 0.9955622255802155,
      "rewards/forensics_sinkhorn_reward": 0.5654981732368469,
      "rewards/forensics_strict_edge_reward": 0.33812472224235535,
      "rewards/forensics_tiou_reward": 0.39647604525089264,
      "step": 1058
    },
    {
      "completion_length": 20.78125,
      "epoch": 2.817576564580559,
      "grad_norm": 29.638771057128906,
      "kl": 6.28125,
      "learning_rate": 2.9466666666666666e-07,
      "loss": 0.4061,
      "reward": 0.45328110456466675,
      "reward_std": 0.2611725330352783,
      "reward_std_all": 0.2999432682991028,
      "rewards-max/forensics_sinkhorn_reward": 0.9270788431167603,
      "rewards-max/forensics_strict_edge_reward": 0.7013481706380844,
      "rewards-max/forensics_tiou_reward": 0.8707509934902191,
      "rewards/forensics_sinkhorn_reward": 0.3312750607728958,
      "rewards/forensics_strict_edge_reward": 0.10917709767818451,
      "rewards/forensics_tiou_reward": 0.17792584002017975,
      "step": 1059
    },
    {
      "completion_length": 15.765625,
      "epoch": 2.8202396804260985,
      "grad_norm": 22.511735916137695,
      "kl": 4.4765625,
      "learning_rate": 2.9399999999999996e-07,
      "loss": 0.2618,
      "reward": 0.5501975417137146,
      "reward_std": 0.16267051175236702,
      "reward_std_all": 0.2027089148759842,
      "rewards-max/forensics_sinkhorn_reward": 0.9282194674015045,
      "rewards-max/forensics_strict_edge_reward": 0.7408639490604401,
      "rewards-max/forensics_tiou_reward": 0.8642104864120483,
      "rewards/forensics_sinkhorn_reward": 0.4724678099155426,
      "rewards/forensics_strict_edge_reward": 0.10311657935380936,
      "rewards/forensics_tiou_reward": 0.20198491215705872,
      "step": 1060
    },
    {
      "completion_length": 17.953125,
      "epoch": 2.822902796271638,
      "grad_norm": 27.971654891967773,
      "kl": 3.484375,
      "learning_rate": 2.933333333333333e-07,
      "loss": 0.3665,
      "reward": 0.5999491214752197,
      "reward_std": 0.21291705965995789,
      "reward_std_all": 0.2091819941997528,
      "rewards-max/forensics_sinkhorn_reward": 0.988151878118515,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9966950714588165,
      "rewards/forensics_sinkhorn_reward": 0.5391420871019363,
      "rewards/forensics_strict_edge_reward": 0.327117420732975,
      "rewards/forensics_tiou_reward": 0.38778552412986755,
      "step": 1061
    },
    {
      "completion_length": 16.890625,
      "epoch": 2.825565912117177,
      "grad_norm": 25.714141845703125,
      "kl": 5.921875,
      "learning_rate": 2.9266666666666667e-07,
      "loss": 0.6221,
      "reward": 0.46889206767082214,
      "reward_std": 0.2745334357023239,
      "reward_std_all": 0.30151569843292236,
      "rewards-max/forensics_sinkhorn_reward": 0.9885382950305939,
      "rewards-max/forensics_strict_edge_reward": 0.9890114367008209,
      "rewards-max/forensics_tiou_reward": 0.9982190430164337,
      "rewards/forensics_sinkhorn_reward": 0.4331375062465668,
      "rewards/forensics_strict_edge_reward": 0.2928761839866638,
      "rewards/forensics_tiou_reward": 0.3704896867275238,
      "step": 1062
    },
    {
      "completion_length": 17.046875,
      "epoch": 2.8282290279627165,
      "grad_norm": 18.820253372192383,
      "kl": 2.3203125,
      "learning_rate": 2.9199999999999997e-07,
      "loss": 0.1905,
      "reward": 0.6264047026634216,
      "reward_std": 0.12189895287156105,
      "reward_std_all": 0.15428156405687332,
      "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": 0.6244475245475769,
      "rewards/forensics_strict_edge_reward": 0.3890373557806015,
      "rewards/forensics_tiou_reward": 0.46625231206417084,
      "step": 1063
    },
    {
      "completion_length": 21.28125,
      "epoch": 2.8308921438082555,
      "grad_norm": 22.985687255859375,
      "kl": 4.21875,
      "learning_rate": 2.913333333333333e-07,
      "loss": 0.4048,
      "reward": 0.6121048033237457,
      "reward_std": 0.17017871886491776,
      "reward_std_all": 0.18663611263036728,
      "rewards-max/forensics_sinkhorn_reward": 0.9762095212936401,
      "rewards-max/forensics_strict_edge_reward": 0.8681221902370453,
      "rewards-max/forensics_tiou_reward": 0.9697146117687225,
      "rewards/forensics_sinkhorn_reward": 0.5294895619153976,
      "rewards/forensics_strict_edge_reward": 0.32137978076934814,
      "rewards/forensics_tiou_reward": 0.34397488832473755,
      "step": 1064
    },
    {
      "completion_length": 17.75,
      "epoch": 2.833555259653795,
      "grad_norm": 20.0809268951416,
      "kl": 5.015625,
      "learning_rate": 2.906666666666667e-07,
      "loss": 0.4097,
      "reward": 0.5781943202018738,
      "reward_std": 0.20512986183166504,
      "reward_std_all": 0.22112341970205307,
      "rewards-max/forensics_sinkhorn_reward": 0.9888132214546204,
      "rewards-max/forensics_strict_edge_reward": 0.9782643616199493,
      "rewards-max/forensics_tiou_reward": 0.9993014335632324,
      "rewards/forensics_sinkhorn_reward": 0.5097517520189285,
      "rewards/forensics_strict_edge_reward": 0.3170907199382782,
      "rewards/forensics_tiou_reward": 0.37742720544338226,
      "step": 1065
    },
    {
      "completion_length": 18.0,
      "epoch": 2.8362183754993344,
      "grad_norm": 31.14272689819336,
      "kl": 6.46875,
      "learning_rate": 2.9e-07,
      "loss": 0.5058,
      "reward": 0.43343059718608856,
      "reward_std": 0.18903623521327972,
      "reward_std_all": 0.2696799635887146,
      "rewards-max/forensics_sinkhorn_reward": 0.974458634853363,
      "rewards-max/forensics_strict_edge_reward": 0.9574736058712006,
      "rewards-max/forensics_tiou_reward": 0.9665934145450592,
      "rewards/forensics_sinkhorn_reward": 0.35513240098953247,
      "rewards/forensics_strict_edge_reward": 0.12587253749370575,
      "rewards/forensics_tiou_reward": 0.1839183047413826,
      "step": 1066
    },
    {
      "completion_length": 17.234375,
      "epoch": 2.8388814913448734,
      "grad_norm": 23.51239013671875,
      "kl": 4.6875,
      "learning_rate": 2.8933333333333333e-07,
      "loss": 0.4615,
      "reward": 0.4627160280942917,
      "reward_std": 0.3063098192214966,
      "reward_std_all": 0.3178521394729614,
      "rewards-max/forensics_sinkhorn_reward": 0.9298782646656036,
      "rewards-max/forensics_strict_edge_reward": 0.7821763753890991,
      "rewards-max/forensics_tiou_reward": 0.8708291053771973,
      "rewards/forensics_sinkhorn_reward": 0.37007056176662445,
      "rewards/forensics_strict_edge_reward": 0.22311784327030182,
      "rewards/forensics_tiou_reward": 0.2505973279476166,
      "step": 1067
    },
    {
      "completion_length": 17.90625,
      "epoch": 2.841544607190413,
      "grad_norm": 38.26982116699219,
      "kl": 6.53125,
      "learning_rate": 2.886666666666667e-07,
      "loss": 0.4687,
      "reward": 0.5466630160808563,
      "reward_std": 0.19653594493865967,
      "reward_std_all": 0.2263353019952774,
      "rewards-max/forensics_sinkhorn_reward": 0.9884622097015381,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9979189336299896,
      "rewards/forensics_sinkhorn_reward": 0.5054586231708527,
      "rewards/forensics_strict_edge_reward": 0.2986403554677963,
      "rewards/forensics_tiou_reward": 0.3655594289302826,
      "step": 1068
    },
    {
      "completion_length": 16.953125,
      "epoch": 2.844207723035952,
      "grad_norm": 29.303424835205078,
      "kl": 6.109375,
      "learning_rate": 2.88e-07,
      "loss": 0.4653,
      "reward": 0.5836523473262787,
      "reward_std": 0.20213141292333603,
      "reward_std_all": 0.22022601962089539,
      "rewards-max/forensics_sinkhorn_reward": 0.987596720457077,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.994513988494873,
      "rewards/forensics_sinkhorn_reward": 0.5698108077049255,
      "rewards/forensics_strict_edge_reward": 0.35161587595939636,
      "rewards/forensics_tiou_reward": 0.433418869972229,
      "step": 1069
    },
    {
      "completion_length": 18.578125,
      "epoch": 2.8468708388814914,
      "grad_norm": 17.28324317932129,
      "kl": 5.171875,
      "learning_rate": 2.8733333333333334e-07,
      "loss": 0.5356,
      "reward": 0.5313185751438141,
      "reward_std": 0.23591995239257812,
      "reward_std_all": 0.26274073123931885,
      "rewards-max/forensics_sinkhorn_reward": 0.9885024130344391,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.998076468706131,
      "rewards/forensics_sinkhorn_reward": 0.504901111125946,
      "rewards/forensics_strict_edge_reward": 0.351699136197567,
      "rewards/forensics_tiou_reward": 0.40432240068912506,
      "step": 1070
    },
    {
      "completion_length": 20.015625,
      "epoch": 2.849533954727031,
      "grad_norm": 30.16874122619629,
      "kl": 6.015625,
      "learning_rate": 2.866666666666667e-07,
      "loss": 0.5052,
      "reward": 0.489267498254776,
      "reward_std": 0.2198307067155838,
      "reward_std_all": 0.2625580057501793,
      "rewards-max/forensics_sinkhorn_reward": 0.9583796262741089,
      "rewards-max/forensics_strict_edge_reward": 0.8612285256385803,
      "rewards-max/forensics_tiou_reward": 0.9286702871322632,
      "rewards/forensics_sinkhorn_reward": 0.4069352149963379,
      "rewards/forensics_strict_edge_reward": 0.20829130336642265,
      "rewards/forensics_tiou_reward": 0.2517574727535248,
      "step": 1071
    },
    {
      "completion_length": 16.453125,
      "epoch": 2.85219707057257,
      "grad_norm": 24.29015350341797,
      "kl": 4.1640625,
      "learning_rate": 2.8599999999999994e-07,
      "loss": 0.3686,
      "reward": 0.6044944822788239,
      "reward_std": 0.13777408003807068,
      "reward_std_all": 0.1661781519651413,
      "rewards-max/forensics_sinkhorn_reward": 0.9883546531200409,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9974933862686157,
      "rewards/forensics_sinkhorn_reward": 0.5805241763591766,
      "rewards/forensics_strict_edge_reward": 0.38003383576869965,
      "rewards/forensics_tiou_reward": 0.46839794516563416,
      "step": 1072
    },
    {
      "completion_length": 16.421875,
      "epoch": 2.8548601864181093,
      "grad_norm": 22.736642837524414,
      "kl": 2.48046875,
      "learning_rate": 2.853333333333333e-07,
      "loss": 0.2803,
      "reward": 0.6179606914520264,
      "reward_std": 0.19773312658071518,
      "reward_std_all": 0.19509422034025192,
      "rewards-max/forensics_sinkhorn_reward": 0.9883137345314026,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9973345994949341,
      "rewards/forensics_sinkhorn_reward": 0.5890591442584991,
      "rewards/forensics_strict_edge_reward": 0.3492726683616638,
      "rewards/forensics_tiou_reward": 0.4287940263748169,
      "step": 1073
    },
    {
      "completion_length": 15.65625,
      "epoch": 2.8575233022636484,
      "grad_norm": 25.426895141601562,
      "kl": 1.62109375,
      "learning_rate": 2.8466666666666665e-07,
      "loss": 0.2274,
      "reward": 0.6235240697860718,
      "reward_std": 0.17935273796319962,
      "reward_std_all": 0.17830321192741394,
      "rewards-max/forensics_sinkhorn_reward": 0.9881477653980255,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9966811537742615,
      "rewards/forensics_sinkhorn_reward": 0.5771365761756897,
      "rewards/forensics_strict_edge_reward": 0.3038543164730072,
      "rewards/forensics_tiou_reward": 0.39713598787784576,
      "step": 1074
    },
    {
      "completion_length": 15.859375,
      "epoch": 2.860186418109188,
      "grad_norm": 37.6321907043457,
      "kl": 4.359375,
      "learning_rate": 2.8399999999999995e-07,
      "loss": 0.4721,
      "reward": 0.5311314463615417,
      "reward_std": 0.19724871218204498,
      "reward_std_all": 0.23916032165288925,
      "rewards-max/forensics_sinkhorn_reward": 0.987975001335144,
      "rewards-max/forensics_strict_edge_reward": 0.9464850425720215,
      "rewards-max/forensics_tiou_reward": 0.9959982335567474,
      "rewards/forensics_sinkhorn_reward": 0.5248581767082214,
      "rewards/forensics_strict_edge_reward": 0.3226751834154129,
      "rewards/forensics_tiou_reward": 0.3902021497488022,
      "step": 1075
    },
    {
      "completion_length": 16.125,
      "epoch": 2.8628495339547273,
      "grad_norm": 21.350204467773438,
      "kl": 4.109375,
      "learning_rate": 2.833333333333333e-07,
      "loss": 0.2816,
      "reward": 0.5779348015785217,
      "reward_std": 0.2087726965546608,
      "reward_std_all": 0.2242172807455063,
      "rewards-max/forensics_sinkhorn_reward": 0.9880708158016205,
      "rewards-max/forensics_strict_edge_reward": 0.9567649364471436,
      "rewards-max/forensics_tiou_reward": 0.9963748455047607,
      "rewards/forensics_sinkhorn_reward": 0.5176472663879395,
      "rewards/forensics_strict_edge_reward": 0.365785151720047,
      "rewards/forensics_tiou_reward": 0.3907778710126877,
      "step": 1076
    },
    {
      "completion_length": 17.21875,
      "epoch": 2.8655126498002663,
      "grad_norm": 21.61695671081543,
      "kl": 5.2890625,
      "learning_rate": 2.8266666666666666e-07,
      "loss": 0.4731,
      "reward": 0.5893527567386627,
      "reward_std": 0.22020982205867767,
      "reward_std_all": 0.22120711207389832,
      "rewards-max/forensics_sinkhorn_reward": 0.9862234592437744,
      "rewards-max/forensics_strict_edge_reward": 0.927969753742218,
      "rewards-max/forensics_tiou_reward": 0.9891560077667236,
      "rewards/forensics_sinkhorn_reward": 0.5521864891052246,
      "rewards/forensics_strict_edge_reward": 0.3215511590242386,
      "rewards/forensics_tiou_reward": 0.3918059915304184,
      "step": 1077
    },
    {
      "completion_length": 16.5625,
      "epoch": 2.8681757656458053,
      "grad_norm": 18.145666122436523,
      "kl": 6.203125,
      "learning_rate": 2.8199999999999996e-07,
      "loss": 0.4407,
      "reward": 0.5277194678783417,
      "reward_std": 0.29564371705055237,
      "reward_std_all": 0.28754591941833496,
      "rewards-max/forensics_sinkhorn_reward": 0.906514972448349,
      "rewards-max/forensics_strict_edge_reward": 0.7510668337345123,
      "rewards-max/forensics_tiou_reward": 0.8165319859981537,
      "rewards/forensics_sinkhorn_reward": 0.4343482702970505,
      "rewards/forensics_strict_edge_reward": 0.15164850652217865,
      "rewards/forensics_tiou_reward": 0.2605131044983864,
      "step": 1078
    },
    {
      "completion_length": 17.671875,
      "epoch": 2.870838881491345,
      "grad_norm": 23.436031341552734,
      "kl": 4.3125,
      "learning_rate": 2.813333333333333e-07,
      "loss": 0.3831,
      "reward": 0.5827215611934662,
      "reward_std": 0.29057684540748596,
      "reward_std_all": 0.27048400044441223,
      "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": 0.5302207320928574,
      "rewards/forensics_strict_edge_reward": 0.31538526713848114,
      "rewards/forensics_tiou_reward": 0.3902568966150284,
      "step": 1079
    },
    {
      "completion_length": 20.0625,
      "epoch": 2.8735019973368843,
      "grad_norm": 24.15324592590332,
      "kl": 6.859375,
      "learning_rate": 2.8066666666666667e-07,
      "loss": 0.5714,
      "reward": 0.4908462017774582,
      "reward_std": 0.3005668520927429,
      "reward_std_all": 0.3033493757247925,
      "rewards-max/forensics_sinkhorn_reward": 0.941174328327179,
      "rewards-max/forensics_strict_edge_reward": 0.8698340952396393,
      "rewards-max/forensics_tiou_reward": 0.897095113992691,
      "rewards/forensics_sinkhorn_reward": 0.3691541999578476,
      "rewards/forensics_strict_edge_reward": 0.23252039402723312,
      "rewards/forensics_tiou_reward": 0.2744217664003372,
      "step": 1080
    },
    {
      "completion_length": 17.6875,
      "epoch": 2.8761651131824233,
      "grad_norm": 32.82736587524414,
      "kl": 5.3125,
      "learning_rate": 2.8e-07,
      "loss": 0.5393,
      "reward": 0.43866604566574097,
      "reward_std": 0.2227434515953064,
      "reward_std_all": 0.2907089367508888,
      "rewards-max/forensics_sinkhorn_reward": 0.9873233437538147,
      "rewards-max/forensics_strict_edge_reward": 0.9075756669044495,
      "rewards-max/forensics_tiou_reward": 0.9934396743774414,
      "rewards/forensics_sinkhorn_reward": 0.3593606948852539,
      "rewards/forensics_strict_edge_reward": 0.21283436566591263,
      "rewards/forensics_tiou_reward": 0.24466826021671295,
      "step": 1081
    },
    {
      "completion_length": 16.9375,
      "epoch": 2.8788282290279628,
      "grad_norm": 29.58588409423828,
      "kl": 4.96875,
      "learning_rate": 2.793333333333333e-07,
      "loss": 0.463,
      "reward": 0.603336364030838,
      "reward_std": 0.23971294611692429,
      "reward_std_all": 0.22332027554512024,
      "rewards-max/forensics_sinkhorn_reward": 0.9605041742324829,
      "rewards-max/forensics_strict_edge_reward": 0.8957232534885406,
      "rewards-max/forensics_tiou_reward": 0.9365669786930084,
      "rewards/forensics_sinkhorn_reward": 0.5076261758804321,
      "rewards/forensics_strict_edge_reward": 0.31123916804790497,
      "rewards/forensics_tiou_reward": 0.35712210834026337,
      "step": 1082
    },
    {
      "completion_length": 17.546875,
      "epoch": 2.881491344873502,
      "grad_norm": 29.303028106689453,
      "kl": 4.7734375,
      "learning_rate": 2.786666666666667e-07,
      "loss": 0.3436,
      "reward": 0.5651264637708664,
      "reward_std": 0.25295013189315796,
      "reward_std_all": 0.2453206107020378,
      "rewards-max/forensics_sinkhorn_reward": 0.9888203740119934,
      "rewards-max/forensics_strict_edge_reward": 0.9782643616199493,
      "rewards-max/forensics_tiou_reward": 0.999329537153244,
      "rewards/forensics_sinkhorn_reward": 0.4958341121673584,
      "rewards/forensics_strict_edge_reward": 0.2737615928053856,
      "rewards/forensics_tiou_reward": 0.3296228349208832,
      "step": 1083
    },
    {
      "completion_length": 16.78125,
      "epoch": 2.8841544607190412,
      "grad_norm": 20.744075775146484,
      "kl": 4.7109375,
      "learning_rate": 2.7800000000000003e-07,
      "loss": 0.3879,
      "reward": 0.49358245730400085,
      "reward_std": 0.3021028935909271,
      "reward_std_all": 0.2968679368495941,
      "rewards-max/forensics_sinkhorn_reward": 0.9877997040748596,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9953099489212036,
      "rewards/forensics_sinkhorn_reward": 0.41962067782878876,
      "rewards/forensics_strict_edge_reward": 0.24421383440494537,
      "rewards/forensics_tiou_reward": 0.2825031131505966,
      "step": 1084
    },
    {
      "completion_length": 14.828125,
      "epoch": 2.8868175765645807,
      "grad_norm": 27.21132469177246,
      "kl": 6.875,
      "learning_rate": 2.7733333333333333e-07,
      "loss": 0.5742,
      "reward": 0.5338549315929413,
      "reward_std": 0.26492922753095627,
      "reward_std_all": 0.2735474780201912,
      "rewards-max/forensics_sinkhorn_reward": 0.9319794476032257,
      "rewards-max/forensics_strict_edge_reward": 0.8503920137882233,
      "rewards-max/forensics_tiou_reward": 0.8758655190467834,
      "rewards/forensics_sinkhorn_reward": 0.4516305774450302,
      "rewards/forensics_strict_edge_reward": 0.23054932057857513,
      "rewards/forensics_tiou_reward": 0.30420079082250595,
      "step": 1085
    },
    {
      "completion_length": 17.390625,
      "epoch": 2.8894806924101197,
      "grad_norm": 22.960786819458008,
      "kl": 4.6953125,
      "learning_rate": 2.766666666666667e-07,
      "loss": 0.3598,
      "reward": 0.6208812296390533,
      "reward_std": 0.2002309337258339,
      "reward_std_all": 0.19073202461004257,
      "rewards-max/forensics_sinkhorn_reward": 0.9880847632884979,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9964329600334167,
      "rewards/forensics_sinkhorn_reward": 0.6038813889026642,
      "rewards/forensics_strict_edge_reward": 0.3627103269100189,
      "rewards/forensics_tiou_reward": 0.49353474378585815,
      "step": 1086
    },
    {
      "completion_length": 16.265625,
      "epoch": 2.892143808255659,
      "grad_norm": 15.462639808654785,
      "kl": 6.15625,
      "learning_rate": 2.7600000000000004e-07,
      "loss": 0.4927,
      "reward": 0.5032916814088821,
      "reward_std": 0.22709441184997559,
      "reward_std_all": 0.26631613075733185,
      "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": 0.4793015271425247,
      "rewards/forensics_strict_edge_reward": 0.34527571499347687,
      "rewards/forensics_tiou_reward": 0.380037397146225,
      "step": 1087
    },
    {
      "completion_length": 16.875,
      "epoch": 2.894806924101198,
      "grad_norm": 30.31568717956543,
      "kl": 7.09375,
      "learning_rate": 2.753333333333333e-07,
      "loss": 0.4528,
      "reward": 0.49804365634918213,
      "reward_std": 0.24759210646152496,
      "reward_std_all": 0.27811436355113983,
      "rewards-max/forensics_sinkhorn_reward": 0.9652668833732605,
      "rewards-max/forensics_strict_edge_reward": 0.8490433394908905,
      "rewards-max/forensics_tiou_reward": 0.9422672092914581,
      "rewards/forensics_sinkhorn_reward": 0.43488219380378723,
      "rewards/forensics_strict_edge_reward": 0.20147064328193665,
      "rewards/forensics_tiou_reward": 0.25305265188217163,
      "step": 1088
    },
    {
      "completion_length": 18.265625,
      "epoch": 2.8974700399467377,
      "grad_norm": 22.46388053894043,
      "kl": 5.328125,
      "learning_rate": 2.7466666666666664e-07,
      "loss": 0.3853,
      "reward": 0.5368056893348694,
      "reward_std": 0.25036315619945526,
      "reward_std_all": 0.25498221814632416,
      "rewards-max/forensics_sinkhorn_reward": 0.9882152080535889,
      "rewards-max/forensics_strict_edge_reward": 0.9364311099052429,
      "rewards-max/forensics_tiou_reward": 0.9969439506530762,
      "rewards/forensics_sinkhorn_reward": 0.44663751125335693,
      "rewards/forensics_strict_edge_reward": 0.20339153707027435,
      "rewards/forensics_tiou_reward": 0.26776690781116486,
      "step": 1089
    },
    {
      "completion_length": 16.21875,
      "epoch": 2.900133155792277,
      "grad_norm": 24.453805923461914,
      "kl": 3.6953125,
      "learning_rate": 2.74e-07,
      "loss": 0.3355,
      "reward": 0.625710517168045,
      "reward_std": 0.18957649171352386,
      "reward_std_all": 0.1913791000843048,
      "rewards-max/forensics_sinkhorn_reward": 0.9882557094097137,
      "rewards-max/forensics_strict_edge_reward": 0.946017861366272,
      "rewards-max/forensics_tiou_reward": 0.9971034228801727,
      "rewards/forensics_sinkhorn_reward": 0.5640874356031418,
      "rewards/forensics_strict_edge_reward": 0.2861451804637909,
      "rewards/forensics_tiou_reward": 0.37714041769504547,
      "step": 1090
    },
    {
      "completion_length": 17.171875,
      "epoch": 2.902796271637816,
      "grad_norm": 20.445995330810547,
      "kl": 3.90625,
      "learning_rate": 2.733333333333333e-07,
      "loss": 0.3377,
      "reward": 0.6162706911563873,
      "reward_std": 0.18633104860782623,
      "reward_std_all": 0.18309718370437622,
      "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": 0.5917401164770126,
      "rewards/forensics_strict_edge_reward": 0.371029295027256,
      "rewards/forensics_tiou_reward": 0.46236251294612885,
      "step": 1091
    },
    {
      "completion_length": 16.1875,
      "epoch": 2.9054593874833556,
      "grad_norm": 25.32705307006836,
      "kl": 4.640625,
      "learning_rate": 2.7266666666666665e-07,
      "loss": 0.3705,
      "reward": 0.6326035559177399,
      "reward_std": 0.20836441218852997,
      "reward_std_all": 0.19418172538280487,
      "rewards-max/forensics_sinkhorn_reward": 0.9872831404209137,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9932704865932465,
      "rewards/forensics_sinkhorn_reward": 0.5782299637794495,
      "rewards/forensics_strict_edge_reward": 0.3626185208559036,
      "rewards/forensics_tiou_reward": 0.4292699992656708,
      "step": 1092
    },
    {
      "completion_length": 16.71875,
      "epoch": 2.9081225033288947,
      "grad_norm": 19.9227294921875,
      "kl": 3.953125,
      "learning_rate": 2.72e-07,
      "loss": 0.3025,
      "reward": 0.6273012459278107,
      "reward_std": 0.20040278881788254,
      "reward_std_all": 0.19077952206134796,
      "rewards-max/forensics_sinkhorn_reward": 0.9819359183311462,
      "rewards-max/forensics_strict_edge_reward": 0.9574736058712006,
      "rewards-max/forensics_tiou_reward": 0.9838172495365143,
      "rewards/forensics_sinkhorn_reward": 0.5768319666385651,
      "rewards/forensics_strict_edge_reward": 0.2940479815006256,
      "rewards/forensics_tiou_reward": 0.41862237453460693,
      "step": 1093
    },
    {
      "completion_length": 17.140625,
      "epoch": 2.910785619174434,
      "grad_norm": 17.362457275390625,
      "kl": 5.9375,
      "learning_rate": 2.713333333333333e-07,
      "loss": 0.3534,
      "reward": 0.5981562733650208,
      "reward_std": 0.1561686396598816,
      "reward_std_all": 0.18015413731336594,
      "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": 0.5810919404029846,
      "rewards/forensics_strict_edge_reward": 0.34599433839321136,
      "rewards/forensics_tiou_reward": 0.4383694678544998,
      "step": 1094
    },
    {
      "completion_length": 17.6875,
      "epoch": 2.9134487350199736,
      "grad_norm": 25.622901916503906,
      "kl": 6.75,
      "learning_rate": 2.7066666666666666e-07,
      "loss": 0.5148,
      "reward": 0.5833812355995178,
      "reward_std": 0.1687970981001854,
      "reward_std_all": 0.19310542941093445,
      "rewards-max/forensics_sinkhorn_reward": 0.9858787953853607,
      "rewards-max/forensics_strict_edge_reward": 0.9062341153621674,
      "rewards-max/forensics_tiou_reward": 0.9877789318561554,
      "rewards/forensics_sinkhorn_reward": 0.5607953369617462,
      "rewards/forensics_strict_edge_reward": 0.2536574602127075,
      "rewards/forensics_tiou_reward": 0.3846242278814316,
      "step": 1095
    },
    {
      "completion_length": 16.875,
      "epoch": 2.9161118508655126,
      "grad_norm": 18.197256088256836,
      "kl": 5.375,
      "learning_rate": 2.7e-07,
      "loss": 0.5197,
      "reward": 0.6119197010993958,
      "reward_std": 0.17178773134946823,
      "reward_std_all": 0.18819130212068558,
      "rewards-max/forensics_sinkhorn_reward": 0.9880793988704681,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9964108169078827,
      "rewards/forensics_sinkhorn_reward": 0.5177872776985168,
      "rewards/forensics_strict_edge_reward": 0.32117025554180145,
      "rewards/forensics_tiou_reward": 0.3409760743379593,
      "step": 1096
    },
    {
      "completion_length": 16.390625,
      "epoch": 2.918774966711052,
      "grad_norm": 25.216419219970703,
      "kl": 5.625,
      "learning_rate": 2.693333333333333e-07,
      "loss": 0.4222,
      "reward": 0.5901800394058228,
      "reward_std": 0.24529461562633514,
      "reward_std_all": 0.23056335747241974,
      "rewards-max/forensics_sinkhorn_reward": 0.9696809351444244,
      "rewards-max/forensics_strict_edge_reward": 0.8915697336196899,
      "rewards-max/forensics_tiou_reward": 0.9548379480838776,
      "rewards/forensics_sinkhorn_reward": 0.47366249561309814,
      "rewards/forensics_strict_edge_reward": 0.3081643208861351,
      "rewards/forensics_tiou_reward": 0.35259823501110077,
      "step": 1097
    },
    {
      "completion_length": 17.984375,
      "epoch": 2.921438082556591,
      "grad_norm": 36.472171783447266,
      "kl": 8.84375,
      "learning_rate": 2.6866666666666666e-07,
      "loss": 0.5457,
      "reward": 0.5416841804981232,
      "reward_std": 0.2884656861424446,
      "reward_std_all": 0.27965667843818665,
      "rewards-max/forensics_sinkhorn_reward": 0.9871109426021576,
      "rewards-max/forensics_strict_edge_reward": 0.9677534997463226,
      "rewards-max/forensics_tiou_reward": 0.9925926029682159,
      "rewards/forensics_sinkhorn_reward": 0.5092913508415222,
      "rewards/forensics_strict_edge_reward": 0.28006044030189514,
      "rewards/forensics_tiou_reward": 0.3753686845302582,
      "step": 1098
    },
    {
      "completion_length": 17.78125,
      "epoch": 2.9241011984021306,
      "grad_norm": 32.16646194458008,
      "kl": 6.515625,
      "learning_rate": 2.68e-07,
      "loss": 0.338,
      "reward": 0.53268763422966,
      "reward_std": 0.19666436314582825,
      "reward_std_all": 0.24655913561582565,
      "rewards-max/forensics_sinkhorn_reward": 0.9882758855819702,
      "rewards-max/forensics_strict_edge_reward": 0.9890114367008209,
      "rewards-max/forensics_tiou_reward": 0.997183084487915,
      "rewards/forensics_sinkhorn_reward": 0.453282967209816,
      "rewards/forensics_strict_edge_reward": 0.2238808572292328,
      "rewards/forensics_tiou_reward": 0.28603221476078033,
      "step": 1099
    },
    {
      "completion_length": 17.9375,
      "epoch": 2.92676431424767,
      "grad_norm": 17.617815017700195,
      "kl": 7.96875,
      "learning_rate": 2.673333333333333e-07,
      "loss": 0.5821,
      "reward": 0.4710860252380371,
      "reward_std": 0.20071562379598618,
      "reward_std_all": 0.26273782551288605,
      "rewards-max/forensics_sinkhorn_reward": 0.9883764684200287,
      "rewards-max/forensics_strict_edge_reward": 0.9574736058712006,
      "rewards-max/forensics_tiou_reward": 0.9975824654102325,
      "rewards/forensics_sinkhorn_reward": 0.4578815698623657,
      "rewards/forensics_strict_edge_reward": 0.281547948718071,
      "rewards/forensics_tiou_reward": 0.3926158547401428,
      "step": 1100
    },
    {
      "completion_length": 16.640625,
      "epoch": 2.929427430093209,
      "grad_norm": 17.53885269165039,
      "kl": 5.4765625,
      "learning_rate": 2.6666666666666667e-07,
      "loss": 0.3861,
      "reward": 0.5692324936389923,
      "reward_std": 0.18377789855003357,
      "reward_std_all": 0.20836547762155533,
      "rewards-max/forensics_sinkhorn_reward": 0.9147929847240448,
      "rewards-max/forensics_strict_edge_reward": 0.7634172439575195,
      "rewards-max/forensics_tiou_reward": 0.8306366503238678,
      "rewards/forensics_sinkhorn_reward": 0.48490211367607117,
      "rewards/forensics_strict_edge_reward": 0.21559399366378784,
      "rewards/forensics_tiou_reward": 0.2671271339058876,
      "step": 1101
    },
    {
      "completion_length": 16.34375,
      "epoch": 2.932090545938748,
      "grad_norm": 18.70501708984375,
      "kl": 4.296875,
      "learning_rate": 2.66e-07,
      "loss": 0.363,
      "reward": 0.6358821988105774,
      "reward_std": 0.20448856055736542,
      "reward_std_all": 0.1933080181479454,
      "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": 0.6066433191299438,
      "rewards/forensics_strict_edge_reward": 0.412392258644104,
      "rewards/forensics_tiou_reward": 0.4837806820869446,
      "step": 1102
    },
    {
      "completion_length": 17.078125,
      "epoch": 2.9347536617842875,
      "grad_norm": 25.690444946289062,
      "kl": 5.21875,
      "learning_rate": 2.653333333333333e-07,
      "loss": 0.291,
      "reward": 0.6224633157253265,
      "reward_std": 0.16686387360095978,
      "reward_std_all": 0.17165520787239075,
      "rewards-max/forensics_sinkhorn_reward": 0.987557053565979,
      "rewards-max/forensics_strict_edge_reward": 0.9185642302036285,
      "rewards-max/forensics_tiou_reward": 0.9943662285804749,
      "rewards/forensics_sinkhorn_reward": 0.5649556219577789,
      "rewards/forensics_strict_edge_reward": 0.2689923495054245,
      "rewards/forensics_tiou_reward": 0.3561669886112213,
      "step": 1103
    },
    {
      "completion_length": 15.65625,
      "epoch": 2.937416777629827,
      "grad_norm": 19.962955474853516,
      "kl": 3.9375,
      "learning_rate": 2.6466666666666663e-07,
      "loss": 0.2524,
      "reward": 0.6210370361804962,
      "reward_std": 0.141245249658823,
      "reward_std_all": 0.16111813485622406,
      "rewards-max/forensics_sinkhorn_reward": 0.9860047698020935,
      "rewards-max/forensics_strict_edge_reward": 0.9890114367008209,
      "rewards-max/forensics_tiou_reward": 0.9931923449039459,
      "rewards/forensics_sinkhorn_reward": 0.5772802829742432,
      "rewards/forensics_strict_edge_reward": 0.342800036072731,
      "rewards/forensics_tiou_reward": 0.3900063931941986,
      "step": 1104
    },
    {
      "completion_length": 16.828125,
      "epoch": 2.940079893475366,
      "grad_norm": 21.444808959960938,
      "kl": 2.96484375,
      "learning_rate": 2.64e-07,
      "loss": 0.2448,
      "reward": 0.5875096917152405,
      "reward_std": 0.2170204445719719,
      "reward_std_all": 0.2170964926481247,
      "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": 0.5681244283914566,
      "rewards/forensics_strict_edge_reward": 0.3287940174341202,
      "rewards/forensics_tiou_reward": 0.3945121318101883,
      "step": 1105
    },
    {
      "completion_length": 16.96875,
      "epoch": 2.9427430093209055,
      "grad_norm": 17.46538734436035,
      "kl": 3.046875,
      "learning_rate": 2.633333333333333e-07,
      "loss": 0.2813,
      "reward": 0.5695272386074066,
      "reward_std": 0.23160075396299362,
      "reward_std_all": 0.23119305074214935,
      "rewards-max/forensics_sinkhorn_reward": 0.9769107401371002,
      "rewards-max/forensics_strict_edge_reward": 0.8445805907249451,
      "rewards-max/forensics_tiou_reward": 0.9722158908843994,
      "rewards/forensics_sinkhorn_reward": 0.4552706182003021,
      "rewards/forensics_strict_edge_reward": 0.24622268974781036,
      "rewards/forensics_tiou_reward": 0.27388226985931396,
      "step": 1106
    },
    {
      "completion_length": 16.53125,
      "epoch": 2.9454061251664445,
      "grad_norm": 24.624942779541016,
      "kl": 5.0234375,
      "learning_rate": 2.6266666666666664e-07,
      "loss": 0.3933,
      "reward": 0.5932815074920654,
      "reward_std": 0.13435163721442223,
      "reward_std_all": 0.17277148365974426,
      "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": 0.5619737505912781,
      "rewards/forensics_strict_edge_reward": 0.32014527171850204,
      "rewards/forensics_tiou_reward": 0.41364750266075134,
      "step": 1107
    },
    {
      "completion_length": 15.671875,
      "epoch": 2.948069241011984,
      "grad_norm": 22.736072540283203,
      "kl": 3.84375,
      "learning_rate": 2.62e-07,
      "loss": 0.4318,
      "reward": 0.6181041598320007,
      "reward_std": 0.19172238558530807,
      "reward_std_all": 0.19015274196863174,
      "rewards-max/forensics_sinkhorn_reward": 0.9819468557834625,
      "rewards-max/forensics_strict_edge_reward": 0.9265981018543243,
      "rewards-max/forensics_tiou_reward": 0.9722630381584167,
      "rewards/forensics_sinkhorn_reward": 0.6167784631252289,
      "rewards/forensics_strict_edge_reward": 0.3161405771970749,
      "rewards/forensics_tiou_reward": 0.46065472066402435,
      "step": 1108
    },
    {
      "completion_length": 15.65625,
      "epoch": 2.9507323568575234,
      "grad_norm": 23.89631462097168,
      "kl": 4.921875,
      "learning_rate": 2.613333333333333e-07,
      "loss": 0.3803,
      "reward": 0.5481034815311432,
      "reward_std": 0.2229311615228653,
      "reward_std_all": 0.2444968819618225,
      "rewards-max/forensics_sinkhorn_reward": 0.9812131822109222,
      "rewards-max/forensics_strict_edge_reward": 0.878633052110672,
      "rewards-max/forensics_tiou_reward": 0.9816499054431915,
      "rewards/forensics_sinkhorn_reward": 0.46944308280944824,
      "rewards/forensics_strict_edge_reward": 0.21374238654971123,
      "rewards/forensics_tiou_reward": 0.2934538349509239,
      "step": 1109
    },
    {
      "completion_length": 16.453125,
      "epoch": 2.9533954727030625,
      "grad_norm": 20.766496658325195,
      "kl": 4.1484375,
      "learning_rate": 2.6066666666666664e-07,
      "loss": 0.2901,
      "reward": 0.5983951985836029,
      "reward_std": 0.17074642330408096,
      "reward_std_all": 0.19033029675483704,
      "rewards-max/forensics_sinkhorn_reward": 0.9872133433818817,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9929953813552856,
      "rewards/forensics_sinkhorn_reward": 0.5857353806495667,
      "rewards/forensics_strict_edge_reward": 0.3717687427997589,
      "rewards/forensics_tiou_reward": 0.45616985857486725,
      "step": 1110
    },
    {
      "completion_length": 17.109375,
      "epoch": 2.956058588548602,
      "grad_norm": 22.374330520629883,
      "kl": 5.34375,
      "learning_rate": 2.6e-07,
      "loss": 0.45,
      "reward": 0.534038782119751,
      "reward_std": 0.23320113122463226,
      "reward_std_all": 0.2580453008413315,
      "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": 0.5230349600315094,
      "rewards/forensics_strict_edge_reward": 0.3776169717311859,
      "rewards/forensics_tiou_reward": 0.43205659091472626,
      "step": 1111
    },
    {
      "completion_length": 16.90625,
      "epoch": 2.958721704394141,
      "grad_norm": 16.63722801208496,
      "kl": 6.546875,
      "learning_rate": 2.5933333333333335e-07,
      "loss": 0.5407,
      "reward": 0.4905453771352768,
      "reward_std": 0.23372063785791397,
      "reward_std_all": 0.27081216871738434,
      "rewards-max/forensics_sinkhorn_reward": 0.9876256585121155,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9946199953556061,
      "rewards/forensics_sinkhorn_reward": 0.48155948519706726,
      "rewards/forensics_strict_edge_reward": 0.3003467172384262,
      "rewards/forensics_tiou_reward": 0.3838420808315277,
      "step": 1112
    },
    {
      "completion_length": 16.09375,
      "epoch": 2.9613848202396804,
      "grad_norm": 32.273231506347656,
      "kl": 6.609375,
      "learning_rate": 2.5866666666666665e-07,
      "loss": 0.4697,
      "reward": 0.5338925719261169,
      "reward_std": 0.19638458639383316,
      "reward_std_all": 0.22048866003751755,
      "rewards-max/forensics_sinkhorn_reward": 0.9874785840511322,
      "rewards-max/forensics_strict_edge_reward": 0.8983768224716187,
      "rewards-max/forensics_tiou_reward": 0.9940570294857025,
      "rewards/forensics_sinkhorn_reward": 0.529903382062912,
      "rewards/forensics_strict_edge_reward": 0.3405957743525505,
      "rewards/forensics_tiou_reward": 0.42097656428813934,
      "step": 1113
    },
    {
      "completion_length": 15.671875,
      "epoch": 2.96404793608522,
      "grad_norm": 17.54153823852539,
      "kl": 3.890625,
      "learning_rate": 2.58e-07,
      "loss": 0.2897,
      "reward": 0.6426649987697601,
      "reward_std": 0.14796015620231628,
      "reward_std_all": 0.15301556140184402,
      "rewards-max/forensics_sinkhorn_reward": 0.9887357950210571,
      "rewards-max/forensics_strict_edge_reward": 0.9890114367008209,
      "rewards-max/forensics_tiou_reward": 0.9989964962005615,
      "rewards/forensics_sinkhorn_reward": 0.6352702677249908,
      "rewards/forensics_strict_edge_reward": 0.39504702389240265,
      "rewards/forensics_tiou_reward": 0.48674771189689636,
      "step": 1114
    },
    {
      "completion_length": 16.71875,
      "epoch": 2.966711051930759,
      "grad_norm": 18.17028045654297,
      "kl": 2.63671875,
      "learning_rate": 2.5733333333333336e-07,
      "loss": 0.0068,
      "reward": 0.6144243478775024,
      "reward_std": 0.13226047158241272,
      "reward_std_all": 0.160368412733078,
      "rewards-max/forensics_sinkhorn_reward": 0.9886972308158875,
      "rewards-max/forensics_strict_edge_reward": 0.9890114367008209,
      "rewards-max/forensics_tiou_reward": 0.9988439381122589,
      "rewards/forensics_sinkhorn_reward": 0.5524243712425232,
      "rewards/forensics_strict_edge_reward": 0.27621033787727356,
      "rewards/forensics_tiou_reward": 0.33613312244415283,
      "step": 1115
    },
    {
      "completion_length": 16.0,
      "epoch": 2.9693741677762984,
      "grad_norm": 15.315509796142578,
      "kl": 4.25,
      "learning_rate": 2.5666666666666666e-07,
      "loss": 0.3575,
      "reward": 0.6564759910106659,
      "reward_std": 0.1906742975115776,
      "reward_std_all": 0.1817413866519928,
      "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": 0.5962639451026917,
      "rewards/forensics_strict_edge_reward": 0.3363167345523834,
      "rewards/forensics_tiou_reward": 0.4323571175336838,
      "step": 1116
    },
    {
      "completion_length": 18.171875,
      "epoch": 2.9720372836218374,
      "grad_norm": 23.43189811706543,
      "kl": 4.046875,
      "learning_rate": 2.56e-07,
      "loss": 0.2847,
      "reward": 0.6509384512901306,
      "reward_std": 0.17408595234155655,
      "reward_std_all": 0.16626977920532227,
      "rewards-max/forensics_sinkhorn_reward": 0.8584176599979401,
      "rewards-max/forensics_strict_edge_reward": 0.8078543543815613,
      "rewards-max/forensics_tiou_reward": 0.8158263564109802,
      "rewards/forensics_sinkhorn_reward": 0.5426121652126312,
      "rewards/forensics_strict_edge_reward": 0.2853672206401825,
      "rewards/forensics_tiou_reward": 0.37341368198394775,
      "step": 1117
    },
    {
      "completion_length": 19.046875,
      "epoch": 2.974700399467377,
      "grad_norm": 20.26163101196289,
      "kl": 4.3984375,
      "learning_rate": 2.5533333333333337e-07,
      "loss": 0.1499,
      "reward": 0.6915807723999023,
      "reward_std": 0.09749564342200756,
      "reward_std_all": 0.10819517821073532,
      "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": 0.6525019109249115,
      "rewards/forensics_strict_edge_reward": 0.39447933435440063,
      "rewards/forensics_tiou_reward": 0.5049923658370972,
      "step": 1118
    },
    {
      "completion_length": 15.875,
      "epoch": 2.9773635153129163,
      "grad_norm": 16.927602767944336,
      "kl": 4.0,
      "learning_rate": 2.546666666666666e-07,
      "loss": 0.3615,
      "reward": 0.553530752658844,
      "reward_std": 0.17283781617879868,
      "reward_std_all": 0.2194371148943901,
      "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": 0.5640464723110199,
      "rewards/forensics_strict_edge_reward": 0.3819948434829712,
      "rewards/forensics_tiou_reward": 0.45467810332775116,
      "step": 1119
    },
    {
      "completion_length": 15.828125,
      "epoch": 2.9800266311584553,
      "grad_norm": 24.87486457824707,
      "kl": 3.390625,
      "learning_rate": 2.5399999999999997e-07,
      "loss": 0.2185,
      "reward": 0.5860985219478607,
      "reward_std": 0.22928158193826675,
      "reward_std_all": 0.2213655188679695,
      "rewards-max/forensics_sinkhorn_reward": 0.927743136882782,
      "rewards-max/forensics_strict_edge_reward": 0.7097077369689941,
      "rewards-max/forensics_tiou_reward": 0.8675743937492371,
      "rewards/forensics_sinkhorn_reward": 0.48311343789100647,
      "rewards/forensics_strict_edge_reward": 0.1740715652704239,
      "rewards/forensics_tiou_reward": 0.26115939021110535,
      "step": 1120
    },
    {
      "completion_length": 16.09375,
      "epoch": 2.982689747003995,
      "grad_norm": 21.145662307739258,
      "kl": 3.265625,
      "learning_rate": 2.533333333333333e-07,
      "loss": 0.1607,
      "reward": 0.6717158854007721,
      "reward_std": 0.15153061598539352,
      "reward_std_all": 0.1518924981355667,
      "rewards-max/forensics_sinkhorn_reward": 0.9888334274291992,
      "rewards-max/forensics_strict_edge_reward": 0.9890114367008209,
      "rewards-max/forensics_tiou_reward": 0.9993807971477509,
      "rewards/forensics_sinkhorn_reward": 0.6272572576999664,
      "rewards/forensics_strict_edge_reward": 0.3514733910560608,
      "rewards/forensics_tiou_reward": 0.44611047208309174,
      "step": 1121
    },
    {
      "completion_length": 17.125,
      "epoch": 2.985352862849534,
      "grad_norm": 24.049367904663086,
      "kl": 4.6640625,
      "learning_rate": 2.526666666666666e-07,
      "loss": 0.1689,
      "reward": 0.5801981687545776,
      "reward_std": 0.19918183982372284,
      "reward_std_all": 0.2097116857767105,
      "rewards-max/forensics_sinkhorn_reward": 0.9841704070568085,
      "rewards-max/forensics_strict_edge_reward": 0.7699888050556183,
      "rewards-max/forensics_tiou_reward": 0.9811376929283142,
      "rewards/forensics_sinkhorn_reward": 0.4639989286661148,
      "rewards/forensics_strict_edge_reward": 0.20734652876853943,
      "rewards/forensics_tiou_reward": 0.2674648314714432,
      "step": 1122
    },
    {
      "completion_length": 18.78125,
      "epoch": 2.9880159786950733,
      "grad_norm": 20.822982788085938,
      "kl": 5.4375,
      "learning_rate": 2.52e-07,
      "loss": 0.4512,
      "reward": 0.5909190475940704,
      "reward_std": 0.2279544249176979,
      "reward_std_all": 0.22253355383872986,
      "rewards-max/forensics_sinkhorn_reward": 0.9420543015003204,
      "rewards-max/forensics_strict_edge_reward": 0.7573865950107574,
      "rewards-max/forensics_tiou_reward": 0.9036043286323547,
      "rewards/forensics_sinkhorn_reward": 0.5335088074207306,
      "rewards/forensics_strict_edge_reward": 0.20262258127331734,
      "rewards/forensics_tiou_reward": 0.3513719066977501,
      "step": 1123
    },
    {
      "completion_length": 15.40625,
      "epoch": 2.9906790945406128,
      "grad_norm": 18.3513240814209,
      "kl": 4.0234375,
      "learning_rate": 2.5133333333333333e-07,
      "loss": 0.2949,
      "reward": 0.6408408284187317,
      "reward_std": 0.17873138189315796,
      "reward_std_all": 0.18245068192481995,
      "rewards-max/forensics_sinkhorn_reward": 0.98814857006073,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9966813623905182,
      "rewards/forensics_sinkhorn_reward": 0.6273533701896667,
      "rewards/forensics_strict_edge_reward": 0.4296092689037323,
      "rewards/forensics_tiou_reward": 0.5083912014961243,
      "step": 1124
    },
    {
      "completion_length": 15.375,
      "epoch": 2.993342210386152,
      "grad_norm": 23.912153244018555,
      "kl": 6.4375,
      "learning_rate": 2.5066666666666663e-07,
      "loss": 0.4661,
      "reward": 0.5935478508472443,
      "reward_std": 0.1835014596581459,
      "reward_std_all": 0.20451730489730835,
      "rewards-max/forensics_sinkhorn_reward": 0.9885534644126892,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9982782006263733,
      "rewards/forensics_sinkhorn_reward": 0.5960820019245148,
      "rewards/forensics_strict_edge_reward": 0.3724013864994049,
      "rewards/forensics_tiou_reward": 0.48269234597682953,
      "step": 1125
    },
    {
      "completion_length": 17.828125,
      "epoch": 2.996005326231691,
      "grad_norm": 22.638761520385742,
      "kl": 3.140625,
      "learning_rate": 2.5e-07,
      "loss": 0.0675,
      "reward": 0.6404826641082764,
      "reward_std": 0.12245658040046692,
      "reward_std_all": 0.1396087110042572,
      "rewards-max/forensics_sinkhorn_reward": 0.9478819668292999,
      "rewards-max/forensics_strict_edge_reward": 0.8553276658058167,
      "rewards-max/forensics_tiou_reward": 0.9093345701694489,
      "rewards/forensics_sinkhorn_reward": 0.56772381067276,
      "rewards/forensics_strict_edge_reward": 0.2752678692340851,
      "rewards/forensics_tiou_reward": 0.3632037341594696,
      "step": 1126
    },
    {
      "completion_length": 15.359375,
      "epoch": 2.9986684420772303,
      "grad_norm": 18.36532974243164,
      "kl": 3.1640625,
      "learning_rate": 2.493333333333333e-07,
      "loss": 0.2472,
      "reward": 0.6532016098499298,
      "reward_std": 0.1731748804450035,
      "reward_std_all": 0.17266342788934708,
      "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": 0.6812042891979218,
      "rewards/forensics_strict_edge_reward": 0.4496179521083832,
      "rewards/forensics_tiou_reward": 0.5434622764587402,
      "step": 1127
    },
    {
      "completion_length": 16.0,
      "epoch": 3.0,
      "grad_norm": 15.447775840759277,
      "kl": 10.5,
      "learning_rate": 2.4866666666666664e-07,
      "loss": 0.1891,
      "reward": 0.08676373958587646,
      "reward_std": 0.41676563024520874,
      "reward_std_all": 0.41752636432647705,
      "rewards-max/forensics_sinkhorn_reward": 0.0,
      "rewards-max/forensics_strict_edge_reward": 0.0,
      "rewards-max/forensics_tiou_reward": 0.0,
      "rewards/forensics_sinkhorn_reward": 0.0,
      "rewards/forensics_strict_edge_reward": 0.0,
      "rewards/forensics_tiou_reward": 0.0,
      "step": 1128
    },
    {
      "completion_length": 15.71875,
      "epoch": 3.0026631158455395,
      "grad_norm": 28.10747528076172,
      "kl": 4.046875,
      "learning_rate": 2.48e-07,
      "loss": 0.3061,
      "reward": 0.5960020124912262,
      "reward_std": 0.23423662036657333,
      "reward_std_all": 0.2279745191335678,
      "rewards-max/forensics_sinkhorn_reward": 0.987274169921875,
      "rewards-max/forensics_strict_edge_reward": 0.905340164899826,
      "rewards-max/forensics_tiou_reward": 0.9932461977005005,
      "rewards/forensics_sinkhorn_reward": 0.5522810816764832,
      "rewards/forensics_strict_edge_reward": 0.2251862734556198,
      "rewards/forensics_tiou_reward": 0.33759117126464844,
      "step": 1129
    },
    {
      "completion_length": 16.9375,
      "epoch": 3.0053262316910785,
      "grad_norm": 19.58733367919922,
      "kl": 4.3359375,
      "learning_rate": 2.473333333333333e-07,
      "loss": 0.2301,
      "reward": 0.6493225991725922,
      "reward_std": 0.138120386749506,
      "reward_std_all": 0.14572268351912498,
      "rewards-max/forensics_sinkhorn_reward": 0.9877222776412964,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9950017035007477,
      "rewards/forensics_sinkhorn_reward": 0.620773434638977,
      "rewards/forensics_strict_edge_reward": 0.3066388815641403,
      "rewards/forensics_tiou_reward": 0.40028248727321625,
      "step": 1130
    },
    {
      "completion_length": 17.125,
      "epoch": 3.007989347536618,
      "grad_norm": 32.90818405151367,
      "kl": 5.84375,
      "learning_rate": 2.4666666666666665e-07,
      "loss": 0.364,
      "reward": 0.530633270740509,
      "reward_std": 0.16498420387506485,
      "reward_std_all": 0.21614858508110046,
      "rewards-max/forensics_sinkhorn_reward": 0.9830209016799927,
      "rewards-max/forensics_strict_edge_reward": 0.9567649364471436,
      "rewards-max/forensics_tiou_reward": 0.9853352010250092,
      "rewards/forensics_sinkhorn_reward": 0.4456530064344406,
      "rewards/forensics_strict_edge_reward": 0.23445898294448853,
      "rewards/forensics_tiou_reward": 0.2830168232321739,
      "step": 1131
    },
    {
      "completion_length": 15.703125,
      "epoch": 3.010652463382157,
      "grad_norm": 31.049394607543945,
      "kl": 3.2109375,
      "learning_rate": 2.46e-07,
      "loss": 0.2044,
      "reward": 0.593395471572876,
      "reward_std": 0.14492622017860413,
      "reward_std_all": 0.1760396957397461,
      "rewards-max/forensics_sinkhorn_reward": 0.9886327385902405,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9985897243022919,
      "rewards/forensics_sinkhorn_reward": 0.5446386635303497,
      "rewards/forensics_strict_edge_reward": 0.2629983276128769,
      "rewards/forensics_tiou_reward": 0.3215567618608475,
      "step": 1132
    },
    {
      "completion_length": 15.28125,
      "epoch": 3.0133155792276964,
      "grad_norm": 22.03500747680664,
      "kl": 3.09375,
      "learning_rate": 2.453333333333333e-07,
      "loss": 0.2242,
      "reward": 0.6070327162742615,
      "reward_std": 0.1328735686838627,
      "reward_std_all": 0.16508949548006058,
      "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": 0.5710671842098236,
      "rewards/forensics_strict_edge_reward": 0.2746240049600601,
      "rewards/forensics_tiou_reward": 0.353070393204689,
      "step": 1133
    },
    {
      "completion_length": 16.8125,
      "epoch": 3.015978695073236,
      "grad_norm": 24.318614959716797,
      "kl": 4.5390625,
      "learning_rate": 2.4466666666666666e-07,
      "loss": 0.2755,
      "reward": 0.6291742622852325,
      "reward_std": 0.17939847707748413,
      "reward_std_all": 0.18399878591299057,
      "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": 0.5801263451576233,
      "rewards/forensics_strict_edge_reward": 0.3865399658679962,
      "rewards/forensics_tiou_reward": 0.476026713848114,
      "step": 1134
    },
    {
      "completion_length": 17.140625,
      "epoch": 3.018641810918775,
      "grad_norm": 27.312419891357422,
      "kl": 4.2421875,
      "learning_rate": 2.4399999999999996e-07,
      "loss": 0.249,
      "reward": 0.6082627773284912,
      "reward_std": 0.22369059920310974,
      "reward_std_all": 0.21459928154945374,
      "rewards-max/forensics_sinkhorn_reward": 0.9819359183311462,
      "rewards-max/forensics_strict_edge_reward": 0.9677534997463226,
      "rewards-max/forensics_tiou_reward": 0.983410507440567,
      "rewards/forensics_sinkhorn_reward": 0.5440466850996017,
      "rewards/forensics_strict_edge_reward": 0.32402026653289795,
      "rewards/forensics_tiou_reward": 0.4274750053882599,
      "step": 1135
    },
    {
      "completion_length": 21.1875,
      "epoch": 3.0213049267643144,
      "grad_norm": 744677312.0,
      "kl": 317442.625,
      "learning_rate": 2.433333333333333e-07,
      "loss": 12743.4141,
      "reward": 0.6109056174755096,
      "reward_std": 0.17197257280349731,
      "reward_std_all": 0.18969684094190598,
      "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": 0.5533716678619385,
      "rewards/forensics_strict_edge_reward": 0.27401939034461975,
      "rewards/forensics_tiou_reward": 0.373960942029953,
      "step": 1136
    },
    {
      "completion_length": 16.0625,
      "epoch": 3.0239680426098534,
      "grad_norm": 19.135969161987305,
      "kl": 3.625,
      "learning_rate": 2.4266666666666667e-07,
      "loss": 0.3021,
      "reward": 0.6858316957950592,
      "reward_std": 0.1787610873579979,
      "reward_std_all": 0.15986453369259834,
      "rewards-max/forensics_sinkhorn_reward": 0.9751991033554077,
      "rewards-max/forensics_strict_edge_reward": 0.8528120815753937,
      "rewards-max/forensics_tiou_reward": 0.9694458544254303,
      "rewards/forensics_sinkhorn_reward": 0.6821626424789429,
      "rewards/forensics_strict_edge_reward": 0.37666648626327515,
      "rewards/forensics_tiou_reward": 0.5543194115161896,
      "step": 1137
    },
    {
      "completion_length": 16.421875,
      "epoch": 3.026631158455393,
      "grad_norm": 19.8276424407959,
      "kl": 7.28125,
      "learning_rate": 2.4199999999999997e-07,
      "loss": 0.514,
      "reward": 0.5999968647956848,
      "reward_std": 0.22101187705993652,
      "reward_std_all": 0.2280411496758461,
      "rewards-max/forensics_sinkhorn_reward": 0.9561372995376587,
      "rewards-max/forensics_strict_edge_reward": 0.8365300297737122,
      "rewards-max/forensics_tiou_reward": 0.9249453842639923,
      "rewards/forensics_sinkhorn_reward": 0.5435295701026917,
      "rewards/forensics_strict_edge_reward": 0.24519222229719162,
      "rewards/forensics_tiou_reward": 0.3718676194548607,
      "step": 1138
    },
    {
      "completion_length": 15.453125,
      "epoch": 3.029294274300932,
      "grad_norm": 27.90178871154785,
      "kl": 3.0546875,
      "learning_rate": 2.413333333333333e-07,
      "loss": 0.1989,
      "reward": 0.6637697219848633,
      "reward_std": 0.08524703979492188,
      "reward_std_all": 0.11138299107551575,
      "rewards-max/forensics_sinkhorn_reward": 0.9882344305515289,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9970229268074036,
      "rewards/forensics_sinkhorn_reward": 0.6217197775840759,
      "rewards/forensics_strict_edge_reward": 0.3575389087200165,
      "rewards/forensics_tiou_reward": 0.4351232498884201,
      "step": 1139
    },
    {
      "completion_length": 15.78125,
      "epoch": 3.0319573901464714,
      "grad_norm": 33.6273193359375,
      "kl": 3.6796875,
      "learning_rate": 2.406666666666667e-07,
      "loss": 0.2904,
      "reward": 0.5861779749393463,
      "reward_std": 0.09986201673746109,
      "reward_std_all": 0.15408798307180405,
      "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": 0.5432733297348022,
      "rewards/forensics_strict_edge_reward": 0.2580811083316803,
      "rewards/forensics_tiou_reward": 0.3674905598163605,
      "step": 1140
    },
    {
      "completion_length": 15.484375,
      "epoch": 3.034620505992011,
      "grad_norm": 26.633352279663086,
      "kl": 5.171875,
      "learning_rate": 2.4e-07,
      "loss": 0.2774,
      "reward": 0.5865113437175751,
      "reward_std": 0.1604311242699623,
      "reward_std_all": 0.1893596425652504,
      "rewards-max/forensics_sinkhorn_reward": 0.9882029294967651,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9968973696231842,
      "rewards/forensics_sinkhorn_reward": 0.5379491150379181,
      "rewards/forensics_strict_edge_reward": 0.26466722786426544,
      "rewards/forensics_tiou_reward": 0.3364461660385132,
      "step": 1141
    },
    {
      "completion_length": 15.75,
      "epoch": 3.03728362183755,
      "grad_norm": 21.65287971496582,
      "kl": 4.1953125,
      "learning_rate": 2.3933333333333333e-07,
      "loss": 0.3104,
      "reward": 0.665790319442749,
      "reward_std": 0.19568860530853271,
      "reward_std_all": 0.18021058291196823,
      "rewards-max/forensics_sinkhorn_reward": 0.9738631546497345,
      "rewards-max/forensics_strict_edge_reward": 0.8719756007194519,
      "rewards-max/forensics_tiou_reward": 0.9644846022129059,
      "rewards/forensics_sinkhorn_reward": 0.6327951550483704,
      "rewards/forensics_strict_edge_reward": 0.34107042849063873,
      "rewards/forensics_tiou_reward": 0.4439725875854492,
      "step": 1142
    },
    {
      "completion_length": 17.078125,
      "epoch": 3.0399467376830893,
      "grad_norm": 31.477079391479492,
      "kl": 3.2421875,
      "learning_rate": 2.3866666666666663e-07,
      "loss": 0.4519,
      "reward": 0.5494418144226074,
      "reward_std": 0.21648389101028442,
      "reward_std_all": 0.2457723617553711,
      "rewards-max/forensics_sinkhorn_reward": 0.9865670204162598,
      "rewards-max/forensics_strict_edge_reward": 0.9464850425720215,
      "rewards-max/forensics_tiou_reward": 0.9904857873916626,
      "rewards/forensics_sinkhorn_reward": 0.5332012176513672,
      "rewards/forensics_strict_edge_reward": 0.2807135581970215,
      "rewards/forensics_tiou_reward": 0.39967116713523865,
      "step": 1143
    },
    {
      "completion_length": 15.859375,
      "epoch": 3.0426098535286283,
      "grad_norm": 19.288511276245117,
      "kl": 4.328125,
      "learning_rate": 2.38e-07,
      "loss": 0.2976,
      "reward": 0.6473874151706696,
      "reward_std": 0.14135992154479027,
      "reward_std_all": 0.14768143743276596,
      "rewards-max/forensics_sinkhorn_reward": 0.9756087064743042,
      "rewards-max/forensics_strict_edge_reward": 0.947419673204422,
      "rewards-max/forensics_tiou_reward": 0.969355970621109,
      "rewards/forensics_sinkhorn_reward": 0.588187962770462,
      "rewards/forensics_strict_edge_reward": 0.29064103960990906,
      "rewards/forensics_tiou_reward": 0.3957003504037857,
      "step": 1144
    },
    {
      "completion_length": 18.171875,
      "epoch": 3.045272969374168,
      "grad_norm": 27.335508346557617,
      "kl": 3.9453125,
      "learning_rate": 2.3733333333333334e-07,
      "loss": 0.2566,
      "reward": 0.6098887324333191,
      "reward_std": 0.17784224450588226,
      "reward_std_all": 0.1806236281991005,
      "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": 0.49014078080654144,
      "rewards/forensics_strict_edge_reward": 0.224745225161314,
      "rewards/forensics_tiou_reward": 0.3200995475053787,
      "step": 1145
    },
    {
      "completion_length": 17.609375,
      "epoch": 3.0479360852197073,
      "grad_norm": 24.548547744750977,
      "kl": 6.28125,
      "learning_rate": 2.3666666666666664e-07,
      "loss": 0.3972,
      "reward": 0.594913899898529,
      "reward_std": 0.23180165141820908,
      "reward_std_all": 0.21972737461328506,
      "rewards-max/forensics_sinkhorn_reward": 0.9701376557350159,
      "rewards-max/forensics_strict_edge_reward": 0.9574736058712006,
      "rewards-max/forensics_tiou_reward": 0.9585616588592529,
      "rewards/forensics_sinkhorn_reward": 0.5148571133613586,
      "rewards/forensics_strict_edge_reward": 0.20060096681118011,
      "rewards/forensics_tiou_reward": 0.3230518102645874,
      "step": 1146
    },
    {
      "completion_length": 17.015625,
      "epoch": 3.0505992010652463,
      "grad_norm": 26.601511001586914,
      "kl": 6.609375,
      "learning_rate": 2.3599999999999997e-07,
      "loss": 0.4758,
      "reward": 0.5566822588443756,
      "reward_std": 0.22094547003507614,
      "reward_std_all": 0.22733189165592194,
      "rewards-max/forensics_sinkhorn_reward": 0.9410161375999451,
      "rewards-max/forensics_strict_edge_reward": 0.7931119501590729,
      "rewards-max/forensics_tiou_reward": 0.8929082155227661,
      "rewards/forensics_sinkhorn_reward": 0.4636538624763489,
      "rewards/forensics_strict_edge_reward": 0.21566612273454666,
      "rewards/forensics_tiou_reward": 0.3126094490289688,
      "step": 1147
    },
    {
      "completion_length": 19.34375,
      "epoch": 3.0532623169107858,
      "grad_norm": 22.150239944458008,
      "kl": 5.46875,
      "learning_rate": 2.3533333333333332e-07,
      "loss": 0.4409,
      "reward": 0.5317448675632477,
      "reward_std": 0.29870525002479553,
      "reward_std_all": 0.2866285741329193,
      "rewards-max/forensics_sinkhorn_reward": 0.9708617031574249,
      "rewards-max/forensics_strict_edge_reward": 0.9357379674911499,
      "rewards-max/forensics_tiou_reward": 0.9610365331172943,
      "rewards/forensics_sinkhorn_reward": 0.4136482775211334,
      "rewards/forensics_strict_edge_reward": 0.16555658727884293,
      "rewards/forensics_tiou_reward": 0.23348063230514526,
      "step": 1148
    },
    {
      "completion_length": 18.90625,
      "epoch": 3.0559254327563248,
      "grad_norm": 18.992395401000977,
      "kl": 5.5625,
      "learning_rate": 2.3466666666666665e-07,
      "loss": 0.3837,
      "reward": 0.5518946051597595,
      "reward_std": 0.24541877210140228,
      "reward_std_all": 0.24814893305301666,
      "rewards-max/forensics_sinkhorn_reward": 0.9876851439476013,
      "rewards-max/forensics_strict_edge_reward": 0.9464850425720215,
      "rewards-max/forensics_tiou_reward": 0.9948640465736389,
      "rewards/forensics_sinkhorn_reward": 0.47327491641044617,
      "rewards/forensics_strict_edge_reward": 0.26772043108940125,
      "rewards/forensics_tiou_reward": 0.3475941866636276,
      "step": 1149
    },
    {
      "completion_length": 18.203125,
      "epoch": 3.0585885486018642,
      "grad_norm": 27.413042068481445,
      "kl": 7.28125,
      "learning_rate": 2.34e-07,
      "loss": 0.5043,
      "reward": 0.5205651819705963,
      "reward_std": 0.28953585773706436,
      "reward_std_all": 0.29067523032426834,
      "rewards-max/forensics_sinkhorn_reward": 0.937440425157547,
      "rewards-max/forensics_strict_edge_reward": 0.5728245973587036,
      "rewards-max/forensics_tiou_reward": 0.8842843174934387,
      "rewards/forensics_sinkhorn_reward": 0.3969274014234543,
      "rewards/forensics_strict_edge_reward": 0.13130780681967735,
      "rewards/forensics_tiou_reward": 0.2272302806377411,
      "step": 1150
    },
    {
      "completion_length": 17.109375,
      "epoch": 3.0612516644474033,
      "grad_norm": 36.51027297973633,
      "kl": 5.828125,
      "learning_rate": 2.3333333333333333e-07,
      "loss": 0.3111,
      "reward": 0.5299700796604156,
      "reward_std": 0.21257196366786957,
      "reward_std_all": 0.23476938903331757,
      "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": 0.45938462018966675,
      "rewards/forensics_strict_edge_reward": 0.23297911882400513,
      "rewards/forensics_tiou_reward": 0.2905337065458298,
      "step": 1151
    },
    {
      "completion_length": 15.921875,
      "epoch": 3.0639147802929427,
      "grad_norm": 36.48210906982422,
      "kl": 7.140625,
      "learning_rate": 2.3266666666666666e-07,
      "loss": 0.5659,
      "reward": 0.5582195520401001,
      "reward_std": 0.2547561228275299,
      "reward_std_all": 0.25409817695617676,
      "rewards-max/forensics_sinkhorn_reward": 0.9877058267593384,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9949361085891724,
      "rewards/forensics_sinkhorn_reward": 0.5213875770568848,
      "rewards/forensics_strict_edge_reward": 0.3265499323606491,
      "rewards/forensics_tiou_reward": 0.3833700120449066,
      "step": 1152
    },
    {
      "completion_length": 16.125,
      "epoch": 3.066577896138482,
      "grad_norm": 28.549360275268555,
      "kl": 4.1015625,
      "learning_rate": 2.32e-07,
      "loss": 0.344,
      "reward": 0.6090835630893707,
      "reward_std": 0.12874094769358635,
      "reward_std_all": 0.16284296661615372,
      "rewards-max/forensics_sinkhorn_reward": 0.9427561163902283,
      "rewards-max/forensics_strict_edge_reward": 0.810933917760849,
      "rewards-max/forensics_tiou_reward": 0.8898183405399323,
      "rewards/forensics_sinkhorn_reward": 0.5055059641599655,
      "rewards/forensics_strict_edge_reward": 0.18330926448106766,
      "rewards/forensics_tiou_reward": 0.25235237181186676,
      "step": 1153
    },
    {
      "completion_length": 19.28125,
      "epoch": 3.069241011984021,
      "grad_norm": 25.5577392578125,
      "kl": 4.640625,
      "learning_rate": 2.3133333333333331e-07,
      "loss": 0.3467,
      "reward": 0.557114452123642,
      "reward_std": 0.24320746958255768,
      "reward_std_all": 0.24119308590888977,
      "rewards-max/forensics_sinkhorn_reward": 0.9883498847484589,
      "rewards-max/forensics_strict_edge_reward": 0.9890114367008209,
      "rewards-max/forensics_tiou_reward": 0.9974779486656189,
      "rewards/forensics_sinkhorn_reward": 0.5302859246730804,
      "rewards/forensics_strict_edge_reward": 0.3148701936006546,
      "rewards/forensics_tiou_reward": 0.4128700792789459,
      "step": 1154
    },
    {
      "completion_length": 15.453125,
      "epoch": 3.0719041278295607,
      "grad_norm": 27.04859161376953,
      "kl": 5.140625,
      "learning_rate": 2.3066666666666664e-07,
      "loss": 0.2993,
      "reward": 0.5933028161525726,
      "reward_std": 0.15501876920461655,
      "reward_std_all": 0.17448324710130692,
      "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": 0.5068162679672241,
      "rewards/forensics_strict_edge_reward": 0.20464354753494263,
      "rewards/forensics_tiou_reward": 0.2856948524713516,
      "step": 1155
    },
    {
      "completion_length": 15.484375,
      "epoch": 3.0745672436750997,
      "grad_norm": 31.54854965209961,
      "kl": 4.3046875,
      "learning_rate": 2.3e-07,
      "loss": 0.2178,
      "reward": 0.5985299050807953,
      "reward_std": 0.19901618734002113,
      "reward_std_all": 0.20816479623317719,
      "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": 0.5442862212657928,
      "rewards/forensics_strict_edge_reward": 0.218744158744812,
      "rewards/forensics_tiou_reward": 0.3011010214686394,
      "step": 1156
    },
    {
      "completion_length": 19.4375,
      "epoch": 3.077230359520639,
      "grad_norm": 19.286548614501953,
      "kl": 4.765625,
      "learning_rate": 2.2933333333333332e-07,
      "loss": 0.443,
      "reward": 0.6311191022396088,
      "reward_std": 0.18252551555633545,
      "reward_std_all": 0.17746400088071823,
      "rewards-max/forensics_sinkhorn_reward": 0.9883166551589966,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9973444640636444,
      "rewards/forensics_sinkhorn_reward": 0.5439874976873398,
      "rewards/forensics_strict_edge_reward": 0.27197425067424774,
      "rewards/forensics_tiou_reward": 0.3809930831193924,
      "step": 1157
    },
    {
      "completion_length": 18.296875,
      "epoch": 3.0798934753661786,
      "grad_norm": 34.616859436035156,
      "kl": 7.171875,
      "learning_rate": 2.2866666666666665e-07,
      "loss": 0.3479,
      "reward": 0.6375041902065277,
      "reward_std": 0.15393535792827606,
      "reward_std_all": 0.15400328487157822,
      "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": 0.559990644454956,
      "rewards/forensics_strict_edge_reward": 0.356603667140007,
      "rewards/forensics_tiou_reward": 0.39639394730329514,
      "step": 1158
    },
    {
      "completion_length": 18.140625,
      "epoch": 3.0825565912117177,
      "grad_norm": 20.92875099182129,
      "kl": 5.328125,
      "learning_rate": 2.28e-07,
      "loss": 0.4128,
      "reward": 0.538802906870842,
      "reward_std": 0.19497571885585785,
      "reward_std_all": 0.2271845042705536,
      "rewards-max/forensics_sinkhorn_reward": 0.9883315861225128,
      "rewards-max/forensics_strict_edge_reward": 0.9782643616199493,
      "rewards-max/forensics_tiou_reward": 0.9974026083946228,
      "rewards/forensics_sinkhorn_reward": 0.5026014447212219,
      "rewards/forensics_strict_edge_reward": 0.28995800018310547,
      "rewards/forensics_tiou_reward": 0.35870276391506195,
      "step": 1159
    },
    {
      "completion_length": 19.96875,
      "epoch": 3.085219707057257,
      "grad_norm": 29.464365005493164,
      "kl": 3.4140625,
      "learning_rate": 2.2733333333333333e-07,
      "loss": 0.322,
      "reward": 0.5849408805370331,
      "reward_std": 0.1784478724002838,
      "reward_std_all": 0.19140931963920593,
      "rewards-max/forensics_sinkhorn_reward": 0.9668294191360474,
      "rewards-max/forensics_strict_edge_reward": 0.8394034504890442,
      "rewards-max/forensics_tiou_reward": 0.9504504203796387,
      "rewards/forensics_sinkhorn_reward": 0.5246277749538422,
      "rewards/forensics_strict_edge_reward": 0.2322055846452713,
      "rewards/forensics_tiou_reward": 0.35052286088466644,
      "step": 1160
    },
    {
      "completion_length": 16.578125,
      "epoch": 3.087882822902796,
      "grad_norm": 32.331687927246094,
      "kl": 6.765625,
      "learning_rate": 2.2666666666666663e-07,
      "loss": 0.5636,
      "reward": 0.5694147050380707,
      "reward_std": 0.21399638801813126,
      "reward_std_all": 0.2319381833076477,
      "rewards-max/forensics_sinkhorn_reward": 0.9882119596004486,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9969325661659241,
      "rewards/forensics_sinkhorn_reward": 0.5414078831672668,
      "rewards/forensics_strict_edge_reward": 0.3364032953977585,
      "rewards/forensics_tiou_reward": 0.41249993443489075,
      "step": 1161
    },
    {
      "completion_length": 16.75,
      "epoch": 3.0905459387483356,
      "grad_norm": 24.61953353881836,
      "kl": 5.375,
      "learning_rate": 2.2599999999999999e-07,
      "loss": 0.3646,
      "reward": 0.5670535564422607,
      "reward_std": 0.17374026775360107,
      "reward_std_all": 0.20069725066423416,
      "rewards-max/forensics_sinkhorn_reward": 0.9692218005657196,
      "rewards-max/forensics_strict_edge_reward": 0.8659839034080505,
      "rewards-max/forensics_tiou_reward": 0.9565639197826385,
      "rewards/forensics_sinkhorn_reward": 0.5043127834796906,
      "rewards/forensics_strict_edge_reward": 0.25224724411964417,
      "rewards/forensics_tiou_reward": 0.33675873279571533,
      "step": 1162
    },
    {
      "completion_length": 17.359375,
      "epoch": 3.0932090545938746,
      "grad_norm": 34.72540283203125,
      "kl": 6.90625,
      "learning_rate": 2.253333333333333e-07,
      "loss": 0.5595,
      "reward": 0.4971468150615692,
      "reward_std": 0.1579844132065773,
      "reward_std_all": 0.23025556653738022,
      "rewards-max/forensics_sinkhorn_reward": 0.9885339736938477,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9982012510299683,
      "rewards/forensics_sinkhorn_reward": 0.4463774263858795,
      "rewards/forensics_strict_edge_reward": 0.22645550221204758,
      "rewards/forensics_tiou_reward": 0.2831062823534012,
      "step": 1163
    },
    {
      "completion_length": 15.875,
      "epoch": 3.095872170439414,
      "grad_norm": 31.22342872619629,
      "kl": 3.875,
      "learning_rate": 2.2466666666666664e-07,
      "loss": 0.1346,
      "reward": 0.6243109405040741,
      "reward_std": 0.21002576500177383,
      "reward_std_all": 0.19831877946853638,
      "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": 0.5680151283740997,
      "rewards/forensics_strict_edge_reward": 0.29784877598285675,
      "rewards/forensics_tiou_reward": 0.37734217941761017,
      "step": 1164
    },
    {
      "completion_length": 18.609375,
      "epoch": 3.0985352862849536,
      "grad_norm": 21.149723052978516,
      "kl": 3.8828125,
      "learning_rate": 2.24e-07,
      "loss": 0.3638,
      "reward": 0.6208570897579193,
      "reward_std": 0.2634574621915817,
      "reward_std_all": 0.23878397047519684,
      "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": 0.5599137842655182,
      "rewards/forensics_strict_edge_reward": 0.32746386528015137,
      "rewards/forensics_tiou_reward": 0.4082519859075546,
      "step": 1165
    },
    {
      "completion_length": 17.609375,
      "epoch": 3.1011984021304926,
      "grad_norm": 27.95280647277832,
      "kl": 7.53125,
      "learning_rate": 2.2333333333333332e-07,
      "loss": 0.4667,
      "reward": 0.5605846643447876,
      "reward_std": 0.21247635781764984,
      "reward_std_all": 0.2291138991713524,
      "rewards-max/forensics_sinkhorn_reward": 0.9876960515975952,
      "rewards-max/forensics_strict_edge_reward": 0.9677534997463226,
      "rewards-max/forensics_tiou_reward": 0.9949109852313995,
      "rewards/forensics_sinkhorn_reward": 0.4834374189376831,
      "rewards/forensics_strict_edge_reward": 0.25625430047512054,
      "rewards/forensics_tiou_reward": 0.3161275386810303,
      "step": 1166
    },
    {
      "completion_length": 15.703125,
      "epoch": 3.103861517976032,
      "grad_norm": 23.508403778076172,
      "kl": 6.796875,
      "learning_rate": 2.2266666666666668e-07,
      "loss": 0.4178,
      "reward": 0.5233874768018723,
      "reward_std": 0.26892784237861633,
      "reward_std_all": 0.27439314126968384,
      "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": 0.4857228249311447,
      "rewards/forensics_strict_edge_reward": 0.2878732830286026,
      "rewards/forensics_tiou_reward": 0.37456807494163513,
      "step": 1167
    },
    {
      "completion_length": 17.859375,
      "epoch": 3.106524633821571,
      "grad_norm": 20.793954849243164,
      "kl": 5.9375,
      "learning_rate": 2.22e-07,
      "loss": 0.4342,
      "reward": 0.5433806777000427,
      "reward_std": 0.22252967953681946,
      "reward_std_all": 0.24839510768651962,
      "rewards-max/forensics_sinkhorn_reward": 0.9614576101303101,
      "rewards-max/forensics_strict_edge_reward": 0.856150895357132,
      "rewards-max/forensics_tiou_reward": 0.9370233118534088,
      "rewards/forensics_sinkhorn_reward": 0.46604713797569275,
      "rewards/forensics_strict_edge_reward": 0.26886117458343506,
      "rewards/forensics_tiou_reward": 0.343226358294487,
      "step": 1168
    },
    {
      "completion_length": 17.125,
      "epoch": 3.1091877496671105,
      "grad_norm": 29.991477966308594,
      "kl": 6.15625,
      "learning_rate": 2.213333333333333e-07,
      "loss": 0.4134,
      "reward": 0.5765966773033142,
      "reward_std": 0.20306870341300964,
      "reward_std_all": 0.21315882354974747,
      "rewards-max/forensics_sinkhorn_reward": 0.988776445388794,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9991561472415924,
      "rewards/forensics_sinkhorn_reward": 0.4861319363117218,
      "rewards/forensics_strict_edge_reward": 0.28336693346500397,
      "rewards/forensics_tiou_reward": 0.3424646109342575,
      "step": 1169
    },
    {
      "completion_length": 17.59375,
      "epoch": 3.11185086551265,
      "grad_norm": 18.862131118774414,
      "kl": 5.65625,
      "learning_rate": 2.2066666666666666e-07,
      "loss": 0.4112,
      "reward": 0.5688831806182861,
      "reward_std": 0.24404596537351608,
      "reward_std_all": 0.23959839344024658,
      "rewards-max/forensics_sinkhorn_reward": 0.9843749105930328,
      "rewards-max/forensics_strict_edge_reward": 0.9890114367008209,
      "rewards-max/forensics_tiou_reward": 0.9901298582553864,
      "rewards/forensics_sinkhorn_reward": 0.5557841658592224,
      "rewards/forensics_strict_edge_reward": 0.3404455631971359,
      "rewards/forensics_tiou_reward": 0.41723765432834625,
      "step": 1170
    },
    {
      "completion_length": 17.015625,
      "epoch": 3.114513981358189,
      "grad_norm": 20.593992233276367,
      "kl": 4.88671875,
      "learning_rate": 2.1999999999999998e-07,
      "loss": 0.3414,
      "reward": 0.612665593624115,
      "reward_std": 0.1492643877863884,
      "reward_std_all": 0.1743037849664688,
      "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": 0.5779799818992615,
      "rewards/forensics_strict_edge_reward": 0.30510516464710236,
      "rewards/forensics_tiou_reward": 0.3821113407611847,
      "step": 1171
    },
    {
      "completion_length": 16.75,
      "epoch": 3.1171770972037285,
      "grad_norm": 19.212886810302734,
      "kl": 5.203125,
      "learning_rate": 2.193333333333333e-07,
      "loss": 0.4233,
      "reward": 0.5690423250198364,
      "reward_std": 0.16456321626901627,
      "reward_std_all": 0.199518583714962,
      "rewards-max/forensics_sinkhorn_reward": 0.9878269135951996,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9954131841659546,
      "rewards/forensics_sinkhorn_reward": 0.5056183934211731,
      "rewards/forensics_strict_edge_reward": 0.27183645963668823,
      "rewards/forensics_tiou_reward": 0.34001632034778595,
      "step": 1172
    },
    {
      "completion_length": 15.953125,
      "epoch": 3.1198402130492675,
      "grad_norm": 29.365575790405273,
      "kl": 3.4609375,
      "learning_rate": 2.1866666666666667e-07,
      "loss": 0.3255,
      "reward": 0.6190643310546875,
      "reward_std": 0.1888679787516594,
      "reward_std_all": 0.19748619198799133,
      "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": 0.5787836909294128,
      "rewards/forensics_strict_edge_reward": 0.39109621942043304,
      "rewards/forensics_tiou_reward": 0.4607004225254059,
      "step": 1173
    },
    {
      "completion_length": 19.90625,
      "epoch": 3.122503328894807,
      "grad_norm": 21.938526153564453,
      "kl": 7.453125,
      "learning_rate": 2.18e-07,
      "loss": 0.5511,
      "reward": 0.45254020392894745,
      "reward_std": 0.26474209129810333,
      "reward_std_all": 0.295906662940979,
      "rewards-max/forensics_sinkhorn_reward": 0.9639594256877899,
      "rewards-max/forensics_strict_edge_reward": 0.8578422963619232,
      "rewards-max/forensics_tiou_reward": 0.9440245032310486,
      "rewards/forensics_sinkhorn_reward": 0.3504768908023834,
      "rewards/forensics_strict_edge_reward": 0.1129644364118576,
      "rewards/forensics_tiou_reward": 0.21327674016356468,
      "step": 1174
    },
    {
      "completion_length": 17.46875,
      "epoch": 3.125166444740346,
      "grad_norm": 35.191932678222656,
      "kl": 4.5390625,
      "learning_rate": 2.1733333333333332e-07,
      "loss": 0.4051,
      "reward": 0.5938585102558136,
      "reward_std": 0.19132135808467865,
      "reward_std_all": 0.21106142550706863,
      "rewards-max/forensics_sinkhorn_reward": 0.9883259832859039,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9973803460597992,
      "rewards/forensics_sinkhorn_reward": 0.5580850541591644,
      "rewards/forensics_strict_edge_reward": 0.31321631371974945,
      "rewards/forensics_tiou_reward": 0.4344744086265564,
      "step": 1175
    },
    {
      "completion_length": 18.65625,
      "epoch": 3.1278295605858855,
      "grad_norm": 22.666114807128906,
      "kl": 5.078125,
      "learning_rate": 2.1666666666666667e-07,
      "loss": 0.4426,
      "reward": 0.5661707520484924,
      "reward_std": 0.21320128440856934,
      "reward_std_all": 0.22313234210014343,
      "rewards-max/forensics_sinkhorn_reward": 0.9868585765361786,
      "rewards-max/forensics_strict_edge_reward": 0.9565287232398987,
      "rewards-max/forensics_tiou_reward": 0.9916075170040131,
      "rewards/forensics_sinkhorn_reward": 0.48339712619781494,
      "rewards/forensics_strict_edge_reward": 0.24256566166877747,
      "rewards/forensics_tiou_reward": 0.28129693120718,
      "step": 1176
    },
    {
      "completion_length": 18.484375,
      "epoch": 3.130492676431425,
      "grad_norm": 27.2955322265625,
      "kl": 4.140625,
      "learning_rate": 2.1599999999999998e-07,
      "loss": 0.3295,
      "reward": 0.6544478535652161,
      "reward_std": 0.16966208070516586,
      "reward_std_all": 0.16480344906449318,
      "rewards-max/forensics_sinkhorn_reward": 0.9871951043605804,
      "rewards-max/forensics_strict_edge_reward": 0.9565287232398987,
      "rewards-max/forensics_tiou_reward": 0.9929230809211731,
      "rewards/forensics_sinkhorn_reward": 0.5388416945934296,
      "rewards/forensics_strict_edge_reward": 0.32285653054714203,
      "rewards/forensics_tiou_reward": 0.3702451288700104,
      "step": 1177
    },
    {
      "completion_length": 17.953125,
      "epoch": 3.133155792276964,
      "grad_norm": 47.69389343261719,
      "kl": 4.796875,
      "learning_rate": 2.153333333333333e-07,
      "loss": 0.2155,
      "reward": 0.600585401058197,
      "reward_std": 0.20634827762842178,
      "reward_std_all": 0.20551562309265137,
      "rewards-max/forensics_sinkhorn_reward": 0.9880286455154419,
      "rewards-max/forensics_strict_edge_reward": 0.9565287232398987,
      "rewards-max/forensics_tiou_reward": 0.996208906173706,
      "rewards/forensics_sinkhorn_reward": 0.5318887829780579,
      "rewards/forensics_strict_edge_reward": 0.22828665375709534,
      "rewards/forensics_tiou_reward": 0.2989092618227005,
      "step": 1178
    },
    {
      "completion_length": 17.65625,
      "epoch": 3.1358189081225034,
      "grad_norm": 26.08708953857422,
      "kl": 5.515625,
      "learning_rate": 2.1466666666666666e-07,
      "loss": 0.3718,
      "reward": 0.6937127709388733,
      "reward_std": 0.13136275857686996,
      "reward_std_all": 0.1286911591887474,
      "rewards-max/forensics_sinkhorn_reward": 0.9661546349525452,
      "rewards-max/forensics_strict_edge_reward": 0.9375866651535034,
      "rewards-max/forensics_tiou_reward": 0.9520201981067657,
      "rewards/forensics_sinkhorn_reward": 0.633285254240036,
      "rewards/forensics_strict_edge_reward": 0.3345993906259537,
      "rewards/forensics_tiou_reward": 0.4228741228580475,
      "step": 1179
    },
    {
      "completion_length": 14.671875,
      "epoch": 3.1384820239680424,
      "grad_norm": 20.301395416259766,
      "kl": 1.5625,
      "learning_rate": 2.1399999999999998e-07,
      "loss": 0.1913,
      "reward": 0.6825176179409027,
      "reward_std": 0.14771553874015808,
      "reward_std_all": 0.14099136739969254,
      "rewards-max/forensics_sinkhorn_reward": 0.9882545471191406,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9971017837524414,
      "rewards/forensics_sinkhorn_reward": 0.6927598416805267,
      "rewards/forensics_strict_edge_reward": 0.4778072386980057,
      "rewards/forensics_tiou_reward": 0.5516242831945419,
      "step": 1180
    },
    {
      "completion_length": 16.578125,
      "epoch": 3.141145139813582,
      "grad_norm": 17.99130630493164,
      "kl": 4.0390625,
      "learning_rate": 2.1333333333333334e-07,
      "loss": 0.2636,
      "reward": 0.5951874256134033,
      "reward_std": 0.1628006175160408,
      "reward_std_all": 0.1882273405790329,
      "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": 0.5887959599494934,
      "rewards/forensics_strict_edge_reward": 0.3883242756128311,
      "rewards/forensics_tiou_reward": 0.4750652015209198,
      "step": 1181
    },
    {
      "completion_length": 17.390625,
      "epoch": 3.1438082556591214,
      "grad_norm": 18.53567886352539,
      "kl": 3.5546875,
      "learning_rate": 2.1266666666666667e-07,
      "loss": 0.2412,
      "reward": 0.6585739254951477,
      "reward_std": 0.21064648032188416,
      "reward_std_all": 0.1925230771303177,
      "rewards-max/forensics_sinkhorn_reward": 0.9886516034603119,
      "rewards-max/forensics_strict_edge_reward": 0.9890114367008209,
      "rewards-max/forensics_tiou_reward": 0.9986648857593536,
      "rewards/forensics_sinkhorn_reward": 0.6309359669685364,
      "rewards/forensics_strict_edge_reward": 0.4389225244522095,
      "rewards/forensics_tiou_reward": 0.4967937171459198,
      "step": 1182
    },
    {
      "completion_length": 17.5,
      "epoch": 3.1464713715046604,
      "grad_norm": 22.84857940673828,
      "kl": 6.09375,
      "learning_rate": 2.12e-07,
      "loss": 0.5093,
      "reward": 0.607864499092102,
      "reward_std": 0.19062937796115875,
      "reward_std_all": 0.19427241384983063,
      "rewards-max/forensics_sinkhorn_reward": 0.9240227043628693,
      "rewards-max/forensics_strict_edge_reward": 0.6899373829364777,
      "rewards-max/forensics_tiou_reward": 0.8529405295848846,
      "rewards/forensics_sinkhorn_reward": 0.5085396319627762,
      "rewards/forensics_strict_edge_reward": 0.17538471519947052,
      "rewards/forensics_tiou_reward": 0.26863573491573334,
      "step": 1183
    },
    {
      "completion_length": 20.0,
      "epoch": 3.1491344873502,
      "grad_norm": 18.971954345703125,
      "kl": 6.15625,
      "learning_rate": 2.1133333333333335e-07,
      "loss": 0.4524,
      "reward": 0.5454016923904419,
      "reward_std": 0.2949676364660263,
      "reward_std_all": 0.28036946058273315,
      "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": 0.48586341738700867,
      "rewards/forensics_strict_edge_reward": 0.30517834424972534,
      "rewards/forensics_tiou_reward": 0.380533903837204,
      "step": 1184
    },
    {
      "completion_length": 17.25,
      "epoch": 3.151797603195739,
      "grad_norm": 33.75282287597656,
      "kl": 5.15625,
      "learning_rate": 2.1066666666666665e-07,
      "loss": 0.2528,
      "reward": 0.5665039122104645,
      "reward_std": 0.1461799256503582,
      "reward_std_all": 0.18206921964883804,
      "rewards-max/forensics_sinkhorn_reward": 0.988383561372757,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9976083338260651,
      "rewards/forensics_sinkhorn_reward": 0.5094554722309113,
      "rewards/forensics_strict_edge_reward": 0.25785408914089203,
      "rewards/forensics_tiou_reward": 0.3094788044691086,
      "step": 1185
    },
    {
      "completion_length": 15.640625,
      "epoch": 3.1544607190412783,
      "grad_norm": 17.228378295898438,
      "kl": 3.265625,
      "learning_rate": 2.0999999999999997e-07,
      "loss": 0.129,
      "reward": 0.6560748517513275,
      "reward_std": 0.09428425505757332,
      "reward_std_all": 0.11512427031993866,
      "rewards-max/forensics_sinkhorn_reward": 0.9862719774246216,
      "rewards-max/forensics_strict_edge_reward": 0.9357379674911499,
      "rewards-max/forensics_tiou_reward": 0.9893210828304291,
      "rewards/forensics_sinkhorn_reward": 0.6071712076663971,
      "rewards/forensics_strict_edge_reward": 0.2470739483833313,
      "rewards/forensics_tiou_reward": 0.371037557721138,
      "step": 1186
    },
    {
      "completion_length": 16.515625,
      "epoch": 3.157123834886818,
      "grad_norm": 22.093141555786133,
      "kl": 4.0859375,
      "learning_rate": 2.0933333333333333e-07,
      "loss": 0.2585,
      "reward": 0.5776878893375397,
      "reward_std": 0.18720989674329758,
      "reward_std_all": 0.20237787812948227,
      "rewards-max/forensics_sinkhorn_reward": 0.9870838224887848,
      "rewards-max/forensics_strict_edge_reward": 0.9464850425720215,
      "rewards-max/forensics_tiou_reward": 0.9925007224082947,
      "rewards/forensics_sinkhorn_reward": 0.5265244841575623,
      "rewards/forensics_strict_edge_reward": 0.2925306707620621,
      "rewards/forensics_tiou_reward": 0.3525290787220001,
      "step": 1187
    },
    {
      "completion_length": 15.515625,
      "epoch": 3.159786950732357,
      "grad_norm": 18.464374542236328,
      "kl": 5.078125,
      "learning_rate": 2.0866666666666666e-07,
      "loss": 0.3003,
      "reward": 0.6290570497512817,
      "reward_std": 0.20633628964424133,
      "reward_std_all": 0.19761023670434952,
      "rewards-max/forensics_sinkhorn_reward": 0.9874483942985535,
      "rewards-max/forensics_strict_edge_reward": 0.9565287232398987,
      "rewards-max/forensics_tiou_reward": 0.9939287304878235,
      "rewards/forensics_sinkhorn_reward": 0.6247256696224213,
      "rewards/forensics_strict_edge_reward": 0.33140213787555695,
      "rewards/forensics_tiou_reward": 0.45208093523979187,
      "step": 1188
    },
    {
      "completion_length": 17.171875,
      "epoch": 3.1624500665778963,
      "grad_norm": 85.6341323852539,
      "kl": 7.34375,
      "learning_rate": 2.0799999999999998e-07,
      "loss": 0.3393,
      "reward": 0.6573644876480103,
      "reward_std": 0.15240908414125443,
      "reward_std_all": 0.1516372114419937,
      "rewards-max/forensics_sinkhorn_reward": 0.9879127442836761,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9957699179649353,
      "rewards/forensics_sinkhorn_reward": 0.5736921727657318,
      "rewards/forensics_strict_edge_reward": 0.3573896586894989,
      "rewards/forensics_tiou_reward": 0.4276326596736908,
      "step": 1189
    },
    {
      "completion_length": 16.28125,
      "epoch": 3.1651131824234353,
      "grad_norm": 24.033226013183594,
      "kl": 3.890625,
      "learning_rate": 2.0733333333333334e-07,
      "loss": 0.3271,
      "reward": 0.6178151667118073,
      "reward_std": 0.149392768740654,
      "reward_std_all": 0.17512784898281097,
      "rewards-max/forensics_sinkhorn_reward": 0.9886096119880676,
      "rewards-max/forensics_strict_edge_reward": 0.9890114367008209,
      "rewards-max/forensics_tiou_reward": 0.9985063076019287,
      "rewards/forensics_sinkhorn_reward": 0.6164839863777161,
      "rewards/forensics_strict_edge_reward": 0.37968941032886505,
      "rewards/forensics_tiou_reward": 0.4791237562894821,
      "step": 1190
    },
    {
      "completion_length": 16.890625,
      "epoch": 3.1677762982689748,
      "grad_norm": 28.692413330078125,
      "kl": 5.9296875,
      "learning_rate": 2.0666666666666666e-07,
      "loss": 0.2547,
      "reward": 0.6025573015213013,
      "reward_std": 0.16825393214821815,
      "reward_std_all": 0.18860753998160362,
      "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": 0.5464658141136169,
      "rewards/forensics_strict_edge_reward": 0.2548355907201767,
      "rewards/forensics_tiou_reward": 0.32111290097236633,
      "step": 1191
    },
    {
      "completion_length": 15.46875,
      "epoch": 3.170439414114514,
      "grad_norm": 24.731201171875,
      "kl": 3.7890625,
      "learning_rate": 2.06e-07,
      "loss": 0.2977,
      "reward": 0.6114871203899384,
      "reward_std": 0.13638099282979965,
      "reward_std_all": 0.16139209270477295,
      "rewards-max/forensics_sinkhorn_reward": 0.926262766122818,
      "rewards-max/forensics_strict_edge_reward": 0.7939938604831696,
      "rewards-max/forensics_tiou_reward": 0.8577475547790527,
      "rewards/forensics_sinkhorn_reward": 0.5554586052894592,
      "rewards/forensics_strict_edge_reward": 0.2454523853957653,
      "rewards/forensics_tiou_reward": 0.31523575633764267,
      "step": 1192
    },
    {
      "completion_length": 16.515625,
      "epoch": 3.1731025299600533,
      "grad_norm": 25.282737731933594,
      "kl": 3.75,
      "learning_rate": 2.0533333333333332e-07,
      "loss": 0.1706,
      "reward": 0.6586540639400482,
      "reward_std": 0.17920373380184174,
      "reward_std_all": 0.17279495298862457,
      "rewards-max/forensics_sinkhorn_reward": 0.9887953698635101,
      "rewards-max/forensics_strict_edge_reward": 0.9782643616199493,
      "rewards-max/forensics_tiou_reward": 0.9992307722568512,
      "rewards/forensics_sinkhorn_reward": 0.6329951882362366,
      "rewards/forensics_strict_edge_reward": 0.32365506887435913,
      "rewards/forensics_tiou_reward": 0.43028439581394196,
      "step": 1193
    },
    {
      "completion_length": 15.296875,
      "epoch": 3.1757656458055927,
      "grad_norm": 19.03988265991211,
      "kl": 2.60546875,
      "learning_rate": 2.0466666666666665e-07,
      "loss": 0.076,
      "reward": 0.6486682891845703,
      "reward_std": 0.15117311477661133,
      "reward_std_all": 0.15280558168888092,
      "rewards-max/forensics_sinkhorn_reward": 0.9882653951644897,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9971436858177185,
      "rewards/forensics_sinkhorn_reward": 0.6090383529663086,
      "rewards/forensics_strict_edge_reward": 0.351261168718338,
      "rewards/forensics_tiou_reward": 0.4564705491065979,
      "step": 1194
    },
    {
      "completion_length": 17.015625,
      "epoch": 3.1784287616511318,
      "grad_norm": 22.168506622314453,
      "kl": 4.9296875,
      "learning_rate": 2.0399999999999997e-07,
      "loss": 0.3803,
      "reward": 0.6625836789608002,
      "reward_std": 0.18409037590026855,
      "reward_std_all": 0.1707904115319252,
      "rewards-max/forensics_sinkhorn_reward": 0.9886549413204193,
      "rewards-max/forensics_strict_edge_reward": 0.9782643616199493,
      "rewards-max/forensics_tiou_reward": 0.9986781179904938,
      "rewards/forensics_sinkhorn_reward": 0.6243528425693512,
      "rewards/forensics_strict_edge_reward": 0.3763623461127281,
      "rewards/forensics_tiou_reward": 0.45235469937324524,
      "step": 1195
    },
    {
      "completion_length": 20.046875,
      "epoch": 3.181091877496671,
      "grad_norm": 25.4213809967041,
      "kl": 4.59375,
      "learning_rate": 2.0333333333333333e-07,
      "loss": 0.232,
      "reward": 0.5999941527843475,
      "reward_std": 0.1822226569056511,
      "reward_std_all": 0.1919747218489647,
      "rewards-max/forensics_sinkhorn_reward": 0.9294722080230713,
      "rewards-max/forensics_strict_edge_reward": 0.7825796902179718,
      "rewards-max/forensics_tiou_reward": 0.8673684597015381,
      "rewards/forensics_sinkhorn_reward": 0.4945897310972214,
      "rewards/forensics_strict_edge_reward": 0.19083420187234879,
      "rewards/forensics_tiou_reward": 0.2863895893096924,
      "step": 1196
    },
    {
      "completion_length": 16.703125,
      "epoch": 3.1837549933422102,
      "grad_norm": 20.771984100341797,
      "kl": 3.09375,
      "learning_rate": 2.0266666666666666e-07,
      "loss": 0.1394,
      "reward": 0.6440703570842743,
      "reward_std": 0.10291292518377304,
      "reward_std_all": 0.12862756848335266,
      "rewards-max/forensics_sinkhorn_reward": 0.9882833659648895,
      "rewards-max/forensics_strict_edge_reward": 0.9185642302036285,
      "rewards-max/forensics_tiou_reward": 0.9972125291824341,
      "rewards/forensics_sinkhorn_reward": 0.6048553884029388,
      "rewards/forensics_strict_edge_reward": 0.2758808732032776,
      "rewards/forensics_tiou_reward": 0.4250285029411316,
      "step": 1197
    },
    {
      "completion_length": 18.1875,
      "epoch": 3.1864181091877497,
      "grad_norm": 25.938331604003906,
      "kl": 4.53125,
      "learning_rate": 2.02e-07,
      "loss": 0.2767,
      "reward": 0.6800304651260376,
      "reward_std": 0.07717549055814743,
      "reward_std_all": 0.10053546726703644,
      "rewards-max/forensics_sinkhorn_reward": 0.9881721436977386,
      "rewards-max/forensics_strict_edge_reward": 0.9574736058712006,
      "rewards-max/forensics_tiou_reward": 0.9967741966247559,
      "rewards/forensics_sinkhorn_reward": 0.6411765813827515,
      "rewards/forensics_strict_edge_reward": 0.28941723704338074,
      "rewards/forensics_tiou_reward": 0.4271761626005173,
      "step": 1198
    },
    {
      "completion_length": 16.671875,
      "epoch": 3.1890812250332887,
      "grad_norm": 24.395526885986328,
      "kl": 3.2890625,
      "learning_rate": 2.0133333333333334e-07,
      "loss": 0.1941,
      "reward": 0.6732544004917145,
      "reward_std": 0.13315127417445183,
      "reward_std_all": 0.13615085929632187,
      "rewards-max/forensics_sinkhorn_reward": 0.9444961845874786,
      "rewards-max/forensics_strict_edge_reward": 0.8582656681537628,
      "rewards-max/forensics_tiou_reward": 0.9021155834197998,
      "rewards/forensics_sinkhorn_reward": 0.6236981451511383,
      "rewards/forensics_strict_edge_reward": 0.32109325751662254,
      "rewards/forensics_tiou_reward": 0.42426636070013046,
      "step": 1199
    },
    {
      "completion_length": 15.578125,
      "epoch": 3.191744340878828,
      "grad_norm": 21.1279296875,
      "kl": 3.625,
      "learning_rate": 2.0066666666666666e-07,
      "loss": 0.2857,
      "reward": 0.6935599446296692,
      "reward_std": 0.13841207697987556,
      "reward_std_all": 0.1307750940322876,
      "rewards-max/forensics_sinkhorn_reward": 0.9788262844085693,
      "rewards-max/forensics_strict_edge_reward": 0.9782643616199493,
      "rewards-max/forensics_tiou_reward": 0.9778402745723724,
      "rewards/forensics_sinkhorn_reward": 0.6489888727664948,
      "rewards/forensics_strict_edge_reward": 0.3292457312345505,
      "rewards/forensics_tiou_reward": 0.4541301280260086,
      "step": 1200
    },
    {
      "completion_length": 16.734375,
      "epoch": 3.1944074567243677,
      "grad_norm": 30.04999351501465,
      "kl": 5.125,
      "learning_rate": 2e-07,
      "loss": 0.3119,
      "reward": 0.6575406491756439,
      "reward_std": 0.15200548619031906,
      "reward_std_all": 0.15312744677066803,
      "rewards-max/forensics_sinkhorn_reward": 0.9881294667720795,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.996605783700943,
      "rewards/forensics_sinkhorn_reward": 0.5843701660633087,
      "rewards/forensics_strict_edge_reward": 0.2842800170183182,
      "rewards/forensics_tiou_reward": 0.37858110666275024,
      "step": 1201
    },
    {
      "completion_length": 17.28125,
      "epoch": 3.1970705725699067,
      "grad_norm": 26.833105087280273,
      "kl": 3.6640625,
      "learning_rate": 1.9933333333333332e-07,
      "loss": 0.3312,
      "reward": 0.5665581822395325,
      "reward_std": 0.15302810817956924,
      "reward_std_all": 0.19734954833984375,
      "rewards-max/forensics_sinkhorn_reward": 0.9659149050712585,
      "rewards-max/forensics_strict_edge_reward": 0.8968285918235779,
      "rewards-max/forensics_tiou_reward": 0.9446727931499481,
      "rewards/forensics_sinkhorn_reward": 0.49237385392189026,
      "rewards/forensics_strict_edge_reward": 0.23784897476434708,
      "rewards/forensics_tiou_reward": 0.28144048154354095,
      "step": 1202
    },
    {
      "completion_length": 17.890625,
      "epoch": 3.199733688415446,
      "grad_norm": 35.628971099853516,
      "kl": 4.6875,
      "learning_rate": 1.9866666666666665e-07,
      "loss": 0.3726,
      "reward": 0.5754318833351135,
      "reward_std": 0.2352202609181404,
      "reward_std_all": 0.23969823867082596,
      "rewards-max/forensics_sinkhorn_reward": 0.9883246421813965,
      "rewards-max/forensics_strict_edge_reward": 0.9782643616199493,
      "rewards-max/forensics_tiou_reward": 0.9973787665367126,
      "rewards/forensics_sinkhorn_reward": 0.5155480802059174,
      "rewards/forensics_strict_edge_reward": 0.2734922915697098,
      "rewards/forensics_tiou_reward": 0.33969762921333313,
      "step": 1203
    },
    {
      "completion_length": 17.75,
      "epoch": 3.202396804260985,
      "grad_norm": 30.876720428466797,
      "kl": 8.84375,
      "learning_rate": 1.98e-07,
      "loss": 0.5299,
      "reward": 0.5756582021713257,
      "reward_std": 0.16378632187843323,
      "reward_std_all": 0.19606424123048782,
      "rewards-max/forensics_sinkhorn_reward": 0.9280877709388733,
      "rewards-max/forensics_strict_edge_reward": 0.9075756669044495,
      "rewards-max/forensics_tiou_reward": 0.8732702136039734,
      "rewards/forensics_sinkhorn_reward": 0.46961838006973267,
      "rewards/forensics_strict_edge_reward": 0.19919777661561966,
      "rewards/forensics_tiou_reward": 0.2488531470298767,
      "step": 1204
    },
    {
      "completion_length": 16.671875,
      "epoch": 3.2050599201065246,
      "grad_norm": 51.249385833740234,
      "kl": 4.1328125,
      "learning_rate": 1.9733333333333333e-07,
      "loss": 0.2359,
      "reward": 0.6249361932277679,
      "reward_std": 0.1396934762597084,
      "reward_std_all": 0.15406456589698792,
      "rewards-max/forensics_sinkhorn_reward": 0.9874871075153351,
      "rewards-max/forensics_strict_edge_reward": 0.9062341153621674,
      "rewards-max/forensics_tiou_reward": 0.9940762221813202,
      "rewards/forensics_sinkhorn_reward": 0.5757100582122803,
      "rewards/forensics_strict_edge_reward": 0.33752500265836716,
      "rewards/forensics_tiou_reward": 0.4117782562971115,
      "step": 1205
    },
    {
      "completion_length": 17.03125,
      "epoch": 3.207723035952064,
      "grad_norm": 20.62799835205078,
      "kl": 3.7265625,
      "learning_rate": 1.9666666666666665e-07,
      "loss": 0.2589,
      "reward": 0.6275280117988586,
      "reward_std": 0.2028188407421112,
      "reward_std_all": 0.19715020805597305,
      "rewards-max/forensics_sinkhorn_reward": 0.98875692486763,
      "rewards-max/forensics_strict_edge_reward": 0.9890114367008209,
      "rewards-max/forensics_tiou_reward": 0.9990796148777008,
      "rewards/forensics_sinkhorn_reward": 0.5630334317684174,
      "rewards/forensics_strict_edge_reward": 0.32243578135967255,
      "rewards/forensics_tiou_reward": 0.4068679064512253,
      "step": 1206
    },
    {
      "completion_length": 19.078125,
      "epoch": 3.210386151797603,
      "grad_norm": 27.800031661987305,
      "kl": 6.453125,
      "learning_rate": 1.96e-07,
      "loss": 0.4906,
      "reward": 0.5842990875244141,
      "reward_std": 0.22406035661697388,
      "reward_std_all": 0.2196914255619049,
      "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": 0.5578969717025757,
      "rewards/forensics_strict_edge_reward": 0.39565131068229675,
      "rewards/forensics_tiou_reward": 0.4539960026741028,
      "step": 1207
    },
    {
      "completion_length": 16.421875,
      "epoch": 3.2130492676431426,
      "grad_norm": 34.60686111450195,
      "kl": 6.015625,
      "learning_rate": 1.953333333333333e-07,
      "loss": 0.377,
      "reward": 0.6412414610385895,
      "reward_std": 0.18210773169994354,
      "reward_std_all": 0.17754080891609192,
      "rewards-max/forensics_sinkhorn_reward": 0.9877589046955109,
      "rewards-max/forensics_strict_edge_reward": 0.9677534997463226,
      "rewards-max/forensics_tiou_reward": 0.9951574206352234,
      "rewards/forensics_sinkhorn_reward": 0.5523562431335449,
      "rewards/forensics_strict_edge_reward": 0.28278952836990356,
      "rewards/forensics_tiou_reward": 0.35537271201610565,
      "step": 1208
    },
    {
      "completion_length": 17.421875,
      "epoch": 3.2157123834886816,
      "grad_norm": 21.997802734375,
      "kl": 6.59375,
      "learning_rate": 1.9466666666666664e-07,
      "loss": 0.4056,
      "reward": 0.5311533510684967,
      "reward_std": 0.21183301508426666,
      "reward_std_all": 0.2380485087633133,
      "rewards-max/forensics_sinkhorn_reward": 0.9194383919239044,
      "rewards-max/forensics_strict_edge_reward": 0.8286563754081726,
      "rewards-max/forensics_tiou_reward": 0.8509148955345154,
      "rewards/forensics_sinkhorn_reward": 0.4153590798377991,
      "rewards/forensics_strict_edge_reward": 0.1604228913784027,
      "rewards/forensics_tiou_reward": 0.2269689217209816,
      "step": 1209
    },
    {
      "completion_length": 18.453125,
      "epoch": 3.218375499334221,
      "grad_norm": 21.02794647216797,
      "kl": 6.46875,
      "learning_rate": 1.94e-07,
      "loss": 0.4768,
      "reward": 0.5492558181285858,
      "reward_std": 0.20604278147220612,
      "reward_std_all": 0.22971989959478378,
      "rewards-max/forensics_sinkhorn_reward": 0.9855321645736694,
      "rewards-max/forensics_strict_edge_reward": 0.9567649364471436,
      "rewards-max/forensics_tiou_reward": 0.9913837909698486,
      "rewards/forensics_sinkhorn_reward": 0.47233909368515015,
      "rewards/forensics_strict_edge_reward": 0.2393401712179184,
      "rewards/forensics_tiou_reward": 0.3340423107147217,
      "step": 1210
    },
    {
      "completion_length": 15.421875,
      "epoch": 3.2210386151797605,
      "grad_norm": 29.389041900634766,
      "kl": 3.1796875,
      "learning_rate": 1.9333333333333332e-07,
      "loss": 0.28,
      "reward": 0.6811283528804779,
      "reward_std": 0.17600426822900772,
      "reward_std_all": 0.1592797338962555,
      "rewards-max/forensics_sinkhorn_reward": 0.9884407222270966,
      "rewards-max/forensics_strict_edge_reward": 0.9357379674911499,
      "rewards-max/forensics_tiou_reward": 0.9978339672088623,
      "rewards/forensics_sinkhorn_reward": 0.6232377886772156,
      "rewards/forensics_strict_edge_reward": 0.2329929694533348,
      "rewards/forensics_tiou_reward": 0.41960379481315613,
      "step": 1211
    },
    {
      "completion_length": 18.1875,
      "epoch": 3.2237017310252996,
      "grad_norm": 29.079265594482422,
      "kl": 4.7109375,
      "learning_rate": 1.9266666666666667e-07,
      "loss": 0.4108,
      "reward": 0.5948489904403687,
      "reward_std": 0.2414693981409073,
      "reward_std_all": 0.22672605514526367,
      "rewards-max/forensics_sinkhorn_reward": 0.9855710566043854,
      "rewards-max/forensics_strict_edge_reward": 0.8983768224716187,
      "rewards-max/forensics_tiou_reward": 0.9865999519824982,
      "rewards/forensics_sinkhorn_reward": 0.5352071225643158,
      "rewards/forensics_strict_edge_reward": 0.3213616758584976,
      "rewards/forensics_tiou_reward": 0.4080556482076645,
      "step": 1212
    },
    {
      "completion_length": 19.34375,
      "epoch": 3.226364846870839,
      "grad_norm": 20.692529678344727,
      "kl": 6.5625,
      "learning_rate": 1.92e-07,
      "loss": 0.5647,
      "reward": 0.5399304330348969,
      "reward_std": 0.24789734929800034,
      "reward_std_all": 0.2576475292444229,
      "rewards-max/forensics_sinkhorn_reward": 0.9866608083248138,
      "rewards-max/forensics_strict_edge_reward": 0.8668389916419983,
      "rewards-max/forensics_tiou_reward": 0.9908174872398376,
      "rewards/forensics_sinkhorn_reward": 0.4540172517299652,
      "rewards/forensics_strict_edge_reward": 0.17189917713403702,
      "rewards/forensics_tiou_reward": 0.291300892829895,
      "step": 1213
    },
    {
      "completion_length": 15.421875,
      "epoch": 3.229027962716378,
      "grad_norm": 23.015939712524414,
      "kl": 4.1796875,
      "learning_rate": 1.9133333333333333e-07,
      "loss": 0.1912,
      "reward": 0.6266287267208099,
      "reward_std": 0.1951952874660492,
      "reward_std_all": 0.18620695918798447,
      "rewards-max/forensics_sinkhorn_reward": 0.9874527752399445,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9939392507076263,
      "rewards/forensics_sinkhorn_reward": 0.5763822495937347,
      "rewards/forensics_strict_edge_reward": 0.30216819047927856,
      "rewards/forensics_tiou_reward": 0.3928109407424927,
      "step": 1214
    },
    {
      "completion_length": 16.703125,
      "epoch": 3.2316910785619175,
      "grad_norm": 24.433534622192383,
      "kl": 6.984375,
      "learning_rate": 1.9066666666666668e-07,
      "loss": 0.5715,
      "reward": 0.5860086977481842,
      "reward_std": 0.26992930471897125,
      "reward_std_all": 0.24966157227754593,
      "rewards-max/forensics_sinkhorn_reward": 0.9882451593875885,
      "rewards-max/forensics_strict_edge_reward": 0.9565287232398987,
      "rewards-max/forensics_tiou_reward": 0.9970638751983643,
      "rewards/forensics_sinkhorn_reward": 0.5609474778175354,
      "rewards/forensics_strict_edge_reward": 0.3582073450088501,
      "rewards/forensics_tiou_reward": 0.4515664726495743,
      "step": 1215
    },
    {
      "completion_length": 19.546875,
      "epoch": 3.2343541944074565,
      "grad_norm": 18.837282180786133,
      "kl": 6.3046875,
      "learning_rate": 1.8999999999999998e-07,
      "loss": 0.5416,
      "reward": 0.5620784163475037,
      "reward_std": 0.22152606397867203,
      "reward_std_all": 0.23766256868839264,
      "rewards-max/forensics_sinkhorn_reward": 0.9762621223926544,
      "rewards-max/forensics_strict_edge_reward": 0.7849180996417999,
      "rewards-max/forensics_tiou_reward": 0.9510802328586578,
      "rewards/forensics_sinkhorn_reward": 0.4501621276140213,
      "rewards/forensics_strict_edge_reward": 0.2351444661617279,
      "rewards/forensics_tiou_reward": 0.3059714436531067,
      "step": 1216
    },
    {
      "completion_length": 17.890625,
      "epoch": 3.237017310252996,
      "grad_norm": 26.946449279785156,
      "kl": 8.0625,
      "learning_rate": 1.893333333333333e-07,
      "loss": 0.5635,
      "reward": 0.556478887796402,
      "reward_std": 0.20592330396175385,
      "reward_std_all": 0.23421815782785416,
      "rewards-max/forensics_sinkhorn_reward": 0.9877296686172485,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9950299859046936,
      "rewards/forensics_sinkhorn_reward": 0.5019257962703705,
      "rewards/forensics_strict_edge_reward": 0.22108732908964157,
      "rewards/forensics_tiou_reward": 0.3836139589548111,
      "step": 1217
    },
    {
      "completion_length": 17.0625,
      "epoch": 3.2396804260985355,
      "grad_norm": 24.65777587890625,
      "kl": 5.671875,
      "learning_rate": 1.8866666666666666e-07,
      "loss": 0.4385,
      "reward": 0.6185529828071594,
      "reward_std": 0.20444023609161377,
      "reward_std_all": 0.19773833453655243,
      "rewards-max/forensics_sinkhorn_reward": 0.98659548163414,
      "rewards-max/forensics_strict_edge_reward": 0.8893801271915436,
      "rewards-max/forensics_tiou_reward": 0.9905882775783539,
      "rewards/forensics_sinkhorn_reward": 0.5347739905118942,
      "rewards/forensics_strict_edge_reward": 0.300188809633255,
      "rewards/forensics_tiou_reward": 0.37072739005088806,
      "step": 1218
    },
    {
      "completion_length": 16.984375,
      "epoch": 3.2423435419440745,
      "grad_norm": 33.04224395751953,
      "kl": 6.125,
      "learning_rate": 1.88e-07,
      "loss": 0.3619,
      "reward": 0.5878406167030334,
      "reward_std": 0.2099798321723938,
      "reward_std_all": 0.21253522485494614,
      "rewards-max/forensics_sinkhorn_reward": 0.9883109331130981,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9973219931125641,
      "rewards/forensics_sinkhorn_reward": 0.5397329330444336,
      "rewards/forensics_strict_edge_reward": 0.27142470329999924,
      "rewards/forensics_tiou_reward": 0.3532594293355942,
      "step": 1219
    },
    {
      "completion_length": 19.5,
      "epoch": 3.245006657789614,
      "grad_norm": 28.3829288482666,
      "kl": 7.796875,
      "learning_rate": 1.8733333333333332e-07,
      "loss": 0.5784,
      "reward": 0.49086688458919525,
      "reward_std": 0.24451259523630142,
      "reward_std_all": 0.2773198038339615,
      "rewards-max/forensics_sinkhorn_reward": 0.942026823759079,
      "rewards-max/forensics_strict_edge_reward": 0.8096016347408295,
      "rewards-max/forensics_tiou_reward": 0.8934427201747894,
      "rewards/forensics_sinkhorn_reward": 0.43281450867652893,
      "rewards/forensics_strict_edge_reward": 0.2387053519487381,
      "rewards/forensics_tiou_reward": 0.31526556611061096,
      "step": 1220
    },
    {
      "completion_length": 15.921875,
      "epoch": 3.247669773635153,
      "grad_norm": 30.731260299682617,
      "kl": 7.046875,
      "learning_rate": 1.8666666666666667e-07,
      "loss": 0.4726,
      "reward": 0.5503799170255661,
      "reward_std": 0.2093988135457039,
      "reward_std_all": 0.23337913304567337,
      "rewards-max/forensics_sinkhorn_reward": 0.9691420197486877,
      "rewards-max/forensics_strict_edge_reward": 0.9048932790756226,
      "rewards-max/forensics_tiou_reward": 0.95464026927948,
      "rewards/forensics_sinkhorn_reward": 0.46667732298374176,
      "rewards/forensics_strict_edge_reward": 0.21460934728384018,
      "rewards/forensics_tiou_reward": 0.2845928445458412,
      "step": 1221
    },
    {
      "completion_length": 18.78125,
      "epoch": 3.2503328894806924,
      "grad_norm": 17.40241241455078,
      "kl": 5.390625,
      "learning_rate": 1.86e-07,
      "loss": 0.3524,
      "reward": 0.5918190777301788,
      "reward_std": 0.11187780648469925,
      "reward_std_all": 0.16074584424495697,
      "rewards-max/forensics_sinkhorn_reward": 0.9884147644042969,
      "rewards-max/forensics_strict_edge_reward": 0.9464850425720215,
      "rewards-max/forensics_tiou_reward": 0.9977304637432098,
      "rewards/forensics_sinkhorn_reward": 0.5458887219429016,
      "rewards/forensics_strict_edge_reward": 0.24616824090480804,
      "rewards/forensics_tiou_reward": 0.35667093098163605,
      "step": 1222
    },
    {
      "completion_length": 18.3125,
      "epoch": 3.2529960053262315,
      "grad_norm": 29.309858322143555,
      "kl": 5.328125,
      "learning_rate": 1.8533333333333333e-07,
      "loss": 0.2154,
      "reward": 0.6181887686252594,
      "reward_std": 0.21088769286870956,
      "reward_std_all": 0.19982929527759552,
      "rewards-max/forensics_sinkhorn_reward": 0.9435405135154724,
      "rewards-max/forensics_strict_edge_reward": 0.8371284604072571,
      "rewards-max/forensics_tiou_reward": 0.8983504176139832,
      "rewards/forensics_sinkhorn_reward": 0.4662743806838989,
      "rewards/forensics_strict_edge_reward": 0.17033156007528305,
      "rewards/forensics_tiou_reward": 0.2613372728228569,
      "step": 1223
    },
    {
      "completion_length": 18.65625,
      "epoch": 3.255659121171771,
      "grad_norm": 22.816774368286133,
      "kl": 5.609375,
      "learning_rate": 1.8466666666666665e-07,
      "loss": 0.4296,
      "reward": 0.5927791595458984,
      "reward_std": 0.20755236595869064,
      "reward_std_all": 0.20634575188159943,
      "rewards-max/forensics_sinkhorn_reward": 0.9871238172054291,
      "rewards-max/forensics_strict_edge_reward": 0.9567649364471436,
      "rewards-max/forensics_tiou_reward": 0.9926637709140778,
      "rewards/forensics_sinkhorn_reward": 0.5314174145460129,
      "rewards/forensics_strict_edge_reward": 0.24575458467006683,
      "rewards/forensics_tiou_reward": 0.3568399101495743,
      "step": 1224
    },
    {
      "completion_length": 16.640625,
      "epoch": 3.2583222370173104,
      "grad_norm": 879.9767456054688,
      "kl": 19.828125,
      "learning_rate": 1.8399999999999998e-07,
      "loss": 0.9184,
      "reward": 0.6572470963001251,
      "reward_std": 0.16371789947152138,
      "reward_std_all": 0.15860754624009132,
      "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": 0.5850523114204407,
      "rewards/forensics_strict_edge_reward": 0.3492797464132309,
      "rewards/forensics_tiou_reward": 0.3924272805452347,
      "step": 1225
    },
    {
      "completion_length": 19.1875,
      "epoch": 3.2609853528628494,
      "grad_norm": 18.319171905517578,
      "kl": 5.484375,
      "learning_rate": 1.833333333333333e-07,
      "loss": 0.2835,
      "reward": 0.5997489094734192,
      "reward_std": 0.17926323413848877,
      "reward_std_all": 0.18750455975532532,
      "rewards-max/forensics_sinkhorn_reward": 0.9850327372550964,
      "rewards-max/forensics_strict_edge_reward": 0.8968285918235779,
      "rewards-max/forensics_tiou_reward": 0.9844632744789124,
      "rewards/forensics_sinkhorn_reward": 0.523393452167511,
      "rewards/forensics_strict_edge_reward": 0.24704834818840027,
      "rewards/forensics_tiou_reward": 0.3598121255636215,
      "step": 1226
    },
    {
      "completion_length": 17.171875,
      "epoch": 3.263648468708389,
      "grad_norm": 25.13123893737793,
      "kl": 4.375,
      "learning_rate": 1.8266666666666666e-07,
      "loss": 0.2993,
      "reward": 0.6032960116863251,
      "reward_std": 0.22061879932880402,
      "reward_std_all": 0.21460938453674316,
      "rewards-max/forensics_sinkhorn_reward": 0.9862519204616547,
      "rewards-max/forensics_strict_edge_reward": 0.947419673204422,
      "rewards-max/forensics_tiou_reward": 0.9892668724060059,
      "rewards/forensics_sinkhorn_reward": 0.5334477722644806,
      "rewards/forensics_strict_edge_reward": 0.23250335454940796,
      "rewards/forensics_tiou_reward": 0.33421941101551056,
      "step": 1227
    },
    {
      "completion_length": 17.109375,
      "epoch": 3.266311584553928,
      "grad_norm": 29.47016716003418,
      "kl": 2.8984375,
      "learning_rate": 1.82e-07,
      "loss": 0.3081,
      "reward": 0.5906400084495544,
      "reward_std": 0.19281378388404846,
      "reward_std_all": 0.20352907478809357,
      "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": 0.5530253052711487,
      "rewards/forensics_strict_edge_reward": 0.34326334297657013,
      "rewards/forensics_tiou_reward": 0.38659901916980743,
      "step": 1228
    },
    {
      "completion_length": 16.640625,
      "epoch": 3.2689747003994674,
      "grad_norm": 17.790956497192383,
      "kl": 3.2421875,
      "learning_rate": 1.8133333333333334e-07,
      "loss": 0.1181,
      "reward": 0.6586337685585022,
      "reward_std": 0.14108359068632126,
      "reward_std_all": 0.1470627635717392,
      "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": 0.6085567772388458,
      "rewards/forensics_strict_edge_reward": 0.26558302342891693,
      "rewards/forensics_tiou_reward": 0.3736872225999832,
      "step": 1229
    },
    {
      "completion_length": 17.15625,
      "epoch": 3.271637816245007,
      "grad_norm": 189.5292510986328,
      "kl": 10.046875,
      "learning_rate": 1.8066666666666667e-07,
      "loss": 0.6019,
      "reward": 0.6328640282154083,
      "reward_std": 0.2151123806834221,
      "reward_std_all": 0.20091626048088074,
      "rewards-max/forensics_sinkhorn_reward": 0.9847122430801392,
      "rewards-max/forensics_strict_edge_reward": 0.9375866651535034,
      "rewards-max/forensics_tiou_reward": 0.9832881093025208,
      "rewards/forensics_sinkhorn_reward": 0.6079829633235931,
      "rewards/forensics_strict_edge_reward": 0.2696617394685745,
      "rewards/forensics_tiou_reward": 0.4386206269264221,
      "step": 1230
    },
    {
      "completion_length": 16.1875,
      "epoch": 3.274300932090546,
      "grad_norm": 25.161846160888672,
      "kl": 4.359375,
      "learning_rate": 1.8e-07,
      "loss": 0.3811,
      "reward": 0.5935841202735901,
      "reward_std": 0.2331017479300499,
      "reward_std_all": 0.22411486506462097,
      "rewards-max/forensics_sinkhorn_reward": 0.987035721540451,
      "rewards-max/forensics_strict_edge_reward": 0.9567649364471436,
      "rewards-max/forensics_tiou_reward": 0.9923133850097656,
      "rewards/forensics_sinkhorn_reward": 0.5534636378288269,
      "rewards/forensics_strict_edge_reward": 0.251481756567955,
      "rewards/forensics_tiou_reward": 0.38217365741729736,
      "step": 1231
    },
    {
      "completion_length": 15.359375,
      "epoch": 3.2769640479360853,
      "grad_norm": 25.36551284790039,
      "kl": 3.1796875,
      "learning_rate": 1.7933333333333332e-07,
      "loss": 0.2539,
      "reward": 0.6430304646492004,
      "reward_std": 0.19536947458982468,
      "reward_std_all": 0.18452690541744232,
      "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": 0.6166074573993683,
      "rewards/forensics_strict_edge_reward": 0.38096268475055695,
      "rewards/forensics_tiou_reward": 0.44662442803382874,
      "step": 1232
    },
    {
      "completion_length": 18.078125,
      "epoch": 3.2796271637816243,
      "grad_norm": 36.27415466308594,
      "kl": 4.1171875,
      "learning_rate": 1.7866666666666665e-07,
      "loss": 0.3377,
      "reward": 0.6112585365772247,
      "reward_std": 0.18949438631534576,
      "reward_std_all": 0.1861392930150032,
      "rewards-max/forensics_sinkhorn_reward": 0.9887149333953857,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.99891397356987,
      "rewards/forensics_sinkhorn_reward": 0.4964735507965088,
      "rewards/forensics_strict_edge_reward": 0.2452763244509697,
      "rewards/forensics_tiou_reward": 0.3415825515985489,
      "step": 1233
    },
    {
      "completion_length": 16.859375,
      "epoch": 3.282290279627164,
      "grad_norm": 23.635009765625,
      "kl": 4.8125,
      "learning_rate": 1.7799999999999998e-07,
      "loss": 0.4081,
      "reward": 0.5982739925384521,
      "reward_std": 0.11768433451652527,
      "reward_std_all": 0.15824590623378754,
      "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": 0.5704380571842194,
      "rewards/forensics_strict_edge_reward": 0.33482062816619873,
      "rewards/forensics_tiou_reward": 0.4618826061487198,
      "step": 1234
    },
    {
      "completion_length": 18.125,
      "epoch": 3.2849533954727033,
      "grad_norm": 19.902036666870117,
      "kl": 5.84375,
      "learning_rate": 1.7733333333333333e-07,
      "loss": 0.363,
      "reward": 0.5806297063827515,
      "reward_std": 0.19144947081804276,
      "reward_std_all": 0.21042266488075256,
      "rewards-max/forensics_sinkhorn_reward": 0.9880830645561218,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9964253306388855,
      "rewards/forensics_sinkhorn_reward": 0.5148962736129761,
      "rewards/forensics_strict_edge_reward": 0.2638101652264595,
      "rewards/forensics_tiou_reward": 0.33875493705272675,
      "step": 1235
    },
    {
      "completion_length": 17.453125,
      "epoch": 3.2876165113182423,
      "grad_norm": 20.987934112548828,
      "kl": 6.53125,
      "learning_rate": 1.7666666666666666e-07,
      "loss": 0.4066,
      "reward": 0.5526963770389557,
      "reward_std": 0.18207576125860214,
      "reward_std_all": 0.2197565883398056,
      "rewards-max/forensics_sinkhorn_reward": 0.9869766533374786,
      "rewards-max/forensics_strict_edge_reward": 0.9890114367008209,
      "rewards-max/forensics_tiou_reward": 0.9955896437168121,
      "rewards/forensics_sinkhorn_reward": 0.4781893193721771,
      "rewards/forensics_strict_edge_reward": 0.2480630874633789,
      "rewards/forensics_tiou_reward": 0.27836549282073975,
      "step": 1236
    },
    {
      "completion_length": 18.03125,
      "epoch": 3.2902796271637818,
      "grad_norm": 29.079986572265625,
      "kl": 8.125,
      "learning_rate": 1.76e-07,
      "loss": 0.5333,
      "reward": 0.5096708834171295,
      "reward_std": 0.2986657917499542,
      "reward_std_all": 0.2957579419016838,
      "rewards-max/forensics_sinkhorn_reward": 0.982568085193634,
      "rewards-max/forensics_strict_edge_reward": 0.9357379674911499,
      "rewards-max/forensics_tiou_reward": 0.9847748577594757,
      "rewards/forensics_sinkhorn_reward": 0.45730060338974,
      "rewards/forensics_strict_edge_reward": 0.21255214512348175,
      "rewards/forensics_tiou_reward": 0.3150571882724762,
      "step": 1237
    },
    {
      "completion_length": 16.265625,
      "epoch": 3.2929427430093208,
      "grad_norm": 25.204120635986328,
      "kl": 5.046875,
      "learning_rate": 1.7533333333333334e-07,
      "loss": 0.3232,
      "reward": 0.6247771084308624,
      "reward_std": 0.17864344269037247,
      "reward_std_all": 0.17817110568284988,
      "rewards-max/forensics_sinkhorn_reward": 0.9875057637691498,
      "rewards-max/forensics_strict_edge_reward": 0.9565287232398987,
      "rewards-max/forensics_tiou_reward": 0.9941495954990387,
      "rewards/forensics_sinkhorn_reward": 0.5992515683174133,
      "rewards/forensics_strict_edge_reward": 0.28439486026763916,
      "rewards/forensics_tiou_reward": 0.4115414619445801,
      "step": 1238
    },
    {
      "completion_length": 17.6875,
      "epoch": 3.2956058588548602,
      "grad_norm": 28.974672317504883,
      "kl": 6.703125,
      "learning_rate": 1.7466666666666667e-07,
      "loss": 0.4526,
      "reward": 0.5069080889225006,
      "reward_std": 0.27817288041114807,
      "reward_std_all": 0.28170420229434967,
      "rewards-max/forensics_sinkhorn_reward": 0.9432019293308258,
      "rewards-max/forensics_strict_edge_reward": 0.774916410446167,
      "rewards-max/forensics_tiou_reward": 0.895800918340683,
      "rewards/forensics_sinkhorn_reward": 0.39191779494285583,
      "rewards/forensics_strict_edge_reward": 0.1618020460009575,
      "rewards/forensics_tiou_reward": 0.20410818606615067,
      "step": 1239
    },
    {
      "completion_length": 16.421875,
      "epoch": 3.2982689747003997,
      "grad_norm": 21.88393211364746,
      "kl": 6.078125,
      "learning_rate": 1.7399999999999997e-07,
      "loss": 0.498,
      "reward": 0.6146014928817749,
      "reward_std": 0.17788584530353546,
      "reward_std_all": 0.18707305938005447,
      "rewards-max/forensics_sinkhorn_reward": 0.9879242777824402,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9957999587059021,
      "rewards/forensics_sinkhorn_reward": 0.5713601261377335,
      "rewards/forensics_strict_edge_reward": 0.27280566096305847,
      "rewards/forensics_tiou_reward": 0.40991684794425964,
      "step": 1240
    },
    {
      "completion_length": 17.8125,
      "epoch": 3.3009320905459387,
      "grad_norm": 25.34548568725586,
      "kl": 5.0,
      "learning_rate": 1.7333333333333332e-07,
      "loss": 0.3369,
      "reward": 0.6044325530529022,
      "reward_std": 0.19863621518015862,
      "reward_std_all": 0.20067795738577843,
      "rewards-max/forensics_sinkhorn_reward": 0.9861200749874115,
      "rewards-max/forensics_strict_edge_reward": 0.9464850425720215,
      "rewards-max/forensics_tiou_reward": 0.9887360334396362,
      "rewards/forensics_sinkhorn_reward": 0.584547221660614,
      "rewards/forensics_strict_edge_reward": 0.390703946352005,
      "rewards/forensics_tiou_reward": 0.471569761633873,
      "step": 1241
    },
    {
      "completion_length": 15.890625,
      "epoch": 3.303595206391478,
      "grad_norm": 25.67864990234375,
      "kl": 5.171875,
      "learning_rate": 1.7266666666666665e-07,
      "loss": 0.2023,
      "reward": 0.6130959987640381,
      "reward_std": 0.1581210047006607,
      "reward_std_all": 0.17344141751527786,
      "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": 0.5373876690864563,
      "rewards/forensics_strict_edge_reward": 0.22064650058746338,
      "rewards/forensics_tiou_reward": 0.2855706959962845,
      "step": 1242
    },
    {
      "completion_length": 18.1875,
      "epoch": 3.306258322237017,
      "grad_norm": 28.1528263092041,
      "kl": 7.171875,
      "learning_rate": 1.7199999999999998e-07,
      "loss": 0.5026,
      "reward": 0.58311727643013,
      "reward_std": 0.17202116549015045,
      "reward_std_all": 0.1926928386092186,
      "rewards-max/forensics_sinkhorn_reward": 0.9832488000392914,
      "rewards-max/forensics_strict_edge_reward": 0.8983768224716187,
      "rewards-max/forensics_tiou_reward": 0.9775905013084412,
      "rewards/forensics_sinkhorn_reward": 0.5487460494041443,
      "rewards/forensics_strict_edge_reward": 0.3489236682653427,
      "rewards/forensics_tiou_reward": 0.4143906831741333,
      "step": 1243
    },
    {
      "completion_length": 14.890625,
      "epoch": 3.3089214380825567,
      "grad_norm": 25.981647491455078,
      "kl": 5.640625,
      "learning_rate": 1.7133333333333333e-07,
      "loss": 0.3274,
      "reward": 0.5642320811748505,
      "reward_std": 0.2082979530096054,
      "reward_std_all": 0.23326610028743744,
      "rewards-max/forensics_sinkhorn_reward": 0.9880536794662476,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9963070154190063,
      "rewards/forensics_sinkhorn_reward": 0.5091368556022644,
      "rewards/forensics_strict_edge_reward": 0.2272331267595291,
      "rewards/forensics_tiou_reward": 0.2923821359872818,
      "step": 1244
    },
    {
      "completion_length": 17.296875,
      "epoch": 3.3115845539280957,
      "grad_norm": 23.031526565551758,
      "kl": 4.3515625,
      "learning_rate": 1.7066666666666666e-07,
      "loss": 0.2962,
      "reward": 0.5353982150554657,
      "reward_std": 0.18034036457538605,
      "reward_std_all": 0.21680963784456253,
      "rewards-max/forensics_sinkhorn_reward": 0.973092257976532,
      "rewards-max/forensics_strict_edge_reward": 0.9677534997463226,
      "rewards-max/forensics_tiou_reward": 0.9656296372413635,
      "rewards/forensics_sinkhorn_reward": 0.471874862909317,
      "rewards/forensics_strict_edge_reward": 0.2102406993508339,
      "rewards/forensics_tiou_reward": 0.3015923798084259,
      "step": 1245
    },
    {
      "completion_length": 16.734375,
      "epoch": 3.314247669773635,
      "grad_norm": 19.621112823486328,
      "kl": 6.421875,
      "learning_rate": 1.7000000000000001e-07,
      "loss": 0.3998,
      "reward": 0.6578772962093353,
      "reward_std": 0.1883433237671852,
      "reward_std_all": 0.17932800948619843,
      "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": 0.6467979550361633,
      "rewards/forensics_strict_edge_reward": 0.35935546457767487,
      "rewards/forensics_tiou_reward": 0.4860704839229584,
      "step": 1246
    },
    {
      "completion_length": 17.109375,
      "epoch": 3.316910785619174,
      "grad_norm": 22.720504760742188,
      "kl": 4.0390625,
      "learning_rate": 1.6933333333333334e-07,
      "loss": 0.2764,
      "reward": 0.6261366605758667,
      "reward_std": 0.12363375723361969,
      "reward_std_all": 0.15784956514835358,
      "rewards-max/forensics_sinkhorn_reward": 0.9878041446208954,
      "rewards-max/forensics_strict_edge_reward": 0.927969753742218,
      "rewards-max/forensics_tiou_reward": 0.9953349232673645,
      "rewards/forensics_sinkhorn_reward": 0.5739218592643738,
      "rewards/forensics_strict_edge_reward": 0.2707481235265732,
      "rewards/forensics_tiou_reward": 0.39781537652015686,
      "step": 1247
    },
    {
      "completion_length": 16.46875,
      "epoch": 3.3195739014647137,
      "grad_norm": 18.75218391418457,
      "kl": 3.9765625,
      "learning_rate": 1.6866666666666664e-07,
      "loss": 0.3105,
      "reward": 0.6263561844825745,
      "reward_std": 0.17096073925495148,
      "reward_std_all": 0.1764160841703415,
      "rewards-max/forensics_sinkhorn_reward": 0.9888060092926025,
      "rewards-max/forensics_strict_edge_reward": 0.9890114367008209,
      "rewards-max/forensics_tiou_reward": 0.9992727339267731,
      "rewards/forensics_sinkhorn_reward": 0.6035741865634918,
      "rewards/forensics_strict_edge_reward": 0.37251582741737366,
      "rewards/forensics_tiou_reward": 0.46235793828964233,
      "step": 1248
    },
    {
      "completion_length": 15.59375,
      "epoch": 3.322237017310253,
      "grad_norm": 19.568151473999023,
      "kl": 3.0390625,
      "learning_rate": 1.68e-07,
      "loss": 0.1284,
      "reward": 0.6810437440872192,
      "reward_std": 0.10448861494660378,
      "reward_std_all": 0.11266409605741501,
      "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": 0.6176020801067352,
      "rewards/forensics_strict_edge_reward": 0.25365814566612244,
      "rewards/forensics_tiou_reward": 0.37034475803375244,
      "step": 1249
    },
    {
      "completion_length": 16.984375,
      "epoch": 3.324900133155792,
      "grad_norm": 19.131267547607422,
      "kl": 4.1953125,
      "learning_rate": 1.6733333333333332e-07,
      "loss": 0.297,
      "reward": 0.6683538854122162,
      "reward_std": 0.13242049142718315,
      "reward_std_all": 0.1373419612646103,
      "rewards-max/forensics_sinkhorn_reward": 0.9878386855125427,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9954647421836853,
      "rewards/forensics_sinkhorn_reward": 0.5937719941139221,
      "rewards/forensics_strict_edge_reward": 0.3337528705596924,
      "rewards/forensics_tiou_reward": 0.3798670619726181,
      "step": 1250
    },
    {
      "completion_length": 16.0625,
      "epoch": 3.3275632490013316,
      "grad_norm": 21.539724349975586,
      "kl": 3.7421875,
      "learning_rate": 1.6666666666666665e-07,
      "loss": 0.2859,
      "reward": 0.641403466463089,
      "reward_std": 0.16118920594453812,
      "reward_std_all": 0.16193218529224396,
      "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": 0.6062043309211731,
      "rewards/forensics_strict_edge_reward": 0.36021101474761963,
      "rewards/forensics_tiou_reward": 0.4521905779838562,
      "step": 1251
    },
    {
      "completion_length": 17.015625,
      "epoch": 3.3302263648468706,
      "grad_norm": 21.599441528320312,
      "kl": 2.640625,
      "learning_rate": 1.66e-07,
      "loss": 0.2741,
      "reward": 0.6436159014701843,
      "reward_std": 0.21198132634162903,
      "reward_std_all": 0.20008091628551483,
      "rewards-max/forensics_sinkhorn_reward": 0.9838880896568298,
      "rewards-max/forensics_strict_edge_reward": 0.9075756669044495,
      "rewards-max/forensics_tiou_reward": 0.9800735116004944,
      "rewards/forensics_sinkhorn_reward": 0.6138569414615631,
      "rewards/forensics_strict_edge_reward": 0.3755764812231064,
      "rewards/forensics_tiou_reward": 0.4685276597738266,
      "step": 1252
    },
    {
      "completion_length": 14.703125,
      "epoch": 3.33288948069241,
      "grad_norm": 25.71774673461914,
      "kl": 3.0,
      "learning_rate": 1.6533333333333333e-07,
      "loss": 0.2578,
      "reward": 0.6519930362701416,
      "reward_std": 0.17638493329286575,
      "reward_std_all": 0.17063453048467636,
      "rewards-max/forensics_sinkhorn_reward": 0.987994372844696,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9960753321647644,
      "rewards/forensics_sinkhorn_reward": 0.5871725082397461,
      "rewards/forensics_strict_edge_reward": 0.3140889033675194,
      "rewards/forensics_tiou_reward": 0.37727488577365875,
      "step": 1253
    },
    {
      "completion_length": 19.078125,
      "epoch": 3.3355525965379496,
      "grad_norm": 15.246038436889648,
      "kl": 4.671875,
      "learning_rate": 1.6466666666666666e-07,
      "loss": 0.2767,
      "reward": 0.5471813976764679,
      "reward_std": 0.2220669761300087,
      "reward_std_all": 0.23844391107559204,
      "rewards-max/forensics_sinkhorn_reward": 0.9881205856800079,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9965714812278748,
      "rewards/forensics_sinkhorn_reward": 0.4972756803035736,
      "rewards/forensics_strict_edge_reward": 0.3204954117536545,
      "rewards/forensics_tiou_reward": 0.3857070952653885,
      "step": 1254
    },
    {
      "completion_length": 14.859375,
      "epoch": 3.3382157123834886,
      "grad_norm": 22.526002883911133,
      "kl": 3.703125,
      "learning_rate": 1.64e-07,
      "loss": 0.3204,
      "reward": 0.6540412306785583,
      "reward_std": 0.13272716850042343,
      "reward_std_all": 0.1417626440525055,
      "rewards-max/forensics_sinkhorn_reward": 0.9887368381023407,
      "rewards-max/forensics_strict_edge_reward": 0.9782643616199493,
      "rewards-max/forensics_tiou_reward": 0.9990000128746033,
      "rewards/forensics_sinkhorn_reward": 0.6409136652946472,
      "rewards/forensics_strict_edge_reward": 0.35239872336387634,
      "rewards/forensics_tiou_reward": 0.4705206900835037,
      "step": 1255
    },
    {
      "completion_length": 16.5625,
      "epoch": 3.340878828229028,
      "grad_norm": 22.017446517944336,
      "kl": 4.328125,
      "learning_rate": 1.6333333333333331e-07,
      "loss": 0.2994,
      "reward": 0.6099773049354553,
      "reward_std": 0.16642332077026367,
      "reward_std_all": 0.1793428435921669,
      "rewards-max/forensics_sinkhorn_reward": 0.9878455102443695,
      "rewards-max/forensics_strict_edge_reward": 0.9265981018543243,
      "rewards-max/forensics_tiou_reward": 0.9954955577850342,
      "rewards/forensics_sinkhorn_reward": 0.5490577220916748,
      "rewards/forensics_strict_edge_reward": 0.26541075110435486,
      "rewards/forensics_tiou_reward": 0.36879822611808777,
      "step": 1256
    },
    {
      "completion_length": 16.3125,
      "epoch": 3.343541944074567,
      "grad_norm": 28.45949363708496,
      "kl": 4.140625,
      "learning_rate": 1.6266666666666664e-07,
      "loss": 0.3396,
      "reward": 0.6247305572032928,
      "reward_std": 0.16964136809110641,
      "reward_std_all": 0.17884408682584763,
      "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": 0.5840726494789124,
      "rewards/forensics_strict_edge_reward": 0.2895520105957985,
      "rewards/forensics_tiou_reward": 0.3914931118488312,
      "step": 1257
    },
    {
      "completion_length": 15.90625,
      "epoch": 3.3462050599201065,
      "grad_norm": 25.146677017211914,
      "kl": 6.53125,
      "learning_rate": 1.62e-07,
      "loss": 0.2841,
      "reward": 0.6724953651428223,
      "reward_std": 0.12337691336870193,
      "reward_std_all": 0.12824072688817978,
      "rewards-max/forensics_sinkhorn_reward": 0.9870195686817169,
      "rewards-max/forensics_strict_edge_reward": 0.9355069994926453,
      "rewards-max/forensics_tiou_reward": 0.9922443330287933,
      "rewards/forensics_sinkhorn_reward": 0.6026785969734192,
      "rewards/forensics_strict_edge_reward": 0.2725376486778259,
      "rewards/forensics_tiou_reward": 0.35308021306991577,
      "step": 1258
    },
    {
      "completion_length": 17.296875,
      "epoch": 3.348868175765646,
      "grad_norm": 45.537208557128906,
      "kl": 8.75,
      "learning_rate": 1.6133333333333332e-07,
      "loss": 0.5897,
      "reward": 0.5652238428592682,
      "reward_std": 0.24089542031288147,
      "reward_std_all": 0.24600188434123993,
      "rewards-max/forensics_sinkhorn_reward": 0.9875503778457642,
      "rewards-max/forensics_strict_edge_reward": 0.9565287232398987,
      "rewards-max/forensics_tiou_reward": 0.9943233132362366,
      "rewards/forensics_sinkhorn_reward": 0.484880656003952,
      "rewards/forensics_strict_edge_reward": 0.24694770574569702,
      "rewards/forensics_tiou_reward": 0.30726462602615356,
      "step": 1259
    },
    {
      "completion_length": 17.046875,
      "epoch": 3.351531291611185,
      "grad_norm": 23.60318374633789,
      "kl": 4.484375,
      "learning_rate": 1.6066666666666668e-07,
      "loss": 0.3909,
      "reward": 0.614342212677002,
      "reward_std": 0.2000296339392662,
      "reward_std_all": 0.2061544954776764,
      "rewards-max/forensics_sinkhorn_reward": 0.9463672637939453,
      "rewards-max/forensics_strict_edge_reward": 0.8915697336196899,
      "rewards-max/forensics_tiou_reward": 0.9084615409374237,
      "rewards/forensics_sinkhorn_reward": 0.5479446351528168,
      "rewards/forensics_strict_edge_reward": 0.2757415324449539,
      "rewards/forensics_tiou_reward": 0.38505876064300537,
      "step": 1260
    },
    {
      "completion_length": 19.015625,
      "epoch": 3.3541944074567245,
      "grad_norm": 24.01484489440918,
      "kl": 6.921875,
      "learning_rate": 1.6e-07,
      "loss": 0.5136,
      "reward": 0.5478952676057816,
      "reward_std": 0.25414905697107315,
      "reward_std_all": 0.2608228698372841,
      "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": 0.5096511542797089,
      "rewards/forensics_strict_edge_reward": 0.3818298280239105,
      "rewards/forensics_tiou_reward": 0.44392454624176025,
      "step": 1261
    },
    {
      "completion_length": 15.734375,
      "epoch": 3.3568575233022635,
      "grad_norm": 30.235931396484375,
      "kl": 3.7578125,
      "learning_rate": 1.5933333333333333e-07,
      "loss": 0.1652,
      "reward": 0.6388593912124634,
      "reward_std": 0.14323583990335464,
      "reward_std_all": 0.16496150195598602,
      "rewards-max/forensics_sinkhorn_reward": 0.9760358929634094,
      "rewards-max/forensics_strict_edge_reward": 0.9185642302036285,
      "rewards-max/forensics_tiou_reward": 0.9706442058086395,
      "rewards/forensics_sinkhorn_reward": 0.5905227661132812,
      "rewards/forensics_strict_edge_reward": 0.2827534079551697,
      "rewards/forensics_tiou_reward": 0.3520139306783676,
      "step": 1262
    },
    {
      "completion_length": 16.734375,
      "epoch": 3.359520639147803,
      "grad_norm": 39.36994552612305,
      "kl": 8.15625,
      "learning_rate": 1.5866666666666666e-07,
      "loss": 0.5443,
      "reward": 0.5347634255886078,
      "reward_std": 0.24005813151597977,
      "reward_std_all": 0.25708942860364914,
      "rewards-max/forensics_sinkhorn_reward": 0.9853793382644653,
      "rewards-max/forensics_strict_edge_reward": 0.927969753742218,
      "rewards-max/forensics_tiou_reward": 0.9858735799789429,
      "rewards/forensics_sinkhorn_reward": 0.49912717938423157,
      "rewards/forensics_strict_edge_reward": 0.21185865253210068,
      "rewards/forensics_tiou_reward": 0.32625406980514526,
      "step": 1263
    },
    {
      "completion_length": 17.421875,
      "epoch": 3.3621837549933424,
      "grad_norm": 20.974706649780273,
      "kl": 6.640625,
      "learning_rate": 1.5799999999999999e-07,
      "loss": 0.5068,
      "reward": 0.5188620239496231,
      "reward_std": 0.2111399993300438,
      "reward_std_all": 0.2519683614373207,
      "rewards-max/forensics_sinkhorn_reward": 0.9658338129520416,
      "rewards-max/forensics_strict_edge_reward": 0.9375866651535034,
      "rewards-max/forensics_tiou_reward": 0.9495626389980316,
      "rewards/forensics_sinkhorn_reward": 0.4479987323284149,
      "rewards/forensics_strict_edge_reward": 0.2308598756790161,
      "rewards/forensics_tiou_reward": 0.27759580314159393,
      "step": 1264
    },
    {
      "completion_length": 17.5,
      "epoch": 3.3648468708388815,
      "grad_norm": 16.499582290649414,
      "kl": 5.421875,
      "learning_rate": 1.573333333333333e-07,
      "loss": 0.4139,
      "reward": 0.60800501704216,
      "reward_std": 0.2131679207086563,
      "reward_std_all": 0.20933566987514496,
      "rewards-max/forensics_sinkhorn_reward": 0.9867768585681915,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9912943542003632,
      "rewards/forensics_sinkhorn_reward": 0.5387015789747238,
      "rewards/forensics_strict_edge_reward": 0.32983657717704773,
      "rewards/forensics_tiou_reward": 0.3623042330145836,
      "step": 1265
    },
    {
      "completion_length": 19.8125,
      "epoch": 3.367509986684421,
      "grad_norm": 23.106651306152344,
      "kl": 7.4375,
      "learning_rate": 1.5666666666666667e-07,
      "loss": 0.5404,
      "reward": 0.5170812904834747,
      "reward_std": 0.23174428939819336,
      "reward_std_all": 0.2538259029388428,
      "rewards-max/forensics_sinkhorn_reward": 0.9868491590023041,
      "rewards-max/forensics_strict_edge_reward": 0.9574736058712006,
      "rewards-max/forensics_tiou_reward": 0.9949440956115723,
      "rewards/forensics_sinkhorn_reward": 0.48269304633140564,
      "rewards/forensics_strict_edge_reward": 0.30505578219890594,
      "rewards/forensics_tiou_reward": 0.3759995698928833,
      "step": 1266
    },
    {
      "completion_length": 18.46875,
      "epoch": 3.37017310252996,
      "grad_norm": 36.496761322021484,
      "kl": 7.59375,
      "learning_rate": 1.56e-07,
      "loss": 0.4166,
      "reward": 0.5905042588710785,
      "reward_std": 0.13828559592366219,
      "reward_std_all": 0.16933859884738922,
      "rewards-max/forensics_sinkhorn_reward": 0.9863327145576477,
      "rewards-max/forensics_strict_edge_reward": 0.9357379674911499,
      "rewards-max/forensics_tiou_reward": 0.9895535111427307,
      "rewards/forensics_sinkhorn_reward": 0.5262929052114487,
      "rewards/forensics_strict_edge_reward": 0.1967082843184471,
      "rewards/forensics_tiou_reward": 0.30749139189720154,
      "step": 1267
    },
    {
      "completion_length": 17.96875,
      "epoch": 3.3728362183754994,
      "grad_norm": 16.55620574951172,
      "kl": 5.078125,
      "learning_rate": 1.5533333333333332e-07,
      "loss": 0.3402,
      "reward": 0.6786829829216003,
      "reward_std": 0.13518329709768295,
      "reward_std_all": 0.13321009278297424,
      "rewards-max/forensics_sinkhorn_reward": 0.9872619807720184,
      "rewards-max/forensics_strict_edge_reward": 0.9169811904430389,
      "rewards-max/forensics_tiou_reward": 0.9932023584842682,
      "rewards/forensics_sinkhorn_reward": 0.6290920078754425,
      "rewards/forensics_strict_edge_reward": 0.30491601675748825,
      "rewards/forensics_tiou_reward": 0.4483693689107895,
      "step": 1268
    },
    {
      "completion_length": 15.921875,
      "epoch": 3.3754993342210384,
      "grad_norm": 15.966964721679688,
      "kl": 3.921875,
      "learning_rate": 1.5466666666666668e-07,
      "loss": 0.3086,
      "reward": 0.6097693592309952,
      "reward_std": 0.11794839985668659,
      "reward_std_all": 0.15299659222364426,
      "rewards-max/forensics_sinkhorn_reward": 0.9883122742176056,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9973292350769043,
      "rewards/forensics_sinkhorn_reward": 0.5968497097492218,
      "rewards/forensics_strict_edge_reward": 0.38193918764591217,
      "rewards/forensics_tiou_reward": 0.45391958951950073,
      "step": 1269
    },
    {
      "completion_length": 16.78125,
      "epoch": 3.378162450066578,
      "grad_norm": 19.633970260620117,
      "kl": 4.859375,
      "learning_rate": 1.54e-07,
      "loss": 0.3449,
      "reward": 0.5917561054229736,
      "reward_std": 0.1970629319548607,
      "reward_std_all": 0.20279698073863983,
      "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": 0.5591360330581665,
      "rewards/forensics_strict_edge_reward": 0.3051953464746475,
      "rewards/forensics_tiou_reward": 0.3729635030031204,
      "step": 1270
    },
    {
      "completion_length": 19.578125,
      "epoch": 3.3808255659121174,
      "grad_norm": 32.497222900390625,
      "kl": 6.203125,
      "learning_rate": 1.533333333333333e-07,
      "loss": 0.3586,
      "reward": 0.5013523548841476,
      "reward_std": 0.17517075687646866,
      "reward_std_all": 0.23214709013700485,
      "rewards-max/forensics_sinkhorn_reward": 0.9880723655223846,
      "rewards-max/forensics_strict_edge_reward": 0.9464850425720215,
      "rewards-max/forensics_tiou_reward": 0.9963839054107666,
      "rewards/forensics_sinkhorn_reward": 0.38899707794189453,
      "rewards/forensics_strict_edge_reward": 0.13186755031347275,
      "rewards/forensics_tiou_reward": 0.19441542029380798,
      "step": 1271
    },
    {
      "completion_length": 14.765625,
      "epoch": 3.3834886817576564,
      "grad_norm": 23.6986083984375,
      "kl": 2.8046875,
      "learning_rate": 1.5266666666666666e-07,
      "loss": 0.1925,
      "reward": 0.6292681694030762,
      "reward_std": 0.1652146875858307,
      "reward_std_all": 0.17397292703390121,
      "rewards-max/forensics_sinkhorn_reward": 0.9792178869247437,
      "rewards-max/forensics_strict_edge_reward": 0.9677534997463226,
      "rewards-max/forensics_tiou_reward": 0.9787491858005524,
      "rewards/forensics_sinkhorn_reward": 0.5739697217941284,
      "rewards/forensics_strict_edge_reward": 0.27925996482372284,
      "rewards/forensics_tiou_reward": 0.3840307593345642,
      "step": 1272
    },
    {
      "completion_length": 15.765625,
      "epoch": 3.386151797603196,
      "grad_norm": 25.429960250854492,
      "kl": 3.234375,
      "learning_rate": 1.5199999999999998e-07,
      "loss": 0.2068,
      "reward": 0.6748658418655396,
      "reward_std": 0.10331057384610176,
      "reward_std_all": 0.11800898611545563,
      "rewards-max/forensics_sinkhorn_reward": 0.98875692486763,
      "rewards-max/forensics_strict_edge_reward": 0.9890114367008209,
      "rewards-max/forensics_tiou_reward": 0.9990796148777008,
      "rewards/forensics_sinkhorn_reward": 0.6205296814441681,
      "rewards/forensics_strict_edge_reward": 0.30154022574424744,
      "rewards/forensics_tiou_reward": 0.3890647441148758,
      "step": 1273
    },
    {
      "completion_length": 16.578125,
      "epoch": 3.388814913448735,
      "grad_norm": 17.435726165771484,
      "kl": 4.0078125,
      "learning_rate": 1.513333333333333e-07,
      "loss": 0.2439,
      "reward": 0.6023592948913574,
      "reward_std": 0.1932571679353714,
      "reward_std_all": 0.19167955219745636,
      "rewards-max/forensics_sinkhorn_reward": 0.9887278974056244,
      "rewards-max/forensics_strict_edge_reward": 0.9890114367008209,
      "rewards-max/forensics_tiou_reward": 0.9989653527736664,
      "rewards/forensics_sinkhorn_reward": 0.5927978157997131,
      "rewards/forensics_strict_edge_reward": 0.36016759276390076,
      "rewards/forensics_tiou_reward": 0.45473022758960724,
      "step": 1274
    },
    {
      "completion_length": 17.0,
      "epoch": 3.3914780292942743,
      "grad_norm": 20.291017532348633,
      "kl": 3.2421875,
      "learning_rate": 1.5066666666666667e-07,
      "loss": 0.2386,
      "reward": 0.6159288585186005,
      "reward_std": 0.1508881188929081,
      "reward_std_all": 0.16944625973701477,
      "rewards-max/forensics_sinkhorn_reward": 0.9347826540470123,
      "rewards-max/forensics_strict_edge_reward": 0.7673945724964142,
      "rewards-max/forensics_tiou_reward": 0.8745266795158386,
      "rewards/forensics_sinkhorn_reward": 0.5416828393936157,
      "rewards/forensics_strict_edge_reward": 0.28439219295978546,
      "rewards/forensics_tiou_reward": 0.3412861227989197,
      "step": 1275
    },
    {
      "completion_length": 15.875,
      "epoch": 3.3941411451398134,
      "grad_norm": 23.20099639892578,
      "kl": 3.421875,
      "learning_rate": 1.5e-07,
      "loss": 0.1747,
      "reward": 0.6980170905590057,
      "reward_std": 0.08819859474897385,
      "reward_std_all": 0.09611432254314423,
      "rewards-max/forensics_sinkhorn_reward": 0.9874123930931091,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9937859773635864,
      "rewards/forensics_sinkhorn_reward": 0.6904652714729309,
      "rewards/forensics_strict_edge_reward": 0.4122115820646286,
      "rewards/forensics_tiou_reward": 0.533497154712677,
      "step": 1276
    },
    {
      "completion_length": 17.25,
      "epoch": 3.396804260985353,
      "grad_norm": 26.319839477539062,
      "kl": 5.65625,
      "learning_rate": 1.4933333333333335e-07,
      "loss": 0.4131,
      "reward": 0.6082740426063538,
      "reward_std": 0.13770408928394318,
      "reward_std_all": 0.16411246359348297,
      "rewards-max/forensics_sinkhorn_reward": 0.9879834055900574,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9960369169712067,
      "rewards/forensics_sinkhorn_reward": 0.5480271279811859,
      "rewards/forensics_strict_edge_reward": 0.22103889286518097,
      "rewards/forensics_tiou_reward": 0.318335123360157,
      "step": 1277
    },
    {
      "completion_length": 15.5,
      "epoch": 3.3994673768308923,
      "grad_norm": 17.743375778198242,
      "kl": 2.1015625,
      "learning_rate": 1.4866666666666667e-07,
      "loss": 0.1533,
      "reward": 0.7135291397571564,
      "reward_std": 0.10342475026845932,
      "reward_std_all": 0.09918815642595291,
      "rewards-max/forensics_sinkhorn_reward": 0.9881440997123718,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9966652989387512,
      "rewards/forensics_sinkhorn_reward": 0.713520884513855,
      "rewards/forensics_strict_edge_reward": 0.3356952667236328,
      "rewards/forensics_tiou_reward": 0.5660473108291626,
      "step": 1278
    },
    {
      "completion_length": 16.359375,
      "epoch": 3.4021304926764313,
      "grad_norm": 18.908185958862305,
      "kl": 4.109375,
      "learning_rate": 1.4799999999999998e-07,
      "loss": 0.341,
      "reward": 0.6850811243057251,
      "reward_std": 0.14581846818327904,
      "reward_std_all": 0.13806888088583946,
      "rewards-max/forensics_sinkhorn_reward": 0.9868179261684418,
      "rewards-max/forensics_strict_edge_reward": 0.9565287232398987,
      "rewards-max/forensics_tiou_reward": 0.9914407730102539,
      "rewards/forensics_sinkhorn_reward": 0.6581770777702332,
      "rewards/forensics_strict_edge_reward": 0.35410381853580475,
      "rewards/forensics_tiou_reward": 0.4816630780696869,
      "step": 1279
    },
    {
      "completion_length": 16.765625,
      "epoch": 3.4047936085219708,
      "grad_norm": 23.417041778564453,
      "kl": 4.1796875,
      "learning_rate": 1.4733333333333333e-07,
      "loss": 0.1973,
      "reward": 0.6036809682846069,
      "reward_std": 0.1950344741344452,
      "reward_std_all": 0.20353343337774277,
      "rewards-max/forensics_sinkhorn_reward": 0.9883900582790375,
      "rewards-max/forensics_strict_edge_reward": 0.9567649364471436,
      "rewards-max/forensics_tiou_reward": 0.9976343810558319,
      "rewards/forensics_sinkhorn_reward": 0.4994353950023651,
      "rewards/forensics_strict_edge_reward": 0.21702267974615097,
      "rewards/forensics_tiou_reward": 0.3135703057050705,
      "step": 1280
    },
    {
      "completion_length": 18.578125,
      "epoch": 3.40745672436751,
      "grad_norm": 22.026592254638672,
      "kl": 3.9375,
      "learning_rate": 1.4666666666666666e-07,
      "loss": 0.1713,
      "reward": 0.6499991416931152,
      "reward_std": 0.22784943878650665,
      "reward_std_all": 0.2039707601070404,
      "rewards-max/forensics_sinkhorn_reward": 0.9671533405780792,
      "rewards-max/forensics_strict_edge_reward": 0.9567649364471436,
      "rewards-max/forensics_tiou_reward": 0.9524356722831726,
      "rewards/forensics_sinkhorn_reward": 0.5291480123996735,
      "rewards/forensics_strict_edge_reward": 0.30606549978256226,
      "rewards/forensics_tiou_reward": 0.34434641897678375,
      "step": 1281
    },
    {
      "completion_length": 16.5,
      "epoch": 3.4101198402130493,
      "grad_norm": 26.058242797851562,
      "kl": 6.15625,
      "learning_rate": 1.4599999999999998e-07,
      "loss": 0.4765,
      "reward": 0.6258521676063538,
      "reward_std": 0.15699831396341324,
      "reward_std_all": 0.17087706923484802,
      "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": 0.6458434760570526,
      "rewards/forensics_strict_edge_reward": 0.3986528068780899,
      "rewards/forensics_tiou_reward": 0.49896201491355896,
      "step": 1282
    },
    {
      "completion_length": 18.84375,
      "epoch": 3.4127829560585887,
      "grad_norm": 18.601633071899414,
      "kl": 5.484375,
      "learning_rate": 1.4533333333333334e-07,
      "loss": 0.3402,
      "reward": 0.5743925869464874,
      "reward_std": 0.2656787782907486,
      "reward_std_all": 0.24982380867004395,
      "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": 0.5290447771549225,
      "rewards/forensics_strict_edge_reward": 0.29066595435142517,
      "rewards/forensics_tiou_reward": 0.39094407856464386,
      "step": 1283
    },
    {
      "completion_length": 17.421875,
      "epoch": 3.4154460719041277,
      "grad_norm": 25.131616592407227,
      "kl": 5.03125,
      "learning_rate": 1.4466666666666667e-07,
      "loss": 0.3637,
      "reward": 0.5794664025306702,
      "reward_std": 0.25278666615486145,
      "reward_std_all": 0.2452811449766159,
      "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": 0.5308918654918671,
      "rewards/forensics_strict_edge_reward": 0.2878424823284149,
      "rewards/forensics_tiou_reward": 0.3676142692565918,
      "step": 1284
    },
    {
      "completion_length": 16.59375,
      "epoch": 3.418109187749667,
      "grad_norm": 23.65653419494629,
      "kl": 4.921875,
      "learning_rate": 1.44e-07,
      "loss": 0.4578,
      "reward": 0.5302155315876007,
      "reward_std": 0.24966100603342056,
      "reward_std_all": 0.2654084041714668,
      "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": 0.4585827887058258,
      "rewards/forensics_strict_edge_reward": 0.27602343261241913,
      "rewards/forensics_tiou_reward": 0.33099478483200073,
      "step": 1285
    },
    {
      "completion_length": 17.484375,
      "epoch": 3.4207723035952062,
      "grad_norm": 25.95047950744629,
      "kl": 6.359375,
      "learning_rate": 1.4333333333333335e-07,
      "loss": 0.4965,
      "reward": 0.5282774865627289,
      "reward_std": 0.24355052411556244,
      "reward_std_all": 0.26669585704803467,
      "rewards-max/forensics_sinkhorn_reward": 0.9370319545269012,
      "rewards-max/forensics_strict_edge_reward": 0.8612285256385803,
      "rewards-max/forensics_tiou_reward": 0.8877713084220886,
      "rewards/forensics_sinkhorn_reward": 0.4404715746641159,
      "rewards/forensics_strict_edge_reward": 0.19525189697742462,
      "rewards/forensics_tiou_reward": 0.2719345986843109,
      "step": 1286
    },
    {
      "completion_length": 16.5625,
      "epoch": 3.4234354194407457,
      "grad_norm": 21.759029388427734,
      "kl": 4.625,
      "learning_rate": 1.4266666666666665e-07,
      "loss": 0.2964,
      "reward": 0.6346924901008606,
      "reward_std": 0.182113915681839,
      "reward_std_all": 0.18346668034791946,
      "rewards-max/forensics_sinkhorn_reward": 0.9140952527523041,
      "rewards-max/forensics_strict_edge_reward": 0.7889257967472076,
      "rewards-max/forensics_tiou_reward": 0.837306559085846,
      "rewards/forensics_sinkhorn_reward": 0.5588363409042358,
      "rewards/forensics_strict_edge_reward": 0.24368472024798393,
      "rewards/forensics_tiou_reward": 0.34842923283576965,
      "step": 1287
    },
    {
      "completion_length": 16.203125,
      "epoch": 3.426098535286285,
      "grad_norm": 20.99517250061035,
      "kl": 4.359375,
      "learning_rate": 1.4199999999999997e-07,
      "loss": 0.2838,
      "reward": 0.5532405078411102,
      "reward_std": 0.2624581605195999,
      "reward_std_all": 0.2560287117958069,
      "rewards-max/forensics_sinkhorn_reward": 0.9858744144439697,
      "rewards-max/forensics_strict_edge_reward": 0.9364311099052429,
      "rewards-max/forensics_tiou_reward": 0.9877206683158875,
      "rewards/forensics_sinkhorn_reward": 0.47859804332256317,
      "rewards/forensics_strict_edge_reward": 0.25978390872478485,
      "rewards/forensics_tiou_reward": 0.32735809683799744,
      "step": 1288
    },
    {
      "completion_length": 16.125,
      "epoch": 3.428761651131824,
      "grad_norm": 27.271759033203125,
      "kl": 4.4140625,
      "learning_rate": 1.4133333333333333e-07,
      "loss": 0.173,
      "reward": 0.6420534253120422,
      "reward_std": 0.16597989946603775,
      "reward_std_all": 0.16606329381465912,
      "rewards-max/forensics_sinkhorn_reward": 0.9845491945743561,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9889996945858002,
      "rewards/forensics_sinkhorn_reward": 0.5764610767364502,
      "rewards/forensics_strict_edge_reward": 0.28533628582954407,
      "rewards/forensics_tiou_reward": 0.36242471635341644,
      "step": 1289
    },
    {
      "completion_length": 16.609375,
      "epoch": 3.4314247669773636,
      "grad_norm": 34.34121322631836,
      "kl": 4.859375,
      "learning_rate": 1.4066666666666666e-07,
      "loss": 0.3765,
      "reward": 0.6097796559333801,
      "reward_std": 0.1681295707821846,
      "reward_std_all": 0.1887487843632698,
      "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": 0.5581006705760956,
      "rewards/forensics_strict_edge_reward": 0.27609819173812866,
      "rewards/forensics_tiou_reward": 0.3538419157266617,
      "step": 1290
    },
    {
      "completion_length": 16.5625,
      "epoch": 3.4340878828229027,
      "grad_norm": 19.843896865844727,
      "kl": 5.0625,
      "learning_rate": 1.4e-07,
      "loss": 0.2314,
      "reward": 0.6312599778175354,
      "reward_std": 0.15441348403692245,
      "reward_std_all": 0.16767698526382446,
      "rewards-max/forensics_sinkhorn_reward": 0.9874953031539917,
      "rewards-max/forensics_strict_edge_reward": 0.9357379674911499,
      "rewards-max/forensics_tiou_reward": 0.9941061735153198,
      "rewards/forensics_sinkhorn_reward": 0.555399626493454,
      "rewards/forensics_strict_edge_reward": 0.23990245163440704,
      "rewards/forensics_tiou_reward": 0.33566105365753174,
      "step": 1291
    },
    {
      "completion_length": 16.015625,
      "epoch": 3.436750998668442,
      "grad_norm": 27.564510345458984,
      "kl": 2.9609375,
      "learning_rate": 1.3933333333333334e-07,
      "loss": 0.1687,
      "reward": 0.6797077059745789,
      "reward_std": 0.12072058022022247,
      "reward_std_all": 0.12830819934606552,
      "rewards-max/forensics_sinkhorn_reward": 0.9883022010326385,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9972867071628571,
      "rewards/forensics_sinkhorn_reward": 0.6292148530483246,
      "rewards/forensics_strict_edge_reward": 0.3701760172843933,
      "rewards/forensics_tiou_reward": 0.43641649186611176,
      "step": 1292
    },
    {
      "completion_length": 17.4375,
      "epoch": 3.4394141145139816,
      "grad_norm": 23.67966651916504,
      "kl": 5.109375,
      "learning_rate": 1.3866666666666666e-07,
      "loss": 0.5491,
      "reward": 0.5703753530979156,
      "reward_std": 0.16809750348329544,
      "reward_std_all": 0.20318979769945145,
      "rewards-max/forensics_sinkhorn_reward": 0.9882398247718811,
      "rewards-max/forensics_strict_edge_reward": 0.9364311099052429,
      "rewards-max/forensics_tiou_reward": 0.9970419406890869,
      "rewards/forensics_sinkhorn_reward": 0.4723481684923172,
      "rewards/forensics_strict_edge_reward": 0.2260017842054367,
      "rewards/forensics_tiou_reward": 0.3228372484445572,
      "step": 1293
    },
    {
      "completion_length": 17.125,
      "epoch": 3.4420772303595206,
      "grad_norm": 19.527822494506836,
      "kl": 5.515625,
      "learning_rate": 1.3800000000000002e-07,
      "loss": 0.3863,
      "reward": 0.5726946890354156,
      "reward_std": 0.22322991490364075,
      "reward_std_all": 0.22372367233037949,
      "rewards-max/forensics_sinkhorn_reward": 0.9876342415809631,
      "rewards-max/forensics_strict_edge_reward": 0.9256840348243713,
      "rewards-max/forensics_tiou_reward": 0.9946552813053131,
      "rewards/forensics_sinkhorn_reward": 0.4783759266138077,
      "rewards/forensics_strict_edge_reward": 0.27954375743865967,
      "rewards/forensics_tiou_reward": 0.3153278976678848,
      "step": 1294
    },
    {
      "completion_length": 15.953125,
      "epoch": 3.44474034620506,
      "grad_norm": 26.907852172851562,
      "kl": 4.875,
      "learning_rate": 1.3733333333333332e-07,
      "loss": 0.3179,
      "reward": 0.631978839635849,
      "reward_std": 0.20466125756502151,
      "reward_std_all": 0.1947474479675293,
      "rewards-max/forensics_sinkhorn_reward": 0.987475723028183,
      "rewards-max/forensics_strict_edge_reward": 0.9574736058712006,
      "rewards-max/forensics_tiou_reward": 0.9940564930438995,
      "rewards/forensics_sinkhorn_reward": 0.5650055110454559,
      "rewards/forensics_strict_edge_reward": 0.30295488238334656,
      "rewards/forensics_tiou_reward": 0.37469199299812317,
      "step": 1295
    },
    {
      "completion_length": 16.421875,
      "epoch": 3.447403462050599,
      "grad_norm": 33.79771423339844,
      "kl": 4.125,
      "learning_rate": 1.3666666666666665e-07,
      "loss": 0.3357,
      "reward": 0.6682111620903015,
      "reward_std": 0.1703856661915779,
      "reward_std_all": 0.15897741168737411,
      "rewards-max/forensics_sinkhorn_reward": 0.9882188737392426,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9969607591629028,
      "rewards/forensics_sinkhorn_reward": 0.6396207213401794,
      "rewards/forensics_strict_edge_reward": 0.35975174605846405,
      "rewards/forensics_tiou_reward": 0.4458082914352417,
      "step": 1296
    },
    {
      "completion_length": 17.46875,
      "epoch": 3.4500665778961386,
      "grad_norm": 26.88910484313965,
      "kl": 3.953125,
      "learning_rate": 1.36e-07,
      "loss": 0.322,
      "reward": 0.6035110950469971,
      "reward_std": 0.21830416470766068,
      "reward_std_all": 0.21516353636980057,
      "rewards-max/forensics_sinkhorn_reward": 0.9864811897277832,
      "rewards-max/forensics_strict_edge_reward": 0.8876297473907471,
      "rewards-max/forensics_tiou_reward": 0.9901179671287537,
      "rewards/forensics_sinkhorn_reward": 0.5286607444286346,
      "rewards/forensics_strict_edge_reward": 0.2555214762687683,
      "rewards/forensics_tiou_reward": 0.3290575444698334,
      "step": 1297
    },
    {
      "completion_length": 16.375,
      "epoch": 3.4527296937416776,
      "grad_norm": 29.76414680480957,
      "kl": 5.609375,
      "learning_rate": 1.3533333333333333e-07,
      "loss": 0.3514,
      "reward": 0.6117670238018036,
      "reward_std": 0.1947779692709446,
      "reward_std_all": 0.19703880697488785,
      "rewards-max/forensics_sinkhorn_reward": 0.987754225730896,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9951337277889252,
      "rewards/forensics_sinkhorn_reward": 0.5614950358867645,
      "rewards/forensics_strict_edge_reward": 0.27798745036125183,
      "rewards/forensics_tiou_reward": 0.39343613386154175,
      "step": 1298
    },
    {
      "completion_length": 18.40625,
      "epoch": 3.455392809587217,
      "grad_norm": 31.220863342285156,
      "kl": 4.765625,
      "learning_rate": 1.3466666666666665e-07,
      "loss": 0.2751,
      "reward": 0.5871379375457764,
      "reward_std": 0.18124642223119736,
      "reward_std_all": 0.207294799387455,
      "rewards-max/forensics_sinkhorn_reward": 0.9569615125656128,
      "rewards-max/forensics_strict_edge_reward": 0.8340697288513184,
      "rewards-max/forensics_tiou_reward": 0.9287327527999878,
      "rewards/forensics_sinkhorn_reward": 0.47489339113235474,
      "rewards/forensics_strict_edge_reward": 0.23394475132226944,
      "rewards/forensics_tiou_reward": 0.28010204434394836,
      "step": 1299
    },
    {
      "completion_length": 17.875,
      "epoch": 3.458055925432756,
      "grad_norm": 19.19270896911621,
      "kl": 4.984375,
      "learning_rate": 1.34e-07,
      "loss": 0.3812,
      "reward": 0.6256613731384277,
      "reward_std": 0.17031323164701462,
      "reward_std_all": 0.1709417849779129,
      "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": 0.5691660642623901,
      "rewards/forensics_strict_edge_reward": 0.3467411398887634,
      "rewards/forensics_tiou_reward": 0.42093174159526825,
      "step": 1300
    },
    {
      "completion_length": 18.171875,
      "epoch": 3.4607190412782955,
      "grad_norm": 20.911718368530273,
      "kl": 5.546875,
      "learning_rate": 1.3333333333333334e-07,
      "loss": 0.3493,
      "reward": 0.5593942701816559,
      "reward_std": 0.26647721230983734,
      "reward_std_all": 0.2614694833755493,
      "rewards-max/forensics_sinkhorn_reward": 0.9333541691303253,
      "rewards-max/forensics_strict_edge_reward": 0.8039028942584991,
      "rewards-max/forensics_tiou_reward": 0.873295247554779,
      "rewards/forensics_sinkhorn_reward": 0.45564818382263184,
      "rewards/forensics_strict_edge_reward": 0.14834477007389069,
      "rewards/forensics_tiou_reward": 0.24095265567302704,
      "step": 1301
    },
    {
      "completion_length": 15.609375,
      "epoch": 3.463382157123835,
      "grad_norm": 22.21153450012207,
      "kl": 4.4375,
      "learning_rate": 1.3266666666666664e-07,
      "loss": 0.3312,
      "reward": 0.6069111526012421,
      "reward_std": 0.19603152573108673,
      "reward_std_all": 0.19708827883005142,
      "rewards-max/forensics_sinkhorn_reward": 0.9826893508434296,
      "rewards-max/forensics_strict_edge_reward": 0.8286563754081726,
      "rewards-max/forensics_tiou_reward": 0.9754062294960022,
      "rewards/forensics_sinkhorn_reward": 0.5233809351921082,
      "rewards/forensics_strict_edge_reward": 0.2001330330967903,
      "rewards/forensics_tiou_reward": 0.3042389452457428,
      "step": 1302
    },
    {
      "completion_length": 16.984375,
      "epoch": 3.466045272969374,
      "grad_norm": 26.194856643676758,
      "kl": 7.359375,
      "learning_rate": 1.32e-07,
      "loss": 0.3983,
      "reward": 0.6190713047981262,
      "reward_std": 0.1707197055220604,
      "reward_std_all": 0.18229468166828156,
      "rewards-max/forensics_sinkhorn_reward": 0.965042918920517,
      "rewards-max/forensics_strict_edge_reward": 0.7852677702903748,
      "rewards-max/forensics_tiou_reward": 0.9427258968353271,
      "rewards/forensics_sinkhorn_reward": 0.5215311646461487,
      "rewards/forensics_strict_edge_reward": 0.17234815657138824,
      "rewards/forensics_tiou_reward": 0.26455287635326385,
      "step": 1303
    },
    {
      "completion_length": 18.09375,
      "epoch": 3.4687083888149135,
      "grad_norm": 19.980947494506836,
      "kl": 5.625,
      "learning_rate": 1.3133333333333332e-07,
      "loss": 0.4509,
      "reward": 0.6260487735271454,
      "reward_std": 0.23329471051692963,
      "reward_std_all": 0.21331332623958588,
      "rewards-max/forensics_sinkhorn_reward": 0.9884831309318542,
      "rewards-max/forensics_strict_edge_reward": 0.9375866651535034,
      "rewards-max/forensics_tiou_reward": 0.9979999959468842,
      "rewards/forensics_sinkhorn_reward": 0.5344756245613098,
      "rewards/forensics_strict_edge_reward": 0.33775241672992706,
      "rewards/forensics_tiou_reward": 0.3908322900533676,
      "step": 1304
    },
    {
      "completion_length": 16.0625,
      "epoch": 3.4713715046604525,
      "grad_norm": 21.032878875732422,
      "kl": 4.53125,
      "learning_rate": 1.3066666666666665e-07,
      "loss": 0.3109,
      "reward": 0.6542590856552124,
      "reward_std": 0.162485983222723,
      "reward_std_all": 0.15867335349321365,
      "rewards-max/forensics_sinkhorn_reward": 0.9433821141719818,
      "rewards-max/forensics_strict_edge_reward": 0.7872911989688873,
      "rewards-max/forensics_tiou_reward": 0.8996151089668274,
      "rewards/forensics_sinkhorn_reward": 0.6316927671432495,
      "rewards/forensics_strict_edge_reward": 0.34605446457862854,
      "rewards/forensics_tiou_reward": 0.47749993205070496,
      "step": 1305
    },
    {
      "completion_length": 17.28125,
      "epoch": 3.474034620505992,
      "grad_norm": 35.89109420776367,
      "kl": 6.140625,
      "learning_rate": 1.3e-07,
      "loss": 0.2558,
      "reward": 0.6339386701583862,
      "reward_std": 0.18979443609714508,
      "reward_std_all": 0.18197641521692276,
      "rewards-max/forensics_sinkhorn_reward": 0.9883260428905487,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9973817765712738,
      "rewards/forensics_sinkhorn_reward": 0.5896112024784088,
      "rewards/forensics_strict_edge_reward": 0.31247539818286896,
      "rewards/forensics_tiou_reward": 0.4424732029438019,
      "step": 1306
    },
    {
      "completion_length": 17.265625,
      "epoch": 3.4766977363515315,
      "grad_norm": 25.9211368560791,
      "kl": 7.0625,
      "learning_rate": 1.2933333333333333e-07,
      "loss": 0.5175,
      "reward": 0.6074480712413788,
      "reward_std": 0.24056345969438553,
      "reward_std_all": 0.22531873732805252,
      "rewards-max/forensics_sinkhorn_reward": 0.9868967235088348,
      "rewards-max/forensics_strict_edge_reward": 0.9357379674911499,
      "rewards-max/forensics_tiou_reward": 0.9917644560337067,
      "rewards/forensics_sinkhorn_reward": 0.518461674451828,
      "rewards/forensics_strict_edge_reward": 0.2559119686484337,
      "rewards/forensics_tiou_reward": 0.3604356646537781,
      "step": 1307
    },
    {
      "completion_length": 17.03125,
      "epoch": 3.4793608521970705,
      "grad_norm": 26.498409271240234,
      "kl": 6.1015625,
      "learning_rate": 1.2866666666666668e-07,
      "loss": 0.4011,
      "reward": 0.5622075796127319,
      "reward_std": 0.26791486144065857,
      "reward_std_all": 0.2561159133911133,
      "rewards-max/forensics_sinkhorn_reward": 0.988337516784668,
      "rewards-max/forensics_strict_edge_reward": 0.9574736058712006,
      "rewards-max/forensics_tiou_reward": 0.9974293112754822,
      "rewards/forensics_sinkhorn_reward": 0.46821992099285126,
      "rewards/forensics_strict_edge_reward": 0.2541821151971817,
      "rewards/forensics_tiou_reward": 0.30679693818092346,
      "step": 1308
    },
    {
      "completion_length": 15.515625,
      "epoch": 3.48202396804261,
      "grad_norm": 26.399044036865234,
      "kl": 2.796875,
      "learning_rate": 1.28e-07,
      "loss": 0.2483,
      "reward": 0.6770155429840088,
      "reward_std": 0.15224780142307281,
      "reward_std_all": 0.14698317274451256,
      "rewards-max/forensics_sinkhorn_reward": 0.9848774373531342,
      "rewards-max/forensics_strict_edge_reward": 0.9574736058712006,
      "rewards-max/forensics_tiou_reward": 0.9839275777339935,
      "rewards/forensics_sinkhorn_reward": 0.6559853553771973,
      "rewards/forensics_strict_edge_reward": 0.40786805003881454,
      "rewards/forensics_tiou_reward": 0.4778763949871063,
      "step": 1309
    },
    {
      "completion_length": 16.765625,
      "epoch": 3.484687083888149,
      "grad_norm": 24.521190643310547,
      "kl": 4.453125,
      "learning_rate": 1.273333333333333e-07,
      "loss": 0.303,
      "reward": 0.6563348770141602,
      "reward_std": 0.1727098971605301,
      "reward_std_all": 0.1704738810658455,
      "rewards-max/forensics_sinkhorn_reward": 0.9722528755664825,
      "rewards-max/forensics_strict_edge_reward": 0.947419673204422,
      "rewards-max/forensics_tiou_reward": 0.9627381563186646,
      "rewards/forensics_sinkhorn_reward": 0.6308155655860901,
      "rewards/forensics_strict_edge_reward": 0.28435076773166656,
      "rewards/forensics_tiou_reward": 0.45429322123527527,
      "step": 1310
    },
    {
      "completion_length": 17.375,
      "epoch": 3.4873501997336884,
      "grad_norm": 23.963788986206055,
      "kl": 5.0625,
      "learning_rate": 1.2666666666666666e-07,
      "loss": 0.2179,
      "reward": 0.5880671143531799,
      "reward_std": 0.2123487964272499,
      "reward_std_all": 0.2169744223356247,
      "rewards-max/forensics_sinkhorn_reward": 0.9706935286521912,
      "rewards-max/forensics_strict_edge_reward": 0.8876297473907471,
      "rewards-max/forensics_tiou_reward": 0.9592528343200684,
      "rewards/forensics_sinkhorn_reward": 0.48456738889217377,
      "rewards/forensics_strict_edge_reward": 0.1846485510468483,
      "rewards/forensics_tiou_reward": 0.27017877995967865,
      "step": 1311
    },
    {
      "completion_length": 16.546875,
      "epoch": 3.490013315579228,
      "grad_norm": 33.5937614440918,
      "kl": 5.5625,
      "learning_rate": 1.26e-07,
      "loss": 0.2947,
      "reward": 0.5864232182502747,
      "reward_std": 0.16001494228839874,
      "reward_std_all": 0.18883801996707916,
      "rewards-max/forensics_sinkhorn_reward": 0.9855501353740692,
      "rewards-max/forensics_strict_edge_reward": 0.8876297473907471,
      "rewards-max/forensics_tiou_reward": 0.986516535282135,
      "rewards/forensics_sinkhorn_reward": 0.5428782999515533,
      "rewards/forensics_strict_edge_reward": 0.26006777584552765,
      "rewards/forensics_tiou_reward": 0.3504050672054291,
      "step": 1312
    },
    {
      "completion_length": 17.640625,
      "epoch": 3.492676431424767,
      "grad_norm": 18.724061965942383,
      "kl": 6.0,
      "learning_rate": 1.2533333333333332e-07,
      "loss": 0.5454,
      "reward": 0.584712564945221,
      "reward_std": 0.22149740904569626,
      "reward_std_all": 0.23347359895706177,
      "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": 0.5749440491199493,
      "rewards/forensics_strict_edge_reward": 0.3227095603942871,
      "rewards/forensics_tiou_reward": 0.45165252685546875,
      "step": 1313
    },
    {
      "completion_length": 15.421875,
      "epoch": 3.4953395472703064,
      "grad_norm": 22.937345504760742,
      "kl": 4.25,
      "learning_rate": 1.2466666666666664e-07,
      "loss": 0.4055,
      "reward": 0.5972082912921906,
      "reward_std": 0.14188618957996368,
      "reward_std_all": 0.1715201959013939,
      "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": 0.5786938965320587,
      "rewards/forensics_strict_edge_reward": 0.3465312831103802,
      "rewards/forensics_tiou_reward": 0.40228863805532455,
      "step": 1314
    },
    {
      "completion_length": 15.84375,
      "epoch": 3.4980026631158454,
      "grad_norm": 19.426345825195312,
      "kl": 7.375,
      "learning_rate": 1.24e-07,
      "loss": 0.3907,
      "reward": 0.6307908892631531,
      "reward_std": 0.13740213960409164,
      "reward_std_all": 0.15392863750457764,
      "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": 0.5955367684364319,
      "rewards/forensics_strict_edge_reward": 0.29383547604084015,
      "rewards/forensics_tiou_reward": 0.3797476887702942,
      "step": 1315
    },
    {
      "completion_length": 16.390625,
      "epoch": 3.500665778961385,
      "grad_norm": 22.38463020324707,
      "kl": 4.84375,
      "learning_rate": 1.2333333333333333e-07,
      "loss": 0.4247,
      "reward": 0.5727401077747345,
      "reward_std": 0.21549122780561447,
      "reward_std_all": 0.22780795395374298,
      "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": 0.544891819357872,
      "rewards/forensics_strict_edge_reward": 0.3405746519565582,
      "rewards/forensics_tiou_reward": 0.4177851229906082,
      "step": 1316
    },
    {
      "completion_length": 18.65625,
      "epoch": 3.5033288948069243,
      "grad_norm": 17.226308822631836,
      "kl": 5.5625,
      "learning_rate": 1.2266666666666665e-07,
      "loss": 0.5437,
      "reward": 0.6190502643585205,
      "reward_std": 0.22487135976552963,
      "reward_std_all": 0.20924298465251923,
      "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": 0.6118767857551575,
      "rewards/forensics_strict_edge_reward": 0.4422624558210373,
      "rewards/forensics_tiou_reward": 0.5272335112094879,
      "step": 1317
    },
    {
      "completion_length": 18.015625,
      "epoch": 3.5059920106524634,
      "grad_norm": 34.25364685058594,
      "kl": 5.625,
      "learning_rate": 1.2199999999999998e-07,
      "loss": 0.209,
      "reward": 0.6019029021263123,
      "reward_std": 0.16132055968046188,
      "reward_std_all": 0.184825137257576,
      "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": 0.5146483182907104,
      "rewards/forensics_strict_edge_reward": 0.27962376177310944,
      "rewards/forensics_tiou_reward": 0.3190961629152298,
      "step": 1318
    },
    {
      "completion_length": 16.8125,
      "epoch": 3.508655126498003,
      "grad_norm": 23.506879806518555,
      "kl": 6.5,
      "learning_rate": 1.2133333333333333e-07,
      "loss": 0.4909,
      "reward": 0.5957365036010742,
      "reward_std": 0.15215438604354858,
      "reward_std_all": 0.17703401297330856,
      "rewards-max/forensics_sinkhorn_reward": 0.9848397672176361,
      "rewards-max/forensics_strict_edge_reward": 0.9158510267734528,
      "rewards-max/forensics_tiou_reward": 0.9837251901626587,
      "rewards/forensics_sinkhorn_reward": 0.49496161937713623,
      "rewards/forensics_strict_edge_reward": 0.27632106840610504,
      "rewards/forensics_tiou_reward": 0.31287239491939545,
      "step": 1319
    },
    {
      "completion_length": 18.609375,
      "epoch": 3.511318242343542,
      "grad_norm": 36.39756393432617,
      "kl": 6.421875,
      "learning_rate": 1.2066666666666666e-07,
      "loss": 0.4686,
      "reward": 0.6928084194660187,
      "reward_std": 0.16950976848602295,
      "reward_std_all": 0.15122536569833755,
      "rewards-max/forensics_sinkhorn_reward": 0.9879485666751862,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9958927631378174,
      "rewards/forensics_sinkhorn_reward": 0.6349018216133118,
      "rewards/forensics_strict_edge_reward": 0.3707757890224457,
      "rewards/forensics_tiou_reward": 0.46808579564094543,
      "step": 1320
    },
    {
      "completion_length": 15.0625,
      "epoch": 3.5139813581890813,
      "grad_norm": 22.45623779296875,
      "kl": 4.109375,
      "learning_rate": 1.2e-07,
      "loss": 0.3449,
      "reward": 0.6852050721645355,
      "reward_std": 0.12257236987352371,
      "reward_std_all": 0.1261277198791504,
      "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": 0.6398763656616211,
      "rewards/forensics_strict_edge_reward": 0.3775116950273514,
      "rewards/forensics_tiou_reward": 0.49272602796554565,
      "step": 1321
    },
    {
      "completion_length": 16.0,
      "epoch": 3.5166444740346208,
      "grad_norm": 19.578935623168945,
      "kl": 2.796875,
      "learning_rate": 1.1933333333333332e-07,
      "loss": 0.1964,
      "reward": 0.7028471231460571,
      "reward_std": 0.14271079376339912,
      "reward_std_all": 0.1319170668721199,
      "rewards-max/forensics_sinkhorn_reward": 0.9889183640480042,
      "rewards-max/forensics_strict_edge_reward": 0.9890114367008209,
      "rewards-max/forensics_tiou_reward": 0.9997156262397766,
      "rewards/forensics_sinkhorn_reward": 0.6403433084487915,
      "rewards/forensics_strict_edge_reward": 0.30660420656204224,
      "rewards/forensics_tiou_reward": 0.42755019664764404,
      "step": 1322
    },
    {
      "completion_length": 18.5625,
      "epoch": 3.51930758988016,
      "grad_norm": 19.873249053955078,
      "kl": 5.28125,
      "learning_rate": 1.1866666666666667e-07,
      "loss": 0.4616,
      "reward": 0.580627828836441,
      "reward_std": 0.22348341345787048,
      "reward_std_all": 0.22132375836372375,
      "rewards-max/forensics_sinkhorn_reward": 0.9858461618423462,
      "rewards-max/forensics_strict_edge_reward": 0.927969753742218,
      "rewards-max/forensics_tiou_reward": 0.9876106083393097,
      "rewards/forensics_sinkhorn_reward": 0.4902840107679367,
      "rewards/forensics_strict_edge_reward": 0.26516616344451904,
      "rewards/forensics_tiou_reward": 0.3507145941257477,
      "step": 1323
    },
    {
      "completion_length": 15.75,
      "epoch": 3.521970705725699,
      "grad_norm": 22.51825714111328,
      "kl": 5.46875,
      "learning_rate": 1.1799999999999998e-07,
      "loss": 0.3837,
      "reward": 0.638322114944458,
      "reward_std": 0.16831254214048386,
      "reward_std_all": 0.16882850974798203,
      "rewards-max/forensics_sinkhorn_reward": 0.9877705574035645,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9951917827129364,
      "rewards/forensics_sinkhorn_reward": 0.5671083033084869,
      "rewards/forensics_strict_edge_reward": 0.2820582687854767,
      "rewards/forensics_tiou_reward": 0.3884526193141937,
      "step": 1324
    },
    {
      "completion_length": 16.25,
      "epoch": 3.5246338215712383,
      "grad_norm": 28.22304344177246,
      "kl": 6.78125,
      "learning_rate": 1.1733333333333333e-07,
      "loss": 0.4971,
      "reward": 0.5750196576118469,
      "reward_std": 0.20987621694803238,
      "reward_std_all": 0.2224130854010582,
      "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": 0.5258187502622604,
      "rewards/forensics_strict_edge_reward": 0.31244035065174103,
      "rewards/forensics_tiou_reward": 0.3715520352125168,
      "step": 1325
    },
    {
      "completion_length": 16.375,
      "epoch": 3.5272969374167777,
      "grad_norm": 16.471643447875977,
      "kl": 4.96875,
      "learning_rate": 1.1666666666666667e-07,
      "loss": 0.2448,
      "reward": 0.6014877557754517,
      "reward_std": 0.15847112238407135,
      "reward_std_all": 0.17097210139036179,
      "rewards-max/forensics_sinkhorn_reward": 0.9542247653007507,
      "rewards-max/forensics_strict_edge_reward": 0.7309097647666931,
      "rewards-max/forensics_tiou_reward": 0.9161719977855682,
      "rewards/forensics_sinkhorn_reward": 0.5019232034683228,
      "rewards/forensics_strict_edge_reward": 0.23648400232195854,
      "rewards/forensics_tiou_reward": 0.3023322969675064,
      "step": 1326
    },
    {
      "completion_length": 16.703125,
      "epoch": 3.5299600532623168,
      "grad_norm": 27.505138397216797,
      "kl": 4.625,
      "learning_rate": 1.16e-07,
      "loss": 0.2999,
      "reward": 0.653016984462738,
      "reward_std": 0.12645653635263443,
      "reward_std_all": 0.14373043924570084,
      "rewards-max/forensics_sinkhorn_reward": 0.9883153736591339,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9973408281803131,
      "rewards/forensics_sinkhorn_reward": 0.6168185472488403,
      "rewards/forensics_strict_edge_reward": 0.2996668666601181,
      "rewards/forensics_tiou_reward": 0.36609169840812683,
      "step": 1327
    },
    {
      "completion_length": 16.6875,
      "epoch": 3.5326231691078562,
      "grad_norm": 25.054136276245117,
      "kl": 5.453125,
      "learning_rate": 1.1533333333333332e-07,
      "loss": 0.296,
      "reward": 0.6091611385345459,
      "reward_std": 0.1897176280617714,
      "reward_std_all": 0.1953100487589836,
      "rewards-max/forensics_sinkhorn_reward": 0.9885756969451904,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9983651041984558,
      "rewards/forensics_sinkhorn_reward": 0.55214424431324,
      "rewards/forensics_strict_edge_reward": 0.29177962243556976,
      "rewards/forensics_tiou_reward": 0.3853924572467804,
      "step": 1328
    },
    {
      "completion_length": 17.078125,
      "epoch": 3.5352862849533953,
      "grad_norm": 23.296428680419922,
      "kl": 4.1953125,
      "learning_rate": 1.1466666666666666e-07,
      "loss": 0.3101,
      "reward": 0.6226895749568939,
      "reward_std": 0.1855836808681488,
      "reward_std_all": 0.18600352108478546,
      "rewards-max/forensics_sinkhorn_reward": 0.9637121856212616,
      "rewards-max/forensics_strict_edge_reward": 0.895060271024704,
      "rewards-max/forensics_tiou_reward": 0.9438068270683289,
      "rewards/forensics_sinkhorn_reward": 0.5080640316009521,
      "rewards/forensics_strict_edge_reward": 0.21812991052865982,
      "rewards/forensics_tiou_reward": 0.30223381519317627,
      "step": 1329
    },
    {
      "completion_length": 16.46875,
      "epoch": 3.5379494007989347,
      "grad_norm": 16.62203025817871,
      "kl": 4.515625,
      "learning_rate": 1.14e-07,
      "loss": 0.3811,
      "reward": 0.6591126918792725,
      "reward_std": 0.17996541038155556,
      "reward_std_all": 0.16788523644208908,
      "rewards-max/forensics_sinkhorn_reward": 0.9885026514530182,
      "rewards-max/forensics_strict_edge_reward": 0.9782643616199493,
      "rewards-max/forensics_tiou_reward": 0.9980769157409668,
      "rewards/forensics_sinkhorn_reward": 0.6500225067138672,
      "rewards/forensics_strict_edge_reward": 0.4646037518978119,
      "rewards/forensics_tiou_reward": 0.5513061434030533,
      "step": 1330
    },
    {
      "completion_length": 15.484375,
      "epoch": 3.540612516644474,
      "grad_norm": 22.37997817993164,
      "kl": 4.0234375,
      "learning_rate": 1.1333333333333332e-07,
      "loss": 0.3572,
      "reward": 0.629082590341568,
      "reward_std": 0.16790468990802765,
      "reward_std_all": 0.17447185516357422,
      "rewards-max/forensics_sinkhorn_reward": 0.9875704050064087,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9944023787975311,
      "rewards/forensics_sinkhorn_reward": 0.6029388606548309,
      "rewards/forensics_strict_edge_reward": 0.33230243623256683,
      "rewards/forensics_tiou_reward": 0.4189899414777756,
      "step": 1331
    },
    {
      "completion_length": 15.5,
      "epoch": 3.543275632490013,
      "grad_norm": 35.41660690307617,
      "kl": 3.8828125,
      "learning_rate": 1.1266666666666666e-07,
      "loss": 0.3711,
      "reward": 0.6524689197540283,
      "reward_std": 0.13694502413272858,
      "reward_std_all": 0.1462503969669342,
      "rewards-max/forensics_sinkhorn_reward": 0.9873168170452118,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9934035539627075,
      "rewards/forensics_sinkhorn_reward": 0.6405392289161682,
      "rewards/forensics_strict_edge_reward": 0.36959871649742126,
      "rewards/forensics_tiou_reward": 0.45150090754032135,
      "step": 1332
    },
    {
      "completion_length": 19.578125,
      "epoch": 3.5459387483355527,
      "grad_norm": 25.762195587158203,
      "kl": 6.84375,
      "learning_rate": 1.12e-07,
      "loss": 0.4164,
      "reward": 0.47497087717056274,
      "reward_std": 0.17271514236927032,
      "reward_std_all": 0.2490084320306778,
      "rewards-max/forensics_sinkhorn_reward": 0.9806618988513947,
      "rewards-max/forensics_strict_edge_reward": 0.9782643616199493,
      "rewards-max/forensics_tiou_reward": 0.9816426634788513,
      "rewards/forensics_sinkhorn_reward": 0.3715825825929642,
      "rewards/forensics_strict_edge_reward": 0.1863480880856514,
      "rewards/forensics_tiou_reward": 0.23628933727741241,
      "step": 1333
    },
    {
      "completion_length": 15.6875,
      "epoch": 3.5486018641810917,
      "grad_norm": 28.922561645507812,
      "kl": 5.875,
      "learning_rate": 1.1133333333333334e-07,
      "loss": 0.5663,
      "reward": 0.608458012342453,
      "reward_std": 0.16717559471726418,
      "reward_std_all": 0.17879604920744896,
      "rewards-max/forensics_sinkhorn_reward": 0.9644637405872345,
      "rewards-max/forensics_strict_edge_reward": 0.9357379674911499,
      "rewards-max/forensics_tiou_reward": 0.9459145069122314,
      "rewards/forensics_sinkhorn_reward": 0.5136115401983261,
      "rewards/forensics_strict_edge_reward": 0.23277079313993454,
      "rewards/forensics_tiou_reward": 0.3080805689096451,
      "step": 1334
    },
    {
      "completion_length": 18.859375,
      "epoch": 3.551264980026631,
      "grad_norm": 36.088871002197266,
      "kl": 4.6953125,
      "learning_rate": 1.1066666666666665e-07,
      "loss": 0.3824,
      "reward": 0.5580702722072601,
      "reward_std": 0.2606223374605179,
      "reward_std_all": 0.2586488574743271,
      "rewards-max/forensics_sinkhorn_reward": 0.9597757458686829,
      "rewards-max/forensics_strict_edge_reward": 0.8404377698898315,
      "rewards-max/forensics_tiou_reward": 0.9279808402061462,
      "rewards/forensics_sinkhorn_reward": 0.47789517045021057,
      "rewards/forensics_strict_edge_reward": 0.23569786548614502,
      "rewards/forensics_tiou_reward": 0.2739035412669182,
      "step": 1335
    },
    {
      "completion_length": 17.828125,
      "epoch": 3.5539280958721706,
      "grad_norm": 26.335886001586914,
      "kl": 5.578125,
      "learning_rate": 1.0999999999999999e-07,
      "loss": 0.4742,
      "reward": 0.5495156645774841,
      "reward_std": 0.23930798470973969,
      "reward_std_all": 0.25019849836826324,
      "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": 0.4935128092765808,
      "rewards/forensics_strict_edge_reward": 0.2554618716239929,
      "rewards/forensics_tiou_reward": 0.35246923565864563,
      "step": 1336
    },
    {
      "completion_length": 17.328125,
      "epoch": 3.5565912117177096,
      "grad_norm": 19.387893676757812,
      "kl": 5.21875,
      "learning_rate": 1.0933333333333333e-07,
      "loss": 0.3573,
      "reward": 0.6043143272399902,
      "reward_std": 0.2165197879076004,
      "reward_std_all": 0.21175270527601242,
      "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": 0.5704877078533173,
      "rewards/forensics_strict_edge_reward": 0.39289991557598114,
      "rewards/forensics_tiou_reward": 0.45001696050167084,
      "step": 1337
    },
    {
      "completion_length": 17.8125,
      "epoch": 3.559254327563249,
      "grad_norm": 19.186540603637695,
      "kl": 5.859375,
      "learning_rate": 1.0866666666666666e-07,
      "loss": 0.3707,
      "reward": 0.5604724884033203,
      "reward_std": 0.15909797698259354,
      "reward_std_all": 0.20237644016742706,
      "rewards-max/forensics_sinkhorn_reward": 0.9877974987030029,
      "rewards-max/forensics_strict_edge_reward": 0.9158510267734528,
      "rewards-max/forensics_tiou_reward": 0.9952972829341888,
      "rewards/forensics_sinkhorn_reward": 0.5150637328624725,
      "rewards/forensics_strict_edge_reward": 0.2764642685651779,
      "rewards/forensics_tiou_reward": 0.35848112404346466,
      "step": 1338
    },
    {
      "completion_length": 19.265625,
      "epoch": 3.561917443408788,
      "grad_norm": 23.730390548706055,
      "kl": 6.984375,
      "learning_rate": 1.0799999999999999e-07,
      "loss": 0.5036,
      "reward": 0.5813319981098175,
      "reward_std": 0.24477021396160126,
      "reward_std_all": 0.23380213975906372,
      "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": 0.5377984046936035,
      "rewards/forensics_strict_edge_reward": 0.3186171352863312,
      "rewards/forensics_tiou_reward": 0.3922095000743866,
      "step": 1339
    },
    {
      "completion_length": 16.625,
      "epoch": 3.5645805592543276,
      "grad_norm": 23.357158660888672,
      "kl": 7.078125,
      "learning_rate": 1.0733333333333333e-07,
      "loss": 0.3186,
      "reward": 0.5559919476509094,
      "reward_std": 0.1358706746250391,
      "reward_std_all": 0.19406907260417938,
      "rewards-max/forensics_sinkhorn_reward": 0.9886516034603119,
      "rewards-max/forensics_strict_edge_reward": 0.9574736058712006,
      "rewards-max/forensics_tiou_reward": 0.9986648857593536,
      "rewards/forensics_sinkhorn_reward": 0.5134590566158295,
      "rewards/forensics_strict_edge_reward": 0.2310175821185112,
      "rewards/forensics_tiou_reward": 0.34230442345142365,
      "step": 1340
    },
    {
      "completion_length": 17.390625,
      "epoch": 3.567243675099867,
      "grad_norm": 17.062742233276367,
      "kl": 5.6640625,
      "learning_rate": 1.0666666666666667e-07,
      "loss": 0.4717,
      "reward": 0.5795726478099823,
      "reward_std": 0.20411261171102524,
      "reward_std_all": 0.21807977557182312,
      "rewards-max/forensics_sinkhorn_reward": 0.9872287809848785,
      "rewards-max/forensics_strict_edge_reward": 0.9158510267734528,
      "rewards-max/forensics_tiou_reward": 0.9930697977542877,
      "rewards/forensics_sinkhorn_reward": 0.5218830108642578,
      "rewards/forensics_strict_edge_reward": 0.31438663601875305,
      "rewards/forensics_tiou_reward": 0.3758547604084015,
      "step": 1341
    },
    {
      "completion_length": 15.984375,
      "epoch": 3.569906790945406,
      "grad_norm": 24.893962860107422,
      "kl": 5.859375,
      "learning_rate": 1.06e-07,
      "loss": 0.4241,
      "reward": 0.5509985387325287,
      "reward_std": 0.1563960239291191,
      "reward_std_all": 0.19611962139606476,
      "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": 0.5173479616641998,
      "rewards/forensics_strict_edge_reward": 0.2810535728931427,
      "rewards/forensics_tiou_reward": 0.3408614695072174,
      "step": 1342
    },
    {
      "completion_length": 17.984375,
      "epoch": 3.5725699067909455,
      "grad_norm": 24.172067642211914,
      "kl": 4.546875,
      "learning_rate": 1.0533333333333332e-07,
      "loss": 0.3431,
      "reward": 0.5836047828197479,
      "reward_std": 0.2650896906852722,
      "reward_std_all": 0.25030840933322906,
      "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": 0.5441579222679138,
      "rewards/forensics_strict_edge_reward": 0.33693838119506836,
      "rewards/forensics_tiou_reward": 0.39493128657341003,
      "step": 1343
    },
    {
      "completion_length": 19.21875,
      "epoch": 3.5752330226364846,
      "grad_norm": 24.31282615661621,
      "kl": 4.734375,
      "learning_rate": 1.0466666666666666e-07,
      "loss": 0.2807,
      "reward": 0.5582239925861359,
      "reward_std": 0.2111499309539795,
      "reward_std_all": 0.23055525124073029,
      "rewards-max/forensics_sinkhorn_reward": 0.9762496948242188,
      "rewards-max/forensics_strict_edge_reward": 0.7988545596599579,
      "rewards-max/forensics_tiou_reward": 0.9700416326522827,
      "rewards/forensics_sinkhorn_reward": 0.48169951140880585,
      "rewards/forensics_strict_edge_reward": 0.2470853552222252,
      "rewards/forensics_tiou_reward": 0.30903999507427216,
      "step": 1344
    },
    {
      "completion_length": 17.046875,
      "epoch": 3.577896138482024,
      "grad_norm": 18.997861862182617,
      "kl": 5.84375,
      "learning_rate": 1.0399999999999999e-07,
      "loss": 0.329,
      "reward": 0.5067426115274429,
      "reward_std": 0.17477139830589294,
      "reward_std_all": 0.22747370600700378,
      "rewards-max/forensics_sinkhorn_reward": 0.9871205985546112,
      "rewards-max/forensics_strict_edge_reward": 0.9565287232398987,
      "rewards-max/forensics_tiou_reward": 0.9926342666149139,
      "rewards/forensics_sinkhorn_reward": 0.4109306037425995,
      "rewards/forensics_strict_edge_reward": 0.12499993294477463,
      "rewards/forensics_tiou_reward": 0.20377449691295624,
      "step": 1345
    },
    {
      "completion_length": 15.84375,
      "epoch": 3.5805592543275635,
      "grad_norm": 28.249988555908203,
      "kl": 3.55078125,
      "learning_rate": 1.0333333333333333e-07,
      "loss": 0.3499,
      "reward": 0.6452495455741882,
      "reward_std": 0.17970824241638184,
      "reward_std_all": 0.17213188856840134,
      "rewards-max/forensics_sinkhorn_reward": 0.987281858921051,
      "rewards-max/forensics_strict_edge_reward": 0.9357379674911499,
      "rewards-max/forensics_tiou_reward": 0.9932758212089539,
      "rewards/forensics_sinkhorn_reward": 0.6106102168560028,
      "rewards/forensics_strict_edge_reward": 0.3524608016014099,
      "rewards/forensics_tiou_reward": 0.46660204231739044,
      "step": 1346
    },
    {
      "completion_length": 17.390625,
      "epoch": 3.5832223701731025,
      "grad_norm": 19.682987213134766,
      "kl": 5.125,
      "learning_rate": 1.0266666666666666e-07,
      "loss": 0.452,
      "reward": 0.5050547271966934,
      "reward_std": 0.2596851885318756,
      "reward_std_all": 0.2707142233848572,
      "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": 0.4584132432937622,
      "rewards/forensics_strict_edge_reward": 0.22858452796936035,
      "rewards/forensics_tiou_reward": 0.2914908826351166,
      "step": 1347
    },
    {
      "completion_length": 18.03125,
      "epoch": 3.5858854860186415,
      "grad_norm": 29.96619987487793,
      "kl": 4.4921875,
      "learning_rate": 1.0199999999999999e-07,
      "loss": 0.2459,
      "reward": 0.4911200851202011,
      "reward_std": 0.2495013251900673,
      "reward_std_all": 0.28281089663505554,
      "rewards-max/forensics_sinkhorn_reward": 0.9451926946640015,
      "rewards-max/forensics_strict_edge_reward": 0.8286563754081726,
      "rewards-max/forensics_tiou_reward": 0.9006274342536926,
      "rewards/forensics_sinkhorn_reward": 0.3820437788963318,
      "rewards/forensics_strict_edge_reward": 0.14942743629217148,
      "rewards/forensics_tiou_reward": 0.19661954790353775,
      "step": 1348
    },
    {
      "completion_length": 17.953125,
      "epoch": 3.588548601864181,
      "grad_norm": 19.496694564819336,
      "kl": 4.2265625,
      "learning_rate": 1.0133333333333333e-07,
      "loss": 0.3861,
      "reward": 0.590889185667038,
      "reward_std": 0.19178813695907593,
      "reward_std_all": 0.21085866540670395,
      "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": 0.5400447845458984,
      "rewards/forensics_strict_edge_reward": 0.2790815383195877,
      "rewards/forensics_tiou_reward": 0.3983256220817566,
      "step": 1349
    },
    {
      "completion_length": 16.71875,
      "epoch": 3.5912117177097205,
      "grad_norm": 28.60277557373047,
      "kl": 5.875,
      "learning_rate": 1.0066666666666667e-07,
      "loss": 0.4484,
      "reward": 0.5336947441101074,
      "reward_std": 0.20504064857959747,
      "reward_std_all": 0.23352710157632828,
      "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": 0.5043561458587646,
      "rewards/forensics_strict_edge_reward": 0.2672423720359802,
      "rewards/forensics_tiou_reward": 0.36927783489227295,
      "step": 1350
    },
    {
      "completion_length": 18.96875,
      "epoch": 3.5938748335552595,
      "grad_norm": 28.458311080932617,
      "kl": 5.46875,
      "learning_rate": 1e-07,
      "loss": 0.4476,
      "reward": 0.5534691214561462,
      "reward_std": 0.2461521476507187,
      "reward_std_all": 0.26056160032749176,
      "rewards-max/forensics_sinkhorn_reward": 0.9882973730564117,
      "rewards-max/forensics_strict_edge_reward": 0.9782643616199493,
      "rewards-max/forensics_tiou_reward": 0.9972714781761169,
      "rewards/forensics_sinkhorn_reward": 0.4707842469215393,
      "rewards/forensics_strict_edge_reward": 0.21810635924339294,
      "rewards/forensics_tiou_reward": 0.325370192527771,
      "step": 1351
    },
    {
      "completion_length": 17.609375,
      "epoch": 3.596537949400799,
      "grad_norm": 23.246490478515625,
      "kl": 3.796875,
      "learning_rate": 9.933333333333332e-08,
      "loss": 0.3146,
      "reward": 0.6024103760719299,
      "reward_std": 0.1570569947361946,
      "reward_std_all": 0.18397897481918335,
      "rewards-max/forensics_sinkhorn_reward": 0.8602488338947296,
      "rewards-max/forensics_strict_edge_reward": 0.6598812341690063,
      "rewards-max/forensics_tiou_reward": 0.7380506098270416,
      "rewards/forensics_sinkhorn_reward": 0.4800664037466049,
      "rewards/forensics_strict_edge_reward": 0.22683560848236084,
      "rewards/forensics_tiou_reward": 0.2592950314283371,
      "step": 1352
    },
    {
      "completion_length": 16.46875,
      "epoch": 3.599201065246338,
      "grad_norm": 27.175800323486328,
      "kl": 3.7734375,
      "learning_rate": 9.866666666666666e-08,
      "loss": 0.2276,
      "reward": 0.6288050711154938,
      "reward_std": 0.2175077646970749,
      "reward_std_all": 0.20575565844774246,
      "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": 0.566019743680954,
      "rewards/forensics_strict_edge_reward": 0.3323451578617096,
      "rewards/forensics_tiou_reward": 0.42009028792381287,
      "step": 1353
    },
    {
      "completion_length": 19.453125,
      "epoch": 3.6018641810918774,
      "grad_norm": 18.69059181213379,
      "kl": 6.0625,
      "learning_rate": 9.8e-08,
      "loss": 0.4246,
      "reward": 0.5099848508834839,
      "reward_std": 0.24461393803358078,
      "reward_std_all": 0.26870593428611755,
      "rewards-max/forensics_sinkhorn_reward": 0.9879729151725769,
      "rewards-max/forensics_strict_edge_reward": 0.9265981018543243,
      "rewards-max/forensics_tiou_reward": 0.9959928095340729,
      "rewards/forensics_sinkhorn_reward": 0.4224112927913666,
      "rewards/forensics_strict_edge_reward": 0.2383756935596466,
      "rewards/forensics_tiou_reward": 0.2823403775691986,
      "step": 1354
    },
    {
      "completion_length": 17.828125,
      "epoch": 3.604527296937417,
      "grad_norm": 23.000873565673828,
      "kl": 4.0703125,
      "learning_rate": 9.733333333333332e-08,
      "loss": 0.2674,
      "reward": 0.6271379292011261,
      "reward_std": 0.18522804975509644,
      "reward_std_all": 0.18417862057685852,
      "rewards-max/forensics_sinkhorn_reward": 0.9154577255249023,
      "rewards-max/forensics_strict_edge_reward": 0.7507967054843903,
      "rewards-max/forensics_tiou_reward": 0.8275088369846344,
      "rewards/forensics_sinkhorn_reward": 0.5407975614070892,
      "rewards/forensics_strict_edge_reward": 0.268424853682518,
      "rewards/forensics_tiou_reward": 0.3854202628135681,
      "step": 1355
    },
    {
      "completion_length": 16.890625,
      "epoch": 3.607190412782956,
      "grad_norm": 23.744163513183594,
      "kl": 5.6171875,
      "learning_rate": 9.666666666666666e-08,
      "loss": 0.506,
      "reward": 0.5957823395729065,
      "reward_std": 0.18621611595153809,
      "reward_std_all": 0.20254766941070557,
      "rewards-max/forensics_sinkhorn_reward": 0.9885318279266357,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9981927871704102,
      "rewards/forensics_sinkhorn_reward": 0.594994306564331,
      "rewards/forensics_strict_edge_reward": 0.39784252643585205,
      "rewards/forensics_tiou_reward": 0.4542558789253235,
      "step": 1356
    },
    {
      "completion_length": 17.6875,
      "epoch": 3.6098535286284954,
      "grad_norm": 22.025644302368164,
      "kl": 4.453125,
      "learning_rate": 9.6e-08,
      "loss": 0.2521,
      "reward": 0.6081858575344086,
      "reward_std": 0.199452705681324,
      "reward_std_all": 0.20016102492809296,
      "rewards-max/forensics_sinkhorn_reward": 0.9377465546131134,
      "rewards-max/forensics_strict_edge_reward": 0.882964164018631,
      "rewards-max/forensics_tiou_reward": 0.8901573121547699,
      "rewards/forensics_sinkhorn_reward": 0.5385462045669556,
      "rewards/forensics_strict_edge_reward": 0.30795541405677795,
      "rewards/forensics_tiou_reward": 0.36756888031959534,
      "step": 1357
    },
    {
      "completion_length": 17.96875,
      "epoch": 3.6125166444740344,
      "grad_norm": 27.528793334960938,
      "kl": 4.2109375,
      "learning_rate": 9.533333333333334e-08,
      "loss": 0.409,
      "reward": 0.632491409778595,
      "reward_std": 0.1955854445695877,
      "reward_std_all": 0.19013123214244843,
      "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": 0.5570907592773438,
      "rewards/forensics_strict_edge_reward": 0.30897940695285797,
      "rewards/forensics_tiou_reward": 0.3856314420700073,
      "step": 1358
    },
    {
      "completion_length": 16.5,
      "epoch": 3.615179760319574,
      "grad_norm": 33.20292663574219,
      "kl": 3.96875,
      "learning_rate": 9.466666666666665e-08,
      "loss": 0.3861,
      "reward": 0.6651857793331146,
      "reward_std": 0.12874462082982063,
      "reward_std_all": 0.14005355164408684,
      "rewards-max/forensics_sinkhorn_reward": 0.9885065257549286,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9980923533439636,
      "rewards/forensics_sinkhorn_reward": 0.653927743434906,
      "rewards/forensics_strict_edge_reward": 0.4556534141302109,
      "rewards/forensics_tiou_reward": 0.5486071109771729,
      "step": 1359
    },
    {
      "completion_length": 20.0,
      "epoch": 3.6178428761651134,
      "grad_norm": 13.7253999710083,
      "kl": 4.5390625,
      "learning_rate": 9.4e-08,
      "loss": 0.4226,
      "reward": 0.6217256188392639,
      "reward_std": 0.19601954519748688,
      "reward_std_all": 0.19111673533916473,
      "rewards-max/forensics_sinkhorn_reward": 0.9875102043151855,
      "rewards-max/forensics_strict_edge_reward": 0.9464850425720215,
      "rewards-max/forensics_tiou_reward": 0.9941733181476593,
      "rewards/forensics_sinkhorn_reward": 0.601534977555275,
      "rewards/forensics_strict_edge_reward": 0.3470635861158371,
      "rewards/forensics_tiou_reward": 0.46965697407722473,
      "step": 1360
    },
    {
      "completion_length": 16.46875,
      "epoch": 3.6205059920106524,
      "grad_norm": 26.25345230102539,
      "kl": 4.3046875,
      "learning_rate": 9.333333333333334e-08,
      "loss": 0.329,
      "reward": 0.5796763300895691,
      "reward_std": 0.20682571083307266,
      "reward_std_all": 0.2092869058251381,
      "rewards-max/forensics_sinkhorn_reward": 0.9878104627132416,
      "rewards-max/forensics_strict_edge_reward": 0.947419673204422,
      "rewards-max/forensics_tiou_reward": 0.9953488409519196,
      "rewards/forensics_sinkhorn_reward": 0.5451785027980804,
      "rewards/forensics_strict_edge_reward": 0.28362249583005905,
      "rewards/forensics_tiou_reward": 0.38023367524147034,
      "step": 1361
    },
    {
      "completion_length": 16.765625,
      "epoch": 3.623169107856192,
      "grad_norm": 19.433563232421875,
      "kl": 4.9765625,
      "learning_rate": 9.266666666666666e-08,
      "loss": 0.4522,
      "reward": 0.592141181230545,
      "reward_std": 0.255703866481781,
      "reward_std_all": 0.23893168568611145,
      "rewards-max/forensics_sinkhorn_reward": 0.9867119193077087,
      "rewards-max/forensics_strict_edge_reward": 0.8528120815753937,
      "rewards-max/forensics_tiou_reward": 0.9910501837730408,
      "rewards/forensics_sinkhorn_reward": 0.5530288517475128,
      "rewards/forensics_strict_edge_reward": 0.31382060050964355,
      "rewards/forensics_tiou_reward": 0.4159005582332611,
      "step": 1362
    },
    {
      "completion_length": 18.640625,
      "epoch": 3.625832223701731,
      "grad_norm": 28.060184478759766,
      "kl": 3.9375,
      "learning_rate": 9.199999999999999e-08,
      "loss": 0.2994,
      "reward": 0.6073687076568604,
      "reward_std": 0.15767142176628113,
      "reward_std_all": 0.17385795712471008,
      "rewards-max/forensics_sinkhorn_reward": 0.9884459972381592,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.997853547334671,
      "rewards/forensics_sinkhorn_reward": 0.5511875748634338,
      "rewards/forensics_strict_edge_reward": 0.310518279671669,
      "rewards/forensics_tiou_reward": 0.398232102394104,
      "step": 1363
    },
    {
      "completion_length": 16.796875,
      "epoch": 3.6284953395472703,
      "grad_norm": 21.50717544555664,
      "kl": 6.21875,
      "learning_rate": 9.133333333333333e-08,
      "loss": 0.4482,
      "reward": 0.6111296713352203,
      "reward_std": 0.25148484110832214,
      "reward_std_all": 0.23031380772590637,
      "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": 0.5536040663719177,
      "rewards/forensics_strict_edge_reward": 0.33273857831954956,
      "rewards/forensics_tiou_reward": 0.4019326865673065,
      "step": 1364
    },
    {
      "completion_length": 16.515625,
      "epoch": 3.63115845539281,
      "grad_norm": 44.48891067504883,
      "kl": 7.46875,
      "learning_rate": 9.066666666666667e-08,
      "loss": 0.5206,
      "reward": 0.6124023795127869,
      "reward_std": 0.1922491118311882,
      "reward_std_all": 0.2009899839758873,
      "rewards-max/forensics_sinkhorn_reward": 0.9879500865936279,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9958986937999725,
      "rewards/forensics_sinkhorn_reward": 0.5780904293060303,
      "rewards/forensics_strict_edge_reward": 0.3205275535583496,
      "rewards/forensics_tiou_reward": 0.40558165311813354,
      "step": 1365
    },
    {
      "completion_length": 16.953125,
      "epoch": 3.633821571238349,
      "grad_norm": 25.256494522094727,
      "kl": 6.859375,
      "learning_rate": 9e-08,
      "loss": 0.5551,
      "reward": 0.5045101344585419,
      "reward_std": 0.21935611218214035,
      "reward_std_all": 0.2531844675540924,
      "rewards-max/forensics_sinkhorn_reward": 0.9479898810386658,
      "rewards-max/forensics_strict_edge_reward": 0.8426913917064667,
      "rewards-max/forensics_tiou_reward": 0.9045924246311188,
      "rewards/forensics_sinkhorn_reward": 0.40041719377040863,
      "rewards/forensics_strict_edge_reward": 0.14774507097899914,
      "rewards/forensics_tiou_reward": 0.20514760166406631,
      "step": 1366
    },
    {
      "completion_length": 16.96875,
      "epoch": 3.6364846870838883,
      "grad_norm": 20.47058868408203,
      "kl": 3.7109375,
      "learning_rate": 8.933333333333333e-08,
      "loss": 0.2467,
      "reward": 0.7178090810775757,
      "reward_std": 0.12945079058408737,
      "reward_std_all": 0.11898856237530708,
      "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": 0.6950218975543976,
      "rewards/forensics_strict_edge_reward": 0.3762269616127014,
      "rewards/forensics_tiou_reward": 0.5092869251966476,
      "step": 1367
    },
    {
      "completion_length": 17.578125,
      "epoch": 3.6391478029294273,
      "grad_norm": 18.209793090820312,
      "kl": 7.640625,
      "learning_rate": 8.866666666666667e-08,
      "loss": 0.5614,
      "reward": 0.5539568066596985,
      "reward_std": 0.24495116621255875,
      "reward_std_all": 0.24154825508594513,
      "rewards-max/forensics_sinkhorn_reward": 0.9839354455471039,
      "rewards-max/forensics_strict_edge_reward": 0.927969753742218,
      "rewards-max/forensics_tiou_reward": 0.98857781291008,
      "rewards/forensics_sinkhorn_reward": 0.4810844957828522,
      "rewards/forensics_strict_edge_reward": 0.24670524895191193,
      "rewards/forensics_tiou_reward": 0.30333347618579865,
      "step": 1368
    },
    {
      "completion_length": 18.0625,
      "epoch": 3.6418109187749668,
      "grad_norm": 25.89984703063965,
      "kl": 8.421875,
      "learning_rate": 8.8e-08,
      "loss": 0.6458,
      "reward": 0.5044728517532349,
      "reward_std": 0.1950920969247818,
      "reward_std_all": 0.23734799772500992,
      "rewards-max/forensics_sinkhorn_reward": 0.9767075181007385,
      "rewards-max/forensics_strict_edge_reward": 0.8283384442329407,
      "rewards-max/forensics_tiou_reward": 0.9687630236148834,
      "rewards/forensics_sinkhorn_reward": 0.44385142624378204,
      "rewards/forensics_strict_edge_reward": 0.208572618663311,
      "rewards/forensics_tiou_reward": 0.28506850451231003,
      "step": 1369
    },
    {
      "completion_length": 15.46875,
      "epoch": 3.6444740346205062,
      "grad_norm": 23.417068481445312,
      "kl": 3.625,
      "learning_rate": 8.733333333333333e-08,
      "loss": 0.3297,
      "reward": 0.6693763136863708,
      "reward_std": 0.1621510237455368,
      "reward_std_all": 0.1522538661956787,
      "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": 0.6497200727462769,
      "rewards/forensics_strict_edge_reward": 0.43986931443214417,
      "rewards/forensics_tiou_reward": 0.5311939716339111,
      "step": 1370
    },
    {
      "completion_length": 15.109375,
      "epoch": 3.6471371504660453,
      "grad_norm": 38.512699127197266,
      "kl": 8.375,
      "learning_rate": 8.666666666666666e-08,
      "loss": 0.5985,
      "reward": 0.4890822470188141,
      "reward_std": 0.22967000305652618,
      "reward_std_all": 0.2612054646015167,
      "rewards-max/forensics_sinkhorn_reward": 0.9884831309318542,
      "rewards-max/forensics_strict_edge_reward": 0.9375866651535034,
      "rewards-max/forensics_tiou_reward": 0.9979999959468842,
      "rewards/forensics_sinkhorn_reward": 0.43035632371902466,
      "rewards/forensics_strict_edge_reward": 0.21299929171800613,
      "rewards/forensics_tiou_reward": 0.2989126071333885,
      "step": 1371
    },
    {
      "completion_length": 15.8125,
      "epoch": 3.6498002663115847,
      "grad_norm": 27.60525131225586,
      "kl": 6.296875,
      "learning_rate": 8.599999999999999e-08,
      "loss": 0.4081,
      "reward": 0.5608445703983307,
      "reward_std": 0.2268206626176834,
      "reward_std_all": 0.2370138093829155,
      "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": 0.5568228662014008,
      "rewards/forensics_strict_edge_reward": 0.37122607231140137,
      "rewards/forensics_tiou_reward": 0.43033677339553833,
      "step": 1372
    },
    {
      "completion_length": 17.921875,
      "epoch": 3.6524633821571237,
      "grad_norm": 60.38167953491211,
      "kl": 11.34375,
      "learning_rate": 8.533333333333333e-08,
      "loss": 0.6789,
      "reward": 0.46634624898433685,
      "reward_std": 0.2856036275625229,
      "reward_std_all": 0.31412671506404877,
      "rewards-max/forensics_sinkhorn_reward": 0.983563631772995,
      "rewards-max/forensics_strict_edge_reward": 0.9265981018543243,
      "rewards-max/forensics_tiou_reward": 0.9788144826889038,
      "rewards/forensics_sinkhorn_reward": 0.4044608175754547,
      "rewards/forensics_strict_edge_reward": 0.25421851873397827,
      "rewards/forensics_tiou_reward": 0.30122673511505127,
      "step": 1373
    },
    {
      "completion_length": 17.953125,
      "epoch": 3.655126498002663,
      "grad_norm": 31.665496826171875,
      "kl": 7.0,
      "learning_rate": 8.466666666666667e-08,
      "loss": 0.4667,
      "reward": 0.5584785044193268,
      "reward_std": 0.23352427035570145,
      "reward_std_all": 0.23995286226272583,
      "rewards-max/forensics_sinkhorn_reward": 0.988534539937973,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9982029497623444,
      "rewards/forensics_sinkhorn_reward": 0.5019065141677856,
      "rewards/forensics_strict_edge_reward": 0.32856012880802155,
      "rewards/forensics_tiou_reward": 0.3988433927297592,
      "step": 1374
    },
    {
      "completion_length": 17.203125,
      "epoch": 3.6577896138482027,
      "grad_norm": 25.159671783447266,
      "kl": 6.25,
      "learning_rate": 8.4e-08,
      "loss": 0.3488,
      "reward": 0.5806022584438324,
      "reward_std": 0.23553466796875,
      "reward_std_all": 0.2301798164844513,
      "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": 0.5338291525840759,
      "rewards/forensics_strict_edge_reward": 0.2943037897348404,
      "rewards/forensics_tiou_reward": 0.3595481142401695,
      "step": 1375
    },
    {
      "completion_length": 17.125,
      "epoch": 3.6604527296937417,
      "grad_norm": 16.357994079589844,
      "kl": 8.6875,
      "learning_rate": 8.333333333333333e-08,
      "loss": 0.6552,
      "reward": 0.531937450170517,
      "reward_std": 0.2519400119781494,
      "reward_std_all": 0.2609880268573761,
      "rewards-max/forensics_sinkhorn_reward": 0.9793817102909088,
      "rewards-max/forensics_strict_edge_reward": 0.9574736058712006,
      "rewards-max/forensics_tiou_reward": 0.9718416929244995,
      "rewards/forensics_sinkhorn_reward": 0.47516731917858124,
      "rewards/forensics_strict_edge_reward": 0.21963566541671753,
      "rewards/forensics_tiou_reward": 0.3563367575407028,
      "step": 1376
    },
    {
      "completion_length": 16.421875,
      "epoch": 3.6631158455392807,
      "grad_norm": 27.864824295043945,
      "kl": 7.0,
      "learning_rate": 8.266666666666667e-08,
      "loss": 0.5093,
      "reward": 0.5722364485263824,
      "reward_std": 0.1795778051018715,
      "reward_std_all": 0.21096375584602356,
      "rewards-max/forensics_sinkhorn_reward": 0.9878188371658325,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9953814446926117,
      "rewards/forensics_sinkhorn_reward": 0.4710579514503479,
      "rewards/forensics_strict_edge_reward": 0.2634393498301506,
      "rewards/forensics_tiou_reward": 0.32578492164611816,
      "step": 1377
    },
    {
      "completion_length": 17.703125,
      "epoch": 3.66577896138482,
      "grad_norm": 30.834440231323242,
      "kl": 8.625,
      "learning_rate": 8.2e-08,
      "loss": 0.4068,
      "reward": 0.5894818305969238,
      "reward_std": 0.21386003494262695,
      "reward_std_all": 0.20960243791341782,
      "rewards-max/forensics_sinkhorn_reward": 0.9856424629688263,
      "rewards-max/forensics_strict_edge_reward": 0.9357379674911499,
      "rewards-max/forensics_tiou_reward": 0.9868722558021545,
      "rewards/forensics_sinkhorn_reward": 0.517689049243927,
      "rewards/forensics_strict_edge_reward": 0.2881162017583847,
      "rewards/forensics_tiou_reward": 0.3387150168418884,
      "step": 1378
    },
    {
      "completion_length": 16.328125,
      "epoch": 3.6684420772303596,
      "grad_norm": 22.20235824584961,
      "kl": 4.125,
      "learning_rate": 8.133333333333332e-08,
      "loss": 0.246,
      "reward": 0.6068356335163116,
      "reward_std": 0.15771130472421646,
      "reward_std_all": 0.17223672568798065,
      "rewards-max/forensics_sinkhorn_reward": 0.9886134266853333,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9985139667987823,
      "rewards/forensics_sinkhorn_reward": 0.566899985074997,
      "rewards/forensics_strict_edge_reward": 0.32117752730846405,
      "rewards/forensics_tiou_reward": 0.38665635883808136,
      "step": 1379
    },
    {
      "completion_length": 19.15625,
      "epoch": 3.6711051930758987,
      "grad_norm": 24.700742721557617,
      "kl": 7.359375,
      "learning_rate": 8.066666666666666e-08,
      "loss": 0.4213,
      "reward": 0.55423304438591,
      "reward_std": 0.2226124033331871,
      "reward_std_all": 0.23631276935338974,
      "rewards-max/forensics_sinkhorn_reward": 0.9581505656242371,
      "rewards-max/forensics_strict_edge_reward": 0.878633052110672,
      "rewards-max/forensics_tiou_reward": 0.9291134476661682,
      "rewards/forensics_sinkhorn_reward": 0.4787517338991165,
      "rewards/forensics_strict_edge_reward": 0.2317412942647934,
      "rewards/forensics_tiou_reward": 0.30515363812446594,
      "step": 1380
    },
    {
      "completion_length": 16.0,
      "epoch": 3.673768308921438,
      "grad_norm": 24.8882999420166,
      "kl": 5.25,
      "learning_rate": 8e-08,
      "loss": 0.4178,
      "reward": 0.6134262382984161,
      "reward_std": 0.18376144021749496,
      "reward_std_all": 0.18353521823883057,
      "rewards-max/forensics_sinkhorn_reward": 0.9840126037597656,
      "rewards-max/forensics_strict_edge_reward": 0.9782643616199493,
      "rewards-max/forensics_tiou_reward": 0.988737165927887,
      "rewards/forensics_sinkhorn_reward": 0.601095587015152,
      "rewards/forensics_strict_edge_reward": 0.4018399566411972,
      "rewards/forensics_tiou_reward": 0.4579373300075531,
      "step": 1381
    },
    {
      "completion_length": 17.765625,
      "epoch": 3.676431424766977,
      "grad_norm": 30.14826774597168,
      "kl": 6.609375,
      "learning_rate": 7.933333333333333e-08,
      "loss": 0.4121,
      "reward": 0.5612289011478424,
      "reward_std": 0.18502119183540344,
      "reward_std_all": 0.20491589605808258,
      "rewards-max/forensics_sinkhorn_reward": 0.9628497362136841,
      "rewards-max/forensics_strict_edge_reward": 0.7607498168945312,
      "rewards-max/forensics_tiou_reward": 0.9428045749664307,
      "rewards/forensics_sinkhorn_reward": 0.4515502005815506,
      "rewards/forensics_strict_edge_reward": 0.14601455628871918,
      "rewards/forensics_tiou_reward": 0.24631114304065704,
      "step": 1382
    },
    {
      "completion_length": 16.3125,
      "epoch": 3.6790945406125166,
      "grad_norm": 19.459373474121094,
      "kl": 4.3671875,
      "learning_rate": 7.866666666666666e-08,
      "loss": 0.3734,
      "reward": 0.5905163884162903,
      "reward_std": 0.1671736016869545,
      "reward_std_all": 0.1906328722834587,
      "rewards-max/forensics_sinkhorn_reward": 0.9881935715675354,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9968591034412384,
      "rewards/forensics_sinkhorn_reward": 0.5770705044269562,
      "rewards/forensics_strict_edge_reward": 0.35842277109622955,
      "rewards/forensics_tiou_reward": 0.43578851222991943,
      "step": 1383
    },
    {
      "completion_length": 18.25,
      "epoch": 3.681757656458056,
      "grad_norm": 24.517587661743164,
      "kl": 6.328125,
      "learning_rate": 7.8e-08,
      "loss": 0.3217,
      "reward": 0.6050854027271271,
      "reward_std": 0.13530953601002693,
      "reward_std_all": 0.15949678793549538,
      "rewards-max/forensics_sinkhorn_reward": 0.9762417674064636,
      "rewards-max/forensics_strict_edge_reward": 0.9185642302036285,
      "rewards-max/forensics_tiou_reward": 0.9722479283809662,
      "rewards/forensics_sinkhorn_reward": 0.5108679383993149,
      "rewards/forensics_strict_edge_reward": 0.23993222415447235,
      "rewards/forensics_tiou_reward": 0.30622710287570953,
      "step": 1384
    },
    {
      "completion_length": 16.484375,
      "epoch": 3.684420772303595,
      "grad_norm": 23.938928604125977,
      "kl": 3.52734375,
      "learning_rate": 7.733333333333334e-08,
      "loss": 0.2043,
      "reward": 0.658231258392334,
      "reward_std": 0.17707481235265732,
      "reward_std_all": 0.1695096641778946,
      "rewards-max/forensics_sinkhorn_reward": 0.9884685277938843,
      "rewards-max/forensics_strict_edge_reward": 0.9782643616199493,
      "rewards-max/forensics_tiou_reward": 0.9979445040225983,
      "rewards/forensics_sinkhorn_reward": 0.6609536707401276,
      "rewards/forensics_strict_edge_reward": 0.3950263112783432,
      "rewards/forensics_tiou_reward": 0.5189866423606873,
      "step": 1385
    },
    {
      "completion_length": 15.765625,
      "epoch": 3.6870838881491346,
      "grad_norm": 18.11216163635254,
      "kl": 3.2734375,
      "learning_rate": 7.666666666666665e-08,
      "loss": 0.2046,
      "reward": 0.6361178159713745,
      "reward_std": 0.15573106333613396,
      "reward_std_all": 0.16231726109981537,
      "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": 0.573716014623642,
      "rewards/forensics_strict_edge_reward": 0.3389825373888016,
      "rewards/forensics_tiou_reward": 0.3737543970346451,
      "step": 1386
    },
    {
      "completion_length": 17.4375,
      "epoch": 3.6897470039946736,
      "grad_norm": 37.393550872802734,
      "kl": 3.32421875,
      "learning_rate": 7.599999999999999e-08,
      "loss": 0.3363,
      "reward": 0.650114119052887,
      "reward_std": 0.15867186337709427,
      "reward_std_all": 0.15834639966487885,
      "rewards-max/forensics_sinkhorn_reward": 0.9842691719532013,
      "rewards-max/forensics_strict_edge_reward": 0.8745477199554443,
      "rewards-max/forensics_tiou_reward": 0.9889333248138428,
      "rewards/forensics_sinkhorn_reward": 0.5797200500965118,
      "rewards/forensics_strict_edge_reward": 0.25599654018878937,
      "rewards/forensics_tiou_reward": 0.3663175404071808,
      "step": 1387
    },
    {
      "completion_length": 15.671875,
      "epoch": 3.692410119840213,
      "grad_norm": 14.2670316696167,
      "kl": 4.3515625,
      "learning_rate": 7.533333333333333e-08,
      "loss": 0.2599,
      "reward": 0.6730118691921234,
      "reward_std": 0.1114714927971363,
      "reward_std_all": 0.12449208274483681,
      "rewards-max/forensics_sinkhorn_reward": 0.9883005917072296,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9972812533378601,
      "rewards/forensics_sinkhorn_reward": 0.5882362425327301,
      "rewards/forensics_strict_edge_reward": 0.292572021484375,
      "rewards/forensics_tiou_reward": 0.4050889015197754,
      "step": 1388
    },
    {
      "completion_length": 15.625,
      "epoch": 3.6950732356857525,
      "grad_norm": 27.194856643676758,
      "kl": 3.890625,
      "learning_rate": 7.466666666666667e-08,
      "loss": 0.2673,
      "reward": 0.6914778351783752,
      "reward_std": 0.12998751178383827,
      "reward_std_all": 0.13206861913204193,
      "rewards-max/forensics_sinkhorn_reward": 0.9879564344882965,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9959279298782349,
      "rewards/forensics_sinkhorn_reward": 0.6387481093406677,
      "rewards/forensics_strict_edge_reward": 0.3117234855890274,
      "rewards/forensics_tiou_reward": 0.4018549621105194,
      "step": 1389
    },
    {
      "completion_length": 19.34375,
      "epoch": 3.6977363515312915,
      "grad_norm": 28.135372161865234,
      "kl": 4.1875,
      "learning_rate": 7.399999999999999e-08,
      "loss": 0.3838,
      "reward": 0.6307209730148315,
      "reward_std": 0.21533722430467606,
      "reward_std_all": 0.19983045756816864,
      "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": 0.6121350228786469,
      "rewards/forensics_strict_edge_reward": 0.3489200323820114,
      "rewards/forensics_tiou_reward": 0.4426395297050476,
      "step": 1390
    },
    {
      "completion_length": 15.640625,
      "epoch": 3.700399467376831,
      "grad_norm": 21.198347091674805,
      "kl": 2.9296875,
      "learning_rate": 7.333333333333333e-08,
      "loss": 0.3691,
      "reward": 0.6358347535133362,
      "reward_std": 0.16956809908151627,
      "reward_std_all": 0.17212190479040146,
      "rewards-max/forensics_sinkhorn_reward": 0.9879744946956635,
      "rewards-max/forensics_strict_edge_reward": 0.946017861366272,
      "rewards-max/forensics_tiou_reward": 0.9959990084171295,
      "rewards/forensics_sinkhorn_reward": 0.5950186848640442,
      "rewards/forensics_strict_edge_reward": 0.3012758791446686,
      "rewards/forensics_tiou_reward": 0.38940583169460297,
      "step": 1391
    },
    {
      "completion_length": 16.28125,
      "epoch": 3.70306258322237,
      "grad_norm": 23.347454071044922,
      "kl": 4.859375,
      "learning_rate": 7.266666666666667e-08,
      "loss": 0.3407,
      "reward": 0.6065048277378082,
      "reward_std": 0.22994324564933777,
      "reward_std_all": 0.2216906100511551,
      "rewards-max/forensics_sinkhorn_reward": 0.9704878032207489,
      "rewards-max/forensics_strict_edge_reward": 0.8681221902370453,
      "rewards-max/forensics_tiou_reward": 0.9548273682594299,
      "rewards/forensics_sinkhorn_reward": 0.5623673796653748,
      "rewards/forensics_strict_edge_reward": 0.28631220757961273,
      "rewards/forensics_tiou_reward": 0.3996056914329529,
      "step": 1392
    },
    {
      "completion_length": 15.765625,
      "epoch": 3.7057256990679095,
      "grad_norm": 20.459749221801758,
      "kl": 3.4921875,
      "learning_rate": 7.2e-08,
      "loss": 0.3556,
      "reward": 0.6770750284194946,
      "reward_std": 0.08205299451947212,
      "reward_std_all": 0.10136303678154945,
      "rewards-max/forensics_sinkhorn_reward": 0.9879628121852875,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9959491491317749,
      "rewards/forensics_sinkhorn_reward": 0.6343422830104828,
      "rewards/forensics_strict_edge_reward": 0.33395107090473175,
      "rewards/forensics_tiou_reward": 0.42865458130836487,
      "step": 1393
    },
    {
      "completion_length": 17.484375,
      "epoch": 3.708388814913449,
      "grad_norm": 43.053775787353516,
      "kl": 5.9375,
      "learning_rate": 7.133333333333332e-08,
      "loss": 0.5515,
      "reward": 0.5604817271232605,
      "reward_std": 0.3037628084421158,
      "reward_std_all": 0.2833261042833328,
      "rewards-max/forensics_sinkhorn_reward": 0.9700173735618591,
      "rewards-max/forensics_strict_edge_reward": 0.8968285918235779,
      "rewards-max/forensics_tiou_reward": 0.9542600214481354,
      "rewards/forensics_sinkhorn_reward": 0.503786027431488,
      "rewards/forensics_strict_edge_reward": 0.294719398021698,
      "rewards/forensics_tiou_reward": 0.36924542486667633,
      "step": 1394
    },
    {
      "completion_length": 21.890625,
      "epoch": 3.711051930758988,
      "grad_norm": 22.337627410888672,
      "kl": 5.875,
      "learning_rate": 7.066666666666666e-08,
      "loss": 0.2878,
      "reward": 0.5740949809551239,
      "reward_std": 0.144317626953125,
      "reward_std_all": 0.19172941893339157,
      "rewards-max/forensics_sinkhorn_reward": 0.9226932227611542,
      "rewards-max/forensics_strict_edge_reward": 0.7768298983573914,
      "rewards-max/forensics_tiou_reward": 0.8568311333656311,
      "rewards/forensics_sinkhorn_reward": 0.474980965256691,
      "rewards/forensics_strict_edge_reward": 0.17562001198530197,
      "rewards/forensics_tiou_reward": 0.2599997967481613,
      "step": 1395
    },
    {
      "completion_length": 16.65625,
      "epoch": 3.7137150466045274,
      "grad_norm": 20.704185485839844,
      "kl": 6.03125,
      "learning_rate": 7e-08,
      "loss": 0.4227,
      "reward": 0.5740087330341339,
      "reward_std": 0.2002672404050827,
      "reward_std_all": 0.22642452269792557,
      "rewards-max/forensics_sinkhorn_reward": 0.9884500503540039,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9978696703910828,
      "rewards/forensics_sinkhorn_reward": 0.5410073101520538,
      "rewards/forensics_strict_edge_reward": 0.29610346257686615,
      "rewards/forensics_tiou_reward": 0.3827662765979767,
      "step": 1396
    },
    {
      "completion_length": 17.921875,
      "epoch": 3.7163781624500665,
      "grad_norm": 28.014299392700195,
      "kl": 6.171875,
      "learning_rate": 6.933333333333333e-08,
      "loss": 0.4361,
      "reward": 0.563235342502594,
      "reward_std": 0.20076143741607666,
      "reward_std_all": 0.23052978515625,
      "rewards-max/forensics_sinkhorn_reward": 0.9703181087970734,
      "rewards-max/forensics_strict_edge_reward": 0.8553276658058167,
      "rewards-max/forensics_tiou_reward": 0.9594551026821136,
      "rewards/forensics_sinkhorn_reward": 0.49922242760658264,
      "rewards/forensics_strict_edge_reward": 0.20727059245109558,
      "rewards/forensics_tiou_reward": 0.333858847618103,
      "step": 1397
    },
    {
      "completion_length": 18.9375,
      "epoch": 3.719041278295606,
      "grad_norm": 27.26203155517578,
      "kl": 6.28125,
      "learning_rate": 6.866666666666666e-08,
      "loss": 0.3216,
      "reward": 0.5781019926071167,
      "reward_std": 0.2343839555978775,
      "reward_std_all": 0.23242971301078796,
      "rewards-max/forensics_sinkhorn_reward": 0.9817340970039368,
      "rewards-max/forensics_strict_edge_reward": 0.9677534997463226,
      "rewards-max/forensics_tiou_reward": 0.9833589792251587,
      "rewards/forensics_sinkhorn_reward": 0.5125190615653992,
      "rewards/forensics_strict_edge_reward": 0.27599263191223145,
      "rewards/forensics_tiou_reward": 0.35784517228603363,
      "step": 1398
    },
    {
      "completion_length": 18.125,
      "epoch": 3.7217043941411454,
      "grad_norm": 17.80704689025879,
      "kl": 6.203125,
      "learning_rate": 6.8e-08,
      "loss": 0.4632,
      "reward": 0.5593351721763611,
      "reward_std": 0.19522513449192047,
      "reward_std_all": 0.21661126613616943,
      "rewards-max/forensics_sinkhorn_reward": 0.9866314232349396,
      "rewards-max/forensics_strict_edge_reward": 0.9464850425720215,
      "rewards-max/forensics_tiou_reward": 0.9907437860965729,
      "rewards/forensics_sinkhorn_reward": 0.5151698589324951,
      "rewards/forensics_strict_edge_reward": 0.3173559755086899,
      "rewards/forensics_tiou_reward": 0.37064971029758453,
      "step": 1399
    },
    {
      "completion_length": 20.890625,
      "epoch": 3.7243675099866844,
      "grad_norm": 17.078487396240234,
      "kl": 4.2421875,
      "learning_rate": 6.733333333333333e-08,
      "loss": 0.3897,
      "reward": 0.6306721866130829,
      "reward_std": 0.18002697825431824,
      "reward_std_all": 0.17749134078621864,
      "rewards-max/forensics_sinkhorn_reward": 0.987069845199585,
      "rewards-max/forensics_strict_edge_reward": 0.9464850425720215,
      "rewards-max/forensics_tiou_reward": 0.9924564957618713,
      "rewards/forensics_sinkhorn_reward": 0.5560996532440186,
      "rewards/forensics_strict_edge_reward": 0.269324854016304,
      "rewards/forensics_tiou_reward": 0.36032022535800934,
      "step": 1400
    },
    {
      "completion_length": 20.328125,
      "epoch": 3.7270306258322234,
      "grad_norm": 19.33246612548828,
      "kl": 5.65625,
      "learning_rate": 6.666666666666667e-08,
      "loss": 0.2607,
      "reward": 0.5564118921756744,
      "reward_std": 0.16193995624780655,
      "reward_std_all": 0.20080328732728958,
      "rewards-max/forensics_sinkhorn_reward": 0.9800664186477661,
      "rewards-max/forensics_strict_edge_reward": 0.9464850425720215,
      "rewards-max/forensics_tiou_reward": 0.9789452254772186,
      "rewards/forensics_sinkhorn_reward": 0.5084368288516998,
      "rewards/forensics_strict_edge_reward": 0.21340462565422058,
      "rewards/forensics_tiou_reward": 0.31682904064655304,
      "step": 1401
    },
    {
      "completion_length": 16.046875,
      "epoch": 3.729693741677763,
      "grad_norm": 16.266271591186523,
      "kl": 5.171875,
      "learning_rate": 6.6e-08,
      "loss": 0.3891,
      "reward": 0.6165761649608612,
      "reward_std": 0.14918629825115204,
      "reward_std_all": 0.167383074760437,
      "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": 0.6235641837120056,
      "rewards/forensics_strict_edge_reward": 0.40681856870651245,
      "rewards/forensics_tiou_reward": 0.5140595138072968,
      "step": 1402
    },
    {
      "completion_length": 16.21875,
      "epoch": 3.7323568575233024,
      "grad_norm": 29.894495010375977,
      "kl": 4.453125,
      "learning_rate": 6.533333333333332e-08,
      "loss": 0.2289,
      "reward": 0.6369777619838715,
      "reward_std": 0.17626383155584335,
      "reward_std_all": 0.17851896584033966,
      "rewards-max/forensics_sinkhorn_reward": 0.916644960641861,
      "rewards-max/forensics_strict_edge_reward": 0.8553276658058167,
      "rewards-max/forensics_tiou_reward": 0.8428114950656891,
      "rewards/forensics_sinkhorn_reward": 0.5656866133213043,
      "rewards/forensics_strict_edge_reward": 0.31357014924287796,
      "rewards/forensics_tiou_reward": 0.3517710268497467,
      "step": 1403
    },
    {
      "completion_length": 15.6875,
      "epoch": 3.7350199733688414,
      "grad_norm": 32.78577423095703,
      "kl": 4.375,
      "learning_rate": 6.466666666666666e-08,
      "loss": 0.2333,
      "reward": 0.6403475999832153,
      "reward_std": 0.17488409578800201,
      "reward_std_all": 0.17505106329917908,
      "rewards-max/forensics_sinkhorn_reward": 0.9882136285305023,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9969388842582703,
      "rewards/forensics_sinkhorn_reward": 0.6026012003421783,
      "rewards/forensics_strict_edge_reward": 0.3413840979337692,
      "rewards/forensics_tiou_reward": 0.4063257873058319,
      "step": 1404
    },
    {
      "completion_length": 18.03125,
      "epoch": 3.737683089214381,
      "grad_norm": 14.328511238098145,
      "kl": 7.890625,
      "learning_rate": 6.4e-08,
      "loss": 0.5826,
      "reward": 0.5476859509944916,
      "reward_std": 0.19479240477085114,
      "reward_std_all": 0.2309989556670189,
      "rewards-max/forensics_sinkhorn_reward": 0.9674592614173889,
      "rewards-max/forensics_strict_edge_reward": 0.8104448914527893,
      "rewards-max/forensics_tiou_reward": 0.9518276453018188,
      "rewards/forensics_sinkhorn_reward": 0.46267104148864746,
      "rewards/forensics_strict_edge_reward": 0.2097979411482811,
      "rewards/forensics_tiou_reward": 0.2727990224957466,
      "step": 1405
    },
    {
      "completion_length": 18.34375,
      "epoch": 3.74034620505992,
      "grad_norm": 29.70140266418457,
      "kl": 7.453125,
      "learning_rate": 6.333333333333333e-08,
      "loss": 0.4749,
      "reward": 0.5983746945858002,
      "reward_std": 0.20859641581773758,
      "reward_std_all": 0.20878823846578598,
      "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": 0.5925964117050171,
      "rewards/forensics_strict_edge_reward": 0.3450136184692383,
      "rewards/forensics_tiou_reward": 0.4357941746711731,
      "step": 1406
    },
    {
      "completion_length": 17.046875,
      "epoch": 3.7430093209054593,
      "grad_norm": 31.736845016479492,
      "kl": 5.734375,
      "learning_rate": 6.266666666666666e-08,
      "loss": 0.293,
      "reward": 0.5551836937665939,
      "reward_std": 0.1531115062534809,
      "reward_std_all": 0.19453084096312523,
      "rewards-max/forensics_sinkhorn_reward": 0.9874184131622314,
      "rewards-max/forensics_strict_edge_reward": 0.946017861366272,
      "rewards-max/forensics_tiou_reward": 0.9938039779663086,
      "rewards/forensics_sinkhorn_reward": 0.470965176820755,
      "rewards/forensics_strict_edge_reward": 0.2411935031414032,
      "rewards/forensics_tiou_reward": 0.3160425275564194,
      "step": 1407
    },
    {
      "completion_length": 16.453125,
      "epoch": 3.745672436750999,
      "grad_norm": 25.78127670288086,
      "kl": 5.40625,
      "learning_rate": 6.2e-08,
      "loss": 0.3472,
      "reward": 0.6560401022434235,
      "reward_std": 0.13408664613962173,
      "reward_std_all": 0.14049237966537476,
      "rewards-max/forensics_sinkhorn_reward": 0.9877142012119293,
      "rewards-max/forensics_strict_edge_reward": 0.9565287232398987,
      "rewards-max/forensics_tiou_reward": 0.9949711263179779,
      "rewards/forensics_sinkhorn_reward": 0.5991096496582031,
      "rewards/forensics_strict_edge_reward": 0.25697900354862213,
      "rewards/forensics_tiou_reward": 0.4080510884523392,
      "step": 1408
    },
    {
      "completion_length": 19.265625,
      "epoch": 3.748335552596538,
      "grad_norm": 19.257394790649414,
      "kl": 8.5625,
      "learning_rate": 6.133333333333333e-08,
      "loss": 0.6335,
      "reward": 0.5267146229743958,
      "reward_std": 0.20002217590808868,
      "reward_std_all": 0.23197387158870697,
      "rewards-max/forensics_sinkhorn_reward": 0.9847935736179352,
      "rewards-max/forensics_strict_edge_reward": 0.9464850425720215,
      "rewards-max/forensics_tiou_reward": 0.9901940822601318,
      "rewards/forensics_sinkhorn_reward": 0.4282333552837372,
      "rewards/forensics_strict_edge_reward": 0.2216692715883255,
      "rewards/forensics_tiou_reward": 0.2696254104375839,
      "step": 1409
    },
    {
      "completion_length": 15.578125,
      "epoch": 3.7509986684420773,
      "grad_norm": 33.04502868652344,
      "kl": 4.2265625,
      "learning_rate": 6.066666666666667e-08,
      "loss": 0.2365,
      "reward": 0.6524627804756165,
      "reward_std": 0.1438198685646057,
      "reward_std_all": 0.14991829544305801,
      "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": 0.6027663946151733,
      "rewards/forensics_strict_edge_reward": 0.2941676080226898,
      "rewards/forensics_tiou_reward": 0.35613858699798584,
      "step": 1410
    },
    {
      "completion_length": 15.453125,
      "epoch": 3.7536617842876163,
      "grad_norm": 23.93330955505371,
      "kl": 5.84375,
      "learning_rate": 6e-08,
      "loss": 0.322,
      "reward": 0.6370912194252014,
      "reward_std": 0.1327625997364521,
      "reward_std_all": 0.14429084211587906,
      "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": 0.6194011867046356,
      "rewards/forensics_strict_edge_reward": 0.3379906713962555,
      "rewards/forensics_tiou_reward": 0.43050773441791534,
      "step": 1411
    },
    {
      "completion_length": 16.40625,
      "epoch": 3.756324900133156,
      "grad_norm": 45.51211929321289,
      "kl": 6.0,
      "learning_rate": 5.9333333333333335e-08,
      "loss": 0.392,
      "reward": 0.5744911730289459,
      "reward_std": 0.17889371514320374,
      "reward_std_all": 0.2110246866941452,
      "rewards-max/forensics_sinkhorn_reward": 0.9762415587902069,
      "rewards-max/forensics_strict_edge_reward": 0.8668389916419983,
      "rewards-max/forensics_tiou_reward": 0.9680993258953094,
      "rewards/forensics_sinkhorn_reward": 0.5045031607151031,
      "rewards/forensics_strict_edge_reward": 0.24639156460762024,
      "rewards/forensics_tiou_reward": 0.31916725635528564,
      "step": 1412
    },
    {
      "completion_length": 17.671875,
      "epoch": 3.7589880159786953,
      "grad_norm": 21.560991287231445,
      "kl": 4.765625,
      "learning_rate": 5.866666666666666e-08,
      "loss": 0.2664,
      "reward": 0.6092088222503662,
      "reward_std": 0.16660013794898987,
      "reward_std_all": 0.1803557127714157,
      "rewards-max/forensics_sinkhorn_reward": 0.9878918528556824,
      "rewards-max/forensics_strict_edge_reward": 0.8719756007194519,
      "rewards-max/forensics_tiou_reward": 0.9956774115562439,
      "rewards/forensics_sinkhorn_reward": 0.5591227114200592,
      "rewards/forensics_strict_edge_reward": 0.2514638900756836,
      "rewards/forensics_tiou_reward": 0.3660321533679962,
      "step": 1413
    },
    {
      "completion_length": 16.28125,
      "epoch": 3.7616511318242343,
      "grad_norm": 16.81740951538086,
      "kl": 4.3203125,
      "learning_rate": 5.8e-08,
      "loss": 0.2273,
      "reward": 0.6251128315925598,
      "reward_std": 0.15443675220012665,
      "reward_std_all": 0.16451626271009445,
      "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": 0.5733422338962555,
      "rewards/forensics_strict_edge_reward": 0.3645597845315933,
      "rewards/forensics_tiou_reward": 0.42218656837940216,
      "step": 1414
    },
    {
      "completion_length": 18.421875,
      "epoch": 3.7643142476697737,
      "grad_norm": 23.26222801208496,
      "kl": 6.3125,
      "learning_rate": 5.733333333333333e-08,
      "loss": 0.5013,
      "reward": 0.5985215604305267,
      "reward_std": 0.17822690308094025,
      "reward_std_all": 0.1941232532262802,
      "rewards-max/forensics_sinkhorn_reward": 0.9074183404445648,
      "rewards-max/forensics_strict_edge_reward": 0.8805811703205109,
      "rewards-max/forensics_tiou_reward": 0.8327738344669342,
      "rewards/forensics_sinkhorn_reward": 0.5114231705665588,
      "rewards/forensics_strict_edge_reward": 0.30016110837459564,
      "rewards/forensics_tiou_reward": 0.3551380932331085,
      "step": 1415
    },
    {
      "completion_length": 15.53125,
      "epoch": 3.7669773635153128,
      "grad_norm": 16.967681884765625,
      "kl": 4.90625,
      "learning_rate": 5.666666666666666e-08,
      "loss": 0.3413,
      "reward": 0.5953090488910675,
      "reward_std": 0.18405013531446457,
      "reward_std_all": 0.19490180164575577,
      "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": 0.5643071532249451,
      "rewards/forensics_strict_edge_reward": 0.2670586407184601,
      "rewards/forensics_tiou_reward": 0.3808263838291168,
      "step": 1416
    },
    {
      "completion_length": 15.53125,
      "epoch": 3.7696404793608522,
      "grad_norm": 43.468631744384766,
      "kl": 8.71875,
      "learning_rate": 5.6e-08,
      "loss": 0.4834,
      "reward": 0.6662087440490723,
      "reward_std": 0.1476038657128811,
      "reward_std_all": 0.14212528616189957,
      "rewards-max/forensics_sinkhorn_reward": 0.9819359183311462,
      "rewards-max/forensics_strict_edge_reward": 0.9782643616199493,
      "rewards-max/forensics_tiou_reward": 0.9845225214958191,
      "rewards/forensics_sinkhorn_reward": 0.6429603099822998,
      "rewards/forensics_strict_edge_reward": 0.30648114532232285,
      "rewards/forensics_tiou_reward": 0.4484565854072571,
      "step": 1417
    },
    {
      "completion_length": 18.0625,
      "epoch": 3.7723035952063917,
      "grad_norm": 26.115039825439453,
      "kl": 3.75,
      "learning_rate": 5.5333333333333326e-08,
      "loss": 0.2175,
      "reward": 0.6206773519515991,
      "reward_std": 0.116254523396492,
      "reward_std_all": 0.15198566019535065,
      "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": 0.5405783653259277,
      "rewards/forensics_strict_edge_reward": 0.2226932942867279,
      "rewards/forensics_tiou_reward": 0.2805878669023514,
      "step": 1418
    },
    {
      "completion_length": 16.921875,
      "epoch": 3.7749667110519307,
      "grad_norm": 24.243967056274414,
      "kl": 4.078125,
      "learning_rate": 5.4666666666666666e-08,
      "loss": 0.2903,
      "reward": 0.6186927258968353,
      "reward_std": 0.21342945843935013,
      "reward_std_all": 0.2044132947921753,
      "rewards-max/forensics_sinkhorn_reward": 0.9858266115188599,
      "rewards-max/forensics_strict_edge_reward": 0.9357379674911499,
      "rewards-max/forensics_tiou_reward": 0.9875307679176331,
      "rewards/forensics_sinkhorn_reward": 0.5275156199932098,
      "rewards/forensics_strict_edge_reward": 0.19154629856348038,
      "rewards/forensics_tiou_reward": 0.2764110416173935,
      "step": 1419
    },
    {
      "completion_length": 17.9375,
      "epoch": 3.77762982689747,
      "grad_norm": 17.891834259033203,
      "kl": 6.21875,
      "learning_rate": 5.3999999999999994e-08,
      "loss": 0.4623,
      "reward": 0.49971799552440643,
      "reward_std": 0.24702215194702148,
      "reward_std_all": 0.27262531220912933,
      "rewards-max/forensics_sinkhorn_reward": 0.9646198749542236,
      "rewards-max/forensics_strict_edge_reward": 0.8915697336196899,
      "rewards-max/forensics_tiou_reward": 0.9434616565704346,
      "rewards/forensics_sinkhorn_reward": 0.41870227456092834,
      "rewards/forensics_strict_edge_reward": 0.2417389675974846,
      "rewards/forensics_tiou_reward": 0.31179333478212357,
      "step": 1420
    },
    {
      "completion_length": 16.953125,
      "epoch": 3.780292942743009,
      "grad_norm": 32.92549514770508,
      "kl": 8.25,
      "learning_rate": 5.3333333333333334e-08,
      "loss": 0.5291,
      "reward": 0.5294244289398193,
      "reward_std": 0.25354307144880295,
      "reward_std_all": 0.266021266579628,
      "rewards-max/forensics_sinkhorn_reward": 0.9826129078865051,
      "rewards-max/forensics_strict_edge_reward": 0.905340164899826,
      "rewards-max/forensics_tiou_reward": 0.9839867055416107,
      "rewards/forensics_sinkhorn_reward": 0.46779631078243256,
      "rewards/forensics_strict_edge_reward": 0.20723392069339752,
      "rewards/forensics_tiou_reward": 0.2815539538860321,
      "step": 1421
    },
    {
      "completion_length": 18.4375,
      "epoch": 3.7829560585885487,
      "grad_norm": 20.935781478881836,
      "kl": 6.578125,
      "learning_rate": 5.266666666666666e-08,
      "loss": 0.4434,
      "reward": 0.527739092707634,
      "reward_std": 0.2578706219792366,
      "reward_std_all": 0.2706267237663269,
      "rewards-max/forensics_sinkhorn_reward": 0.9420217871665955,
      "rewards-max/forensics_strict_edge_reward": 0.8205901980400085,
      "rewards-max/forensics_tiou_reward": 0.8970248401165009,
      "rewards/forensics_sinkhorn_reward": 0.42615124583244324,
      "rewards/forensics_strict_edge_reward": 0.2039644569158554,
      "rewards/forensics_tiou_reward": 0.25735362619161606,
      "step": 1422
    },
    {
      "completion_length": 15.34375,
      "epoch": 3.785619174434088,
      "grad_norm": 23.291505813598633,
      "kl": 4.78125,
      "learning_rate": 5.1999999999999996e-08,
      "loss": 0.2497,
      "reward": 0.6222732961177826,
      "reward_std": 0.2238323837518692,
      "reward_std_all": 0.21374057978391647,
      "rewards-max/forensics_sinkhorn_reward": 0.9878039658069611,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9953241944313049,
      "rewards/forensics_sinkhorn_reward": 0.5781637132167816,
      "rewards/forensics_strict_edge_reward": 0.25415314733982086,
      "rewards/forensics_tiou_reward": 0.3788304924964905,
      "step": 1423
    },
    {
      "completion_length": 16.703125,
      "epoch": 3.788282290279627,
      "grad_norm": 16.18073844909668,
      "kl": 4.984375,
      "learning_rate": 5.133333333333333e-08,
      "loss": 0.3808,
      "reward": 0.6116417050361633,
      "reward_std": 0.2292700856924057,
      "reward_std_all": 0.21602730453014374,
      "rewards-max/forensics_sinkhorn_reward": 0.9880759418010712,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9963962137699127,
      "rewards/forensics_sinkhorn_reward": 0.587756797671318,
      "rewards/forensics_strict_edge_reward": 0.35325001180171967,
      "rewards/forensics_tiou_reward": 0.43688875436782837,
      "step": 1424
    },
    {
      "completion_length": 17.46875,
      "epoch": 3.790945406125166,
      "grad_norm": 19.57366943359375,
      "kl": 5.265625,
      "learning_rate": 5.0666666666666664e-08,
      "loss": 0.3955,
      "reward": 0.595108836889267,
      "reward_std": 0.11679337173700333,
      "reward_std_all": 0.16148684173822403,
      "rewards-max/forensics_sinkhorn_reward": 0.9865851700305939,
      "rewards-max/forensics_strict_edge_reward": 0.9782643616199493,
      "rewards-max/forensics_tiou_reward": 0.9944144189357758,
      "rewards/forensics_sinkhorn_reward": 0.49198848009109497,
      "rewards/forensics_strict_edge_reward": 0.24848081171512604,
      "rewards/forensics_tiou_reward": 0.29443617165088654,
      "step": 1425
    },
    {
      "completion_length": 16.75,
      "epoch": 3.7936085219707056,
      "grad_norm": 23.70380210876465,
      "kl": 4.140625,
      "learning_rate": 5e-08,
      "loss": 0.2183,
      "reward": 0.6125172972679138,
      "reward_std": 0.20366913825273514,
      "reward_std_all": 0.19982010126113892,
      "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": 0.5674935281276703,
      "rewards/forensics_strict_edge_reward": 0.2895151823759079,
      "rewards/forensics_tiou_reward": 0.40919849276542664,
      "step": 1426
    },
    {
      "completion_length": 18.765625,
      "epoch": 3.796271637816245,
      "grad_norm": 24.245817184448242,
      "kl": 5.46875,
      "learning_rate": 4.933333333333333e-08,
      "loss": 0.3971,
      "reward": 0.571793794631958,
      "reward_std": 0.2131091132760048,
      "reward_std_all": 0.2224409505724907,
      "rewards-max/forensics_sinkhorn_reward": 0.9875989556312561,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9945243000984192,
      "rewards/forensics_sinkhorn_reward": 0.49084535241127014,
      "rewards/forensics_strict_edge_reward": 0.24235548824071884,
      "rewards/forensics_tiou_reward": 0.302901029586792,
      "step": 1427
    },
    {
      "completion_length": 17.34375,
      "epoch": 3.798934753661784,
      "grad_norm": 16.569494247436523,
      "kl": 4.6328125,
      "learning_rate": 4.866666666666666e-08,
      "loss": 0.3587,
      "reward": 0.540743499994278,
      "reward_std": 0.1831030175089836,
      "reward_std_all": 0.20908305048942566,
      "rewards-max/forensics_sinkhorn_reward": 0.8788491785526276,
      "rewards-max/forensics_strict_edge_reward": 0.7790428102016449,
      "rewards-max/forensics_tiou_reward": 0.7727727591991425,
      "rewards/forensics_sinkhorn_reward": 0.44392921030521393,
      "rewards/forensics_strict_edge_reward": 0.23383206129074097,
      "rewards/forensics_tiou_reward": 0.28044160455465317,
      "step": 1428
    },
    {
      "completion_length": 16.84375,
      "epoch": 3.8015978695073236,
      "grad_norm": 26.872848510742188,
      "kl": 7.140625,
      "learning_rate": 4.8e-08,
      "loss": 0.5681,
      "reward": 0.5536260306835175,
      "reward_std": 0.14655854180455208,
      "reward_std_all": 0.2027256265282631,
      "rewards-max/forensics_sinkhorn_reward": 0.9885909557342529,
      "rewards-max/forensics_strict_edge_reward": 0.9565287232398987,
      "rewards-max/forensics_tiou_reward": 0.9984255135059357,
      "rewards/forensics_sinkhorn_reward": 0.49640586972236633,
      "rewards/forensics_strict_edge_reward": 0.2929431423544884,
      "rewards/forensics_tiou_reward": 0.3821537345647812,
      "step": 1429
    },
    {
      "completion_length": 19.125,
      "epoch": 3.8042609853528626,
      "grad_norm": 27.029151916503906,
      "kl": 5.0546875,
      "learning_rate": 4.733333333333333e-08,
      "loss": 0.2415,
      "reward": 0.5421715974807739,
      "reward_std": 0.18675320595502853,
      "reward_std_all": 0.22291965782642365,
      "rewards-max/forensics_sinkhorn_reward": 0.8967550694942474,
      "rewards-max/forensics_strict_edge_reward": 0.783246785402298,
      "rewards-max/forensics_tiou_reward": 0.8088340759277344,
      "rewards/forensics_sinkhorn_reward": 0.40330246090888977,
      "rewards/forensics_strict_edge_reward": 0.14740955084562302,
      "rewards/forensics_tiou_reward": 0.18550366908311844,
      "step": 1430
    },
    {
      "completion_length": 15.25,
      "epoch": 3.806924101198402,
      "grad_norm": 37.37334442138672,
      "kl": 4.609375,
      "learning_rate": 4.666666666666667e-08,
      "loss": 0.1657,
      "reward": 0.6812737584114075,
      "reward_std": 0.09919316321611404,
      "reward_std_all": 0.11094451695680618,
      "rewards-max/forensics_sinkhorn_reward": 0.9616763889789581,
      "rewards-max/forensics_strict_edge_reward": 0.9093653857707977,
      "rewards-max/forensics_tiou_reward": 0.9384812414646149,
      "rewards/forensics_sinkhorn_reward": 0.6335307955741882,
      "rewards/forensics_strict_edge_reward": 0.32782283425331116,
      "rewards/forensics_tiou_reward": 0.42063264548778534,
      "step": 1431
    },
    {
      "completion_length": 16.15625,
      "epoch": 3.8095872170439415,
      "grad_norm": 22.646114349365234,
      "kl": 6.0625,
      "learning_rate": 4.5999999999999995e-08,
      "loss": 0.4589,
      "reward": 0.5792442262172699,
      "reward_std": 0.220509834587574,
      "reward_std_all": 0.22059470415115356,
      "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": 0.5197022259235382,
      "rewards/forensics_strict_edge_reward": 0.2732735127210617,
      "rewards/forensics_tiou_reward": 0.33768168091773987,
      "step": 1432
    },
    {
      "completion_length": 16.0625,
      "epoch": 3.8122503328894806,
      "grad_norm": 41.759159088134766,
      "kl": 5.5546875,
      "learning_rate": 4.5333333333333336e-08,
      "loss": 0.4404,
      "reward": 0.6355550289154053,
      "reward_std": 0.1351892240345478,
      "reward_std_all": 0.15429198369383812,
      "rewards-max/forensics_sinkhorn_reward": 0.987949401140213,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9958972632884979,
      "rewards/forensics_sinkhorn_reward": 0.5650106072425842,
      "rewards/forensics_strict_edge_reward": 0.2640148997306824,
      "rewards/forensics_tiou_reward": 0.32055698335170746,
      "step": 1433
    },
    {
      "completion_length": 16.90625,
      "epoch": 3.81491344873502,
      "grad_norm": 26.041976928710938,
      "kl": 7.8125,
      "learning_rate": 4.466666666666666e-08,
      "loss": 0.3909,
      "reward": 0.5757961869239807,
      "reward_std": 0.16861113160848618,
      "reward_std_all": 0.20018847286701202,
      "rewards-max/forensics_sinkhorn_reward": 0.9824369251728058,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.984756737947464,
      "rewards/forensics_sinkhorn_reward": 0.5200716257095337,
      "rewards/forensics_strict_edge_reward": 0.29652640223503113,
      "rewards/forensics_tiou_reward": 0.3521603047847748,
      "step": 1434
    },
    {
      "completion_length": 18.5625,
      "epoch": 3.817576564580559,
      "grad_norm": 16.3914737701416,
      "kl": 5.8125,
      "learning_rate": 4.4e-08,
      "loss": 0.3432,
      "reward": 0.5445276200771332,
      "reward_std": 0.19750377535820007,
      "reward_std_all": 0.2227071076631546,
      "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": 0.5230383723974228,
      "rewards/forensics_strict_edge_reward": 0.3225739747285843,
      "rewards/forensics_tiou_reward": 0.38510552048683167,
      "step": 1435
    },
    {
      "completion_length": 14.6875,
      "epoch": 3.8202396804260985,
      "grad_norm": 34.682594299316406,
      "kl": 3.7421875,
      "learning_rate": 4.333333333333333e-08,
      "loss": 0.3482,
      "reward": 0.7093550264835358,
      "reward_std": 0.1333811953663826,
      "reward_std_all": 0.1213541105389595,
      "rewards-max/forensics_sinkhorn_reward": 0.988188773393631,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.996839702129364,
      "rewards/forensics_sinkhorn_reward": 0.7340298593044281,
      "rewards/forensics_strict_edge_reward": 0.45075005292892456,
      "rewards/forensics_tiou_reward": 0.5904676020145416,
      "step": 1436
    },
    {
      "completion_length": 16.828125,
      "epoch": 3.822902796271638,
      "grad_norm": 25.2484188079834,
      "kl": 4.453125,
      "learning_rate": 4.2666666666666665e-08,
      "loss": 0.376,
      "reward": 0.6267991364002228,
      "reward_std": 0.20542402565479279,
      "reward_std_all": 0.20072170346975327,
      "rewards-max/forensics_sinkhorn_reward": 0.9784559607505798,
      "rewards-max/forensics_strict_edge_reward": 0.9169811904430389,
      "rewards-max/forensics_tiou_reward": 0.9748593866825104,
      "rewards/forensics_sinkhorn_reward": 0.5608888566493988,
      "rewards/forensics_strict_edge_reward": 0.29393937438726425,
      "rewards/forensics_tiou_reward": 0.3966413289308548,
      "step": 1437
    },
    {
      "completion_length": 16.296875,
      "epoch": 3.825565912117177,
      "grad_norm": 22.558467864990234,
      "kl": 4.265625,
      "learning_rate": 4.2e-08,
      "loss": 0.3237,
      "reward": 0.6865726113319397,
      "reward_std": 0.15482497215270996,
      "reward_std_all": 0.14585530012845993,
      "rewards-max/forensics_sinkhorn_reward": 0.9880908727645874,
      "rewards-max/forensics_strict_edge_reward": 0.9782643616199493,
      "rewards-max/forensics_tiou_reward": 0.9964601993560791,
      "rewards/forensics_sinkhorn_reward": 0.6861090362071991,
      "rewards/forensics_strict_edge_reward": 0.47039277851581573,
      "rewards/forensics_tiou_reward": 0.5655648410320282,
      "step": 1438
    },
    {
      "completion_length": 17.25,
      "epoch": 3.8282290279627165,
      "grad_norm": 24.386253356933594,
      "kl": 4.203125,
      "learning_rate": 4.133333333333333e-08,
      "loss": 0.296,
      "reward": 0.635200172662735,
      "reward_std": 0.21929819136857986,
      "reward_std_all": 0.20090515166521072,
      "rewards-max/forensics_sinkhorn_reward": 0.9887720048427582,
      "rewards-max/forensics_strict_edge_reward": 0.9890114367008209,
      "rewards-max/forensics_tiou_reward": 0.9991390407085419,
      "rewards/forensics_sinkhorn_reward": 0.5840552449226379,
      "rewards/forensics_strict_edge_reward": 0.3716532737016678,
      "rewards/forensics_tiou_reward": 0.46378201246261597,
      "step": 1439
    },
    {
      "completion_length": 18.15625,
      "epoch": 3.8308921438082555,
      "grad_norm": 20.02346420288086,
      "kl": 6.65625,
      "learning_rate": 4.066666666666666e-08,
      "loss": 0.5016,
      "reward": 0.5931431949138641,
      "reward_std": 0.24303187429904938,
      "reward_std_all": 0.22983253747224808,
      "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": 0.5356476157903671,
      "rewards/forensics_strict_edge_reward": 0.3165188431739807,
      "rewards/forensics_tiou_reward": 0.39982110261917114,
      "step": 1440
    },
    {
      "completion_length": 16.234375,
      "epoch": 3.833555259653795,
      "grad_norm": 19.521753311157227,
      "kl": 4.71875,
      "learning_rate": 4e-08,
      "loss": 0.3483,
      "reward": 0.5353298038244247,
      "reward_std": 0.20977284014225006,
      "reward_std_all": 0.23997588455677032,
      "rewards-max/forensics_sinkhorn_reward": 0.9884760975837708,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9979734718799591,
      "rewards/forensics_sinkhorn_reward": 0.49191413819789886,
      "rewards/forensics_strict_edge_reward": 0.2584572806954384,
      "rewards/forensics_tiou_reward": 0.37551017105579376,
      "step": 1441
    },
    {
      "completion_length": 18.578125,
      "epoch": 3.8362183754993344,
      "grad_norm": 23.498512268066406,
      "kl": 6.78125,
      "learning_rate": 3.933333333333333e-08,
      "loss": 0.435,
      "reward": 0.47147709131240845,
      "reward_std": 0.2667696103453636,
      "reward_std_all": 0.2936938852071762,
      "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": 0.38281087577342987,
      "rewards/forensics_strict_edge_reward": 0.18893320113420486,
      "rewards/forensics_tiou_reward": 0.22554119676351547,
      "step": 1442
    },
    {
      "completion_length": 19.625,
      "epoch": 3.8388814913448734,
      "grad_norm": 24.250240325927734,
      "kl": 6.15625,
      "learning_rate": 3.866666666666667e-08,
      "loss": 0.4295,
      "reward": 0.4950515478849411,
      "reward_std": 0.23572711646556854,
      "reward_std_all": 0.26889193058013916,
      "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": 0.4587358087301254,
      "rewards/forensics_strict_edge_reward": 0.30063828080892563,
      "rewards/forensics_tiou_reward": 0.35525764524936676,
      "step": 1443
    },
    {
      "completion_length": 15.34375,
      "epoch": 3.841544607190413,
      "grad_norm": 18.372774124145508,
      "kl": 5.015625,
      "learning_rate": 3.7999999999999996e-08,
      "loss": 0.3839,
      "reward": 0.587264895439148,
      "reward_std": 0.18369802087545395,
      "reward_std_all": 0.20560133457183838,
      "rewards-max/forensics_sinkhorn_reward": 0.9886156022548676,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9985236823558807,
      "rewards/forensics_sinkhorn_reward": 0.5454622209072113,
      "rewards/forensics_strict_edge_reward": 0.2606535702943802,
      "rewards/forensics_tiou_reward": 0.3733253628015518,
      "step": 1444
    },
    {
      "completion_length": 17.84375,
      "epoch": 3.844207723035952,
      "grad_norm": 24.544811248779297,
      "kl": 6.703125,
      "learning_rate": 3.733333333333334e-08,
      "loss": 0.653,
      "reward": 0.4724861979484558,
      "reward_std": 0.25020118802785873,
      "reward_std_all": 0.2824118584394455,
      "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": 0.40939170122146606,
      "rewards/forensics_strict_edge_reward": 0.17506089061498642,
      "rewards/forensics_tiou_reward": 0.2757290452718735,
      "step": 1445
    },
    {
      "completion_length": 16.515625,
      "epoch": 3.8468708388814914,
      "grad_norm": 32.46628952026367,
      "kl": 5.875,
      "learning_rate": 3.6666666666666664e-08,
      "loss": 0.384,
      "reward": 0.6121791303157806,
      "reward_std": 0.23653147369623184,
      "reward_std_all": 0.2206258624792099,
      "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": 0.5708833932876587,
      "rewards/forensics_strict_edge_reward": 0.3446069657802582,
      "rewards/forensics_tiou_reward": 0.4554212689399719,
      "step": 1446
    },
    {
      "completion_length": 15.65625,
      "epoch": 3.849533954727031,
      "grad_norm": 22.750783920288086,
      "kl": 6.6171875,
      "learning_rate": 3.6e-08,
      "loss": 0.4531,
      "reward": 0.6096031367778778,
      "reward_std": 0.2057114839553833,
      "reward_std_all": 0.1973080039024353,
      "rewards-max/forensics_sinkhorn_reward": 0.9864572286605835,
      "rewards-max/forensics_strict_edge_reward": 0.9364311099052429,
      "rewards-max/forensics_tiou_reward": 0.9900169372558594,
      "rewards/forensics_sinkhorn_reward": 0.6061467230319977,
      "rewards/forensics_strict_edge_reward": 0.36114636063575745,
      "rewards/forensics_tiou_reward": 0.47178518772125244,
      "step": 1447
    },
    {
      "completion_length": 16.765625,
      "epoch": 3.85219707057257,
      "grad_norm": 56.40885543823242,
      "kl": 10.265625,
      "learning_rate": 3.533333333333333e-08,
      "loss": 0.6593,
      "reward": 0.5397012233734131,
      "reward_std": 0.25735047459602356,
      "reward_std_all": 0.2565690279006958,
      "rewards-max/forensics_sinkhorn_reward": 0.9881677031517029,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.996760368347168,
      "rewards/forensics_sinkhorn_reward": 0.45763295888900757,
      "rewards/forensics_strict_edge_reward": 0.22117038816213608,
      "rewards/forensics_tiou_reward": 0.31081536412239075,
      "step": 1448
    },
    {
      "completion_length": 14.65625,
      "epoch": 3.8548601864181093,
      "grad_norm": 22.938274383544922,
      "kl": 5.046875,
      "learning_rate": 3.4666666666666666e-08,
      "loss": 0.3471,
      "reward": 0.6263564825057983,
      "reward_std": 0.17132353782653809,
      "reward_std_all": 0.17049439996480942,
      "rewards-max/forensics_sinkhorn_reward": 0.9888220727443695,
      "rewards-max/forensics_strict_edge_reward": 0.9890114367008209,
      "rewards-max/forensics_tiou_reward": 0.9993362128734589,
      "rewards/forensics_sinkhorn_reward": 0.6118191778659821,
      "rewards/forensics_strict_edge_reward": 0.3179868310689926,
      "rewards/forensics_tiou_reward": 0.3927464932203293,
      "step": 1449
    },
    {
      "completion_length": 17.5,
      "epoch": 3.8575233022636484,
      "grad_norm": 31.569171905517578,
      "kl": 10.328125,
      "learning_rate": 3.4e-08,
      "loss": 0.7401,
      "reward": 0.4600286930799484,
      "reward_std": 0.2290618121623993,
      "reward_std_all": 0.28729045391082764,
      "rewards-max/forensics_sinkhorn_reward": 0.9876099824905396,
      "rewards-max/forensics_strict_edge_reward": 0.9565287232398987,
      "rewards-max/forensics_tiou_reward": 0.9945634603500366,
      "rewards/forensics_sinkhorn_reward": 0.41851066052913666,
      "rewards/forensics_strict_edge_reward": 0.31049781292676926,
      "rewards/forensics_tiou_reward": 0.3376648351550102,
      "step": 1450
    },
    {
      "completion_length": 16.828125,
      "epoch": 3.860186418109188,
      "grad_norm": 23.096500396728516,
      "kl": 6.125,
      "learning_rate": 3.3333333333333334e-08,
      "loss": 0.4603,
      "reward": 0.5492876172065735,
      "reward_std": 0.23958713561296463,
      "reward_std_all": 0.2517513632774353,
      "rewards-max/forensics_sinkhorn_reward": 0.9873640239238739,
      "rewards-max/forensics_strict_edge_reward": 0.9574736058712006,
      "rewards-max/forensics_tiou_reward": 0.993610292673111,
      "rewards/forensics_sinkhorn_reward": 0.49788305163383484,
      "rewards/forensics_strict_edge_reward": 0.24260485172271729,
      "rewards/forensics_tiou_reward": 0.3328928202390671,
      "step": 1451
    },
    {
      "completion_length": 16.921875,
      "epoch": 3.8628495339547273,
      "grad_norm": 18.28312873840332,
      "kl": 6.171875,
      "learning_rate": 3.266666666666666e-08,
      "loss": 0.4526,
      "reward": 0.5957518219947815,
      "reward_std": 0.21991613507270813,
      "reward_std_all": 0.21962504088878632,
      "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": 0.5556739866733551,
      "rewards/forensics_strict_edge_reward": 0.36357173323631287,
      "rewards/forensics_tiou_reward": 0.43056249618530273,
      "step": 1452
    },
    {
      "completion_length": 16.375,
      "epoch": 3.8655126498002663,
      "grad_norm": 24.189096450805664,
      "kl": 7.40625,
      "learning_rate": 3.2e-08,
      "loss": 0.5511,
      "reward": 0.5560029447078705,
      "reward_std": 0.1882450133562088,
      "reward_std_all": 0.21074645221233368,
      "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": 0.5098024010658264,
      "rewards/forensics_strict_edge_reward": 0.2932004779577255,
      "rewards/forensics_tiou_reward": 0.3563365042209625,
      "step": 1453
    },
    {
      "completion_length": 19.3125,
      "epoch": 3.8681757656458053,
      "grad_norm": 17.12511444091797,
      "kl": 4.828125,
      "learning_rate": 3.133333333333333e-08,
      "loss": 0.3164,
      "reward": 0.6101788580417633,
      "reward_std": 0.21580547839403152,
      "reward_std_all": 0.2099262699484825,
      "rewards-max/forensics_sinkhorn_reward": 0.9882899522781372,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9972399473190308,
      "rewards/forensics_sinkhorn_reward": 0.5591908097267151,
      "rewards/forensics_strict_edge_reward": 0.30980390310287476,
      "rewards/forensics_tiou_reward": 0.4060560315847397,
      "step": 1454
    },
    {
      "completion_length": 19.4375,
      "epoch": 3.870838881491345,
      "grad_norm": 22.78488540649414,
      "kl": 6.328125,
      "learning_rate": 3.0666666666666663e-08,
      "loss": 0.404,
      "reward": 0.5393407344818115,
      "reward_std": 0.16612878441810608,
      "reward_std_all": 0.21483652293682098,
      "rewards-max/forensics_sinkhorn_reward": 0.9555778503417969,
      "rewards-max/forensics_strict_edge_reward": 0.8096016347408295,
      "rewards-max/forensics_tiou_reward": 0.9232383668422699,
      "rewards/forensics_sinkhorn_reward": 0.4657919108867645,
      "rewards/forensics_strict_edge_reward": 0.23919397592544556,
      "rewards/forensics_tiou_reward": 0.2988215759396553,
      "step": 1455
    },
    {
      "completion_length": 15.5625,
      "epoch": 3.8735019973368843,
      "grad_norm": 30.554197311401367,
      "kl": 4.515625,
      "learning_rate": 3e-08,
      "loss": 0.3777,
      "reward": 0.6707192361354828,
      "reward_std": 0.16967764124274254,
      "reward_std_all": 0.15912960469722748,
      "rewards-max/forensics_sinkhorn_reward": 0.9879263937473297,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9958052337169647,
      "rewards/forensics_sinkhorn_reward": 0.6471492648124695,
      "rewards/forensics_strict_edge_reward": 0.3788750320672989,
      "rewards/forensics_tiou_reward": 0.4764532148838043,
      "step": 1456
    },
    {
      "completion_length": 15.71875,
      "epoch": 3.8761651131824233,
      "grad_norm": 27.619356155395508,
      "kl": 5.109375,
      "learning_rate": 2.933333333333333e-08,
      "loss": 0.4007,
      "reward": 0.5645313262939453,
      "reward_std": 0.2028460130095482,
      "reward_std_all": 0.22063319385051727,
      "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": 0.514639139175415,
      "rewards/forensics_strict_edge_reward": 0.2587391585111618,
      "rewards/forensics_tiou_reward": 0.3344593942165375,
      "step": 1457
    },
    {
      "completion_length": 17.703125,
      "epoch": 3.8788282290279628,
      "grad_norm": 22.99434471130371,
      "kl": 4.0546875,
      "learning_rate": 2.8666666666666665e-08,
      "loss": 0.2562,
      "reward": 0.6722424626350403,
      "reward_std": 0.09951411932706833,
      "reward_std_all": 0.11768221110105515,
      "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": 0.6344869434833527,
      "rewards/forensics_strict_edge_reward": 0.2947767376899719,
      "rewards/forensics_tiou_reward": 0.38078172504901886,
      "step": 1458
    },
    {
      "completion_length": 17.125,
      "epoch": 3.881491344873502,
      "grad_norm": 22.096229553222656,
      "kl": 5.3671875,
      "learning_rate": 2.8e-08,
      "loss": 0.4917,
      "reward": 0.6013890504837036,
      "reward_std": 0.18225930631160736,
      "reward_std_all": 0.19580981880426407,
      "rewards-max/forensics_sinkhorn_reward": 0.9864528179168701,
      "rewards-max/forensics_strict_edge_reward": 0.9375866651535034,
      "rewards-max/forensics_tiou_reward": 0.9900000095367432,
      "rewards/forensics_sinkhorn_reward": 0.5187939405441284,
      "rewards/forensics_strict_edge_reward": 0.2896885871887207,
      "rewards/forensics_tiou_reward": 0.36329932510852814,
      "step": 1459
    },
    {
      "completion_length": 16.390625,
      "epoch": 3.8841544607190412,
      "grad_norm": 22.228939056396484,
      "kl": 7.03125,
      "learning_rate": 2.7333333333333333e-08,
      "loss": 0.6087,
      "reward": 0.503407746553421,
      "reward_std": 0.24361690133810043,
      "reward_std_all": 0.2753439396619797,
      "rewards-max/forensics_sinkhorn_reward": 0.9573951959609985,
      "rewards-max/forensics_strict_edge_reward": 0.8698340952396393,
      "rewards-max/forensics_tiou_reward": 0.9258435666561127,
      "rewards/forensics_sinkhorn_reward": 0.45409145951271057,
      "rewards/forensics_strict_edge_reward": 0.2688589245080948,
      "rewards/forensics_tiou_reward": 0.3041663393378258,
      "step": 1460
    },
    {
      "completion_length": 16.078125,
      "epoch": 3.8868175765645807,
      "grad_norm": 20.70585060119629,
      "kl": 4.8828125,
      "learning_rate": 2.6666666666666667e-08,
      "loss": 0.3899,
      "reward": 0.5603727102279663,
      "reward_std": 0.28334905207157135,
      "reward_std_all": 0.2702649235725403,
      "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": 0.5488800406455994,
      "rewards/forensics_strict_edge_reward": 0.35487715899944305,
      "rewards/forensics_tiou_reward": 0.4230075031518936,
      "step": 1461
    },
    {
      "completion_length": 17.484375,
      "epoch": 3.8894806924101197,
      "grad_norm": 22.818174362182617,
      "kl": 5.671875,
      "learning_rate": 2.5999999999999998e-08,
      "loss": 0.4714,
      "reward": 0.493664026260376,
      "reward_std": 0.24937184900045395,
      "reward_std_all": 0.2728554606437683,
      "rewards-max/forensics_sinkhorn_reward": 0.9349924921989441,
      "rewards-max/forensics_strict_edge_reward": 0.7293311953544617,
      "rewards-max/forensics_tiou_reward": 0.8840223252773285,
      "rewards/forensics_sinkhorn_reward": 0.40329766273498535,
      "rewards/forensics_strict_edge_reward": 0.19537591561675072,
      "rewards/forensics_tiou_reward": 0.2633910998702049,
      "step": 1462
    },
    {
      "completion_length": 17.140625,
      "epoch": 3.892143808255659,
      "grad_norm": 19.278085708618164,
      "kl": 7.65625,
      "learning_rate": 2.5333333333333332e-08,
      "loss": 0.5391,
      "reward": 0.529737114906311,
      "reward_std": 0.23561497032642365,
      "reward_std_all": 0.2531069219112396,
      "rewards-max/forensics_sinkhorn_reward": 0.9887371361255646,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9990014731884003,
      "rewards/forensics_sinkhorn_reward": 0.47243717312812805,
      "rewards/forensics_strict_edge_reward": 0.26137620210647583,
      "rewards/forensics_tiou_reward": 0.34314195811748505,
      "step": 1463
    },
    {
      "completion_length": 16.96875,
      "epoch": 3.894806924101198,
      "grad_norm": 44.7537727355957,
      "kl": 5.171875,
      "learning_rate": 2.4666666666666666e-08,
      "loss": 0.3285,
      "reward": 0.6112985014915466,
      "reward_std": 0.1830076053738594,
      "reward_std_all": 0.19165586680173874,
      "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": 0.5650307536125183,
      "rewards/forensics_strict_edge_reward": 0.29457975924015045,
      "rewards/forensics_tiou_reward": 0.37186892330646515,
      "step": 1464
    },
    {
      "completion_length": 16.859375,
      "epoch": 3.8974700399467377,
      "grad_norm": 26.485361099243164,
      "kl": 7.234375,
      "learning_rate": 2.4e-08,
      "loss": 0.5308,
      "reward": 0.546905905008316,
      "reward_std": 0.2497241050004959,
      "reward_std_all": 0.261577270925045,
      "rewards-max/forensics_sinkhorn_reward": 0.9668141305446625,
      "rewards-max/forensics_strict_edge_reward": 0.9256840348243713,
      "rewards-max/forensics_tiou_reward": 0.9496714174747467,
      "rewards/forensics_sinkhorn_reward": 0.48092329502105713,
      "rewards/forensics_strict_edge_reward": 0.25651441514492035,
      "rewards/forensics_tiou_reward": 0.2998741716146469,
      "step": 1465
    },
    {
      "completion_length": 17.734375,
      "epoch": 3.900133155792277,
      "grad_norm": 30.1179141998291,
      "kl": 7.125,
      "learning_rate": 2.3333333333333334e-08,
      "loss": 0.5314,
      "reward": 0.567706435918808,
      "reward_std": 0.2489648461341858,
      "reward_std_all": 0.2445487454533577,
      "rewards-max/forensics_sinkhorn_reward": 0.98875692486763,
      "rewards-max/forensics_strict_edge_reward": 0.9890114367008209,
      "rewards-max/forensics_tiou_reward": 0.9990796148777008,
      "rewards/forensics_sinkhorn_reward": 0.5260346531867981,
      "rewards/forensics_strict_edge_reward": 0.28469762206077576,
      "rewards/forensics_tiou_reward": 0.3672879636287689,
      "step": 1466
    },
    {
      "completion_length": 17.078125,
      "epoch": 3.902796271637816,
      "grad_norm": 20.38372802734375,
      "kl": 3.6953125,
      "learning_rate": 2.2666666666666668e-08,
      "loss": 0.2852,
      "reward": 0.689227819442749,
      "reward_std": 0.1441512405872345,
      "reward_std_all": 0.1372256800532341,
      "rewards-max/forensics_sinkhorn_reward": 0.9883083999156952,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9973110854625702,
      "rewards/forensics_sinkhorn_reward": 0.6384597718715668,
      "rewards/forensics_strict_edge_reward": 0.394386351108551,
      "rewards/forensics_tiou_reward": 0.4491945952177048,
      "step": 1467
    },
    {
      "completion_length": 16.296875,
      "epoch": 3.9054593874833556,
      "grad_norm": 22.699295043945312,
      "kl": 4.421875,
      "learning_rate": 2.2e-08,
      "loss": 0.3962,
      "reward": 0.5348891615867615,
      "reward_std": 0.24019506573677063,
      "reward_std_all": 0.25419315695762634,
      "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": 0.4818597882986069,
      "rewards/forensics_strict_edge_reward": 0.2914334684610367,
      "rewards/forensics_tiou_reward": 0.31469225138425827,
      "step": 1468
    },
    {
      "completion_length": 16.171875,
      "epoch": 3.9081225033288947,
      "grad_norm": 36.65154266357422,
      "kl": 4.484375,
      "learning_rate": 2.1333333333333332e-08,
      "loss": 0.3849,
      "reward": 0.5901907682418823,
      "reward_std": 0.14565163105726242,
      "reward_std_all": 0.1741398200392723,
      "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": 0.4985189139842987,
      "rewards/forensics_strict_edge_reward": 0.19924454391002655,
      "rewards/forensics_tiou_reward": 0.2862890064716339,
      "step": 1469
    },
    {
      "completion_length": 16.953125,
      "epoch": 3.910785619174434,
      "grad_norm": 18.54768943786621,
      "kl": 3.921875,
      "learning_rate": 2.0666666666666666e-08,
      "loss": 0.1801,
      "reward": 0.6132287085056305,
      "reward_std": 0.2297370582818985,
      "reward_std_all": 0.21631145477294922,
      "rewards-max/forensics_sinkhorn_reward": 0.9883759617805481,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9975789189338684,
      "rewards/forensics_sinkhorn_reward": 0.5690062344074249,
      "rewards/forensics_strict_edge_reward": 0.2748000845313072,
      "rewards/forensics_tiou_reward": 0.41033847630023956,
      "step": 1470
    },
    {
      "completion_length": 16.140625,
      "epoch": 3.9134487350199736,
      "grad_norm": 28.373926162719727,
      "kl": 9.15625,
      "learning_rate": 2e-08,
      "loss": 0.6141,
      "reward": 0.5219016373157501,
      "reward_std": 0.2824186682701111,
      "reward_std_all": 0.2780527174472809,
      "rewards-max/forensics_sinkhorn_reward": 0.9887395799160004,
      "rewards-max/forensics_strict_edge_reward": 0.9782643616199493,
      "rewards-max/forensics_tiou_reward": 0.999011367559433,
      "rewards/forensics_sinkhorn_reward": 0.4570769816637039,
      "rewards/forensics_strict_edge_reward": 0.17135339975357056,
      "rewards/forensics_tiou_reward": 0.24734065681695938,
      "step": 1471
    },
    {
      "completion_length": 18.0625,
      "epoch": 3.9161118508655126,
      "grad_norm": 25.924379348754883,
      "kl": 4.21875,
      "learning_rate": 1.9333333333333334e-08,
      "loss": 0.3306,
      "reward": 0.6347250044345856,
      "reward_std": 0.1914147287607193,
      "reward_std_all": 0.18222155421972275,
      "rewards-max/forensics_sinkhorn_reward": 0.9885894358158112,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9984190165996552,
      "rewards/forensics_sinkhorn_reward": 0.5939687490463257,
      "rewards/forensics_strict_edge_reward": 0.35212019085884094,
      "rewards/forensics_tiou_reward": 0.44501280784606934,
      "step": 1472
    },
    {
      "completion_length": 16.703125,
      "epoch": 3.918774966711052,
      "grad_norm": 20.303730010986328,
      "kl": 4.734375,
      "learning_rate": 1.866666666666667e-08,
      "loss": 0.3689,
      "reward": 0.610520213842392,
      "reward_std": 0.18117982149124146,
      "reward_std_all": 0.18656865507364273,
      "rewards-max/forensics_sinkhorn_reward": 0.9882866442203522,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9972255229949951,
      "rewards/forensics_sinkhorn_reward": 0.5474323928356171,
      "rewards/forensics_strict_edge_reward": 0.29365992546081543,
      "rewards/forensics_tiou_reward": 0.36936840415000916,
      "step": 1473
    },
    {
      "completion_length": 16.59375,
      "epoch": 3.921438082556591,
      "grad_norm": 19.867877960205078,
      "kl": 5.78125,
      "learning_rate": 1.8e-08,
      "loss": 0.3205,
      "reward": 0.5893760025501251,
      "reward_std": 0.18427354097366333,
      "reward_std_all": 0.19240841269493103,
      "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": 0.5577114224433899,
      "rewards/forensics_strict_edge_reward": 0.28951898217201233,
      "rewards/forensics_tiou_reward": 0.3956492245197296,
      "step": 1474
    },
    {
      "completion_length": 18.15625,
      "epoch": 3.9241011984021306,
      "grad_norm": 21.19430923461914,
      "kl": 7.109375,
      "learning_rate": 1.7333333333333333e-08,
      "loss": 0.471,
      "reward": 0.5458544194698334,
      "reward_std": 0.18333350867033005,
      "reward_std_all": 0.21181854605674744,
      "rewards-max/forensics_sinkhorn_reward": 0.9864251911640167,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.99337437748909,
      "rewards/forensics_sinkhorn_reward": 0.48950186371803284,
      "rewards/forensics_strict_edge_reward": 0.2125435173511505,
      "rewards/forensics_tiou_reward": 0.3249589055776596,
      "step": 1475
    },
    {
      "completion_length": 15.140625,
      "epoch": 3.92676431424767,
      "grad_norm": 14.881291389465332,
      "kl": 2.5703125,
      "learning_rate": 1.6666666666666667e-08,
      "loss": 0.246,
      "reward": 0.6464681327342987,
      "reward_std": 0.16676375642418861,
      "reward_std_all": 0.16941827535629272,
      "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": 0.7019890248775482,
      "rewards/forensics_strict_edge_reward": 0.5518229007720947,
      "rewards/forensics_tiou_reward": 0.616946280002594,
      "step": 1476
    },
    {
      "completion_length": 15.90625,
      "epoch": 3.929427430093209,
      "grad_norm": 30.12491798400879,
      "kl": 3.59375,
      "learning_rate": 1.6e-08,
      "loss": 0.2233,
      "reward": 0.5740542113780975,
      "reward_std": 0.20826219767332077,
      "reward_std_all": 0.21211273223161697,
      "rewards-max/forensics_sinkhorn_reward": 0.9881497621536255,
      "rewards-max/forensics_strict_edge_reward": 0.9565287232398987,
      "rewards-max/forensics_tiou_reward": 0.9966856837272644,
      "rewards/forensics_sinkhorn_reward": 0.4993419349193573,
      "rewards/forensics_strict_edge_reward": 0.23978261649608612,
      "rewards/forensics_tiou_reward": 0.31775452196598053,
      "step": 1477
    },
    {
      "completion_length": 17.328125,
      "epoch": 3.932090545938748,
      "grad_norm": 22.625730514526367,
      "kl": 5.875,
      "learning_rate": 1.5333333333333332e-08,
      "loss": 0.3378,
      "reward": 0.5464487969875336,
      "reward_std": 0.26219768822193146,
      "reward_std_all": 0.2676382064819336,
      "rewards-max/forensics_sinkhorn_reward": 0.8715921342372894,
      "rewards-max/forensics_strict_edge_reward": 0.7353169620037079,
      "rewards-max/forensics_tiou_reward": 0.7907498180866241,
      "rewards/forensics_sinkhorn_reward": 0.46124792098999023,
      "rewards/forensics_strict_edge_reward": 0.22822348028421402,
      "rewards/forensics_tiou_reward": 0.27661949396133423,
      "step": 1478
    },
    {
      "completion_length": 17.375,
      "epoch": 3.9347536617842875,
      "grad_norm": 38.017826080322266,
      "kl": 6.859375,
      "learning_rate": 1.4666666666666666e-08,
      "loss": 0.3661,
      "reward": 0.5352017283439636,
      "reward_std": 0.2670134902000427,
      "reward_std_all": 0.26314200460910797,
      "rewards-max/forensics_sinkhorn_reward": 0.9550743997097015,
      "rewards-max/forensics_strict_edge_reward": 0.8394034504890442,
      "rewards-max/forensics_tiou_reward": 0.9251529276371002,
      "rewards/forensics_sinkhorn_reward": 0.45546670258045197,
      "rewards/forensics_strict_edge_reward": 0.23449477553367615,
      "rewards/forensics_tiou_reward": 0.300515279173851,
      "step": 1479
    },
    {
      "completion_length": 18.296875,
      "epoch": 3.937416777629827,
      "grad_norm": 20.288848876953125,
      "kl": 6.140625,
      "learning_rate": 1.4e-08,
      "loss": 0.38,
      "reward": 0.5704959034919739,
      "reward_std": 0.20246028900146484,
      "reward_std_all": 0.21095171570777893,
      "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": 0.5054204314947128,
      "rewards/forensics_strict_edge_reward": 0.25939588248729706,
      "rewards/forensics_tiou_reward": 0.3379950374364853,
      "step": 1480
    },
    {
      "completion_length": 16.125,
      "epoch": 3.940079893475366,
      "grad_norm": 34.94266891479492,
      "kl": 5.171875,
      "learning_rate": 1.3333333333333334e-08,
      "loss": 0.2734,
      "reward": 0.5535732507705688,
      "reward_std": 0.214305579662323,
      "reward_std_all": 0.2280547097325325,
      "rewards-max/forensics_sinkhorn_reward": 0.9091213047504425,
      "rewards-max/forensics_strict_edge_reward": 0.6703200340270996,
      "rewards-max/forensics_tiou_reward": 0.8328148126602173,
      "rewards/forensics_sinkhorn_reward": 0.43099264800548553,
      "rewards/forensics_strict_edge_reward": 0.08650030940771103,
      "rewards/forensics_tiou_reward": 0.17358119040727615,
      "step": 1481
    },
    {
      "completion_length": 15.71875,
      "epoch": 3.9427430093209055,
      "grad_norm": 18.788860321044922,
      "kl": 2.3984375,
      "learning_rate": 1.2666666666666666e-08,
      "loss": 0.1634,
      "reward": 0.6632292568683624,
      "reward_std": 0.12246029451489449,
      "reward_std_all": 0.1345965415239334,
      "rewards-max/forensics_sinkhorn_reward": 0.988050788640976,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9962992370128632,
      "rewards/forensics_sinkhorn_reward": 0.6308431029319763,
      "rewards/forensics_strict_edge_reward": 0.32533547282218933,
      "rewards/forensics_tiou_reward": 0.45616787672042847,
      "step": 1482
    },
    {
      "completion_length": 16.15625,
      "epoch": 3.9454061251664445,
      "grad_norm": 26.961734771728516,
      "kl": 4.65625,
      "learning_rate": 1.2e-08,
      "loss": 0.3905,
      "reward": 0.5880156755447388,
      "reward_std": 0.20911240577697754,
      "reward_std_all": 0.21580949425697327,
      "rewards-max/forensics_sinkhorn_reward": 0.9870611131191254,
      "rewards-max/forensics_strict_edge_reward": 0.9565287232398987,
      "rewards-max/forensics_tiou_reward": 0.9924023449420929,
      "rewards/forensics_sinkhorn_reward": 0.516539067029953,
      "rewards/forensics_strict_edge_reward": 0.2710965573787689,
      "rewards/forensics_tiou_reward": 0.3183610141277313,
      "step": 1483
    },
    {
      "completion_length": 16.609375,
      "epoch": 3.948069241011984,
      "grad_norm": 24.983827590942383,
      "kl": 5.4375,
      "learning_rate": 1.1333333333333334e-08,
      "loss": 0.3086,
      "reward": 0.6218604147434235,
      "reward_std": 0.2117295265197754,
      "reward_std_all": 0.20190448313951492,
      "rewards-max/forensics_sinkhorn_reward": 0.9874922931194305,
      "rewards-max/forensics_strict_edge_reward": 0.9672757983207703,
      "rewards-max/forensics_tiou_reward": 0.9940958619117737,
      "rewards/forensics_sinkhorn_reward": 0.6099609136581421,
      "rewards/forensics_strict_edge_reward": 0.3792024999856949,
      "rewards/forensics_tiou_reward": 0.4755425751209259,
      "step": 1484
    },
    {
      "completion_length": 17.0625,
      "epoch": 3.9507323568575234,
      "grad_norm": 15.981608390808105,
      "kl": 4.0390625,
      "learning_rate": 1.0666666666666666e-08,
      "loss": 0.3546,
      "reward": 0.5285218954086304,
      "reward_std": 0.19437125325202942,
      "reward_std_all": 0.22939980030059814,
      "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": 0.48509612679481506,
      "rewards/forensics_strict_edge_reward": 0.3127150684595108,
      "rewards/forensics_tiou_reward": 0.36082151532173157,
      "step": 1485
    },
    {
      "completion_length": 17.25,
      "epoch": 3.9533954727030625,
      "grad_norm": 26.142578125,
      "kl": 4.0,
      "learning_rate": 1e-08,
      "loss": 0.3433,
      "reward": 0.5777537524700165,
      "reward_std": 0.23372675478458405,
      "reward_std_all": 0.2324173003435135,
      "rewards-max/forensics_sinkhorn_reward": 0.9869881868362427,
      "rewards-max/forensics_strict_edge_reward": 0.946017861366272,
      "rewards-max/forensics_tiou_reward": 0.9921095669269562,
      "rewards/forensics_sinkhorn_reward": 0.5188242793083191,
      "rewards/forensics_strict_edge_reward": 0.30515846610069275,
      "rewards/forensics_tiou_reward": 0.3826661556959152,
      "step": 1486
    },
    {
      "completion_length": 15.125,
      "epoch": 3.956058588548602,
      "grad_norm": 17.756662368774414,
      "kl": 4.0078125,
      "learning_rate": 9.333333333333334e-09,
      "loss": 0.3454,
      "reward": 0.5562315434217453,
      "reward_std": 0.2531522214412689,
      "reward_std_all": 0.25646571815013885,
      "rewards-max/forensics_sinkhorn_reward": 0.9880072474479675,
      "rewards-max/forensics_strict_edge_reward": 0.9780228734016418,
      "rewards-max/forensics_tiou_reward": 0.9961277842521667,
      "rewards/forensics_sinkhorn_reward": 0.5323933213949203,
      "rewards/forensics_strict_edge_reward": 0.35197940468788147,
      "rewards/forensics_tiou_reward": 0.41456231474876404,
      "step": 1487
    },
    {
      "completion_length": 15.65625,
      "epoch": 3.958721704394141,
      "grad_norm": 22.263916015625,
      "kl": 4.2578125,
      "learning_rate": 8.666666666666667e-09,
      "loss": 0.4259,
      "reward": 0.5994058847427368,
      "reward_std": 0.18276017159223557,
      "reward_std_all": 0.1947297528386116,
      "rewards-max/forensics_sinkhorn_reward": 0.9646883606910706,
      "rewards-max/forensics_strict_edge_reward": 0.9003686904907227,
      "rewards-max/forensics_tiou_reward": 0.9453645944595337,
      "rewards/forensics_sinkhorn_reward": 0.5628247708082199,
      "rewards/forensics_strict_edge_reward": 0.37058350443840027,
      "rewards/forensics_tiou_reward": 0.4283420145511627,
      "step": 1488
    },
    {
      "completion_length": 18.0625,
      "epoch": 3.9613848202396804,
      "grad_norm": 20.336334228515625,
      "kl": 6.2109375,
      "learning_rate": 8e-09,
      "loss": 0.4549,
      "reward": 0.5962561368942261,
      "reward_std": 0.2442258670926094,
      "reward_std_all": 0.23433813452720642,
      "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": 0.5763977020978928,
      "rewards/forensics_strict_edge_reward": 0.38979965448379517,
      "rewards/forensics_tiou_reward": 0.46253180503845215,
      "step": 1489
    },
    {
      "completion_length": 16.890625,
      "epoch": 3.96404793608522,
      "grad_norm": 20.82293128967285,
      "kl": 5.15625,
      "learning_rate": 7.333333333333333e-09,
      "loss": 0.5117,
      "reward": 0.5733269155025482,
      "reward_std": 0.1796315610408783,
      "reward_std_all": 0.20214027911424637,
      "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": 0.5399667024612427,
      "rewards/forensics_strict_edge_reward": 0.32051563262939453,
      "rewards/forensics_tiou_reward": 0.3997676819562912,
      "step": 1490
    },
    {
      "completion_length": 15.984375,
      "epoch": 3.966711051930759,
      "grad_norm": 21.827484130859375,
      "kl": 5.640625,
      "learning_rate": 6.666666666666667e-09,
      "loss": 0.2944,
      "reward": 0.6406621932983398,
      "reward_std": 0.16566627472639084,
      "reward_std_all": 0.1652096062898636,
      "rewards-max/forensics_sinkhorn_reward": 0.9881614446640015,
      "rewards-max/forensics_strict_edge_reward": 0.9677534997463226,
      "rewards-max/forensics_tiou_reward": 0.9967373609542847,
      "rewards/forensics_sinkhorn_reward": 0.6217168569564819,
      "rewards/forensics_strict_edge_reward": 0.3308796137571335,
      "rewards/forensics_tiou_reward": 0.4726877808570862,
      "step": 1491
    },
    {
      "completion_length": 17.328125,
      "epoch": 3.9693741677762984,
      "grad_norm": 16.090389251708984,
      "kl": 5.0625,
      "learning_rate": 6e-09,
      "loss": 0.4008,
      "reward": 0.6207523345947266,
      "reward_std": 0.14665485173463821,
      "reward_std_all": 0.17571943253278732,
      "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": 0.5899623930454254,
      "rewards/forensics_strict_edge_reward": 0.39312055706977844,
      "rewards/forensics_tiou_reward": 0.48633891344070435,
      "step": 1492
    },
    {
      "completion_length": 18.65625,
      "epoch": 3.9720372836218374,
      "grad_norm": 20.010791778564453,
      "kl": 4.5,
      "learning_rate": 5.333333333333333e-09,
      "loss": 0.1612,
      "reward": 0.5856397151947021,
      "reward_std": 0.18782681226730347,
      "reward_std_all": 0.19926904141902924,
      "rewards-max/forensics_sinkhorn_reward": 0.9188109636306763,
      "rewards-max/forensics_strict_edge_reward": 0.8227206170558929,
      "rewards-max/forensics_tiou_reward": 0.852524071931839,
      "rewards/forensics_sinkhorn_reward": 0.47499579191207886,
      "rewards/forensics_strict_edge_reward": 0.2445787787437439,
      "rewards/forensics_tiou_reward": 0.2700165584683418,
      "step": 1493
    },
    {
      "completion_length": 18.203125,
      "epoch": 3.974700399467377,
      "grad_norm": 20.251178741455078,
      "kl": 6.484375,
      "learning_rate": 4.666666666666667e-09,
      "loss": 0.4599,
      "reward": 0.5900021344423294,
      "reward_std": 0.18054477870464325,
      "reward_std_all": 0.19590627774596214,
      "rewards-max/forensics_sinkhorn_reward": 0.9864940941333771,
      "rewards-max/forensics_strict_edge_reward": 0.9464850425720215,
      "rewards-max/forensics_tiou_reward": 0.9902027249336243,
      "rewards/forensics_sinkhorn_reward": 0.5289282202720642,
      "rewards/forensics_strict_edge_reward": 0.2952760383486748,
      "rewards/forensics_tiou_reward": 0.3656983822584152,
      "step": 1494
    },
    {
      "completion_length": 15.5625,
      "epoch": 3.9773635153129163,
      "grad_norm": 18.380096435546875,
      "kl": 5.796875,
      "learning_rate": 4e-09,
      "loss": 0.4451,
      "reward": 0.5555476248264313,
      "reward_std": 0.18710419535636902,
      "reward_std_all": 0.2183711752295494,
      "rewards-max/forensics_sinkhorn_reward": 0.9873566925525665,
      "rewards-max/forensics_strict_edge_reward": 0.9464850425720215,
      "rewards-max/forensics_tiou_reward": 0.9935687482357025,
      "rewards/forensics_sinkhorn_reward": 0.5043125599622726,
      "rewards/forensics_strict_edge_reward": 0.2791558802127838,
      "rewards/forensics_tiou_reward": 0.3396558314561844,
      "step": 1495
    },
    {
      "completion_length": 17.0,
      "epoch": 3.9800266311584553,
      "grad_norm": 21.08037757873535,
      "kl": 5.65625,
      "learning_rate": 3.3333333333333334e-09,
      "loss": 0.5327,
      "reward": 0.5849929451942444,
      "reward_std": 0.20330508053302765,
      "reward_std_all": 0.20761727541685104,
      "rewards-max/forensics_sinkhorn_reward": 0.9882056415081024,
      "rewards-max/forensics_strict_edge_reward": 0.9464850425720215,
      "rewards-max/forensics_tiou_reward": 0.9969088137149811,
      "rewards/forensics_sinkhorn_reward": 0.5511322021484375,
      "rewards/forensics_strict_edge_reward": 0.2509419471025467,
      "rewards/forensics_tiou_reward": 0.41130946576595306,
      "step": 1496
    },
    {
      "completion_length": 16.390625,
      "epoch": 3.982689747003995,
      "grad_norm": 16.905155181884766,
      "kl": 5.6875,
      "learning_rate": 2.6666666666666666e-09,
      "loss": 0.4782,
      "reward": 0.5710006356239319,
      "reward_std": 0.21324586123228073,
      "reward_std_all": 0.2294028326869011,
      "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": 0.538374736905098,
      "rewards/forensics_strict_edge_reward": 0.3004342317581177,
      "rewards/forensics_tiou_reward": 0.3782586455345154,
      "step": 1497
    },
    {
      "completion_length": 16.546875,
      "epoch": 3.985352862849534,
      "grad_norm": 56.86882781982422,
      "kl": 6.84375,
      "learning_rate": 2e-09,
      "loss": 0.408,
      "reward": 0.5821929275989532,
      "reward_std": 0.200160451233387,
      "reward_std_all": 0.20763956010341644,
      "rewards-max/forensics_sinkhorn_reward": 0.9635455012321472,
      "rewards-max/forensics_strict_edge_reward": 0.7387099117040634,
      "rewards-max/forensics_tiou_reward": 0.9412756264209747,
      "rewards/forensics_sinkhorn_reward": 0.4870176315307617,
      "rewards/forensics_strict_edge_reward": 0.2103695087134838,
      "rewards/forensics_tiou_reward": 0.3088250905275345,
      "step": 1498
    },
    {
      "completion_length": 15.453125,
      "epoch": 3.9880159786950733,
      "grad_norm": 22.554147720336914,
      "kl": 4.3359375,
      "learning_rate": 1.3333333333333333e-09,
      "loss": 0.4062,
      "reward": 0.6924674808979034,
      "reward_std": 0.1330495961010456,
      "reward_std_all": 0.12650812044739723,
      "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": 0.639966607093811,
      "rewards/forensics_strict_edge_reward": 0.3848976492881775,
      "rewards/forensics_tiou_reward": 0.4971211403608322,
      "step": 1499
    },
    {
      "completion_length": 17.15625,
      "epoch": 3.9906790945406128,
      "grad_norm": 22.80509376525879,
      "kl": 3.484375,
      "learning_rate": 6.666666666666666e-10,
      "loss": 0.3374,
      "reward": 0.5975591540336609,
      "reward_std": 0.22972357273101807,
      "reward_std_all": 0.23222684860229492,
      "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": 0.5149797052145004,
      "rewards/forensics_strict_edge_reward": 0.30511363595724106,
      "rewards/forensics_tiou_reward": 0.3794189542531967,
      "step": 1500
    }
  ],
  "logging_steps": 1.0,
  "max_steps": 1500,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 4,
  "save_steps": 50,
  "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
}